DECEMBER 2015

MRTEN BLIX

The economy and digitalization

opportunities and challenges

Arkitektkopia AB, Bromma, 2015

Frfattare: Mrten Blix
Email: [email protected]
www.martenblix.com

Denna studie r gjord p uppdrag av Svenskt Nringsliv. Frfattaren str sjlv fr allt innehll.
This report is written on behalf of the Confederation of Swedish Enterprise. The author is alone
responsible for all content.
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Preface

If a software robot, such as IBMs Watson or some future variant thereof were to
write a report on the economic implications of digitalization, what would it look like?

Unlike the author, it would be able to scan all the current works that reference auto-
mation and technology, from Keynes and the Luddites to the myriad of reports from
think tanks, management consultancies, and armchair philosophers. In this I fear the
robot would win in terms of sheer volume and endurance. What would the scan look
like? It would find the current economic literature a little unsure of how to tackle the
issue of digitalization, recently scarred from having misread and otherwise having been
blindsided by the financial crisis in 2007. Could economists once again get it wrong?

Perhaps the first step for the software robot would be to specify keywords to use in
the search. In his book Superintelligence: Paths, Dangers, Strategies, Nick Bostrom
discusses the path to artificial intelligence (AI), but for the foreseeable future, robots
are not true AI and the keywords would have to be supplied by a human being, such
as myself. Indeed, if computers were true AI, it is unclear why they would want to
write a report about the economic effects of digitalization in the first place. Taking
the economists approach of focusing on incentives, we might suppose that AI robots
would be most concerned about the steady supply of electricity and raw materials for
replacement parts, just as human beings are preoccupied with extending the span and
enhancing the quality of their lives.

After having been given the keywords to search for, the robot would scan the literature
and arrange the arguments in some order. Using statistical techniques and amassing
frequencies, it would compile a list of data up the current day with all the arguments
that support various positions. For example, the argument that robots will take over
is supported by facts x, y, and z and so on. In the jargon of the economic profession,
this kind of search might be called data mining, which is very useful for finding corre-
lations and discovering trends.

But it also brings several pitfalls. On the plus side, the robot software would not be
afraid of amassing evidence that did not support a particular position. As long as the
program it follows is neutral, it stands to reason that the output would be neutral as
well. A human being writing a report inevitably has prior beliefs (as Bayesians might
say) that slip into the writing and influence the report; we might call them biases. All
writings, from newspaper articles to books to academic papers are to some extent
colored by prior beliefs, as is the selection of reading material. For example, a person
who selects one newspaper rather than another makes a choice to be more interested
in the prior belief of that newspaper. It is then perhaps no surprise that that the early
robot-written texts tend to be in areas of sports and competition, where the presenta-
tion of goals and times compared to the past are the key outcomes.

So, having an unbiased report would surely be a good thing? It is not clear that even
if the researcher or the robot software is unbiased that this would automatically
make the report unbiased. If the main driving force for finding results would be the
frequency of how many times various keywords are mentioned, then there is always

1
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

the risk that frequently mentioned nonsense would be assigned high value (or in the
jargon, a high probability mass), whereas the really good stuff might be discarded.
In my reading of the literature, I keep coming back to some of the great writings
and thinkers Keynes, Schumpeter, Coase that stand the test of time; even when
they are wrong, they are wrong in helpful ways that give us the wherewithal to think
better about the issues that matter. Of course, no robot would dismiss Keynes, but
what other good material might fall below the radar?

In the end, I would be quite happy to let robot software do the scut work of a literature
review, not unlike how we send out spacecraft to amass data from faraway planets
while we go about our lives, so that I could focus on the other stuff: structuring the
report, trying to make abstract ideas understandable, and addressing policy issues of
how the economy works and how it affects people. In the future, they might be able
to help with these things too, but then I would likely feel more inclined to overrule
the robot.

I would especially like the robot to create all my charts and put the references in the
correct order and format and I doubt I am alone there. Just as machines in the past
made the need for arduous and dangerous physical work redundant across broad
activities mining and road/railway construction, for instance we can only hope
that some of the hitherto inevitably dreary parts of writing or the equivalent in
other professions might be easier.

How I try to address the policy issues raised by digitalization is the outcome of this
report. I believe the amassing and structuring of data is an important phase but that it
should not distract from scenario or policy analysis. Human beings are not particularly
good at forecasting far into the future and there is no reason to believe robots would
be better. Instead, one of our strengths is to have scenario analyses that outline different
future paths often taking somewhat extreme roads to more clearly illustrate the
implications of going in some particular direction in order to raise the issues con-
fronting policymakers. Ultimately, the question is what kind of policies roughly support
good outcomes and avoid the really bad ones.

It is especially important to think about policies that are grounded in the way our
institutions currently work because systems change slowly and, absent a crisis, there is
little chance of public acceptance of major upheavals. Since we cannot be completely
sure what the right policy is anyway, slow changes may be a good thing, while always
being open to adjusting the path if it turns out prior beliefs were ill-advised.

Perhaps as a matter of conscience, I should also state my own prior beliefs. I find it
hard to believe doomsday predictions of a future without work, as outlined in any
number of recent books, for example Martin Fords Rise of the Robots. People adjust
and new jobs are created; this happened despite the fears of the Luddites during the
Industrial Revolution. As stressed by David Autor, most of the benefits from digitaliza-
tion come from viewing the complementarity in work between computers and people.

But the worry is that the period of adjustment may be tough. In particular, it might
be hard for broad groups in society the lower and middle classes as the changes
brought by digitalization are both wide in scope and fast. To be sure, in contrast to
the adjustments that occurred during the Industrial Revolution affecting workers,
artisans, and even the aristocracy we now have social safety nets such as unemploy-
ment insurance and sick pay; in Sweden, we have one of the most expansive social
safety nets in the world and so the issues confronting us are different from those in
the US, where the social safety nets are riddled with holes.

2
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

I believe the question we should be asking is how institutions in different countries

need to adapt to make the structural change from digitalization as unrugged as pos-
sible. This involves social safety nets, how wage bargaining works, the rules governing
how firms operate, and the education system. In contrast to life during the Industrial
Revolution, people today have much higher expectations on government, and life is
not just about surviving but also about quality living the good life. The British
television drama Downton Abbey, set in the latter years of the Industrial Revolution
illustrates well the clash between the old and new. We are now entering an age faced
with our own, digital, version of this challenge.

In writing this report, I have received useful input and suggestions from many people
but all opinions expressed are my own. I would especially like to thank Jonas Arnberg,
Stig Bengtsson, Jan Brockmann, Hugo Brndstrm, Jrgen Bdmar, Mia Chennell,
Stefan Eklund, Eva Erlandsson, Ann-Marie Fransson, Stefan Flster, Dan Grannas,
Christine Grahn, Ola Landstrm, Christian Levin, Erik Ljungberg, Bjrn Lundqvist,
Anders Hektor, Carl-Gustaf Leinar, Assar Lindbeck, David Mothander, Magnus Nytell,
Claudia Olsson, Ulf Pehrsson, Paul Palmstedt, Patrik Regrdh, Therese Rosen Lfstedt,
Pr Nygrds, Anna Sabelstrm, Fredrik Sand, Christian Sandstrm, Rene Summer,
sa Sterte, Magnus Thynell, Fredrik Sderqvist, Markus Tibblin, Katarina Tobe,
Anita Vahlberg, and Urban Wass.

A special thanks to all those that have commented on various aspects in the report:
Marianna Blix Grimaldi, Anna Breman, Jimmy Boumediene, Lena Carlsson, Gabriela
Chirico Willstedt, Enrico Deiaco, Irene Ek, Ingemar Eriksson, Robert Gidehag, Magnus
Henrekson, Fredrik Heyman, Pehr-Johan Norbck, Lars Persson, Jesper Roine, Eva
Udden Sonnegrd, Victor Snellman, Susanne Spector, Daniel Wiberg, and Sara hrvall.
A very special thanks also to Hiu Yee Fan and Sebastian Jvervall for excellent research
assistance, to Bo Nordstrom for translation and to Rosemary Nordstrom for superb
editing of the manuscript.

I would also like to express special appreciation to Magnus Henrekson. Above all,
Iwould like to express my thanks to Ann berg for her unwavering support and help
throughout the project.

I am also grateful to my parents Eva Kettis and Hans Blix, for comments and for
proofreading. Finally, I would like to thank my wife Marianna and my children
Melanie and Marco, who, I suspect, may have come to see digitalization more as
a manual activity with me in front of the computer rather than anything remotely
resembling software automation in the cloud. Maybe next time

Mrten Blix

Tyres, November 30, 2015

3
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Executive summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Exekutiv sammanfattning psvenska . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Extended summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Sammanfattning p svenska . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1. Creative destruction in a digital world . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
1.2 Speed and scope of structural change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
1.3 Long-run trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
1.4 Productivity growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
1.5 Structure of this report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2. Economics of digitalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.2 Foundations of economic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.3 Features of the workhorse model that are changing: production . . . . . . . . . . . . . . . . 78
2.4 Demand side changes: prices and competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.5 Changing how the economy works the sharing economy . . . . . . . . . . . . . . . . . . . . . . 93
2.6 Output gap and sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.7 Can digitalization explain flatline inflation? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3. Implications for the labor market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.2 Lesson from previous periods of structural change in Sweden . . . . . . . . . . . . . . . . . 110
3.3 The effect of automation on employment and wages . . . . . . . . . . . . . . . . . . . . . . . . . 118
3.4 Humans and robots...or humans vs robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.5 Changing nature of work and its implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4. Regulation and obstacles in the digital economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.2 Legal and regulatory hurdles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.3 Taxation and the digital economy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.4. Regulatory hurdles for digital goods and services need more attention . . . . . . . . . 153
4.5 Big data and Internet of things . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
5. I robot, you employed, he or she freelance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.2 Sweden as a great IT nation, is the past any guide? . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.3 Human capital, digital skills and lifelong learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
5.4 Risk sharing in the economy, digital version . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
5.5 A future with jobs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184

4
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Executive summary

Will digitalization destroy half the available jobs in the next decade or so, as suggested
by Oxford researchers Carl Benedict Frey and Michael Osborne? Or will we experience
structural change as usual that our institutions are well poised to handle? Will the
current lackluster productivity growth, burdened by aging demographics and high public
debt, bounce back when the full effects of digitalization emerge? How our institutions
governments, central banks and labor market organizations respond to these challenges
will be key to the welfare effects.

In this report we survey the evidence of the effects of digitalization, by which we

mean the broad implementation of digital technologies across many different dimen-
sions of society. We contend that it is not enough to only look at past experience of
technological change to understand what digitalization may bring; we also need to
consider the effects of long-run trends, such as aging populations and urbanization.
Itis also important to critically assess what new technological developments mean.

Against the background of slow productivity growth in OECD countries, there

is some urgency about improving the structural features to not hold back growth
unnecessarily due to aging populations and public and private debt overhang. But
thepace of regulatory reform for product and labor markets has slowed in OECD
countries, including Sweden, in recent years.

It is fair to say that there is consensus that technology had positive impact on the
content of work and has not, despite forebodings to the contrary, reduced the total
amount of work. Famously, the Luddites destroyed machines during the Industrial
Revolution in the UK, fearing that jobs of artisans would fade away with the rise of
machines. Jobs were indeed destroyed, but new ones were created in industry and
in services. Throughout the 20th century, especially the latter half, advances in tech-
nology led to better jobs and higher productivity. Many bad jobs, requiring arduous
or dangerous manual labor such as working in mines or building roads were taken
over by machines. Household work also improved tremendously with electricity and
availability of modern appliances.

We have experienced what economists sometimes label skill-biased technological

change. Improvements in technology led to higher demands on labor, which in turn
improved output and generated real wage growth. The question then becomes: is the
good streak of technology and work set to continue? Will computers and robots lead
to the kind of upheaval experienced during the Industrial Revolution or will events
unfold more like they have in recent decades? The jury is still out, but in this report
we raise concerns that we may be in for a difficult transition period and the outcomes
will crucially depend on the quality of response from governments and institutions.

Today we tend to look at the Industrial Revolution as the basis of our welfare. The
invention of the steam engine, the car, and the train, along with clean water and
advances in medicine made tremendous improvements in quality of human life. We
also saw the development of labor unions to balance the power of factory owners.
Nevertheless, economic historian Joel Mokyr of Northwestern University has pointed
out that for most of the period 17501850, real wages did not improve much.

5
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The challenges today, of course, are not as dramatic as during the Industrial Revolution.
But the modern equivalent is job polarization and the shrinking middle class, which
have been documented in many OECD countries. The experience in Sweden is different
from for example the US, where broad groups have had little real wage growth. Quite
the contrary, real wage growth has been strong in Sweden since the beginning of the
1990s, coinciding with the introduction of a set of institutional reforms for fiscal
discipline and an independent central bank.

It is likely that we will continue to see job polarization in OECD countries. So far,
job polarization in Sweden has been characterized by high growth of the most skilled
jobs, a diminished share in the middle and an about constant share of the lower skilled
jobs. Future developments will depend on how policy makers respond. With protec-
tionist responses that undermine the benefits of digitalization, we are likely to see
much stronger job and wage polarization. In other words, income inequality may rise,
especially since the growth of new jobs will be slower if regulation developed for the
analog world is not updated to account for the different mechanisms and potentials
of digitalization. It is especially important to reduce the uncertainty in regulatory
change that may constrain investment.

One way to formulate the challenge is thus to aim for policies that will help realize
the productivity gains from digitalization but with as few consequences on welfare as
possible; that is, ensuring that general prosperity increases and that as largest share
of the population as possible are along for the ride. Thus far, policies in Sweden have
not taken the challenges seriously enough and the risks of adverse outcomes in the
labor market have increased.

To better understand the magnitude of the policy challenge, one place to start isto
review the outlook for automation of jobs. The incentive to automate work remains
strong and is driven by the will to improve products and services to make things better.
This incentive is now being strengthened by demography. In many OECD countries,
populations are aging rapidly; in Sweden the number of young people entering the
labor market will soon be fewer than those who exit for retirement. German carmakers
are already dealing with shortages of production workers, increasing theneed to
automate tasks and improve working conditions for older people to promote longer
working lives. In some sectors or regions, the lack of skilled workers may be especially
acute even as there is an oversupply of unskilled workers in some areas.

The extent to which it is possible to automate tasks would likely come as a shock to
many people. Often the idea of automation is displayed as a robot sitting at a desk,
often with a coffee mug perhaps to appear less threatening. But the image is mis-
leading, as most of automation, especially of more skilled work, will be done in the
cloud, from faceless server halls physically located in especially favorable locations.
For example, Facebook has located such a center in Lule in the north of Sweden to
draw benefits from the natural cooling provided by the low average temperatures.

We are seeing further automation in services, such as hotels and restaurants. With the
advent of self-driving vehicles, many fewer professional drivers will be needed. There
is also a next level of automation in finance and insurance. The tasks in the middle are
increasingly subject to competition from computer code or from on-demand services
in the cloud. For example, a Dutch insurance company has streamlined insurance
claims processes, removing several human steps, which has resulted in faster results
and higher customer satisfaction. Banking is particularly amenable to programming
because a great deal of content originates in numbers and many intermediary func-
tions are disappearing.

6
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Work automation revolves around subdividing tasks into different parts, similar to
the process that led to outsourcing of work to low-income countries in Asia and else-
where. Suitably defined, tasks can be automated and improved, often in cloud-based
services; they may also be outsourced to the sharing economy via digital platforms.
The new feature of job automation is that more skilled work is subject to automation
as well, such as legal research, administration, writing, and grading of written exams
and essays.

A constraint on work automation has been the difficulty of coding tasks that require
cognitive ability or creativity, an argument especially espoused by Professor David
Autor of MIT, one of the foremost scholars in the field of technology and labor mar-
kets. For example, humans know how ride a bicycle but writing down the rules for
this activity is inordinately difficult. Another often used example is the difficulty of
making programs recognize physical objects based on definable characteristics, such
as that a chair has legs or a cat has whiskers. While humans can no longer beat com-
puters at chess or Jeopardy, it provides false comfort that humans are still betterat,
for example, not mistaking a traffic cone for an easy chair. Notably, overcoming these
issues is a challenge for programmers, but it is not clear that they pose an insurmount-
able obstacle. Indeed, advances in machine learning allow software to mimic human
behavior even when the understanding of purpose is not coded. That automationwill
hit an insuperable obstacle when it comes to tackling tacit skills remains to be seen.
Rather than being a brick wall beyond which automation cannot venture, tacit
knowledge might be reshaped or subject to circumvention and redefinition.

We may already be seeing advances in this direction. Even creative work, hitherto firmly
in the human domain, is also affected by ones and zeroes. For example, computer
programs have written music performed at concerts. Even the disrupters themselves
the programmers who write code to automate work are also beginning to be at risk.
The reason for this is that a large share of code is a bit like the building pieces for a
piece of IKEA furniture; the basic components could be pulled off the shelf and do
not have to be reinvented every time.

While the possibilities of automation have increased tremendously, this is not tosay
that the path of automation is a straight line or somehow inevitable in all areas. There
is strong institutional inertia that slows automation, but other factors are also impor-
tant and sometimes dominate. Tasks will not be automated if there is no profit to be
made. And profits will depend on the costs of automation compared to human work,
demand factors, and regulation.

Taken together, what are the major forces most likely to affect the pace of automation?
Aging populations and lack of skilled workers in some areas will increase the incentive
for automation. Rigid labor markets have the same effect, making it more attractive
to either automate tasks or buy them in the sharing economy. A force in the other
direction for a small country such as Sweden is that of scale. The benefits of automation
depend on costs of investment; in some areas where investment costs are substantial,
automation may be slower. The net effect of various forces is impossible to predict
and will depend on a variety of factors, including policy responses from government
and from the labor market partners. However, to conclude that small countries are
safe from automation from big platforms would be a mistake. A more proactive
view would be that there may be opportunities for entrepreneurs to build clever ways
to account for local idiosyncrasies, in Europe and elsewhere, before the big platforms
come knocking.

7
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

With a protectionist response and slow regulatory reform, the path may be rockier.
Welfare gains may be slower in coming or even fail to materialize. More specifically,
we would likely to see more technological unemployment, increases in income ine-
quality, and slower productivity growth. When the economys ability to shift resources
between sectors is interfered with, the adjustment will be tougher, most especially on
labor market outsiders.

But there are some policies that would help smooth adjustment in the labor market
and help realize the benefits of digitalization:
Lower the tax on (human) labor. High taxes on labor in Sweden further strengthen
the already strong incentive to automate tasks. Tax deductions for household services
should be expanded, not reduced.
Ease the regulatory uncertainty that surrounds the sharing economy. This will pro-
vide flexibility for adjustment and lower risks of higher structural unemployment.
Improve opportunities for lifelong learning. Longer working lives combined
with rapid technological change increase the risk that some skills will reach their
best-before date earlier than before. To reduce the risk of poor prospects in the
labor market, updated skills will be crucial.
Reduce the asymmetry in social security between being employed or working
freelance.
Establish principles for regulation of the digital economy that can be used to
speed reform in different sectors, reducing the need to reinvent the wheel. This
will require focused collaboration between lawyers and economists and not at
the usual slow pace. Otherwise, productivity growth may remain stagnant and
themacroeconomic headwinds from demography and public debt will continue
todampen growth.

The digital revolution will likely improve quality of life, efficiency at work, and con-
tinue to transform leisure. But there is nothing inevitable about how smooth the ride
will be. The institutional response will be key to improving productivity growth
without risking higher unemployment and rising inequality.

8
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Exekutiv sammanfattning
psvenska

Kommer digitaliseringen att f hlften av dagens arbetstillfllen att frsvinna inom de

nrmaste tjugo ren, som Oxford-forskarna Carl Benedict Frey och Michael Osborne
pstr? Eller kommer vi att f uppleva en vanlig strukturfrndring som vra insti-
tutioner har erfarenhet av hur man hanterar? Kommer dagens svaga produktivitets
tillvxt, hmmad av ldrande befolkningar och hga statsskulder, att terhmta sig
nr digitaliseringen slr igenom fullt ut? Hur vra institutioner regering, riksbank
och arbetsmarknadens parter reagerar p sdana utmaningar lr f avgrande
effekt p vlfrden.

Den hr rapporten redogr fr digitaliseringens effekter, och vi avser implementeringen

av digital teknik i samhllet i bred bemrkelse. Vi menar att det inte rcker med att
bara titta p tidigare erfarenheter av tekniska frndringar fr att frst vad digita-
liseringen kan komma att innebra. Vi mste ocks vga in effekterna av lngsiktiga
trender, som ldrande befolkningar och urbanisering. Det r ocks viktigt att kritiskt
analysera vad teknologiska innovationer har fr implikationer.

Mot bakgrund av den lngsamma produktivitetstillvxten i OECD-lnderna r det

angelget att snarast genomfra strukturreformer i ekonomin s att tillvxten inte
iondan bromsas av ldrande befolkningar och hga privata och offentliga skulder.
Dessvrre har takten p strukturreformer p produkt- och arbetsmarknader saktat
ned i OECD-lnderna, inklusive Sverige, p senare r.

Det rder konsensus om att tekniken har pverkat arbetets innehll p ett positivt
stt och inte, trots varningar om motsatsen, minskat den totala mngden jobb. Det r
vl knt att de s kallade Ludditerna frstrde maskiner under den industriella revolu-
tionen i Storbritannien av rdsla fr att hantverkarna skulle frlora sina jobb. Deras
jobb frsvann mycket riktigt, men det skapades ocks nya inom industrin och tjnste-
sektorn. Under hela 1900-talet, srskilt under de sista 50 ren, skapade den tekniska
utvecklingen bttre jobb och hgre produktivitet. Mnga farliga eller mdosamma
arbetsuppgifter till exempel gruv- eller vgarbete vertogs av maskiner. Elektrici-
teten ochtillgngen till moderna hushllsmaskiner innebar ocks en kraftig frenk-
ling avhushllsarbetet.

Vi har upplevt en teknologisk utveckling som gynnat personer med hg produktivitet

framfr personer med lg produktivitet. De tekniska frbttringarna innebar att kraven
p arbetskraften kade, vilket i sin tur innebar hgre produktion och skapade real-
lnekningar. Frgan blir d: kommer den hr positiva trenden fr teknik och jobb
att fortstta? Kommer datorer och robotar att leda till lika omvlvande frndringar
som under den industriella revolutionen eller kommer utvecklingen att bli ungefr
likadan som under de senaste decennierna? Den frgan finns det inget svar p nnu,
men i den hr rapporten stller vi oss frgan om det kan vara s att vi str infr en
svr vergngsperiod och att utgngen till avgrande del beror p hur vl myndig-
heter och institutioner kommer att hantera utmaningarna.

9
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Idag betraktar vi ofta den industriella revolutionen som grunden fr vrt vlstnd.
Uppfinnandet av ngmaskinen, bilen och jrnvgen, kombinerat med rent vatten och
framstegen p det medicinska omrdet, innebar stora frbttringar av mnniskornas
tillvaro. Samtidigt utvecklades fackfreningsrrelsen och fungerade som en motvikt
mot fabriksgarna. Men ekonomihistorikern Joel Mokyr p Northwestern University
framhller nd att reallnerna inte frbttrades i ngon strre utstrckning under
merparten av perioden 17501850.

Idag str vi naturligtvis inte infr lika dramatiska utmaningar som under den indu-
striella revolutionen. Men den moderna motsvarigheten r jobbpolarisering och en
krympande medelklass, som dokumenterats i mnga OECD-lnder. Sverige skiljer sig
drvidlag t.ex. frn USA, dr reallnerna nstan inte har kat alls fr stora grupper,
genom att vi sett betydande reallnekningarna sedan brjan av 1990-talet, och
de sammanfll med att regering och riksdag genomfrde ett antal institutionella
reformer fr budgetterhllsamhet samt en oberoende riksbank

Jobbpolariseringen p arbetsmarknaden kommer sannolikt att fortstta i OECD-

lnderna. Hittills har jobbpolariseringen i Sverige i huvudsak karakteriserats av en
hg tillvxt av de mest kvalificerade jobben, en minskad andel av jobben i mitten
och ett konstant antal jobb i den nedre delen av frdelningen. Hur utvecklingen fort-
stter beror p vad beslutsfattarna gr. Om reaktionerna blir protektionistiska, och
underminerar digitaliseringens frdelar, kommer jobb- och lnepolariseringen sanno-
likt att bli betydligt mer omfattande. Lneskillnaderna kan med andra ord komma
att ka, i synnerhet eftersom kningen av antalet nya arbetstillfllen kommer att bli
lngsammare om de regelverk som utvecklats fr den analoga vrlden inte uppdateras
s att de tar hnsyn till de nya mekanismer och mjligheter som digitaliseringen medfr.
Det r srskilt viktigt att minska oskerheten i utformningen av regelverk fr att inte
hmma investeringar i ondan.

Ett stt att sammanfatta utmaningen blir drfr att sikta p en politik som bidrar till
att tillgodogra sig produktivitetstillvxten frn digitaliseringen med s f nackdelar
fr vlfrden som mjligt, dvs. att skerstlla en generell vlstndskning i vilken en
s stor andel av befolkningen som mjligt fr ta del. n s lnge har politiken i Sverige
inte tagit utmaningarna tillrckligt allvarligt och riskerna fr negativa terverkningar
p arbetsmarknaden har kat.

Ett stt att frst vilka policyutmaningar man str infr r att vrdera den kunskap
som finns kring utsikterna fr automatisering av jobb. Drivkraften fr sdan auto-
matisering r stark och kommer av en vilja att frbttra produkter och tjnster att
gra saker och ting bttre. Den drivkraften frstrks nu av att befolkningen ldras
snabbt i mnga OECD-lnder. I Sverige r det snart frre unga som gr sitt intrde
p arbetsmarknaden n antalet personer som gr i pension. Biltillverkarna i Tyskland
har redan problem med brist p arbetskraft i produktionen, vilket kar behovet av
att automatisera arbetsuppgifter och frbttra arbetsfrhllandena fr ldre personer
s att de kan frlnga sitt yrkesliv. Inom vissa sektorer eller regioner kan bristen p
yrkesutbildad arbetskraft vara stor medan det finns ett verskott p okvalificerad
arbetskraft p andra omrden.

Mnga mnniskor skulle sannolikt bli frvnade om de fick klart fr sig i hur hg
grad det r mjligt att automatisera arbetsuppgifter. Automatiseringen beskrivsofta
med en bild av en robot som sitter vid ett skrivbord, grna med en kaffekopp kanske
fr att figuren ska framst som mindre hotande. Men det r en vilseledande bild
eftersom automatiseringen, srskilt av mer krvande arbetsuppgifter, oftast kommer
att ske i molnet med fysiskt hemvist i opersonliga serverhallar som r frlagda till

10
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

srskilt utvalda platser med frmnliga villkor. Facebook har till exempel frlagt
ensdan datacentral till Lule i norra Sverige fr att kunna utnyttja den naturliga
kylning som de lga genomsnittstemperaturerna ger tillgng till.

Exemplen p automatisering av tjnster, till exempel i hotell och restauranger, blir allt
fler. D sjlvkrande fordon infrs kommer behovet av yrkesfrare att minska kraf-
tigt. Automatiseringen kommer ocks att fortstta inom finanssektorn och frskrings-
industrin. Datorprogram och tjnster i molnet konkurrerar i allt hgre utstrckning
om arbetsuppgifterna som medelklassen utfr. Ett hollndskt frskringsbolag har till
exempel rationaliserat handlggningen av frskringsrenden genom att eliminera en
rad mnskliga faser och det har bland annat resulterat i snabbare handlggningstider
och njdare kunder. Banktjnster passar srskilt bra fr digitalisering, eftersom en stor
del av arbetsmaterialet bestr av siffror och mnga mellanled frsvinner.

Automatisering av jobb handlar om att dela upp arbetsuppgifterna i olika moment,

ungefr p samma stt som den process som ledde till outsourcing av jobb till lg
inkomstlnder i Asien och andra delar av vrlden. Nr arbetsmomenten definierats
p lmpligt stt kan de automatiseras och frbttras, ofta i molnbaserade tjnster
eller outsourcas till delningsekonomin via digitala plattformar. En nyhet r att ven
mer kvalificerade arbetsuppgifter kan automatiseras, till exempel juridiska utred-
ningar, administration och betygsttning.

Svrigheten att programmera arbetsuppgifter som krver kognitiv frmga eller kreati-
vitet har rests som begrnsningar fr automatiseringen av arbetslivet, ngot som frts
fram i synnerhet av professor David Autor frn MIT, en av vrldens frmsta forskare
p omrdet teknik och arbetsmarknader. De flesta mnniskor vet till exempel hurman
cyklar, men att skriva ner reglerna fr den aktiviteten har visat sig vara svrt. Ett
annat exempel som ofta beropas r svrigheten f program att knna igen fysiska
freml med utgngspunkt frn bestmda egenskaper, till exempel att en stol har ben
eller att en katt har morrhr. Mnniskor kan inte lngre sl datorerna i schack eller
Jeopardy, och det r en falsk trst att mnniskor till exempel fortfarande r bttre
p att inte frvxla exempelvis en trafikkon och en stol. Att lsa sdana problem r
frsts en utmaning fr programmerare, men om de utgr overstigliga hinder eller
ej har nnu inte klarlagts. Maskiners frmga att lra sig sjlva har idag kommit s
lngt att programvara kan hrma mnskligt beteende ven utan att frstelsen fr
syftet har programmerats. Om frmgan att digitalisera undermedvetna frdigheter
kommer att mta overstigliga hinder terstr att se. Kanske undermedvetna kun-
skaper inte utgr en ogenomtrnglig mur som automatiseringen inte kan ta sig ver,
utan snarare ngot som kan omformas eller kringgs och omdefinieras.

Redan nu brjar vi sknja framsteg i denna riktning. Till och med kreativt arbete,
som tidigare var frbehllet mnniskor, pverkas numera ocks av ettor och nollor.
Datorprogram har till exempel skrivit musik som spelats p konserter. Till och med
de som sjlva driver frndringen programmerarna som skriver koden som auto-
matiserar jobb har brjat hamna i riskzonen. Det beror p att en stor del av koden
fungerar p samma stt som delarna till en IKEA-mbel, grundkomponenterna kan
man hmta frn hyllan, utan att teruppfinna dem varje gng.

Mjligheterna till automatisering har kat dramatiskt, men det innebr inte att vgen
till automatisering r s gott som sjlvklar p alla omrden. Det finns starka institu-
tionella trgheter som frdrjer automatiseringen men ocks andra faktorer som r
viktiga och ibland dominerande. Arbetsuppgifter automatiseras inte om det inte finns
ngon vinst att hmta. Och vinsten beror p kostnaderna fr automatiseringen jm-
frt med kostnaderna fr mnskligt arbete, efterfrgan och regelverk.

11
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Vilka r sammantaget de krafter som med strsta sannolikhet frmst kommer att
pverka tempot fr automatiseringen? ldrande befolkningar och brist p kvalificerad
arbetskraft inom vissa omrden kommer att ka efterfrgan p automatisering. Stel-
benta arbetsmarknader har samma effekt, de gr det mer attraktivt att antingen auto-
matisera arbetsuppgifter eller att kpa dem frn delningsekonomin. En kraft som
verkar i motsatt riktning fr ett litet land som Sverige r frgan om skala. Frdelarna
med automatisering r beroende av investeringskostnaderna och p en del omrden
med hga investeringskostnader kan automatiseringen komma att g lngsammare.
Nettoeffekten av de olika krafterna r omjlig att frutsp och kommer att bero p
mnga olika faktorer, bland annat vilken politik som frs p central och regional niv
samt hur arbetsmarknadens parter agerar. Men att dra slutsatsen att sm lnder r
skyddade frn automatisering r ett misstag. Mer proaktivt r att inse att det kan
finnas mjligheter fr entreprenrer att utveckla smarta metoder som tar hnsyn till
de lokala frhllandena i Europa och p andra platser, innan digitala fretag med
hemvist i Silicon Valley knackar p drren.

Med en protektionistisk policy-respons och lngsam reformering av regelverken

kommer vlfrdsvinsterna att bli svagare eller riskera att helt utebli. Vi skulle till
exempel sannolikt f hgre teknikdriven arbetslshet, strre inkomstskillnader och
lngsammare produktivitetstillvxt. Om ekonomins frmga att frflytta resurser
mellan olika sektorer strs kommer anpassningen att bli svrare, i synnerhet fr
outsiders p arbetsmarknaden.

Men det finns tgrder som kan underltta en smidigare anpassning p arbetsmark-
naden och bidra till att man tillgodogr sig frdelarna med digitaliseringen
Att ta ut lgre skatt p (mnskligt) arbete. Hg skatt p jobb i Sverige strker
det redan starka incitamentet att automatisera arbetsuppgifter ytterligare. Skatte
avdragen fr hushllstjnster br utkas, inte minskas.
Att minska oskerheten kring de regler som pverkar delningsekonomin. Det ger
flexibla mjligheter till justeringar och lgre risk fr hgre strukturell arbetslshet.
Att skapa strre mjligheter fr livslngt lrande. Ett lngre yrkesliv och snabba
tekniska frndringar kar risken att vissa kompetenser tidigare passerar bst
fre datum. Uppdatering av yrkeskompetens kommer att bli av stor betydelse
fratt minska risken fr ofrdelaktiga framtidsutsikter p arbetsmarknaden.
Att minska skillnaderna i socialfrskringsskydd mellan anstllda och egenfretagare.
Att infra principer fr reglering av den digitala ekonomin som kan anvndastill
att snabba upp reformerna p mnga skilda omrden, och drmed minska behovet
av att utarbeta srskilda regler fr varje omrde. Detta krver ett nra, smidigt
och kontinuerligt samarbete mellan jurister och ekonomer. Fr vi inte det, kan
den lga produktivitetstillvxten bli seglivad och demografins och statsskuldens
makroekonomiska motvindar fortstta att dmpa tillvxten.

Den digitala revolutionen kommer sannolikt att frbttra svl livskvaliteten som
effektiviteten p arbetsplatsen och fortstta att frbttra vr fritid. Men ingenting r
sjlvklart nr det gller hur smrtfri frndringen kommer att bli. Institutionernas
reaktion r en nyckelfaktor fr kad produktivitetstillvxt utan risk fr hgre arbets-
lshet och vxande klyftor.

12
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Extended summary
Technologys impact will feel like a tornado...No government is prepared for it.
Leader, The Economist (2014b).

Digitalization is affecting most of human interaction in one way or another and our
institutions are belatedly trying to catch up. No sector will be impervious to changes
and the impacts will be felt throughout the private sector, the public sector, and areas
hitherto less affected by technology, such as cultural institutions.

There are widely different beliefs about what this means all from business as usual to
a tornado whipping through an office landscape on the front cover of the Economist.1
Is digitalization in which the robots finally take over the end of jobs or will job crea-
tion keep pace with job destruction? Will inequality rise further and will income growth
be concentrated to a few winners-take-all, who acquire enormous wealth, while others
may work freelance and on-demand with more precarious and lower income streams?
We will argue that the outcomes will crucially depend on the policy responses.

This report is aimed at policymakers and others who are interested in the possibilities
and challenges of digitalization. It sketches the qualitative economic effects of digital
ization and aims to fill the gap between the academic literature and think tanks. The
think tanks and management consultancies have written extensively about the Internet-
of-Things (IoT), big data, 3D printers, and the sharing economy, in which people
and firms conduct business via digital platforms. All of these developments are likely
to be significant, but the sharing economy has, in our assessment, the biggest potential
to change how the economy works in the next few years and be the most disruptive.
Although some work in academia is discussing the economics of digitalization,most
scholars are preoccupied with understanding the past effects of technology. In bridging
these two perspectives, the challenge for any reader is to decide what aspects of histor-
ical experiences apply and what to make of all the hype, of which there is, alas, plenty.

Digitalization is one of several trends affecting the world economy and it is important
to stress that there is nothing inevitable about the welfare consequences in the years to
come. The outcomes good or bad will depend on the choices of institutions and their
ability to adapt to technology, demography, globalization, and other factors. Indeed,
akey insight from research is that nations that have managed to become rich have had
institutional features that supported incentives for value creation while ensuring that the
ways insiders and special interest groups can extract monopoly rents are limited.2 The
strongest driving force behind inequality is probably not technology, but the response
ofinstitutions when they erect barriers to entry, especially in the labor market.

Technological changes are coming in an era of many other undercurrents in the world
economy. Although there are some doubts about the quality of official statistics, pro-
ductivity growth has slowed in OECD countries. Other broad changes are clearly
apparent: populations are aging and public debt and unemployment are high in the
aftermath of the financial crisis. This implies that the welfare consequences of poor
choices or no choices at all may be particularly pernicious and have long-lasting
social and economic effects. Are our institutions well-poised for the challenges? Instead
of answering this question directly, we will use three scenarios below to illustrate the
possible economic consequences for productivity growth, employment, and income ine-
quality. We argue that the outcomes will crucially depend on policy responses. If the

1
See the Economist (2014a).
2
See, for example, Acemoglu and Robinson (2013).

13
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

responses lean more towards protectionism in various forms, inequality and unemploy-
ment are likely to increase (see scenario 2 below in Box 2. Scenarios for Sweden). We
also need to reduce the discrepancies in access to social security between employees and
the self-employed.

It is undeniably the case that our economies have been able to adapt to a great many
changes in the past, moving from being primarily based on agriculture to mass pro-
duction and the increasing dominance of services. In hindsight, all these changes
have led to increased prosperity for all of society, but life was tough for many people
amidst the upheaval. The economic historian Joel Mokyr noted that welfare did not
improve much between 1750 and 1850.3 Today, modern social safety nets and stand-
ards alleviate some of the burden of change on individuals and our economies have
developed ways to share risks among groups and generations. Among OECD coun-
tries, Japan may be the only one where younger generations are not better off than
older ones due to two decades of poor growth and the rapid aging of the population,
combined with strong rights for insiders.4

Technology may now change the way risk is spread throughout the economy, further
widening the gap between insiders and outsiders, incumbents and startups. If we are
going to learn one lesson from economic history, one way to formulate the policy chal-
lenge is to make the period of structural change as smooth or at least as unrugged as
possible while ensuring conditions for productivity growth, which is especially impor-
tant given the previously mentioned macroeconomic headwinds. In terms of social wel-
fare, the key challenge is how institutions respond to maintain incentives forinnovation
and job creation while not giving in to special interest groups and protectionism. The
outcome of inequality may hinge on the way these challenges areultimately resolved.

Speed of technological change vs institutional inertia

The dotcom bubble that burst a decade ago was not dissimilar in pattern to the hype
that often surrounds new technology. Surely there are reasons to be skeptical today that
significant changes are underway and that the pace may be swifter than during previous
periods of transformation? A case in point for being wary of hype was the spectacular
failure of the Swedish company Boo.com just after the millennium shift. Headquartered
in London, the company tried to sell designer clothes via a digital platform and was
thus one of the first to attempt something that has now become commonplace, but the
market was not mature enough and the conditions for success not yet in place.

Despite the experience of Boo.com and other startups that failed, there are good
arguments that things are different this time round. Boo.com failed at something that
has now become ubiquitous via, for example, Zalando, Wish, and other e-commerce
platforms. Today, we have an expanding number of digital companies with global
reach. Similar to all startups, some of these digital firms are struggling, some are
failing, but a few are extremely profitable. The capacity to quickly grow the scale of
a business has become key. As expressed by Reid Hoffman, co-founder of LinkedIn:
First-scaler advantage beats first-mover advantage.5 In contrast to the situation
about a decade ago, a number of key conditions have evolved and matured:
The emergence of large platforms with standards that attract consumers and
producers alike (the web, Apple with iOS and Google with Android).
Trust mechanisms for digital transactions, both for goods and services.

3
See Mokyr (2004).
4
See IMF Article IV consultation with Japan from 2012.
5
Hoffman (2015).

14
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Digital payment systems with low transaction costs.

Ubiquitous use of smart phones and tablets.

The world is now ready for transactions with digital goods and services. The changes
are mainly consumer-driven. Consumers search the internet, use their smart phones,
and share information, pictures, and experiences, much of which has implications for
commerce or social discourse. Successive generations are likely to increasingly use
digital tools for consumption and leisure; this means that consumer-driven activism
will remain a major force and may continue to increase in importance. Each new ser-
vice only needs to be available in one of the major app stores and/or on the web to
reach a large number of consumers. The best comparison with earlier technological
change is with electricity, also an all-purpose technology that allows a lot of other
machines to operate. The difference with electricity is that digital technology can dis-
rupt businesses more or less continuously: expanding the grid and lighting the way
for modern appliances and factories, after which many of the steps were evolutionary
rather than revolutionary. It has also been argued that digital technology is the only
all-purpose technology that has increasing returns to scale.6

Firms that do not adapt to changing consumer demands on products or information

risk falling behind. At the same time, our institutions are built around a slower world,
with inertia in laws and labor contracts. Most of the inertia is probably good, espe-
cially when it comes to democracy and core features of how rules and social welfare
are applied. But the labor market has developed a set of rigidities that are now under
assault from digitalization. Apart from digitalization, it would still be under pressure
to change from the rapid aging of populations and the need to extend working lives
to finance public welfare. Digitalization simply adds to the urgency of institutional
reform. The challenge is to maintain those elements that are conducive to stability
while not holding back growth and innovation.

Digitalization provides many new ways to circumvent rigid rules and sidestep regu-
lation that has been built up over the years. Many of the rules may be there for good
reason, especially consumer protection, but quite a few serve to protect insiders and
protect monopolies from competition. This leads to higher prices for consumers and
makes it harder for young people and the unemployed to find jobs. There will be
an assault on much of this rigidity, especially when digital solutions can be used to
explore pockets of existing inefficiencies and unused resources to create profits. These
pressures on institutions and firms to adapt will come in force as they are driven by
profits that can be remarkably small per unit sold, as long as the volume of sales
through network effects is high. In short, the reason change is likely to be rapid is
that profits will be in the drivers seat and the technology not only allows it, but is
actually built into the accelerator. The ultimate outcome for productivity growth,
the labor market, and inequality will depend on how governments and regulators
respond to these challenges.

We find it unlikely that digitalization will imply mass unemployment where robots
take over many if not most jobs. But not everyone agrees with this. There are
respected researchers, famous scientists, and business mavens who contend that robots
will replace labor on a massive scale: essentially, the end of work.7 The arguments
will be discussed at some length and are briefly summarized below in Box 1. Summary
of arguments, but the main lesson from economic history is that new jobs (some which

6
See Edquist and Henrekson (2006).
7
For example, physicist Stephen Hawkins espouses this view; Bill Gates, one of the founders of Microsoft has also
expressed asimilar position.

15
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

we cannot even imagine) develop all the time. The challenge is instead that bad policy
responses can make the period of transition tough and the adjustment period may
result in higher structural unemployment and rising inequality.

The many different views on what digitalization implies are probably also leading to
some confusion. On the one hand, people are being asked to work longer and there
are shortages of skilled labor in many OECD countries, including Sweden and Germany.
On the other hand, the robots may be taking over. What should we make of this? In
this report we try to address these issues and hopefully reduce some of the confusion.
Essentially, we are seeing the long-run trend of aging populations that is bringing
challenges for financing public welfare potentially colliding with the advance of tech-
nology. If robots indeed were to take over most work from humans, which we think
unlikely, the existing ways of financing public welfare would not work. Significant
government revenues come from taxing labor. If, on the other hand, jobs continue to
exist, we will have to extend our working lives to finance welfare, since there will be
fewer young people of working age compared to those in retirement.8

Box 1. A summary of arguments for and against the end of jobs

Arguments that support structural change as usual

Jobs have not disappeared despite vast changes in technology that have transformed the
economy from agriculture to manufacturing and services.
Tendency to underestimate the complementarity between humans and machines that allows
for more productive work rather than replacing labor.
Cars, the steam engine, electricity, clean water, telephones etc. have been more disruptive to
way of life and have improved quality more than existing digital innovations.
Significant inertia in institutions and regulation slow rate of adoption and may undermine new
business models.
Internet of Things and big data have many elements of legal uncertainty that will delay benefits.
Preferences for human interaction in schools, hospitals, etc.

Arguments that support significant transformation but not the end of jobs
Digitalization has strong network effects, marginal cost is small or zero for many digital services.
Transaction costs can become much lower (on-demand economy, 3D printers).
Digital knowledge is non-exclusive and information spreads seamlessly.
Many sectors affected more swiftly and simultaneously.
Demography and the problem of mismatching in the labor market increase incentives to automate.
Digital platforms can match buyers and sellers even for small goods and services, and similarly
for investment capital to find entrepreneurs.
Sharing economy provides more flexibility in labor markets and stronger competition for labor.
Rapid advancement in intelligent software that can replace humans in analyzing non-
structured data, perform empirical work and write texts that are hard to distinguish from
thosefrom written by the human hand.

Arguments that support the end of jobs

Machine learning is advancing rapidly.
Human ability to adapt is slower than the increase in computing power that can be used to
solve analytical tasks and apply tacit knowledge.
Unskilled labor will have lower productivity than machines and hence be unemployable or
have low real wage growth.

8
See Blix (2013a, b).

16
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

1. The labor market

What has happened?
Most people may not be aware of quite how far software has come in making inroads
into human activity. Although self-driving cars have been on the radar for some time,
but surely in other areas, especially for highly skilled tasks, the human monopoly still
stands? That computers have won against the best chess and Jeopardy players in the
world may be vaguely familiar to many people. But the full range of human activity,
especially creativity, writing, and analysis, cannot be coded into software they
require cognitive and tacit skills that are beyond computers right?

While artificial intelligence is not in sight in the near future, software has become
sophisticated enough to replicate humans in a vast array of areas, including highly
skilled ones.9 Due to the longevity of Moores law of increasing computer speed,
processing power is now so powerful that in many tasks previously reserved for
humans, such as writing and research, computers are indeed faster and more accu-
rate; increasingly, they can also write with style and it is no longer evident what is
computer-generated or created from human hand.10 The popular notions from movies
and books about walking robots doing ominous things is misleading. Work can now
be outsourced to the cloud, either to clever software or to a global labor market via
digital platforms.11 The modern replacement for an office worker is not a humanoid
thing sitting in a chair with a coffee mug (for our benefit). It is one of many stacks
of hardware in an anonymous server hall, located in a more or less remote area with
good fiber connections. For example, Facebooks servers in Sweden are located in the
northern city of Lule, where the climate helps keep energy costs down.

One of the foremost scholars today in the field of technology and jobs, MIT professor
David Autor12, argues that the fear of automation underestimates two forces: the degree
of complementarity between human and machine, that is, how technology makes us
better at a wide range of tasks; and the difficulty of automating tacit knowledge. Tacit
knowledge refers to the things we can do without explaining exactly how, such as
riding a bicycle or recognizing a chair from thousands of pictures. The argument is that
we cannot automate an activity unless it follows specific rules that can be transcribed
incode.

On the question of complementarity there is a lot of evidence that we will discuss

below, of which some raises concerns as to the number and kind of jobs that will
remain. But the issue of tacit knowledge is of a different kind, something between a
philosophical and practical impediment to automation.13 It should be clear, however,
that the idea that tacit knowledge cannot be coded is an assumption and not a lawof
nature. For example, some years ago it was widely assumed that autonomous vehicles
were impossible because driving includes too many elements of tacit knowledge. Whether
automation will hit an insuperable obstacle when it comes to tackling tacit skills remains
to be seen. Rather than being a brick wall beyond which automation cannot venture,
tacit knowledge might be reshaped or subject to circumvention and redefinition.

Machine learning is a prime example of ways to avoid the difficulty of coding tacit
knowledge. Machines can be taught to imitate and learn from humans and from
observing physical events, and this an area of rapid progress and development. The
software today is already clever enough to replace a lot of human activity but may

9
For an overview of advances in artificial intelligence and where we stand today see, for example, Bostrom (2014).
10
See Clerwall (2014).
11
See, for example, OConnor (2015b).
12
See, for example, Autor (2014).
13
For a discussion of measurement issues on tacit knowledge, see Ambrosini and Bowman(2001)

17
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

beslowed by obstacles other than technology (more on this below). For example,
while there are driverless trains in Asia, pilotless airplanes are not yet in the real clouds.
The technology is already here, but it is unclear whether there is an appetite for it at
least for now. But if pilotless planes are safe but much cheaper than the regular variety,
this may change too. We cannot easily know in advance what activities will be accept-
able for automation and exploring these limits is a challenge for entrepreneurs. It is
likely that public preferences will change over time for what kind of activity we are
comfortable automating. While the World Values Surveys14 indicate strong beliefs in
core values, especially in the importance of the family, beliefs about how we interact
with each other and how we consume goods and leisure are more likely to be malleable
and change with technology.

One of the more significant changes from digitalization is the rise of the sharing
economy, which despite its name, is more about market forces and entrepreneurs
finding new business opportunities. Other, perhaps more descriptive terms are the
on-demand economy and the gig economy, but in this report we will continue to
use the term the sharing economy. With digital platforms, even minor demand for
goods and services can be matched to supply at low costs. The sharing economy is a
major macroeconomic change. Although neighbors and colleagues may have bartered
goods and favors throughout human history, this activity is now being reshaped in a
way not dissimilar to how industrial manufacturing forever changed the landscape of
production. Knowing that there is a person asking for a business or household service at
a specific time is no longer the lottery it used to be in matching supply and demand.
As a consequence, we will see more jobs in sharing economy with more flexibility but
also more insecurity, an issue we will return to further below.

Further automation
The automation of low-skilled and manufacturing jobs has become a driving force
and perhaps also a potent symbol of productivity growth in market economies. It does
not require much imagination to see that driverless cars and vehicles will change a
lot of activity; there will be less need for taxi drivers, truck drivers, driving schools,
insurance companies, and service personnel in hotels and restaurants. On the upside,
doctors will have fewer traffic accident victims to contend with. But there are likely
to be ripple effects, many of them unpredictable. Of the more predictable ones, we
may venture to guess that the rise of autonomous vehicles may affect rural areas.
With fewer humans delivering goods to rural areas, a lot of restaurants and hotels
will have fewer guests, putting a strain on their business.

In agriculture, some human activity has hitherto remained in handling sensitive fruits
but this is also beginning to change. Machines are now sophisticated enough to take
over such tasks. Logistics warehouses, such those of Amazon, Wal-Mart in the US, and
Clas Ohlson in Sweden, have become increasingly semi-automatic with more and more
functions taken over by machines. There are now machines that can produce two
hundred or more hamburgers in an hour; there are fully-automated sushi restaurants
and hotels in Japan where service personnel have been largely replaced by machines
in much the same way that airline check-in is more about interacting with machines
these days than with harried clerks.

It is true that many of the jobs that have disappeared are not necessarily the good
ones. Indeed, much of modern work has improved in construction, mining, and

14
See Halman et al. (2008).

18
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

other physically demanding occupations as machines have replaced human labor.

Seeing children in developing countries coming out of dark holes in the ground after
grueling work to extract coal or minerals is a reminder of this.

Some clues for interaction between machines and humans in modern economies
come from banks. For bank tellers, there was an initial fear that their jobs would dis-
appear with the introduction of ATMs, but in fact such jobs increased in proportion
and required more skill as tellers began providing more services and advice to clients.
Complementarity between machines and humans has been in evidence in other areas as
well, such as medicine, where doctors have used instruments, computers, and machines
to improve their clinical and surgical skills. But in some areas, such as clerical work,
computers have displaced labor the same way that artisans jobs were reduced at the
dawn of industrialization. This has been the case throughout history with some jobs
displaced by machines and new ones created with higher skills and elements of comple-
mentarity.

Digitalization has now reached such a point of speed, maturity, and saturation that its
effects may come faster and be more widespread than any other technology to date. The
most relevant comparison is with the introduction of electricity, also an all-purpose
technology, in the 20th century. Although electricity also affected many sectors at the
same time, one difference is the speed of adoption. This may hamper the ability of
workers whether skilled or unskilled to switch from a sector in decline to one on
the rise.

There are strong incentives to discover which parts of a global value chain can be
automated. For large Swedish firms that are already mostly geared towards exports,
the constant striving for efficiency is already in their DNA. For them, outsourcing
or automating parts of their global value chains is second nature. The difficulty of
replacing workers who retire due to aging populations and the problem of skills mis-
match in the labor market may well strengthen incentives to automate. The incentives
to automate are even stronger in Sweden and other countries with rigid labor laws
and high taxes on labor.

The new and forceful changes underway imply that more middle class jobs are being
affected by digitalization. For example, media companies (film, music, and newspapers)
have long been exposed to the economics of the low marginal cost of digital distri-
bution and the challenge of charging money for content. In the next wave of auto-
mation, thousands of texts are being automated, everything from corporate earnings
reports to football scores.

Likewise for legal and business research, software is automating the sorting, analysis,
and presentation of huge amounts of data.15 For legal applications where precedents
are particularly important, software can now comb through millions of documents
that would have taken many years to go through. Thus far, many of these concern
entry level jobs, such as paralegals, but the technology is already available to grade
written exams and essays in schools, for instance.

The tasks performed by computers are climbing the skills ladder. IBM and Google are
buying health care companies with a view to applying software to improve diagnos-
tics.16 With vast databases at their disposal and sophisticated data analysis routines,

15
For an overview of the tasks that can now be performed by machines, see, for example, Ford (2015a).
16
See Crow (2015) and Lohr (2015a, c).

19
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

software has the potential to improve medical assessments. The volume is simply too
vast for any human doctor to follow broad medical research outside their own domain.
Using statistical techniques and inference, the software may give predictions and recom-
mendations based on information that surpasses the capability of any one person.

Would we trust software recommendations in medical, legal, and other spheres?

Many people likely dislike black box answers, however mighty and prescient they
may seem. Therefore, developments in IBMs WatsonPaths that show not only the
recommended option, but also the steps taken to arrive at the conclusion may pro-
vide the trust boost required for the next wave of automation in highly skilled work.

In thinking about the future of jobs how many and what kind we should look
to the past and determine what experiences are relevant. In economic history, there
are useful lessons on the effects of structural change and what challenges should be
expected. While the first part of the Industrial Revolution was deskilling, most of
the 20th century saw a combination of increasing technological prowess concordant
with higher skills. But recently, in the last few decades, we have also seen increasing
polarization in the labor markets of OECD countries: middle income jobs have been
decreasing whereas the overall shares of the labor market have been increasing for
both lower and higher end jobs. This has not occurred in Sweden in the same way;
Sweden has instead seen an increase of higher paid jobs. Moreover, Sweden has not
experienced the kind of wage polarization that has taken place in the US, where many
lower and middle income jobs have had little or no real wage growth. In fact, quite
the opposite is the case and real wage growth in Sweden has been strong in the last
two decades.

Labor market polarization may well give rise to tensions in any economy, but, as
emphasized by Raguran Rajan, former chief economist at the IMF and now governor
of the Central Bank of India, the rapid increase of US house prices may have contrib-
uted to making middle income earners feel like they obtained part of the wealth created,
at least until the financial crisis.17 The US is now seeing increasing tension, especially
among lower paying jobs, such as in fast food chains. A number of people in low-
paying jobs in the US are also receiving food stamps, thus in effect providing a subsidy
from the government to low-wage employers, for example in fast-food restaurants.18

Sweden has had a very different experience, having reformed its institutions in the
1990s in the aftermath of a collapsed real estate bubble. With radically overhauled
institutions and solid public finances, Sweden was able to cope with the financial crisis
in 2007 and the increase in unemployment was nowhere near the large shocks in
the early 1990s. Though the fallout from the financial crisis was challenging, Swedens
reforms in the 1990s paved the way for institutional resilience to shocks that served
its people well.19 In contrast to many other OECD countries, public finances in Sweden
remain sustainable and public debt was low throughout and after the financial crisis.
Indeed, Sweden was one of the few EU countries that was able to maintain public
debt well below 60 percent of GDP, as mandated by the original EU:s Stability and
Growth Pact.

The reforms to the budgeting process in the 1990s were key elements that helped
achieve this outcome; especially significant were the medium term expenditure ceiling
for government spending and improved top-down budgeting in parliament, which

17
See Rajan (2010).
18
See, for example, Cohen (2015) and Jacobs et al. (2015).
19
See, for example, Calmfors (2013); Heyman et al. (2015b).

20
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

were valuable tools for reducing the strong deficit bias otherwise inherent in public
spending. Moreover, labor market reforms initiated after 2006 increased the incentive
to work by reducing income taxes and strengthening the weak control mechanisms in
benefit systems to reduce misuse.20 Another key reform was giving independence to
the Riksbank in 1999 for monetary policy to pursue the goal of price stability. One
lesson from Sweden is thus that institutional reform that improves the way different
forces interact is crucial to managing the economy, and especially so in times of crisis.

But it is a mistake to describe the institutions as having reached a zenith after which
they can handle anything. It will always be difficult to choose a tradeoff between dif-
ferent goals in the economy, but the institutional set-up can mitigate and facilitate. In
other words, institutions can make it easier to pursue good long-term policies but this
does not mean that good enough institutions make it possible to have policy on auto-
pilot. Ultimately the hard choices have to be made and defended by policymakers in
government and elsewhere.

The labor market in Sweden is characterized by centralized bargaining between the

labor market partners (unions and employers organizations). The government has no
direct control over the wage-setting process, in contrast to countries where govern
ment is involved directly in public sector pay.21 Though employers are permitted to
lay off workers, they are required to negotiate departures with the unions based on
first-in, last-out rules. The OECD has often recommended Sweden to increase the
flexibility of the labor market. The current system benefits labor market insiders with
indefinite contracts, making it harder for others to get jobs; it also reduces the incen-
tive to switch jobs even when economic conditions change, for example if a sector
is in decline and prospects are better elsewhere. Moreover, people may be less prone
to voluntarily giving up the safety of an indefinite employment contract even if their
personal preferences were to suggest switching jobs or sector.

The market economy often finds ways to deal with inflexibility, sometimes at great cost
and at other times more easily. In Sweden, specialized employment firms that supply
temporary labor on demand have become popular across the range, from clerical ser-
vices to management. Overall when it comes to temporary workers, Sweden has among
the weakest protections for workers in the OECD. All told, this means that Sweden has
one of the most pronounced dual labor markets among OECD countries.22

Limits to automation of jobs and implications for the Swedish labor market
Recent work has highlighted that a large share, 50 percent, of current jobs may be
automated.23 Though it may be obvious, it is still worth stressing that the rate of job
destruction presents only part of the story of how the labor market develops. That
jobs disappear and that the nature of existing jobs changes is nothing new. New jobs
are created all the time and so the crux is really the rate at which new jobs appear
compared to the number of people looking for work and whether those people have
the right skills. Moreover, there are also countervailing forces that may slow down
the rate of automation.

20
See the Spring Budget Bill 2014, Appendix 4.
21
Public sector pay in Sweden follows a mechanical formula that uses productivity gains in the private service sector as a
benchmark for mandating cost efficiency in salaries and operations.
22
See Cahuc (2010, pages 150153, and OECD (2015b).
23
See Frey and Osborne (2013) and Flster (2015).

21
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

First, the fact that technology allows the automation of many jobs does not neces-
sarily mean it will happen. The fundamental driver behind automation is the will
to make something better, leading to higher quality, new goods or services, or lower
costs. In some areas, automation will be costlier than human labor. For example,
while new cars have windshields inserted by robots, when it comes to windshield
repair, a human will typically do the work. In some areas, there may be little demand
for automation. People may grudgingly accept having to go through automated
responses to reach a customer service representative, but there are many jobs where
automation will not be welcome or accepted, at least not in the immediate future.
Automation of education is an example, where teachers to a fairly large extent
could be replaced by software but parents may not be enthused about this and
organized labor might also hold back such a development.

Second, there are many other trends that affect demand for jobs that have little to
do with automation. As an overall driver of automation, demography may be signif-
icant. But in other areas, aging populations may increase the need for human work
in health care and elder care services. Granted, automation can perform more of
such tasks now than before; there are already robots that help the elderly in Japan
and there are trial versions in Sweden. But it is hard to see that these changes will be
major compared to the overall need within the foreseeable future. In other areas, such
as law, automation technology has made it possible to vastly reduce the number of
paralegals and junior lawyers and yet in Sweden, lawyers have increased as a share
of the labor market. There may be several explanations for this, increased complexity
of regulation among them.24 For example, the strengthening of requirements on
financial services has led to a need for much more work on compliance. Thus, other
changes in the economy can from time to time be more significant for developments
than technology alone.

Third, small open economies, such as that in Sweden, whose language is spoken by
only about ten million people, provide less scale for the benefits of automation than
the US and China. In many instances, it is not enough to take a US/UK technical solu-
tion and paste it onto a smaller country. There are institutional idiosyncrasies in law,
culture, and consumer preferences that need to be addressed. Some of the institutional
features reflect strongly held constitutional rules, the division of responsibility between
local and central government among them. For example, local government is respon-
sible for many services where digitalization has the potential to improve efficiency,
such public transport and health care, but also for systems that make it easier to find
and pay for parking spaces. But local autonomy also means that each area may have
its own special solutions, thus impeding the ease of developing software that aims to
be nationwide.

This issue holds true for most countries to a varying extent and can, of course, be
alleviated by voluntary agreements to apply common formats and standards. Never-
theless, reaching consensus may take time and be subject to interpretation or change
with ensuing uncertainty delaying action. While none of these are insurmountable
obstacles, they may slow developments because profits from scale take longer to achieve
or are more uncertain. The same money invested generates a lower return when the
scale is limited. Language is another issue, conferring an advantage on English and
other major languages. But to conclude that small countries are safe from automa-
tion from big platforms would be a mistake. A more proactive view would be that
there may be opportunities for entrepreneurs to build clever ways to account for local
idiosyncrasies, in Europe and elsewhere, before the big platforms come knocking.

24
This argument is made in Flster (2015).

22
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Finally, existing regulation and the threat of new regulation may stymie entrepre-
neurs. This point is discussed further below, but in this context it suffices to note that
for some firms, their very business model may hinge on small details in legislation and
on how the law is interpreted. For example, are platform providers in the sharing
economy accountable as employers, as a recent California court decision implied?
This would vastly limit, if not completely undermine, many actors in the sharing
economy. Likewise, who is liable in case of an accident caused by a 3D printed object,
a self-driving vehicle, or the loss of personally sensitive data (on health or illnesses)?

Taken as a whole, what does this imply for the speed of automation? On the one
hand, Europes more rigid labor markets make further automation attractive as a way
to increase flexibility. Moreover, demography makes automation attractive, as it may
be the answer to shortages of skilled workers. On the other hand, regulatory hurdles
and institutional idiosyncrasies work against Europe, at least for all but the biggest
countries where the lack of scale is not an issue.

But, as stressed above, the extent of automation will critically depend on the policy
responses. Creating conditions that are conducive to job creation in the private sector
will be key to smoothing changes. Recent research shows that job creation in Sweden
has been substantial. A study of Swedish data finds that 190,000 new jobs were created
in the period 19902009, the net result of about 3.4 million created jobs against
3.2 million eliminated.25 In an average year about one-fifth of the jobs in the labor
market were turned over, giving some indication of the economys ability to adjust
to demand and new technologies. Most of the jobs were created in the service sector,
while the manufacturing industry exhibited decline in employment.

2. Economic forces
Changing economic forces gaining hold
The sharing economy is lowering transaction costs across a broad spectrum of ser-
vices, all from self-publishing of books and searching the web for information toper
forming household tasks. This increases the flexibility of work, especially in European
countries where there is strong protection for labor market insiders. But as more people
work freelance or on-demand in the gig economy, they are also exposed to a higher
level of risk as the social security systems are tailored to full time work as employees
covered by collective bargaining.

Of all the forces of digitalization, the economic effects of the sharing economy
maywell be the fastest change and one of the more significant. Unlike evolutionary
improvements, the sharing economy is a big step that shifts the way we work and
organize our lives. One implication already evident is that there is less need toown
things when rental is but a click away. This also means fewer idle resources on standby,
as exemplified by cars that typically are parked for most of the day or unused rooms
in houses or apartments. Entrepreneurs are on the lookout for new pockets of unused
resources and overpriced goods, such as formal evening gowns or that drill you use
once or twice a year. They are also using technology to change the way demand for
household services is met by saving time and increasing convenience in areas such
asgrocery shopping, cleaning, or sending flowers.

Of course, these services are not new, but the ease of matching supply and demand
combined with low costs has major implications for the economy. A recent survey
by the Freelancers Union found that 34 percent of the American workforce has been

25
Heyman et al. (2013).

23
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

involved in sharing economy activities. In Europe, the sharing economy is trailing

the US but also becoming significant, especially so in the UK. Due to stricter labor
markets in Europe, the sharing economy may have even more long-run impact on
Europes rigid labor market unless new regulation chokes development.

One major implication of the sharing economy is that the raison dtre for ownership
is shifting. Why own a car if a seat is available on demand, only a click away? Why
buy a party dress and wear it only once or twice? Why not rent out spare capacity in
apartments? The incentives to use resources better are omnipresent, but are coming
tolife in ways not seen before due to substantially lower transaction costs.

Digital reputations on the rise?

A novel thing about the sharing economy is how technology is also demonstrating
new ways to develop trust in transactions. After all, renting out your home or a
seat in your car to a stranger requires a lot of trust. But just like the bumblebee, the
sharing economy manages to fly.26 Business acumen and technology combined have
provided ways to mitigate the risks and establish trust in such transactions. Using
various ways to monitor quality and reliability, technology provides ways for both
sides of a transaction to rate each other a two-way process of establishing reputa-
tions. Digital reputations are becoming as valuable as digital data and probably more
valuable/useful than analog forms of identification. After all, a drivers license may
say who you are, but it says nothing about reliability and trustworthiness. Indeed,
there are reports of people finding it difficult to get Uber rides after having obtained
negative ratings in previous trips.27

Digital reputations are growing in importance and so are ways to establish confidence
in online reputations. For example, if someone establishes a solid reputation in one
area, should that history be transferable to other areas? There may be market failure
here, in a not dissimilar way to how it is difficult to transfer pension accounts from
one bank to the other. Hence, governments might conceivably play a role in facili-
tating digital reputations. So far they have largely stayed away, probably for good
reason, but there may well be a coordination role to serve as a conduit for estab-
lishing standards. Why not a single digital market for reputations as well?

Digital technologies putting capital to better use

One other major implication of digitalization going beyond just the sharing economy
is that there is less need for capital to start firms. Of course, heavy industry still requires
substantial investment and advertising costs are set to remain significant. But a lot of
activities can be simulated on the computer, tweaked, and improved. With a decent
computer and an internet connection, entrepreneurs can reach consumers much more
easily today at lower costs. Architects have long been able to design houses in 3D
programs but the real power of computers is now taking another leap. For example,
architects can have more complete control of all aspects of design, everything from
the placement of electrical wiring to plumbing and calculating all the equipment
needed for production. Stylistic changes (type, color, etc.) can be consistently applied
across the design, not unlike the way text formatting is handled automatically in
word processing software.

26
The comparison between the bumble bee and the welfare state is made in Thakur. (2003).
27
See, for example, Streitfeld (2015b).

24
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Distribution costs are also likely to decrease with the increasing use of 3D printing
and the rise of the sharing economy. The long ongoing trend to reduce the middle-side
of transactions in banks, industry, logistics, publishing, etc. will get another push.
Why store and ship expensive parts when they can be printed on demand? 3D printing
brings down distribution costs and increases the speed of both design and delivery. It
has applications in industry, consumer products, and health. 3D printers have been
used to build full-size houses in China and the Netherlands.28 But there are also appli-
cations of 3D printing that improve quality of life through better health care. For
example, 3D printing is bringing down the cost of customized prosthetic limbs. There
are also examples where surgeons make 3D replicas ahead of complicated procedures
in order to anticipate problems that might only otherwise be discovered while the
patient is on the operating table.29

Internet of Things (IoT), and big data have the potential to take much firmer control
of all aspects of production. Together with 3D printing, these are transformational
technologies that affect all aspects of production, but IoT progress may be slowed
by legal obstacles, most notably consumer safety concerns and issues of intellectual
property and patent rights (more on this below).

On the demand side, technology is changing the ways businesses reach out to con-
sumers with information and advertising. All online activities generate troves of data,
much of it very valuable to firms wanting to target specific groups. Indeed, personal
data has been called a new asset class to signify its rising importance, alongside tra-
ditional assets traded in financial markets. This has raised a lot of privacy concerns.
Telecoms operators have vast amounts of information about peoples movements and
can draw inferences about their lifestyles and habits from the data.30 Consumers are
giving a lot of the data away willingly to get free services such as email, networking,
and maps. But it may well be the case that people are not actually aware of quite how
much information they divulge simply by logging in and moving the cursor across
thescreen.31

The economic impact of personal data is that firms can target products with cam-
paigns with much higher accuracy. They can also give different prices to consumers,
lower prices to those about to exit without closing the deal while charging people
who are less sensitive to price more. The specter of price discrimination is increasing
and yet is a sensitive topic that firms are careful not to flaunt. On the one hand, there
are efficiency gains in using technology to find each persons reservation price above
which they would not buy. On the other hand, with their notions of fairness and
equal treatment, this may make consumers angry and firms need to be careful about
balancing the efficiency provided by technology with the human reaction to seeing it
in practice. There are reports of consumers being annoyed at being left out of cam-
paigns geared to affluent consumers. The issues are not new, but technology is putting
a new spin on things and adding to the urgency of dealing with consumer reactions,
especially in areas where consumers can easily shift from one supplier to another.

Microeconomic revolution fueled by digitalization has macroeconomic consequences

not yet understood in policy circles

28
See, for example, Davison (2015).
29
See Weintraub (2015).
30
In 2011, Malte Spitz, a German member of parliament, requested and received the data generated by his mobile phone
from Deutsche Telekom, which showed his activities in great detail; in Sweden rstadius and Larsson (2015) have done a
similar analysis.
31
See for example Sveriges Konsumenter (2014) and Bylund (2013).

25
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The changes discussed in the previous sector will have major implications for growth
and jobs. Entrepreneurs in the sharing economy are already affecting overall resource
allocation and increasing efficiency in existing areas as well as picking up new pockets
of tasks that were previously below the radar due to difficulty of matching and high
transaction costs. The rate of technological change is increasing compared to previous
periods and the effects are going to be felt in broad sectors of the economy.

All the changes discussed above will have potentially major implications for assessing
the stance of the business cycle and its effects on the economy, such as on employ-
ment and inflation. In particular, fewer resources in terms of capital and labor will
be needed to produce the same output. The qualitative effect of this improvement
isclear, but the magnitude and scope remains to be seen and will among depend on
several factors, not least the regulatory response.

One implication of this development is that there will be lesser inflation impulses
from a given level of resources in the economy. A car that is idle most of its life can
be used to a much higher degree; a 3D printer saves on storage and distribution. Less
capital produces more output. But the implications go beyond better use of existing
resources and may also have effects on productivity growth.

Many OECD countries have experienced lower productivity growth in the last few
years. Part of this may be due to measurement problems in the statistics32 and the
long-lasting cyclical effects of deleveraging following the financial crisis. But Professor
Robert Gordon at Northwestern University also highlights several headwinds that
dent growth, including aging populations and high public debt.33 It may be that IoT,
big data, 3D printers, and all the other new technologies in the pipeline will improve
potential growth over time and balance the headwinds. But this process may take
time and its speed will depend crucially on the regulatory response.

In the medium term, in the next decade or so, it is likely that the benefits of digitalization
will raise potential growth and thus the forces that may hold back inflation are likely
to be active in one way or another for a long period of time. Evidently, digitalization
is not the only medium-term force affecting inflation, but it may be one of the more
misunderstood by governments and central banks. The direct price effects on con-
sumer goods, especially electronics, are clear enough. But the indirect effects on the
economy on resource utilization are likely to be more significant and persistent. It is
a challenge for central banks and policy institutions to understand the conditions for
firms that affect the economy; otherwise policy may be based on shaky ground.

3. Economic challenges in a digital world

When it comes to the future and the effects of digitalization, institutional resilience
in Sweden will be put to a different test than during the financial crisis of 2007 or
the home-brewed crisis of the early 1990s. While inordinately challenging, the strong
focus and drama ensuing from these events contributed to making policy adjustments
easier. When it comes to digitalization, the effects on the labor market are rather sim-
ilar to those consequent upon the aging population: the changes come gradually, from
year-to-year, and there will probably be no big event that heralds a major change. It
may be more difficult for governments to change course under conditions of gradual
change even when the ultimate effects on welfare may be even more far-reaching.

32
See, for example, Coyle (2015).
33
See Gordon (2014).

26
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Challenges for governments regulation

Against the background of slow productivity growth in OECD countries, there
is some urgency about improving the structural features to not hold back growth
unnecessarily due to aging populations and public and private debt overhang. But
thepace of regulatory reform for product and labor markets has slowed in OECD
countries, including Sweden, in recent years.

It is of course not so that little regulation is always better. Regulation serves a key
role in establishing safety and trust in markets. Without adequate regulation and
standards, market economies would likely be constrained by lack of trust and higher
transaction costs. Nevertheless, a lot of regulation has expanded over the years and
likely over and above the need to protect consumers. The European Union is working
hard to improve the free exchange of goods and services. While important progress
has been made, the language and cultural obstacles in the EU make the remaining
regulatory hurdles, such as VAT reporting requirements and safety and information
standards that vary across 28 Member States, more burdensome than in the US.

Moreover, there are lingering protectionist forces aiming to raise entry barriers for
new firms, since this allows incumbents to enjoy monopoly profits by crippling
competition. In practice, the constant demand on regulators in standardizing and
increasing safety requirements tend to be more powerful than the interests working
to keep things simple. As well, complexity can also be the result of compromise at
theexpense of overall coherence. Thus, there is a bias in favor of creating ever more
regulation that adds to this complexity.

Digitalization is entering this fray of regulation that is already complex and raising
new issues that were not previously addressed or important. Granted, patents and
intellectual property are areas that have been discussed for a long time. Music and
publishing industries were the first to struggle against digital technology that allowed
easy copying and distribution. It has taken a long time for new business models to
emerge that allow content producers to charge money for their work, notably through
online music, newspaper, and book subscription services. The changes are still ongoing.

Two important regulatory challenges stand out among all others. First, there are
significant regulatory obstacles that need to be addressed in the EU Single Digital
Market. The key benefits of digital technology through low marginal cost and ease
of distribution are hollowed out by the 28 different regulatory environments in the
EU. It is not so much the different levels of VAT that are problematic but reporting
requirements, restrictions on data transfer across borders, and procedures. Rules on
safety notices also vary, as do rules on how to store and handle consumer data.

Second, the expansion of the sharing economy is hampered by the question of which
regulatory model to apply. With a minor change in interpretation or change in existing
regulation, a business model may even collapse. Throughout the economy there are
entrepreneurs busily figuring out how to make money and provide value to other
businesses and consumers. They are likely to take note of the treatment awarded to
Uber and Airbnb, the standard bearers of disruption in the sharing economy. The
most damaging aspect of regulatory uncertainty is this fog of unpredictability that
prevents new ideas from being implemented.

There should be some urgency to resolve the legal status of people who work in the
sharing economy and improve social safety nets for freelancers. In Sweden, the entire
structure of the economy is geared towards sustaining large-scale bargaining for full

27
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

time employees in agreements between the labor union and employers organization.
Should the number of freelancers increase substantially, the model may be put under
strain and its legitimacy questioned. There is a challenge in how institutions update
agreements and systems to manage the risks different individuals are exposed to
while at the same time promoting labor market flexibility that is needed to reduce
unemployment and remain competitive.

In relation to the sharing economy, various sectors are confronted with a myriad of
other regulatory question marks. A common thread is whether professional standards
should apply to more amateur-type activities and, if so, where should the line be drawn?
What safety standard should apply to Airbnb rentals compared to hotels? Should
small-scale activity be taxed at the same rate as large corporations? In peer-to-peer
(P2P) or peer-to-business (P2B) finance platforms, what liability do the platform holders
have in the event of default? Why do households face less favorable tax rules for P2P
loans than for listed financial assets held at a bank in the Swedish Investment Savings
Account?

One particularly onerous example of how legislation can actually work against safety
is that related to self-driving vehicles. It is widely believed that self-driving cars will
cause vastly fewer accidents and reduce the number of dead and injured. Yet, the ques-
tion of legal liability in the event of an accident is constraining this development.

If governments try to resolve digital regulation in each separate market, it is going

to take a long time. Moreover, technology evolves swiftly and new businesses may dis-
cover ways of doing things that might need to fall under regulation still not on the table.
For these reasons, governments should work hard to establish some basic principles
of regulation that can be applied to all areas. It is beyond the scope of this report to
make specific recommendations for such principles, but a few comments are possible.
First, safety should remain a concern but should be reasonable in terms of expected
outcomes, as is not the case with self-driving vehicles; second, goods and services
should become more neutral in taxation. For example, the differences between books
(a good) and e-books (a service) is flagrantly large in almost all countries. In Sweden,
the difference is almost twenty percentage points on VAT. This is an impediment for
many other areas as well, including 3D printers. Third, liability in the event of mal-
function or injury should be clarified in general terms. Fourth, zoning laws and other
municipal prerogatives need to be harmonized to allow benefits of scale.

With the establishment of high-level principles, the speed of adapting regulation to

specific areas that cater to heterogeneity would vastly improve. If, on the other hand,
different sectors in the economy get their own set of idiosyncratic rules with little
commonality, it stands to reason that the benefits of digital technology will be held
back, denting productivity growth. It is of the utmost importance that regulation does
not support rent-seeking by special interest groups and instead caters to innovation
and Schumpeterian creative destruction.

On a more specific note, the way the Swedish government organizes it work poses a
specific challenge for removing obstacles that hold back the benefits of digitalization.
The responsibility for digital matters rests with the ICT minister attached to the Min-
istry of Enterprise and Innovation. But most of the obstacles and challenges that may
slow progress for digitalization are under the purview of other departments, especially
the Ministry of Justice (data protection, legal liability, patents, and copyright), the
Ministry of Finance (taxation, financial regulation) and the Ministry of Employment.
Each of the ministries is very powerful in its own domain and unlikely to be easily

28
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

swayed. The benefits, costs and risks of digitalization need to be strategically assessed
at the highest level. To speed up regulatory overhaul that slows down digitalization,
it is likely that some ICT matters should be coordinated from the Prime Ministers
Office, at least for the years it takes to make more substantial progress on regulation.

Challenges for governments lifelong learning and digital skills

People are living longer than before, yet the norm of spending the early years of life
in education with comparatively less resupply of formal schooling in mid-life is per-
sistent. Granted, there are numerous courses in all sorts of subjects, covering the
gamut from business writing to computer coding and vocational skills. But with rapid
advancement in technology, there is a need for a more systematic replenishment of
knowledge throughout our lives. Correspondingly, we need to think about whether
some tertiary education lasts too long in relation to its utility in the job market. These
issues have become increasingly important in light of aging populations, but are
becoming more urgent with the advent of digitalization.

While skills supply is crucial to firms, more education is not necessarily the answer.
Indeed, Professor Alison Wolf at Kings College has argued that we invest too much
in education and that the quality in some areas is too low.34 Notably, one year in edu-
cation is also one year not in work, which puts additional burdens on financing social
welfare systems that were designed in times with a more favorable ratio of working
to retired people.

Instead, we need more education and skills that are in immediate demand in the labor
market to improve the matching between vacancies and jobseekers. We also need ways
to replenish that knowledge throughout life, not only so that we keep abreast of tech-
nology but also so that we are equipped to switch careers in mid-life, a point emphasized
in the report on demography from the Commission for the Future of Sweden.35 We need
better skills in general and more digital skills in particular, a point also emphasized
in recent report from the Swedish Digitalization Commission to the government.36
Arecent survey by Eurostat shows that a large number of citizens lack basic skills.37
To some extent, this involves a generational gap that exemplifies the need for contin-
uous skills development.

There are two distinct challenges. First, how can improvement of skills be better
organized to cater to the needs of firms? Second, how should it be financed? The
incentive for firms to finance learning for its staff is dented by the risk that employees
may leave and take a job elsewhere, taking their new skills with them, but this is also
the kind of agility required for efficient functioning of the labor market.

As regards the organization of learning digital skills, the public/private UK Tech

Partnership is a model that might be useful for Sweden to consider. With a relatively
small staff, the Tech Partnership coordinates education needs in existing firms through
co-financing shared by government and business. They also focus on improving digital
skills in schools, and especially on increasing the participation of girls.

34
See Wolf (2011).
35
See Blix (2013a, b).
36
See SOU (2015a).
37
See Figure 5.1 in Chapter 5 of this report.

29
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

To the extent that digital skills should become part of our general toolkit in the same
way as mathematics and science, there are arguments that the public sector should
finance some aspects, especially those that are more general in nature; the more spe-
cific the skills, the more they should be financed by firms or individuals. Where to
draw the line is difficult question, but the issue is well worth pursuing.

More lifelong learning in appropriate doses is also likely to benefit productivity growth
and make it less likely that people who stay in their jobs for a long time will become
unemployable elsewhere in the event of company restructuring.

Challenges for firms regulation, new competition and skills

Periods of rapid technological change accentuate several distinct challenges for firms:
adapting to regulation, updating skills and human capital, and readiness to change
the business model or core products if demand shifts.

The challenge of regulation is in some sense the mirror image of the discussion for
government above. There are distinct challenges for incumbents compared to digital
entrepreneurs. For entrepreneurs, the most basic issue is the continued viability of the
business model if regulation changes or is interpreted in an unexpected way. The most
damaging aspects of how Uber is being investigated by authorities concerns the signals
sent to all other actors in the economy that are considering entry, and of course espe-
cially into the sharing economy. One particular concern is whether or not freelancers
are to be considered employees. In the former case, businesses need to pay social
security contributions and other taxes. But other forms of uncertainty also abound,
such as the legal liability of platform owners for the actions of those who use or
provide the service, be it rental of rooms or peer-to-peer loans.

It is difficult to quantify the effect of this uncertainty on entrepreneurs considering

starting a business in the sharing economy. But the extensive media coverage of the
challenges faced by Uber is hard to miss. The small profit margins or indeed mostly
initial losses of start-ups make this kind of risk particularly damaging. From an
overall perspective, easing entry and reducing business model risk for the sharing
economy brings significant benefits to the economy as a whole: resources are better
used, easing the strain on the environment, low and middle income earners can
reduce their consumption and ownership to essentials and rent other resources
(especially cars) as needed, and jobs in the gig economy provide flexibility in working
hours. Finally, the sharing economy may also ease the challenge of job automation.

The challenge for existing firms, especially those in mature industries, is different. In
some areas, new digital firms will try to enter and compete. Having digital first in the
business model typically means lower costs, faster operations, more ease of experimen-
tation and reaching customers. Being large is no iron-clad protection against com-
petition from digital firms. Sony has been struggling for some time to transform its
products and services in response to the rise of the internet and is still trying to find
its way; Microsoft was also taken by surprise and its virtual monopoly on operating
systems now looks much more vulnerable.

Successful firms in industry are less exposed than start-ups because investment costs
are large and so entry typically requires large coffers and the ability to accept rather
uncertain returns on that investment. But carmakers are not being threatened by
start-ups, but rather large firms such as Google and Tesla. Banks are facing compe-

30
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

tition from fintech companies, now receiving increasing funding from institutional
investors and banks; IBM is entering the health care market with recent acquisitions;
big law and management firms that typically charge premium fees for their services are
being threatened by the sharing economy and the automation of knowledge work.
Especially for legal services, a lot of work can now be automated.

Lessons learnt from business failures show clearly that technological changes bring
strategic challenges, known as the incumbents curse. In the US, Kodak is a prime
example of this but there are many more. In Sweden, Hasselblad and Facit each had
strong market positions but their employees skills reflected outmoded technology
and the sales organizations were geared towards providing a service that would no
longer be needed with the shift to electronic technology.38 A particular challenge in
turning a company around when technology shifts is that vested interests in the firm
may protect the existing product line and focus on evolution only. Asking the cus-
tomers may not be the answer, because they may not know how technology can
change the service or product they need. And when the change comes, the transition
window may be too small.39 This happens time and time again.

To mitigate strategic risks, senior management cannot rely only on structures for
existing product lines to decide where to place their bets. While this in itself is hard
enough, it is compounded by the likelihood that new technology may initially be
lossmaking and may also ultimately cannibalize profits from existing products. The
auto industry is a prime example. With the advent of self-driving cars, it is likely that
fewer cars will be bought in total. The same applies to other aspects of the sharing
economy, such as rental of tools, formalwear, etc. where total consumption may be
at least initially reduced. While it is possible that consumers may buy more of
thesame good, they may be just as likely pocket the savings or spend the money
onother services.

Finally, as with regulation, the need to find and retain skills, especially digital skills,
mirrors the discussion above on the challenges for government policy. Firms will
have a key role in communicating business demands to the education sector as well
as being more active in lifelong learning. The rapid shift in technology implies that
workers who do not upgrade their digital skills will be held back in productivity and
wage growth. In countries with rigid labor laws and first-in, last-out collective bar-
gaining clauses, the need to ensure continuous skills development becomes all the more
important. Attending courses for a day or two will not be enough and skills will need
to be more strategically upgraded. The tax system should be adjusted to make it more
advantageous for firms to pay for upgrading workforce skills.

Potentially serious mistakes can be made by confusing correlation with causality

Another risk that may increase is that of confusing correlation with causality when
making business decisions or conducting policy analysis. The huge amounts of data
available on consumer trends, lifestyles, and online habits are valuable to firms and
provide information that can be used to target specific groups through advertising
and sales pitches. In relation to policy, there is a vast trove of behavior data available
that can inform policymakers how people will react to tax changes or other changes.
The irresistible lure may be that the huge amounts of data provide seemingly precise
estimates. Large data sets typically imply less uncertain measurement and when the
whole population is measured, there is no remaining sampling uncertainty. More and

38
See Sandstrm (2011, 2013).
39
See Bower and Christensen (1995).

31
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

more firms have access to big data or sell access to big data. But the apparent preci-
sion of such analysis may be a castle built on sand and may not withstand a change
in conditions.

Those conducting empirical work needs to be mindful to not interpret correlations as

showing causality. The experience with Google that used search frequencies to predict
the onset of flu is an illustration of how fragile correlations can be.40 When conditions
change, consumer behavior can change. The danger of over-interpreting correlations
is aggravated by that correlations can be stable for a long time and change only sud-
denly to due some unforeseeable trigger. For example, banks that provided loans in
the US prior to the financial crisis used vast amounts of past foreclosure data to pre-
dict future risk. But the many sub-prime foreclosures, while very small compared to
the total market, created a domino effect that cascaded into the rest of the market,
blowing historical correlations into the trashcan. The same may happen to consumer
surveys and other analysis based on big data.

The consequences of presuming causality when none exists is thus not one of an aca-
demic faux pas; it can have serious economic consequences on society and lead to bad
business decisions. One example of where this danger lurks is in the rising area of
now-casting, using vast hoards of online data (such as peoples search habits) as input
for macroeconomic forecasts. For example, do more online searches for unemploy-
ment benefits imply unemployment may be on the rise? Such a correlation may quite
possibly seem strong, but just as with the Google flu, peoples behavior may alter
over time and the correlation become weaker. The point is not to deny the value of
drawing inferences from online searches but to highlight the importance of combining
it with other types of corroborating data and models.

4. What might happen?

Digitalization is an unstoppable force, comparable to globalization. But its speed
and effectiveness may change significantly depending on the regulatory response. It
is instructive to illustrate what might happen through stylized scenarios. Ideally, such
scenarios should be done in a macroeconomic model but this is difficult as the ques-
tion is really about how changes in policy parameters may affect behavior. But past
models are based on historical patterns that may be only partly relevant, especially
for macro models that aggregate numerous different responses. To narrow the ques-
tions somewhat, we focus on only three central variables: productivity, employment,
and income inequality.

In Box 2, we discuss three different scenarios. They are not meant as forecasts but
asillustrations. The point of departure is that outcomes depend on regulation as well
as how society manages to smooth the period of adjustment in the labor market. In
particular, the objective should be to promote productivity growth and incentives to
work while reducing the severity of household income volatility. As emphasized in the
interim report from the Digitalization Commission improving skills, especially digital
skills, will likely make a big difference for employability, productivity growth, and
wages.41

40
See Lazer et al. (2014).
41
See SOU (2015a).

32
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Box 2. Scenarios for Sweden

In this box, we sketch the effects of policy on the economy. This thought experiment is
motivated by the literature that emphasizes the crucial role of how policy institutions react
to technological change.42 While the scenarios are constructed for Sweden, they are relevant to
other EU countries with extensive public welfare systems.43 Starting from the current levels
of productivity, employment, and inequality, we trace the effects of different policy respon-
ses. The time perspective is in the medium-to-long term, roughly the next two decades.

Macroeconomic assumptions for all scenarios

Macroeconomic headwinds continue to weigh down productivity growth
Unchanged framework for labor markets, including strong protection for regular employment
contracts
Social security systems remain unchanged

Scenario 1. Policy as usual: continued slow productivity growth, job creation slower than
job destruction.
This is a muddling-through scenario in which policy adapts slowly. A piecemeal approach to
regulation is likely associated with slower productivity growth. Essentially, the benefits of
digitalization may be slower in coming than the costs of adjustments in the labor market. The
macroeconomic headwinds discussed by Robert Gordon are then not sufficiently outweighed
by benefits from innovations and the value thereby created. The aging populations and the
high levels of debt in OECD countries are already affecting the economies, whereas the bene-
fits from new technology and its pace of adoption depend on regulation. If each area of regu-
lation is addressed separately liability for self-driving cars, patent and intellectual property
rights related to 3D printers, and the extent of liability of platform owners for transactions
they facilitate progress is likely to be measured in minihertz rather than megahertz. Issues
that need to be incorporated into laws and negotiated in treaties tend to take time for good
reason, since the credibility of and trust in the system depend on factors including its fairness
(real and perceived) and predictability.
Continued automation in industry and services mean that more jobs disappear and structural
unemployment may be higher during a transition period, since many sectors are affected at
the same time and with greater intensity than before. The process of automation is facilitated
by workers retiring and thus provides a continued opportunity for firms to shed costs without
actively having to reduce employees working under regular contracts. At the same time, the
challenges of finding people with the right skills remain an obstacle to hiring, especially for
highly skilled jobs. In Sweden, the additional rigidity of the housing market with the lack of
available rental homes imply that structural adjustments are harder. Workers cannot easily
move to new locations if there is no housing to be had.
New jobs are, of course, created all the time, but entrepreneurs are stymied by uncertainty
about regulatory changes that affect their business model. The question at stake is whether
their business model will be pushed into an existing analog regulatory box that undermines
the core idea or, less dramatically, whether the changes will introduce an element that raises
costs over and above those in the business plan.
In this scenario, we may see a gradual, but perhaps not marked, increase in income inequality.
Some jobs will have slower wage growth and some people will not be able to work in the
sharing economy due to regulatory obstacles. But we are still likely to see more superstar
42 winners, entrepreneurs who are able to capture large market shares and then remain dominant
43 due to brand recognition, but also due to obstacles for others to enter the market and compete.

42
See, for example, Acemoglu et al. (2005).
43
The mechanisms described are also relevant to the US, but the policy institutions and the starting points are different.
Both EU countries and the US share challenges in removing obstacles to benefiting from digitalization, but the US labor
market functions differently enough that the adjustments will likely be of a different kind and beyond the scope of this
report. Notably, the US has already experienced strong increases in income inequality.

33
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Scenario 2. Protectionist backlash: markedly slower productivity growth, more

unemployment, more income inequality.
One might regard this scenario as Scenario 1 combined with an economic policy reaction that
emphasizes the rights of insiders and incumbents and where destructive rent-seeking dominates
at the expense of value creation. Incumbent firms are able to use regulation to erect hurdles
that make entry harder and thus limit competition. This means that fewer jobs are created
and the incentive to shift away from human labor towards further automation becomes stronger.
This results in greater labor market polarization where middle level jobs are pushed down
the skills ladder but the slow productivity growth will imply poor real wage growth for broad
groups; that is, there will be both job and income polarization.

Scenario 3. Embracing change: better productivity growth, more job creation, unchanged
income inequality.
If regulatory overhaul is initiated strategically with a few guiding principles on consumer
safety and other concerns, this will reduce the need to reinvent the wheel for every area
subject to digitalization. Using those principles on a case-by-case basis in each sector makes
it possible to speed up regulatory review yet allow proper discussion of the idiosyncrasies
of each area. For example, principles and thresholds for how and at what rates freelancers
should pay tax in the sharing economy and what rules should apply with a view to reasonable
levels of consumer safety.
Some existing regulation should be reviewed to determine whether the requirements are
unnecessarily high with respect to risks and outcomes. For example, self-driving vehicles are
likely to save many lives from fewer traffic accidents and yet are held back by uncertainty,
notably on liability. Regulation should aim at neutrality between different alternatives, but
the outcomes should stem more from relaxing unnecessary rules rather than raising bars for
potential entrants. Raising the bars of entry may deter competition and lead to monopoly-type
rents to the detriment of consumers.
In this scenario, many more services will be created in the sharing economy and jobs may
thus keep pace with job destruction. Combining ease of adjustment between sectors with
improved skills, especially digital skills, will ease the changes and make welfare more stable
for large groups.
With the rise of the sharing economy as a more substantial part of the labor market, the dif-
ference in rights accrued to insiders compared to outsiders with insecure jobs needs to be
addressed. To maintain the legitimacy of the welfare system, freelance workers also need
tobetter covered by general social security safety nets. The safety nets need to be calibrated
to even out income over time but not so generous as to discourage work.

Summary of effects over the medium to long term (next 20 years)

Productivity Employment Income distribution
Scenario 1. Policy as usual 0 - -
Scenario 2. Protectionist backlash - -- --
Scenario 3. Embracing change + + 0

Note: The scenarios described in this box are stylized. A plus sign indicates an improvement, a zero roughly
unchanged, and a minus sign implies a deterioration. A double minus sign implies a stronger deterioration.

34
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

In the scenarios, we do not need to assume a dramatic adjustment in the labor market
in response to further automation. Instead, through the ongoing process of workers
retiring, firms will choose whether to replace them or to automate the tasks. Due to
shrinking working-age populations, automation may also be a response to lack of
skilled labor.

5. Closing words
How well countries fare in times of technological change will much depend on their
institutions. Very rigid institutions may tend to push the possibilities brought by dig-
italization into established analog regulatory boxes and thus hinder progress andmake
adjustment unnecessarily difficult. Governments that do not find the right balance
between leadership and cooperation with business may likewise lose out. The costs
will be paid in higher unemployment and slower productivity growth. Getting policy
roughly right matters a lot in times of change, and conversely, errors will be more
damaging than in normal times. The errors may be of omission as well as commission.
The absence of legal clarification of sharing economy business models, 3D printers, big
data, and other technical advancements is burdening the march of progress and pre-
venting entrepreneurs from improving business and consumer experiences.

On the other hand, there are some policies that would help to smooth adjustment in
the labor market and help realize the benefits of digitalization:
Lower the tax on (human) labor. High taxes on labor in Sweden further strengthen
the already strong incentive to automate tasks. Tax deductions for household
services should be expanded, not reduced.
Ease the regulatory uncertainty that surrounds the sharing economy. This will pro-
vide flexibility for adjustment and lower risks of higher structural unemployment.
Improve opportunities for lifelong learning. Longer working lives combined
with rapid technological change increase the risk that some skills will reach their
best-before date earlier than before. To reduce the risk of poor prospects in the
labor market, updated skills will be crucial.
Reduce the asymmetry in social security between being employed or working
freelance.
Establish principles for regulation of the digital economy that can be used to
speed reform in different sectors, reducing the need to reinvent the wheel. This
will require focused collaboration between lawyers and economists and not at
the usual slow pace. Otherwise, productivity growth may remain stagnant and
themacroeconomic headwinds from demography and public debt will continue
todampen growth.

The digital revolution is likely to improve quality of life, efficiency at work, provide
new goods and services and transform leisure. But there is nothing inevitable about
how smooth the ride will be. The institutional response will be key to improving pro-
ductivity growth without further growth in inequality. If the digital challenge were to
be summarized in one sentence it would be to remove the obstacles for digital inno-
vation to improve productivity growth while ensuring that incentives to work remain
strong and people are not left by the wayside with increasing inequality as a result.

35
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Sammanfattning p svenska
Technologys impact will feel like a tornado...No government is prepared for it.
Leader, The Economist, 18 januari 2014.

Digitaliseringen pverkar nstan all mnsklig interaktion p ngot stt och vra insti-
tutioner frsker nu sent omsider hinna ifatt utvecklingen. Inga sektorer r immuna
mot frndringarna och effekterna kommer att bli mrkbara inom hela den privata
sektorn, den offentliga sektorn och omrden som tidigare inte pverkats av tekniken,
som till exempel kulturinstitutioner.

Uppfattningarna om vad detta innebr skiljer sig kraftigt t frn tron att allt kommer
att frbli som det r till bilden av en tornado som drar fram genom ett kontorslandskap
p framsidan av The Economist.44 r digitaliseringen d robotarna tar ver slutet
fr jobben eller kan vi skapa tillrckligt mnga nya arbeten fr att erstta de som
frsvinner? Kommer klyftorna att vxa ytterligare och inkomstkningen att koncen-
treras till ngra f vinnare som frvrvar enorma rikedomar samtidigt som andra
fr arbeta som frilansare eller on demand med strre otrygghet och lgre inkomster?
Vi hvdar att utfallet frst och frmst kommer att bero p vilken politik som frs.

Den hr rapporten riktar sig till beslutsfattare och andra som r intresserade av digi-
taliseringens mjligheter och utmaningar. Den ger en bild av digitaliseringens kvalita-
tiva ekonomiska effekter och syftar till att fylla tomrummet mellan den akademiska
litteraturen och tankesmedjorna. Tankesmedjor och managementkonsulter har skrivit
mngder om Internet-of-Things (IoT), big data, 3D-skrivare och delningsekonomin,
dr mnniskor och fretag gr affrer via digitala plattformar. Alla sdana frndringar
kommer sannolikt att bli vsentliga, men delningsekonomin har, enligt vr uppfatt-
ning, strst frutsttningar att frndra hur ekonomin fungerar under de nrmaste
ren och f de mest djupgende effekterna. Ngra akademiska rapporter diskuterar
visserligen digitaliseringens ekonomi, men de flesta akademiker fokuserar huvudsak-
ligen p att frska frst de effekter som tidigare teknik givit upphov till. Utmaningen
fr en lsare som frsker skapa ett samband mellan de bda perspektiven r att avgra
vilka aspekter p de historiska upplevelserna som r relevanta och hur man ska stlla
sig till alla verdrifter som frekommer.

Digitaliseringen r en av flera olika trender som pverkar vrldsekonomin och det

rviktigt att understryka att ingenting r sjlvklart nr det gller konsekvenserna fr
den framtida vlfrden. De slutliga utfallen bra eller dliga kommer att bero p
vilka beslut institutionerna regering, riksbank och arbetsmarknadens parter fattar
och p deras frmga att anpassa sig till tekniken, demografin, globaliseringen och
andra faktorer. En avgrande lrdom frn forskningen r att de nationer som lyckats
bli rika har haft institutioner som gav std t incitament fr vrdeskapande och sam-
tidigt begrnsade mjligheterna fr insiders och srskilda intressegrupper att utnyttja
sin srstllning.

Teknikskiftet sker i en era med mnga olika understrmmar inom vrldsekonomin.

Produktivitetstillvxten har minskat i OECD-lnderna, ven om det rder viss tvekan
om den officiella statistikens kvalitet. Det finns ocks andra genomgripande och tyd-
liga frndringar: ldrande befolkningar och hg statsskuld och arbetslshet i finans-

44
Economist 18 januari 2014.

36
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

krisens klvatten. Det innebr att olmpliga beslut eller avsaknaden av beslut kan
f konsekvenser fr vlfrden som r mycket skadliga och ger lngsiktiga sociala och
ekonomiska effekter. r vra institutioner redo att ta sig an de hr utmaningarna?
Istllet fr att besvara den frgan direkt kommer vi att beskriva de mjliga ekonomiska
konsekvenserna fr produktivitetstillvxten, sysselsttningen och inkomstskillnaderna
med hjlp av tre olika scenarier i slutet av denna sammanfattning. Vi hvdar att utfallet
frst och frmst kommer att bero p de politiska besluten. Om besluten gr i protektio-
nistiskt riktning kommer klyftorna och arbetslsheten sannolikt att ka (se scenario 2 i
Ruta 2. Scenarier fr Sverige). Vi mste ocks minska skillnaderna mellan anstllda
och egenfretagare nr det gller tillgngen till socialfrskringar.

Det r obestridligen s att vra ekonomier tidigare har lyckats anpassa sig till mnga
olika frndringar och utvecklats frn att primrt vara jordbruksbaserade till att
baseras p massproduktion och en alltmer dominerande tjnstesektor. En tillbaka
blick visar att alla sdana frndringar har inneburit hgre vlstnd fr hela sam-
hllet, men att livet kunde vara svrt fr mnga mnniskor under sjlva frndringen.
Ekonomihistorikern Joel Mokyr har noterat att vlfrden inte frbttrades nmn-
vrt mellan 1750 och 1850.45 Idag finns sociala skyddsnt och regler som avlastar en
del av frndringens brda frn individen och vra ekonomier har utvecklat metoder
fr riskfrdelning mellan olika grupper och generationer. Bland OECD-lnderna r
Japan antagligen det enda landet dr de yngre generationerna inte har ftt det bttre
n de ldre p grund av tv rtionden med vikande tillvxt och en snabbt ldrande
befolkning, kombinerat med starka rttigheter fr insiders.46

Tekniken kan nu frndra sttet att sprida riskerna inom ekonomin och ytterligare
vidga klyftan mellan insiders och outsiders, mellan etablerade och nystartade fretag.
Om det finns en lxa som vi br lra oss av den ekonomiska historien s r det att ett
stt att formulera den politiska utmaningen r att gra den strukturella frndrings-
perioden s smidig som mjligt eller i mjligaste mn undvika stora svngningar
och samtidigt garantera frutsttningar fr produktivitetstillvxt. Nr det gller
social vlfrd r den strsta utmaningen hur institutionerna agerar fr att upprtt-
hlla incitament fr innovationer och fr att skapa nya arbetstillfllen samt att inte
ge vika fr srskilda intressegrupper och protektionism. Hur man lser sdana utma-
ningar kan vara avgrande fr om klyftorna vidgas eller ej.

Frndringstakt kontra institutionell trghet

Dotcom-bubblan som brast fr ett rtionde sedan knnetecknades av samma sorts
verdrifter som ofta omger ny teknik. Idag finns det finns skerligen anledning att se
kritiskt p verdrivna pstenden att stora frndringar r p gng och att tempot
kan bli hgre n under tidigare frndringsperioder. Ett exempel som visar att man
br vara vaksam p verdrifter r den spektakulra konkursen fr det svenska bolaget
Boo.com strax efter millennieskiftet. Fretaget, med ste i London, frskte slja
designerklder via en digital plattform och var drmed ett av de frsta som frskte sig
p ngot som idag r vardagsfreteelse, men marknaden var inte tillrckligt mogen
och frutsttningarna fr framgng fanns inte nnu.

Trots erfarenheterna frn Boo.com och andra nystartade fretag som misslyckats finns
det mycket som talar fr att frhllandena r annorlunda den hr gngen. Boo.com
misslyckades med ngot som idag r blivit vardagsfreteelse hos bland andra Zalando,
Wish och andra e-handelsplattformar. Idag har vi ett stndigt kande antal digitala

45
Se Mokyr (2004).
46
Se IMF:s artikel IV-konsultation med Japan frn 2012.

37
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

fretag med global verksamhet. En del av de hr fretagen har svrigheter, och vissa
misslyckas, som vanligt r bland nystartade fretag, men ngra f har blivit extremt
framgngsrika. Frmgan att snabbt utka bolagets verksamhet har blivit kritisk.
Eller som Reid Hoffman, en av LinkedIns medgrundare, uttrycker det: First-scaler
advantage beats first-mover advantage.47 Dagens situation r annorlunda n den
var fr tio r sedan och en rad av de viktigaste frutsttningarna har frndrats och
mognat:
Genom utvecklingen av stora plattformar med standarder som lockar till sig bde
kunder och producenter (webben, Apple med iOS och Google med Android).
Genom att frtroendemekanismer tillkommit fr digitala transaktioner, bde fr
varor och tjnster.
Genom digitala betalningssystem med lga transaktionskostnader.
Genom utbredd anvndning av smarta telefoner och surfplattor.

Vrlden r nu redo fr transaktioner med digitala varor och tjnster. Frndring-

arna har frmst drivits fram av konsumenterna. De sker p internet, anvnder sina
smarta telefoner och delar information, bilder och upplevelser p ett stt som i hg
grad pverkar handeln och samhllsdiskursen. De nya generationerna kommer sanno
likt att anvnda digitala verktyg fr konsumtion och fritidsaktiviteter i allt hgre
grad, vilket innebr att den konsumentdrivna aktivismen frblir en viktig drivkraft
som kan komma att bli allt viktigare. Varje ny tjnst behver bara vara tillgnglig
frn en av de strsta app-butikerna och/eller p webben fr att n ett stort antal kon-
sumenter. Infrandet av elektriciteten ger den bsta jmfrelsen med tidigare teknik-
skiften, eftersom den ocks var en universalteknik som gav mjlighet att anvnda mnga
andra maskiner. Skillnaden mot elektriciteten r att digital teknik kan hota andra verk-
samheter mer eller mindre kontinuerligt: Fr elektriciteten var det frsta steget vikti-
gast: att utka elntet och skapa frutsttningar fr moderna hushllsapparater och
fabriker. De mnga fljande stegen var av evolutionr snarare n revolutionr natur.
En del hvdar ocks att den digitala tekniken r den enda universaltekniken som ger
stordriftsfrdelar.48

Fretag som inte anpassar sig till konsumenternas frndrade krav p produkter eller
information riskerar att hamna p efterklken. Samtidigt r vra institutioner upp-
byggda kring en lngsammare vrld, dr lagarna och gngse regler p arbetsmark-
naden innebr trgheter. Trgheten r sannolikt frmst en frdel, srskilt nr det
gller demokratin och grundfrutsttningarna fr tillmpningen av regler och social
vlfrd. Men arbetsmarknaden har utvecklat en stelbenthet som nu utmanas av digi-
taliseringen. Den snabbt kande ldrande befolkningen och behovet av att frlnga
arbetslivet fr att finansiera den offentliga vlfrden skulle nd tvinga arbetsmark-
naden att frndras, men digitaliseringen gr reformeringen av institutionerna allt
mer angelgen. Utmaningen r att behlla komponenter som bidrar till stabilitet
utanatt hmma tillvxt och innovation.

Digitaliseringen erbjuder mnga nya stt att kringg stelbenta regler och undvika
regelverk som har utvecklats under mnga r. Det kan finnas goda skl fr mnga
av reglerna, srskilt vad gller konsumentskyddet, men mnga av dem leder till fr
mycket skydd fr insiders och monopol frn konkurrens. Det innebr hgre priser fr
konsumenterna och gr det svrare fr unga mnniskor och arbetslsa att hitta jobb.
Mycket av den hr stelbentheten kommer att ifrgasttas, srskilt nr de digitala ls-

47
Hoffman (2015).
48
Se Edquist och Henrekson (2006).

38
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

ningarna ger mjlighet att utforska nischer med ineffektivitet och outnyttjade resurser
som kan skapa vinst. Frndringskraven fr institutioner och fretag kommer att bli
mycket starka eftersom de motiveras av vinster som kan vara anmrkningsvrt sm
per sld enhet, bara ntverkseffekten skapar en hg frsljningsvolym. Anledningen
till att frndringen sannolikt kommer att bli snabb r kort sagt att den motiveras av
vinster som tekniken bde skapar mjligheter till och utgr drivkraften fr. Hur pro-
duktivitetstillvxten, arbetsmarknaden och klyftorna slutligen kommer att pverkas
beror p hur regeringar och myndigheter reagerar p de hr utmaningarna.

Vi ser det som osannolikt att digitaliseringen kommer att innebra massarbetslshet
p grund av att robotarna vertar mnga eller de flesta jobb. Men alla hller inte
med om det. Det finns vlrenommerade forskare, bermda vetenskapsmn och affrs-
experter som hvdar att robotarna kommer att erstta nstan all arbetskraft, vilket i
allt vsentligt innebr slut fr jobben.49 Ruta 1 innehller en kort sammanfattning av
de olika argumenten, men erfarenheterna frn den ekonomiska historien visar att nya
jobb (som vi ofta inte ens kan frestlla oss) skapas hela tiden. Problemet r i stllet
att missriktade politiska beslut kan frsvra vergngsperioden och anpassnings
perioden kan ge upphov till hgre strukturell arbetslshet och kande klyftor.

De mnga olika uppfattningarna om vad digitaliseringen innebr skapar sannolikt

ocks viss frvirring. ena sidan uppmanas mnniskor att fortstta arbeta lngre och
det finns en brist p yrkesutbildad arbetskraft i mnga OECD-lnder, bland annat
Sverige och Tyskland. andra sidan kan det hnda att robotarna tar ver. Hur ska
man stlla sig till detta? I den hr rapporten frsker vi att svara p sdana frgor och
frhoppningsvis bidra till mindre frvirring. I grunden r det den lngsiktiga trenden
med ldrande befolkningar som skapar problem med finansieringen av den offentliga
vlfrden. Om robotarna faktiskt skulle verta merparten av all jobb frn mnniskorna,
ngot som vi betraktar som osannolikt, skulle de befintliga finansieringsmetoderna
fr den offentliga vlfrden inte fungera. En mycket stor andel av statens intkter
kommer frn beskattningen av arbete. Men om jobben istllet fortstter att finnas
kvar mste vi frlnga vra arbetsliv fr att finansiera vlfrden, eftersom det kommer
att finnas fler pensionrer n unga mnniskor som kan frvrvsarbeta.50

Ruta 1. En sammanfattning av argumenten fr och emot slutet fr jobben

Skl som talar fr vanlig strukturomvandling

Arbetena har inte frsvunnit trots enorma teknikskiften som ersatt jordbruket med tillverk-
ning och tjnster som bas fr ekonomin.
Tendensen att underskatta komplementariteten mellan mnniskor och maskiner som ger
mjlighet till produktivare arbete utan att erstta arbetskraften.
Bilen, ngmaskinen, elektriciteten, rent vatten, telefonen osv. pverkade vr livsstil i
mycket hgre grad och har frbttrat livskvaliteten mer n dagens digitala uppfinningar.
Betydande trgheter i institutioner och regelverk och lngsam anpassning kan under
minera nya affrsmodeller.
Internet of Things och big data har ett stort inslag av juridisk oskerhet som kommer att
frdrja frdelarna.
Preferenser fr mnsklig interaktion p skolor, sjukhus, osv.

49
Fysikern Stephen Hawkins stdjer till exempel den uppfattningen, och Bill Gates, en av grundarna till Microsoft, har
uttryckt liknande sikter.
50
Se Blix (2013a).

39
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Skl som talar fr omfattande frndringar, men inte slutet fr jobben

Digitaliseringen har starka ntverkseffekter och marginalkostnaden r lg eller obefintlig
fr mnga digitala tjnster.
Transaktionskostnaderna kan bli mycket lgre (delningsekonomi, 3D-skrivare).
Digital kunskap r icke-exklusiv och information sprids utan begrnsningar.
Mnga sektorer pverkas snabbare och samtidigt.
Demografin och matchningsproblemen p arbetsmarknaden kar drivkrafterna att
automatisera.
Digitala plattformar kan matcha kpare och sljare ven fr sm varor och tjnster och
gradet lttare fr investeringskapital att hitta entreprenrer.
Delningsekonomin ger flexiblare arbetsmarknader och starkare konkurrens fr arbetskraften.
Snabb utveckling av intelligent programvara som kan erstta mnniskor och analysera
ostrukturerad information, utfra empiriskt arbete och skriva texter som inte med ltthet
kan identifieras som maskinproducerade.

Skl som talar fr slutet fr jobben

Maskiner som brjar lra sig sjlva utvecklas snabbt.
Mnniskans frmga att anpassa sig r lngsammare n kningen av datorkraften som kan
anvndas till att lsa analytiska arbetsuppgifter och tillmpa undermedveten kunskap.
Okvalificerad arbetskraft kommer att ha lgre produktivitet n maskiner och drmed inte
vara anstllningsbar eller ha en svag reallneutveckling.

1. Arbetsmarknaden
Vad har hnt?
De flesta mnniskor r sannolikt inte medvetna om hur lngt programvarorna har
utvecklats i sin frmga att utfra mnskliga aktiviteter. Sjlvkrande bilar har det
talats om sedan lnge, men p andra omrden, srskilt nr det gller mycket kvalifi-
cerade arbetsuppgifter, r det vl fortfarande mnniskan som har monopol? Mnga
mnniskor r nog vagt medvetna om att datorerna har slagit vrldens bsta schack-
spelare och Jeopardy-deltagare. Men alla mnskliga aktiviteter, i synnerhet kreativt
arbete, skrivfrmga och analys, kan inte kodas i programvara de krver kognitiv
och undermedveten kunskap som datorerna saknar eller hur?

Artificiell intelligens r inte i sikte inom en nra framtid, men dagens programvaror r
s avancerade att de kan utfra mycket kvalificerade mnskliga arbetsuppgifter p
mnga olika omrden.51 Tack vare Moores seglivade lag om allt snabbare datorer r
datorkapaciteten nu s hg att mnga arbetsuppgifter som tidigare var frbehllna
mnniskor, till exempel att skriva och forska, nu kan utfras snabbare och exaktare av
datorer, och de har dessutom stilistisk frmga som gr att det inte lngre r uppen-
bart att en text har skrivits av en dator.52 Populrkulturens filmer och bcker inne-
hller ofta ondskefulla robotar som gr p tv ben, men det ger en missvisande bild.
Arbete kan nu lggas ut i molnet, antingen till intelligent programvara eller till en
global arbetsmarknad via digitala plattformar.53 Den moderna ersttaren fr en kon-
torist r inte en mnniskoliknande manick som sitter med en kaffekopp i en stol (fr
att framst som mnskligare). Den r istllet en av de mnga stackar med maskinvara

51
Se till exempel Bostrom (2014) fr en versikt av framstegen p omrdet artificiell intelligens och lget idag.
52
Se Clerwall (2014).
53
Se till exempel OConnor (2015b).

40
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

som finns i anonyma serverhallar p mer eller mindre avlgsna platser med snabba
fiberanslutningar. Facebooks servrar i Sverige finns till exempel lngst upp i norr,
iLule, dr det kyliga klimatet bidrar till lga energikostnader.

En av dagens frmsta forskare p omrdet teknik och arbetsmarknad, MIT-professorn

David Autor54, hvdar att oron fr automatisering inte tillrckligt tar hnsyn tilltv
krafter: graden av komplementaritet mellan mnniska och maskin, dvs. hur tekniken
gr oss bttre p mnga olika arbetsuppgifter, och svrigheten att automatisera under
medveten kunskap. Undermedveten kunskap avser saker som vi gr utan att vi kan
frklara exakt hur, till exempel att cykla eller knna igen en stol bland tusentals bilder.
Argumentet r att vi inte kan automatisera en aktivitet om den inte fljer specifika
regler som kan verfras till kod.

Nr det gller komplementaritet finns det en hel del material som vi kommer att dis-
kutera nedan, varav en del r oroande fr hur mnga och vilka jobb som kommer
finnas kvar. Men frgan om undermedveten kunskap r av ett annat slag, ngot av ett
mellanting mellan ett filosofiskt och ett praktiskt hinder fr automatisering.55 Det br
dock vara tydligt att idn om att undermedveten kunskap inte kan kodas r ett anta-
gande och inte en naturlag. Fr ngra r sedan var det till exempel mnga som utgick
ifrn att sjlvkrande fordon var en omjlighet eftersom det ingr alltfr mnga olika
typer av undermedveten kunskap fr att kra bil. Om frmgan att digitalisera under-
medvetna frdigheter kommer att mta overstigliga hinder terstr att se. Kanske
undermedvetna kunskaper inte utgr en ogenomtrnglig mur som automatiseringen inte
kan ta sig ver, utan snarare ngot som kan omformas eller kringgs och omdefinieras.

Maskiner som brjar lra sig sjlva exemplifierar hur man kan undvika svrigheterna
med att koda undermedveten kunskap. Man kan lra maskiner att imitera mnniskor
och observera fysiska hndelser. Detta r ett omrde som snabbt utvecklas och fr-
bttras. Dagens programvara r redan s intelligent att den kan ta ver mnga mnsk-
liga aktiviteter, men det finns andra hinder n tekniken som kan st i dess vg (mer
om detta nedan). Ett exempel r att det nnu inte finns pilotlsa passagerarflygplan,
medan det finns frarlsa tg till exempel i Asien. Tekniken finns redan, men om det
finns en efterfrgan p den r oklart tminstone idag. Men detta kan ocks komma
att frndras om pilotlsa plan blir lika skra men betydligt billigare n vanliga flyg-
plan. Det r inte ltt att veta p frhand vilka aktiviteter som lter sig automatiseras
och att utforska var grnserna gr r en utmaning fr entreprenrerna. Allmnhetens
uppfattning om vilka aktiviteter som tryggt kan automatiseras kommer sannolikt att
frndras ver tid. World Values Surveys56 indikerar visserligen en stark tro p vissa
grundvrderingar, srskilt nr det gller familjens betydelse, men uppfattningen om
hur vi samverkar med varandra och hur vi konsumerar varor och fritidsaktiviteter
kan sannolikt formas och frndras i takt med tekniken.

En av de strsta frndringarna som digitaliseringen givit upphov till r framvxten

av delningsekonomin, som trots sitt namn handlar mer om marknadskrafter och
entreprenrer som hittat nya affrsmjligheter. Andra, kanske ngot mer beskrivande
termer, r engelskans on-demand economy och gig economy, och uppdrags
ekonomin, men i den hr rapporten kommer vi att fortstta anvnda termen delnings-
ekonomin. Med digitala plattformar kan ven en liten efterfrgan p olika varor och
tjnster matchas och levereras till lg kostnad. Delningsekonomin r en viktig makro
ekonomisk frndring. Grannar och arbetskamrater har visserligen bytt varor och

54
Se till exempel Autor (2014).
55
Se Ambrosini and Bowman(2001) fr en diskussion om mtningsproblematik fr undermedveten kunskap
56
Se Halman et al. (2008).

41
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

tjnster med varandra under hela mnsklighetens historia, men den sortens aktiviteter
hller nu p att omvandlas p ett stt som liknar hur industritillverkning frndrade
produktionen fr alltid. Vetskapen om att det finns en mnniska som efterfrgar en
fretags- eller hushllstjnst vid en viss tidpunkt innebr att det inte lngre r samma
lotteri att matcha tillgng och efterfrgan. Fljaktligen kommer det att finnas fler och
flexiblare arbeten i delningsekonomin, men ocks strre otrygghet, en frga som vi
kommer att tervnda till nedan.

Ytterligare automatisering
Automatiseringen av lgkvalificerade jobb och arbeten inom tillverkningsindustrin
har blivit en drivkraft och kanske ocks en potent symbol fr produktivitetstillvxten
inom marknadsekonomierna. Det behvs ingen strre fantasi fr att frst att sjlv
krande bilar och fordon kommer att frndra mnga aktiviteter, och minska behovet
av taxi- och lastbilschauffrer, bilskolor, frskringsbolag och servicepersonal fr hotell
och restauranger. En positiv effekt r dock att lkarna kommer att ha frre trafik-
skadade att ta hand om eftersom de lr bli frre olyckor. Men det uppstr sannolikt
ocks dominoeffekter, och mnga av dem r ofrutsgbara. Ett exempel p en ngor-
lunda frutsgbar dominoeffekt r att infrandet av sjlvkrande fordon kommer att
pverka landsbygden. Om frre mnniskor levererar varor till landsbygden minskar
antalet gster fr mnga restauranger och hotell, vilket ger smre frutsttningar fr
deras verksamhet.

Inom jordbruket har det hittills krvts viss mnsklig aktivitet fr hantering av kns-
liga frukter, men detta brjar ocks att frndras. Maskinerna r nu s avancerade
att de kan ta ver sdana arbetsuppgifter. Logistiklager, som anvnds av Amazon,
Wal-Mart i USA, och Clas Ohlson i Sverige, har i allt hgre grad halvautomatiserats
och allt fler funktioner har vertagits av maskiner. Idag finns det maskiner som kan
producera tvhundra hamburgare eller mer p en timma och i Japan finns helauto-
matiserade sushi-restauranger och hotell dr servicepersonalen i stort sett har ersatts
av maskiner, ungefr p samma stt som att incheckningen p vra flygplatser mest
handlar om att interagera med maskiner istllet fr med stressad incheckningspersonal.

Det r sant att mnga av arbetena som har frsvunnit inte ndvndigtvis var srskilt
tilltalande Mnga av dagens arbeten har faktiskt blivit bttre bland annat byggnads-
arbete, gruvarbete och andra fysiskt pfrestande yrken i takt med att maskinerna
har ersatt en del av den mnskliga arbetsinsatsen. Att se barn i utvecklingslnder stiga
ut ur mrka hl i marken efter slitsamt arbete med att utvinna kol eller mineraler ger
oss en pminnelse om det.

En del av ledtrdarna fr interaktionen mellan maskiner och mnniskor i dagens

ekonomier kommer frn bankvsendet. I brjan fruktade bankkassrerna att deras
arbeten skulle frsvinna nr bankomaterna infrdes, men antalet arbetsuppgifter
kade faktiskt och krvde hgre kompetens: kassrerna brjade erbjuda mer service
och rdgivning till kunderna. Det finns exempel p komplementaritet mellan maskiner
och mnniskor ven p andra omrden, bland annat inom sjukvrden, dr lkarna
har frbttrat sin kliniska och kirurgiska yrkeskompetens med hjlp av instrument,
datorer och maskiner. Men p vissa omrden, bland annat nr det gller kontorsarbete,
har datorer ersatt arbetskraften p samma stt som hantverkarna trngdes undan nr
industrialiseringen brjade. Maskiner har vertagit vissa arbeten och nya, mer kvalifi-
cerade jobb med hgre grad av komplementaritet har skapats under hela historien.

42
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Digitaliseringen har nu ntt en sdan hastighet, mogenhet och spridning att den kan
sl igenom snabbare och f mera omfattande effekt n ngon annan teknik hittills.
Infrandet av elektriciteten, som ocks var en universalteknik, under 1900-talet
r kanske den bsta jmfrelsen. ven elektriciteten pverkade mnga olika sektorer
samtidigt, men skillnaden r att den inte infrdes lika snabbt. Detta kan gra det svrare
fr arbetskraften kvalificerad eller ej att ta sig frn en sektor p tillbakagng till
en sektor p frammarsch.

Det finns starka incitament att ta reda p vilka delar av den globala vrdekedjan
som kan automatiseras. Stora svenska bolag, som till stor del r exportinriktade, har
redan en stndig strvan efter effektivitet inbyggd i sitt DNA. Fr dem r det helt
naturligt att outsourca eller automatisera delar av den globala vrdekedjan. Svrig-
heten att erstta personal som i en ldrande befolkning gr i pension och matchnings
problem p arbetsmarknaden kan mycket vl komma att strka incitamenten att
automatisera. Sdana incitament r srskilt stark i Sverige och andra lnder med
rigidarbetslagstiftning och hg skatt p arbete.

De nya och kraftiga frndringar som pgr antyder att fler medelklassarbeten
kommer att pverkas av digitaliseringen. Tidnings-, film- och musikbranscherna har
till exempel sedan lng tid tillbaka varit pverkade av att den digitala distributionens
har lga marginalkostnader och reser svrigheter att ta betalt fr innehll. Nsta
automatiseringsvg kommer att automatisera tusentals texter, allt frn bolagens
resultatrapporter till fotbollsresultat.

ven nr det gller juridiskt utredningsarbete och affrsanalys hller programvara p

att automatisera sortering, analys och presentation av enorma datamngder.57 Fr juri-
diska tillmpningar, dr prejudikat r viktiga, kan programvara numera finkamma
miljoner dokument som tidigare skulle ha tagit mnga r att arbeta sig igenom. nnu
s lnge avser mnga av dessa arbeten p instegsniv, till exempel av assistenter p
advokatbyrer, men det finns redan teknik fr betygsttning av skriftliga tentamina
och uppsatser i skolor och liknande.

Arbetsuppgifterna som kan utfras av maskiner blir alltmer kvalificerade. IBM och
Google kper upp sjukvrdsbolag i hopp om att kunna frbttra diagnostiken med
hjlp av programvara.58 Tillgngen till stora databaser och avancerade rutiner fr
dataanalys ger programvaran mjlighet att frbttra de medicinska bedmningarna.
Volymerna r helt enkelt fr stora fr att en mnsklig lkare ska kunna flja den breda
medicinska forskningen utanfr sitt eget kompetensomrde. Med hjlp av statistiska
metoder och slutledningsfrmga kan programvaran tillhandahlla prognoser och
rekommendationer baserade p information som en enskild person helt enkelt inte
kan ta till sig.

Br vi ha frtroende fr programvarurekommendationer nr det gller medicin, juridik

eller andra omrden? Mnga mnniskor kommer sannolikt inte att uppskatta svar
frn svarta ldor, oavsett hur mktiga och allvetande de verkar vara. Hr har till
exempel IBMs WatsonPaths potential att skapa frtroende fr automatisering av ven
kvalificerade arbetsuppgifter, eftersom den inte bara visar det rekommenderade alter-
nativet utan ocks stegen fr att n dit.

57
Se till exempel Ford (2015a) fr en versikt av arbetsuppgifter som idag kan utfras av maskiner.
58
Se Crow (2015) och Lohr (2015a, c)

43
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Nr vi funderar p framtidens jobb hur mnga de blir och av vilket slag br vi

gra en terblick och frska komma underfund med vilka historiska erfarenheter
som r relevanta. Den ekonomiska historien ger exempel p praktiska erfarenheter
som visar effekterna av strukturfrndringar och vilka utmaningar vi kan frvnta
oss. Frsta delen av den industriella revolutionen prglades av s kallade deskilling,
men strre delen av 1900-talet kombinerade allt hgre teknisk expertis med allt hgre
kompetens. Men under de senaste decennierna har OECD-lndernas arbetsmarknader
polariserats alltmer: antalet arbeten p medelinkomstniv har minskat medan de hg-
och lgkvalificerade arbetena utgr en allt strre andel av arbetsmarknaden. Detta
har nnu inte skett i Sverige p samma stt; Sverige har istllet sett en kning av andelen
hgavlnade arbeten. I Sverige har lnerna inte heller polariserats p samma stt som
i USA, dr reallnerna fr mnga arbeten p lg- och medelinkomstniv har kat fga
eller inte alls. I Sverige har utvecklingen istllet varit den motsatta, med en stark real
lneutveckling under de senaste tjugo ren.

Polariseringen av arbetsmarknaden kan mycket vl ge upphov till spnningar i en

ekonomi, men som Raguran Rajan, tidigare chefsekonom fr IMF och idag Chef
fr den Indiska centralbanken, framhller kan den snabba kningen av bostads-
priserna i USA ha bidragit till att f medelinkomsttagarna att knna sig som om de
fick ta del av det vlstnd som skapades, tminstone fre finanskrisen.59 I USA kar
idag spnningarna, srskilt bland de som har lgavlnade arbeten, till exempel inom
snabbmatskedjorna. Ett antal mnniskor med lgavlnade arbeten i USA fr ocks
matkuponger, och drmed subventionerar staten i realiteten en del av de arbetsgivare
som betalar lga lner, till exempel i snabbmatsrestaurangerna.60

I Sverige har utvecklingen varit helt annorlunda. Hr reformerade man institutionerna i

efterdyningarna av den spruckna bostadsbubblan p 1990-talet. Radikalt frndrade
institutioner och starka statsfinanser gjorde att Sverige klarade finanskrisen 2007och
kningen av arbetslsheten blev inte alls lika stor som efter den stora chocken p 1990-
talet. Efterverkningarna av finanskrisen blev visserligen en utmaning, men Sveriges
reformer p 1990-talet strkte institutionernas frmga att motst chocker p ett stt
som gynnat landets innevnare.61 I motsats till mnga andra OECD-lnder r stats
finanserna i Sverige hllbara och landets statsskuld har varit lg bde under och efter
finanskrisen. Sverige var faktiskt ett av de f EU-lnderna som kunde hlla sin stats-
skuld en bra bit under 60 procent av BNP, som den ursprungliga stabilitets- och till-
vxtpakten i EU krvde.

Reformerna av budgetprocessen p 1990-talet var nyckelkomponenter som bidrog till

detta resultat, och utgiftstaket p medellng sikt fr statsutgifterna och frbttringen av
hur budgeten hanteras i riksdagen var srskilt vrdefulla verktyg som kundeanvndas
till att minska den starka tendensen mot underskott som i vrigt prglade statsutgif
terna. Arbetsmarknadsreformerna som inleddes efter 2006 gav dessutom ett starkare
incitament fr arbete genom att snka inkomstskatterna och strka bidragssystemens
svaga kontrollmekanismer fr att minska missbruket.62 En annan nyckelreform var nr
man gjorde Riksbanken oberoende r 1999 fr att penningpolitiken skulle inriktas mot
prisstabilitet. En erfarenhet frn Sverige r allts att institutionella reformer som fr
olika krafter att samverka p ett bttre stt r helt avgrande fr att klara ekonomin,
i synnerhet i kristider.

59
Se Rajan (2010).
60
Se till exempel Cohen (2015) och Jacobs et al. (2015).
61
Se till exempel Calmfors (2013); Heyman et al. (2015b).
62
Se Vrbudgeten 2014, bilaga 4.

44
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Men det r ett misstag att tro att institutionerna ngonsin blir fullndade och klarar
av vad som helst utan att reformeras. Det kommer alltid att vara svrt att gra avvg-
ningar mellan olika angelgna ndaml i ekonomin, men institutionernas struktur
kan underltta och skapa frutsttningar. Institutionerna kan med andra ord gra
det lttare att fra en bra lngsiktigt politik, men det betyder inte att tillrckligt bra
institutioner mjliggr policy p auto-pilot. Till syvende och sist r det de offentliga
beslutsfattarna andra parter som mste genomfra och frsvara svra vgval.

Arbetsmarknaden i Sverige knnetecknas av centrala avtalsfrhandlingar mellan arbets

marknadens parter och staten har ingen direkt kontroll ver lnefrhandlingarna, till
skillnad frn andra lnder dr staten kan vara direkt involverad i lnesttningen fr
den offentliga sektorn.63 Arbetsgivarna har visserligen rtt att fristlla arbetstagare,
men de mste frhandla om uppsgningar med fackfreningarna enligt principen frst
in, sist ut. OECD har ofta rekommenderat att Sverige br ka flexibiliteten p arbets-
marknaden. Det nuvarande systemet r gynnsamt fr insiders p arbetsmarknaden
med avtal fr tillsvidareanstllning. Systemet gr det svrt fr andra att komma in
parbetsmarknaden. Det ger ocks mindre incitament att byta arbete ven d de
ekonomiska frhllandena frndras, till exempel om en sektor r p tillbakagng
och utsikterna bttre p annat hll. Dessutom kan mnniskor vara mindre bengna
att frivilligt ge upp den trygghet som ett avtal fr tillsvidareanstllning erbjuder, ven
om de skulle vara gynnsamt att byta arbete eller sektor.

Marknadsekonomin hittar ofta metoder att hantera bristen p flexibilitet, ibland till
hg kostnad och ibland till lgre. I Sverige har speciella bemanningsfretag, som p
begran tillhandahller tillfllig arbetskraft fr allt frn kontorstjnster till fretags-
ledning, blivit populra inom flera branscher. Nr det gller tryggheten fr tillfllig
personal r Sverige ett av de lnder som har det svagaste skyddet fr personal inom
OECD. Sammantaget innebr detta att Sverige har en av de mest utprglade tudelade
arbetsmarknaderna bland OECD-lnderna.64

Grnser fr automatiseringen av jobb och konsekvenser fr den svenska

arbetsmarknaden
Ngra forskare argumenterar att en stor andel, 50 procent, av dagens arbeten kan
komma att automatiseras.65 Det kanske framstr som sjlvklart, men det r nd vrt
att framhlla att de jobb som frsvinner endast ger en partiell bild av hur arbets-
marknaden kan utvecklas. Att jobb frsvinner och att befintliga arbeten frndras
r inget nytt. Nya jobb skapas hela tiden och problemet r egentligen hur snabbt de
uppstr i frhllande till antalet mnniskor som sker efter jobb och om de mnnis-
korna har rtt kvalifikationer. Dessutom finns det motkrafter som kan frdrja
automatiseringen.

Fr det frsta innebr inte det faktum att tekniken ger mjlighet att automatisera
arbeten att detta ndvndigtvis kommer att ske. Den frmsta drivkraften bakom
automatiseringen r viljan att skapa ngot som r bttre och ger hgre kvalitet, nya
varor eller tjnster eller lgre kostnader. Inom vissa omrden innebr automatisering
hgre kostnader n mnsklig arbetskraft. Nya bilar har till exempel vindrutor som
installeras av robotar, men nr en vindruta ska ersttas r det normalt en mnniska
som utfr arbetet. P en del omrden kanske det inte finns ngon strre efterfrgan

63
Lneutrymmet inom den offentliga sektorn i Sverige fljer en mekanisk regel som anvnder produktivitetstillvxten
inom den privata tjnstesektorn som en mttstock fr krav rrande kostnadseffektivitet vad gller lner och verksamhet.
64
Se Cahuc (2010), sid. 150153, och OECD (2015b).
65
Se Frey och Osborne (2013) och Flster (2015).

45
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

p automatisering. Mnniskor kanske motvilligt accepterar att besvara automatiska

frgor fr att n en kundservice. Det finns mnga jobb dr automatisering inte kommer
att vlkomnas eller accepteras, tminstone inte inom en snar framtid. Automatiserad
utbildning r ett exempel dr lrare i tmligen stor utstrckning skulle kunna
ersttas av programvara. Frldrarna kanske inte skulle bli s entusiastiska ver det
och fackfreningar kanske ocks skulle hlla tillbaka en sdan utveckling.

Fr det andra finns det mnga andra trender som pverkar efterfrgan p jobb och som
har lite att gra med automatisering. Demografin kan vara vsentlig som en generell
drivkraft fr automatisering. Men en ldrande befolkning kan samtidigt ka behovet
av mnskligt arbete inom sjukvrd och ldreomsorg. Fler sdana arbetsuppgifter kan
visserligen automatiseras idag robotar som hjlper ldre personer finns till exempel
redan i Japan och det finns frsksversioner av dem i Sverige. Men det r svrt att tro
att sdana frndringar kommer att bli avgrande inom verskdlig tid, om man ser
till de behov som finns. Inom juridiken har automatiseringstekniken till exempel gjort
det mjligt att minska antalet juridiska assistenter och juniora tjnster, men trots det
har andelen jurister kat p arbetsmarknaden. Det kan finnas flera olika frklaringar till
detta, en av dem kan vara att vi ftt allt komplexare regler.66 Skrpningen av kraven
fr finansiella tjnster har till exempel inneburit ett behov av betydligt mer arbete
med regelefterlevnad. Andra frndringar inom ekonomin kan allts emellant vara
mer betydelsefulla fr utvecklingen n teknik.

Fr det tredje erbjuder sm, ppna ekonomier som Sverige, med ett sprk som endast
talas av ca tio miljoner mnniskor, mindre utrymme fr automatiseringens frdelar
n USA och Kina. Ofta kan man inte bara ta en amerikansk eller brittisk teknisk ls-
ning och infra den direkt i ett litet land. Det finns institutionella srdrag nr det gller
lagstiftningen, kulturen och konsumenternas nskeml, som man mste ta hnsyn till.
En del av den institutionella strukturen speglar starkt rotade konstitutionella regler,
bland annat uppdelningen av ansvar mellan regionala och centrala myndigheter. Regi-
onala myndigheter, som kommuner och landsting, ansvarar fr mnga tjnster dr
digitaliseringen har potential att ka effektiviteten, till exempel fr kollektivtrafik
och sjukvrd, men ven fr system som underlttar att hitta och betala fr parkerings-
platser. Men lokalt sjlvstyre innebr ocks att varje omrde kan ha sina egna speciella
lsningar, vilket gr det svrare att utveckla programvara som omfattar ett helt land.

Det hr r ett problem som i olika hg grad pverkar nstan alla lnder och det kan
naturligtvis lsas genom frivilliga verenskommelser om att anvnda gemensamma
format och standarder. Men att n konsensus kan ta tid och bli freml fr tolkning
eller frndring som ger upphov till oskerhet och frdrjer tgrderna. Inga av de
hr hindren r dock overstigliga, men de kan sakta takten i utvecklingen, eftersom
skalavkastningen tar lngre tid att n eller r osker. Smskalighet ger lgre avkast-
ning frn samma investeringskapital. Sprket r ett annat problem, som skapar fr-
delar fr engelska och andra stora sprk. Men att dra slutsatsen att sm lnder r
skyddade frn automatisering frn digitala fretag med hemvist i Silicon Valley r
ett misstag. En mer proaktiv instllning r att det kan finnas mjligheter fr entre-
prenrer att utveckla smarta metoder som tar hnsyn till de lokala frhllandena,
iEuropa och p andra platser, innan de stora plattformarna knackar p drren.

Och till sist kan befintliga regler och hotet om nya regler avskrcka entreprenrerna.
Den hr frgan diskuteras vidare nedan. I det hr sammanhanget rcker det att notera
att en del fretag kan ha en affrsmodell som r helt beroende av sm detaljer i lag-

66
Se Flster (2015).

46
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

stiftningen och av hur lagen tolkas. r det till exempel rtt att de som tillhandahller
plattformarna i delningsekonomin har arbetsgivaransvar, som ett aktuellt domstols
beslut Kalifornien antyder? Sdana frgor kan allvarligt begrnsa, eller till och med
helt underminera, mnga aktrer inom delningsekonomin. Frgan r likaledes vem
som har ansvaret fr en olycka orsakad av ett freml som skapats av en 3D-skrivare,
av ett sjlvkrande fordon eller av frlorade knsliga personliga data (om till exempel
hlsa och sjukdomar)?

Vad innebr detta sammantaget fr hur snabbt automatiseringen framskrider? ena

sidan gr Europas rigida arbetsmarknad ytterligare automatisering attraktiv som
ett stt att ka flexibiliteten. Demografin gr dessutom automatiseringen attraktiv
eftersom den kanske kan lsa problemet med brist p kvalificerad arbetskraft.
andra sidan utgr regelverken och de institutionella skillnaderna ett hinder fr alla
lnderna i Europa, utom de strsta, dr bristen p storskalighet inte r en faktor.

Men, som vi framhller ovan, hur omfattande automatiseringen blir kommer i

grunden att bero p de politiska besluten. Att de bidrar till att skapa arbeten inom
den privata sektorn r en absolut frutsttning fr smidiga frndringar. Aktuell
forskning visar att ett stort antal nya jobb har skapats i Sverige. En underskning av
svenska data visar att 190,000 nya jobb skapades under perioden 19902009. Det
var nettoresultatet av att ca 3,4 miljoner jobb skapades medan 3,2 miljoner jobb
frsvann.67 Under ett genomsnittligt r omsattes ca en femtedel av jobben p arbets-
marknaden, ngot som visar ekonomins frmga att anpassa sig till efterfrgan och
ny teknik. De flesta jobben skapades inom tjnstesektorn, medan sysselsttnings
andelen minskade inom tillverkningsindustrin.

2. Ekonomiska krafter
Frndrade ekonomiska krafter fr fotfste
Delningsekonomin snker transaktionskostnaderna fr mngder av olika tjnster. De
omfattar allt frn egenpublicering av bcker och webbskningar efter information till
utfrandet av hushllstjnster. Det ger flexiblare arbete, srskilt i europeiska lnder
som har ett starkt skydd fr insiders p arbetsmarknaden. Men i takt med att allt fler
mnniskor arbetar som frilansare eller p begran inom delningsekonomin utstts de
ocks fr hgre risker, eftersom de sociala skyddsnten frmst r avsedda fr heltids-
anstllda som tcks av kollektivavtal.

Av alla krafterna som digitaliseringen ger upphov till kommer delningsekonomins

ekonomiska effekter sannolikt att st fr den snabbaste frndringen och bli en av
de viktigaste. Delningsekonomin r ett stort steg som frndrar vrt stt att arbeta
och organisera vra liv p ett helt annat stt n andra mer gradvisa frbttringar.
Eneffekt som redan mrks r att behovet av att ga saker har minskat eftersom man
enkelt kan hyra dem med en musklickning. Det innebr ocks frre resurser som str
outnyttjade , till exempel i form av bilar som normalt str parkerade nstan hela
dagen eller rum som inte anvnds i villor eller lgenheter. Entreprenrer r p jakt
efter nya nischer med outnyttjade resurser och varor som kostar mycket, till exempel
formella aftonklnningar eller borrmaskiner som bara anvnds en eller tv gnger
varje r. De frndrar ocks hur efterfrgan p hushllstjnster uppfylls med hjlp
avteknik som sparar tid och bland annat gr det lttare att handla matvaror, f
stdning utfrd och skicka blommor.

67
Heyman et al. (2013).

47
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Sdana tjnster r naturligtvis inte nya, men de kan smidigt matcha tillgng och efter-
frgan och erbjuder lga kostnader p ett stt som har stor betydelse fr ekonomin.
En aktuell underskning utfrd av Freelancers Union fann att 34 procent av alla som
arbetar i USA har deltagit i aktiviteter inom delningsekonomin. Den europeiska del-
ningsmarknaden r inte lika utvecklad som den amerikanska, men blir allt viktigare,
srskilt i Storbritannien. Europas rigida arbetsmarknader innebr att delningsekonomin
p sikt kan f nnu starkare effekt, svida inte utvecklingen stryps av nya regler.

En viktig faktor som kan pverka delningsekonomin r att motivet fr gande hller
p att frndras. Varfr ga en bil om man med ett enkelt klick kan bestlla en trans-
port nr man behver den? Varfr kpa en festklnning om man bara ska anvnda
den en eller tv gnger? Varfr inte hyra ut lediga rum i lgenheten? Incitamenten att
utnyttja resurser effektivare finns verallt, men de hller p att f ett helt nytt genom-
slag p grund av betydligt lgre transaktionskostnader.

Digital renomm allt viktigare?

Nytt med delningsekonomin r ocks hur tekniken kan skapa frtroende fr transak-
tioner. Att hyra ut sitt hem eller en plats i sin bil till en frmling krver ju en hel del
frtroende nr allt kommer omkring. Men delningsekonomin lyckas utvecklas nd,
precis som humlan trots sin storlek faktiskt kan flyga.68 En kombination av affrs-
sinne och teknik har skapat metoder att minska riskerna och utveckla frtroende i
sdana transaktioner. Tekniken kan kontrollera kvaliteten och plitligheten p olika
stt och lter bda parterna i en transaktion recensera och betygstta varandra en
tvvgskommunikation fr att skapa frtroende. Digitala recensioner hller p att bli
lika viktiga som digitala data och r sannolikt vrdefullare/anvndbarare n analoga
identifikationsmetoder. Ett krkort visar ju vem man r, men sger ingenting om hur
plitlig och frtroendeingivande man r. Det finns faktiskt uppgifter om mnniskor
som har svrt att f Uber-resor fr att de har ftt lg rating fr tidigare resor.69

Digitala recensioner blir allt viktigare och r drmed ett stt att etablera en frtroende
ingivande renomm p ntet. Men frgan r om man br kunna verfra historiken
fr en person som har god renomm p ett visst omrde till andra omrden? Det kan
tnkas att marknaden inte fungerar p det hr omrdet och ger upphov till ungefr
samma problematik som gr det svrt att verfra pensionskonton mellan olika banker.
Drfr kan det behvas policytgrder som banar vgen fr digitala recensioner. nnu
s lnge har man i stort sett lmnat det hr omrdet i fred, sannolikt av goda skl,
men myndigheterna kan behva anta en samordnande roll och leda arbetet med att
skapa normer. Varfr inte skapa en gemensam digital marknad fr recensioner?

Digital teknik ger bttre kapitalutnyttjande

En annan viktig effekt av digitaliseringen, som inte r begrnsad till delningsekonomin,
r att det krvs mindre kapital fr att starta ett fretag. Tung industri kommernatur
ligtvis fortfarande att krva omfattande investeringar och reklamkostnaderna kommer
att frbli hga. Men mnga aktiviteter kan simuleras med datorer, finjusteras och fr-
bttras. Idag kan entreprenrerna n ut till sina kunder mycket enklare och till lgre
kostnad med hjlp av en hygglig dator och en internetanslutning. Arkitekter har sedan
lnge ritat byggnader med hjlp av 3D-program, men nu fr dagens kraftfulla datorer
utvecklingen nnu ett steg framt. En arkitekt kan till exempel f strre kontroll ver
alla aspekterna p en konstruktion, frn placeringen av elledningarna till rrdragningen

68
Jmfrelsen mellan en humla och vlfrdsstaten finns bland annat i Thakur m. fl. (2003).
69
Se till exempel Streitfeld (2015b).

48
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

och berkningen av all utrustning som krvs fr produktionen. Estetiska frndringar

(typ, frgsttning, osv.) kan infras p ett enhetligt stt i hela konstruktionen, ungefr
p samma stt som formatering av text hanteras automatiskt i en ordbehandlare.

Distributionskostnaderna kommer sannolikt ocks att bli lgre i takt med att anvnd-
ningen av 3D-skrivare kar och delningsekonomin blir allt populrare. Den sedan
lng tid tillbaka pgende trenden att minska antalet mellanhnder fr transaktioner
inom bankvsende, industri, logistik, bokutgivning, etc. kommer att bli nnu
starkare. Varfr lagra och frakta kostsamma reservdelar nr de kan skrivas ut p
begran? 3D-utskrifter ger lgre distributionskostnader och snabbar upp bde kon-
struktion och leverans. De kan anvndas inom industrin och sjukvrden och fr
konsumentprodukter. I Kina och i Nederlndarna har man byggt fullskaliga hus
med hjlp av 3D-skrivare.70 Men det finns ocks tillmpningar fr 3D-skrivare som
ger bttre livskvalitet genom effektivare sjukvrd. 3D-utskrifter bidrar exempelvis
till lgre kostnader fr specialkonstruerade arm- och benproteser. Det finns ocks
exempel dr kirurger skapar 3D-kopior innan komplicerade ingrepp utfrs fr att
frutse komplikationer som kanske inte skulle kunna upptckas frrn patienten
lgp operationsbordet.71

Internet of Things (IoT) och big data ger mjlighet att f betydligt strre kontroll
ver alla aspekter av produktionen. Detta och 3D-utskrifter r revolutionerande
teknik som pverkar alla led i produktionen, men utvecklingen av IoT kan frdrjas
av rttsliga hinder, frmst frgor om konsumentskerhet, upphovs- och patentrtt
(mer om detta nedan).

Nr det gller efterfrgan hller tekniken p att frndra fretagens metoder att n
ut till konsumenterna med information och reklam. Alla aktiviteter p internet gene-
rerar mngder med data och en del av dem r mycket vrdefulla fr fretag som
vill n speciella mlgrupper. Personuppgifterna har faktiskt kallats fr en helt egen
typ av tillgng (asset class) fr att ange hur viktiga de hller p att bli jmte traditio-
nella tillgngar som man handlar med p finansmarknaderna. Det har skapat mnga
sekretessproblem. Telekomoperatrerna har enorma mngder med information om
hur mnniskor frflyttar sig och kan dra slutsatser om deras livsstilar och vanormed
utgngspunkt frn sdana data. 72 Konsumenterna sknker frivilligt bort mnga sdana
data fr att f tillgng till kostnadsfria tjnster till exempel e-post, ntverk och kartor.
Men det kan ocks vara s att mnniskor faktiskt inte r medvetna om hur mycket
information de faktiskt frmedlar genom att bara logga in och flytta muspekaren
ver skrmen.73

Den ekonomiska effekten av personuppgifterna r att fretagen kan mlinrikta kam-

panjer fr specifika produkter p ett mycket trffskrare stt. De kan ocks erbjuda
olika priser till konsumenterna, lgre priser fr konsumenter som just str i begrepp
att lmna webbplatsen utan att kpa ngot och hgre priser fr konsumenter som
r mindre prisknsliga. Omfattningen av prisdiskriminering kar, men det r likvl
en knslig frga som fretagen r varsamma med. ena sidan kan man ka effek-
tiviteten om man med hjlp av tekniken kan hitta det hgsta pris ngon r beredd
att betala. andra sidan kan detta reta upp konsumenterna, som naturligtvis har en
knsla fr rttvisa och lika behandling, och fretagen mste lra sig att skapa balans

70
Se till exempel Davison (2015).
71
Se Weintraub (2015).
72
Malte Spitz, en tysk riksdagsman, begrde och fick lagrade mobiluppgifter 2011 frn Deutsche Telekom som gav
mycket detaljerad information om hans frehavanden; i Sverige har rstadius och Larsson (2015) gjort en liknande
underskning.
73
Se till exempel Sveriges Konsumenter (2014) och Bylund (2013).

49
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

mellan den effektivitet som tekniken erbjuder och mnniskornas reaktioner nr de ser
hur den anvnds i praktiken. Det finns rapporter om konsumenter som blir arga nr
de utelmnas frn kampanjer riktade till vlbestllda personer. De hr frgorna r inte
nya, men tekniken vinklar dem p ett nytt stt och gr det angelget att hantera kon-
sumenternas reaktioner, srskilt p omrden dr konsumenterna ltt kan byta till en
annan leverantr.

Den mikroekonomiska revolution som digitaliseringen driver fram har

makroekonomiska konsekvenser som beslutsfattarna nnu inte frstr
Frndringarna som vi diskuterat i avsnittet ovan kommer att f avgrande effekter
fr tillvxt och arbetstillfllen. Entreprenrer i delningsekonomin hller redan p att
pverka det vergripande resursutnyttjandet och ka effektiviteten p vissa omrden
samtidigt som den hittar nya nischer med arbetsuppgifter som tidigare helt enkelt inte
utfrdes p grund av matchningsproblem och hga transaktionskostnader. Den tek-
niska frndringen blir allt snabbare om man jmfr med tidigare perioder och effek-
terna kommer att bli mrkbara inom breda sektorer i ekonomin.

De frndringar som diskuteras ovan kan komma att ha stor pverkan p hur vi
bedmer lget i en konjunkturcykel och dess effekter p ekonomin, inte minst p
sysselsttningen och inflationen. Mer konkret kommer mindre resurser i termer av
arbete och kapital kunna ge samma output. De kvalitativa effekterna r sjlvklara,
men deras storleksordning och omfattning kommer att pverkas av flera faktorer
ochinte minst hur regelverken reformeras.

En implikation r att det kommer att bli svagare inflationsimpulser frn en given
resursniv i ekonomin. En bil som mestadels str oanvnd under sin livstid kan
anvndas i mycket hgre utstrckning och en 3D-skrivare snker kostnaderna fr
lagring och distribution. Man kan allts f mer gjort med mindre kapital. Men resul-
tatet kan blir mer n bara ett bttre utnyttjande av befintliga resurser och kan ven
komma att pverka produktivitetstillvxten.

Mnga OECD-lnder har haft en dmpad produktivitetstillvxt under de senaste

ren. Detta kan delvis bero p mtproblem i statistiken74 och de lngsiktiga cykliska
effekterna av skuldnedvxlingen efter finanskrisen. Men professor Robert Gordon
vid Northwestern University beskriver ocks en rad andra faktorer som hmmar till-
vxten, bland annat ldrande befolkningar och hga statsskulder.75 Det kan vara s
att IoT, big data, 3D-skrivare och all annan ny teknik som r p gng kommer att
frbttra tillvxtpotentialen ver tid och balansera de motverkande faktorerna. Men
detta r en process som kan ta lng tid. Hur fort det gr kommer i hg grad att bero
p hur myndigheterna anpassar regelverken.

P medellng sikt, kanske inom de nrmast tio till tjugo ren, kommer frdelarna
med digitaliseringen sannolikt att ka tillvxtpotentialen och drmed kommer de
krafter som kan hlla tillbaka inflationen sannolikt att verka p ett eller annat stt
under en lng tidsperiod. Digitaliseringen r uppenbarligen inte den enda kraften som
p medellng sikt pverkar inflationen, men den kan vara en av de mest missfrstdda
i regeringskanslier och centralbanker. Den direkta priseffekten fr konsumentvaror,
i synnerhet elektronik, r frsts tydlig. Men de indirekta effekterna av resursutnytt-
jandet i ekonomin kommer sannolikt att bli n mer signifikanta och varaktiga. Det

74
Se till exempel Coyle (2015).
75
Se Gordon (2014).

50
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

r en utmaning fr centralbankerna och de politiska institutionerna att lra sig frst

vilka frhllanden fr fretagen som pverkar ekonomin, annars riskerar man att
formulera en politik som str p skakig grund.

3. Ekonomiska utmaningar i en digital vrld

Nr det gller framtiden och effekterna av digitaliseringen kommer institutionernas
frmga att hantera chocker i Sverige att testas p ett annat stt n under finanskrisen
2007 eller den inhemska krisen i brjan av 1990-talet. Hndelserna ovan var visser-
ligen mycket svra och problematiska, men det starka fokus och den dramatik som
de gav upphov till bidrog till att underltta den politiska anpassningen. Nr det gller
digitaliseringen liknar effekterna p arbetsmarknaden i hg grad effekterna i samband
med den ldrande befolkningen: de uppstr gradvis, frn r till r, sannolikt utan ngon
avgrande hndelse som frebdar en stor frndring. Det kan bli svrare fr myndig-
heterna att ndra kurs, nr frndringar intrder gradvis trots att de slutliga effek-
terna p vlfrden kan bli betydligt mer omfattande.

Utmaningar fr myndigheterna regelverk

Mot bakgrund av den lngsamma produktivitetstillvxten i OECD-lnderna r det
angelget att snarast genomfra strukturreformer i ekonomin s att tillvxten inte
iondan bromsas av ldrande befolkningar och hga privata och offentliga skulder.
Dessvrre har takten p strukturreformer p produkt- och arbetsmarknader saktat
ned i OECD-lnderna, inklusive Sverige, p senare r.

Det r naturligtvis inte s att f regler alltid r bttre. Regler r ndvndiga fr att
skapa trygghet och frtroende p marknaderna. Utan adekvata regler och normer
skulle marknadsekonomierna sannolikt begrnsas av en brist p frtroende och
hgre transaktionskostnader. Men regelverken har utkats under rens lopp, sanno-
likt i betydligt hgre grad n vad som krvs fr att skydda konsumenterna. Europe-
iska unionen gr stora anstrngningar fr att frbttra det fria utbytet av varor och
tjnster. Och man har gjort viktiga framsteg, men sprkliga och kulturella hinder inom
EU gr merarbetet med de terstende reglerna, bland annat krav p momsredovisning
och normer fr skerhet och information som varierar inom de 28 medlemsstaterna,
mer betungande n vad motsvarande regler gr i USA.

Dessutom kvarstr protektionistiska krafter som gr det svrare att starta nya fretag
och minska regelbrdorna. Det frsvarar monopolvinster fr etablerade fretag nr
konkurrensen begrnsas. De stndiga kraven p tillsynsmyndigheterna att standardi-
sera och skrpa skerhetskraven har i praktiken en tendens att vga ver de intressen
som strvar efter att inte ka komplexiteten. Det finns en bengenhet att skapa nnu
fler regler vilket ytterligare spr p komplexiteten.

Digitaliseringen gr sitt intg samtidigt som den redan snriga och komplexa regel-
striden pgr och den skapar nya frgor som tidigare inte tagits upp eller uppfattats
som intressanta. Det r sant att patent och immateriell egendom r omrden som har
diskuterats under lng tid. Musik- och frlagsbranscherna var de frsta som kmpade
emot digital teknik som gav mjlighet till enkel kopiering och distribution. Det har
tagit lng tid att utveckla nya affrsmodeller som ger innehllsproducenterna mj-
lighet att ta betalt fr sitt arbete, frmst genom prenumerationstjnster fr musik,
dagstidningar och bcker p internet. Frndringen pgr fortfarande.

51
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Fr lagstiftningen finns det tv utmaningar som r viktigare n alla andra. Fr det

frsta finns det regler som utgr ett stort hinder p den inre digitala marknaden och
som mste undanrjas. De frmsta frdelarna med den digitala tekniken i form av
lga marginalkostnader och enkel distribution urholkas av att det finns 28 omrden
med olika regler inom EU. Det r egentligen inte frmst de olika momssatserna som
r problemet, utan redovisningskraven, begrnsningarna fr dataverfring ver
grnserna och rutinerna. Reglerna fr skerhetsinformation varierar ocks, p samma
stt som reglerna fr lagring och hantering av konsumentdata.

Fr det andra hmmas expansionen av delningsekonomin av oskerhet om vilken

modell fr regelverken som ska tillmpas. Det kan rcka med en liten frndring av
befintliga regler eller av tolkningen av dem fr att en affrsmodell ska omintetgras.
Det finns entreprenrer inom hela ekonomin som r i frd med att fundera ut hur
de ska kunna tjna pengar och leverera vrde till andra fretag och konsumenter. De
kommer sannolikt att uppmrksamma behandlingen av Uber och Airbnb, som r de
ledande rebellerna inom delningsekonomin. Den skadligaste sidan av oskra regler
rden dimma av oberknelighet som hindrar att nya ider blir genomfrda.

Det borde vara angelget att klargra den rttsliga stllningen fr personer som arbetar
i delningsekonomin och frbttra det sociala skyddsntet fr frilansare. I Sveriger
hela ekonomin strukturerad fr att behlla de omfattande avtalsfrhandlingarna fr
heltidsanstllda mellan fackfreningarna och arbetsgivarorganisationerna. Om antalet
frilansare kar kraftigt kan frutsttningarna fr modellen frsmras och dess legi-
timitet ifrgasttas. Det r en utmaning fr vra institutioner att utforma avtal och
system p ett sdant stt att de hanterar hur risker pverkar olika individer och sam-
tidigt befrmjar den flexibilitet p arbetsmarknaden som krvs fr att minska arbets-
lsheten och behlla konkurrenskraften.

Sektorer dr delningsekonomin brjar utvecklas str infr mngder av regelproblem

som nnu r olsta. En vanlig frga r om yrkesnormer ska tillmpas fr mer amatr
mssiga typer av aktiviteter och var man i s fall ska dra grnsen. Vilka skerhets-
normer br tillmpas fr Airbnb-uthyrning jmfrt med hotell? Br skattesatserna
fr smskaliga verksamheter vara lika hga som fr stora bolag? Vilket ansvar har
innehavare av peer-to-peer- (P2P) eller peer-to-business (P2B)-fretag i hndelse av
en konkurs? Varfr har hushll smre skatteregler fr P2P-ln n fr noterade finan-
siella tillgngar placerade i investeringssparkonton (ISK) hos en bank?

Hinder fr sjlvkrande fordon r ett srskilt besvrande exempel p hur lagstift-

ning faktiskt kan motverka skerheten. Mnga r vertygade om att sjlvkrande
bilar kommer att orsaka betydligt frre olyckor och minska antalet dda och skadade
i trafiken. Men frgan om det juridiska ansvaret i hndelse av en olycka hmmar
utvecklingen av sjlvkrande fordon.

Om myndigheterna frsker skapa separata digitala regler fr varje marknad kommer

arbetet att ta mycket lng tid. Dessutom utvecklas tekniken snabbt och nya fretag
kan hitta metoder att gra saker som kanske behver omfattas av regler som fort
farande inte nnu finns p plats. Drfr borde myndigheterna anstrnga sig fr att
etablera regler i form av grundprinciper som kan tillmpas p alla omrden. Specifika
rekommendationer om sdana principer faller utanfr ramen fr den hr rapporten,
men ngra kommentarer kan vara p sin plats. Fr det frsta mste skerhetskrav
givetvis ven fortsttningsvis beaktas, men vara rimliga i frhllande till frvntade
utfall, vilket inte r fallet fr sjlvkrande fordon. Fr det andra br varor och tjnster
beskattas mer likformigt. Skillnaderna mellan bcker (en vara) och e-bcker (en tjnst)
r alldeles fr stor i nstan alla lnder. I Sverige r moms-skillnaden nstan tjugo pro-

52
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

centenheter. Det r ett hinder som ven pverkar mnga andra omrden, bland annat
3D-skrivare. Fr det tredje br ansvaret i hndelse av fel eller personskada klarlggas
i generella termer. Fr det fjrde br lagar om byggnation och andra kommunala
regler harmoniseras s att det mjliggr vinster med stordriftsfrdelar.

Om principer av generell bring infrdes skulle utformningen av regler som tar hnsyn
till heterogenitet p specifika omrden kunna g betydligt snabbare. Om varje sektor
inom ekonomin istllet fr egna uppsttningar med regler som inte har mycket gemen
samt med varandra r det uppenbart att frdelarna med den digitala tekniken inte
kommer att kunna utnyttjas och produktivitetstillvxten kommer att hmmas. Det r
naturligtvis viktigt att reglerna inte ger otillbrliga frdelar fr skilda intressegrupper
utan istllet stder innovation och bidrar till den kreativa frstrelse som blev ett
begrepp genom den bermda ekonomen Schumpeter.

Mer specifikt kan man sga att hur regeringen organiserar sitt arbete utgr en en
utmaning fr att undanrja hinder som hller tillbaka frdelarna frn digitaliseringen.
Ansvaret fr digitala frgor vilar hos en bostads-, stadsutvecklings och it-minister
som r anknuten till Nringsdepartementet. Men ansvaret fr de flesta hindren och
utmaningarna som kan frdrja utvecklingen av digitaliseringen ligger hos andra
departement, frmst Justitiedepartementet (dataskydd, rttsligt ansvar, patent och
upphovsrtt), Finansdepartementet (skatter, finansregler) och Arbetsmarknads
departementet. Varje departement har stor makt inom sitt eget omrde och lter sig
sannolikt ogrna rubbas frn sina positioner. Vinsterna, kostnaderna och riskerna
med digitalisering br vgas strategiskt p hgsta niv. Fr att snabba upp versynen
av regelverk som frdrjer digitaliseringen br ngra it-frgor sannolikt samordnas
inom statsministerns kansli, tminstone under de r som arbetet med att reformera
regelverken pgr.

Utmaningar fr myndigheterna livslngt lrande och digital kompetens

Mnniskor lever lngre n frr, men normen att man ska utbilda sig i unga r och
drefter endast i liten omfattning behver fylla p kunskap senare i livet r svr att
bryta. Det finns visserligen mngder med utbildningskurser i alla tnkbara mnen
frn affrskommunikation till datorprogrammering och specialkunskaper i olika
yrken, men den snabba tekniska utbildningen skapar ett behov av mjligheter till mer
systematisk kunskapspfyllning under hela vr livstid. Man ocks stlla frgan om
en del hgre utbildning har blivit fr tidskrvande i relation till dess gngbarhet p
arbetsmarknaden. Sdana frgor har blivit allt viktigare i takt med att befolkningarna
ldras och digitaliseringen frstrker dem.

Tillgng till kvalificerad arbetskraft r en viktig frga fr fretagen, men det r inte
sjlvklart att bara mer utbildning r den rtta lsningen. Professor Alison Wolf vid
Kings College hvdar faktiskt att vi investerar fr mycket i utbildning samtidigt som
kvaliteten r fr lg p vissa omrden.76 Ett r i utbildning r ocks ett r d man
inte arbetar, vilket ytterligare belastar finansieringen av de sociala vlfrdssystemen
som utvecklades i en tid d andelen av befolkningen som arbetade var strre i rela-
tion till antalet pensionrer.

Vi skulle snarare behva mer utbildning och kompetens som direkt efterfrgas p
arbetsmarknaden fr att f ett bttre frhllande mellan antalet lediga platser och
antalet arbetsskande. Vi behver ocks metoder att komplettera de inhmtade kun-
skaperna under hela livet -- inte bara fr att hlla jmna steg med den tekniskautveck

76
Se Wolf (2011).

53
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

lingen utan ocks fr att kunna byta karrir mitt i livet, en frga som betonas ien
rapport om demografin frn Framtidskommissionen.77 Vi behver bttrekvalifika
tioner rent generellt och i synnerhet mer digital kompetens, ngot som understryks
i en aktuell rapport till regeringen frn Digitaliseringskommissionen.78 En aktuell
underskning frn Eurostat visar att mnga medborgare inte har ens elementr it-
kunskap.79 Till viss del handlar det om en generationsklyfta, men det illustrerar ocks
behovet av kontinuerlig kompetensutveckling.

Det sagda reser tv stora utmaningar. Den frsta r hur frbttringen av kompetensen
ska stadkommas fr att bttre ta hnsyn till fretagens behov. Den andra r hur
detta ska finansieras. Incitamentet fr fretagen att finansiera utbildningen av sin
personal frsvagas av risken att de anstllda ska lmna fretaget och ta med sig sin
kompetens till en annan arbetsgivare, men det r ocks den sortens rrlighet som
krvs fr att arbetsmarknaden ska fungera effektivt.

Nr det gller hur inlrningen av digital kompetens ska anordnas r den statliga/
privata modellen UK Tech Partnership en modell som kanske kan vara lmplig fr
Sverige. Dess verksamhet samfinansieras av staten och nringslivet och med en rela-
tivt liten personalstyrka samordnar den utbildningsbehoven hos befintliga fretag.
Den fokuserar ocks p att utveckla den digitala kompetensen i skolorna och i
synnerhet p att f fler flickor att delta.

En srskild frga gller huruvida digital kompetens br ing i vra baskunskaper,

p samma stt som matematik och naturvetenskap. Vissa argument talar fr att den
offentliga sektorn br st fr en del av finansieringen, srskilt av utbildning som r
mer generell, men att utbildning som ger mer specifik kompetens i hgre grad br
finansieras av fretag eller privatpersoner. Var man br dra grnsen r en svr frga
som behver belysas.

Mer lrande i lmpliga doser under livet r sannolikt till nytta fr produktivitetstill-
vxten och gr det mindre sannolikt att mnniskor som behller sina jobb under lng
tid inte kommer att kunna hitta nya uppgifter om fretag omstruktureras.

Utmaningar fr fretagen regler, ny konkurrens och nya kompetenser

Perioder med snabba tekniska frndringar reser flera tydliga utmaningar fr fretagen:
beredskap att frndra affrsmodell eller krnprodukter om efterfrgan frndras,
uppdatering av kompetens och personal . och anpassning till gllande regler,

Frndringsproblematiken ser till viss del ut som en spegelbild av diskussionen om

myndigheterna ovan. Etablerade fretag str infr helt andra utmaningar n digi-
tala entreprenrer. Den mest grundlggande frgan fr entreprenrer r om affrs-
modellen fortfarande r hllbar, om reglerna skulle frndras eller tolkas p ett
ovntat stt. De mest skadliga verkningarna av myndigheternas utredning rrande
Uber r de signaler som den snder till alla andra aktrer som funderar p att starta
fretag srskilt frsts inom delningsekonomin. En srskild frga r om frilansare
ska betraktas som anstllda. Om s r fallet mste fretagen betala socialfrskrings-
avgifter och andra skatter. Men det finns ocks mnga andra oskra faktorer, bland
annat om de digitala plattformarnas gare r juridiskt ansvariga fr handlingar som
utfrs av personer som anvnder eller tillhandahller tjnster, exempelvis fr uthyr-
ning av lokaler eller peer-to-peer ln.

77
Se Blix (2013a).
78
Se SOU (2015a).
79
Se Figur 5.1 i kapitel 5 i denna rapport.

54
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Det r svrt att ge ngot mtt p hur den hr oskerheten pverkar entreprenrer
som funderar p att starta fretag inom delningsekonomin. Men den omfattande
mediarapporteringen om utmaningarna fr Uber r svr att missa. De sm vinst
marginalerna eller initiala frluster, som det oftast handlar om fr nystartade
fretag gr den sortens risker extra skadliga. Lgre risker fr affrsmodellerna och
enklare insteg i delningsekonomin kan rent generellt skapa stora frdelar fr ekonomin
som helhet: resurserna utnyttjas effektivare, miljbelastningen minskar, arbetstagare p
lg- och medelinkomstniv kan minska sin konsumtion och sitt gande och vid behov
hyra resurser (srskilt bilar), och jobb inom delningsekonomin ger flexibla arbetstider.
Och delningsekonomin kan slutligen ocks minska omstllningssvrigheter som fljer
i klvattnet av fortsatt automatisering av arbete.

Etablerade fretag, srskilt fretag inom mogna branscher, str infr helt andra
utmaningar. P vissa omrden kommer nya digitala fretag att frska ta sig in och
konkurrera. Att ha ordet digital i affrsmodellen innebr vanligen lgre kostnader,
snabbare verksamhet och att fretaget lttare kan experimentera och n sina kunder.
Att vara stor ger inget ogenomtrngligt skydd mot konkurrens frn digitala fretag.
Sony kmpar sedan lng tid fr att frvandla sina produkter och tjnster och mta
utmaningen frn internet och sker sig fortfarande fram, och Microsoft verrumplades
ocks p ett stt som gr att deras kvasimonopol p operativsystem nu ser mycket
srbarare ut.

Framgngsrika industrifretag r inte lika srbara som nystartade fretag eftersom

investeringskostnaderna r hga och att komma in p marknaden krver ofta stora
kapitaltillgngar och frmga att acceptera relativt osker avkastning frn en sdan
investering. Biltillverkarna hotas inte av nystartade fretag, utan istllet av stora
fretag som Google och Tesla. Bankerna utstts fr konkurrens frn fintech-bolag,
som nu i allt strre utstrckning finansieras av institutionella placerare och banker.
IBM har gjort insteg p sjukvrdsmarknaden genom sina senaste frvrv och stora
advokatfirmor och frvaltningsbolag som normalt tar mycket bra betalt fr sina tjnster
brjar hotas av delningsekonomin och automatiseringen av kunskapsintensiva jobb.
Mnga arbetsuppgifter kan idag automatiseras, srskilt fr juridiska tjnster.

Erfarenheterna frn vissa fretag som misslyckats visar hur tekniska frndringar
kan innebra strategiska utmaningar, ngot som p engelska brukar benmnas
incumbents curse. I USA r Kodak ett utmrkt exempel p detta, men det finns
ocks mnga andra. I Sverige hade bde Hasselblad och Facit starka positioner
p marknaden, men personalen hade specialistkompetens p frldrad teknik och
frsljningsorganisationerna fokuserade p att leverera tjnster som inte lngre
behvdes nr den elektroniska tekniken infrdes80. Ett av de strsta problemen med
att frndra ett fretag under ett teknikskifte r risken att redan etablerade intressen
ifretaget frsker skydda fretagets befintliga produktutbud och endast fokuserar
p utveckling. Att frga kunderna r sannolikt ingen lsning eftersom de kanske inte
vet hur tekniken kan frndra tjnsten eller produkten som de behver. Och nr fr-
ndringen kommer kan det hnda att det inte lngre finns tillrckligt med tid att
frndras81. Detta hnder gng p gng.

Fretagsledningen kan inte eliminera de strategiska riskerna genom att enbart fr-
lita sig p strukturer fr befintliga produkter, nr man beslutar vad man ska satsa p.
Detta r en mycket svr omstllning som frsvras ytterligare av att den nya tekniken
inledningsvis kan innebra frluster och i ett senare skede ven kan komma att ta

80
Se Sandstrm (2011, 2013)
81
Se Bower and Christensen (1995)

55
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

upp vinster frn existerande produkter. Bilindustrin r ett utmrkt exempel p detta.
Infrandet av sjlvkrande bilar innebr sannolikt att det totala antalet bilar som
kps kommer att minska. Det gller ven andra aspekter p delningsekonomin, bland
annat uthyrning av verktyg och hgtidsdrkter, dr den totala konsumtionen, tmins-
tone inledningsvis, kan komma att minska. Det finns en mjlighet att kunderna kper
mer av samma varor, men det r lika sannolikt att de stoppar besparingarna i fickan
eller spenderar dem p andra tjnster.

Slutligen terspeglas behovet av att hitta och behlla kompetens, i synnerhet digital
kompetens, i diskussionen ovan om utmaningarna fr politiken. Fretagen kommer
att ha en nyckelroll i arbetet med att informera utbildningssektorn om nringslivets
krav och mste ocks engagera sig mera aktivt i frgorna kring livslngt lrande. Det
snabba teknikskiftet innebr att personal som inte utvecklar sina digitala kunskaper
kommer att hamna p efterklken nr det gller produktivitet och lneutveckling.
Ilnder med rigida regler p arbetsmarknaden och kollektivavtal med klausuler som
fljer principen om frst in, sist ut kommer det att finnas ett nnu strre behov av
attskerstlla kontinuerlig kompetensutveckling. Att delta i utbildningskurser en eller
attpar dagar kommer inte att rcka utan kompetensen mste utvecklas p ett mera
strategiskt stt. Skattesystemet br ocks frndras s att det blir frdelaktigare fr
fretagen att betala fr personalens kompetensutveckling.

Att frvxla korrelation med kausalitet kan leda till allvarliga misstag
Ytterligare en risk som kan komma att bli mer markerad r att frvxla korrelation
med kausalitet nr man fattar affrsbeslut eller analyserar policy. De enorma mng-
derna med data om konsumenttrender, livsstilar och ntvanor r vrdefulla fr fre-
tagen och ger tillgng till information som kan anvndas till att n specifika grupper
med reklam och frsljningsargument. Nr det gller policy finns det mngder med
vrdefulla data om mnniskors beteende som kan ge beslutsfattarna en bild av mn-
niskors reaktioner p skattereformer eller andra frndringar. En oemotstndlig lock-
else kan vara att lta de enorma datamngderna ge upphov till bedmningar som
frefaller att vara mycket exakta. Stora datamngder innebr vanligen mindre oskra
mtningar och mtningar p hela populationen eliminerar all stickprovsoskerhet.
Allt fler fretag har tillgng till big data eller sljer tkomst till big data. Men den
synbarliga precisionen frn sdana analyser kan vara falsk och kanske inte tl en fr-
ndring av frutsttningarna.

De som utfr kvantitativa studier behver vara frsiktiga med att inte tolka korrela-
tioner som kausala samband. Erfarenheterna frn Google, som frutsg en influensa-
epidemi med hjlp av skfrekvenser, r ett bra exempel p hur brckliga korrelationer
kan vara.82 Nr frhllandena frndras kan konsumenternas beteende ocks frndras.
Riskerna med vantolkning av korrelationer kar om de r stabila under lng tid och
sedan pltsligt frndras p grund av en ofrutsedd hndelse. Banker som tillhanda
hll ln i USA fre finanskrisen frutsg framtida risker med hjlp av enorma mngder
data om tidigare tvngsfrsljningar av fastigheter. Men de omfattande problemen
med s kallade sub-prime ln, som visserligen var en en liten del av den totala mark-
naden, skapade sedan en dominoeffekt till resten av marknaden, och de historiska
korrelationerna var som bortblsta. Samma sak kan hnda med konsumentundersk-
ningar och andra analyser som baseras p big data.

82
Se Lazer et al. (2014)

56
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Konsekvenserna av att anta att det finns kausalitet, trots att den inte existerar, r inte
bara ett rent akademiskt felsteg, det kan f allvarliga ekonomiska fljdverkningar och
ven leda till felaktiga affrsbeslut. Ett exempel som visar var sdana risker finns r
det nya fenomenet now-casting (nutidsprognoser), som anvnder enorma mngder
data frn webben (bland annat om mnniskors skvanor) som indata fr makro
ekonomiska prognoser. Ett exempel: r fler skningar efter arbetslshetsunderstd
pntet ett tecken p att arbetslsheten kommer att ka? En sdan korrelation kan se
trovrdig ut, men mnniskors beteenden kan frndras ver tid s att korrelationen
blir svagare, som i exemplet med Google och influensaepidemin. Slutsatsen r inte att
korrelationerna saknar vrde men att de br kombineras med andra typer av data
och modeller fr att stdja de slutsatser som dras.

4. Vad kan komma att hnda?

Digitaliseringen r en kraft som inte gr att stoppa, ungefr p samma stt som globa-
liseringen. Men dess hastighet och effektivitet kan pverkas kraftigt beroende p hur
reglerna frndras. Det kan vara p sin plats att visa vad som kan hnda med hjlp
av enkla scenarier. Sdana br helst utfras i en makroekonomisk modell, men det r
svrt eftersom frgan egentligen handlar om hur frndringar av policy-parametrar kan
pverka beteenden. Men de tidigare modellerna r baserade p historiska mnster
som kanske endast delvis r relevanta, srskilt makromodeller som aggregerar mnga
olika beteenden. Fr att ngot begrnsa frgorna fokuserar vi endast p tre centrala
variabler: produktivitet, sysselsttning och inkomstskillnader.

I Ruta 2 diskuterar vi tre olika scenarier. De r inte avsedda att tolkas som prognoser
utan som illustrationer. Utgngspunkten r att utfallen beror bde p regelverken och
hur samhllet lyckas underltta anpassningsperioden p arbetsmarknaden. Mlet r
frmst att befrmja produktivitetstillvxten och incitamenten fr arbete och samtidigt
minska problemen med volatila hushllsinkomster. Som Digitaliseringskommissionens
delrapport framhller kommer kompetensutveckling, i synnerhet av digital kompe-
tens, sannolikt att innebra en stor frbttring av mjligheterna till sysselsttning,
produktivitetstillvxten och lnerna.83

I scenarierna behver vi inte utg ifrn dramatiska anpassningar p arbetsmarknaden

som en reaktion p ytterligare automatisering. I takt med pensionsavgngen kommer
fretagen istllet att vlja om de ska erstta personal som gr i pension eller automa-
tisera arbetsuppgifterna. Och eftersom andelen personer i arbetsfr lder minskar i
befolkningen kan automatiseringen ocks vara en reaktion p bristen p kvalificerad
arbetskraft.

83
Se SOU (2015a).

57
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Ruta 2. Scenarier fr Sverige

I den hr rutan skisserar vi policy-effekterna p ekonomin genom. Det hr tankeexperimentet

motiveras av den forskningslitteratur som understryker hur avgrande de politiska institutio-
nernas reaktion p den tekniska frndringen r.84 Scenarierna har visserligen konstruerats
med utgngspunkt frn frhllandena i Sverige, men de r nd relevanta fr andra EU-lnder
med omfattande offentliga vlfrdssystem.85 Vi diskuterar effekterna av olika politiska tgr-
der med utgngspunkt frn dagens niver fr produktivitet, sysselsttning och ojmlikhet. Vi
anvnder ett tidsperspektiv p medellng till lng sikt, en tidsperiod p ca tjugo r.

Makroekonomiska antaganden fr alla scenarier

Den makroekonomiska motvinden fortstter att hmma produktivitetstillvxten.
Ofrndrade strukturer fr arbetsmarknader, inklusive starkt skydd fr ordinarie anstll-
ningsavtal.
De sociala skyddssystemen kvarstr ofrndrade.

Scenario 1. Politik som vanligt: fortsatt svag produktivitetstillvxt, frre jobb skapas n
de som frsvinner
Detta r ett trampa-p scenario, dr politiken frndras lngsamt. Fragmentarisk frnd-
ring av regelverken ger sannolikt upphov till lngsammare produktivitetstillvxt. Frdelarna
med digitaliseringen uppstr i huvudsak lngsammare n kostnaderna fr anpassningen av
arbetsmarknaden. De makroekonomiska motvindarna som Robert Gordon diskuterar uppvgs i
s fall inte i tillrckligt hg grad av innovationsvinster och vrdet som de skapar. De ldrande
befolkningarna och de hga skuldsttningsniverna i OECD-lnderna pverkar redan ekono-
mierna, samtidigt som frdelarna med den nya tekniken och hur snabbt den infrs beror p
regelverken. Om reglerna fr olika omrden tgrdas separat ansvarsfrgan fr sjlvkrande
bilar, patent och intellektuell egendom i samband med 3D-skrivare och graden av ansvars-
skyldighet som garna av digitala plattformar har fr transaktionerna de frmedlar kan
utvecklingen sannolikt mtas i minihertz i stllet fr megahertz. Frgor som mste infras i
lagstiftningen och avgras genom avtalsfrhandling brukar bli tidsdande och det finns en
god anledning till det eftersom trovrdigheten och frtroendet fr systemet beror p faktorer
som dess rttvisa (verklig och upplevd) och frutsgbarhet.
Fortsatt automatisering av industrier och tjnster innebr att fler jobb frsvinner och struk-
turarbetslsheten kan bli hgre under en vergngstid eftersom mnga sektorer pver-
kas samtidigt och med starkare kraft n tidigare. Automatiseringsprocessen underlttas av
arbetstagare som gr i pension och drmed ger fretagen fortsatta mjligheter att snka
kostnaderna utan att aktivt behva minska andelen anstllda med sedvanliga anstllnings-
avtal. Problemen med att hitta personer med rtt kvalifikationer kommer dock ven fortstt-
ningsvis att utgra ett hinder fr rekryteringen, srskilt fr hgt kvalificerade jobb. I Sverige
kommer den stelbenta bostadsmarknaden, med brist p hyresbostder, ocks att frsvra
strukturanpassningen. Det r svrt fr dem som arbetar att flytta till nya platser om det inte
finns ngra bostder dr.
Nya jobb skapas naturligtvis hela tiden, men entreprenrerna hmmas av oskra regelfr-
ndringar som kan pverka deras affrsmodeller. Frgan r om deras affrsmodeller kommer
behva anpassas till regelverk som underminerar sjlva grundidn eller, mindre dramatiskt,
om frndringarna tillfr en komponent som fr kostnaderna att ka s att de verstiger
84
kostnaderna i affrsplanen.
85

84
Se till exempel Acemoglu et al. (2005).
85
Mekanismerna som beskrivs r ocks relevanta fr USA, men de politiska institutionerna och startpunkterna skiljer
sig t. EU-lnderna och USA str infr liknande utmaningar nr det gller att undanrja hinder till frmn fr digitali-
seringen, men frhllandena p arbetsmarknaden i USA r spass annorlunda att anpassningarna sannolikt kommer att
vara av ett annat slag, och faller utanfr ramen fr den hr rapporten. Noteras br att i USA har inkomstskillnaderna
redan kat kraftigt.

58
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Det scenariot kan komma att innebra en successiv, men kanske inte srskilt markant, kning
av inkomstskillnaderna. En del jobb kommer att ha lngsammare lneutveckling och fr en del
mnniskor kommer det att finnas regler som hindrar dem frn att arbeta i delningsekonomin.
Men vi kommer nd att f se fler superstjrnor som blir vinnare, entreprenrer med frmga
att ervra stora marknadsandelar och sedan frbli dominanta med hjlp av vlknda varumr-
ken, men ocks p grund av hinder som gr det svrt fr andra att ta sig in p marknaden och
konkurrera.

Scenario 2. Protektionistisk motreaktion: betydligt lgre produktivitetstillvxt, hgre

arbetslshet, strre inkomstskillnader.
Man kan betrakta det hr scenariot p samma stt som Scenario 1, kombinerat med en ekono-
misk politisk reaktion som understryker rttigheterna fr insiders och etablerade fretag dr
destruktiva otillbrliga frdelar dominerar p bekostnad av vrdeskapandet. Etablerade fre-
tag kan med hjlp av regelverken skapa hinder som gr det svrare att ta sig in p marknaden
och drmed begrnsar konkurrensen. Det innebr att frre jobb skapas och att incitamentet
att gra sig av med den mnskliga arbetskraften till frmn fr automatisering blir starkare.
Det kar polariseringen p arbetsmarknaden, s att jobb p medelinkomstniv hamnar lngre
ner p kompetensskalan samtidigt som den lngsamma produktivitetstillvxten innebr en
svag reallneutveckling fr stora grupper, dvs. en polarisering av bde jobb och inkomster.

Scenario 3. Bejaka frndring: bttre produktivitetstillvxt, fler arbeten, ofrndrade

inkomstskillnader.
Om versynen av reglerna inleds strategiskt med ngra f ledande principer om konsument-
skerhet och andra frgor, minskar behovet av att utarbeta nya regler fr varje omrde som
kan bli freml fr digitalisering. Om sdana principer tillmpas fr varje enskild frga inom
alla sektorer kan versynen av regelverket snabbas upp samtidigt som egenheterna som
finns inom varje omrde kan diskuteras igenom ordentligt. Till exempel principer och trsk-
lar fr hur och med vilka skattesatser frilansare br betala skatt i delningsekonomin och vilka
regler som br glla fr att garantera rimlig konsumentskerhet.
Vissa befintliga regler kan granskas fr att avgra om kraven r ondigt strnga nr det gl-
ler risker och utfall. Sjlvkrande fordon kommer till exempel sannolikt att rdda mnga liv
eftersom antalet trafikolyckor minskar, men lanseringen av dem drjer nd p grund av os-
kerheten, srskilt vad avser ansvarsfrgan. Reglerna br strva efter neutralitet mellan olika
alternativ, men utfallen br sikta p att frmst ltta p ondiga regler istllet fr att hja rib-
ban fr nya potentiella aktrer. Att hja ribban fr tilltrde kan hindra konkurrens och leda till
monopolartad prissttning som drabbar konsumenterna.
Det hr scenariot kommer att skapa betydligt fler tjnster inom delningsekonomin s att
antalet befintliga jobb kan hlla jmna steg med antalet arbeten som frsvinner. Genom att
kombinera enkel anpassning mellan sektorerna med frbttrad yrkeskompetens, i synnerhet
digital kompetens, underlttas frndringarna s att vlfrden blir stabilare fr stora grupper.
Infrandet av delningsekonomin som en vsentlig del av arbetsmarknaden innebr att skill-
naden i rttigheter som tillkommer insiders jmfrt med outsiders med oskra jobb mste
tgrdas. Fr att upprtthlla vlfrdssystemets legitimitet mste socialfrskringssyste-
mens trygghetsnt erbjuda ett effektivare generellt skydd ven fr frilansare. Skyddsnten
mste kalibreras s att de jmnar ut inkomsterna ver tid, men fr inte vara s genersa att
de minskar viljan att arbeta.

Sammanfattning av effekterna p medellng till lng sikt (de nrmaste 20 ren)

Produktivitet Sysselsttning Inkomstfrdelning
Scenario 1. Sedvanlig politik 0 - -
Scenario 2. Protektionistisk motreaktion - -- --
Scenario 3. Bejaka frndring + + 0

OBS: Scenarierna som beskrivs i den hr rutan r stiliserade. Ett plustecken indikerar en frbttring, en nolla ofrnd-
rade frhllanden och ett minustecken anger en frsmring. Ett dubbelt minustecken indikerar kraftigare frsmring.

59
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

5. Ngra avslutande ord

Hur bra det kommer att g fr olika lnder i tider av teknologisk frndring beror i
hg grad p deras institutioner. Mycket stelbenta institutioner kan komma att trycka
in digitaliseringens mjligheter i redan etablerade analoga regelfack p ett stt som
hindrar utvecklingen och gr anpassningen ondigt svr. Regeringar och myndigheter
som inte lyckas hitta rtt balans mellan ledarskap och samarbete med nringslivet
kan ocks komma att misslyckas. Priset fr detta blir hgre arbetslshet och lng-
sammare produktivitetstillvxt. I tider av frndring r det mycket viktigt att fra en
politik som i grova drag r rtt. Misstagen fr betydligt allvarligare konsekvenser n
i normala tider. Felen kan vara resultat av aktiv politik eller av avsaknad av sdan
politik. Avsaknaden av tydliga rttsliga regler fr delningsekonomins affrsmodeller,
3D-skrivare, big data och andra tekniska framsteg hmmar utvecklingen och hin-
drar entreprenrerna frn att utveckla och frbttra sina verksamheter och f njda
kunder.

Men det finns tgrder som kan underltta en smidigare anpassning p arbetsmark-
naden och bidra till att man tillgodogr sig frdelarna med digitaliseringen
Att ta ut lgre skatt p (mnskligt) arbete. Hg skatt p arbete i Sverige strker
det redan starka incitamentet att automatisera arbetsuppgifter ytterligare. Skatte-
avdragen fr hushllstjnster br utkas, inte minskas.
Att minska oskerheten kring de regler som pverkar delningsekonomin. Det ger
flexibla mjligheter till justeringar och lgre risk fr hgre strukturell arbetslshet.
Att skapa strre mjligheter fr livslngt lrande. Ett lngre yrkesliv och snabba
tekniska frndringar kar risken att vissa kompetenser blir frldrade tidigare i
vra yrkeskarrirer. Uppdatering av yrkeskompetens kommer att bli av stor bety-
delse fr att minska risken fr ofrdelaktiga framtidsutsikter p arbetsmarknaden.
Att minska skillnaderna i socialfrskringsskydd mellan anstllda och
egenfretagare.
Att infra principer fr reglering av den digitala ekonomin som kan anvndas
till att snabba upp reformerna p mnga skilda omrden, och drmed minska
behovet av att utarbeta srskilda regler fr varje omrde. Detta krver ett nra,
smidigt och kontinuerligt samarbete mellan jurister och ekonomer. Fr vi inte
det, kan den lga produktivitetstillvxten bli seglivad och demografins och stats
skuldens makroekonomiska motvindar fortstta att dmpa tillvxten.

Den digitala revolutionen kommer sannolikt att frbttra svl livskvaliteten som
effektiviteten p arbetsplatsen och fortstta att frbttra vr fritid. Men ingenting r
sjlvklart nr det gller hur smrtfri frndringen kommer att bli. Institutionernas
reaktion r en nyckelfaktor fr kad produktivitetstillvxt utan risk fr hgre arbets-
lshet och vxande klyftor.

60
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

1. Creative destruction in
a digital world
Thou aimest high, Mister Lee... Consider thou what the invention could do to my poor subjects.
It would assuredly bring to them ruin by depriving them of employment, thus making them
beggars. Queen Elizabeth I, 1589, on rejecting the request of William Lee for a patent on a knitting
machine, quoted in Acemoglu and Robinson (2013).

1.1 Introduction
Digital techniques, advances in robotics and neuroscience are increasing opportunities
and in some cases improving quality of life for many people. Self-driving cars, already
old technology, are navigating through government regulatory obstacles. Autonomous
cars have the potential to vastly reduce the number of accidents, improve mobility for
everyone who either cannot drive or prefers not to, and to help commuters and travelers
spend time more productively or pleasurably.

But technological advancements are also bringing major disruptions to the economy
within a compressed time period. While the disruptions are not unprecedented the
changes during the Industrial Revolution were undoubtedly more significant they
are occurring in a time of higher expectations for the level of welfare that society
should provide its citizens. In the aftermath of the financial crisis, unemployment and
public debt are significantly higher in many OECD countries and productivity growth
has slowed. Many European countries have seen the rise of populist extreme left and
right-wing parties. Long-term trends of aging populations are affecting the outlook
for growth. Might digitalization86 be the succor that boosts growth and improves
prosperity?

Digitalization presents both opportunities and challenges. It will be a balancing act to

harness the benefits of digitalization while mitigating the downsides. In Why Nations
Fail, Acemoglu and Robinson (2013) discuss the institutional features that support
growth and can explain why some nations become rich while others remain poor.87
Notably, it is important to have political institutions that balance the interests of var-
ious groups, the rule of law, protection of property rights, and predictability so that
people and firms can expect to reap the benefits of long-term plans or investments.

What does this imply looking forward? How well we cope with the effects of digital-
ization will depend on striking a balance among institutional features conducive to
stability and openness to recasting policies that get in the way of growth.

While this may sound easy, it is not. In essence, it is a clash between old and new. While
the old will have to change in some ways, it is not superfluous to note that that the
new is not always right. Those who argue that digitalization changes everything
tend to underestimate the importance of norms and traditions (Chapter 3) and the
role of regulation (Chapter 4). Although the impact of technology depends on how
institutions react by adapting rules and regulations for goods and services, other
trends and changes may matter as much and sometimes more.

86
Digitalization is typically defined as the representation of objects into zeroes and ones that can be stored on computers.
We use the term more broadly for the ways that computers affect work, consumption and leisure.
87
It is beyond the scope of this report to discuss these issues in detail. For a reference to the academic literature see
Acemoglu (2005). For an overview of economic growth, see for example Aghion and Howitt (2006).

61
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

It is a mistake to underestimate how technology may have political and economic

impact, for example, with respect to how social security and labor law affect the dis-
tribution of risk in society, such as that between insiders and outsiders (Chapter 5).
Moreover, aging populations will make it more difficult to find new workers in some
sectors and regions, thereby strengthening the case for increased automation. Germany
is already beginning to experience this challenge in some sectors, notably the auto
industry. But the challenge may also be felt keenly in sparsely populated regions in
Sweden, for example in the need for health care and elder care.

Challenges from digitalization and inertia in institutions

Digitalization has a number of major implications for:
Individuals as workers and consumers.
Firms competing in the global economy.
Governments regulating firms and setting the rules for competition, public welfare
and social security.

Our institutions and habits produce considerable inertia to change and that is as it
should be up to a point. Rules and regulations are there to protect us, but the situa-
tion becomes somewhat incongruous when, in some instances, digital alternatives are
likely to save a lot of lives. For example, self-driving vehicles are widely thought to
be less prone to accidents, because, unlike humans, they are not impaired by lack of
sleep, overconfidence or intoxication. Indeed, one of the technical challenges for self-
driving vehicles is that they obey traffic rules, which becomes more of an issue when
interacting with the average driver.

There is a disconcerting tendency to underestimate the challenges involved, especially

by governments in Sweden and elsewhere, as if the myriad of small changes underway
somehow might gel into a coherent whole. In the Swedish government, there is also
a tendency to put the legal aspects above all other matters, such as economic effi-
ciency, when it comes to adapting rules that allow us to benefit from digital technolo-
gies (Chapter 4). Perhaps some of this insouciance about the impact of digitalization
comes from the dotcom crash of 2000. Things were blown out of proportion then, so
surely they are again?

There are several sobering reasons why things are different now. With the ubiquitous
use of the smart phone, digital technology is making inroads in most aspects of our
lives. Many of the fastest changes underway are consumer driven. A large share of the
population uses smart phones and tablets in their daily lives, for work and for leisure.
Consumers making choices via digital media are changing the conditions for firms.
Consumers are becoming more powerful as a collective through the use of rating
systems for services combined with the ease of switching providers.

Markets are also becoming more mature and payment systems for digital transactions
are available that people are beginning to trust. Digital technology is providing net-
work effects, speed, and buyers and sellers matching services at low or zero marginal
cost. A lot of business models are being disrupted or ubered in the digital parlance
everything from taxis to education and financial institutions and media.

62
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Economists on the other hand, borrowing some jargon from finance, have a tendency
to look at digitalization as a plain vanilla structural change, no different from what is
always in progress. Somewhat unhelpfully, the impacts of digitalization are becoming
more significant just when the economics profession is still trying to recover from
being blindsided by the financial crisis. The new research being conducted is primarily
in microeconomics, on understanding network effects, competition and different
markets in isolation. There is an urgent research agenda to draw insights from the
microeconomic literature as to what this may imply for the economy as a whole,
the macroeconomy.

Policy institutions and central banks may overestimate inflation pressures

Digitalization may lead policy institutions, such as government agencies, ministries
and central banks to misjudge the state of the economy. One reason for this is related
to how economists measure the degree of unused resources in the economy and infla-
tion pressures. In particular, digitalization may improve the efficiency of the economy,
leading to better use of existing resources, and may imply higher degrees of price
competition than before. Wage competition, especially for low-skilled jobs, may also
increase with the use of digital platforms. Inflation has already been subdued in many
countries for several years; see Figure 1.1. Although many other developments have
impact on inflation, part of the explanation may well be digitalization. This is dis-
cussed further in Chapter 2.

Figure 1.1. Inflation in selected countries, annual average change in percent.

5
Sweden

Germany
4
Italy

United Kingdom
3 France

United States

1
Source: Eurostat. HICP for EU countries.

-1

-2
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

63
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Job creation may be slower than job destruction during a transition period
While the change in inflation pressures on its own is a significant development, the
risks and issues for the labor market as a whole are more multifaceted. Althoughthe
changes are not coming overnight, ignoring the possibility of major upheaval may be
risky; small policy errors can easily compound over time with unnecessarilyonerous
effects on social welfare, especially since short bouts of unemployment all too easily
can become long ones, or even permanent, when skills are lost or self-confidence
erodes. In recent decades, technology and humans have been largely complementary;
see Autor (2014). While some jobs are lost, machines have by and large enabled
people to work more efficiently, which has been accompanied by a need for higher
skills. The effect has been productivity and real wage growth, improving disposable
income and raising living standards for broad groups over time.

But even assuming that the long-run effect of digitalization on employment is neutral,
which we believe, it may still be the case that new jobs are created at a slower pace
than old ones are destroyed. Aging populations and shrinking workforces will make
the transition somewhat easier at the overall level, but in some sectors we may see an
increased challenge in matching available jobs to the skill sets of people entering the
labor market. During a transition period, probably counted in decades rather than in
years, it will be a challenge to help smooth adjustment in the labor market and miti-
gate the downsides in terms of increased inequality while not putting shackles on the
impetus for growth.

Job polarization set to increase further

There is strong evidence today of widespread job polarization, in which middle
income earners are being squeezed; see for example Goos et al. (2014) for an exam-
ination of several countries. Essentially, the middle class is becoming thinner. This
has also occurred in Sweden, but only on the positive side with an increase in high
income earners; see Adermon and Gustavsson (2015). Job polarization is likely to
increase further because digital technology is beginning to also replace highly skilled
workers. The effect may be either low wage growth for large groups and/or increased
risk of unemployment, topics that we explore further in Chapter 3. Low wage growth
for broad groups may in turn risk further political polarization and social unease.
The US has already experienced significant wage polarization for broad groups that
have received little real wage growth. By contrast, real wage growth in Sweden has
been strong for the last two decades.

Secular stagnation and digitalization are different mechanisms

The notion of secular stagnation has been raised recently by Summers(2014). This
hypothesis contends that we have entered a new era in which demand remains below
potential but that due to excess savings, interest rates have to be more negative to
boost the economy. This idea is somewhat controversial, see for example Dolan (2015),
Sandbu (2015a) and Wolf (2015), but it is undoubtedly so that economists need to
adjust their models and estimates in an era of negative interest rates. It should be noted
that digitalization and secular stagnation are different mechanisms; digitalization in
this regard concerns mainly microeconomic efficiency that leads to more price com-
petition and better use of resources in the economy, while secular stagnation is an
explanation of the macroeconomic effects of a savings glut. Thus even though they
both ultimately concern output, the issues and channels differ.

64
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

If the nature of work changes, there may be pressures on social welfare institutions
The legitimacy of the modern welfare state, such as in Sweden, hinges on peoples
acceptance of the tax burden in order to enjoy a high level of universal public wel-
fare, such as schooling, health care and infrastructure. Looking ahead, the combined
forces of digitalization and aging populations imply challenges for how to maintain
the trust and legitimacy underpinning the welfare state.

Though challenges from an aging population are by now fairly well understood, see
for example Blix (2013a,b), the effects of digitalization remain relatively unexplored.
Admittedly, there has been considerable discussion about artificial intelligence (AI),
not least the question of people being replaced by robots. While AI might be possible
some time in the future, robots are far from exhibiting the versatility of humans. The
famous DARPA challenge in the US has teams of robots competing in an obstacle
course. While there have been tremendous improvements in recent years, self-driving
cars not least, robots need a lot of help to operate. Robots tend to be bad at things
that humans are good at, such as dexterity and cognitive skills. Even the huge develop
ments in self-driving vehicles, while very impressive, are not evidence of human cog-
nitive skills. Self-driving cars do not drive on roads so much as they drive on digital
maps, a point emphasized by MIT professor David Autor.

In the most recent DARPA competition, robots had to perform an obstacle course
akin to performing salvaging operations inspired by measures that might have saved
lives during the Japanese Fukushima Daiichi nuclear meltdown. The performance of
the robots is still substantially subpar to that of humans on the obstacle course, but
descriptions of performance were upgraded from watching grass grow in previous
competitions to watching golf, see Markoff (2015a, b).

However, the lackluster performance of these robots should not provide false com-
fort for what may happen to labor. The idea of a robot sitting a desk (possibly with
a coffee mug) perhaps serves to make them more human and less threatening. But
in practice, jobs can be divided into parts and some of those can be outsourced to
cloud computing or to the sharing economy with on-demand labor, see for example
OConnor (2015b). The extent to which this is possible is already considerable and
on the increase, a topic discussed further in chapter 3.

In this report, we will focus on the economic and social consequences of advances that
have already been made or are in the pipeline in the medium term. Arguably, the idea
of robots taking over at some future point tends to distract from the more pressing
issues that have not yet received enough attention in the public debate. What regulations
are barriers to benefiting from technological change? Do we need to make changes
to improve the labor market and reduce the risks of further job polarization and
increased unemployment induced by technology?

One salient issue is that digitalization may lead to more freelance work compared to
the situation today, especially in relatively rigid European labor markets with strong
rights for insiders, see for example OConnor (2015a). More freelance work and an
on-demand economy will help provide flexibility that will make structural change
easier. In the so-called sharing economy, the owners of digital platforms typically
do not see themselves as employers only as conduits between those who want to
buy and sell through digital platforms. Thus, these employers mostly do not provide
health insurance and other benefits. Are government institutions and labor market
partners prepared for such an economy if it becomes much larger than today? Since
the norm is designed around full-time work for a single employer, this may lead to

65
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

more people unable to fully qualify for social security benefits or unable to draw on
those benefits because of time constraints and the loss of business that absence might
imply for the self-employed. The sharing economy will be discussed in various con-
texts in Chapters 2, 3 and 5.

The challenge ahead

Adaptability has long been a hallmark among humans and there are good reasons
for thinking we will be able to cope with the changes brought by digitalization. But
coping can come in different forms, all from muddling through to actively and stra-
tegically making informed choices about the direction of change and how we might
best go about it. One could be forgiven for thinking that Europe has had quite enough
of muddling through during the years of the financial crisis. Alas, many institutions
do not seem sufficiently prepared for the changes underway.

But what exactly is the alternative to muddling through? In this report we will out-
line the challenges and indicate policies that would make the structural change easier in
terms of employment and welfare while contributing to improving productivity growth
(outlined in Box 2. Scenarios for Sweden in the Extended Summary). There is some
urgency in addressing the challenges as bad policies can have especially onerous effects
in times of rapid change. The changes from digitalization are broad, affecting social
behavior, leisure and politics, but we will primarily focus on economic effects of
digitalization.

Even if governments were to choose a muddling-through policy, people will still cope.
But the social costs may be higher. The starting point for adjusting to vast technolog-
ical challenges is far from ideal for many European countries: public debt and youth
unemployment are high; refugees from sectarian violence and civil war are coming in
great numbers. It should be high on the policy agenda to make the utmost effort to
benefit from digitalization while smoothing the worst downside risks.

1.2 Speed and scope of structural change

There is an implicit view in policy circles, especially in central banks and ministries of
finance, that digitalization is similar to previous structural change. This view implies
a risk of underestimating the broad scope of changes and the speed of adjustment.
Societies have a long tradition of adapting to change. The advance of electricity with
all its benefits and improvements for households and firms is a case in point. It was
a very broad change, affecting the whole economy, but its effects took a long time
to materialize (see Chapter 2). Grids had to be built, machines invented and so on.
There was plenty of time to adjust, even if some people were surely caught unawares
in the process.

If we take an even broader view, urbanization and the Industrial Revolution provide
a more dramatic transformative change. In Sweden, about 95 percent of the popula-
tion worked on farms or lived in rural regions around 1800, see Figure1.2; now, the
situation is virtually reversed. This is a massive change, but it occurred over a period
of two hundred years, thus allowing people and institutions time to adjust. From the
1970s on, urbanization has continued unfettered but changed character. In the last
few decades, people have been moving from smaller cities to major cities and so, if
anything, Figure 1.2 understates the continued increase in major metropolitan areas.

66
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 1.2. Urbanization, share of population in cities and rural areas, Sweden
100
Cities
90 Rural areas

80

70

60

50

40

Source: Statistics Sweden.

30

20

10

0
1800 1820 1840 1860 1880 1900 1920 1930 1940 1950 1960 1970 1980 1990 1995 2000 2005 2010

By contrast, a swift change that affects a specific area is difficult but as long as
the rest of the economy is unaffected, there are many opportunities for the people
involved to adapt and find new jobs and opportunities in other areas. In Sweden,
major changes occurred in the textile industry with the onset of a steady decline
starting in the 1950s and even more dramatically in the shipbuilding industry in the
late 1970s. We will discuss these and draw some lessons from them in Chapter 3.

In recent times, some large Swedish companies have cut their workforces rather dra-
matically. Ericsson, for example, implemented mass layoffs after the dotcom bubble
burst and in response to fierce competition from other network operators, such as
Nokia in Finland and Blackberry in Canada. Sometimes the reasons for the lack of
competitiveness are long in coming, but the fallout can be fast once the tipping point
is reached.

While these changes have been significant for specific sectors or locations, the possibil-
ities for new jobs within an industry or elsewhere often work out relatively well over
time. When Astra Zeneca decided to close its research department with about 1,200
employees in the Swedish town of Sdertlje, population about 90,000, this was a
major event. The people affected were highly skilled workers performing non-routine
and creative tasks. Although a major upheaval, the final tally turned out rather well.
A report from TRR (2015) shows that almost 95 percent of those affected found new
employment or became self-employed within a two-year period of turbulence and
adjustment.

How will digitalization affect highly skilled workers in other areas if several sectors
are affected simultaneously? The impetus to change will not be as fast as it was for
Ericsson, Astra Zeneca or Nokia but probably faster than what we are used to com-
pared, for example, to the introduction of electricity. In contrast to the researchers at
Astra Zeneca, most of whom found other jobs elsewhere, what if all researchers were
affected at the same time? This is what digitalization could bring.

The changes that have been underway in music, news and video are instructive. The
music industry has seen its business model uprooted with the advent of digital tech-
nology; journalism is still battling strong cost pressures and video streaming is replacing
physical discs sold in shops and available for rent.

67
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Perhaps the likeliest scenario for other skilled workers would be similar to the process
in journalism: unlike the closing of a factory or industry, there will be no dramatic
event that heralds a new age. But cost pressures and changing production processes
will bring gradual changes every year. Firms that fail to adapt the business model
may find a gradual erosion of profits. Such a process may thus increase the tendency
towards job polarization that is already under way.

1.3 Long-run trends

To assess the effects of digitalization, we will argue that it is necessary to combine the
evidence from established long-run trends, economic models, and anecdotal evidence.
Using only economic models would risk missing the specter of a structural shift, while
using only anecdotes risks over-interpreting a noisy signal. The long-run trends inthe
economy help us link the anecdotes to economic theory and make better sense of the
changes underway. Apart from technological progress, the most relevant ongoing
trends in this context are:
Aging populations
Continued globalization and increased trade
Further urbanization
Further growth of the service sector compared to industry
Slowdown in productivity growth

The slowdown in productivity growth may or may not be a long term trend. This is
akey topic for assessing the impact of digitalization and we will discuss it more
thoroughly in the next section.

Aging populations
The age structure of societies affects everything from the available work pool to the
size of preschool classes and demand for health care services and funding of pensions is
a big topic in most countries. One issue in particular is a challenge for welfare states,
that of sustaining growth in tax-financed public welfare services when the working
populations aligned with existing norms are shrinking. While an enormous literature
is devoted to aging from a policy perspective it is hard to muster political impetus
for appropriate policies because the changes occur slowly from year to year without
drama; see for example Blix (2013a,b). But the cumulative demands of digital skills
are a challenge for many older people unaccustomed to using computers, smart
phones and tablets.

But there are actually some similarities in the challenges stemming from aging popu-
lations and digitalization. Most obviously, digital technologies are making substantial
improvements in the lives of older people in areas ranging from health care to self-
driving cars. For the German auto industry, the lack of new workers as older ones
retire has led to robots taking over even more functions; see Bryant(2015).

There are also other similarities. While digitalization has an image of being fast moving,
its effect on the economy takes time to materialize: in 1987, for example Robert Solow
famously quipped, You can see the computer age everywhere but in the productivity
statistics. This issue is arguably still relevant today, as discussed in Section 1.4 below.
Moreover, just as life-long learning is key to a longer productive working life, learning

68
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

and retraining are key to employment and wage growth as the economy becomes
more digital (Chapter 5). People have to continue to learn new software and how to
operate it or risk wage stagnation or worse unemployment. Employees who do
not update their skills can survive for a few iterations, but with big technological
leaps, the skills mismatch may become more pressing, as artisans discovered with
theinvention of machines at the beginning of the Industrial Revolution.

The lack of workers in the German auto industry has increased demand for automating
work. This experience is likely to hold true for broad sectors of the economy. Many
OECD countries, including Sweden, have aging populations. In some sectors, finding
workers especially skilled workers will be hard. Thus, demography is increasing
the demand for automating work. In Chapter 3 we will explore this topic further and
argue that the challenge is not that work is disappearing but rather that the skills mis-
match is problematic and the risks of technology-induced unemployment for those
with inadequate skills will increase. Thus, the main concern is that income inequality
may rise further and increase tensions in the labor market.

Globalization
Globalization and continued increases in world trade have brought higher standards
of living and improved conditions, but also increased competition and pressure to
specialize, particularly for small, open economies. There are many parallels between
the effects of globalization and digitalization. Indeed, some of the literature on glo-
balization that discusses how some jobs in industrialized countries were outsourced
to countries with lower wages are very similar to current arguments about which
jobs are at risk of being automated; see Autor(2014), Blinder (2009), and Frey and
Osborne (2013). Simple assembly jobs and technical support are among the jobs that
have been outsourced, notably to Asian countries. This process has led to lower cost
of production in industrialized countries and development in the countries to which
tasks have been outsourced.

For some years, the trend to outsource continued. Though most of the high-end develop
ment of technology is retained in corporate headquarters in OECD countries, wages
have been pushed up in some Asian countries. With the advent of further digitalization
and reprogrammable robots, the nature of the outsourcing trend may change, although
it is too early to draw definite conclusions.

One possible shift is a change in outsourcing from Asian countries to other regions
where wage costs are still low, such as Africa or South America. But political instability
and barriers to trade may be an obstacle to this development. Another possible shift is
home-sourcing, a process in which previously outsourced tasks return to industrial-
ized countries but are performed by robots instead of manual labor and which could
lead to especially challenging welfare implications for Asian countries. For example,
some countries may lose their manufacturing base before highly skilled services evolve,
leading to so called premature deindustrialization, see Box 3.4. Labor market dis-
ruptions in chapter 3.

The effects of digitalization would likely be much less if the world was not already
highly interconnected through foreign trade and investment. Globalization has thus
paved the way for the network effects and increasing returns to scale that digital tech-
nology thrives on. A digital revolution without global reach would not generate the
leverage that is the elixir of life to digital platforms.

69
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Urbanization and the rise of the cities

There are two somewhat contradictory trends that concern urbanization and digi-
talization. Urbanization was a key ingredient in the Industrial Revolution; the influx
of people from farms to cities made for a climate of innovation. The literature on
growth emphasizes the role of cities as magnets for talent; see for example Glaeser
and Gottlieb (2009). Ambitious people are more likely to meet in cities and be stimu-
lated by interacting with other talented people. Also, there are larger pools of workers
available for employers looking to hire.

The dynamic cocktail of people with good ideas colliding to make something
even better is an ingredient in tech clusters. Arguably, however, it is really the same
mechanism as the advantages of having a city in the first place, but more focused on
technology.

While urbanization is thus part of the progress that has made technological inno-
vation possible, it may also, paradoxically, reduce the need for people to move to
cities. With the capacity to work remotely, more tasks can be performed off site. For
example, some aspects of education and simple interactions with doctors might be
accomplished with digital technologies. Machines in connected mines can also be
controlled remotely; see Ek (2014).

Service sector growth

Productivity gains in industry have driven overall productivity for the whole
economy; for a long time, industry has been able to increase output while decreasing
employment. Correspondingly, the service sector has increased in size and repre-
sents three out four jobs in many OECD countries, if both private and public sectors
are included. The private service sector accounts for almost half of employment and
has increased steadily over time; see Figure 1.3. By contrast, manufacturing has been
declining in terms of employment while maintaining strong productivity growth.

Figure 1.3. Employment in different sectors, Sweden

50 Manufacturing
Services

45 Public sector

40

35

30

25
Source: Statistics Sweden

20

15
80

82

84

86

88

90

92

94

96

98

00

02

04

06

08

10

12

14
19

19

19

19

19

19

19

19

19

19

20

20

20

20

20

20

20

20

70
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

At a cursory glance, industry seems very efficient and service sector productivity
growth somewhat lackluster. But it may very well be the case that part of the pro-
ductivity gain in industry is due to outsourcing functions that were previously done
in-house to the service sector. For example, clerical services, food catering, adminis-
tration, and sanitation might be turned over to outside contractors, enabling industry
to focus more on core functions.

In terms of the effect of digitalization, the feature to highlight is that manufacturing

has continued to improve efficiency for a long period of time and digitalization is
now making inroads towards improving efficiency in services as well. Looking ahead,
much of the hope for future productivity resides in improved services. We will return
to this point in Chapter 2 when we discuss the sharing economy, but first we turn to
an overall view of where productivity is headed.

1.4 Productivity growth

Productivity growth now and in a historical context
Productivity growth has slowed in major industrialized countries. The slowdown
occurred even before the financial crisis (see Figure 1.4). This compares to historical
GDP per capita growth, which was around 2 percent per year in Sweden during 1820
2000, a number on the same order in magnitude as many other industrialized countries.

Research from Angus Maddison shows that historical growth prior to the Industrial
Revolution was virtually non-existent and hence a growth rate of 2 percent is still
rather good. Using the Geary-Khamis dollar (see Figure1.5) we see that growth
only became significantly positive at the beginning of the Industrial Revolution.
Starting in the mid 19th century, growth picked up with a steady positive trend.88

Figure 1.4. Productivity growth in Sweden, Germany, France, UK, US

8
France

Germany

6 Sweden

United Kingdom

4 United States

6
Source: OECD.

8
1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013

88
Some BRIC countries, such as China, have managed to sustain even higher growth in the last decades by leapfrogging,
thereby skipping some steps on the development ladder by adopting the latest technology from OECD countries.

71
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 1.5. GDP per capita growth in Geary-Khamis dollars

30 000
France

Germany

Sweden
25 000
UK

20 000

15 000

Source: The Maddison Project (2013).

10 000

5 000

1955
1960
1965
1970
1975
1980
1985
1990
1995
2000
2005
1
1820
1825
1830
1835
1840
1845
1850
1855
1860
1865
1870
1875
1880
1885
1890
1895
1900
1905
1910
1915
1920
1925
1930
1935
1940
1945
1950
Note: The Geary-Khamis dollar is a hypothetical currency with the same purchasing power as the US dollar and is often
used to compare countries over longer time periods. For comparisons in the last few decades, the OECD publishes GDP
per capita corrected for purchasing power that accounts for changes in exchange rates and inflation rates.

Slowdown in productivity: temporary or permanent?

Is the recent slowdown in productivity growth a temporary setback or a return to a
norm that held prior to the Industrial Revolution? Robert Gordon of Northwestern
University argues that it is permanent or at least long-lasting. In a series of papers, see
for example Gordon (2014), he discusses headwinds for growth stemming from aging
populations, high public debt, diminished returns on education and outsourcing as well
as environmental factors. Using simple calculations, Gordon argues that most of the
productivity growth dwindles down to pre-Industrial Revolution levels of around zero.

That this is controversial is a bit of an understatement; see for example Bartelsman

(2013). While Gordon has evidence in the form of the recent slowdown to support
his case, many others contend his findings; see for example Brynjolfsson and McAfee
(2014), Ford (2009, 2015a), and in a Swedish context, see Blix(2015) and Boumediene
and Grahn (2015). Joel Mokyr, a colleague of Gordons at Northwestern University,
has for many years argued the opposite case; see for example Mokyr (2013). Gordon
essentially contends that recent advancements in digital technologies are much less
significant to productivity growth than the major discoveries of the last century, such
as electricity, cars, and airplanes. Further, he argues that the gains from digitalization
have already largely been realized, such as electronic banking, word processing, and
so on.

Ultimately, the validity of these arguments rests on data. Alas, the difficulties of
measuring quality and its implications have likely become more serious in recent
years; see Feldstein (2015) and Coyle (2015). Current GDP and inflation figures
might underestimate productivity and quality improvements. As physical goods are
transformed into digital services, the shift in consumption bundles in indices becomes
more problematic. Even with modern statistical techniques, it is difficult to adjust
goods, and especially services, for the improvements that occur over periods of time.

72
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

For example, a car with ABS brakes and modern safety features is a much different
product than a car made only a decades ago; likewise, self-driving cars of the future
will be of higher quality than cars today. Statisticians try to account for these differ-
ences but the methodological questions are difficult.

Gordons case rests on clearly identifiable macroeconomic factors that are likely to
impact industrialized countries in the years to come. Although not all countries have
high debt, the overall debt burden is high and may be a drag on the world economy
for quite some time; likewise, aging populations are common to most OECD countries
and regions of Asia, whereas countries in Africa tend to have younger populations.

The key question is whether the microeconomic innovations from digitalizationand

other technology can outweigh the headwinds identified by Gordon. We do not have
a complete answer to this, of course, but a few points in Gordons favor deserve to
be highlighted. First, the recent productivity slowdown has to be explained in some
other way (such as secular stagnation or measurement problems as mentioned above);
while lingering uncertainty from the financial crisis may surely delay recovery: the
slowdown occurred even before the crisis. Second, as the historical account above
shows, a high level of growth is not a law of nature. We cannot simply presume that
all things will return to high growth.

At first, it may seem that those who proclaim the fantastic new effects of new tech-
nology and Gordons pessimistic view are diametrically opposed but (with some effort)
they can be reconciled. If it takes a long time for digitalization to have broad effects
on the economy, productivity growth may be sluggish for some time. The productivity
boost from technology may be slower and weaker unless regulation removes the
obstacles standing in the way (see Chapter 4). A few highly efficient firms and inno-
vative firms may not be enough to carry the whole economy. Without broad diffusion
of technology, the effects on the economy may be limited; OECD (2015a).

New technology that may boost productivity

What is the effect of new digital technologies like smart phones, 3D printers, and new
services with digital platforms that can transform work and leisure? This report is in
large part motivated by the potential of such innovations. For example, 3D printers
are now capable of building large buildings and bridges, see Davison(2015) and the
Economist (2015i), replace limbs, see Mrozfeb (2015), and enable doctors to practice
before performing complex surgery, see Weintraub (2015). The costs of such replication
were previously either too high or simply impossible. For example, when an astro-
naut needed to repair the spaceship, the requisite tool could simply be sent to the 3D
printer on board. Other examples of digital technologies that save time or improve
efficiency are:
Swifter communication
Helping navigation with maps
Remote work
The sharing economy (see Chapters 2 and 3).

Technology is already improving the quality of life for people with disabilities, wounded
veterans, and senior citizens by helping them remain active. Most people will agree
that there are great innovations around the corner the question is when they will
have effect on the broader economy. And will they be in time to reduce the dampening
effects of aging populations etc. identified by Gordon?

73
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Moores law and productivity growth

Productivity growth is a broad topic, see for example Syverson (2011) and Jones(2015),
and we will cover only the discussion of digitalization. It is instructive to recapitulate
Moores law, which in many ways has come to symbolize the optimism about all things
digital. The original chart from Moores paper is depicted in Figure1.6, reproduced
from Poeter (2015). Essentially, it holds that the speed of microchips doubles every year
or every 1 years. So far Moores law has held true.

Figure 1.6. Moores law 19591975

A doubling in speed or some measure thereof implies a growth rate of one hun-
dred percent every year compared to the paltry 2 percent post-Industrial Revolution
average or the even more lackluster 0.2 percent in Sweden between 20072012 (see
Figure1.4).

If the speed of computer chips increases so much, why is productivity growth still so
slow? Even asking this question might seem a bit unfair, but the difference between one
input factor, the microchip, in parts of the economy does not automatically translate
to enormous output growth in the whole economy. Not even if the microchip indeed
was a key input into all products and services would the output growth automatically
be high. Supposing the speed of microprocessors were to grow by 1,000 percent, the
effect on productivity growth might still not increase by much in the short run.

The reason, of course, is that there are other input factors, notably human labor, other
factors of production etc. that only benefit from increased speed up to a point. Nord-
haus (2007) makes a similar point in noting that computers represented about 2.3
percent of total capital stock in 2000. An analogy might be a Ferrari: it can go very
fast on the roads, but if it has to cross a lake or river, it will have to slow down like
other cars. Productivity growth in the economy as a whole is likely constrained by
avariety of factors unrelated to technology:
High public debt or inflation.
Corruption, red tape and inefficiencies in legal institutions.
Poor education and lack of human capital.
Poor infrastructure, roads, communications and internet infrastructure etc.

74
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Recent research has tried to go beyond these factors. Knowledge-based capital (KBC)
concerns various forms of soft knowledge, management and organizational skills; see
OECD (2013). When it comes to digitalization, one key factor concerns how IT tech-
nology is used in the firm. Bloom et al. (2012a, b) have shown that American multi
nationals tend to have higher gains from IT investments than other firms due to better
people management. They are better able to integrate IT into the work process and
management can use the technology as a tool to improve efficiency. Experiences from
India with control groups have illustrated the key importance of how management
works in order to derive benefits from using IT. Having a fast computer in an other-
wise disorganized operation does little for productivity or efficiency; the benefits of
technology can be held back by the physical environment and poor work procedures.

1.5 Structure of this report

This report is organized as follows. The next chapter is devoted to discussing how
digitalization may change the way economists look at the economy. Chapter3 is devoted
to the labor market, and in particular the creation and destruction of jobs. Chapter4
covers challenges for firms and regulatory issues. Finally, Chapter 5 discusses Swedens
standing as an IT nation and in terms of, education and inequality. It concludes with
some reflections on how policy can smooth structural change. In particular, what
policies are conducive to productivity growth while limiting the increase in inequality?
The Extended Summary of the report contains some scenarios that illustrate the
effects that different policy choices may have on productivity, unemployment, and
inequality.

75
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2. Economics of digitalization
...digitalization has some features that suggest that many well studied economic models may
not apply... Goldfarb et al. (2015)

2.1 Introduction
Economics at its core is about understanding incentives and tradeoffs between different
choices and their implications for the economy and welfare. Digitalization has not
changed this but many other aspects of economics may be affected, a topic we will
explore in this chapter.

We will argue that digitalization brings many of the stylized features of basic economic
models closer to reality, making models based on competition and instantaneous and
frictionless market clearing more relevant than before. Essentially, digitalization implies
improved efficiency in a number of dimensions. But in bringing economic models
closer to the real world, paradoxically, some of the relationships estimated in histor-
ical data may need to be reevaluated and may lead to misleading policy conclusions.

In this chapter we organize the effects of digitalization on the economy into demand
and supply side effects at the micro level: that of the firm or the individual. On the
supply side, network effects, increasing returns to scale, and 3D printers are among
the factors leading to faster, cheaper production. It is well known that not all these
improvements are accounted for in official statistics. It is especially hard to measure
quality improvements, which has long been a challenge for statistical agencies. But
the mismeasurement may be more serious and systemic today than before, withthe
consequence that it is harder to assess the growth rate of the economy; see for example
Byrne and Pinto (2015), Coyle(2015), and Feldstein (2015). On the demand side,
increased price transparency and more global access are improving efficiency and
competition. Digital platforms may be thought of as combining the elements of both
improved supply and demand: by smoother matching of buyers and sellers they are
creating markets and opportunities even for goods and services for which transaction
costs and frictions were hitherto prohibitively expensive.

All of these microeconomic improvements have a variety of indirect effects on the

macroeconomy, total output (GDP), employment, inflation, etc., but these effects are
hard to measure. More importantly, we argue that they may confound models currently
used for policy analysis at fiscal institutions and central banks. Admittedly, somecon
servatism in not changing models too often is advisable. Moreover, it may be especially
difficult to separate the effects of digitalization from other factors in the aftermath of
the financial crisis. However, the improvements are gradual and therefore those who
want solid econometric evidence may have to wait for many years, by which time
policy errors may have accumulated and done untold harm to the economy.

What might these errors be? At the end of this chapter, we discuss the policy impli-
cations of issues primarily related to the measurement of capacity utilization (the
output gap), whereas we discuss matters related to employment in Chapter 3 and
those related to firms and regulation in Chapter 4. Although this chapter focuses on
the somewhat narrow topic of capacity utilization, this in itself brings central eco-
nomic policy issues to the fore.

76
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Digitalization is set to bring faster structural change at the microeconomic level com-
pared with recent decades, moving models in use especially macro models onto
shakier ground. First, digitalization will bring about more efficient use of existing
resources both capital and labor. We are probably experiencing this effect already
and it is likely to grow in importance in the years ahead. The explanation is partly
related to the sharing economy exemplified by more efficient use of the existing
total vehicle fleet. But it is also related to little or zero marginal costs and network
effects of expansion in several digital areas. Secondly, new technologies will also help
us make better use of existing resources. They may also be a countervailing force on
productivity growth weighed down by the macroeconomic headwinds discussed in
Chapter 1 (demography, public debt, etc.). But the strength and timing of productivity
improvements will depend on several factors, most notably on government regulation.

Even if as we believe digitalization will be employment-neutral in the long run, it

is likely that more jobs will be destroyed than created during a period of adjustment;
see Chapter 3. This would imply that medium-term equilibrium unemployment might
be higher than accounted for in current models that assume a largely unchanged rate
of structural change.

It is worthwhile to stress that digitalization may affect estimates of economic behavior,

output gaps, and medium-term equilibrium employment regardless of whether or not
preferences for work, leisure, and consumption change. The way firms produce output,
the way consumers select goods and how jobs are found may be altered because tech-
nology is changing the overall environment for work and transactions.

Of course, we cannot know how serious this issue is. As discussed in Chapter 1, pro-
ductivity growth is currently low in most developed countries, which may be partly due
to persistent effects of the financial crisis, partly cyclical, partly related to uncertainty
about future financial regulation, fiscal consolidation, or secular stagnation. As dis-
cussed in Chapter1, the effects of digitalization cannot easily be separated from
existing long run trends of ageing populations, globalization, and the increasing
employment share of the service sector.

But whatever the contribution of digitalization, it is likely to act as an accelerator for

other changes already underway. For example, when robots become even more versatile
than today, reprogrammable, or closer to artificial intelligence (AI), they will be able
to replace even more jobs than before, an issue we will touch on in the next chapter.
But before discussing how digitalization may be changing the economy, it is worth-
while to review the basic economic assumptions that underpin many models.

2.2 Foundations of economic models

Let us briefly discuss the standard models used in economics as a starting point to the
way digitalization may affect behavior and what it may mean for the models. This is
of course a very broad topic, but the ambition is to sketch some arguments that may
be useful when thinking about the broader effects of digitalization.

The standard microeconomic model taught in basic economics makes a number of

assumptions that are used to understand how incentives and actions can lead to
different outcomes depending on preferences, attitudes to risk, and other factors.

For the individual, the workhorse model assumes some form of utility maximization,
typically involving a tradeoff between work, leisure, and consumption. We work in

77
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

order to be able to consume, but there is assumed to be some disutility to working

and in equilibrium the marginal disutility from work relative to that of consumption
is directly related to the real wage. Of course, the set-up is highly stylized and modern
research has developed more realistic assumptions and investigated various ways of
relaxing these assumptions. But the fundamental idea behind the individuals utility
maximization remains one of the core principles of economics. Digitalization does
not change or alter the value of using the utility-maximization framework, but in some
dimensions, further discussed below, it makes the workhorse model less unrealistic
or more realistic depending on ones perspective.

For firms, the standard model assumes cost minimization of production and that
prices of goods and services are set at marginal cost when the market features perfect
competition and at marginal revenue when the firm is a monopolist. Between the
extremes of perfect competition and a monopoly, models explore various forms of
imperfect competition and price behavior with different outcomes depending on
factors such as market structure, demand elasticity, brand names, advertising, and
research and development.

2.3 Features of the workhorse model that are changing: production

There are several aspects of the economic workhorse model that may be affected by
digitalization. At the most basic level, it seems reasonable that peoples preferences
and attitudes to risk will not change due to digitalization though they may changefor
other reasons over time.89 But digitalization may be changing the exogenous factors
that affect economic agents decisions about work, leisure, trade, and so on. In this
section, we discuss how digitalization may change parts of the production function
compared to the standard model with regard to how goods are produced and dis
tributed. A stylized summary of all the potential ways that digitalization can affect
the economy is given in Box 2.1.

2.3.1 Increasing returns to scale due to network effects

Economists have long discussed increasing returns of scale due to network effects
but these have only been a minor part of the toolkit, mainly a theoretical curiosity:
most models are based on decreasing returns or constant returns to scale.90 For most
of physical production, this assumption is often a good starting point, but not so for
digital technologies. With increasing returns to scale, the fixed costs, such as invest-
ments, quickly become relatively small when more units or services are sold.

Increasing returns to scale are perhaps most easily understood in the context of tele-
phone services, where the parallel to digital technologies is informative. When there
are only two telephone users, they can only call each other; with three or more users,
the combinations rise very fast and so does the usefulness of the service. Compared
to telephone services with digital services such an effect is magnified because once the
computers are connected, each activity benefits from the network. Telephones, instead,
allow only one activity. It is as if the telephone network is only one slice from an
infinite number of activities enabled by digital technology.

89
Research has indicated, for example, that values tend to changes as societies become richer; for example, richer
countries in the OECD region may ascribe greater value to safety than do developing countries.
90
Increasing returns to scale may also give rise to explosive solutions that are awkward in mathematical models.

78
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Box 2.1. How digitalization can affect the economy

Output (GDP)
Lower transaction costs
Network effects and increasing returns to scale for digital goods and services
Improved matching: digital platforms also enable minor transactions in goods and services
Speed of technical diffusion increasing in some dimensions
Non-exclusive nature of digital goods
Lower cost of tweaking production/innovation.

Capital
Many free digital tools available
Cutting out the middleman reduces cost
Cloud computing, vast processing power available
Easier available capital with peer-to-peer (P2P) and peer-to-business (P2B) financing
New payment systems; bitcoin, apple pay, swish, etc.
More efficient use of existing capital stock (cars, apartments, tools etc.), sharing economy
More seamless information

Pricing power
Easier price comparisons and transparency
Younger generations more prone to price comparisons online
Easier entry
Increased global competition in value chains
More information about consumers and targeted advertising, price discrimination
Brand names established faster but also higher risk of negatives

Wages
Online labor markets services more global
On demand economy and self-employment increase competition
More self-employment may change wage-bargaining processes in the economy

This feature of digital technology is probably the best understood aspect and the one
that has the most dramatic impact, notably in terms of accelerating how the economy
works. Whatever service can be delivered to 10 people may in theory be as easily
delivered to 10 million or 10 billion. In practice, there are regulatory obstacles, issues
of taxation and language barriers that prevent the scaling from being entirely seam-
less from local, to city, to country or global, discussed further in Chapter 4.

The importance of network effects hinges on how prevalent they are in the economy
and the magnitude of the effects. There is a small but growing empirical literature
that investigates specific markets; for an overview see Belleflamme and Peitz (2015,
pp. 581583). For example, for spreadsheet packages, consumers were willing to
pay almost 50 percent more for software compatible with a well-known standard;
for direct network effects, it has been found that without network effects, German
mobile penetration might be 50 percent lower with observed prices.

Better knowledge about the size and scope of network effects in the economy would
be a useful research agenda. How much of network effects from digital firms spill over
to industry and what is the rate of change? This information could be used to better
understand potential GDP in the economy. For example, if network effects are increas-
ingly prevalent, potential GDP might increase compared to a less digital economy.

79
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.3.2 Speed of technological diffusion and the digital shift

Innovations used to take a long time from conception to commercial use. Steven
Johnson in How We Got to Now: Six Innovations That Made the Modern World
describes the path of progress for technologies that irrevocably changed welfare across
the world; see Johnson (2014). For example, Angel Barovier discovered how to make
crystal on Murano Island, Venice around 1450, but the understanding of why it worked
took another half millennium and led to the discovery of the mirror and later to fiber
optic cables used for fast internet transmissions. Likewise, William Carrier invented air
conditioning in 1902, but its commercial use took another half century to take hold.

In the last three centuries, the speed at which innovations spread throughout the world
economy has vastly increased; see for example Comin and Mestieri Ferrer (2013).
As shown in Figure 2.1, the average number of years for technological diffusion has
more or less decreased continuously, with the introduction of spindles taking on
average 119 years to spread compared to seven years for the internet.

With the advent of smart phones and the web, the speed of diffusion is likely increasing
yet again. It is now much easier to seamlessly reach many users, for example, through
the web, via the Apple app store or Googles android. Exactly how much faster digital
technologies are spreading is not easy to determine. Reports have been circulating
comparing the number of days that the game Angry Birds needed to reach 50 million
users (35 days) compared with television, for example (75 years), but the veracity
and origin of such numbers have been questioned; see Aeppel (2015) and Hannemyr
(2003). The comparisons are not straightforward, because while games may spread
like wildfire, the time these games are played may have a half-life considerably shorter
than other technologies in the 20th century.

What is undoubtedly true is that technological dispersion has increased manyfold over
the centuries. The speed and scaling of digital services means that some companies
that become successful can grab an enormous share of the market, including the recent
notable examples of Amazon, Facebook, and Google, as illustrated in Figure 2.2.

Figure 2.1. Technology is spreading faster, mean time of adoption in years

Spindles

140
Steam

Railway passengers

120
Railway freight

100
Electric arc furnace

Source: Comin and Mestieri Ferrer (2013, page 14).

Aviation passengers

80
Aviation freight
Tractor
Steel
Telephone

Synthetic Fiber
Electricity
Telegraph

Fertilizer

Blast oxygen furnace

Harvester

60
Liver transplant
Kidney transplant
Mail

Trucks
Cars

Heart surgery
Cellphones

40
Internet
PC:s

20

0
1779
1788
1825
1825
1835
1840
1855
1876
1882
1885
1885
1892
1903
1903
1907
1910
1912
1924
1950
1954
1963
1968
1973
1973
1983

Note: The time to adoption is the mean number of years for a large sample of countries; for details see the explanation
in the paper.

80
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 2.2. Years needed to reach 50 million users

Sources: Telephone, radio, internet, Facebook Citi strategy; for others, see links in note.87
Ipad

Tinder

Facebook

Internet

Dropbox

linkedin

TV

PC

Netflix

Radio

Telephone

91
0 10 20 30 40 50 60 70 80

The network effect and the possibility of fast scaling are the fundamental aspects
underlying digital effects on the economy. Almost everything stems from these features.
The effects of scaling are evident from the number of companies that go from startup
to global in a few short years. Indeed, it has given name to a new term unicorns
coined by the venture capitalist Aileen Lee; see Manjoo(2015a). The appellation
denotes a company that reaches $1 billion in sales in just a few years. There are five
companies with Swedish origins among the unicorns, notably King, Klarna, Mojang,
Skype, and Spotify. While some digital companies clearly have reached extraordinary
market value in very short periods of time, there is concern that some of this may be
exaggerated due to accounting treatments and the terms of private investments set
before initial public offerings; see Smith (2015).

What do the network effect and increasing returns imply for how the economy is
working? For all areas where services are delivered digitally, the benefits are huge.
Though digital services are growing rapidly, they are still a small part of the economy
and much of the benefits come from finding and addressing inefficiencies in existing
business models. Most of the economy is using the computer in one way or another,
but so far only a fraction can claim to enjoy the pure network effects discussed above.
This may change over time, but is still likely to be a gradual process.

For example, hairdressers use websites to allow customers to make electronic appoint-
ments and perhaps include texting to send reminders, thus making the business cheaper
and less time consuming to interact with customers before the service is provided. But
the actual haircut does not enjoy low marginal cost, increasing returns, or network
effects. Similarly, the physical production of cars or appliances can benefit from digital
designs and experimentation; spare parts can be produced with 3D printers and cars
can be ordered online. But most of the production is still subject to physical limitations
and hence deceasing returns to scale; it is really only largely digital services, such as
games, newspapers, books, music, and video streaming that enjoy full network effects
and increasing returns to scale.

91
iPad www.quora.com/How-long-time-did-it-take-Apple-to-reach-50-million-iPad-users; Tinder http://mobile.
nytimes.com/2014/10/30/fashion/tinder-the-fast-growing-dating-app-taps-an-age-old-truth.html?referrer=&_r=1; Dropbox
http://www.fastcompany.com/3029699/bottom-line/lessons-in-hyper-growth-from-the-man-who-scaled-engineering-at-
dropbox-and-faceb; linkedin https://ourstory.linkedin.com/#year-2003; netflix http://money.cnn.com/2014/07/21/tech-
nology/netflix-subscribers/.

81
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.3.3 Lower marginal cost, non-exclusiveness

If we consider the parts of the economy that are mostly non-digital, such as manu-
facturing, they typically include machinery, inventory, sales, etc. All features of produc-
tion are benefiting to some extent from digital technology, which has some dampening
effect on the marginal cost of production. The larger the digital content, the more
significant the effect. Newspapers and journalism are especially illustrative examples.
Delivery of newspapers via digital services (pressreader, etc.) is seamless and enjoys
near-zero marginal cost but the production of news does not. Having journalists in
different countries to report events is expensive, especially in war or conflict zones,
when risk premiums have to be paid. But the larger the distribution of newspapers, the
lower the marginal cost: the fixed component of the firms costs becomes a smaller
share of overall costs if distribution is digital. The same argument applies to many
other activities that involve physical production but digital delivery, such as movies
and music. Notably, for digital books most of the costs of production remain the same
for publishers: the physical printing of the books is a fairly small share of total costs.

All things that can be delivered by digital means share another important feature that
distinguishes them from the rest of the economy: the non-exclusive nature of consump-
tion. When you buy a product in a supermarket or a book in a store, no one else can
use it at the same time. With digital delivery, the consumption of one person doesnot
restrict that of another. This is essentially the other side of the network effect discussed
above. It has especially important implications for the distribution of knowledge and
it is instructive to discuss why. Before written language, knowledge had to be passed
down orally, which was slow and prone to error/misinterpretation. With printed
media, knowledge is passed much more easily from generation to generation, but only
with digital media and the internet has virtually all information become available to
everyone. Libraries with classical works can burn down and unique works destroyed
but if the works are stored digitally, there may be backups available. But while digital
knowledge does not burn down, there are other issues instead.

When all information is available digitally, the amount of noise also increases and
without strategies for dealing with this, some efficiency gains may be held back. More-
over, for information and knowledge to become seamless, there is also an issue of
backward compatibility. As standards are changing, retaining the ability to read old
documents and pictures may be low down on the priority list of firms that are looking
to be profitable. Indeed, for many people, accessing their own personal files, pictures,
and videos from many years ago might not be seamless with the latest software.

There are also physical limitations to digital storage of media, as deterioration of hard
drives, etc. may occur over time. With the upgrade logic of changing phones and com-
puters every few years, the degradation of hardware may not be a big issue but for
the stuff on our desks or in our pockets, for tape and hard drives, degradation of data
may be an issue perhaps not the same order of magnitude as books being destroyed
by wars, fire, or simply old age. Admittedly, digital storage in the cloud may be more
secure than previous storage methods, but while books can be destroyed they cannot
be corrupted the same way data can. Overall, backward compatibility remains an issue
with unclear ramifications.

2.3.4 Information becomes more seamless in capital markets

The standard model assumes that all relevant information is reflected in the price of
assets and that new information instantaneously translates into price changes when
conditions change. Much research in the field of behavioral finance, see for example

82
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

De Grauwe and Grimaldi(2006), is devoted to relaxing the assumption that infor-

mation is infinitely available and that individuals are able to process it seamlessly to
compute complicated optimization problems and reallocate portfolios.

While digital information processing has been available for decades, automatedhigh-
speed trading has become possible only with the advent of fast computers, digital
infrastructure, and high-speed internet. With sophisticated algorithms, high frequency
trading finds arbitrage opportunities in the smallest price movements. Indeed, it may
no longer be possible for traders to find arbitrage opportunities without using high-
speed computers. In this way, information processing and trading has gradually moved
closer to the workhorse models of finance. To the extent that deviations from the
workhorse model also implied imperfections and inefficiencies, digitalization has
improved the price mechanism of financial instruments and hence, at least from this
perspective, contributed to making financial markets operate more efficiently.92

The efficiency gains may not be easy to measure, but digitalization has arguably brought
a different type of vulnerability to the financial system. First, the financial crisis of
2007-09 showed how the vast complex interdependencies made the financial system
vulnerable. Second, technical mishaps can cause serious problems. Third, the systems
may also be sensitive to manipulation. While insider trading and rumors have long
been part of illegal activity in financial markets, the manipulation of trading aimed
at nudging automated trading algorithms in one direction or another is a new form
of vulnerability, exemplified by the flash crash of May 2010. It is contended that a
single day-trader based in London on his own wreaked havoc on the Chicago exchange
by first placing a large batch of orders to nudge algorithms in one direction only to
cancel those orders in the next instant.

2.3.5 Capital investment less required to start digital companies

Some of the first benefits of the Industrial Revolution lay in replacing manual labor
by machines, known as capital deepening. But once the major shift from a labor-
intensive economy is made, productivity improvements from capital deepening become
lower; that is, the marginal return from an additional unit of capital decreases over
time. For OECD countries, most of the growth differentials in recent decades are
explained by how well the productivity factors interact together in total or multi-
factor productivity.

Does digitalization change the economic logic of the current shares of labor and
capital? For some parts of the economy, notably manufacturing, digitalization has
involved a gradual shift over long periods of time. Fewer people have been needed
to perform the same functions. But for those sectors involved in digital production
theshifts are more dramatic.

First, by using existing ICT platforms, digital firms can implement new business
ideas and spread them very widely at low cost. For instance, Apples app store is an
example of an ecosystem where, with a modest amount of programming, it is possible
to reach almost the whole world.

Second, the amount of capital needed to start a digital firm is often much less than
that of physical production. A moderately fast computer and an internet connection
may be all that is needed. Platforms, software, storage, and computing power are free

92
Such efficiency gains may well be dwarfed by other problems with how financial markets operate, an issue outside the
scope of this report.

83
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

or can be rented at low cost. Powerful and sophisticated tools are available through
cloud computing. One effect of this is that even small startup firms can swiftly
become large based on very little capital.

2.3.6 Quicker production, tweaking, innovation, and cutting out the middleman
Digital technologies make it possible to speed up production and improve quality.
The initial stages of the Industrial Revolution entailed building specialized machines
that performed one task and one task only. New products typically involved costly
changes to existing machines or, indeed, the need to build new machines. As a result,
product cycles tended to last longer than they do now.

Firms that have not updated their IT systems for a long time may find themselves sud-
denly facing intense competition from those that have modern platforms and are able
to more clearly discern customer demand and adjust production accordingly. Banks are
especially vulnerable in this regard with old IT platforms stemming from acombina-
tion of conservative upgrade cycles, legacy systems, and mergers. One consequence for
banks is that some of them may not have a complete picture of their client base in con-
trast to competition from new firms that use digital platforms, which is further dis-
cussed below in Box 2.3. What is fintech? The same may hold true for shops and
chain stores that use outdated IT systems, with vast potential for improvement.

With advances in robotics, see for example Bryant (2014), Harding (2015), and
Markoff(2012, 2015a, b), machines are becoming more versatile and reprogrammable,
even in areas where highly skilled labor interacts with physical development and so
has broader relevance for the manufacturing sector. We are not yet at the point where
machines can replicate the full range of human physical and cognitive skills, but in
many sub-areas, computers consistently outperform people (see Chapter3). Some of
the work in building electronic components that was previously outsourced to Asian
countries may even return to OECD countries, though probably in the form of robot
fleets rather than manual labor, see the Economist(2012b).

As robots become more programmable, the fixed costs become smaller and their uses
increase. Architects are working on ways to send their drawings directly to industrial
robots for production. Even sewing of garments may soon be performed by robots
and may in a few years lead to the reshoring of production from low-wage countries;
see for example the Economist (2015c). When design can be sent directly to industrial
robots, another intermediate function of production is removed, thus speeding up
production and making it economically feasible to have smaller production lines,
retooling more frequently, and making goods more personalized, similar to the
advantages of 3D printers.

3D printers are being used to build bridges and houses; see the Economist(2015i).
With complete design of a bridge in the computer, it is possible to reduce the amount
of material used. In the example of the bridge, this resulted in 75 percent reduction
in weight, which is a major cost saving only possible through software integration
ofdesign with the printer.

It is difficult to assess how far digitalization has cut out layers of production and stream-
lined processes. The changes are occurring in broad sectors of the economy, including
in areas hitherto largely unaffected. For example, restaurants are now able to use soft-
ware such as table 8, Zurvu and SeatMe to improve efficiency and reduce the number
of no-show reservations that can amount to 510 percent of all reservations; see
Buchanan(2014). But even if the magnitude is so far unclear, the direction is not. During
the Industrial Revolution, machines were replacing workers; today all sorts of functions,

84
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

but especially those that are primarily rule-based are being increasingly automated. In
the years to come. advanced services such as legal analysis, design, and writing may also
increasingly be subject to automation, a topic we will return to in Chapter 3.

Indeed, eliminating the intermediary has been a trend in many areas: word processing
reduced the need for secretaries; online booking, the need for travel agents; and online
banking the need for bank tellers. A recent example is InShared, a Dutch insurance
company, that has recently made the whole claims process digital, thus eliminating
several layers of human interaction.93 This overall trend has consequences for the
individuals who are the recipients of those services, notably in that households need
to do more of the work; see, for example, Lindbeck and Wikstrm(2003).

But reducing the need for the intermediary has other effects as well. A digital design
can be changed more easily than a physical one. With few strokes on the keyboard,
the initial design can be easily improved and production processes quickly updated.
This is good news for innovations. The possibility to experiment and explore becomes
cheaper and easier; with 3D printers, designs can be shown not just in concept but
also physically, which cuts costs in the sensitive start-up phase for companies.

We are not aware of any research that systematically analyzes how far intermediary
services have been squeezed. Thus, it is also hard to say how much of efficiency and
productive growth comes from this trend. But looking ahead, it stands to reason that
digitalization may bring further efficiency gains, especially in areas that have hitherto
been less affected, especially in services, including highly skilled services in the profes-
sions (more on this in Chapter 3).

2.3.7 Business model disruption, subscriptions becoming more common

Digitalization is changing the calculus for how firms can charge money, notably in
choosing between subscription-type services or payment per good or service. While
itis hard to generalize, subscription services are associated with loyalty and steady
revenues and are thus attractive for firms whenever possible. For example, news
papers and gyms traditionally offer both, but in rapid changing markets, businesses
that are firmly entrenched in either model may also have to rethink their strategy or
risk being left at the station as consumers jump onboard. Digital companies that ini-
tially provided a lot of content free of charge are tweaking their payment models in
avariety of ways. This will affect traditional businesses in a lot of different areas.

Digital media music, film, newspapers, books, and magazines are in this category. Sub-
scriptions for music via Spotify vie to attract customers with free offers combined with
advertising or ask customers to pay a monthly fee for virtually unlimited access. Other
streaming services provide free trial periods, similar to apps and software. Netflix used
to be a content provider only but has now shifted into content creation, thus becoming
a slightly different kind of business. Apple has until now been a platform to buy media
but is now going into competition with Spotify and offering streaming music.

Some businesses are facing strategic choices about where they want their revenues to
grow or how to find new ways to make money. Publishers that offer subscription ser-
vices for books, music, and the like may be squeezed between consumers and content
creators. Content creators want to be paid for each time their work is accessed, be
it a song or a book. Thus, the more times a customer accesses a particular work, the
less profit for the subscription service that charges a fixed fee.

93
The whole claims process is done online. The customers report the claim on a web page, where they are even able to
book contractors.

85
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

What this development will mean in the longer term is unclear. What is clear is that
the relationship between content provider and content creator is changing in complex
ways and so are their relative bargaining positions. Much of the bargaining strength
traditionally resides with the content providers but this is also changing, most prom-
inently recently when a tweet from Taylor Swift made Apple reverse its position that
artists would not be paid for songs played during the free three-month trial music
subscription.

Payments are also changing in other areas. For example, Dollar Shave Club ships razor
blades to customers every month, thus competing against Gillette; see Hill (2015). Car
rental firms are having to compete with providers such as Car2go that charge per use,
which in turn may be competing with leasing alternatives and ownership. Apple and
Facebook are beginning to provide newspaper content directly in their feed rather than
sending customers to news sites via links. Netflix used to distribute content but is now
successfully producing original material. Google used to be an internet-search com-
pany until it designed Android, its own operating system for smart phones and is now
exploring venturing into cars, smart homes, and other areas. Overall, digitalization has
blurred the line between content creation and distribution. Several companies are in the
throes of deciding exactly what kind of business they are doing.

For businesses that are locked into a particular way of interacting with customers, it
will be important to be able to change how it charges money in response to shifting
consumer preferences and behavior. For example, some newspapers have opted in
to the new Facebook and Apple services, while others remain out; see Garrahan and
Kuchler (2015). While the choice is complex, at least two factors stand against each
other: by providing content to digital platforms with many visitors, newspapers can
potentially broaden their market and reach more customers. But at the same time,
there is a risk that the brand name will be eroded and that loyal subscribers will leave
to instead enjoy lower cost access via the digital platforms. Overall, it seems the digital
disruption in media is only beginning and spreading to other areas as well.

2.3.8 Working hours divided into small pieces and more self-employment
Digitalization increases possibilities of moving employment contracts more towards
small pieces of time rather than the full time-part time-unemployed spectrum. For
example, someone might work part-time in an office, drive a car-share in the evenings,
and rent out an unused room, thus being employed and self-employed simultaneously
and to varying extents. When this choice is involuntary, however, it may have negative
consequences, some of which have already been observed in the US. For example,
some low-paid workers in fast food chains in the US have seen their working hours
become more unpredictable when their firms optimize schedules according to demand.
One effect of this is that it becomes harder for people to plan their lives for second
(or third) jobs, studies, child care, and free time; see Kantor (2014) and Singer(2014).
Recent strikes for better conditions in some fast food chains in the US could be a sign
of such frictions in conjunctions with issues related to low pay.

Wages in the US have not increased much for median workers in recent decades;
instead, many households may have perceived their wealth going up due to increasing
house prices; see for example Rajan (2010). Many less skilled workers have also had
low wage growth, as the pool of available workers to fill positions is substantial; see,
for example, Leubsdorf and Hilsenrath(2015). Although the explanations may be
several, digitalization could be a contributing factor. For example, digital platforms
can provide a wider pool of applicants from which to fill positions and competition
for available jobs may increase. But it can also be a demand issue, with firms still

86
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

recovering from the financial crisis. For the Federal Reserve and other central banks,
assessing how long to keep interest rates low, the question of low wages stemming
from supply (digitalization) or demand is important.

Among OECD countries, multiple simultaneous forms of involuntary part time work
at minimum wage is much less common than in the US. Instead, more rigid labor mar-
kets in Europe often result in higher levels of unemployment. But the technical pos-
sibilities to have more flexible working schedules may also become more common in
other countries with more rigid labor markets. While this increases efficiency in the
labor market, it also raises issues about how the labor market operates with respect
to regulation, wage bargaining, pensions, and social security. The nature of work is
of course not going to change overnight, but the trends towards flexibility may be
achallenge for some countries that have strong institutional features for how work
isorganized, Sweden among them.

2.4 Demand side changes: prices and competition

Digitalization is increasing competition in some ways, but it has not necessarily led to
greater competition across the board in all areas. Economists have long noted that com-
petition is a complicated thing. Once departures from simple but unrealistic models of
perfect competition or monopoly are made, it is typically the case that predictions of
outcomes are much harder to make. Early literature in industrial organization empha-
sized the structure-conduct-performance hypothesis in which the characteristics of the
market, such as physical aspects, geography, etc. set the stage for the incentives of firms
and how they behave. These factors were deemed to determine how much competi-
tion there would be in the market. Later research has used insights from game theory to
model and understand the incentives involved. Typically, outcomes of competitive pro-
cesses can depend on a multitude of factors, which are difficult to disentangle in practice.

Recent research has investigated the factors that can lead to more realistic outcomes
than those simple models. Some examples of these more realistic factors are conditions
of entry into the market, regulation on collusion in the form of cartels, or predatory
pricing.94 There are also various ways to affect competition. For example, a firm can
potentially change the competitive landscape through advertising. As well, firms can
make substantial investments in R&D to deter potential competitors from entering
the market. Even relatively homogenous goods, such as mineral water, can reap close
to monopoly prices under some circumstances.

It is fair to say that all the elements that characterize competition in the non-digital
economy are also present in the digital economy. Indeed, in the digital economy some
elements are extreme, notably in the time scale of events, which is often extremely
compressed. Even within a short span of years, a company can go from garage to
global and become an enterprise with near-monopoly power.

2.4.1 Increasing price transparency and changing consumer behavior

There are many parallels between price competition arising from digitalization and
the forces of globalization. Clearly, competition in the digital world relies much on
the ability to reach out across the world to new markets and consumers. In this sense,
digitalization allows more globalization. However, more digitalization does not imply
that each and every market will become global. Indeed, some markets for services

94
Predatory pricing is when an incumbent intentionally sets a price below marginal cost to pressure competitors to give
in or to scare them off from entering the market.

87
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

remain stubbornly local. In these markets, new forms of competition may develop, a
topic we will return to further below. Therefore, even though closely related, digitali-
zation and globalization need to be discussed separately.

One especially important aspect of digitalization is the potential to change consumer

behavior in a way that increases competition. Many of the now common elements
of consumer power were anticipated some time ago; see for example Lindbeck and
Wikstrm(2000). Better informed consumers can put more pressure on retailers and
increase competition, which in turn gives retailers stronger incentives for product
differentiation and brand-building.

In practice, digitalization has changed the way in which consumer search for and
compare goods, which is increasingly done outside the bricks-and-mortar store.But
itcan also occur on the spot with consumers entering a store and using their mobile
phones to search for a lower price online, increasing consumer power. A report on
Swedish data shows that almost half of all consumers in 2014 reported that they search
the internet in a process called webrooming before making a purchase even when
the purchase is not made online, see Svensk Handel and HUI Research (2014). One
out of four Swedish consumers have compared prices online and one in ten have used
their mobile phone to check stock availability. In the US, nearly 70 percent of consumers
search and compare goods online according to Deloitte (2014a) and close to a half
of all consumers plan to shop online; see Luna (2014). Although online shopping is
flourishing, it appears that digitalization has expanded the search and compare phase
of a buy, but that the actual purchase of the goods happen in the bricks-and-mortar
store. Not only do consumers have access to unfathomable amounts of data to make
comparisons, they are also using it more and more.

In the UK, a study of 5,900 retail bankruptcies showed that stores in the central busi-
ness district (high street shops) have higher resilience and capacity for reinvention
than shopping centers and retail parks; see Deloitte (2014b). This shows that price is
not the only factor affecting consumer behavior. Other factors such as the convenience
of nearby shopping still matter too.

The resilience of high street shops in the UK notwithstanding, if online internet sales
continue to increase, retailers in central business districts will face even tougher times.
Younger generations that are already now more exposed to digitalization will likely be
more at ease with shopping online compared to older generations. We can only specu-
late as to the outcome of such a development, but it is likely that many retailers will
manage to survive this trend only if they will maintain the ability to reinvent them-
selves and adapt to changing consumer behavior. For example, stores may try to sell
services connected to their goods instead of only the product itself: you might not
actually buy a running shoe, you pay for a running shoe experience with various
forms of individualization, such as trying out the shoes on a treadmill. You might
not buy an electronic tool, but instead use a sharing service and purchase the nails
and things you need, such as with toolpool. The sharing economy, discussed further
below, may also change consumer behavior. Does this mean that price competition is
increasing and that price levels are always and everywhere under serious assault?

2.4.2 Price discrimination on the rise through sophisticated algorithms

Somewhat paradoxically, the scope for price discrimination is also increasing despite
greater transparency. Just as there can be a mismatch in the labor market with a simul-
taneous lack of skills and surfeit of education, so too can prices be both transparent
and discriminatory at the same time, perhaps a special feature of the digital economy.

88
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

There is increasing evidence of sophisticated algorithms that use very large datasets
to make inferences. It has been reported that Facebook likes are a better guide for
predicting a persons behavior than asking their spouse; see Ahmed (2015a). Tracking
likes also makes it possible to gather a myriad of information about users that can
then be used, for example, to create tailored advertisement. There is indeed a large
increase in third-party services that collate and analyze consumer data; see the Econ-
omist (2014a). Algorithms are increasingly geared towards improving sales and busi-
ness by assessing online behavior. For example, a consumer may be offered a discount
when it appears she is likely to leave the website without closing the deal; see Tanner
(2014). Notably, the algorithms can infer a wealth of personal information from var-
ious sources. A change in the price an online consumer can receive may depend on
many factors. Exactly how these algorithms work is clouded in some mystery and is
aclosely guarded secret of the trade.

The fear of a consumer backlash against amassing and exploiting personal data may dis-
courage some firms from using the full potential of the data for improving their business.
Tony Weisman of DigitasLBi a digital data provider was quoted in the Economist
(2014a) as saying, We can do more technologically than were permitted culturally.

It is well-known that Amazon may change the price of the same product multiple
times during the day.95 The price change may be driven by what the competitors
are doing but also by the type of consumers who are searching the Amazon site; see
Peterson(2014). Granarolo, an Italian dairy products company, uses sophisticated
algorithms based on machine learning to predict demand with almost complete accu-
racy; see Dempsey(2014). This allows their executives to approve recommendations
on supply of dairy products in close to real time.

Although sophisticated algorithms that use personal data are currently employed,
a wider and broader use may have unpredictable consequences. In some countries,
some forms of algorithm-based price discrimination may already be illegal, but the
public perception about the use of personal data for price discrimination may be an
even more significant obstacle. Consumers have long become accustomed to coupons
and various forms of price rebates on goods. But if rebates come at the cost of allowing
personal data to be used, consumers may react negatively.

Other forms of sophisticated price discrimination have also been observed. For
example, someone who uses a price comparison engine to find the lowest price may
get a better deal at the same site than a customer who went directly to the site.

While the advent of a consumer backlash is possible, we are likely to see rising use
of sophisticated algorithms to assess and influence consumers. When competition is
strong, firms need to find ways to reach consumers and differentiate their products
from those of their competitors. This brings strong impetus to expand and deepen the
use of personal data. As more personal data becomes available and the algorithms
become even more sophisticated, the general price level may be affected. Digitalization
may be one factor among others affecting inflation, as discussed further below. It is
still too early to determine whether the ease of price comparisons will outweigh the
ability of firms to price discriminate. There are also questions about what the regula-
tory authorities may do, especially since some prices may be different depending on
country of residence or the location of the server.

95
See for example http://www.businessinsider.com/amazon-price-tracking-2014-8?IR=T

89
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.4.3 Physical goods becoming digital services

Books, movies, and music have already been transformed from physical goods to digital
services. For example, e-books are providing the same material as a service compared
to buying a book. In many instances, the digital service is cheaper and has more content.
For example, some e-books contain video and audio recordings; digital magazines like-
wise blend still pictures and video. 3D printers can also blur the difference between
goods and services, a topic we discuss in the context of dentistry in Chapter4.

But goods are becoming services in many other areas as well. Notably, vehicles are no
longer just trucks and cars, they are essentially big computers on wheels. Commercial
trucks have long been fitted with black boxes similar to airplanes. The data generated
from the trucks can be used to control fleets, their location, speed, and so on. Scania
and Volvo have been supplying such services for a long time.96 The data can also be used
for other purposes, such as giving feedback to drivers on fuel efficiency and predictive
maintenance, which can provide advance notice if the truck needs new spare parts
or other service. Electrolux is exploring integrating cameras into its ovens, which can
then be connected to a smart phone for access.97

2.4.4 First mover advantage not always present when winner takes all
It is often thought that being the first with a new product brings an advantage, but
there are plenty of contrarian examples. Ericsson was close to launching a device
not unlike the iPad many years before Apple introduced it. The website Boo.com
selling designer clothes was launched with much hullabaloo around 2000 but sub
sequently went bankrupt. Today, being able to quickly increase the scale of a business
has become key. As expressed by Reid Hoffman, co-founder of LinkedIn: First-scaler
advantage beats first-mover advantage; see Hoffman (2015). In contrast to the situa-
tion about a decade ago, a number of key conditions have evolved and matured:
The emergency of large platforms with standards that attract consumers and
producers alike (the web, Apple with iOS and Google with Android).
Trust mechanisms for digital transactions, both for goods and services.
Digital payment systems with low transaction costs.
Ubiquitous use of smart phones and tablets.

Just as in the analog world, being first is not always the best. If the product at launch
is not of good quality or the market infrastructure is insufficiently developed, the
second, third, or tenth entrant may be the one to capture the whole market. Again, a
key difference with the analog world is that this process can be extremely fast, so that
an obscure company can become very large, very quickly. Some videogames, such as
Angry Birds and Minecraft, are examples. In contrast to the norm for firms that pro-
duce physical goods, digital firms can become tremendously successful in a very short
period of time and not seldom having only one product. To survive after an initial
success, they face a tremendous challenge in quickly building a new profitable line,
not least since consumer loyalty to any specific digital product can be ephemeral.

96
See, for example, http://www.scania.com/products-services/services/fleet_management/.
97
See http://www.electroluxgroup.com/en/aeg-launches-the-worlds-first-connected-steam-oven-with-integrated-
camera-21191/.

90
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.4.5 Brand names matter as much as before but speed is (again) higher
One feature that decidedly remains the same in the digital as in the physical goods
world is the value of brand names. Coca-Cola, H&M, Volvo and all other multi
national companies spend a great deal of time, money, and effort building and
developing their brands.

For digital firms, the brand matters just as much if not more. There is little time to
build a brand; compared to existing companies that have nurtured their brands over
decades, digital firms typically need a global view of how their brand will be perceived
from the outset. Building brand names also faces new challenges in the digital age if
exposure primarily comes from apps that aggregate services or provide concierge-style
services that expose consumers less to brands; see for example Waters (2015a).

Speed in building brand recognition and consumer loyalty is essential for the survival
of digital firms especially because services in some areas may be very similar and the
need to stand out from the competition may therefore be even stronger.

Moreover, the reputation effect on a digital firms brand firm can be very damaging,
very quickly. A wrong tweet, an offensive Facebook update, a misplaced email, leaked
information or passwords can quickly blow up into a maelstrom of negative pub-
licity. When Sonys email system was hacked in 2014 and confidential emails were
published online, the repercussions were wide and long-lasting. It even affected the
launch of a comedy movie about North Korea and had implications for foreign
policy, with even President Obama weighing in on how the debacle was handled.

In Sweden, a survey of 10,000 respondents conducted by researchers at the Stockholm

School of Economics and Askus Consulting found that consumers were particularly
unforgiving about the misuse of private information; see Dellham et al. (2013). But
some analysts and executives interviewed in Twentyman (2015a) argue otherwise,
identifying the possibility of breach fatigue as an explanation. The two views seem
irreconcilable at first, but both can be right if attitudes have evolved over time and
become more forgiving as breaches are more commonly reported. However, a survey
of US attitudes shows that people trust neither the government nor private corpora-
tions to handle their personal data; see Goel(2015). Another recent survey in the US
showed that 91 percent of consumers disagree that being given a discount is a fair
exchange for firms to collect information about them without their knowledge; at the
same time, about 65percent of respondents asked appear resigned to this develop-
ment, see Singer (2015).98

Whatever the overall view, some areas are more likely to be simply more sensitive in
terms of how data are handled. Notably for e-commerce and online financial firms,
electronic fraud and impersonation for economic gain may be especially damaging
and expose them to the risk of enduring consumer wrath.

Trust is the sine qua non of financial transactions. In this regard, some digital com-
panies may share elements with the financial sector. Once trust in the solvency of a
financial institution is in doubt, investors dash to withdraw their cash to save what
they can. Even the unfounded perception of solvency problems can be enough to
trigger a bank run, as evident during the financial crisis, for example with Northern
Rock in the UK or Bear Stearns in the US. Thus, digital firms will likely need to be

98
While surveys are indicative of trends and attitudes, they should be interpreted with care. A hypothetical tradeoff
between privacy and an unknown discount may be less revealing than a real situation with a specified discount. It is likely
that personal data has some price at which consumers are willing to cede it; the problem is that the transaction is opaque.

91
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

as careful as financial institutions with their brand names and reputations. Just as
abank run can be the end of a financial institution, a brand run can be the death of
adigital firm.

2.4.6 Unassailable monopoly one day, gone the next

The nature of competition has changed in the last few decades. As technology has
evolved, the scope and nature of markets have also changed. In the 1980s, IBM
dominated both the market for mainframes and personal computers. As competi-
tion increased, IBMs struggles to manage and innovate led to the outsourcing of the
operating system to what was then a small, unknown company, Microsoft. Without
this decision, Microsoft would not have become a dominant force in the last decades.
Indeed, it started the era of Wintel the dominance of Microsoft through its Windows
operating system and Intel through its virtual monopoly on microprocessors. During
the 1990, Microsoft was accused of anti-competitive behavior in the US on several
occasions, as well as by the European Commission.

The charges were complicated and involved several elements. On one count, Microsoft
was accused of bundling, a practice in which it used its near monopoly on operating
systems to force out competing web browsers. For example, Netscape had an early
alternative to Internet Explorer, the Microsoft browser. Forcing consumers to accept
IE with Windows was prosecuted as anti-competitive behavior by the European
Commission.

But in the end, it is not so much the competition authorities as changing technology
and markets that began to rattle Microsoft. Apples introduction of the iPhone, and
later the iPad, upset a business model based on personal computer dominance. Today,
Microsoft is still a huge company with the Windows and Office suites ubiquitous but
no longer unassailable. The future for the behemoths of recent decades looks much
less certain. The competition for business on the internet left Microsoft wrong-footed
and the company is trying to regain its prominence.

The battle today between Apple and Google illustrates the nature of changing com-
petition in digital firms. Only a few years ago, Google and Apple were competing in
different markets. Google was primarily a web search company with revenues from
advertising while Apple was a computer company entering the mobile phone market.
Eric Schmidt, then CEO of Google, was even on the board of Apple as an outside
director up until 2009. But Googles venture into mobile phones through the Android
operating system made them direct competitors. Indeed, sharing board seats today
would be inconceivable. This illustrates how the nature of competition can change as
markets evolve, a factor that is especially pertinent with digital firms. Even when the
product may evolve only gradually akin to non-digital products, the market in which
digital firms compete can change so dramatically that even a near monopoly can
wither in short order.

2.4.7 Has overall competition increased?

A survey of Swedish e-commerce by Svensk Digital Handel and HUI (2014) indicates
that almost half of firms believe competition is increasing; see Figure 2.3. Swedish
consumers bought goods from foreign e-sites worth about SEK 11.5 billion in 2014,
still a small percentage of overall sales.

92
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 2.3. Has competition changed? Response from firms in Sweden, 2014

Source: HUI (2014, page 24). Lower competition has zero response and is not displayed.
0,35
0,31
0,3
0,27

0,25 0,23

0,2 0,19

0,15

0,1

0,05

0
Yes, competition has Yes, competition has Competition is unchanged Unkown
increased substantially increased somewhat

But is overall competition increasing from digitalization? More price transparency

increases competition but the effects of brand names and the winner-takes-all
phenomenon may be forces in the other direction. Ultimately, this is an empirical
question and a subject for further research.

2.5 Changing how the economy works the sharing economy

In the previous sections, we have discussed how digitalization alters aspects of produc-
tion, labor, price setting, and consumer behavior. Many times the changes accelerate
processes from design to production and delivery. The changes are so significant
that they have an effect in their own right but many are matters of degree rather than
fundamental shifts. For example, replacing manual labor with machines has been going
on for a long time; the question now is rather different in nature: is the pace faster
and the scope more encompassing than before? But other aspects of digitalization are
making even bigger inroads on humans and the economy, even before considering the
possibility of artificial intelligence.

2.5.1 The sharing economy

Throughout history, sharing in various forms has been central to progress. Sharing
of tools and services prior to fiat money was necessary, as the logic of specialization
implies advantages to productivity that have to be balanced against the risks of relying
on others for factors of production or for sustenance. Civic society and other trends
have been vibrant throughout modern history. Indeed, it is hard to think of the market
economy without also viewing the strong bonds and commonality that constitute the
nation-state. In Levay (2013), the role and challenges of civic society are discussed in
Sweden. Bergh and Bjrnskov (2011) and others have shown how trust is important
in society for supporting the modern welfare state.

But just as civic society is integral to the modern welfare state, sharing as a phenomenon
is overshadowed by ownership in modern capitalist societies. To be sure, ownership
and rental have always coexisted, but the digital economy is making a fundamental
shift away from the logic and indeed the benefits of ownership. This shift is per-
haps the most dramatic change brought about by digitalization and has arguably just
begun. While industrialization replaced manual labor with machines, this new wave
may be redrawing the landscape of how capitalist economies organize themselves.

93
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.5.2 What is the sharing economy?

The sharing economy has been given many names, such as collaborative consumption,
the mesh, the access economy, freelancing, on-demand work, and the gig economy.
We will use the sharing economy, but it may be appropriate to first define what
we mean. Throughout the economy, there are idle resources anything from spare
capacity in the form of a room to a seat in a car or a vacant parking place. Idle resources
are bad in the economy because by and large they imply waste. Unemployment, due
to its social costs and the risk of erosion of human capital, is a particularly onerous
form of idle resources.

The sharing economy provides a way for these idle resources to find their use; see for
example Fellnder et al. (2015) and the Economist (2015a,b). The key feature is that
digital platforms allow buyers and sellers to be matched in ways that were not possible
before. Essentially, the digital platforms reduce the transaction costs or the frictions
involved to the extent that it may be beneficial to sell even miniscule idle resources.
For example, finding a ride-sharing service to commute from the suburbs to the city
center used to be possible to organize only within a relatively small circle of neighbors
or colleagues. But with digital platforms, it is much easier today to find someone selling
the service at least in densely populated areas. In few large metropolises, carpooling
is even supported by the local authorities with special carpool lanes, similar to bus lanes.

The sharing economy is characterized by:

An idle resource for sale either a physical good, knowledge or a service.
Someone willing to buy said service or resource.
A digital platform that matches buyer and seller with a very low transaction charge.

In Box 2.2 we summarize some of the sharing economy services; there are a huge
number already and they are expanding fast.

While the sharing economy certainly connotes ideas of social benevolence or

improving the environment, most times the driving force is the classical capitalist
profit motive. The incentive to become the next Facebook, Airbnb or Uber is very
strong. Striking the next mother lode can imply immense riches for the owners.

It is useful to divide the sharing economy into three categories: purely digital services
that enable sharing of knowledge, those that pertain to unused physical resources,
and those that involve providing labor. Sometimes the sharing activities involve all of
these to varying degrees but they are still useful ways to discuss the issues involved.
Some of the sharing is done peer-to-peer (P2P), some peer-to-business (P2B), and
some business-to-business (B2B) each with distinct features.

94
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Box 2.2 A few firms in the sharing economy

Fintech, peer-to-peer lending

LendingClub, Toborrow*, Lendify, Prosper, Funding Circle, Zopa, Payoff, Estimize, Kabbage,
Karrot, Vouch, Lendico, Zencap

Cars, transportation, and parking

Uber*, Blablacar, Tripda, Hailo, Car2go*, Stockholms bilpol*, Justpark, EasyCar, Zipcar, Drive-
way, Lyft, Sidecar, RelayRides

Lodging
Airbnb*, Housing Anywhere

Tools and miscellaneous items

Peerby, Rentmyitems.com, Ecomodo, Streetbank, Streetclub, Freecycle

Tasks
eLance, Upwork, Taskrabbit, Dog vacay, Handy, Washio, Bloom That, Fancy Hands, Spoon
Rocket, Fiverr, Hassle, Shyp, Tongal, Amazons Mechanical Turk

Legal services
Invest4justice, Crowdjustice, Upcounsel, Axiom

Business applications, professional and knowledge sharing

Simplist, Fon, Textio, Gild, LinkedIn, Wikipedia, Coursera, Studybuddy, NearDesk, BrandGathe-
ring, Nimber, Topcoder, Eden McCallum, Hoffice.

Marketplaces, fashion clothing, charity

Amazon, Etsy, Yelp, Tradera, E-bay, Blocket, Girl meets Dress, Rent the Runway, Freegle, Yer-
dle, Poshmark, Kidizen, Rocksbox, Just Giving

Medical
Crowdmed, Medicast, Heal, Pager, Retracehealth, MedZed, Dispatch Health

Hobbies, training, sports gear

Book Crossing, Vint*, Velib, Splinster

Food
FarmDrop, Food Assembly, Grub Club, Casserole Club, Instacart, Pallafrukt*

Note: For more explanations, go to the respective app or website. Two good sources on the sharing economy are
Wosskow (2014) and the Economist (2015b). Some sharing economy platforms deliver to other countries, some are
geo-based. A star denotes a service available in Sweden.

95
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.5.3 Knowledge sharing

Purely digital services have all the benefits of networks and scaling at low or zero
marginal cost. For example, IBMs Watson computer, famous for winning at Jeopardy
against the best humans a few years ago, has found uses in medical and legal services.
With a huge database, be it road-street mapping for navigation, medical research for
diagnosis, or legal case histories, knowledge sharing is making it much simpler and
cheaper to perform even advanced services. It is also these changes that have the most
potential to be disruptive for middle-class jobs, a topic further explored in Chapter 3.

The medical profession is a case in point; see for example Lohr (2015a,c). Each medical
discipline contains a huge amount of medical research. Doctors may struggle to actively
follow research even in their own area of expertise. Applying that knowledge is yet
another challenge, for example, with new statistical inferences on likely illnesses and
diagnoses. With huge medical databases available, doctors have information and tools
at their fingertips to interpret the information that would be beyond the capacity of
humans. Pertaining to the discussion in Chapter 1 of looking at computers as comple-
ments or substitutes, this is primarily an example where computers can serve as a tool
for greater efficiency and thus support the medical doctor in making better inferences
rather than replacing her. This is an example of the complementarity discussed in
Autor (2014).

But knowledge sharing in the legal sector may be an example of the opposite. IBM
Watson has also been used by some researchers to create a legal database that can
handle a great deal of paralegal work; see Jackson (2015). Here the possibility of
substitution appear greater, and we will return to these issues in Chapter 3.

2.5.4 Sharing of physical resources

Items that were hitherto too cumbersome to share or rent are now more readily
available. The first wave of sharing applications allowed people to sublet rooms with
ease. Car rental is also very big. There is now an explosion of firms that provide dif-
ferent types of homes all from luxury rentals for tourists to sharing of office space
through, for example, Second Home. Sharing office space can bring big advantages
to startups that do not want to commit to renting more space than necessary; there
are also benefits to be gained by providing meeting places for people with ideas. The
same driving forces that make cities attractive and drivers of economic growth, see
Shapiro (2006), are also relevant to sharing of office space. The notion of ideas
colliding is similar to the benefits of cities, albeit on a smaller scale. Here there are
existing parallels to the makerspace movement; see Sleigh et al. (2004).

However, small items, such as tools and parking spaces, can also be shared for shorter
periods of time. There is a wealth of new ideas and entrepreneurs discovering which
idle resources can be made available on the market. We are now only in the early
stages of this development and how far we can go will depend on regulation, trust,
and convenience but also on other factors that are not foreseeable now.

The sharing of tools is a particularly illustrative example, as it holds both the poten-
tial to disrupt as well as to increase the market. Homeowners and DIYers have long
acquired various tools that they use only rarely, such as drills, saws, etc. If these tools
are instead rented, the owners acquire income, the tools are used more, and rental
costs are lower than the cost of ownership. But what does this mean for retailers that
sell tools? Will they sell fewer if the sharing market thrives? And what does this mean
for their businesses? These are questions that remain to be answered.

96
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.5.5 Sharing of services

The bulk of service sharing has hitherto been focused on relatively unskilled jobs.
Carsharing has become a vibrant area for new apps that provide a variety of sharing
services all from an empty seat, in which case it is a service, to actually renting a car,
in which it becomes sharing of a physical resource.

Many services provided in the sharing economy are low skilled, but not all. There are
especially big changes in financial services or fintech, where small companies can
provide specialized services that are cheaper, faster, and easier to use than traditional
banking services.

P2B lending illustrates some of the issues involved; see the Economist(2015d). For
example, Funding Circle in the UK, Ondeck in the US and Toborrow in Sweden pro-
vide a platform for private individuals to loan their money to businesses. Typically,
P2B firms provide a platform for the financial service but essentially never hold on to
the cash. For lenders, this provides an opportunity to place their savings at a higher
interest rate than in the bank at (presumably) moderate risk; for businesses, it often
saves time and effort in trying to get bank loans. Moreover, the loans extended are
typically smaller than those banks are interested in since the banks have almost the
same marginal costs for doing due diligence and processing small loans as compared
to a big one. In this way, fintech firms are supplying credit to parts of the economy
that previously had no access to the capital they wanted; see Box 2.3. What is fintech?
Potentially, this could be a very significant boost to how capital is put to use in the
economy, enabling more small firms to get started or expand. The overall effect may
be that the economy works much better.

So far crowdfunding, P2P and P2B services may not be in direct competition with
banks when the focus is on smaller loans, but this may change, as expressed by Jamie
Dimon, CEO of JP Morgan Chase & Co, Silicon Valley is coming, see Dimon (2015).
The nature of investment in fintech may also be changing, with more venture capital
firms supplying the capital; see the Economist (2015d). In the UK, one P2P platform
gets clients from a conventional retail bank in exchange for free advertisement; see
Alloway and Dunkley(2015).

What this financial disruption will mean for the financial sector is yet to be seen.
Margins and profits have long been high in banking and the potential for efficiency
improvements is likely considerable; many banks have had their systems updated only
piecemeal or created a patchwork of more-or-less incompatible systems acquired
through mergers. Fees on credit and bank cards should make entry into payments
systems attractive with services such as Paypal, Apple Pay and Square. Digital curren-
cies allow money transfer at virtually no cost; see for example Mims (2015). Digital
receipts have yet to make a dent in physical receipts. Thus, a great deal may happen
with fintech in the years to come.

What we know much less about are the new types of risks that amass when individ-
uals and firms use these kinds of services. Traditional banking has a long history of
failures and bailouts, but also of being bailed out by taxpayers, most recently in the
aftermath of the financial crisis in 2007; the financial platforms are not covered by
deposit guarantees and other safety nets: any losses occurred are private.

This implies that risk is distributed in new ways in the economy, which may have
implications in the next downturn that we may not yet fully understand. For example,
could negative news lead to quicker dampening effects on consumption and invest-

97
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

ment than before and what happens to consumers when liquidity dries up in P2P
markets? How does the nature of systemic risk change and how well are the regula-
tors monitoring developments? How will consumers who have lost money invested
infintech companies react? These are urgent questions for macroprudential authori-
ties, such as the Financial Supervisory Authority in Sweden.

Size and scope of the sharing economy still small but not a marginal change
The forces and implications of the sharing economy are not a marginal shift of a few
percentage points more efficiency in production. Admittedly, the sharing economy is
still small. According to PwC (2014), the use of sharing economy services is still in
the single digits based on a sample of 1,000 interviews, see Figure 2.7. In the UK, a
Nesta study by Stokes et al. (2014) finds that 25 percent of the population has used
sharing economy services; see also Jacob (2015). Wosskow (2014) reports that about
20,000 property owners in the UK rent out their driveways for parking through Just-
Park. According to another estimate by Vision Critical and Crowd Companies based
on 90,000 customers, about 23 million people have used sharing services; see Stokes
et al. (2014). PwC estimates that the sharing economy is currently worth about $335
billion, which includes P2P finance, online staffing, P2P accommodation, carsharing,
and video streaming. Baeck, Collins and Zhang (2014) estimate that fintech, the part
of the sharing economy devoted to financial services, is worth 1.7 billion, barely a
drop in the UK financial sector ocean.

Since there are many platforms for sharing services, it is difficult to know how big the
overall phenomenon is becoming, as indicated by the various estimates cited. But we
are likely seeing only the beginning of the sharing economy. As current generations
become more used to such platforms, the platforms will become an increasingly
established part of the economy.

What is already known, however, is that some of the established sharing economy
companies are very big. According to PwC (2014), Airbnb averages 425,000 guests
per night or 155 million annually, which is 22 percent more than Hilton, which has
127 million annually. Uber, the carsharing service, operates in 250 cities worldwide and
was valued at $41.2 billion in February 2015 more than Delta Airlines, for example.

2.5.6 Nature of trust, limitations

There can be no sharing without trust. Indeed, almost all activities involving multiple
agents, individuals, or firms involve some degree of trust. In the worst of worlds,
sometimes trust is replaced by mutual need, but this is a fragile basis of relationships
and is wrought with dangers and inefficiencies.

Trust is key to the sharing economy and to all well functioning economies, but along-
side regulation it may be its biggest challenge. How do we trust the other party in
the transaction or deal? How do we know they will deliver the good or the service
we ordered? And how does the seller know the buyer will pay and is not otherwise
engaged in illicit activities?

Government regulation of course has a key role in setting standards of what is per-
mitted and what is not. When parties disagree, they try to reach an agreement in some
cases they may compromise, in others they part ways or resort to legal action. How
well the economy handles disputes not suppressing all conflicts but not necessarily
giving incentives to resort to costly court battles either is one factor that affects the

98
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

efficiency of the economy and ultimately productivity growth. It is perhaps no accident

that countries with weak institutions and lack of trust tend to grow less and become
less wealthy; see Acemoglu and Robinson(2013).

But is government regulation enough to establish trust? Clearly, government regulation

has evolved in tandem with peoples beliefs and the outcome of elections. Sometimes
accidents or events cause new legislation and sometimes there are drives to simplify.
But above all, government regulation tends to be behind the curve, that is, reacting to
events rather than smoothing the way forward.

In the sharing economy, this is especially true. Much government regulation originated
decades ago, before anyone knew the meaning of selfie or status update. That is not to
say that all regulation is outdated. There is sometimes a tendency to view all regulation
as an obstacle to growth. But the absence of regulation for firms trying to innovate
does not equal lack of friction, it is more likely to mean anarchy or unpredictability.
So, government regulation at its best balances various interests against each other
andour values in a way that does not put undue strain on progress.

Government regulation of the sharing economy a wobbling question mark

For the sharing economy, government regulation is if not clueless at least a slowly
wobbling question mark. Instead, trust in the sharing economy is being established
largely through self-regulation, mainly using various forms of home-brewed reputation
systems.

Economists have long viewed reputation effects as key to understanding outcomes.

In the traditional Nash bargaining problem in the prisoners dilemma, both prisoners
have an incentive to deviate and rat out the other, thus creating the worst possible
equilibrium in which they are both found guilty. If instead, choices were also affected
by reputation, the incentive to deviate might lose out to other future concerns, perhaps
the expectations of future income or friendships.

The sharing economy largely relies on self-evaluation to build reputation both the
sellers and the buyers evaluate each other with a simple click immediately following
atransaction. Presumably, people with poor evaluations have difficulties entering into
new transactions; few will trust them.

In theory, these kind of reputation-making technologies make eminent sense. Why not
amass in data what in the pre-industrial age used to be known in the village? Hopefully
the system is more unbiased and more forgiving than reputation building in the past.

The importance of self-regulation has been stressed by Cohen and Sundararajan (2015)
but there are some drawbacks. For one, each platform uses different technologiesfor
establishing reputation. There is as yet no common reputation history stored for each
person that can be used across platforms in the way that credit history checks are
made when people seek loans. This implies that whenever you want to switch to a
new service, you might have to rebuild a good reputation from scratch, thus poten-
tially strengthening lock-in into existing contracts. How much this matters is unclear,
but the absence of a common reputation log, perhaps a history trail of transactions
like bitcoin, implies new types of frictions that reduce incentives to change. While, as
discussed above, overall competition from digitalization is likely to increase, it may
not increase to its full potential, depending on how reputation effects evolve.

99
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Another issue is that a common reputation log or even a few central ones from
the services people mostly use may raise ethical questions and practical problems,
some of which we will briefly discuss in Chapter 6. The possibility of fraud with
digital signatures is a risk. When banks trade billions of dollars, B2B transactions
canbe very sizeable and the downsides significant.

It is fair to say that we are only in the early stages of understanding how the digital
economy and the sharing economy can accommodate and deal with the issue of repu-
tation to smooth transactions and maintain high levels of trust without resorting to
legal measures. We will return to this topic in Chapter 4.

2.5.7 Changing logic of ownership

While renting and owning are central parts of our economies, the logic of deciding
between them is changing in favor of the former. The advantages of owning something
are well known; the item is always there when you need it, no one else can tamper
with it. But there have always been some disadvantages to ownership as well:
Ownership ties up capital that might be more productive elsewhere
Depreciation and the need for some kind of maintenance to reduce wear and tear
Periods of idleness or underutilization

Digital technologies are making rental feasible for a great number of items that were
previously impractical. All the arguments in the previous section about the sharing
economy have a corresponding positive impetus for a shift in favor of more rental
and less ownership for individual use. All sorts of items are now available for rent
via digital technologies: cars (perhaps eventually the self-driving kind), clothes, tools,
etc. While these have always existed, the scale may become more significant when
themarginal cost of matching approaches zero.

Car ownership is an especially interesting area. Despite all the disadvantages of car
ownership in terms of cost and maintenance, it is very widespread; many people want
the freedom of having a car at their disposal despite ubiquitous traffic jams and the
difficulty of finding parking in big cities. Might the sharing economy shift the calculus?
Throw in a self-driving car that parks itself? It is hard to predict whether consumers
will react to this with minor adjustments or if we will see major changes in behavior.

There is an instructive parallel between owning/renting on the one hand and firms
choosing what to produce in-house on the other. The initial idea of Adam Smiths
invisible hand was that of a huge number of single-producer firms selling goods or
services; it was not until Ronald Coase (who later received the Nobel Prize in eco-
nomics) published his theory on the firm that economists had an explanation for
what should determine whether a firm buys a service on the market or performs it
in-house; essentially, the firm itself supersedes the price mechanism when items are
not traded and sold on an open market. When the marginal cost of in-house pro-
duction is lower than buying from the outside, it makes sense to build the function
in-house. For a discussion of Coase in the context of the sharing economy, see the
Economist (2015b).

What factors influence the marginal cost of buying a service or renting something
on the market? Or, in other words, what are the limits of the rental market? One
factor often presented is that of transaction costs, see Williamson(1979), who quotes
Stanley Fischer (then an academic, currently the vice-chair of the Federal Reserve):

100
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Transaction costs have a well-deserved bad name as a theoretical device...because there

is a suspicion that almost anything can be rationalized by invoking suitably specified
transaction costs.

Whatever the merits of the transaction cost argument, digital technologies are reducing
them. They are also making inroads on another problem with rental: that people tend
to take less good care of items they rent rather than own (a classic principal-agent
problem). This has hitherto presented an obstacle, though not an insurmountable one,
to renting a spare room, the family car, or anything held to be close to the private
domain. For example, how well will a renter take care of a sublet room or apartment?
Typically, landlords ask for security deposits to cover risk of misuse and wear and
tear. But digital technologies are giving renters and landlord more tools to mitigate
this incentive problem. As discussed above, the possibility of rating both parties to
atransaction establishes a reputation. The value of the reputation loss may be much
worse in the digital economy and this may potentially resolve the problem of a broad
but sweepingly unidentified transaction cost. Indeed, there are reports of people having
a difficult time hailing an Uber car, likely due to poor ratings in previous transactions;
see Streitfeld (2015b).

How much will the sharing economy grow and how much will rental increase com-
pared to owning? While guesses abound in reports, there is little foundation for such
forecasts and those that exist tend to be rather speculative. As consumers become
more used to the sharing economy, the direction of growth is unquestioned but the
speed and scope will depend on regulatory responses. Notably, the sharing economy
has evolved from areas with large fixed costs, such as housing and cars, to other areas.
But at some point it is likely that the marginal benefit of using sharing services will
reach a plateau, not dissimilar to the argument made by Coase about the size of the
firm with respect to deciding what to do in-house and what to outsource. People are
unlikely to want to rent items that cost little and are used every day. The economy is
now in a transition phase where producers and consumers are exploring the limits of
sharing and where that new plateau might be.

Box 2.3 What is fintech?

Silicon Valley is coming, wrote Jamie Dimon, head of JP Morgan in a recent letter to share
holders; see Dimon (2015). He was referring to the rise of peer-to-peer (P2P), peer-to-busi-
ness (P2B) and other acronyms popping up like weeds to describe the services offered by
platforms like Funding Circle, Lending Club, and Prosper (see Box 2.2. A few firms in the
sharing economy.99 Stockholm has become a major hub of fintech in Europe, second only
toLondon, with 18 percent of all European fintech investments amounting to $532 million
during 2010-14.100 Fintech is still small compared to volumes in global finance estimated at
$4.7 trillion with profits of $470 billion; see the Economist (2015m). So far, P2P platforms
have been issuing a comparatively small share of overall lending volume with about $10 billion
in loans last year; see the Economist (2015l). But fintech is increasing rapidly, with the poten-
tial to disrupt the big banks if they allow it.
Banking is particularly amenable to digitalization, as the main commodity is already represen
99 ted by numbers. Fintech companies use digital platforms for all their activities, thus largely
100 replacing physical bank branches with websites and cloud computing.

99
For an overview of fintech, see, for example, Baeck et al. (2014), Morse (2015), and McKinsey (2015).
100
See http://www.investstockholm.com/globalassets/2.-understartsidor-investment-opportunities/fintech/stockholm_fin-
tech_report.pdf

101
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Fintech thus has the potential to vastly increase the efficiency of matching available capital
to entrepreneurs and households. A small firm or startup looking for modest amounts of capi-
tal to invest in new machines or to expand are able to directly reach investors via digital plat-
forms with low transaction costs. Also, the digital process of vetting and due diligence on risk
is often much quicker than the methods used by traditional banks, especially time-consuming
face-to-face interviews. Also noteworthy is that the traditional banks costs of doing due dil-
igence on a small-to-medium loan are not appreciably lower than those associated with a large
loan; see the Economist(2015k). Since small loans typically generate much less profit in rela-
tion to the time required to arrange them, they tend to be less attractive to the banks. The
advent of fintech thus heralds potential improvement of capital availability in a market seg-
ment where it has been lacking while enhancing competition and efficiency at the same time.
Initially, fintech firms were catering to needs that were not being met by the banks, but that
is beginning to change and they are beginning to compete with banks in other segments of
the financial sector as well. In Sweden, Klarna, the electronic payments company, has recently
applied to the Financial Supervisory Authority for a permit to offer banking services, such as
issuing credit cards.

Banks weighed down by legacy IT systems with spaghetti complexity

Many traditional banks have old IT systems, so-called legacy systems, that have evolved over
the years and successively patched and upgraded. Or as expressed by Francisco Gonzlez,
CEO of the Spanish financial services group BBVA: more spaghetti on the spaghetti, quoted
in Arnold(2015). Another unflattering characterization is that asset management [is] stuck
in the digital Ice Age. See Grene (2015).
The software often uses old code written in COBOL, which makes upgrades to modern IT sys-
tems major undertakings replete with risks of faults and downtime that may be very costly.
Moreover, the popularity of bank mergers also implies that many banks have several differ-
ent IT systems with various degrees of compatibility issues. In theory, simply starting afresh
with a new and modern IT system would be the obvious solution but this is not easy in prac-
tice: while getting the inventory of widgets in stock in a firm would be problematic, getting
the numbers wrong in a bank could put the whole edifice at risk. For example, technology
glitches at RBS caused 600,000 payments and direct debits to go missing. See Arnold and
Braithwaite(2015). A botched upgrade could have profound impact on reputation, which is
especially serious in an area where trust is sine qua non.

Benefits
Digital technology can be deployed to make existing ways of providing finance more efficient.
In a sense, fintech is democratizing finance, opening up more investment opportunities and
risk to smaller players, such as mom and pop operations. A key benefit to the economy is
catering to the segment of investment needs for small and medium sized enterprises (SMEs).
While SMEs capital requirements are smaller, the time required to acquire capital from tradi-
tional banks may present a substantial hurdle. Fintech thus overall provides smoother func
tioning of capital needs and improve the capacity of SMEs to invest and grow. This should
ultimately contribute to better productivity growth for the whole economy.
Digital technologies enable a myriad of new ways of providing capital and services:
Virtual investment advice or robo-advisers can offer automated services at lower cost,
24/7, and in different languages; see for example Flood (2015). Clearly, building trust can
be a challenge, but substantially lower fees could compensate for this, especially since the
track record and quality of human investment advice is not held in universally high regard.
Refinancing of loans to poor students based on future earnings potential calculated accord-
ing to academic history; see the Economist (2015l, n).
Extend credit to people with poor credit ratings based on guarantees from friends and
family, thereby relying on potential social pressure to repay; see Economist (2015n).

102
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Searching vast amounts of unstructured financial data to detect fraud; see the Economist
(2012a).
Lending money based on assessments of creditworthiness derived from cash-flows, past
bills, invoices, employment history, and social media history; see the Economist (2015n).
Invoices issued to large corporations may as creditworthy as corporate bonds, see the
Economist (2015k). Some lenders require intrusive monitoring of current accounts to mon-
itor activity and substantial changes may trigger a new payment plan; see the Economist
(2015n).

How threatened are traditional banks?

Banks make money on their net interest margins, charging fees for payments and services,
and on the differential between buy and sell rates for currencies and financial instruments.
Most of these are now subject to competition from fintech firms in one way or another. The
effects on banks have been evident for some time, with continued reductions in employment
as transactions and analysis migrate to digital platforms. In its latest Global Annual Banking
Review for 2015, McKinsey(2015) emphasizes that cutting costs in banking is the main if
not the only driver of profits.
While the competition is real, how large the threat is to traditional banks is an open question.
Gapper (2014) argues that technology will hurt the banks, not kill them. One argument is
that there are still barriers to competition on the core business of deposits and other highly
regulated areas of finance, an area that is not profitable when interest rates are extremely
low. The Economist (2015j) notes that banks may not be as incapable of change as claimed,
noting that fintech firms have so far not broken into everyday banking, such as current
accounts. In Sweden, the Financial Supervisory Authority has reassessed oversight and may
require fintech firms to acquire permits as credit institutions with stricter requirements; see
Wolf-Watz(2015). In Germany, new rules may stifle investments in crowdfunding because
investors are required to issue a statement about their financial status for investments above
1,000 and these are only permitted if the investor has liquid assets in excess of 100,000
or monthly income of more than twice the amount invested; see Vasagar (2015b).

Risks
The rise of fintech is bringing a myriad of new interconnectedness into the economy and
there is little data on or understanding of its size and magnitude. Individual fintech firms
may have control over their own lending data, but loans and credits may be passed on to third
parties or along even longer chains. This could change the sensitivity to financial shocks in
ways we do not yet fully understand. Notably, the effects on investors and consumer behavior
have yet to be tested in a new downturn; see, for example, Skypala (2015). Moreover, it remains
to be seen what will happen to profits in fintech when the current super-low interest environ
ment ends and interest rates rise to historical levels.
These and other questions related to understanding the implications for monitoring finan-
cial stability are emerging. Tett (2015) argues that the elements of financial innovation, arbi-
trage, and cheap money in fintech bear some resemblance to events and actions leadingup
to the financial crisis in 2007; see also Gapper (2015). While the amount of loans is still small,
the new tentacles are creating new interdependencies that may become systemically impor-
tant; see Corkery (2015).
Another set of questions concern risks to low-income borrowers. Just as sub-prime borrow-
ing allowed consumers to buy houses they essentially could not afford, fintech can provide
easy credit in a matter of minutes to individuals that may be beyond their ability to repay, for
example, in the event of unemployment. With interest-on-interest accumulation on bad loans,
the situation may quickly become untenable, especially for people who have accumulated sev-
eral loans or who take out new loans to finance repayments of old ones; see Corkery(2015).

103
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

2.6 Output gap and sustainability

2.6.1 Misidentifying potential GDP, missteps in policy?
No one quite knows what potential GDP is, but it is used a lot in forecasting. To be
fair, the difficulties with potential GDP are well known and recognized; few use the
concept without caveats and other supporting statistics. It is typically thought of as
the level of GDP where inflation pressures are neutral or, for inflation-targeting cen-
tral banks such as the Riksbank and the Bank of England, the level of GDP consistent
with being on the inflation target.

But what does it have to do with digitalization? Arguably, it is becoming increasingly

important. It is typically not potential GDP directly that is used by central banks,
governments, and forecasting institutions, but rather the difference between poten-
tial and actual GDP the output gap. The output gap is a measure of the degree of
resource utilization in the economy: being below the potential implies that there are
free resources available and vice versa. This matters because in the conceptual frame-
work of stabilization policy, being below potential GDP typically implies that infla-
tion pressures are weak and monetary policy can be expansionary (i.e. low interest
rates); conversely, if the economy is above potential, resources are being well utilized
and inflation pressures may be building up, an argument for the central bank to have
more restrictive monetary policy (i.e. higher interest rates).

What, then, is the connection to digitalization? In this chapter we have set out argu-
ments for how the economy is subject to more competition, works faster and more
efficiently, and more goods reach more people at lower cost. These are all factors that
pertain to supply side efficiency. In some cases, we get better utilization of existing
resources and in other cases, productivity may rise.

Methods to calculate potential GDP

Before discussing the detection of structural shifts in potential GDP driven by digital-
ization, it is instructive to first review methods used to calculate the measure. There
are three main approaches, all with various strengths and weaknesses: the production
function approach, the unobserved components approach, and pure statistical filtering
(such as the HP filter). All the methods use historical data to estimate potential GDP
but they differ in how much economic theory they impose; pure filtering techniques
typically use no theory at all, but instead just smooth fluctuations and essentially draw
a trend line through GDP, thereby creating the business cycle.

It is well known that GDP and hence potential GDP is measured with considerable
uncertainty; see for example Orphanides and van Norden (2002), Sometimes, revised
statistics can radically change the position in which the economy was thought to be
several years hence. It is also known that some measures, such as the HP filter, are the
most inaccurate for the start and end points in the series, where arguably interest is
the strongest.

Structural shifts in potential GDP?

If indeed structural change in potential GPD is underway, what would it look like?
The most helpful structural change is of course a big spike from one year to the next,
clearly observable from an identifiable event. For example, for variables such as Foreign
Direct Investments (FDI), at least for small countries, big shifts can often be attributed
to single events, such as a merger or relocation of corporate headquarters, and so on.

104
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

No such easy visual clue exists for GDP and digitalization. A shift induced by digi-
talization is likely to be gradual and last a long time, perhaps decades. There are also
issues with measurement; see Coyle (2015). As production using digital techniques
continues to improve efficiency in the economy, there may well be improvements from
year to year that are very difficult to detect, given the considerable uncertainty that
already surrounds GDP measurement. Even small improvements by one tenth of a
percentage point of GDP or small multiples thereof is a substantial shift but is
dwarfed by measurement uncertainty. One conclusion from this is that policy anal-
ysis needs to pay particular attention to other sources of information about shifts
in potential GDP to make an informed judgment about the effects of digitalization.
When it comes to effects on inflation, however, there are some further issues involved,
which are discussed in the penultimate section of this chapter.

2.6.2 Digitalization and sustainability

The idea of using resources more efficiently is closely linked to the idea of sustain-
ability in a broader sense especially for the climate and for population growth. If
the same welfare and consumption can be obtained through better use of existing
resources, this should improve prospects for sustainability. Digitalization has perhaps
most environmental impact if it can lead to more efficient transportation and less waste
in travel. Truckmakers, like Volvo and Scania in Sweden, are exploring ways to have
several trucks driven in a convoy with minimum distance between each, thus reducing
air resistance to all but the first truck in the line, potentially leading to substantial fuel
savings. Intelligent traffic lights and other infrastructure that help cars and trucks
maintain a steady speed may also reduce fuel consumption.

The sharing economy is also associated with better use of resources, especially when
pockets of idle resources are put to use. For example, when fewer seats are left empty
in cars less carbon is used per passenger; intelligent software can improve traffic flows
and public transport and so on.

It is thus tempting to think that there might be a one-to-one correspondence between

the rise of the sharing economy through digital services that improve efficiency and
other goals of sustainable climate change. While some contribution by digitalization
to ameliorating climate change is likely, the effect may not be quite so simple and
there may be forces moving in the opposite direction. Economists have long discussed
the effects of price changes in terms of substitution and income effects. In this con-
text, the implication could be that better use of existing resources leads to spare
resources at given consumption levels, which can translate either into higher savings
or investment but could just as well translate into higher consumption. Suppose, for
example, that a household saves money by renting out a spare room. This money can
of course be used for whatever purpose, including higher consumption of energy, for
example.

Thus, the effects on the environment from better use of existing resources are ambig-
uous. But even if all of the gain is used for consumption, this may bring considerable
benefits. Notably, lower income families will also be able to enjoy a higher standard of
living due to better availability of cheap goods and services. For example, Fraiberger and
Sundararajan(2015) use a model to estimate the benefits of peer-to-peer (P2P) rental
for cars and find a significant shift away from ownership to rental. Their results indi-
cate that consumption shifts are more significant for below- median income users, to
whom most of the consumer surplus also accrues. Or put differently, those with lower
incomes can benefit more from the sharing economy because spending less money
makes for the possibility of a better consumption bundle.

105
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

While this result is based on estimates from P2P rental of cars, it is likely that the
larger welfare effects for low income groups holds much more broadly when it comes
to consumption. Many more people will be able to afford to buy services that were
hitherto the province of high income earners, or as expressed by Hal Varian, chief
economist at Google: A simple way to forecast the future is to look at what rich
people have today, quoted in McAfee (2015). Of course, higher consumption is only
one aspect of economic wellbeing and the overall result also hinges on how the labor
market adapts, the topic of Chapter 3.

2.7 Can digitalization explain flatline inflation?

Inflation may be affected by a host of factors, all from oil shocks and the price of
coffee to wage setting in the labor market. It is beyond the scope of this report to dis-
cuss all of these issues. Instead, we will outline how digitalization may be contribute
to low inflation. Can a mismeasured output gap, discussed above, in conjunction with
more price competition explain how inflation has flatlined in Sweden and other coun-
tries despite the efforts of the central banks to kick-start the economy with quantita-
tive easing and negative interest rates?

Breman and Fellnder (2014) argue that a fair share of the Swedish consumer basket,
about 37 percent, is either directly or indirectly affected by digitalization. Apel et al.
(2014) find some support for a structural shift in how firms set prices. In its monetary
policy reports from February this year, the Riksbank posited low capacity utilization
as a main explanation, along with the effects of import prices and exchange rates. In
the June report, the Riksbank emphasized that prices of domestic goods and services
are rising and that resource utilization is beginning to increase.

The Riksbank also discussed factors other than macro. Jonsson (2007) discusses
increased competition and inflation. The possible channels through which digitalization
is affecting the economy are outlined in the February 2015 report; some of the micro
factors discussed in this chapter are illustrated among these. Moreover, the Riksbank
mentions the possibility that structural change and technological unemployment
willconstrain wage growth and hence inflation. One particular channel is through
e-commerce and greater price transparency holding back inflation. The European
Central Bank, on the other hand, discusses the effect of e-commerce on euro area
data and finds only limited evidence of price effects (Monthly Bulletin, Feb 2015).
But, as for Sweden, durable goods have declined in the last ten years, especially due
to competition in electronics.

These are useful first steps in the analysis but they miss the broader picture. Overall,
what we are seeing is many markets undergoing microeconomic improvements that,
as a whole, are changing how the whole economy behaves: microeconomic changes
are having macroeconomic impact. In this chapter, we have discussed the range of ways
in which the economy is becoming more efficient: digital technologies are improving
the match of supply and demand in product and labor markets; price comparisons
and transparency are changing the way consumers consume. Demand for labor is
probably also changing, with more demand for special skills.

More efficiency in matching supply and demand can raise potential growth, as does
lower capital needed to start firms. But better use of existing resources can also give
lower inflation impulses at the same level of resource utilization. In some cases, for
example with carsharing, the need to own a car declines and hence one effect may
be to dampen consumption of durable goods. Depending on what happens with the

106
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

resulting higher disposable income, the effects on GDP can be different. If consumers
use the freed resources to buy other goods and services, the overall effects may be
neutral; but if they use some part to increase savings, growth may be dampened
fromthe demand side.

The supply-side effects of digitalization increase the efficiency of production but the
speed of the improvement depends on a variety of regulatory and institutional obsta-
cles. It is likely that digitalization will improve productivity growth but the effects
may take long time to materialize, especially in light of the known dampening macro
trends on growth discussed in Chapter 1. How much of the micro changes from dig-
italization can explain inflation is an empirical question that has received too little
focus from central banks. Indeed, the microeconomic changes are largely absent from
the typical macro models used by central banks and this is a serious omission in times
of rapid technological change.

2.8 Summary
Digitalization is bringing a microeconomic revolution to the economy. Some of the
changes are fast, notably those that are consumer-driven. Other changes are subtler,
especially those related to resource utilization. Technological improvements have been a
recurrent theme since the Industrial Revolution and the changes have had huge impact
on living standards and welfare. While the changes now underway are not as funda-
mental as during the Industrial Revolution, they are affecting several sectors within
acompressed period of time. They have the potential to improve productivity growth
against the known macroeconomic headwinds of demography and high public debt.

In this chapter we have also explored how digitalization can affect capacity utilization,
a concept that underpins a great deal of the work of forecasting and simulations carried
out by policy institutions. The effects discussed here are of a qualitative nature, but
a lot more work needs to be done to understand and estimate the size of the changes
already in motion. Even small changes on the order of one, or even a few tenths of
one, percentage point on GDP growth may be significant and can alter policy conclu-
sions. The challenge is compounded by measurement in official statistics. As goods
are becoming digital services and many popular digital services are free or available
at low cost, the uncertainty surrounding estimates of inflation and GDP is becoming
more serious.

Digitalization is changing the way the economy works by reducing frictions, improving
knowledge transfer, and reducing the required capital needed to start companies. It
also changes how business is conducted and the nature of competition. But while
competition is increasing on many fronts, we are also seeing the rise of companies
with near-monopoly power due to rapid expansion with network effects, putting a
new face on challenges for competition policy and regulation. Current regulation has
evolved over the years in an analog world and need to be updated. Indeed, how well
we are able to benefit from digitalization will to largely depend on how regulation
tackles these challenges (further discussed in Chapter 4). The sharing economy is an
especially big shift and may have wide-ranging implications for how we buy goods
and services. It may also have profound implications for employment and increase
the prevalence of self-employment in the economy (discussed in Chapters 3 and 5).
Moreover, with the expansion of the sharing economy, the pros and cons of owning
vs renting may shift dramatically. We are now only embarking on the process of
understanding how far-reaching these changes will be.

107
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

3. Implications for the labor market

Ive come down in the world.Mr. Molesley in the TV series Downtown Abbey

3.1 Introduction
In Thinking Fast and Thinking Slow by Daniel Kahneman, a main theme is how two
parts of the brain interact with each other to perform different functions. The first
part is autonomous, governing functions such as walking, face recognition, and other
things that we do mostly effortlessly; the other part is the analytical one that requires
conscious effort from the brain. With some oversimplification, computers and robots
are good at the latter but bad at the former, while for humans, it is the reverse. These
two worlds are increasingly colliding and causing frictions. And it is occurring not in
a dramatic Hollywood fashion but affecting everyday work in broad sectors of the
economy ranging from manufacturing to legal research and creative writing. Does
this mean that the robots are taking over our jobs?

One conclusion to emerge is that the last few centuries of technological change have
not, in fact, led to mass unemployment; that is, technological change has been neu-
tral for the amount of work in the economy in the long run. Moreover, as emphasized
in Autor (2014), there is evidence for complementarity between machines and human
work, especially in the latter half of the 20th century: machines are improving worker
productivity rather than reducing the total amount of human work. But, as we will
explore in this chapter, there are some compelling arguments for why this might not
be sustained during a period of transition in the coming years, looking one to two
decades ahead.

Even though long-run neutrality for technology and jobs is a reasonable starting posi-
tion, it is not a law that holds true under all circumstances, especially not in the short
to medium run. There are two reasons to be concerned about the pace of automation
and its effects on labor:
Digitalization has broad effects on virtually all sectors at roughly the same time
The changes today are much faster the previously technological disruptions

The first point makes digitalization different from previous periods of structural
change. The market economy is quite capable of shifting resources from waning
sectors to waxing sectors but the challenge is greater when the driving force affects
most of the economy virtually simultaneously.

The multi-shock challenge is compounded by the high speed of technological dis-

persion, currently much faster than previous technological disruptions, as illustrated
in Figures 2.1 and 2.2 in Chapter 2. It is not that Adam Smiths invisible hand has
stopped working but rather that the labor market and our institutions may be too
slow to adapt due to various frictions. The effects of digitalization on the labor market
over decades may be quite dramatic even if the year-on-year changes are small. It is
difficult to predict the magnitude of the effects, but they are likely to depend on how
firms respond and on how political institutions adapt. A muddling through, or even
worse, protectionist response may increase the risk for a prolonged period of bad
labor market outcomes.

108
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

It may be worthwhile to recall that the Industrial Revolution contained episodes of

difficult adjustment. People had to adapt or be impoverished, especially without the
social welfare protection that is available today. As emphasized by Mokyr (2004),
Life did not improve all that much between 1750 and 1850. Mr. Molesley, the
fictional character in the TV series Downtown Abbey set in the early 20th century,
finds his job as footman gradually eroded in pay and status. The modern equivalentof
this could well happen to professionals, office workers, and skilled labor. As elucidated
by McAfee and Brynjolfsson, there has never been a better time than the Digital Age to
be a worker with special technological skills, but it is not such a great time to have
ordinary skills, quoted in Bernsteinand Raman (2015). Ford(2015a) and Cowen
(2015) make similar points and argue that weak wage growth for large groups in
the US is only the beginning of wage polarization in the labor market. Susskind and
Susskind(2015) go even further and argue that technology will imply an end to
theprofessions as we know them, as new digital tools will be able to more cheaply
incorporate expert skills in law, medicine, architecture, etc.

In this chapter, we first discuss how the Swedish economy has coped with structural
change in the past. We then draw on the academic literature on the effects of technology
on jobs. Next, we turn our attention to advances in robotics and the implications for
the labor market, especially the way robotics may affect the amount of human labor.
We examine the literature that predicts that a large share of current jobs may be auto
mated through digitalization, notably Frey and Osborne (2013). While the results are
intriguing, the risk that particular professions may be automated may distract from
amore pertinent question. The issue is rather how to design policies that make the
consequences of structural change as unrugged as possible without damaging produc-
tivity growth. More specifically, there is a challenge in improving conditions in the
private sector to ease job creation, a topic we also return to in the last chapter.

Indeed, it is crucial to look at both job destruction and creation in the labor market,
not just the former as in Frey and Osborne (2013). We draw on new developments
in robotics and software automation to discuss features that are likely to affect the
impact on the labor market that have not received much attention to date a more
holistic approach. We then show that some professions that might, according to pre-
vious estimates, have diminished have in fact increased. Other forces, such as demo-
graphics, demand, and profitability can be crucial. Even if digitalization is one strong
driving force in the economy, it is not the only one.

We also highlight some aspects of the ongoing technological change that are particu-
larly relevant for small, open economies. While the automation of manual work may
easily spread from the US to other countries, the same may not hold true for certain
non-manual work, such as business analysis, legal research, and journalism, thatare
increasingly being automated in the US. The potential gain of automation may be
muted in smaller countries because smaller scale weakens network effects. Or, in
other words, the automation of work done in the English language is going to be
stronger due to its bigger scale.

For all these reasons, the US is likely to be far ahead in automation for non-manual
work. Moreover, in areas where institutional features are important, such as law,
automation software from the US will also not translate directly. But it would likely
be wrong to conclude that small countries are safe from automation. A more con-
structive view is that there are opportunities for entrepreneurs to find clever ways to
handle idiosyncrasies and language in smaller countries and not wait to be swooped
up by a big (US) platform, a point we discuss in Section 3.4.

109
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

3.2 Lesson from previous periods of structural change in Sweden

The models, empirical work, and statistical analysis of how jobs may be automated,
or how humans and new technology can complement each other are somewhat lim-
ited, because the period we know most about, the latter half of the 20th century and
onwards, is relatively short. It may very well be somewhat special too, in that is dom-
inated by increased demand for skills arising from technological change.

In this section, we draw on lessons from economic history to give a broad picture of
structural change and its implications, at the expense of being somewhat less empiri-
cally precise. This sets the stage for the next section, where we review the academic
literature on jobs and technology.

In addition to reviewing broad changes in this section, we also examine the specific
experiences from periods of major upheaval, or job destruction, to draw some lessons
that may be applicable in the future. In particular, we discuss the experiences of the
slow erosion in textile production as well as the more dramatic implosion of the ship-
building industry. The reason for selecting these two sectors is twofold. First, they
constitute major changes within their respective sectors, the most significant disrup-
tions in the Swedish economy in the 20th century. Second, they illustrate paths of dis-
ruption one fast and one slow that span the space of possible outcomes but with
the significant difference that digitalization is set to be much broader in scope and
to affect several sectors at the same time. Or put differently, it is worthwhile to learn
from all periods of structural change and determine what aspects may be relevant to
how we adjust to digitalization.

It is worthwhile to stress that Sweden has a long tradition as a small, open economy
with significant trade. Schn (2014) paints a picture of a trading nation going through
waves of technical change and being subjected to fluctuations driven by external shocks,
such as development in raw materials. From the 1950s, Sweden did rather well and
developed heavy industries and manufacturing multinationals, notably Atlas Copco,
Alfa Laval, Brown Bovery (later ABB), Ericsson, Volvo, and Saab, to name a few.

3.2.1 Electricity adoption disrupted jobs

Electricity is a general technology that impacted on all aspects of the economy, and
therefore may contain information as to what we might expect from digitalization.
Barroso Morin(2014a) uses a newly digitized dataset for the concrete industry from
19291935 in the US and finds that electricity adoption did induce a shift away from
labor. If this holds true for electricity, it may be even more so for computers, also an
all-purpose technology. Indeed, Edquist and Henrekson(2006) argue that ICT is the
only all-purpose technology that has increasing returns to scale, due to Moores law.
Moreover, Barroso Morin(2014a) also shows that the labor share of income declines
and there is some evidence for jobless recovery in the US.

More research is needed to better understand the extent to which the result is appli-
cable in a broader context and over longer time periods. The effects of electricity in
the last century were fundamental and occurred over a long period of time. As we
discussed in Chapter 2, the speed of digitalization is an argument that supports larger
effects than from other technological advances, including that of electricity.

3.2.2 Slow decline in textiles in Sweden

The textile industry is one example of a prolonged decline and may be relevant to
other industries exposed to global competition or disruption from new technology. At

110
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

the outset, the textile industry was fairly sizeable. Employment in the Swedish textile
industry rose fairly steadily from the beginning of the 20th century and reached its peak
in the 1950s at around 115,000 workers, see Figure3.1. In terms of total private-sector
industry, textiles accounted for about 15 percent in the 1950s, see SIND (1991) and
Gustafsson (1983).

Figure 3.1. Employment in the Swedish textile industry 19151980

Source: Gustafsson (1983, page 366).

From the 1950s, due to strong competition from abroad, textile production was put
under pressure and declined fairly steadily. For some communities, the changes were
rather drastic but these changes are not well documented and thus mainly anecdotal.
Many of those employed were women, for whom employment opportunities in other
industries were likely limited at the time. In 1989, only about 20,000 or 2 percent of
the total manufacturing jobs were in textiles.

Unlike the experience from the decline in textile production, digitalization will likely
allow a relatively short time to manage adjustment.

3.2.3 The Swedish shipbuilding crisis of 19751982

The latter half of the 1970s turned out to be a period of crisis and decline for the
Swedish economy. A series of political choices led to lack of fiscal discipline and ten-
sions in the labor markets, all aggravated by inflation from high oil prices. Mining,
paper, and steel mills were affected, drastically reducing the demand for iron ore. But
in terms of severity, the downturn in the shipbuilding industry was unprecedented
in Swedish history; see Schn (2014) and Magnusson (2014). In a few short years,
19741977, the value of the shipbuilding industry was cut in half. At its peak in the
mid-1970s, shipbuilding employed about 40,000 people, see Figure 3.2.

At its peak in the mid-1970s, shipbuilding accounted for about 1.6 percent of the total
private sector, a share that continued to fall up until the 1990s when it accounted for
about 0.3 percent. While one should be careful in comparing timelines, the adjustment
in the textile industry occurred over roughly three decades, while the decline of ship-
building was much more compressed, over a period of roughly 15 years.

111
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 3.2. Employment in Swedish shipbuilding 19451990

Sources: Olsson (1995) and Svensson (1983).

Note: There are some minor differences in definitions, for explanations see respective source.

Sweden was not the only country in crisis in the end of the 1970s. The UK also
experienced a period of decline, with unrest and strikes in the mining industry. The
response of the Swedish authorities was different from elsewhere in Europe, notably
Germany and France. In Sweden, the government assumed ownership of shipbuilding
companies, thus prolonging the period of painful adjustment and incurring consid-
erable losses for the taxpayers. In other countries, prices in declining industries were
sometimes subsidized instead; see Schn (2014).

We cannot know what would have happened without government nationalization,

but swifter bankruptcy seems a likely scenario.

3.2.4 Experiences from larger disruptions at the firm level

It is instructive to review major events in Swedish corporate history to discern the
economys capacity for structural change, regardless of whether or not develop-
ments were directly related to digitalization. Big events can tell us how workers are
absorbed by the same sector or how they move to another area. In this regard, one
of the largest defaults in modern times is the one that of the automaker Saab in
December 2011. The company had been making losses for a long time, had changed
owners, and had tried various ways to restructure.

Although the end of Saab should not have been a surprise due its lossmaking and
against the backdrop of global overcapacity in the auto industry, for many people it
was; see TRR(2014). The default was a major event with 3,300 jobs lost in the city
of Trollhttan, with a population of about 56,000. Despite the scale of the job losses,
most of the people affected found new employment. Assisted by Trygghetsrdet101,
within a two year period, 91 percent of non-manual workers found new work, 4 per-
cent started new companies, 4 percent went back to school, and 1 percent chose to
exit the program. Of those gaining new employment, about 4 out of 5 got the same
or higher pay; see TRR (2014).

101
Trygghetsrdet is a non-profit foundation owned by labor market partners and helps displaced non-manual workers
find new jobs.

112
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Another major event in Sweden was the decision of pharmaceutical giant Astra Zeneca to
eliminate research activities in Sdertlje in February 2012. Most of the people affected
were highly skilled workers and within a two-year period most, about 85 percent, found
new jobs afterwards, although only about 55 percent of those workers received the
same or higher pay; see TRR(2015). The examples of Saab and Astra Zeneca illus-
trate how the labor market manages to adjust to large firm-specific shocks. Figure 3.3
illustrates the general trend for numbers of displaced workers. Prior to the 1980s, non-
manual workers in Sweden were rarely laid off, a trend that changed thereafter.

Figure 3.3. Workers losing jobs

Non-manual workers
70 000
Manual workers
60 000

Source: Trygghetsrdet TRR (roughly, the job security

council), for non-manual workers; Trygghetsrdet
50 000

40 000

30 000

TRS for manual workers.

20 000

10 000

0
80

82

84

08

10

12

14
86

88

90

92

94

96

98

00

02

04

06
19

19

19

19

19

19

19

19

19

19

20

20

20

20

20

20

20

20
Note: no statistics are available for manual workers prior to 2004. The diagram shows only people who were
eligible for TRR or TRS and not all redundant workers.

3.2.5 Structural change 19702012, overall changes in employment

In the 1970s, the agricultural sector had already undergone radical decline, see
Table3.1, but it would shrink still further by nearly 5 percentage points as a share
ofoverall employment until 2012. The most noteworthy change over this period is
the decline of employment within the industrial sector by about 10 percentage points.
Unpaid work inhouseholds also declined, as more women entered the labor force.

Table 3.1. Overall changes in employment, Sweden

Sector Share 1971/75 Share 1991/95 Share 2008/12
Agriculture 7.1 3.9 2.5
Industry 28.5 20.5 18.8
Construction 8.6 6.1 6.9
Transport 6.6 7.0 7.3
Trade 15.5 15.3 13.3
Other private services 9.7 13.8 22.8
Public sector 24.0 33.4 28.5
Total 100 100 100
Paid work 74.8 85
Unpaid household 25.2 15
Total 100 100 -
Source: Schn (2014).

113
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Better data is available for later periods. In a comprehensive study of the Swedish
labor market, Heyman et al. (2013) investigate job flows using micro data. They find
that 190,000 new jobs were created in the period of 19902009, with a net result of
about 3.4 million jobs created against 3.2 million jobs destroyed. In an average year,
about one-fifth of all jobs in the labor market were turned over, giving some indica-
tion of the Swedish economys ability to adjust to changing demand and new tech-
nologies. Most of the jobs were created in the service sector, while the manufacturing
industry exhibited a decline in employment. In a study of job reallocation in Swedish
manufacturing during 19721998, Andersson (1999) finds considerable flows. But
there are periods of large scale movement hidden behind the smaller average change.
There is also evidence that most of the reallocation occurred within narrowly defined
sectors within manufacturing.

Some other trends during those years are also worthwhile to highlight. Small and
medium size enterprises accounted for most of the new jobs while large companies
tended to reduce the size of their workforce, especially during the crisis in the 1990s.
During these years, educational requirements increased and the number of people
without high school diplomas (secondary school leaving certificates) decreased by
about half; see Heyman et al. (2013). Thus, the demand for skilled labor increased.

3.2.6 Empirical estimates for risk of automation

A new branch of the literature tries to estimate the probability that some tasks will
be automated and hence the likelihood of jobs disappearing. In this research agenda,
jobs are assessed according to tasks, whether requiring mainly creative, repetitive,
empathetic, or cognitive skills. Using statistical techniques in static context, the head-
line numbers indicate the probability that a robot or a computer may replace a human.

Frey and Osborne (2013) have received much attention for their paper, in which they
assess that almost half of the jobs in the US today may be automated in the not too
distant future, the next two decades or so; for the UK, the figure is somewhat lower
at about 35 percent, see Knowles-Cutler et al. (2014); for Sweden, slightly over half
the jobs are subject to automation, according to Flster(2014) who uses similar
methods but different data. Also based on Swedish data, Heyman et al. (2015) esti-
mate that about 40 percent of the total workforce are in the high risk category, about
50 percent for the private sector, a result similar to the Frey and Osborne(2013).

Apart from the headline result that almost half of US jobs may be automated in the
near future, Frey and Osborne (2013) also give detailed probabilities for each profes-
sion. They identify 117 out of 702 professions as having above 90 percent chance of
automation, among them fashion models, library technicians, telemarketers, parale-
gals, and manicurists. Heyman et al. (2015) identify 14 such professions out of 114
Swedish occupations, including fashion models, numerical, library, and filing clerks,
shop and stall salespersons, and assemblers.

These models provide one way to address the risk of automation that is based on
assessing the task according to categories but this is only a partial view of how auto-
mation may proceed. In Section 3.4.3 below we will discuss other factors that may be
just as important, notably demography, regulation, and the costs of automation com-
pared to human labor. Moreover, the analysis focuses only on job destruction even
though we know from experience that new jobs are created all the time as a result of
new technology and changing demand. For instance, personal trainers, professional
computer gamers, hackers-cum-data-security engineers are all jobs that, not so very
long ago, were either non-existent or very few in number.

114
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

On a methodological note, the assessed risks of automation are forced into fairly uni-
form slots on the number line between 0 and 1. For example, in Frey and Osborne
(2013), we find probabilities such as 0.98, 0.98,...,0.99 in neat order. But the ordering
is purely a result of distributional assumptions. If indeed there were objective prob-
abilities of automation, based on all factors that mattered, there is no inherent
reason for them to be ordered nicely.102 Thus, there is no basis for making comparisons
between professions that are close to each other in terms of probability of automation.

Evidence of effects of digitalization and automation but other factors matter too
The labor market is in constant flux, with employment increasing in some professions
and decreasing in others. The reasons are sometimes complex, other times easier to
isolate. The effects of digitalization are among the more complicated. We know that
digitalization has caused disruption in photography, music, movies, and the publishing
businesses, but more gradual changes can also be discerned. For example, when older
workers retire, firms sometimes use these opportunities to change work processes
rather than replicating previous methods. This implies that the effects of digitalization
on the demand for labor might also come gradually, especially in mature industries
and services. But which effect will dominate?

In order to illustrate how specific professions are changing, we have used data from
Statistics Sweden with three-digit classification covering the period 20012013103.
This period covers the aftermath of the dotcom bubble at the turn of the century
and a few years of high productivity growth. It also encompasses the recent finan-
cial crisis. Moreover, some professions are small in terms of total employment, and so
changes in percentage terms can be very large. For all these reasons, it is important to
be careful when drawing inferences from the data.

A question of particular interest is in which professions employment have changed

the most over this time period, as this can give an indication of the broader changes
in the labor market, some of which stem from digitalization. In Figure 3.4 we illustrate
the six occupations that have decreased the most in absolute numbers of workers
employed; shares of total employment for each are displayed above the bars. The big-
gest decreases have occurred for blacksmiths and toolmakers, a decline of two-thirds
over the whole period, from 25,869 to 8,995 workers with a share of total employ-
ment in 2013 of about 0.2 percent. In all likelihood, this change is due to other changes
in the labor market. The decline in machine operators in the wood products and metal
working industries may be due to changes in global and domestic demand as well
as the perennial drive for productivity growth in manufacturing, some of which is
probably related to increased use of robots and machines. But the decline in manu-
facturing laborers may be related to increased use of robots. The decline is also con-
sistent with the high risk of automation estimated in Heyman et al. (2015) that we
discuss more fully later in this chapter.

Office secretaries, data entry operators, and numerical clerks have been more directly
influenced by digitalization, declining by almost half and just over one-fourth, respec-
tively, during the period 20012013, each representing roughly one percent of total
employment. These are the occupations that should decline even more according to the
probability assessments in Frey and Osborne (2013). It could well be that most of the
decline has already occurred and that further changes will be slower in the making.

102
We would see probabilities such as 0.45, 0.57, and 0.83 with various forms of discrete jumps between different professions.
103
It would be not appropriate to focus only on sectors that are dominated by digital platforms, as those account only
for a small part of the economy.

115
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 3.4. Occupations in Sweden that have decreased the most, 20012013, percent
0,2% 0,2% 0,7% 0,6% 1,1% 0,9%
0%

-10%

-20%

Metal & mineral

-30% Numerical products machine
clerks operators

-40% Manufacturing
labourers
Office secretaries &
-50%
data entry operators

-60%
Wood products
machine operators
-70% Blacksmiths, tool-
makers & related
trades workers
-80%

-90%
Note: Bars show percentage change from 2001 to 2013 in the occupations that have decreased the most in
absolute numbers. The numbers above the bar represent the size of the occupation as a share of all workers
employed in 2013. Own calculations. Source: Statistics Sweden.

The occupations that have increased the most in absolute numbers are displayed
in Figure 3.5. Due to its size, with more than 12 percent of total employment, the
increase in the sector for personal care and related workers is significant (and does
not include the more modest increase in qualified nurses of about 16 percent with
a share of employment just below two percent). Here, the driving forces are most
likely demography and increased demand for health care as incomes rise. It cannot
be ruled out that digitalization has dented the increase in health care employment but
the overriding conclusion is that other factors may at times be much more important
than technology.

Figure 3.5. Occupations in Sweden that have increased the most, 20012013, percent
90%

80%
Production &
operations managers
70%

60%

50%
Shop & stall Building frame
salespersons workers etc. Business
40% professionals
Finance & sales
30% associate
professionals

20% Personal care &

related workers
Source: Statistics Sweden.

10%

0%
2,0% 5,1% 2,5% 2,6% 4,6% 12,3%

-10%
Note: Bars show percentage change from 2001 to 2013 in those professions that have increased the most in absolute
numbers. The numbers below the bar represent the size of the profession as share of all employed 2013.

116
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Rather significant increases in IT professionals occurred during 20012013, where

computing professionals and associated computing specialists increased by (not dis-
played) 34 and 22 percent respectively.

It is noteworthy that the number of business professionals increased during this

period, a trend that might be reversed if some of these functions are taken over by
software, as discussed above. Another occupation that has increased against the tide
of automation is shop and stall salespersons retail workers. Selling via digital plat-
forms like Amazon, provides pressure towards fewer workers but the increase might
reflect the changing nature of such jobs, for example, upskilling in the way that the
invention of automatic tellers actually led to demand for higher skilled bank staff;
see for example Bessen (2015). Production and operations managers represent 2 per-
cent of employment in 2013 and the occupation has increased over the whole period,
which is consistent with a low chance of automation in Heyman et al.(2015a).

What happens if we compare the probability of automation from Heyman etal.

(2015a) with the trend in data since 2001? While these probabilities should be seen
from a forward-looking perspective it stands to reason that we might find some evi-
dence already in the data for the last decade. In Figure 3.6, we use the estimations from
Heyman et al.(2015a) to select the occupations that have a higher than 90 percent
chance of automation. As can be seen, the picture is a bit mixed. Fashion and other
models are supposed to be among those in high risk and yet the occupations share of
total employment is virtually unchanged, albeit a very small group. One explanation
for this pattern is outlined in Box 3.2. Demand related issues, namely that demand
for the profession prefers real people, perhaps because they are deemed more effective
in generating interest in the products they showcase.

Figure 3.6 Change in employment, occupations with high risk of automation, 20012013
100%

Other machine
80%
Agricultural, operators
fishery etc.
60%

Fashion & Salespersons etc.

40%
other models

20%

0%

-20%
Cashiers, etc.
Numerical
-40% clerks

Library &
Source: Statistics Sweden.

-60% filing clerks

Machine
operators (wood)
-80%

-100%
0,0% 1,1% 0,2% 0,1% 0,5% 0,1% 0,7% 5,1%

Note: The occupations selected are those from Heyman et al. (2015a) that have a higher than 90 percent chance
of automation within the next decades. For each, we compute its share as a percentage of total employment and
calculate the change over the whole period, 20012013. The red bars are consistent with declining employment
shares while the blue bars denote increases in employment shares, thus illustrating that other forces can be more
important than the technical possibility of automation. Own calculations.

117
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Three other occupations at high risk have actually increased over the period: agri-
cultural workers, machine operators, and salespersons. But library and filing clerks
have decreased, as one would expect. While the figure gives broad support to prob-
able effects of digitalization it also clear that there are other changes that may explain
these changing shares, which may have little to do with computers or robots, such as
aging populations and changes in demand as incomes go up. It is a mistake to consider
digitalization as the only force that matters.

What can we learn from past periods of structural change?

Two features stand out from the foregoing discussion. First, the economy is always
undergoing changes. Second, there are periods of rapid change, often caused by external
events, such as new technology, changes in demand, or fluctuations in prices of raw
materials. For example, the Swedish mining industry is sensitive to world prices of
minerals. If you take a walk in the countryside, you are likely to stumble across an
abandoned mine that has become unprofitable due to declining prices of minerals
orraw materials in world markets.

Most economists would agree that the economy is resilient and adaptable. To be sure,
some of the periods involved difficult times, but the economy as an aggregate is able
to shift workers from one sector to another. Hall and Krueger(2015) argue that the
sharing economy may provide a cushioning role by easing the transition from different
forms of work and reducing the risk of unemployment. This adaptability is likely the
result of how institutions cope with changing conditions, discussed for example in
Acemoglu (2002). In Davis andHaltiwanger (2014), evidence is found that the US
labor market has deteriorated in its ability to re-allocate workers between sectors,
orlabor market fluidity.

Arguably, the ability to shift work from one sector to another will be crucial in obtaining
the benefits of digitalization while reducing its downsides, notably the risk that tem-
porary unemployment will become long-term unemployment. Experience indicates
that the risk of being permanently unemployed increases dramatically after about six
months out of work, as skills and self-confidence begin to erode.

3.3 The effect of automation on employment and wages

The academic literature seeks to understand the driving forces behind how technology
affects the labor market and especially numbers and types of jobs. These issues are
important because only when we understand them can we adapt and determine the
appropriate policy responses. There is a broad consensus about the empirical facts:
Deskilling of labor occurred during the Industrial Revolution in the 19th century.
Increased demand for skills was prevalent during much of the 20th century.
Job polarization the shrinking of the middle class has taken place in the labor
market over the last two or three decades. In Sweden, only the better paid jobs have
so far increased, an experience that differs from many other OECD countries.
Wage polarization has occurred in the US with stagnant real wage growth for
broad groups, but not in Sweden. In fact, developments in Sweden are rather the
opposite with good real wage growth in the last twenty or thirty years.

Above all, one conclusion stands out from the literature: there is nothing inevitable
about how technology impacts on the labor market. As posited by Acemoglu (2002),
how institutions respond to changes from technology may alter the paths in good
or in bad ways. Moreover, the nature of technological change is likely of crucial

118
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

importance. Notably, general purpose technologies, such as electricity and computers,

may have broader effects than other innovations; see Barroso Morin (2014a, b) for
an account of the effects of electricity. When reading the literature on previous tech-
nological changes and responses of the labor market, it is important to keep in mind
that digitalization differs from previous periods in several ways. As discussed in
Chapter 2, these are:
High speed of adoption at low marginal cost
Increasing returns to scale and network effects
Non-rival and non-exclusive nature to some but not all parts of digital services

Industrial Revolution was mainly deskilling

During the first phase of the Industrial Revolution, technological innovations were
by and large deskilling; see for example Acemoglu (2002), and Acemoglu and Autor
(2011). Unskilled labor was hired to operate machines and thus replaced skilled artisans
at lower wages, a chain of events that led to social upheaval. In particular, the Luddite
movement tried to resist change by destroying machines, efforts that ultimately proved
futile. Remnants of the fear expressed by Luddites have never disappeared and have
now reappeared in public discourse, although so far without the drama ofthe past.

The period of deskilling during the Industrial Revolution is largely a consensus view,
but Katz and Margo (2013) use micro data to argue that the picture of technology
displacing jobs should be more nuanced. In particular, while this effect occurred in
manufacturing, it did not happen in the aggregate economy: starting in the 1850s, highly
skilled workers increased as a share of total workers. Therefore, they conclude that
the picture is mixed: only some reallocations between sectors caused unemployment.

Complementarity between jobs and machines, skill-biased technical change

In contrast, much of the development during the last couple of decades has been
skill-biased technical change. This means that recent innovations have led to an
increased demand for skills in the labor market. Autor et al. (2003) find that com-
puterization can explain about 60 percent of the estimated demand shift favoring
employees with university education during 1970-98. Acemoglu (2002) surveys the
evolution of skills and concludes that most of the 20th century was characterized by
an acceleration in demand for skills.

Much of this literature emphasizes the complementarity between humans and machines;
see the seminal work by Acemoglu (2002, 2003) and a recent account by Autor (2014).
Labor productivity rises over time as employees work with machines as opposed to
the experience during the Industrial Revolution when machines indeed replaced some
people. Graetz and Michaels(2015) find that industrial robots have contributed to
increasing wages and total factor productivity. Akerman et al. (2015) confirm com-
plementarity in Norwegian data.

To varying extents, complementarity between human and machine is evident every-

where. When jobs are destroyed due to machines, they are often the physically arduous
and dangerous jobs. For example, extracting minerals from mines used to be a dark
and perilous business; today much of the work, at least in OECD countries, is done by
machines operated remotely; see Ek(2014). Bessen (2015) uses the example of ATMs
and bank tellers to illustrate the same point. When people no longer had to seea teller
to withdraw cash, bank tellers might have become superfluous. Instead, they began per-
forming higher quality services requiring more skills, such as providing financial advice.

119
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Despite all the evidence of complementarity between humans and machines, there
has always been some angst about robots taking over work, at least outside the eco-
nomics profession. Recently, for example, famous physicist Stephen Hawkins and
Microsoft founder Bill Gates articulated such views and there are several books by
economists and non-economists that express these views, for example Cowen (2015),
Ford (2015a), and Susskind and Susskind (2015). Regardless of the chosen view on
the ultimate effects of digitalization on jobs, the existence of complementarity does
not rule out difficult adjustments in the short- to medium-term.

Decline in demand for skills over the past decade?

Recent work by Beaudry et al, (2015) challenges the trend towards demand for higher
skills arising from technological progress. They argue that this trend was reversed in
2000 and obscured by the aftermath of the Great Recession, thus painting a rather
dark picture. They argue that having a college degree is less a path to management
than being able to compete for entry-level jobs against applicants with lower educa-
tion. In other recent work, Graetz and Michaels(2015) find no effect of robots on
aggregate employment but some reduction in demand for mid-level jobs, similar to
the trend during the Industrial Revolution.

A more detailed analysis of the demand for skills is estimated in MacCrory etal.(2015).
Using US O*NET occupational data for the period 20062014, they detect a significant
reduction in skills that compete with machines and an increase in skills that comple-
ment machines. While not providing conclusive proof, these findings are indicative
that a fundamental shift in the types of skills demanded is underway.

Polarization can increases in inequality and wages be explained by technology?

The literature finds overwhelming support for labor market polarization and lower
wage growth for unskilled workers in the US. For example, Autor and Dorn(2013)
find that job polarization stems from the interaction between consumer preferences
and falling costs of automation. They find evidence of reallocation from other sectors
to those with low pay during 19802005. Notably, employment increased for secu-
rity guards, food operatives, assembly, and mining. In these occupations, wage growth
outpaced other lower level service jobs. Autor (2014) argues that these are jobs that
require a skill set that comes easy to humans and where robots are not yet capable
or too costly; that is, jobs that are neither complements nor substitutes for machines.
For example, robots easily fit the windshield on new cars in factory assembly but
repairing a broken windshield is still a manual job.

In a study covering 16 countries, including the US, the UK, and Sweden, Goos et al.
(2014) find evidence of labor market polarization during 19932010. The results
are replicated for Sweden in Adermon and Gustavsson (2015). They find that mid-
level jobs have so far mostly evolved into more high paying jobs rather than the low
paying kind. Akerman et al.(2015) find evidence that access to broadband internet
increases the wage bill share of skilled workers in Norway. Based on the Scandinavian
experience, employment polarization is not necessarily followed by wage polarization;
see Adermon and Gustavsson(2015). However, this wage polarization has occurred
in the US. Acemoglu (2002) finds that average wages for low skilled workers have
declined in real terms since the 1970s and he conjectures that the outcomes in different
countries stem from how institutions cope with technological shocks. Moreover,
Acemoglu et al.(2014) show that productivity growth in the IT sector is driven
mostly by reduced employment rather than better labor performance. Goos etal.
(2014) stress that technological advances are more important for the labor market
than the effects of offshoring.

120
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The abovementioned studies concern the economy as a whole but research is also
emerging that analyzes the effects of digital platforms on the labor market. Agrawal
et al.(2015) find evidence of frictions in an online labor market as measured by the
value of information. In particular, they find that small pieces of information in the
form of employment history for offshore workers can substantially increase their
future pay compared to those without such as history; also, they find that digitaliza-
tion primarily benefits workers with higher wages.

Theoretical models where machines destroy jobs

Above, we discussed the development during the latter half of the 20th century where
technology led to demand for higher skills. This trend is now beginning to be ques-
tioned in the academic literature. For example, Beaudry et al. (2013) find a shift in
the relationship between machines and skills since the beginning of 2000. Throughout
history, the effects of technological change on the labor market have been different
depending on the rate of structural change and on institutional factors; it would be
rather optimistic to assume robot and human complementarity will prevail in the
short-to-medium term no matter what kind of technology shock makes an appear-
ance. There are periods of change when employment effects have been negative, for
example following the introduction of electricity; see Barroso Morin(2014b).

Indeed, Acemoglu and Autor (2011) illustrate a model where technical change does
not raise wages of all workers and where there can be substitution between human
labor and machines. Models with even darker implications for mankind are found in
Sachs et al. (2015), Benzell et al. (2015), and Sachs and Kotlikoff(2012). With some
variations, these models result in lower welfare for future generations due to human
and robot substitution. Even though robot innovations can be welfare-enhancing in
the long run under some parameterization, the short-run effects tend to be bad for
humans. The wage share of income is declining in these models, consistent with new
findings in the literature.

3.4 Humans and robots...or humans vs robots

We have discussed lessons from the academic literature concerning the effect of
technology on employment as well some features of structural change in the Swedish
economy, drawing on economic history and recent economic research. Arguably, con-
sensus among economists is that as machines destroy jobs, new ones will be created
especially in the service sector. The implication of this view is that things are normal
and no special action is required from firms or governments. Bessen(2015, page 3)
captures this view in the following excerpt:

New technology will surely take over more tasks that humans perform, but many human
qualities will remain important in global commerce. Although computers can pick stock
portfolios, financial advisors provide reassurance when markets are down. Although com-
puters can recommend which products to buy, salespeople understand consumer needs
and inspire confidence that unforeseen contingencies will be handled fairly. Although
computers can make accurate medical prognoses, they dont yet have the bedside manner
to guide patients through difficult medical choices. And computer scientists dont foresee
computers acquiring such capabilities anytime soon.

121
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

In this section we summarize recent advances in robotics and knowledge-based systems,

drawing especially on Ford (2015a), and we take issue with the view summarized by
Bessen. Already today most of the services that are claimed to be beyond technology
in the above paragraph are already possible or on the way to being implemented.
The speed at which this occurs may indeed hinge more on other aspects than on tech-
nological impediments.

3.4.1 Humans make mistakes and some may lack social skills...
In Thinking Fast, Thinking Slow, Daniel Kahneman discusses many examples of
experts poor assessments. This is especially true in the business of forecasting, where
the anecdotal evidence of bad performance is rather overwhelming and the culprits
easy to find by comparing forecasts to outcomes. The internet is awash with quotes
about the future of computers and the internet, many of which have been spectacu-
larly wrong. But even experts can be mistaken when the data is hard to interpret. For
example, then Chairman of the Federal Reserve Ben Bernanke initially diagnosed the
subprime meltdown in the US as being too small to have an impact on the rest of the
economy.

Top experts in their fields are of course not alone in getting their assessments wrong,
the risk of overconfidence in predictions is a very human trait. As emphasized by
Kahneman, humans tend to be bad at some things they think they excel at:
In finance, actively managed funds rarely do better than index funds, sometimes
worse.
Wine experts can be worse at predicting good vintages than formulas that use
weather, soil, and other factors as inputs.
Risk of infant mortality is higher based on clinical assessment than based on
simple formulas for heart rate, respiration, and reflexes.

As the examples illustrate, the value of expert assessments is both an amusing and
serious subject. It is fun when experts get it wrong but alas the matters at hand are
serious, sometimes concerning the health of the economy, sometimes the health of
anewborn.

Experts may have a vested interest in highlighting and exaggerating their own impor-
tance. In recruitment, there is evidence of bias in that employers tend to hire those
similar to themselves. So far, it has been difficult for non-experts to question the experts,
but the ubiquity of the internet is beginning to correct this imbalance. For example,
for health-related issues it is easy for to search for others with similar symptoms and
assess the quality of advice.

Health care is sometimes used as an example of where human contact is desirable

with doctors, psychiatrists, etc. There is a perception that people do not want to
interact with a computer concerning highly emotional issues. After all, computers
cannot feel empathy and even if they are programmed to simulate it, people may
object, knowing that the feelings are not genuine. There may be some truth to this
objection, but it is not self-evident that these obstacles will prevent the takeover of
certain functions by software. When given a choice of interacting with a computer
toget an immediate response or waiting a long time to see a human doctor, some
people may prefer the immediate response, especially if the software has a good track
record. What will people choose: an immediate response from Dr. Apple, Dr. Google
or Dr. IBM or having to wait an unspecified time for a human doctor? Being con-

122
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

fronted with this choice may come sooner than we think; see Crow (2015). Moreover,
even doctors who are good at clinical assessments may not necessarily be good at
interacting with patients. The point is that it is difficult to predict how far technology
will make inroads in these areas as well, especially since values may change. Younger
generations are more used to interacting with software for other tasks, after all, so
why not for health care too?

What does this imply for the discussion of tasks that may be taken over by software?
The main takeaway is that software may increasingly be better than humans at finding
patterns in data and presenting them in an objective way, not clouded by ambition or
embarrassed by a change of mind on a topic. Indeed, the whole issue of automation of
knowledge-based systems that can take over non-manual jobs concerns this question
(more on this below). How far such expert systems will go and how much they will
replace human experts is impossible to know, but may depend on some identifiable
factors that we will discuss further below (Section 3.4.3). Those who argue that many
tasks are intrinsically better done by humans also need to take into account that
humans are not always as good as claimed. Indeed, Susskind and Susskind (2015)
argue this point and posit that the professions, as we see them today, will gradually
disappear and be largely replaced by technology. It is likely that developments are
moving some way in this direction, but it is a mistake not to consider the inertia of
institutions as well as the crucial role of tax on labor, a point we return to at the end
of Chapter 5.

3.4.2 Role of robots on the labor market today

Machines and robots are now migrating from factory floors to logistics centers and
offices and into knowledge management tasks and journalism. The ongoing changes
are very broad and the last couple of years have shown tremendous advances.

Ford (2015a) discusses these changes and makes a compelling argument that the speed
of adoption and its effects on the labor market will be quite significant. The combina-
tion of emerging standards for software code as well as broad applications make it pos-
sible to explore automation in a wide variety of areas previously largely unaffected.

Automation of low-skilled jobs

As discussed above, employment in manufacturing has declined and new jobs have
been created in the service sector, but automation is beginning to make further inroads
into simple service sector jobs. Frey and Osborne (2013) show that many tasks in the
service sector are subject to automation if they are routine and do not require much
in the way of personal or social skills.

How far this development has progressed may be a surprise:

Semi-autonomous warehouses where many if not most tasks are done by robots
Restaurants can be automated
Next level of automation in agriculture
Self-driving trucks are now on the roads in some US states

The examples in Ford (2015a) show how the technical possibilities are rapidly
expanding beyond the most well-known case of self-driving vehicles. Costs are
decreasing in several areas and the drive for profits is inducing further shifts away
from labor to machines. For example, logistic centers and warehouses are designed

123
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

with robot automation as an integral part of design, laying out barcodes for the
machines to follow, somewhat akin to building roads that allow cars to travel
efficiently. Warehouse robots made by companies like Industrial Perception are
improving all the time, with speed, dexterity, and object recognition.

Jobs in fast-food restaurants that have long been the province of low-skilled workers
are now subject to automation. Momentum Machines have designed machines that can
make 360 hamburgers per hour, exactly to customer specifications; the Kura Sushi
restaurant has pioneered automation, including all elements from making the sushi
and putting it on a conveyor belt to automated billing. Monitoring by management
isdone off-site.

In agriculture, automation has already had tremendous impact in developed countries

but has so far left some manual labor for tasks such as picking grapes, strawberries,
almonds, and oranges. Not only are these tasks now being automated but new tech-
niques are making inroads towards improving the entire agricultural process, ena-
bling control of fertilizers, pesticides, and other elements affecting yield, a topic we
return to in Chapter 4 in the context of regulatory obstacles. The Economist (2014c,
d) gives further examples of how big data and technology can be used to improve
crop yield and increase farm productivity, although the methods remain controversial.

Automation of high-skilled jobs

During the last few years, we have seen advancements in hardware and programming
that enable robots and computers to perform jobs that are typically deemed as requiring
high skills and/or a college degree. Indeed, the incentive to replace high-wage earners
might be as strong if not more so in lieu of larger potential cost savings and rigid
labor markets, notably in Europe. Moreover, in the examples mentioned below, anec-
dotal evidence indicates improved productivity.

A starting point, at least a symbolic one, for the vast potential of computers to also
perform highly skilled work originates with two well-known competitions of human
vs. machine. First, the win of IBMs Deep Blue against then-reigning chess world
champion Gary Kasparov in 1997; and second, the win of IBMs Watson in Jeopardy
against the reigning champions in 2011. Especially the latter became a powerful but
somewhat overused symbol of what computers are capable of. After all, great poten-
tial to solve such problems is only part of the human skill set in knowledge-based
organizations. But as the book by Ford(2015a) points out, this is beginning to change:
Automated newspaper reports from sports events and releases of economic data.
Automated knowledge work, and including elements of in-house work and
crowdsourcing.
Automated data modeling work, finding good fit to empirical data in science,
economics or any other area.
Legal briefs that scan enormous amounts of legal documents to find precedents
orother supporting material for cases.
Software robots scan financial markets for investment opportunities and eliminate
the need for trading pits.
Automatic grading of school essays.
Automated music programs that generate quality high enough to be played
inconcerts.

124
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The current state of computer writing may not (yet) win stylistic prizes, but can
produce accurate and grammatically correct reports; they are also faster and already
difficult to distinguish from human texts, see Box 3.1. Indeed, computer-generated
texts may in some instances be better and contain fewer grammatical errors. For
example, Quill, a new program from Narrative Science, can produce all sorts of reports,
including business and other analysis, and argues that the results are indistinguishable
from those written by the human hand. Automated Insights can generate text from
spreadsheets. One study that uses a survey to compare the perceptionof computer-
generated articles with human texts finds the former a bit boring but also largely
indiscernible from texts written by journalists; see Clerwall (2014). In a contest between
an experienced journalist and a machine, the latter was much faster but with less style
for now; see Vanek Smith (2015). The Associated Press is already using software to
generate more than 3,000 financial reports per quarter, see Mullin (2015), and auto-
mated reports are used in banks such as Credit Suisse; see Yang (2015). Some books
are also software-generated, for example, the Insead marketing professor Philip Parker
has by his own estimation more than one million such books, of which more than
100,000 are available on Amazon, see Bosker(2013).

Box 3.1 Guess the computer-generated text

Example 1
Things looked bleak for the Angels when they trailed by two runs in the ninth inning, but
Los Angeles recovered thanks to a key single from Vladimir Guerrero to pull out a 7-6 victory
over the Boston Red Sox at Fenway Park on Sunday.

Example 2
The University of Michigan baseball team used a four-run fifth inning to salvage the final
game in its three-game weekend series with Iowa, winning 7-5 on Saturday afternoon
(April24) at the Wilpon Baseball Complex, home of historic Ray Fisher Stadium.

Source: Podolny (2015).104

Ford (2015a) provides other examples that illustrate how the possibilities have been
extended in many areas. Legal analysis sometimes requires huge numbers of documents
to be analyzed. With e-discovery software, it is possible to replace large numbers of
(entry-level) lawyers or paralegals and indeed perform work that was previously pro-
hibitively costly or time-consuming. Markoff (2011) illustrates this by comparing a
lawsuit in 1978 that involved hordes of lawyers who examined six million documents
at the cost of $2.2 million to the capabilities and costs of software available today.
The e-discovery software, for example, has been used to scan 1.5 million documents
at the cost of $100,000 in a very short time, probably less likely to miss relevant
information than humans attempting the same task.

Missing relevant information is one potential human weakness, biases another. These
are well documented in the literature; see, for example, Kahneman. In recruitment,
sometimes employers hire people who are most like themselves instead of the appli-
cant who is actually most suitable for the team. Here, software can suggest matches
that might not otherwise take place and potentially reduce biases, a development also
occurring with online dating. Recruitment is an area where there are many strong
views and the software will have to prove itself. But LinkedIn is already becoming
an often used tool; other software, such as Gild, Entelo, Textio Doxa, and Gapjumpers

104
The first example is written by a computer.

125
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

are providing digital hiring services; see Cain Miller(2015). If the software becomes
widely used, hiring may become cheaper and less prone to biases and may lessen
the role of human resources. A pertinent ethical question in this context is how much
humans should override the machine, especially with decisions concerning life-changing
events; see Lohr (2015d).

It may be instructive to compare the potential effect of digitalization on lawyers to

that of economists (disclaimer: the author is an economist). Basic legal research can
now be done by computers, thus potentially displacing junior to mid-level lawyers;
for economists, technology has instead been complementary: allowing researchers to
build mathematical models and run estimations that were simply not possible before.
Indeed, advances in computers have likely spurred the degree to which economics has
become a more quantitative subject.

Changes are also occurring in areas where creativity is important, for example,
through automated music generation. Software has also produced paintings. Digital-
ization is thus entering into areas that are also considered art. It remains to be seen
how much progress will be made but the results so far are astounding. The notion
that a computer could come up with a symphony, such as the Iamus program, which
can be played at a concert is extraordinary.

It is fair to say that the extent to which computers are actually creative is controver-
sial. But in the end, what matters for the impact on the economy is if output is treated
in a similar way and that remains to be seen. The notion that creative outcomes
can be coded is not as bizarre as it might sound at first. Computers use random-
ization and mathematical routines to come up with answers. Good randomization
combined with sufficiently clever rules may well be indistinguishable from human
creativity. Nobel Prize-winning physicist Richard Feynman sometimes described his
theorizing in this way: ... for Feynman, the essence of the scientific imagination was
a powerful and almost painful rule. What scientists create must match reality. It must
match what is already known. Scientific creativity is imagination in a straitjacket, as
described in Gleick (1993).105

Overall, the advancements of clever software into cognitive and creative tasks can
make a lot of entry-level and mid-level analysts, such as journalists and writers, super
fluous; the computers can collect and amass the data, and increasingly, they can also
analyze the data. The program Eureqa has been demonstrated to be able to use math-
ematical techniques to find physical laws and is know available in the cloud for any
kind of model analysis that tries to fit a model to data. Sometime in the future these
kinds of programs will be competing against PhD researchers. Since researchers tend
to specialize in some particular area, the software may be more open to trying models
and approaches that researchers may not know about. When there are fewer low-
to-mid-level analysts, potentially there will also be a correspondingly lesser need for
senior management.

One hurdle for software is the mistrust of the black box nature of some recommen-
dations. People want to understand why a model produces a particular result; a pure
correlation based forecast or recommendation is typically not held in high esteem.

105
Economists sometimes use the concept of observational equivalence to denote situations when the data is congruent
with several different explanations of what drives the result. If an output generated by a computer a book, a painting,
or music is treated in the economy as if were a creative piece of art, then its impact will be similar to real art and its
origin will not matter. But the laws of economics still hold: if a computer generates large numbers of creative outputs,
these will become less valuable due to effects similar to inflation.

126
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Indeed, without understanding why a model produces a particular result, people

(managers, policymakers, doctors etc.) tend not to trust the result, quite regardless of
past forecasting prowess. This is why, for example, automated forecasting routines in
institutions are typically complemented with anecdotes and finding the right narrative.

Distrust of the quality of software recommendations may be lessened, or even over-

come, by showing not only the conclusion but also the underlying reasoning. Hosea
(2015) makes this point in the context of medical diagnosis but the arguments apply
much broadly. IBMs WatsonPaths can do this by showing the path the computer fol-
lowed to arrive at the result. The experience of WatsonPaths in health underscores
this. For example, in a cancer diagnosis, if the software can list the research papers
and statistics used to generate the recommended option, the physician can then decide
whether or not to trust the results, which is an example of complementarity between
human and machine. Indeed, in some instances, the software might do better. Ford
(2015a) documents cases where the right diagnoses baffled the physicians but were
resolved with the software. When the software is right most of the time and can pro-
duce coherent reasoning behind its answers, it can take over quite a lot of sophisti-
cated tasks, including forecasting and business analysis.

There are at least two important takeaways from these examples and anecdotes that
matter for their impact on the labor market and the economy.
There are many changes occurring at the same time in different sectors that can
replace labor: all from making hamburgers and sushi to writing business reports.
The drive to improve existing goods and services, or invent new ones, will lead soft-
ware programmers into new territory, including areas that have so far remained
less affected.

Polarization and skills

One important question is whether the coming decades will see more complementarity
or more substitution between humans and machines. Brynjolfsson and McAfee (2014),
and Cowen (2015) and Ford (2015a), argue that the skills needed to thrive in the future
labor market will lead to even further wage polarization. Cowen (2015) poses the
following question:

Are your skills a complement to the skills of the computer, or is the computer doing better
without you? Worst of all, are you competing against the computer?

It is likely that having some digital skills will lead to better opportunities in the labor
market. By contrast, for someone with little understanding of how computers work,
the risks of worse wage prospects or unemployment are evident. Clearly, many jobs
do not require programming skills and technology is becoming simple to use but then
the complementarity between human and machine is low. Cowen(2015) also argues
that wage polarization in the US would have been more severe if protected jobs in the
public sector are discounted. In Chapter 5 we will explore education and digital skills
in more detail.

Ripple effects of automation

When one type of job is automated, it also affects jobs that are, in an economic sense,
nearby. For example, it is obvious that self-driving vehicles can ultimately make taxi

127
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

and truck drivers superfluous. Moreover, ripple effects may also be considerable. For
example, Jain et al.(2015) discusses how insurance premiums for autonomous vehi-
cles will decrease as accidents become fewer. This would reduce the need for staff in
insurance companies, coming in addition to further automation of insurance claims
already underway; see the discussion in Section 2.3.6 in Chapter 2 of how the Dutch
insurance company Inshared has eliminated several layers of human interaction through
software automation. Moreover, the need for driving schools will also be fundamentally
eroded. Perhaps even more significantly, hotels and restaurants along major trans-
portations arteries, where truck drivers make stops en route may lose many customers
and some may fail, thus reducing employment opportunities outside of cities. Thus,
self-driving vehicles may indirectly cause disruption to the economic sustainability
ofsparsely populated regions, thus increasing the effects of urbanization.

Other ripple effects may be impinge upon the relationship between junior and senior
staff. If more junior staff are replaced by software or robots, fewer managers are
needed as well. Moreover, the long hours of preparatory work that junior staff perform
in knowledge-based organizations to qualify for senior positions is part of a learning
process to become better at the trade. If the share of this on-the-job training goes down,
then the skills of senior staff may gradually erode, unless compensated with some other
factor. This could well represent a more prevalent form of deskilling of the professions
(lawyers, accountants, doctors, etc.) argued in Susskind and Susskind (2015).

It may also be the case that some skills will no longer be needed and that institutions
will adapt. Evidently, that has already happened across some skilled manual work.
For example, car trouble is diagnosed not by a mechanic poking around under the
hood, but by plugging the car into a computer and so the need for comprehensive
knowledge of the engine is correspondingly less today. Similar skills transformations
may also occur for non-manual workers. The point here is that further digitalization
may have broader effects on the need for skill sets and the circumstances within a
firm. How to deal with these issues may depend on the sector and the company, but
itmay be useful for senior managers to have a view on the skills needed in the organ-
ization and not only the short-run benefits of automation.

3.4.3 Forces that limit automation

The implications from the previous sections constitute a bleak future for employment
but there are several factors that may limit the speed of job displacement and allow
more time for adjustment for workers and for policymakers. Autor (2014) emphasizes
the complementarity between machine and computers in historical data; Frey and
Osborne (2013) and Bakhshi, Frey and Osborne (2015) argue that creative jobs will
have less risk of automation because creativity is different from rules that can be imple
mented by computers. But as we have seen above, software with the appropriate mix
of rules and randomization can appear to be quite creative. And so, if robots are not
to replace all jobs, what are the forces that may prevent this?

The effects of technology on the economy and the labor market will arguably depend
on at least three key factors that may in some instances slow the technological feasi-
bility of automation:
The cost of automation compared to human labor
When peoples preferences favor interacting with another person rather than
amachine
Existing and future regulation

128
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Automation still involves high costs in many areas, compared with human labor
With the basis of Moores Law (discussed in Chapter 1), there is a general presump-
tion that the costs of computing speed are going down as chips are becoming ever
faster. If Moores Law continues to hold, however, it does not necessarily imply that
costs of automation will decline at the same rate. For example, the first self-driving
vehicles used state-of-the-art electronics, sensors, and computers that were quite
costly and could quite easily be worth more than the rest of the car. In the early
stages, the Google car is reported to have included $75,000 worth of electronics, but
Ionut Budisteanu, a Romanian youth, was able to design a self-driving vehicle at a
fraction of that cost using more off the shelf electronic components and clever pro-
gramming, see Foy (2014), a point echoed by Ragunathan Rajkumar, an expert; see
Waters and Sharman(2015). Major car manufacturers can do the same and push
down the costs of autonomous vehicles, thus making them increasingly attractive.
Butthis will take time, especially considering regulation and safety issues.

Thus, the process of reducing costs may be slow in many areas, as exemplified by
theauto industry. New technology eventually trickles down to mid-range cars but
theprocess takes time and tweaking. Scale of operations may help to reduce marginal
costs, a path further fueled by strong competition.

The scope for cutting costs with digital technology is now being explored in broad
sectors of the economy by new and existing firms. The sharing economy exemplifies
the drive to find new markets and reduce costs in existing ones with digital platforms
and tools. But are costs coming down fast enough so that we will see robot chefs in
our kitchens anytime soon? Moley Robotics is developing such a robot that can learn
and use thousands of different recipes and is expected to be available commercially in
2017 at a price tag of about $15,000; see the Economist (2015e). The safety require-
ments imposed on such a robot are likely to be significant, which will contribute to
high costs in the foreseeable future. Some people who can afford them might buy
them, but in the meantime most everyone else will just have to put up with using the
numerous semi-autonomous kitchen appliances already available. As discussed by
Autor and Dorn (2013), relative costs matter. Overall, the cost of factors other than
computer processing speed may put a significant brake on the speed of automating
tasks in industry and in homes.

Demand for human interaction likely to remain in some areas

Autor and Dorn (2013) discuss how consumer preferences may impact the labor market,
notably if consumption has close complements or substitutes. Demand is also about
preferences and these may be quite important for the jobs that may be automated.
People who can afford a $15,000 robot chef may still prefer to go without one: per-
haps they like cooking or prefer to eat out or order in. Culinary activities exemplify
an area with strong traditions and cultural norms that may supersede cost benefits
from automation or at least act as a strong counterweight.

Indeed, preference for human contact may in some instances be an impediment to

automating many service-sector jobs. How people react to buying a service that is
automated versus provided by a person may depend on the context, culture, and
other factors. Most people have grudgingly come to accept that customer service
viatelephone requires navigating through several layers of menus, all to weed out as
many simple questions as possible, but with the result that reaching a person on the
other end is cumbersome and sometimes not even possible. Moreover, the person on
the other end may physically be in another country and time zone.

129
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The logic of cutting costs dictates reducing staff, but the logic of customer loyalty and
brand reputation may go the other way. The technical possibility of interacting with dig-
ital media will increase in the future but raises further issues. What do people value and
where do we draw the line for when we want to interact with a person? If indeed there
is such a line, will it shift over time as attitudes change? Might there even be a backlash
against automation? In Box 3.2, we illustrate some of these points with examples.

Box 3.2. Examples of demand-related issues with further automation

Education
Education has a low probability of automation according to Osborne and Frey (2013). Yet mas-
sive online courses (MOOCs), education apps, and other software are changing the funda
mentals in education. When children are learning the to read, write, and do arithmetic, they
can interact with software, such as EdQu, Razkids, and Mathletics, with high pedagogical con-
tent; for higher education, the best teachers can design and implement courses with poten-
tially global reach. Thus, it is in theory possible to reduce the number of teachers and reward
the superstars at the best universities.
In such a system, the software would do most of the work and the teacher might help primarily
if someone gets stuck. The question is not if it can be done clearly it can. In Korea, students
are already learning English with the help of robots. Rather, the question is: do people want
this or not? Some further automation of grading and improving mathematics and science edu-
cation are likely inevitable, but how far is an open question: our norms and cultural values may
vastly limit the scope for schools with few teachers.

Transportation
Transportation is among the sectors with high risk of automation. Self-driving vehicles can
clearly challenge jobs for drivers of taxis, trucks, trains, airplanes, boats, and buses. Self-driv-
ing trucks are already on the roads in California and Nevada, see Kessler (2015), and it is likely
only a matter of time before we see autonomous trucks and other cargo carriers.
Here, the demand for interaction with the counterparts may not be enough to save such jobs.
People will probably accept self-driving taxis, but when it comes to self-driving trains, the world
is more divided: in many countries, trains are still driven manually, but when it comes to self-
driving trains, the world is experiencing a trend of introducing more of them. The majority of
trains are still driven manually but some trains in, for example, Tokyo, Seoul, Singapore, Paris,
and Copenhagen have self-driven trains servicing the metros. And nowhere are completely
automated airplane pilots used for passenger transport. Despite the possibility of pilot error,
the world of travel does not yet seem prepared to accept pilotless planes.

Fashion models, etc

People who work as models have a high probability of seeing their jobs automated according
to Frey and Osborne (2013). In theory, current technology allows computer-generated people
to appear in advertising both in video or in still pictures; even catwalks could probably be auto-
mated with current technology or at least soon in the future. But even though we can make
any kind of computer generated object, the modeling profession may well be safe from digitali-
zation. Simply, there is demand for real men and women as models, even though some of them
are highly paid. The ultimate object of models is to generate demand and as long as models do
this, they may be safe.

Health workers
In some instances, it is very hard to envisage machines ever substituting for humans, such as
for childbirth and such. But it may be possible in other areas, clinical diagnosis for example,
especially if the cost of using the automated system is lower and the waiting time is shorter.
Doctors are no different from other professions in that technical skill, in surgery for example,
need not be accompanied by superb social skills. It is conceivable that software could be pro-
grammed to display more empathy and sensitivity than a stressed and tactless surgeon.
Surgeons already guide robots in some operations that require great precision.

130
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Regulation can slow digitalization

Regulation affects all areas of business and, of course, aspects of digitalization as well.
As digital firms try to disrupt existing businesses, there is an ongoing concern as to
what extent regulation applies (see also Section 3.5 on the sharing economy).

Existing regulation is there to protect workers and consumers, but not all regulation is
necessary and some may be outdated. In New York, questions have been raised about
Airbnb and the extent to which Airbnb hosts should comply with regulations that
apply to hotels, as they may be in competition for some of the same customers. Uber
is in conflict with taxi drivers and regulation in many countries. Overall, existing reg-
ulation and future regulation can have major impact on how digitalization affects the
labor market and the economy (more on this in Chapter 4).

Steps towards a more holistic view on the risk of job automation

Frey and Osborne (2013) and Heyman et al. (2015a), discussed above, estimate the
risk of automation based on the nature of the tasks, basically the input factors in pro-
duction. These provide a perspective on the challenges for the economy but, as noted
above, tell only part of the story.

In this section, we illustrate another perspective based on the tasks that can already
be performed with robots and software. Frey and Osborne (2013) focus on the poten-
tial for automation of input tasks in production while we argue that it is important
to also consider the output of production and other economic driving forces. For
example, journalists are at low risk of automation, just below one-fifth, according
to Heyman et al. (2015a). But we have seen software already exists that automates
sports writing and business news. Thus, considering only the potential for job auto-
mation may lead the analysis astray and also be unhelpful when designing a policy
response. Ultimately, the possibility to automate a task performed by a certain occu-
pation is only one aspect of whether the good or service will be automated.

In Figure 3.7, we show the change in some selected occupations in Sweden based on
the reading of the material in Section 4.2. We see that many of the occupations where
software has made significant inroads have declined substantially over the period.
Granted, the size of each of these occupations is relatively small as a share of total
employment. For example, librarians declined by about 50 percent from 2001 to
2013 but constitute only 0.1 percent of employment. But many of these small occu-
pations add up to numbers that matter to the whole labor market. According to
Flster(2015), somewhere between 915 percent of all jobs disappeared due to digi-
talization during 20062011.

The bars at the far right display two occupations that deserve some explanation.
First, fashion models have very high probability of automation, a whisker below 100
percent according to both Frey and Osborne (2013) and Heyman et al. (2015a). Yet
fashion modeling jobs have increased over this time period. Some possible explanations
of this development illustrate the limitations of focusing only on the technological fea-
sibility of automation. Fashion models might not be automated because of demand
issues, for example brand recognition and such. If a model increases sales more than
a computer generated image, the former will prevail: after all, the choice is about
profitability.

Second, lawyers have increased considerably in Sweden. Flster (2015) discusses

one possible explanation, arguing that complexity is increasing and thus justifies the
need for more lawyers. Complexity stems from regulation and the changing environ

131
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

ment, all which are exogenous (excluded) from probability assessments in Frey and
Osborne (2013). One notable example of greater complexity is financial regulation fol-
lowing the strengthening of oversight and implementation after the financial crisis.
But greater complexity can also stem from changes in the tax code. Also, higher employ
ment numbers for lawyers might be related to anticipated changes in regulation and
lobbying, for example with increases in staff working with compliance with new
financial rules and risk assessments.

Figure 3.7. Selected occupations at risk of automation, percentage change, 20012013

80%

60%

Legal professionals

40%

Fashion & other models

20%
Adm. Associate

0%
Archivists etc.

Other office clerks

-20% Mail carriers etc.

Cashiers etc.

Numerical clerks

Source: Statistics Sweden.

-40%

Library & filing clerks

-60%

-80%
0,1% 1,1% 0,5% 0,5% 2,0% 0,2% 1,8% 0,0% 0,4%
Note: The occupations have been selected as prone to digitalization on the basis of the literature review in
Section 3.2.6 in this chapter. For each occupation, the percentage change in the number employed is computed
over the whole time period. The total employment share of each occupation is shown at the bottom. Own calculations.

Small countries may have slower rate of automation for knowledge-based work
Automation of knowledge-based work, like that of lawyers, business analysts, and
journalists may be much slower in small countries, such as Sweden, as compared to
the US. The scale of the US economy means that investments in automation software
can bring large rewards compared to input costs. The same may not hold to the same
extent for small countries. For example, the ediscovery software for legal research is
designed for the US market but cannot be used as easily for other countries with dif-
ferent legal systems and traditions. The software either has to be adapted or made
from scratch, which may be costlier when the scale is small. Automated writing is
another example where English is dominant and the quality of automated English
texts is likely to be superior to most other languages, at least for a long while.

None of the small-country issues/scale are insurmountable obstacles to automation.

Once a platform has been established in the US, the steps to translate into other areas
are smaller and the platform may come in with a bang. The Swedish social networking

132
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

site Shortcut was popular, but was nevertheless eclipsed by Facebook. It may well be
that methods to adapt language will also improve quickly, similar to how GPS navi-
gation is becoming ubiquitous even in areas with low population density. But idiosyn-
cratic rules and local government regulation are still likely to slow down and reduce
some competition from international platforms (see more on this in Chapter4).
However, to conclude that small countries are safe from automation based on big
platforms would be a mistake. A more proactive view would be that there may be
opportunities for entrepreneurs to build clever ways to account for local idiosyncrasies,
in Europe and elsewhere, before the big platforms come knocking.

3.5 Changing nature of work and its implications

Throughout this report, we have highlighted the importance of institutions for how
well the economy copes with structural change. Few institutions are as important
as those related to the labor market. In Sweden, the model with strong insiders with
close to permanent employment contracts is still persistent but has also evolved over
time. Figure 3.8 shows that most private sector employment is covered by collective
bargaining agreements in the labor market. It is also evident that this share has
declined somewhat during the last couple of years. The share of self-employment in
Sweden, with its own set of rules, is largely unchanged at around ten percent. Sim-
ilarly, there are no changes for the coverage of those employed in the public sector
(not shown), who are all covered by collective bargaining agreements. The share of
workers on temporary contracts has fluctuated somewhat but is somewhat higher
today, about one percentage point, compared with 2005.

Much of the structure of the labor market focuses on providing a social safety net for
the insiders with jobs. Also, collective agreements stipulate agreements over vacation,
pensions, unemployment insurance, and many include top-ups to parental leave. To
be entitled to unemployment benefits, for example, people have to have been in work
for six out of the last twelve months.

Figure 3.8. Types of employment contracts in the Swedish labor market,

percent of total
95 25

All (LHS)

90 20

Private sector (LHS)

85 15
Sources: SCB (LNS), Kjellberg (2015).

Self-employed (RHS)
80 10

75 5
2007 2008 2009 2010 2011 2012 2013

133
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Sweden has one of the most pronounced dual labor markets in the OECD
In comparisons with other OECD countries, Sweden is the epitome of a rigid labor
market, featuring strong rights for insiders on permanent contracts; see OECD (2015b).
There are, however, possibilities for unions and employers organizations to bypass
first in-last-out clauses in negotiations, which reduces some of the inflexibility. But
when it comes to temporary workers, the Swedish system offers among the weakest
protections in the OECD. Taken as a whole, this means that Sweden has one of the
most pronounced dual labor markets among OECD countries; see Cahuc (2010,
pages 150153).

Some recent developments are noteworthy. As shown in Figure 3.9, union membership
has declined fairly dramatically throughout this time period driven by both non-manual
and manual workers up until 2006. After the center-right government came into office
in 2006, differentiated fees for unemployment insurance were put into force in 2007,
after which the share of union membership among manual workers continued to
steadily decline while non-manual worker membership resurged somewhat. Overall,
membership went down by over ten percentage points over the period 20012014.

Figure 3.9. Union density in decline 20012014

Manual workers
85
All employees
Non-manual workers

80

75

70
Source: Kjellberg (2015).

65

60
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

The trend in the Swedish labor market has been towards more flexibility with tem-
porary workers, reinforcing the dual labor market with severe job insecurity for out-
siders and strong protection for insiders. As digitalization affects the way we work, it
may also alter the structure of employment in more fundamental ways by potentially
increasing the share of self-employment, the topic of the next section.

Digital disruption to the labor market model?

The shift to digital could fundamentally change the labor market. The Economist
illustrated this possibility with a front-cover image of office workers being thrown
asunder by a tornado.106 The implications are two-fold:
Workers are being replaced by machines
Labor market agreements may be subject to radical overhaul as labor is being
sliced, diced, and atomized via digital platforms

106
June 18, 2014.

134
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

There is no doubt that digitalization and especially the sharing economy discussed
in Chapter 2 may have major impacts on the organization of labor, but it is also
likely that the effect will depend on the choices of political institutions. In Box 3.3
below, we discuss some scenarios for how the sharing economy might affect the labor
market.

Digital platforms that connect freelancers with consumers have built their business
model on contracting work and typically do not prove employment status. By acting
as a platform for providing contracts without the other commitments associated
with being an employer, firms within the sharing economy can keep their costs much
lower, and thus be very competitive against firms that pay social security, pensions,
on-the-job training, and other employee benefits. The competitiveness is magni-
fied many times over by the network effect: only comparatively few employees are
required to run a platform that serves millions and thus the cost per unit of service
can be very low.

The possibility to use technology to outsource both simple work and work that
requires high skills can have considerable impact on wages and hiring decisions. Why
take on permanent staff when more tasks can be bought on the market as needed?
Potentially, this fundamentally shifts the raison dtre for the firm that won Robert
Coase the Nobel Prize (see Chapter 2). Jobs that can be subdivided into separate
parts can then either be outsourced or automated. So far, this has so far had the
greatest impact on low- and mid-level jobs but highly skill professionals may also be
affected, such as people working in law firms, management consultancies, or fund
managers and other services in the financial sector. Their reputations and standards
have enabled them to charge high fees but that may change in the future. Indeed, the
incentive to outsource tasks to algorithms and to sharing platforms or cloud services
should be the strongest where the biggest cost savings are to be made.

While the sharing economy provides low costs and high efficiency and enables people
to work part time with great flexibility, it is also increasingly under assault on legal
and moral grounds, as well as for exacerbating inequality. While these issues have
been prevalent throughout periods of rapid technological change, digitalization may
vastly increase the effects over a short period of time, making it more difficult for
people and institutions to adapt.

Legal challenges to the sharing economy

The arrangement as contract worker vs. employee status has recently been challenged
in a California court that decided in favor of the plaintiff against Uber, see for example
Bradshaw(2015), which ruled that the freelancer should be considered an employee.
Should this ruling have wide applicability, the whole sharing economy model may
be severely affected. Uber is being challenged not only in US but across the world,
notably in France with taxi drivers staging protests in Paris and with negative rulings
from the Constitutional Court, see Reuters (2015), and recent arrests of senior Uber
representatives; in Germany, a nationwide ban was subsequently overturned and
again challenged; see Vasagar et al. (2014). A search of the internet for Uber and
legal will return countless examples of challenges in countries all over the world.

Moral and ethical challenges to the sharing economy

While the challenges to Uber and other sharing economy platforms are trave-
ling their course in the courts, the arguments often used against them are moral and
ethical. The labor markets in the US and in other countries have increased in polar-

135
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

ization in recent years; in the US in particular, labor market polarization has also
resulted in wage polarization. Jobs in the service sector have increased, but mainly
those with low pay, see Section 3.3 in this chapter.

The research on job polarization in the US has received a great deal of media coverage.
Many news stories discuss the difficulties for low-skilled employees to survive on
their wages; see for example Porter (2015a), Leubsdorf and Hilsenrath (2015), and
Keen (2015). McDonalds workers have been protesting in several cities. The work
of Ashenfelter (2012) documenting McDonalds pay across countries, found that real
wages declined in the US during 20002007. Jacobs et al. (2015) document that some
low-paid workers in the US need public assistance in the form of food stamps despite
working full-time, and calculate that the cost to the US taxpayers amounts to $153
billion per year. They contend that this amount is a subsidy to firms with low wages.
Kantor(2014) discusses how some families are unable to plan their week of study
or childcare due to erratic schedules, thus making family life more difficult as well
as preventing upskilling and realization of the American dream through hard work
and education. The erratic work schedules are the fruit of software that can optimize
when staff is needed at very short notice, depending on factors that affect demand for
fast food, cappuccinos, or other services in the low-wage sector.

Should the wage disparities continue to increase in the US and other countries, the
debate about these low paying jobs will increase, as well as the potential for conflict.
The actor Charlie Chaplin in Modern Times captured the zeitgeist of discontent with
the Industrial Revolution with the worker struggling but failing to keep up with the
machines in the automated factory; likewise, the image of struggling dockworkers
waiting in line for day work was memorably captured in On the Waterfront with
Marlon Brando. There is a perception that our times equivalent are the baristas
juggling coffee cups to affluent customers; contractors in the sharing economy waiting
for work online, the dockside being replaced by a computer. Whatever the merits of
the perception, the issue of stagnant wages and increasing inequality are well docu-
mented and the challenge they pose may increase in the digital economy. In Sweden,
the relatively high salary levels at the lower end of the pay scale pose a particular
challenge for the employability of low-skilled workers, for whom entry into the labor
market poses a rather high hurdle to overcome. This difficulty is likely to be accen-
tuated in the years ahead due to the large inflow of immigrants seeking asylum in
Sweden and other affluent countries.

Risk sharing in the labor market

In Sweden, labor market institutions are centered on workers with indefinite employ-
ment contracts, so-called insiders. The sharing economy may shift the balance between
insiders and outsiders. Being self-employed typically involves a higher level of risk
than being employed, notably being responsible for finding business opportunities as
well as requiring considerable administrative work for taxes and such. Some people
want the degree of autonomy associated with being self-employed and are prepared
to accept the higher level of risk and administrative burden. In Sweden, the share of
self-employed workers has been fairly constant at around 10 percent of the labor
force, see Figure 3.8.

To be sure, even in Sweden, a country with fairly strong employment legislation, workers
are laid off in the private sector all the time. But, by and large, self-employed people
tend to have higher levels of risk than employees. Should the number of self-employed
workers increase significantly, they will become more important to politicians. This
may result in pressures to change institutions to reduce the level of risk they are
exposed to. As noted by the Economist(2015b):

136
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

...governments will have to rethink institutions that were designed in an era when
contract employers were a rarity.

Economic constraints to the sharing economy

The sharing economy and hiring freelance work is still subject to some of the con-
straints of traditional economic models. As discussed in Chapter 2, the logic of
Robert Coase has shifted but not disappeared. At some point, the marginal cost of
hiring temporary workers in the digital labor market will be higher than having
thefunction in-house.

What factors might affect the tradeoff between marginal cost and marginal revenue?
At least two issues stand out. For one, firms that rely on their reputation need high
quality staff who enhance the trust in which the firm is held. For example, MyClean
discovered that contract workers got lower ratings than those who were employed;
see the Economist(2015b). With reputation such a potentially fragile thing, firms
must be mindful of quality. Another traditional upside of employment is that firms
also take some responsibility for training employees and for lifelong learning. These
features are important in the knowledge economy but also potentially for lower skilled
workers, for example in the efficient operation of logistics distribution centers to be
able to update and upgrade systems.

A second major issue is the risk of downtime. When tasks are optimized, sliced, and
outsourced as much as possible, small disruptions in one part of the chain can cause
production to grind to a halt altogether. Faced by such a specter, management has to
decide how much to weigh cost reductions from outsourcing against the benefits of
control and reliability of operations.

The risk of downtime may be higher the more interconnected things are, either phys-
ically via global supply-side chains or in digital networks. Small errors in software
or indeed malevolent hacking can spread faster and more widely. For example, with
the Internet of Things, the vulnerabilities to the system may have parallels to banking,
in which one shock can unravel and cause domino effects. Indeed, such fears were
present, but never materialized, concerning the Y2K millennium bug, which may well
be thought of as a historical example of a systemically important event for digital
vulnerability. With the advent of smart cities where cars, subways, energy systems,
elevators, and all other things are connected, small shocks can have big consequences,
especially as the code to optimize functions becomes ever more complicated.

A particular example is the financial risk connected to digital systems, which stem
from payment systems in firms. Administrative work connected to paying salaries,
bills, and other outgoes tends to be repetitive and thus subject to automation; see
Frey and Osborne (2013). But if things break down, wages and bills do not get paid.
Depending on the type of operation, the risks to the business may justify keeping more
functions in-house than pure cost efficiency arguments would entail, exemplifying the
tradeoff between efficiency and risk management. For some businesses, retaining more
functions in-house may be likened to paying an insurance premium in exchange for
lower risk of downtime.

137
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Box 3.3. How much will the sharing economy disrupt the traditional labor model?

Some argue that digitalization will imply the end of existing labor markets as we know them.
In such a scenario, most things are automated and the available jobs are mostly in the service
sector, on digital standby for meager wages; inequality is rampant and unemployment is
involuntary rather than by choice. How far away from such a bleak picture are we?
Some arguments give credence to the idea that we are moving in this direction, but as dis-
cussed above there are forces that slow down the change as well. Overall, however, there
are other forces, such as demography and changing demand, that may be just as important.
The degree of automation will be affected by the possibility of generating profits, and not by
technical feasibility alone. Moreover, the path is not a deterministic law of nature and will be
affected by political choices and their implementation by institutions.
Perhaps at the outset, we might state the obvious possibility that the sharing economy model
could be found incompatible with existing labor laws or other regulations. This is not as
extreme as it may sound. The US court in California that decided an Uber contractor should
be given the status of employee is a case in point; some sharing economy platforms have
already taken steps in this direction, notably Hello Alfred, Instacart, Munchery and Shypall
converted their contractors into employees; see Waters (2015b, c) and the Economist (2015f).
In Sweden, the taxi business is being investigated in a governmental inquiry, which is essen-
tially about Uber.
If the sharing economy model is curtailed, it need not take the form of outright legislation,
but the courts and other government agencies might create expectations of rights for cont-
ractors that increase the potential liability of firms using this labor model. Either by legislation
or through expectations of future legislation, the use of the sharing economy model might
expand only a limited amount. We do not believe this to be the most likely scenario, but it
cannot be ruled out.
Perhaps a more likely scenario is that the world takes steps towards a larger sharing economy
and our institutions adapt as well. How this will unfold will be largely path-dependent and
country-specific.

How far will development towards more self-employment go? As discussed in

Box3.3, the outcomes will depend on choices and how we respond to technology
and changing possibilities in the labor market. What is clear is that there are great
opportunities to improve the functioning of labor markets so that they become
more flexible. There are many groups that would benefit from more flexibility,
not least among them parents with young children, students wanting to supplement
their incomes etc. For firms, the greater flexibility will imply possibilities to provide
more and better services at lower cost. And yet, a substantially increased amount of
self-employment involved in the sharing economy is not a sure thing. While many
economic forces are pulling in this direction, these forces are not the only ones that
matter. The legal battles now underway belie some political choices already on the
table and which will have to be addressed in one way or another.

138
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Box 3.4 Labor market disruptions in developing countries

This discussion above has centered on developed countries in the OECD with a special focus
on the US as the technology leader and with implications for Sweden. But it is worthwhile
to also briefly discuss the effects on other countries, especially those in Asia and parti-
cularly China. China was able to leapfrog several steps in technological development and was
not inhibited by having old systems in place or the need to cater to political forces the way
democratic countries must. As a result, some southeast Asian countries have more modern
capital stock and infrastructure than many OECD countries.
In the last decades of this rapid development, southeast Asian countries have benefited from
work outsourced from OECD countries, all from manufacturing of iPhones to support cen-
ters or software development. With plenty of cheap labor available, outsourcing was attrac-
tive and allowed global companies to keep costs low. The same logic of keeping costs low
may now lead to a wave of reshoring manufacturing of electronics and other products may
return to OECD countries but in the form of jobs for robots instead of people, especially as
labor costs in some of those countries have been rising.
If jobs in manufacturing and textiles are indeed reshored, the upheaval in southeast Asia may
be considerable. With little or no social safety nets, the unskilled workers in factories will find
it hard to find other work. Also, the political processes that may have prevented some of the
outsourcing from OECD countries may even accelerate this process.
There is a risk that reshoring may be much more disruptive for developing countries wit-
hout adequate social safety nets than outsourcing was in the OECD; see for example Ford
(2015b). It is sometimes said of China that it may grow old before it grows rich; reshoring due
to digitalization may intensify this trend. As expressed by Dani Rodrik, the emerging world
may have to cope with premature industrialization or, in the words of Raghuram Rajan with
regard to India, premature non-industrialization, see the Economist (2014e). This captures
the challenge of having a manufacturing base without other functioning features of econo-
mies in the developed world, especially services. Should manufacturing in China and emer-
ging economies become uncompetitive, there are too few other exports to support growth.

3.6 Summary
This chapter discusses how the labor market may be affected by digitalization. We
draw on experience from research, previous periods of rapid structural change, and
a discussion of how robots and software are evolving. An overall reflection is that
several centuries of technological development have not led to mass unemployment.
Another feature is that the latter half of the 20th century has involved complementarity
between human and machine, although there is some evidence that this has changed
in the last decade. Moreover, there is overwhelming evidence of labor market polar-
ization, which has also translated also to wage polarization in the US, but so far not
inSweden.

Overall, the recent changes have also led to higher wage inequality, especially in the
US. Also in Sweden, Gini coefficients have increased somewhat, albeit an increase
from the lowest levels of inequality in the world. There are likely several explanations
for these changes, of which technological change is but one. But digital technology
can create enormous wealth with winner-takes-all features from network effects
and increasing returns to scale. The most pressing issue is not the super-rich, who are
likely to increase in number, but about the wealth and welfare for the other 99 percent
of the population, a topic we return to in Chapter 5.

139
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The question at the core of this chapter is whether things are different this time. We
argue that they are, in the sense that the labor market may be facing decades of difficult
adjustments for large parts of the economy. There are substantial trends in various
occupations that are likely driven by digitalization but no evidence of an avalanche
about to bury jobs and wages. Of the more major changes, the workforce in personal
care and allied occupations (the health sector) stands out with substantial increases in
employment, driven by changing demographics and increasing demand for such ser-
vices. Other professions in Sweden have also increased, such as lawyers, despite the
fact that technical possibilities to automate should be considerable.

A key message of this chapter is thus that automation and digitalization are present
but can be superseded by other economic forces, institutions, and demand responses.
Without potential profits, automation will not occur even in areas where the technical
possibilities allow them to happen. It is a mistake to only view the risk of automation
from a technical feasibility perspective and not consider other mechanisms that may
be equally important.

Digitalization implies a potential for higher efficiency and flexibility in labor markets
at lower costs. Depending on how governments respond, productivity may either con-
tinue at the present lackluster pace or return to previous higher growth paths. Getting
policy wrong at this juncture can have dire consequences for the labor market and for
medium- to long-run welfare. Indeed, considering the magnitude of change underway,
the muddling-through approach may lead to unnecessarily high costs in terms of
unemployment and lower productivity growth. The risk of populist policies that lead
to protectionism would only make matters worse.

In the next chapter, we turn to questions about the role of regulation that are condu-
cive to future productivity growth and job creation.

140
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

4. Regulation and obstacles in the

digital economy
Essentially we want to see how we can regulate these new business models in a way that
would protect consumers and not hinder innovation. Marina Lao, director of FTC policy planning
office in the US, quoted in Financial Times, 2015-05-11.

When there is competition through new technology, we must examine whether the regulation
is still necessary. Daniel Zimmer, chairman of the German Federal Competition Commission, quoted
in Financial Times, 2015-06-02.

4.1 Introduction
Ask any entrepreneur and they are likely to talk about the difficulty of navigating
regulation. Governments regulate a wide variety of businesses but there are consid-
erable differences in regulation between countries and even within countries. Some-
times local government is responsible and sometimes central government. In many
areas, there is also a substantial degree of self-regulation, a matter that is becoming
increasingly important in digital markets.

While regulation is inherently complicated, the underlying rationale is not. From an

economic perspective, the notion of market failure is what motivates regulation.107
The main practical raison dtre is to ensure consumer safety that the medicine we
take has the effect it claims with acceptable risks and side-effects, that our food is
safe to eat and that the elevator does not fall down when we use it.

In addition to safety, regulation serves a multitude of other purposes, such as worker

safety and tax and data collection, ensuring a level playing field in competition, environ
mental protection, etc. Many of those rules serve valuable purposes and may even
contribute to welfare over and above the improvement in safety standards. For example,
effective price competition brings consumers more quality for less money, while environ
mental protection can contribute to mitigating pollution and improving quality of life.

But a lot of regulation is complicated and does not achieve its goals efficiently.108
Moreover, regulation sometimes changes incentives in unpredictable ways that may
not meet the objective of consumer protection and thus may sometimes not fulfill its
original purpose. Moreover, even when the purpose of each area of regulation on its
own has some merit, there may still be considerable issues with the totality of reg-
ulation. Rules tend to accumulate over time as values and demands change, but the
mechanisms for keeping the overall burden of regulation at a reasonable level are
generally weaker than the forces that crave new rules.

Successive governments in Sweden, as in other countries, have often pledged to reduce

the overall regulatory burden. But even with the best of intentions, this goal often
proves elusive when confronted with a cacophony of conflicting demands. The main
explanation is, of course, found with powerful special interest groups that advocate

107
Especially asymmetric information, a situation in which the buyer in a transaction is at a disadvantage, and
externalities, when the actions of one party affect a third party, such as with pollution or traffic congestion.
108
The OECD often divides rules into four different categories: rules that prevent competition; rules that prevent
firms from growing or expanding; rules that generate high costs of compliance, and rules that make adaptation to new
technology more difficult. For a discussion of this in a Swedish context, see Tillvxtanalys (2010).

141
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

increased regulation, sometimes with the noblest of intentions, but sometimes with
the intent of protecting their market power by raising the entry hurdle. Safety con-
cerns can all too easily be misused to prevent competition. There is a considerable
incentive for special interest groups to increase regulation stemming from two inter-
related factors:
Keeping prices and wages high
Excluding competition by raising requirements and/or erecting other entry hurdles

Both factors tend to prevent competition and benefit those firms or individuals that
are already well established in the sector or market.

It should be stressed that not all new rules are bad and the absence of rules would
bring havoc. In fact, quite a lot of rules and their application are crucial to creating
trust in markets. Without trust and standards, much fewer transactions would take
place. Regulation of banks engenders trust in banking, regulation of cars ensures
safety for drivers and pedestrians. The European Single Market has been instrumental
in harmonizing rules, setting standards, and removing obstacles to free trade in goods
and services. The European Single Digital Market likewise has the potential to help
bring the benefits of digitalization to the wider market and consumers. However, the
outcome crucially depends on how the regulation is designed; the details matter.

In this chapter, we focus on the factors that hinder productivity growth discussed
in Chapter 1. Productivity and the amount of work (Chapter 3) together constitute
GDP, thereby forming the foundation for consumption and welfare. Digitalization is
now entering this already complex maze of economic relationships. In this chapter
we also shed light on the strategic challenges faced by firms. In times of technological
quantum leaps, senior management must weigh the risks and benefits of investing in
new ways of doing business, often in the face of strong corporate inertia that protects
existing business lines.

4.2 Legal and regulatory hurdles

4.2.1 How regulated is the economy?
Before we discuss regulatory obstacles for the digital economy, let us briefly review
the general stance on regulatory reform. This topic is broad but nonetheless a useful
starting point to discuss how the digital economy may be seen through the prism of
existing regulation.

Regulatory reform slowing down in OECD economies

OECD (2014a) discusses how reform of product market regulation has slowed down
in OECD countries with respect to previous assessments in 2003 and 2008. In the last
few years, OECD countries have implemented reforms by reducing state ownership,
reducing price controls, and easing administrative burdens for start-ups. Nevertheless,
substantial room for improvement remains, especially in the governance of state-
owned enterprises and concerning regulatory barriers to entry in network industries.

Sweden has the lowest regulation of professional services and retail trade; see OECD
(2014a, pages 7778). Major institutional reforms were introduced in the late 1990s
and until 2008, after which the pace of reform decelerated. In its recent recommen-
dations to Sweden, OECD (2015b) highlights the countrys good performance during
the financial crisis but also identifies some challenges. In the context of regulation,

142
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

one recommendation is to simplify procedures for licenses and permits; see OECD
(2015b, page 11). In their flagship report Doing Business 2015, the World Bank
(2015a, page 4) places Sweden in the top 6th percentile among 189 countries but also
notes few regulatory improvements since the last measurement the previous year.
Similar to the OECD (2015b), the World Bank (2015a) also stresses bureaucracy, for
example concerning zoning restrictions for housing, as a hurdle. This places indirect
obstacles for achieving the benefits of digitalization (see further below) and is also an
issue for the overall functioning of the economy.

In a recent microeconomic assessment of Sweden, World Bank (2015b) describes

restrictive labor laws as an obstacle for economic growth. In addition, firms with
fewer than 250 employees report taxation as a hurdle more often than large firms
do. The lack of appropriate skills in the labor market is another commonly reported
theme across firms of different sizes.

Increasing demands for professional licensing raises costs and makes it harder for
low-income workers to secure employment
Licensing of professions is a broader question than digitalization but it is important
for at least two reasons. First, as discussed in Chapter 3, digitalization will disrupt the
labor market. If displaced workers find it harder to switch jobs into another firm or
sector due to regulation, it is important to assess how reasonable those restrictions
are in the first place. Second, the growth of new jobs may be hampered by unnecessary
licensing requirements.

Professional licensing serves several functions, such as creating a standard that sends
a signal about competence. It can also safeguard consumers and can thus create con-
fidence so that markets can thrive. But professional licensing becomes a regulatory
obstacle when it protects a profession from competition and market entry. Depending
on the profession, the standard may be very high such as for lawyers, or relatively
low, such as for florists.109

If consumer safety is the main concern, a mapping onto licensing requirements would
likely not yield a list that could easily be understood as measures that are necessary
to protect consumers; see Porter (2015b). For example, unlicensed tour guides in
Washington are illegal but still manage to provide a quite adequate service according
to the Economist(2014f). Some have even been subject to court injunctions, which
were later overturned; see the Economist(2014g). A recent report has found that today
almost three out of ten professions in the US require a license, an increase from about
one in twenty in the 1950s; see Kleiner (2015). With some assumptions, the report esti-
mates an implied loss of about 2.8 million jobs. From a survey, Kleiner and Kruger
(2013) estimate that 38 percent of all professions either have to be licensed by the
government now or soon will be and that this results in an average wage premium
compared to unlicensed work of about 18 percent.

The increase in occupational licensing in the US also recently received attention from
policymakers; see Fleming (2015). A recent report highlights that too often the cost
and benefits of licensing are not properly weighed: see White House (2015) and
that the costs of the entry requirements disproportionately fall on weaker groups in
society. The report also sheds light on the arbitrary nature of licensing requirements
and how they are updated.

109
This is not intended to imply that florists do not need skills, only that there are no clear consumer safety issues involved.

143
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

In the EU, Koumenta et al. (2014) on the one hand find that occupational regula-
tions cover about 1024 percent of the labor force, much lower than in the US. On
the other hand, they find evidence that intra-EU migrants are less likely to be in these
professions, which might be indicative of barriers to migration. Free movement of
labor is a pillar of EU membership and the European Commission is doing a great
deal of work and monitoring to improve it. Some professions remain regulated but
many have been opened to EU-wide competition. Some barriers in terms of difficul-
ties in recognizing professional qualifications remain, especially in health and social
care, education and construction; see GHK(2011). For these, and especially for health
care professions, automatic recognition of qualifications might promote better align-
ment between the supply of skilled workers and the likely increase in future demand.

Considerable variations in restrictions to trade services in the OECD

Trade in services is subject to a variety of regulations. Despite harmonization in the
EU, some differences remain. As Figure 4.1a shows, air transport services have the
most remaining restrictions while banking, distribution services, and sound recording
have the least.

Figure 4.1a. Restrictions on traded services in the OECD organized by service, 2014

Italy
3
Spain
UK
2,5 Germany
USA
OECD
2 Sweden

Source: OECD Services Trade Restrictiveness

1,5

Indices database.110

0,5

0
ite e

n
s

on
e

rs

ht

ica g
ns
s

rt
ga
ie
ce

e
ic

nc

im
io
ur

tio
in

in

n
Ro spo
te

ig
ur

tio

si
di
ur
rv

Le
ut
nk

er
ct
i

ra

ot arit
rv

uc
pu

vi
or
ct
Co
se

ne

n
rib

fr
Ba

su
se

le
tr

pi

c o re c
ra
m

ad
gi
rt

Te
un
st

In
ch

ns
Co

tt
n
g
po

En
Di

io

nd
tin

Ar

Co

m
gh
ns

m
u
un

ei

So
M
ra

Fr
co

rt

le
Ac

Te
Ai

Note: Each sector is evaluated on the degree of restrictiveness from 0 to 1, the latter being the most restrictive. Each
countrys scores are added to each bar for a total score to give an indication of overall restrictiveness. The OECD average
(displayed in black) also includes six major emerging economies: Brazil, China, India, Indonesia, Russia, and South Africa.

Figure 4.1b displays the same information as Figure 4.1a but organized by country
instead. Among the selected countries, Italy has the most restrictions among the Euro-
pean countries but it is below the OECD average (that has been expanded to also
include six major emerging markets which often tend to increase the average degree
of restrictiveness). The US has a fairly high degree of trade restrictiveness, just below
Italy. Germany and the UK are about on par and the lowest in this sample. Sweden
isabove Spain, Germany, and the UK but below Italy in terms of trade restrictiveness
in services.110

110
See www.oecd.org/tad/services-trade/services-trade-restrictiveness-index.htm

144
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 4.1b. Restrictions on traded services in the OECD

organized by country, 2014

Television
4,5
Telecommunications

Sound recording
4
Road freight

3,5 Freight transport

Motion pictures

3 Maritime

Legal

2,5 Insurance

Engineering

2 Distribution

Courier

1,5 Construction

Source and note: See figure 4.1a.

1

0,5

0
UK
USA

Italy
OECD

Spain
Sweden

Germany

Finally, on a sector-by-sector basis: see Figure 4.1c. For construction services, Sweden
is actually above the OECD average. For all other services included in the sample,
Sweden is at or below the OECD average, but in some areas not by much such as
for air transport, computer services, road freight, and telecommunications. Sweden
has low restrictions in architecture, banking, and engineering.

Figure 4.1c. Restrictions on trade in services, Sweden compared to the OECD, 2014

Sweden
0,45
OECD
0,4

0,35

0,3

0,25

0,2
Source and note: See Figure 4.1a.

0,15

0,1

0,05

0
g

ite t

Ba re

rs

Co n

rib r

In ing

ht

ht

un nd

Te ion

on
ga
ie
or

nc

m
tio

io
tin

io
in

te
u

Ro eig

ig

ou

si
ur
sp

Le

iti
ut

ot

t
nk

er
ct

ra
uc

ica
pu
un

vi
S
ar
ne
an

Fr

fr
su

le
tr
m
co

ad
gi
tr

st
ch

ns
Co
Ac

m
En
Di
r
Ar

Co
Ai

m
co
le
Te

145
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Before we turn to issues of digital regulation, we will briefly discuss two topics that
concern legal aspects of digitalization: intellectual property and patents. While they
have had a huge impact on progress and productivity and indeed still do for the
sake of brevity, we will only give an overview.

Ongoing challenges for intellectual property rights and copyright in the digital economy
Few issues are as clearly associated with the challenges of production and distribution
of digital goods as copyrights and intellectual property (IP) rights. It is not that IP was
not an issue in the days of analog music, movies, books, newspapers, and magazines;
but rather the issue is that IP comes to center stage when digital material can be distrib-
uted at low cost. Indeed, digitalization has shaken these industries deeply.

The first wave of digitalization of music brought a tremendous amount of litigation

as music producers confronted the rising challenge of pirated goods. We saw P2P
sharing of music, for example via the company Napster that started in 1999. The ini-
tial response in the industry was to litigate, which was made more difficult by the
scale and global nature of the piracy. Some of the piracy was commercial in nature
and some of it for personal use. Even though Napster was shut down by court order
and subsequently liquidated, the events rocked the music industry and propelled
questions about how to make money in the shift from analog to digital. Notable fears
were that all music would be free so that artists could no longer make a living and
that diversity would disappear.

A recent report that has studied how creative artists have been affected by digitaliza-
tion has, perhaps somewhat surprisingly, found that the industry may be in as good
shape and in some ways better than before; see Johnson (2015). Although the
path has not been straight and the process is still ongoing, the industry has adapted
and has found ways to make money from creative works in a variety of ways,
including sale via digital platforms and subscription services.

While the handling of digital piracy is a concern for producers, it has also given rise
to new and developing business models in order to be able to continue to earn money
from digital distribution and it has created value for artists and producers. To be sure,
subscription is not new but digital distribution brought new challenges to the existing
business models. Nowadays, we are seeing subscription-type services for:
Books (Amazon Kindle, Oyster, Readly, Scribd)
Magazines and newspapers (Next issue, pressreader, Magzter, Readly)
Movies (Netflix, HBO)
Music (Spotify, Apple)

Technology is evolving as consumers are adapting their behavior. People accustomed

to reading print books and newspapers may simply prefer this format and be slow to
change or perhaps not change at all. But predicting a further rise in the overall pop-
ulation who consume media primarily via digital avenues is not overly bold.

While those creative businesses have already changed dramatically in how artistic
content is distributed and how money is made, further changes are nonetheless
underway. It will likely be many years before the industry begins to stabilize. In the
meantime, legal issues of copyright and IP are likely to remain at the forefront.

146
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Patents, a huge area of law in search of a new model

Patents are another big issue for producers. The original motivation behind patents
was to balance the incentive to innovate against the consumer benefits therefrom
accrued. With very strong and long lasting patents rights conferring monopoly rents,
much benefit is given to the incumbent. On the other hand, with too little protection,
consumers may benefit in the short run but the incentive for firms to innovate and
make new products may be dampened, thus diminishing or destroying long-term ben-
efits. The balance between these two dimensions is a delicate matter and there are
those who question whether the balance is not too heavily tilted in the interests of
producers; see the Economist (2015g). OECD(2015f) stresses that the context in
which intellectual property operates has changed substantially, for example, with the
advent of cloud computing, and that investment in intellectual property-protected
capital is growing faster than investment in tangible capital; see also Hargreaves (2011).

The issues are made more complicated by a number of interrelated factors. First, dif-
ferent technologies have different lifespans, meaning that the length of time they accrue
benefits varies considerably. Development of new drugs typically takes years of research
and large costs and can give income streams for a long time. By contrast, technical
advances in phones and computers quickly reach consumer markets but their life-spans
tend to be much shorter. Where can we draw the line between different industries in
ways that are not arbitrary and yet manage to balance conflicting goals? This is an
open question.

Second, the monopoly rights from patents have given rise to a whole industry of
patent applications, all from the mundane and frivolous to the truly original. The
old saying publish or perish that applied to books may also characterize patents.
Famously, even huge companies such as Apple apply for patents for items that seem
somewhat unrelated to core innovation, perhaps most famously patenting the round-
ness of the corners on the iPhone. Firms own not only technologies but libraries
of patents, some of which will never see the light of day and others which are core
standards in smart phones, computers, or other electronic devices. Many of the big
tech companies are involved in litigation on patents in one way or another. For example,
Ericson is suing Apple for patent infringement and Apple is doing the same against
Samsung and so on.

Third, the vastness of the patent system has given rise to a legal froth for so-called
trolls. These are patent lawyers who earn considerably by finding patent infringe-
ments whether real or not. Trying to recoup a small percentage of some technical
standard used in millions of products can give large rewards. It may well be the case
that the incentives are too much tilted towards legal wrangling and that our current
system actually hampers innovation; see OECD (2015f) for a discussion. The danger
posed by the patent trolls is likely to be most severe for small companies in the
start-up phase when they may be the most vulnerable.

Ambitions in the European digital single market (DSM)

The European Commission has recently launched an agenda for a Digital Single
Market for Europe; see European Commission (2015). The aim is to improve access
to services from music to movies and books and more, to harmonize consumer pro-
tection and data protection with the ultimate goal of facilitating job creation. There
are specific proposals in several areas:
Improving trust in cross-border e-commerce rules
Ensuring parcel delivery is affordable also across borders

147
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Preventing geo-blocking, whereby online content is restricted or not sold to some

member states
Modernizing copyright protection
Reducing VAT-related burdens across borders

These ambitions are now being discussed and the timetable includes finalized proposals
that are supposed to be delivered this year and next. In this chapter, we touch on
mainly the last two bullet points.

4.2.2 From traditional regulation to digital regulation

New and better products can outcompete old; this happens all the time, unless new
regulation prevents it by creating enough hurdles to deter market entry and compe-
tition. Governments and authorities need to be mindful of this risk. Since the digital
innovations have the potential to upset quite a lot of businesses, the forces that may
stifle competition are going to be strong and ubiquitous.

Nowhere is the pandemonium greater than in and around the services provided by
Uber, the ridesharing/taxi service. 111 As discussed in Chapter 3, Uber has accumulated
more legal headlines in France, Germany, and other countries within a short period of
time than most other companies do in their lifetimes. Airbnb has also attracted consid-
erable legal scrutiny, especially in New York where a recent report from the attorney
general has alleged that almost three-quarters of their listings are illegal; seeStreitfeld
(2014). But the legal challenges to Uber are in a class of their own.

The challenge for regulators around the world is to infer what kind of company
Uber actually is. Does it provide taxi services or is it a way for people to share rides?
Although driving a passenger in a car is a simple activity, the rules and regulations
concerning the activity are far from simple. As expressed by Commissioner Catherine
Sandoval, California Public Utilities Commission, the first regulatory body to license
the kind of services provided by Uber:

So the bigger issue really came around what is, not Uber Black, but Uber X, and Lyft, and
the others, where you have just regular people who have drivers licenses, not commercial
licenses, and are not licensed limo drivers. So one of the regulatory questions is, is the new
school app-based match up ridesharing exempt from regulation, or indeed, are they a charter
party carrier, or a taxi, or if theyre operating in a different way, are you a passenger stage
corporation? So among our goals were and remain to balance public safety, consumer
protection, reliability, innovation, competition, as well as be mindful of privacy.112

It is likely that the real reason for the attention Uber receives has more to do with
political economy. Those who lose the most from competition with ridesharing services
are in only one profession taxi drivers. The issue is not black and white, however,
and is more complicated in markets where taxi drivers had to pay large sums for their
license (or badge), a cost ride-sharing services tend to avoid. Moreover, the payment
of tax is a controversial issue: ride-sharing services, especially uber-pop, are sometimes
accused of tax avoidance. The ride-sharing services tend to retort that their drivers are

111
The irony of the backlash against Uber is that the when the self-driving car finally arrives as a consumer product, the
need for drivers might be much reduced anyway.
112
Quoted verbatim at the Federal Trade Commission seminar on the sharing economy, 2015-06-09, see https://www.ftc.gov/
system/files/documents/videos/sharing-economy-workshop-part-3/ftc_sharing_economy_workshop_-_transcript_segment_3.pdf.

148
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

not their employees and hence the responsibility of paying income and payroll tax rests
on the individual driver. Here the challenge for the regulators and for the ride-sharing
services is to use technology to make it easy to pay tax when mandated but also to
review the threshold points at which tax becomes payable. In particular, it is not reason-
able to put equality of tax bracket on an occasional ride-sharer with a professional driver.

Getting the appropriate balance between different concerns and reforming tax
payments matters a lot for how the sharing economy will evolve. Consumers will
benefit from the convenience and lower cost of goods and services provided online.
But the issue is much broader than that. At stake is not only regular or extra income
for those working in the sharing economy, but also their possibility to shift between
freelancing/education/regular employment; see Hall and Krueger(2015). Working in
the sharing economy may be a way to avoid unemployment and to retain a foothold
in the labor market in times of fast technological change, especially if combined with
wider social security coverage and lifelong learning (see further in Chapter 5).

Regaining productivity growth will require realizing the benefits from the digital
economy, which will depend on how rules are adapted in several areas:
Labor markets, such as differences in risk for employment/self-employment
(seeChapter 5)
Ease of entry for new firms balanced against consumer safety
How taxes are collected, especially when goods become digital services
Striking a balance between not stifling innovation and protecting producer interests;
How we treat privacy

While the discussion of ridesharing is important, the real crux is the signal sent to
the rest of the economy. Entrepreneurs and innovators searching for a new way to
improve businesses and disrupt old ways may be less inclined to risks if high regulatory
hurdles are erected or perceived to be possible ex post. In this section, we discuss
a few examples of issues that prevent the full benefits of digitalization while high-
lighting the importance of good government as a keystone in times of great techno-
logical change.

Box 4.1 Digital challenges for firms

Regulation
Businesses have to comply with a myriad of existing regulations but digitalization has some
special features that stand out. The core of digital technology is the low marginal cost through
scalability and the network effects that enable a firm to reach not only its local market, but
the region, country, continent or indeed the entire world. All regulation that hinders expansion
is an obstacle both for new and existing businesses and diminishes the benefits of the tech-
nology. Geo-blocking with specific country restrictions, asymmetric taxation between digi-
tal and non-digital services and different regulatory environments are all examples of this.
The European Union is working to reduce regulatory differences through the single market
and harmonize standards for qualifications and services but important differences remain.
Since cultural and language barriers are significant, regulatory obstacles over and above
these hurdles can imply the difference between expanding or staying put. Facebook is sub-
ject to regulatory investigation by the European Commission and has argued that the inten-
tion of one EU regulatory environment is degenerating into 28 different jurisdictions; see
Allen (2015).

149
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The EU is of course not alone in having varied regulatory environments. The US gives consi-
derable freedom to the states, resulting in an environment on par in complexity with the EU
but less weighed down by language barriers and different legal frameworks.

Incumbents curse
It has long been observed that existing businesses even or perhaps especially successful
ones can succumb to what has colloquially become known as the incumbents curse.One
Swedish example is Ericsson, thetelecoms giant that suffered the curse but was able to
recover after significant reductions in force. Nokia in Finland was once the dominant mobile
phone maker, Microsoft missed the shift to smartphone devices only to try to jump on board
after the train had already left thestation, and Kodak went into bankruptcy a few years ago.
Research paints a common theme for these failings that, as a matter of course, mostly are
not due to myopia or incompetence; see for example Bower and Christensen(1995). Success-
ful firms may find it hard to diverge from a profitable line of business on which the organi-
zation was built. The sales force, skilled workers, and indeed - actually especially existing
customers may all prefer the old technology and work towards incremental change. Such a
development holds true until a disruptive technology gains a foothold and creates a new
market, attracting attention and growth. At this point, it may be too late to make a strategic
shift. Incumbent firms need to create management structures that can place bets on disrup-
tive technologies and withstand internal opposition when dictated by strategic concerns.
Although easy to describe in theory, this is hard to do in practice, as the bets may cannibalize
existing profits. New production methods or new products can make old ones obsolete, just as
5-inch floppy disks were replaced by 3 inch disks, which were then replaced by CD/DVD,
which gave way to network storage and cloud services. The sharing economy may also decrease
the need for ownership when many more items can be rented as needed; see Chapter 2.5.
Overall, the chance of long-term survival may all too often lose out against the certainty of
lower profits in the short-to-medium term. Digitalization may well be the perfect storm for
the incumbents curse.

Digital skills and labor market mismatch

Big technological shifts imply challenges for the skill set of people in existing businesses as
well as for the education system trying to catch up. Facit, the Swedish mechanical calculator
company, had its core competence in mechanical skills. When electronic calculators came, the
company was not able to regroup; see Sandstrm(2013).
The mismatch in the Swedish labor market has become more significant in recent years, with
an increase in both unemployment and vacant positions (the Beveridge curve). In Europe,
many people lack digital skills (see Section 5.3 in Chapter 5). To stay competitive, firms need
to ensure that workers have adequate training. On the job learning and a life-long perspec-
tive on acquiring skills has never been more important, be it for the individual or the firm.

Intellectual property and patents

Digital technology has had profound impact on all items that can be reproduced, such as music,
film, and print items. For all those businesses, market conditions have changed and intellectual
property remains a crucial concern. Piracy and illegal distribution were difficult to prevent
but new business models are emerging to enable producers to charge fees. Still, the changes
underway will continue to impact on competition, especially for newspapers, video, and music
content.
The handling of patent rights both the firms own as well as avoiding infringement of the
rights of others is another complex area. Legal, business, and technical questions overlap as
markets and products evolve quickly; see OECD (2015f) for an overview. Patent infringements
can be costly, but the area also attracts legal arbitrage and so called trolls that try to extract
rents, which may dissuade smaller firms from entry and be a cost burden for larger firms.

150
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

4.3 Taxation and the digital economy

All the factors that make taxation difficult in the economy are even more challenging
for digital goods and services; see OECD (2015g) for a comprehensive overview of
taxation in the digital economy. The location of the platform, the use and delivery
of the good or service, may all be in different locations within different jurisdictions.
Similar physical goods and digital equivalents are taxed at vastly different rates, distort
the market, and create peculiar incentives. The case of books vs e-books discussed
below is a case in point. This issue is likely to become more important as physical
goods are increasingly transformed into digital services.

In Gaspar et al. (2014), an independent expert group for the European Commission,
recommendations are given on taxation of the digital economy. Their report stresses
anumber of central features for how the taxation framework should progress:
No special tax regime for digital companies: general rules should be applied to all
Steps toward stable and predictable rules
Highly restrictive use of tax incentives; and apply them only where there are
market failures

These recommendations likely represent consensus among economists but the challenge
lies in their implementation. To strive for neutrality, it is important to avoid distorting
incentives to less productive uses. The predictability of rules is key to creating condi-
tions for long-term investment but extends well beyond taxation includingregula
tions discussed in other parts of this chapter as well as corruption. The expected return
on an investment can change not only due to taxation but also if requirementsto
operate change or costs imposed through graft or red tape. For example, a tax increase
can have a similar effect on costs as a requirement to introduce additional safety or
licensing requirements.

For the digital economy, the above recommendations are a good starting point but
agreat deal of work remains to be done on a technical level that is beyond the scope
of this report. Instead, we will illustrate the discrepancy between the digital economy
and the rest of the economy with the example of books. A variety of forces as well as
historical accidents can result in taxation that is difficult to understand and much
less justify.

Tax discrepancies between books and e-books

Navigating VAT rules is not for the faint-hearted; there are many differences across
countries and various rates within countries, which makes it difficult for businesses.
From an efficiency perspective, having a flat VAT would be easiest for firms to admin-
ister but political considerations often result in lower VAT on some categories, such
as books, food, and childrens clothing, depending on the country. Among the tax
heterogeneities, one of the most difficult to understand is the difference between VAT
on printed books and e-books; see Heyman (2015) and the International Publishers
Association (2015). Reading a book or an e-book are arguably just different ways
toconsume the same good.

For countries in Europe, Figure 4.2 shows the differences in VAT on books and e-books
in the vertical bars, along with the general level of VAT. Overall, the price difference
is quite significant in Europe. Although only a few countries have no differences and
a fairly low applicable VAT, such as France, Iceland, and Italy, most other countries
have considerable differences. Ireland has one of the largest discrepancies at 23 per-
centage points, with zero VAT on print books but 23 percent on e-books; Denmark

151
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

has the same VAT for both media, but at the high rate of 25 percent. The median dif-
ference in VAT is just over 12 percentage points. In Sweden, the difference is 19 per-
centage points. For the 79 countries covered in the survey, 35 apply a higher rate of
VAT to e-books than to print books.113

Figure 4.2. Difference in VAT on print books and e-books in Europe 2015, percentage points
and the general level of VAT

30

25

Source: International Publishers Association (2015).

20

15

10

Portugal
Norway

Serbia
Bulgaria
Albania
Austria
Belgium

Croatia
Cyprus
Czech Republic
Denmark
Estonia

France
Germany
Greece
Hungary
Iceland
Ireland
Italy

Lithuania

Malta
Netherlands

Poland

Romania
Russia

Slovakia
Slovenia
Spain

Switzerland
Turkey
UK
Ukraine
Finland

Latvia

Sweden
Luxembourg

Note: The bars show the difference between VAT on print books compared to e-books. Some countries have no difference
(and hence the bar is at zero, in Italy, for example). The square indicates the general VAT level in that country.

Trade barriers in e-commerce

In the EU, the basic principle for commerce is freedom of movement for goods and
labor. Although a high-level principle, it is not without exceptions and Member States
are permitted to impose their own restrictions, motivated, for example, by consumer
safety concerns. Such restrictions may be subject to (re-) assessment by the European
Court of Justice and are occasionally overturned.

For new digital trends, there is considerable regulatory uncertainty. But for more
established areas, such as e-commerce, NBT(2015a) notes significant improvements
inregulation during the last couple of years. There has been full harmonization of
EUrules on distance-selling and abolition of some establishment requirements. None-
theless, some obstacles remain for regular e-commerce, such as labeling requirements
on products and complying with procedures for local VAT payment. There are also
issues with knowing which national rules actually apply in cross-border transactions;
for example, which standards apply for marketing, data processing, and sales con-
tracts; see NBT(2015a, page 41) for a summary. For example, barriers are set up by
requiring top-level web domains and a physical presence in the country, which raises
costs and administration.

By contrast with the tangible remaining obstacles for e-commerce, regulation for new
digital trends is murkier; see NBT (2015a). The new business models emerging in the
digital economy do not all fit into established regulatory boxes:

113
The US is not included in the International Publishers Association survey.

152
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

What is the responsibility of the platform owner? Full liability, partial liability,
or none as only a conduit for the transaction? Relevant to most of the sharing
economy, exemplified by Uber and Airbnb.
How should product information requirements be applied to digital devices?
Small screens easily become overburdened with information, which is exacerbated
by varying national requirements.
Rules on data protection and cloud services are vague.
Taxation is a challenge, especially when physical goods become digital services,
such as print books vs e-books.

4.4. Regulatory hurdles for digital goods and services need more
attention
The legal hurdles for firms in the sharing economy are discussed in Chapter 3. In this
section, we instead turn to matters related to regulation of digital goods and services.
The issues concerned are broad and affect many aspects of how digital firms compete
in markets, ranging from financial regulation to consumer safety. Although copious
material exists in each area of regulation, work that provides an overall perspective
is sparse. The challenge is that each area of regulation requires specialized knowledge
and it may be hard for a non-expert to assess the importance of technical standards
and what kind of hurdle they present.

One overall issue for countries in the EU compared to the US is that of the scale of
the market. In Chapters 2 and 3 we discussed how the digital economy benefits most
from low cost of expansion; the marginal cost of a new user is virtually nil and as
users increase, network effects generate increasing returns to scale.

In Europe, the language barrier presents an obvious, major obstacle but arguably it
is becoming less so over time. Browsers that can translate text to ones preferred lan-
guage instantaneously lowers the bar. The difficulty for businesses is instead the cost
of adapting their services to various institutional requirements in EU countries, espe-
cially since technology expands quickly and regulation lags behind developments.
The issue of harmonization is a challenge in the US as well, with varying regimes
throughout the states and strong degrees of autonomy. But the EU has to contend
with both harmonization and the language barrier.

In the absence of a broad overview of technical hurdles, we instead provide a few

examples that illustrate the issues at stake:
Legal uncertainty about liability for self-driving cars and other transport.
Legal uncertainty about liability for the Internet of Things (IoT) where gadgets
large and small have internet connectivity.
Asymmetric standards for items built with 3D printers.
Heterogeneity of local government data and rules.
Strengthened investor protection deters funding for fintech; see Vasagar(2015b).

Legal obstacles for self-driving vehicles despite vast potential to save human lives
The example of self-driving vehicles is the most widely known. The question of lia-
bility in the event of a crash is an important one, but may be one peculiar instance
where existing obstacles may be deleterious to consumer safety. Worldwide, more

153
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

than a million people die in car accidents every year. There are compelling arguments
that self-driving vehicles would result in far fewer accidents; the social and human
gains would be sizeable. Edith Ramirez, chair of the Federal Trade Commission:

These potential benefits are immense, but so too are the potential risks...We have an
important opportunity right now to ensure that new technologies, with the potential to
provide enormous benefits, develop in a way that is also protective of consumer privacy.114

Notably, there are ethical and quasi-philosophical issues at stake. For example, a
person involved in a crash may make split-second decisions about where to steer the
vehicle, either at minimum risk to himself or the more altruistic choice of veering into
a physical obstacle with greater risk of personal injury. Such choices would have to
be pre-programmed by the manufacturer supplying the software that controls the car,
and thus a technical solution that improves safety becomes an ethical question that
society may not be comfortable with. Essentially, the issue may be weighing the small
risk of injuries from self-driving cars against the big potential gains from lives saved
through automation. While the potential gains in terms of saved lives are not much
disputed, the slowness of regulation implies that the process may nevertheless take
many years.

Challenges for liability and privacy for the Internet of Things (IoT)
Other legal uncertainty about IoT may also retard its adoption. When everything is
connected, there are great potential benefits but also new risks. If the toaster mal-
functions, will the error spread to the pacemaker? If the traffic light breaks down,
will the lights in the tunnel work? Essentially, this concerns the risk of contagion or
the butterfly effect, referring to how seemingly unrelated events (a butterfly flap-
ping its wings) can have major impact in non-linear/chaotic systems. If most things
are connected, it stands to reason that the risk of contagion may be greater, especially
because it is difficult to maintain an overview of all technical aspects that may cause
a single system to malfunction.

For firms that build functionality with IoT, the consumer safety issues and liability in
the event of an unforeseen event may be a significant hurdle. The IoT also encom-
passes many issues of privacy where progress may have outstripped regulation. Many
devices will be built by manufacturers unused to cybersecurity. Even when security
is built into devices, detection might be hard as we do not (yet) interact with gadgets
the way we interact with computers. Moreover, the issues may be more serious with
IoT than for other data. The hacking of private data in the form of passwords and
even financial data may be considerably less upsetting to people than if information
about their medical history is leaked, because this can damage job opportunities and
affect insurance premiums. For example, a known genetic predisposition to a par-
ticular disease may lead to instances where private insurance might be very expensive
or unavailable.

No specific laws regulate IoT, which is instead governed by existing legal frameworks;
see Croft(2015). This is one explanation for how IoT may be slow to catch hold.
Since interconnected devices may contain a lot of personal data, regulators are in new
territory as to what kind of consent is needed because while some uses of IoT may

114
Quoted in Kuchler (2015).

154
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

be frivolous, others might be a matter of life and death. For example, it has been sug-
gested that personal technology that monitors health can warn of impending heart
attack and thus help save lives, but the same data can also be scanned and analyzed
to suggest potential medicines, possibly blurring the line between welcome health
monitoring to intrusive personal advertising. The current way of giving consent (such
as quickly checking some boxes to participate in networks or download updates)
may be problematic when the data is generated by our bodies. The practical issues
involved in obtaining consent may be substantial. For these reasons, IoT is likely
totake time to implement and the benefits may be a long time coming.

Idiosyncrasies in local government prevent economies of scale for digital services

The right to local self-government is a key feature of democracies and most commu-
nities enjoy some autonomy in decisions at the local level; see Hooghe et al. (2010).
This helps keep local politics vibrant and ensures that people can vote on matters
that affect the local community. Nevertheless, there are several factors that would
bring efficiency gains if organized by central government. The trade-off between local
accountability and central government economies of scale is a perennial question that
is particularly relevant to reaping the benefits of digitalization. Let us consider three
examples:
Taxation
Schedules, flow of traffic and government data
Parking zone regulations

If each local government uses the same open data format, app developers can con-
struct software that can be used in the whole country, in every US state, or across the
EU. But local rights to self-government mean that such standards often need to be
voluntary. However, if each of Swedens 290 municipalities, or each of Frances nearly
37,000, has its own data format the difficulty of incorporating idiosyncrasies into the
code may be insurmountable when the cost and benefits are weighed. Each country
has its own set of institutions and responsibilities. Nesta, an innovation charity in the
UK, has been pioneering various forms of initiatives for open data in governments and
the uses to which it can be put. Nonetheless, the value of accelerating work towards
voluntary standards that enable the possibility of scale should be considerable, but
the supposition that this will take time is not a risky one.

Parking regulations are useful to illustrate the potential benefits of common data
standards. With common data, an app developer can write code that can be used in
the whole country and eliminate the need for different software for different areas.
Moreover, if cities and boroughs within cities can set common rules for parking,
parksharing can help to ease the challenge of finding a parking space, an unwelcome
obstacle for people coming home or away on business. Someone who owns a free
parking space can simply rent it out to whoever wants it via a sharing app; see Wosskow
(2014). Moreover, cars that people rent or share can park nearby and be called when
needed, which will be especially useful in a future with autonomous vehicles.

Common data standards make the integration of public transport information into
commonly used software, such as Google Maps, much more seamless. The benefits
to households and consumers would be considerable, especially in congested cities
where people want to drive alternative routes to avoid congestion or find alternatives
in public transportation.

155
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

In Sweden, governmental institutions are not allowed to collate data from different
sources, such as tax returns and social security records, to improve services or detect
abuse. Ekholm(2015) argues that governmental institutions will find it more difficult
to perform their tasks because of outdated views on data. More specifically, it is diffi-
cult to process and use data more efficiently to learn from experience and find patterns
if every step must be monitored by lawyers.

A final example concerns the tax code. The difficulty of declaring income varies con-
siderably from one country to the next. For most households in Sweden, the process is
very streamlined and, for the most part, people can simply confirm the pre-populated
tax return form provided by the tax agency via text message or online. Business tax
returns are more complex but software is available that automates much of the pro-
cess. In contrast, the US tax code is one of the more complicated and the benefit of
pre-populated tax forms for households would likely be considerable. Nonetheless,
due to lobbying there is still primarily only one software package to automate the
task; see Manjoo (2015b).

Obstacles for realizing benefits from 3D printing

3D printing of objects can save firms the expense of storing spare parts and reduce
the need for costly and timely transportation. Factories, carmakers and others who
need to repair broken machines or vehicles need only download a blueprint of the
item in question. Clearly, 3D printing has the potential to vastly improve efficiency
and reduce costs. On a more non-terrestrial note, astronauts printed a tool needed
intheir space station that was sent digitally from earth.

The comparatively low cost of 3D printing offers the prospect of small series or
highly customized products that would otherwise be too expensive. For people who
need prosthetic replacements for missing limbs, 3D printing opens a world of lower
cost, convenience and individual fit; see Mrozfeb(2015). Moreover, 3D printing has
successfully been used by surgeons to create a practice replica ahead of complex sur-
gery, thus reducing the risks during the actual surgery; see Weintraub (2015). Some-
time in the future, it may be possible to print spare organs, such as kidneys, livers,
orhearts; see Twentyman (2015b).

Against the background of large benefits to firms and numerous health benefits to
patients, allowing 3D printing should be a regulatory priority, but there is anecdotal evi-
dence to suggest that approval in the US is slow; see Wright(2014). In part, the issue
is that regulators are unfamiliar with the issues and every printed medical product is
potentially different than every other.115 The unfamiliarity extends across wide areas
such as catheters for coronary artery procedures regulated by the US FDA and air-
plane parts regulated by the US FAA.

In the European Union, the challenges from the regulatory framework are somewhat
different. Prosthetic limbs made in regular manufacturing are subject to high quality
standards and CE-marking, but those made by 3D printers are not subject to any par-
ticular legislation; see NBT(2015b). This creates an asymmetry in regulation that
could become more problematic as 3D printing becomes more widespread.

NBT (2015b) finds no evidence of specific regulatory hurdles as goods and services
become more intertwined in the servification of the economy. But there are questions

115
The US Food and Drug Administration controls the approval and marketing of medical devices, such as catheters for
coronary artery procedures. The US Federal Aviation Administration regulates airplane parts and such.

156
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

as to how to interpret liability for the final product. Dentists using 3D printers to
make dental implants provide an interesting example. Is it the software maker, the
supplier of the raw material, or the manufacturer of the 3D printer who bears ulti-
mate liability for the safety of the implant?

A similar issue arises when consumers rent 3D printers online; see NBT(2015b).
Forregular manufactured goods (toys, etc.), the producer is liable for conforming to
CE-marking standards, copyrights, patents etc. It is reasonable to ask whether such
demands could be put on a firm supplying a 3D-printing service with no particular
interest in the object being printed. If this were to be the case, the business model
would not be viable and the potential benefits would be lost. This raises a host of
questions that are so far unanswered in the EU regulatory framework and that may
hinder innovation; see OECD (2015f).

3D printing has tremendous potential but the obstacles are likely to be considerable
and it is not yet a disruptive technology; see Sandstrm (2015).

New ways of self-regulation to create confidence in market transactions

As we discussed in Chapter 2, trust and confidence in transactions are key factors
underpinning the market economy. For some time, payments and delivery of goods
were challenges that stunted the growth of the digital economy. Some of the first busi-
ness models were unsuccessful; boo.com, for example, was a fashion apparel digital
marketplace that went bankrupt after the dot.com bubble burst. Nowadays, there
are many firms successfully selling goods online; the sharing economy is another
keyexample of how digital interactions have matured and found ways to establish
confidence.

By and large, confidence has been built up by means of self-regulation and assessment
without much influence from governments. Many service providers integrate evalua-
tions into the process surrounding the actual transaction. For providers of services,
such as Uber drivers, Airbnb hosts and others in the sharing economy, a good rating
is essential to obtaining new customers. By the same token, consumers also may need
to maintain a good rating. For example, there have been reports of people unable to
get Uber drives because of low ratings; see Streitfeld (2015b). The same might extend
to other types of work or services in the sharing economy.

Indeed, ratings in the sharing economy form a system of self-regulation that has been
instrumental in building trust and confidence in their services. We may well go in a
direction where digital reputations and footprint are more important to business than
other forms of identification and recognition. The striking feature of this development
is the palpable absence of government. In many areas important to consumers, govern
ments usually exercise some form of regulation but not in all areas. For example, credit
ratings of firms and individuals are typically handled by private companies, such as
credit reporting agencies.

As digital reputations become more important, whether or not we will see convergence
towards some standard is an open question. Will there be a plethora of different plat-
forms where firms and individuals have to establish reputations one-by-one or will
there be some transferable digital reputation, like a passport? Perhaps in the future
reputations might be aggregated or leveraged in some way. We are already seeing
some early signs of how digital reputations matter. For example, in Agrawal et al.
(2015) contract labor that has performed certain tasks are generally able to charge

157
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

higher prices the next time, while some small companies are able to borrow money
on the basis of invoices sent to large firms with strong credit ratings; see Economist
(2015d).

Whatever the evolution of digital reputations, it remains a key area for trust and
without establishing broad and credible ways to build up and monitor reputations,
businesses and transactions are likely to be constrained.

4.5 Big data and Internet of things

The Internet of Things (IoT) and big data are topics that have received copious atten-
tion and voluminous publications and books; see for example Mayer-Schnberger and
Cukier(2013), MGI (2011, 2015a) and Greengard(2015) but any search generates
numerous hits in books, magazines, and newspapers. To do these topics justice, we
will focus only on the economic forces at work rather than the many possible future
uses, which are important but also tend to be rather speculative. It is also necessary to
discuss personal privacy and the hype surrounding these technologies. Indeed, MGI
(2015a) in a major flagship publications purports to go beyond the hype on IoT, but
then writes:

Our central finding is that the hype may actually understate the full potential of the
Internet of Things...(sic)- MGI (2015a, page vi).

Personal privacy surrounding these technologies is also a big issue, involving ethics,
judgment, political freedom, and a host of other aspects. Again, to do the topic justice
we will focus only on those forces that may affect how the technology is used in the
economy.

4.5.1 IoT and big data: where we stand today

Big data is arguably already here but less so the IoT. MGI (2015a) gives the example
of oil rigs equipped with thousands of sensors that generate data but only 1percent
is analyzed. Even though IoT is largely the physical realization of digital technology
and big data is the intangible result, ultimately both are about making current tasks
better, everything from health care and manufacturing to science. They also tend to
share the same obstacles, especially personal privacy and legal issues.

Big data
It would be useful to have a definition of big data, but developments are moving so
fast that it may be more trouble than it is worth. What is clear is that big data... is
big. Gigabytes are long out of fashion and we are having to use new and non-terres-
trial sounding prefixes to describe size: tera-, peta-, exa-, zetta- and yotta- (the latter
being the fathomless 10008) bytes. Our ability to store media has increased dramat-
ically almost every year and costs have been coming down. Much of this is due to
Moores law about the size of microprocessors but it is also due to expansions in
broadband internet, smart phones, and consumer behavior. The technologies have
developed in tandem to make big data a reality. For example, without fast internet
access, tracking internet behavior would be less feasible and people would also be
less inclined to spend time online. Without fast connections, shopping would be a
lesser experience, streaming of movies or uploading pictures might not be happening.
Online content draws people in, but it is what people do that generates the data. In
big data, we the people are the commodities.

158
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

All the actions we do online can and are tracked by various means. Our surfing
habits, the computer we use, what we buy, and our social network footprint are all
valuable information that can be used to generate value and make more direct adver-
tising. While many services allow people to opt out, for example by not accepting
cookies, but in reality, stronger privacy controls often diminish the online experience
and most applications assume a fair amount of exposure. Mostly, the battle for online
privacy was lost long ago, if indeed it was fought at all. It is fair to say thatonline
collection of data is an area where regulation is far behind. Even if regulators were
todecide on some standard for what data could be collected, it might not be possible
toturn the clock back. Even limited regulatory changes are difficult enough. The
European Union has recently forced Google to implement the right to be forgotten,
by removing links to peoples unwanted past. While it may be costly and technically
difficult to implement, it is really only a small part of online personal privacy.

The recent development of ad blockers, whereby advertising can be prevented from

appearing on tablets and smart phones, has raised a whole set of new questions. Ini-
tially, these developments were welcomed as a means to increase privacy and make
web browsing less bogged down by unnecessary data. The downside is that a lot
of internet content provided free of charge depends on advertising revenue. Should
ad blocking grow in size, so might the availability of free content be reduced. These
developments are still recent and it may be the case that some new way is found to
reduce nuisance ads while not undermining the free content model on much of
theinternet. But the evolution is still ongoing and it is too early to tell.

Deriving benefits from the IoT

Defining IoT is difficult for a similar reason that big data is somewhat elusive: what
is classified as IoT is evolving; see OECD (2015c, page 242). For example, IoT may
be intelligent thermostats that help reduce energy consumption, sensors in fridges that
monitor available food and best-before dates, or sensors in cars that monitor various
aspects of the trip and driving style. But IoT is also about sensors in oil rigs and pro-
duction processes. The gains from IoT may differ depending on the sector in question,
but for consumer goods it may well be that the most benefit is derived once the product
is being used generating data and improving whatever process it is a part of. So far,
IoT is mostly used for control and detection, not optimization or improvement, but
that is likely to change in the years to come.

While IoT is in its infancy, further developments will depend on regulation and agree-
ments of standards for communication and how platforms communicate. Legal and
ethical issues will have to be addressed. On the technical side, the proliferation of
sensors will require much more reliable bandwidth, an issue that is addressed with
the next generation of mobile networks 5G. While 4G is already fast for many IoT
applications, some of the benefits will only be possible with much more reliable con-
nections. There is a big difference in quality demand, comparing a few minor blips in
a streamed movie as opposed to the possible consequences of an electronic hiccup for
an industrial appliance performing precision work. Ericsson and other telecommu-
nication companies will roll out 5G in the next few years; we may also see the first
phones without SIM cards.

These technical changes imply different challenges in the economy. We will discuss a
few to exemplify the issues involved: industry, consumer intelligence, and agriculture.
But there are vast applications outside these areas, such as health, the public sector,
and environmental protection that are not discussed; the interested reader is referred
to MGI(2011,2015) and the references therein.

159
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Big data and IoT in industry

The manufacturing industry has long been pushing the forefront of productivity growth
ahead of other sectors, especially services. Future advances in robotics, improved
organization and management and other innovations will likely continue this trend.
Whether this will be enough to counter negative demographic trends and other factors
discussed in Chapter 1 is an open question.

In many respects, manufacturing is ideally suited to using IoT. Big data and IoT have
the potential to further boost manufacturing performance and productivity, but as
we will discuss further below there are obstacles along the way and at present a
great deal of data receives scant attention. The potential lies in expanding industrys
long tradition of measuring and gathering data about production. A more complete data
set for all operations could be used to improve processes and even improve throughput
in real time, controlling all aspects of production and saving production time.

MGI (2015a) estimates that product development time can be drastically cut and that
delays can be reduced by using simulation. Powerful computers can model events and
use 3D to streamline performance. While big data is often associated with people,
machines generate vast quantities. Self-driving cars have a huge number of sensors that
are continuously updated. Commercial flights generate hundreds of terabytes of data.
This makes it possible to draw on real world experiences to improve production design.

Many production lines are already digital but often do not communicate with other
parts of the process; see MGI (2011). When all aspects of production are integrated
with digital technology, it is possible to get a full picture rather than only trying to
improve the parts. Each sequence can be properly modeled and strengthened. Major
changes in production can be simulated across several factories in models before imple-
mentation. A digital map of production might also point the way to areas where research
and innovation would do the most good.

The data generated might also be used for so-called predictive maintenance, reducing
unscheduled production downtime and reducing, if not eliminating, the need for sched-
uled maintenance. IoT sensors used in logistics operations show how inventory can be
tightly controlled, which is also relevant to production lines. The next level in automa-
tion may also further reduce the risk of human errors and their possible consequences.

All the arguments point to a considerable upside in productivity but it is hard to assess
the outcome and how serious the obstacles are. MGI (2015a) estimates that IoT will
have a huge impact. Following the success of Watson, IBM has invested billions of
dollars in IoT, cloud computing, weather services, auto insurance, and health; see Lohr
(2015a, b, c). Amazon web services provide vast computing power to firms to apply
to their production.

For companies in Europe, going more digital is a competitive race against the US,
South Korea, and Japan. Notably, Sweden and Germany have several successful
multinationals in industry becoming more digital. Acatach (2013, 2015) describe the
challenges and issues for German industry to adapt in this regard, which are also rel-
evant to Sweden despite the differences in size. Notably, Germanys Industrie 4.0 pro-
vides a vision for the country up to 2025, outlining issues and challenges.

A particular challenge for Germany is that the very success and fame of German engi-
neering from heavy machinery to cars may make it hard to adapt. Changing a
successful business model is difficult, especially if the changes initially do not generate
profit the incumbents curse discussed above. Acatech(2013, 2015) argues that digi-

160
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

talization will require a change of focus from product to client, with more individual-
ization according to what customers want. Thus, a continued change towards making
production more like services is likely. With digital platforms as the key component,
the owner of the data may increasingly come into the drivers seat. This may require
adapting management structures and procedures and perhaps also collaborating more
with other stakeholders, even competitors:

Competitors will cooperate with each other and employees will engage in automated
interactions with platform operators, meaning that they will no longer be managed by
thetraditional in-house management structures of the company that employs them.
Acatech (2015, page 13).

The notion that competitors may need some cooperation in digital platforms raises
new anti-trust questions that need to be addressed. The technology in itself is already
bringing some new aspects to competition policy that we are only beginning tounder
stand; see van Gorp and Batura(2015). The combination of increasing returns to
scale from network effects and the rapidly evolving nature of these markets raises
new issues of how to maintain competition. Markets are contestable in ways that we
have not seen before, making entry easy but hard to sustain. The technical nature of
standards means that small changes can have big impact on competition. For example,
bundled packages of digital applications may increase competition in some instances
but reduce it in others; see OECD (2015c, page 19).

The changes are driven by consumer demand, but a particular challenge for industry is
that it may be hard to charge more money for IoT connectivity. The costs of investing
in the technology are paid up front, while the benefits of improving production with
big data may come some years down the road. The issue is further compounded by the
sensitivity of data (more on this below) and the legal issues involved. This means that
US companies that are already digital first have an advantage once they enter. German
carmakers have been taking an incremental approach, consistent with long traditions
of quality and safety, but this also means being less prepared for the next level of auto-
mation, such as for autonomous vehicles. In recent interviews, Google indicates such a
direction, wanting to eat the carmakers lunch, see Waters and Sharman (2015). This
refers to the profits that may arise from delivering content and services to the driver
or to the passengers in the car. The smaller the profit margins on the vehicles, the more
important it will be to charge for services rendered to the passengers.

Industry has a long tradition of using robots, but it has often been a time-consuming
process to get all aspects right when setting up such lines, especially safety issues with
very fast movements that could injure humans. With IoT and big data, robots can
be programmed to learn and setting up factories can be much quicker; see OECD
(2015c). Indeed, robots are beginning to take over production of electronics that
has previously been outsourced to Asia. For example, one of the biggest investors
in robots is Foxconn in China/Taiwan, which has been involved in production of
iPhones and tablets.

The overall effects of digitalization on industry are hard to assess but a few things are
clear, in qualitative terms. Industry may be poised for continued productivity growth
if the obstacles can be overcome. IoT and big data bring potential benefits in auto-
mation, improving production from huge amounts of empirical data and energy savings.
There are also vast synergies possible between different areas of transport, infrastruc-
ture, and sanitation.

161
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Increasing possibilities for price discrimination in consumer markets

One of the clear uses of big data already underway is that of understanding customers
and their needs. Information on online behavior and profiles can be used to tailor
goods and advertising to specific groups in ways that were not possible before. Prices
can be set according to willingness to pay based on sophisticated algorithms, what
economists otherwise call price discrimination. For example, algorithms can predict
when a customer is likely to leave a website without purchase and hence provide
asteeper discount to induce a sale anyway; see Tanner (2014).

In Chapter 2, we discussed how Amazon has been noted to change the price of some
goods many times during the day, depending on a range of factors, such as what com-
petitors are doing, etc. While the algorithms that analyze data are becoming cleverer,
the underlying change driving price discrimination is big data, extracted from our
emails and from our actions online. Every search we perform, every phone call, and
every mouse click can be tracked and provide information that may be helpful in
designing personal deals; see for example MGI(2011) and the Economist (2010).

In 2011 Malte Spitz, a German member of parliament, requested and ultimately

received the records from Deutsche Telekom that logged his mobile activities.116
Therecords turned out to be quite comprehensive and able to provide a fairly detailed
account of his life and whereabouts using meta data about his location, time of call
etc. Knowing the content would provide another layer of information. In Sweden,
journalists at Dagens Nyheter accessed similar information and received detailed
information about their own whereabouts; see rstadius and Larsson (2015).

But big data cannot only be used to detect and target consumers to sell goods, they
can also be used for services. Drivers in the sharing economy can use data to assess
where demand is likely to appear, with services such as SherpaShare, and choose to
bein the right spot at the right moment, perhaps making it less difficult to find a car
on a rainy evening; see Singer and Isaac (2015).

Big data can also be used by management to monitor employees to an extent not
previously possible, through email and performance assessments in real time; see
for example Streitfeld(2015a). This raises some ethical issues that we will discuss
inChapter 5.4.

Agriculture low productivity growth may be boosted by IoT and big data
Agriculture is an area that was most affected by the Industrial Revolution, with
machines replacing human labor. After many years of gradual changes in the last
century, the sector is poised for some significant changes once again. As we discussed
in Section 3.4.2 in Chapter 3, automation is making further inroads into farming:
robots are now sufficiently flexible to also harvest more delicate fruits and displace
more manual labor; tractors are self-driving. Indoor farms are making it possibleto
exercise a much higher degree of control independent of seasons on all aspects of
crop growth, such as light conditions, water flow, and fertilizers; see the Economist
(2014d). Robots are also replacing simple milking machines, allowing the animals
more freedom and autonomy to decide themselves when it is time for delatte; see
McKinley(2014). Using various sensors and transponders, the quality and quantity of
milk is monitored, how much the animals walk, and their overall condition. Smaller
farms that donot upgrade may have a hard time competing. Although the operation

116
His discussion of what led him to request the records and what transpired afterwards can be seen on TED talks,
http://www.ted.com/talks/malte_spitz_your_phone_company_is_watching.

162
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

may seem a bit Orwellian, the next level of automation may actually improve animal
welfare, giving them more freedom and humans more time to care for them.

But the changes in motion are more encompassing than just the next level of automa-
tion; see the Economist (2014c) and Lohr (2015e). The notion of planting crops and
using experience to improve conditions and yield are particularly well suited to using
big data and IoT. Large databases make it possible to derive information about soil
quality, weather patterns, and other data that affect yields. Using cheap (IoT) sensors
combined with satellite information, farmers can adjust and improve crop yield to a
much larger degree. Big data can play a particularly significant role in understanding
the features that improve yield by drawing on vastly different data sets: large varia-
tion in the underlying variables reduces the uncertainty of their effect on output.117 By
using all available data, the farmer can assess how much nitrogen is in the fields and
thus how much fertilizer is needed and by one estimate yield can increase by 5 per-
cent; see Harding(2014). According to another estimate in MGI (2015a), the increase
in crop yield from better use of IoT sensors could be in the range of 25 percent. Though
such numbers are clouded in uncertainty, it is unequivocally so that much of the guess
work about what works in agriculture is eliminated while inputs are based more on
available science.

The benefits of big data in farming are not limited to better crop yield and output,
but also better use of land and water. Better use of available resources may improve
sustainability and the environment. Making better use of inputs may be particularly
important in areas where water is scarce. California, for example, has experienced
many years of drought, affecting farming and consumers alike. In developing coun-
tries, where unfavorable weather can result in famine, wiser use of resources is not
only a matter of productivity gains but also improving basic living conditions.

4.5.2 Personal privacy

Scott McNealy, co-founder of Sun Microsystems, is alleged to have said: You have
zero privacy...get over it. The last few years has been an inexorably march towards
people divulging more and more information about themselves, both willingly and
unwittingly. The amount of information that can be mined through our online activi-
ties easily lends itself to notions of Big Brother in George Orwells 1984. The difference
compared to the ominous vision from fiction is that it is not the state collecting the
information but private companies. For the most part, people willingly share certain
information through their activities in trade for the benefits on the internet, such as
free services, networking, and so on.

It is not clear whether a typical user actually is aware of just how much informa-
tion becomes available through online activities, just as it is doubtful that people
(inSweden) are aware of how much tax is paid on their employment when social
security contributions are included.

There may be a privacy backlash in the future, but at least for now, it seems that if
ever there was a battle for privacy, it was lost before it was fought. There are many
issues related to personal privacy that fall outside the scope of this report; see the dis-
cussion in Bylund(2013). A report from Sveriges Konsumenter (2014) finds several
examples of firms that do not comply with the Personal Data Act that regulates the
processing of personal data.118 There are also big differences in how much informa-
tion the user is explicitly asked to share without explicit consent.

117
To get good parameter estimates of a dependent variable, the more variation there is in the independent variables the
better.
118
The Swedish Consumers Association.

163
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

World Economic Forum (2011) argues that personal data is a new asset class and
should be treated as such. Meyer(2015) stresses that consumers should get more
value from it than simply giving it away. For this to occur, there will have to be a
clearer market transaction than is currently the case and new standards will have
tobe developed. But as technology is evolving rapidly, it is likely to take time before
new standards emerge. Advertising on smart phones and tablets is now in its infancy.
Google and other firms are vying to supply content. At the same time, ad blocking
software is also being created and should its adoption become widespread, perhaps
as a response to issues of personal privacy, the fundamentals will change once again.
Personal privacy and its effects will be on the agenda for the foreseeable future. In
Sweden, government regulation manages the trick of being simultaneously strict and
far behind the times.

4.5.3 The hype

Big data and the IoT have great potential. Indeed, the technologies are being imple-
mented in a variety of sectors with promising results. But they are also surrounded
by tremendous hype and it is difficult to forecast the implications. Not everything
that counts can be counted, and not everything that can be counted counts is an
old saying that may well apply. There is an incentive for the people who work with
these technologies to sell them regardless of how useful they may be, no different
than selling overcapacity or surplus items in the past. Firms and consumers need to
embark on a learning-by-doing path to discover the measurements that matter. MGI
(2015a) highlights the example of the oil industry that has thousands of sensors on
its platforms yet analyzes data for only a few. Perhaps more measurement would
improve production but it might also be that only a few of the measures are impor-
tant and the others mainly noise. Of course, until the data is analyzed, we cannot
know which.

The complexity of what works has not been an impediment to very precise and bold
forecasts:
$11.1 trillion per year potential economic and consumer surplus from IoT,
MGI(2015a).
Ericsson predicted 50 billion connected devices by 2020, which in June 2015 was
roughly halved to 26 billion, see Ericsson (2015).
Cisco also predicted 50 billion devices by 2020, see Evans (2011).
OECD (2015c, page 255) predicts 14 billion by 2022.

The range of forecasts of the vast potential of IoT and big data are part of what
the tech analyst firm Gartner labels the hype cycle.119 It is reasonable that there will
be considerable growth in IoT applications in the next few years, but forecasts for
long-run trends of this kind are notoriously difficult to make. The pace of IoT and
big data will depend on the pace of overcoming obstacles for the technologies, dis-
cussed in the next section.

But apart from forecast uncertainty, which for most institutions is considerable even
for the next few months, there are other issues about the value of big data whether
generated by IoT or consumers.

119
http://www.gartner.com/newsroom/id/2819918.

164
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Big data correlations are not the same as causal relationships

In most econometrics courses, usually somewhere in the middle, there will be a small
section about correlations and causality; there may also be references to data mining,
that is, searching for and manipulating data until the hypothesis being investigated is
fulfilled. Econometricians also sometimes talk about collective data mining. This occurs
when several researchers test the same hypothesis on similar data (thus appearing inde-
pendent but implicitly acting in cohort and thus not truly independently).

The problem with data mining be it individual or collective is that the confidence
level at which estimates are confirmed or rejected is incorrect. Statistical tables for
confidence bands would need to be adjusted for the entire path of estimation choices
the researcher has made to arrive at the conclusion.

Moreover, even when a stable correlation is found, it might have nothing to do with
an underlying relationship; see Mayer-Schonberger and Cukier (2013). It could either
be spurious occurring just by chance or the variables may be connected through
other variables that are in fact driving developments. The problem is especially pro-
found with time series: take any two times series with trends: they are likely to be
correlated simply by either increasing or decreasing over the years; the underlying
trend is time and not an inherent stability in their relationship.

The issue of correlation and causality is also tricky for professional economists, but
is becoming increasingly important as more data is collected and analyzed. Businesses
are going to get a wealth of correlation analysis, some of which might work well for
a while but then suddenly become misleading. A lot of digital firms that base assump-
tions on correlations may be at risk and the huge amount of available data does not
help; this may provide false comfort about the precision of estimates. Large data sets
can return small error bands but if the underlying test is unsound, the result may be
misleading. To alleviate the problem of over interpreting big data, Peysakhovich and
Stephens-Davidowitz(2015) stress the need to also base decisions on small data
from surveys or other sources.

Nobel laureate Robert Lucas highlighted one particular problem related to predicting
behavior based on data in a famous 1976 paper: when underlying factors (such as
policy or taxes) change, consumer behavior may also change, which may well be the
case for data extracted from online behavior. One example is Googles flu predictor,
launched in 2008, which was based in part on the number of flu-related searches on
their web site. It was initially successful but then overstated the results for several
years; see Lazer et al. (2014).

Why were the results wrong and what might be the implications for other big data anal-
ysis? Two possible explanations stand out. First, as Google was becoming more widely
used, people might search for flu not because of their own symptoms but to find out if
flu was approaching their area (i.e., the behavior change that Robert Lucas wrote about
in an economic context). Second, as Google continuously tweaks its search engine, the
comparison to past data may change. Lazer et al. (2014) also note that it is difficult to
improve on the simple model used by the US Center for Disease Control (CDC).

While errors in predicting the flu might seem relatively minor, as IoT and big data
become larger parts of our lives measuring our health, driving habits, consumer
preferences, and so on the potential missteps can become more serious. Research
and science have a slow and laborious way of verifying claims and big data with
higher precision does not downgrade the importance of being mindful of whether
theresults are correlations or actual causal relations.

165
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

4.5.4 Obstacles
There are three main obstacles that may slow the use and adoption of IoT:
Institutional inertia
Government regulation, including in the area of privacy
Proliferation of platforms difficulty agreeing on standards

Institutional inertia and risks to business: lessons from the past

While the digital revolution is faster than previous technical changes, not all features
involving digital will adapt at the same time. Many of the fastest changes, like the
sharing economy and social networks, are consumer-driven. By contrast, some of
theIoT benefits accrue to established firms and then changes may be slower. Existing
businesses have management and organizations built around a profit generating line;
if the line has been successful to date, willingness to change may be muted.

The prime example of this in modern times is Kodak, which practically invented digital
photography. From what little is known about internal deliberations, Kodak manage-
ment knew they needed to change, but was unable to steer the company away from
its cash cow of analogue pictures, though the risks were clear even then. Several dec-
ades ago, Facit, a Swedish company that made mechanical calculators, was similarly
unable to change as technology went electronic. Facit had its core human skills in
mechanical engineering and the location of a small Swedish town were obstacles that
would have been difficult under most circumstances, and ultimately proved insur-
mountable; see Sandstrm (2013). Even with the experience of Facit as a warning
sign, a similar fate almost beset Hasselblad, a Danish/Swedish camera manufacturer;
see Sandstrm(2011).

In general, successful businesses will find it especially hard to change when the deci-
sions de facto imply cannibalizing the existing business and main profit lines. This
dilemma has been at the forefront of print newspapers that suddenly found them-
selves faced with the question of how much material to put on their websites for free.
Of course, restricting all traffic to pay-per-view was possible but in practice meant
fewer viewers, especially newer generations. Making too much free content available,
on the other hand, meant losing revenues and risking the alienation of existing print
subscribers why pay when most content is free anyway? The dilemma of how much
to charge for media and what to provide free remains to this day and successful mag-
azines and firms have managed to strike a balance, notably the Economist and the
Financial Times. Two examples in Sweden are the evening paper Aftonbladet, and
Dagens Samhlle, a weekly paper mainly aimed at local government that has a dif-
ferent payment model, providing very little free of charge mainly the debate forum
and is able to charge for its print copies.

The free vs pay question lingers over other consumer goods as well, especially gadgets
that were previously sold separately but are now often free of charge on smart phones,
such as GPS navigation, running apps, and radios.

For industry, the speed of IoT adoption will depend on a number of factors. For one,
transforming businesses to IoT and using new technology will be risky and require
investments that may only reap benefits far in the future if at all. This may result
in skepticism and risk-aversion at senior management levels. Moreover, technology
moves fast and it is easy to be left by the wayside. Ford, for example, recently discon-
tinued its investment in car seats that could measure data from the driver, such as for

166
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

the purpose of warning of an impending heart attack. The decision was ostensibly
based on an assessment of being too far behind competitors; see Sharman(2015).
Moreover, as noted by Acatech(2015), clients may not initially be prepared to pay
apremium for IoT applications until they are convinced of the business case a sort
of classic chicken/egg problem.

A second factor is the quality of data. Management in some firms may not trust the
existing data; see the Economist (2010). For example, data may be poor due to dupli-
cates or incorrect entries. The risk is then that big data would build a bigger haystack
with the same poor hay, making it harder to find the needle.

Third, old-tech companies may face competition from new purely digital firms.
Thisis already occurring in some areas, such as fintech, automobiles, and some home
appliances. Firms that are digital from get-go are not bogged down by legacy systems
and face no organizational inertia of the type that drove Kodak to the grave. Googles
recent reorganization as a holding company for a variety of businesses can be seen
in the light of the business maturing and requiring more management focus on indi-
vidual ventures to assuage investors concerns. Google is known for making big bets
and investors may want more transparency concerning their viability, for example
for autonomous cars and home appliances; Samsung is already a giant in consumer
electronics and can leverage this to IoT applications. In insurance industries, new
startups such as Oscar in New York are drawing from the advantages of big data
to make better business and provide a simpler interface for its customers; see De la
Merced(2015).

Although some industry, such as automobiles, may be subject to competition from

digital firms (Google), it seems unlikely that the heavy manufacturing industry will
have much to fear at least in the first phases of the digital transformation. Investors
may not have the patience for even extraordinarily successful firms like Google and
Amazon to foray into areas far outside their domain unless they generate profits; see
Garrahan and Bond(2015); Google Glass (with internet connectivity and camera)
was recently ditched and the future may imply greater discipline and focus on core
business. Amazon is another example of a firm trying to enter new markets and rev-
olutionize businesses, most recently with plans to deliver ordered goods via drones.
IBMs Watson is venturing into health care analytics with its recent acquisition of
Merge Healthcare; see Lohr(2015c). We are now seeing a new range of collaboration
between tech and medicine, for example, robot surgeons and automated medical advice:
Dr. Google (or Dr.Apple) will see you now may well become a reality in the not
too distant future; see Crow (2015). Depending on the ability to go from concept
to consumer acceptance, the profits for digital firms may give more or less leeway
to enter other areas. But for manufacturing and pharmaceuticals, where the start-
up-costs of entry are large, these specters do not seem imminent.

Government regulation, privacy, and safety

If one goes searching for an area that holds a lot of promise but also an inordinate
amount of regulatory miasma, IoT would surely qualify. It has privacy issues, safety
issues, commercial applications, military applications, and health issues. Practically
all the controversial topics one can think of apply to the IoT in one way or another.

One perception of IoT is that of connected household appliances, such as the fridge,
coffeemaker, etc. The idea of an intelligent fridge that knows when you are running
low on milk and can order delivery or not, depending on your schedule, is surely a
fun gimmick which may or may not be viable commercially. When our heating systems

167
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

and thermostats are hooked up, whoever owns or has access to the data knows when
we are home and has a great deal of information about our habits. What happens if
the fridge and other gadgets are hacked? When gadgets in our homes can be read and
manipulated, the temperature on all sorts of privacy issues is bound to rise. Knowing
our habits at home could be a gold mine of information and not all of it will be
innocuous.

But even more worrisome is the possibility of hacking IoT gadgets for fun or for
nefarious purposes. Only the imagination can limit the havoc that could be wreaked
everything from disrupting pacemakers to taking over control of cars. Recently,
hackers demonstrated how they could take over a car armed essentially only with
the IP address; see Tufekci(2015). Another central privacy concern related to IoT
gadgets is that of disclosure of medical histories and the issues this poses to insurance
premiums and future employment opportunities. Because data may be processed by
a lot of companies that are digital security neophytes, data security is an issue; even
firms whose core business is actually IT routinely have to issue security patches and
updates.

We are likely to see firms having to try different ways to discover what kind of IoT
devices are viable for businesses and households alike. For a lot of people, the benefit
of divulging private information via their surfing habits or on social networks perhaps
outweighs the costs, at least on the evidence that the practice is so widespread; for the
IoT, this remains an open question. The benefits of remote controlling the coffee
maker or washing machine may be small for many people and the convenience
marginal compared to the risks posed.

The challenge of establishing rules for privacy and protection versus the need for
business to use data for new services will remain relevant in the years to come. The
European Court of Justice recently ruled against the Safe Harbor agreement that
has allowed firms such as Google and Facebook to transfer data between Europe and
the US on the implicit understanding that it was adequately protected by local legis-
lation; see Robinson et al.(2015). The implications of tearing up this consensus on
data protection are still being explored. Morozov(2015) argues that we are just in
the beginning of a worldwide fight on how to handle data. These developments make
it more difficult to extract benefit from the data, a topic we discuss in a more specific
context in the next section that of data owned or administered by local government.

Obstacles to obtaining benefits from open data in local government

Governmental bodies collect vast amounts of data for a variety of purposes, such as
monitoring, tax collection etc. Only some of the data is provided free of charge; most
is available only for a fee and some not at all, either for privacy reasons or because
it is stored in analog form. The benefits of some data can be substantial, for example
information about geography and maps, historical information, and art.

Pricing of the data is sometimes an obstacle. In many countries, governmental bodies

charge considerable amounts, notably statistical agencies. Koski (2011) uses data on
15 countries over the period of 20002007, including major EU countries, Sweden,
and the US, to estimate the effects of pricing. The result is that freely available data
or data charged at marginal cost is associated with 15 percent stronger growth for
firms; the effects are particularly evident for small and medium sized enterprises, less
so for larger firms. In Denmark, one study found that free and unrestricted access to
addresses generates high monetary value and the benefits accrue about 30 percent to
the public sector and 70 percent to the private sector; see Danish Enterprise (2010).

168
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The value of this data is also emphasized in Vickery (2011) for the European Union
as a whole and estimated at about 27 billion. Although there is some uncertainty
about how much the market and services for data could grow, most estimates point
to substantial benefits; see also SOU (2014b).

There are several obstacles to harvesting benefit from data. To be sure, some of the
data may contain sensitive information. Even when data is anonymized on the sur-
face, it may nevertheless still be possible to identify individuals by correlating the
information with other sources. These issues are real and need to be addressed, but
at the same time are not an argument for not making concerted efforts to extract the
benefits for the public and firms. For example, digital maps and information about
public transportation and traffic can help commuters and tourists alike. The Stockholm
Chamber of Commerce argues that this and other data should be available electronically
at little or no charge with clear disclosure of data sources; see Stockholms Handels
kammare(2014).120

Data protection in the EU the General Data Protection Regulation

For some businesses the flow of data between borders is important to operations
and is likely to become even more so in the future. The benefits of IoT and big data
will not accrue to firms if data has to be scrutinized, sanitized or in a manner of
speaking present a passport at the border. At the same time, there are legitimate
questions about how such data is handled, in terms of privacy, copyright etc. The
European Court of Justice recently ruled against the Safe Harbor Directive; see
Robinson et al. (2015). The potential impact is that data may no longer be sent
backand forth between US and Europe, thereby limiting the potential economic
benefits while, at least for now, placing emphasis on personal privacy.

This ruling is not likely to be the end of how the overcome the challenge of balancing
economic interests against those of privacy. It is vital that the EU continues to address
this issue and brings clarity. Data-driven innovation will be essential for new business
models; see for example World Economic Forum (2011). Svensk Handel (2015) also
emphasizes the importance of:121
Bringing clarity to the rules about what constitutes personal data.
Reducing uncertainty about joint liability of data processors and data controllers.
Having a set of sanctions for breaches that is proportional to the severity of
infractions and making allowances for warnings if breaches are unintended. The
current proposal may unduly deter innovation.

Proliferation of platforms and standards

The hype of connecting data on everything and optimization is based on data being
in compatible formats to enable seamless exchange between the traffic system in a
city, public transportation, energy and water utilities in the public sector. In the pri-
vate sector, the benefits also depend on communication up and downstream from
suppliers of intermediate goods to final producers and retail stores. But there is a
clear risk that there will be a proliferation of competing platforms. Granted, the pro-
cess of establishing a winning platform is necessary to find a good market-based
solution, but it may take time. The experience from previous platform competitions
is mixed. Betamax video was likely a superior format in quality to VHS, but lost

120
Stockholm Chamber of Commerce.
121
The Swedish Trade Federation.

169
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

anyway. After some years of competition, Blu-ray movies won the HD format war.
There is nothing inevitable about the swift emergence of a single platform; even
whenthis is the case, there are also complicated anti-trust issues to contend with;
seeVan Gorp and Batura(2015).

The winner of the platform race is set to reap big benefits. But clearly, if it takes many
years to agree on standards to apply to IoT communication, the benefits of inter
connectedness may be a long time coming; see Thomas (2014).

4.6 Summary
Against the background of slow productivity growth, it should be an imperative to
ease the obstacles for businesses to grow and create value. The European Commission
is working to strengthen the single market and solidify the foundations of the single
digital market. While significant progress has been made on the free movement of
goods, challenges remain for new business models associated with the digital economy,
such as the sharing economy and peer-to-peer platforms, liability, copyright, informa-
tion requirements, and processing VAT payments.

The pace of reform and clarification on these issues may be too slow for the benefits
to help productivity in the short-to-medium term. The dampening macro trends dis-
cussed in Chapter 1 arising from aging populations and high public debt are aggra-
vated by the aftermath of the financial crisis, affecting young people trying to enter
the labor market as well as middle-aged people who have been laid off or had poor
wage development. There is too little urgency to resolve the obstacles and allow
productivity the full benefit of the potential of rapid technological change.

Regulators are belatedly trying to set tough standards for privacy, but this arguably
amounts to closing the barn door after the horse has bolted. An enormous amount
of information is already voluntarily given away to get free services, such as email,
maps, etc. Turning the behemoth cruise liner around and away from free digital ser-
vices in exchange for private meta-data would be inordinately difficult and prob-
ably not even possible. A more interesting question for businesses and regulators
to discuss is whether there can be a market for personal information and how that
might be balanced against privacy concerns. Can individuals decide to sell varying
parts of their information and get value in return? And what is the market price for
such data? These are open questions that deserve further attention.

IoT, big data, and 3D printing all hold major promise for future disruption but it is
difficult to separate the substance from the hype. The regulatory questions and privacy
issues attached to IoT, especially related to peoples health, are considerable. Perhaps
we will first see IoT being introduced in industry and especially in agriculture where
it is already in use and improving productivity.

Another major obstacle for the benefits of digitalization is the lack of skills. Education
systems need to improve and be more responsive to the needs of business. Periods of
rapid change also motivate greater focus on in-house training and life-long learning.
The notion of one education followed by a lifetime at the same job is more incongruent
with each year of increasing life expectancy in Sweden and elsewhere, a topic we dis-
cuss further in the next chapter.

170
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

5. I robot, you employed, he or she

freelance
In the twenty-first century, stable, long-term employment with a single employer will no
longer be the norm, and unemployment or underemployment will no longer be a rare and
exceptional situation. Intermittence will increasingly prevail, with individuals serving as wage
earners, freelancers, entrepreneurs, and jobless at different stages of their working lives.
Colin and Palier (2015).

5.1 Introduction
In this chapter we connect some of the issues in this report to discuss challenges for
education, lifelong learning and what this may imply for future work and inequality.
Sweden has fallen behind in a succession of PISA studies and skills improvement will
be important in the future to Swedish competitiveness in the world economy. Educa-
tion and inequality are big topics, of course, and for reasons of brevity, we will discuss
them primarily in the context of digitalization. But given the risk that more automation
may increase unemployment and inequality, this focus brings some clarity to choices
concerning regulation and taxation that may make a profound difference for welfare
in the next decades.

The effects of digitalization on the economy are broad and we have touched on some
of the main developments in this report. In the final section, we will also outline some
policy issues that are crucial to the smooth development of welfare while adjusting to
a more digital economy. In particular, what features stand out that would mitigate the
risk of rising unemployment and inequality? Highlighting the areas that matter most
is worthwhile, as this also excludes areas that are less significant and indeed indicate
policies that might even be counterproductive.

But first we will begin with a brief overview of Swedens position as an IT nation and
the features that facilitated a strong standing despite being a small country on the
periphery of Europe. Will these strengths endure or will other countries catch up?

5.2 Sweden as a great IT nation, is the past any guide?

How did Sweden become a prominent digital nation?
Sweden is among the countries with the highest internet penetration in the world.
Fixed broadband is also widely available, though the goal of full coverage has not
been achieved and would be prohibitively expensive in sparsely populated areas.

Was the success due to explicit policies or was Sweden just lucky? Frick(2015)
discusses the events that led to Swedens current position:
Starting in the 1970s, certain key individuals played a significant role in promoting
standards and technology
Among the significant events that gave Sweden a head start was the role these
individuals played in developing internet protocol (IP) standards

171
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The Commodore 64 with 64 kB of memory was hugely popular among early

geeks who wanted to tinker and experiment with code (including the author
ofthis report)
The Compis school computer introduced in the mid 1980s flopped, but may have
had positive long term benefits by contributing to the wider spread of digital curiosity
The Home PC reform subsidized household purchases of computers and the
government made broadband for all a political priority, which, although it met
with some success, did not accomplish its goal
Symbolism at the highest political level when Swedish Prime Minister Carl Bildt
sent an email to US President Bill Clinton in 1994

Swedens IT history is thus not a straight and self-evident path to its destination as a
successful IT nation exemplified by firms such as Mojang, Skype, and Spotify. Instead,
a combination of various initiatives, often with unclear short-term benefit, led to long-
term digital maturity. Frick (2015) stresses the importance of certain key individuals
who played a central role from the 1970s and onwards, among them Jacob Palme,
Bjrn Eriksen, Peter Lthberg, and Ragnar Lnn. Also, the period before Sweden
acceded to the EU was characterized by a mixture of competition and cooperation
between private companies, notably Ericsson, and the state-owned telecom company, in
which KTH, the Royal Institute of Technology, played a vital role, see Kaijser(2015).

Benchmark measures of Swedens standing as IT nation

Sweden typically ranks among the leaders in IT development, sharing the top five
or ten positions with Finland, Denmark, Norway, South Korea, the UK, the US, and
New Zealand; see SOU (2014a, page 37). The relative rankings of these leading coun-
tries have moved only slightly in the last few years and are fairly stable. Sweden
remains among the top countries.

Sweden has attracted increasing attention as an IT nation; see for example Powell
(2012) and Milne (2014). Established industry has a long tradition of engineering
and excellence through ABB, Electrolux, Ericsson, Volvo, and Scania. The largest IT
company in Sweden may well be Volvo. As regards new firms in the digital economy,
Facebook has located its data center to Lule where it benefits from low energy
costs and a favorable climate for computers that keeps cooling costs down. The data
centers provide considerable investment in the local area, estimated by BCG (2014)
at SEK 9 billion in total and involving 4,500 full-time workers over the course of
ten years throughout the country. However, once built, data centers typically do
notrequire a lot of people in their day-to-day operations.

For a small country, Sweden has produced quite a few successful IT startups. Otherwise,
the US and China tend to dominate in digital startups that become very successful; see
Foley(2015). The relative success of Swedish startups is explained by a combination of
factors. Strong institutions and strong technical community gave Sweden a head start.
As a relatively small market, digital Swedish companies often need to consider global
expansion from the very beginning; if the company is successful, fewer steps are needed
to expand globally. Moreover, the success of the first wave of Swedish digital firms
created a strong brand name, attracting more venture capital; see Ahmed (2015b).

Overall, Swedens standing as a tech nation is well beyond its size in the world
economy. But the features that led to this strong development might not necessarily
endure. For one thing, other nations are catching up. The value of the initial head

172
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

start with Internet Protocol (IP) and network standards will not last forever, especially
in the context of the European Union where discussions are underway to establish
standards in a variety of different areas, notably in mobile broadband and reducing
roaming charges on smart phones.

The increasing dominance of platforms, such as Android and iOS, imply that signifi-
cant network benefits accrue to the US by default. When digital companies in smaller
countries find pockets of inefficiency and expand, they will in many cases have to rely
on platforms built in Silicon Valley. A recent survey by Chakravorti et al. (2015) puts
Sweden among countries that may either succeed or stall in digital expansion, along
with the UK and Canada. One should not give too much credence to the inevitability
of decline. The point is rather that Swedish success cannot be taken for granted and
Swedens future standing depends on how institutions adapt and how regulation is
updated to accrue greater benefit from digitalization.

5.3 Human capital, digital skills and lifelong learning

Skills in core subjects, such as mathematics, language, and science, is important in
thelabor market and vital for highly skilled jobs. The lack of a high school/secondary
education makes entry into the labor market more difficult and dampens lifetime
wage prospects. As discussed in Chapter 3, there are still going to be many jobs that
require less formal education, but the possibility of making a career with increasing
lifetime prospects will be correspondingly harder. The same holds true to an even
greater extent the higher up the skills ladder one aspires.

While core subjects are the foundation of requisite skills, digital skills are becoming
increasingly important. The number of jobs that require interaction with computers
and smart phones is on the rise and includes everyone from truck drivers to baristas
and hairstylists. Indeed, there will be few jobs that do not require some form of dig-
ital skill, such as handling electronic booking systems, giving feedback en route, or
taking advantage of tools that increase productivity at work.

What is the current level of digital skills?

Both the European Commission and the OECD measure the level of digital skills
among member states. OECD (2014b) has found large variation, with high internet
usage among the adult population in Luxembourg, the Netherlands and the Nordic
countries (including Sweden), but low rates in southern Europe, notably Greeceand
Italy. The Nordic countries are also among the highest in measurements of user
sophistication; OECD (2014b, page 81).

Eurostat measures digital skills and in the 2014 survey continue to find that poorer
EU countries dominate with no or low digital skills; see Figure 5.1. Another
glaring feature of digital skills in Europe is the large number of people identified as
having basic skills. Sweden is among the countries with relatively good skills, with
about 80 percent of the population in the basic or above basic categories. While
such categories are sometimes sensitive to cutoff points and definitions, the overall
view is that there is considerable room for improvement. The middle category of
digital competence may be particularly vulnerable to the next wave of automation.
The countries with lower GDP per capita in the EU also tend to be those with lower
digital skills, making the structural change ahead more difficult for those countries.

173
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 5.1 Digital skills in the EU ( percent of individuals), 2014

Above basic
100%
Basic
90%
Low
80%
No skill
70%

60%

50%

40%

30%

20%

Source: Eurostat.
10%

0%
IS DK NO LU NL SE FI UK DE FR EE BE SK AT CZ IE LV ES HU SI MT LT HR CY PL PT IT EL BG RO

The first publication of the OECD PISA assessment of digital skills on 15 September
2015 give cause of concern for Sweden, as did the poor results in the overall survey;
see OECD (2015d). The general findings are that:
Even countries that have invested heavily in ICT for education have not seen any
noticeable improvement in reading, mathematics or science
The gap between advantaged and disadvantaged children is similar to the gap in
traditional PISA tests
Students who use computers extensively in school do much worse in PISA even
when adjusted for social and demographic background, suggesting that computer
time is not being spent productively

These findings are worrisome and one conclusion emphasized by OECD is that to
reduce inequalities, it is imperative to reduce knowledge gaps in traditional fields first,
notably in mathematics, writing and science. Swedens performance in digital reading
declined markedly between 2009 and 2012; see OECD(2015d, page 22), and is at
about median performance in the OECD despite higher than average internet access.
Swedens decline in digital skills thus mirrors its downward trend in the overall PISA.

Another key lesson from the OECD report is that improving digital skills is not a
matter of spending more public money, but how that money is spent. This result is
consistent with Vigdor and Ladd (2010), who report no improvement in skills for
disadvantaged groups despite subsidies.

Digital skills and the labor market today

In 2013, just under 40 percent of individuals in the EU labor force on average assessed
their own computer skills as adequate to allow them to change jobs within a year.
Here as well, the Nordic countries and the Netherlands are on top. Just below 60 per-
cent of Swedish respondents judged their digital skills to be sufficient for a job change.
The lack of IT skills is, as expected, especially concentrated among people with low
education; see OECD (2014b, page 89). OECD (2012) finds that people with the
lowest skills are 1.8 times more likely to be unemployed and 1.4 times more likely
compared to a reference group.

174
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

It is likely that small improvements in skills might make a big difference for employ-
ability and wage growth for people who have more rudimentary skills. For example,
commercial trucks a physical good are increasingly being transformed into services
rendered through IT. Drivers who are more able to use this technology will have an
advantage,122 a factor that is likely to hold across many areas in the labor market.
Experience from an online labor market shows that even small improvements in
skills and reputation can have significant impact on the ability to get contracts; see
Agrawal et al. (2015).

The lack of digital skills is of course only one aspect of human capital, but remains
important. OECD (2012) estimates that 25 percent of firms in member states are
concerned about the availability of adequately trained workers, with markedly worse
numbers for Africa, the Pacific, and East Asia. MGI (2015b) discusses the persistent
mismatch between jobs and skills at the global level, with about 3045 percent of
the adult population unemployed. Poorer countries especially would benefit from
increasing digital skill levels in the population combined with infrastructure that
supports digital communication. With online platforms for work, such as Upwork,
people from all open economies can supply their services even if there is no demand
in the local market. Access to such online platforms could thus provide substantial
benefits, but will of course not solve other structural problems in developing economies.

Digital skills and inequality

Inequality as measured by Gini coefficients has increased somewhat in recent years,
but Sweden remains among the countries with the lowest level of inequality among
all OECD countries. In the OECD as a whole, the gap between rich and poor is at its
highest level in 30 years; see OECD (2015e). The biggest driver of inequality is unem-
ployment especially long-term unemployment. Moreover, unemployment spells that
last longer than roughly six months substantially increase the risk of eroding human cap-
ital and self-confidence and creating a vicious cycle that further damages employability.

This matters a great deal during regular business cycles but is likely to become even
more significant for overall welfare in periods of rapid technological change when the
difference between having the right skills or inadequate skills can result in unemploy
ment or slow real wage growth. Since wage growth should follow productivity, people
who fall behind in skills may have poor lifetime earning prospects. Notably, in the
US this has already occurred for broad groups of low and middle income earners
in recent decades (see Chapter 3) with increasing tension, especially among lower
paying jobs (the working poor). A number of people in low-paying jobs in the US
are also receiving food stamps, thus a de facto subsidy from the government to low-
wage employers, for example in fast-food restaurants; see Cohen (2015) and Jacobs
et al. (2015).

New evidence from Norway shows that broadband adoption in firms complements
skilled workers but replaces unskilled workers performing routine tasks; see Akerman
et al. (2015). Other findings indicate that lower income children fall even further behind
in IT skills and this occurs even when they are provided subsidized computers. Although
these children improve their general IT skills, research shows that grades tend to go
down; see Stross (2010) and Vigdor and Ladd (2010). One possible explanation for
this outcome is that in environments with little parental support for learning, the
computers are used too much for entertainment; see also Richtel (2012). This result is
also reinforced by the results of the digital PISA study in OECD (2015d).

122
For example, trucks that have a black box, similar to that of airplanes, can give feedback on driving/fuel efficiency
in real time.

175
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

This implies that simple policies aimed directly at children in poorer families do not
work on their own and need to be supplemented by other tools or redesigned. It is
beyond the scope of this report to discuss specific policies in schools, but it stands
to reason that if the environment for studies is inadequate at home, perhaps more
organized homework should be done in school at the end of the school day. OECD
(2015d) emphasizes that the digital divide in skills between disadvantaged and advan-
taged children reflects differences in traditional skills in mathematics, language, and
science. Notably, more ICT in schools is not enough by itself and the results of the
OECD study highlight the importance of teaching children to be critical judges of
online content. Moreover, teachers and parents also need to mitigate the risks of internet
use related to bullying, privacy concerns and over-consumption of entertainment. Using
the computer for more than six hours per day was associated with markedly worse
academic performance, not unexpectedly since that means less time is spent on
homework.

General skills and digital skills are key to job opportunities and better lives. Professor
Alison Woolf has argued in series of books and articles that the general drive for
more education has resulted in a quality problem; see for example Wolf(2002, 2011).
One argument is that a relaxation of selection criteria and grade inflation have led
to lower educational standards and overly optimistic expectations of job opportunities
after graduation. Many graduates find jobs for which they may feel they are overqual-
ified or in fact are overqualified. Or in the words of Beaudry et al. (2013), ...having
a BA [bachelors degree] is less about obtaining access to high paying managerial and
technology jobs and more about beating out less educated workers for the barista or
clerical job.

The overqualified barista seems so far to be primarily a US problem. For Sweden,

the results in Adermon and Gustavsson (2015) indicate that job polarization has
primarily involved an increase in highly paid jobs and not the increase in low paid
jobs observed in other OECD countries; see chapter 3. While not conclusive, it is at
least indicative that the overeducated barista is not a structural problem in Sweden.
Notably, the outcomes of wage negotiations depend on the institutional formsfor wage
bargaining. The institutions in Sweden are geared towards setting wages according to
productivity growth, especially in the export sector subject to global competition.
There are also strong elements of equity concerns that so far have likely contributed
to preventing wage polarization, but the serious mismatch in available skills identi-
fied by firms, especially in highly skilled workers, is a warning of potential changes
inthe labor market. The lack of workers with the right skills combined with aging
populations may increase the incentive to automate work.

Digital skills in the future, complements or substitutes?

Economic history has exhibited periods of deskilling, where workers with lower
skills using machines have replaced humans, as well periods in which technology has
enhanced and complemented human work, see Chapter 3. Although the last few dec-
ades have been dominated by complementarity between work and machine, this is
not a law of nature.

What does this mean for the labor market? Brynjolfsson and McAfee (2014) argue
that those with low to medium skills face dimmer prospects in the labor market
unless their skills imply productivity higher than machines. Those whose productivity
is lower than that of the machines risk poor real wage growth or unemployment
unless they can improve their skills.

176
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Having the right skill has of course been imperative for good prospects in the labor
market for a long time but two trends are now changing the nature of this dynamic.
The speed of technological change implies that the best-before date for certain
skills acquired in school comes sooner in our working lives. In addition, the fact that
we live longer implies a longer share of our working lives may be lived with a rusty
skill set, especially when it comes to digital skills.

A key lesson from the recent OECD (2015d) survey is that more computers are not
always better and that some students have done worse in other subjects when they
use computers too much or unproductively. This taken together with the results of
Akerman et al. (2015) showing that digital skills tend to become complementary for
advantaged groups, but not for disadvantaged, groups could well be an indication
offurther polarization in the labor market.

The Swedish Digitalization Commission emphasizes the variety of digital competence

in its report SOU (2015a), including digital skills required to interact socially, with
local government, and at work. The report also stresses the need for lifelong learning
and improving digital skills. It also notes that computers are used widely in the private
sector and that this has increased markedly in the last decade. More than 98 percent of
firms with more than ten employees use computers and digital tools. Earlier reports
on the challenges facing smaller and medium sized firms highlighted the lack of skills
and the time it takes to acquire those skills as a problem.

While we concur with the recommendation on lifelong learning in SOU(2015a), the

figures presented above on computer use in business may provide false comfort.
Having a computer at work may mean using it for anything from writing documents
only (as an advanced typewriter) to using it for 3D modelling or tweaking production
processes. If computers are used as in the former case, Professor Robert Gordon would
be right about the future of productivity growth; see Chapter 1. The benefits of elec-
tronic documents have already been gained long ago and will not budge future growth.

Instead, we need to find ways to replenish knowledge, especially digital skills, throughout
our working lives. But the benefits of digital skills will be hampered unless the school
system as a whole is able to reverse its negative trend. As highlighted by OECD (2015d,
f), digital skills do not exist in a vacuum and need to be supported by other knowledge.

How we improve skills and lifelong learning, again, especially digital skills, is a topic
beyond the scope of this report, but some general comments can be made. With longer
working lives, it is more reasonable to spend time in education throughout lives, either
to acquire new knowledge or prepare for changing careers; see Blix (2013a, b) for an
extensive discussion of this. The more general the skills, the more reasonable it is to
have public financing as the main option. Sweden is already one of the countries that
places the most emphasis on adult education and thus has many of the elements in
place that need to be utilized. For skills that are more specialized or firm-specific, pri-
vate sector funding should continue to be the main source of financing in conjunction
with more favorable tax treatment of the investment.

The UK Tech Partnership is one example of public/private cooperation aimed at raising

skills in firms and boosting digital learning in schools, especially for girls. The initi-
ative is something that could very well be explored for Sweden. Anecdotal evidence
from the US suggests that people who have general skills and upgrade them with some
digital training in areas like Web design or programming, improve their chances of
finding well-paid work; see Manjoo(2015c). It is likely that people with skills in

177
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

specific areas can strengthen their chances in the labor market by improving their
digital skills and that this lesson holds widely true in the economy as a whole.

To highlight the importance of improving skills in the workforce, the tax system
could be much more supportive of its treatment of investments in human capital.
More specifically, investment in physical capital and human capital can be put on
more equal footing; see Almega (2014). Today, firms have to expense the costs of
investments in training education the same year and cannot spread the cost across
several years as is possible with other types of investment.

5.4 Risk sharing in the economy, digital version

How many jobs will be automated in the future is a key question in the debate, the
answer to which will have impact on the extent of job and wage polarization. In
Chapter 3 we discussed the various forces that will accelerate or dampen the speed
ofautomation, including profits and institutional inertia. From an overall perspective,
demographic trends may have huge impact.

In countries with aging populations, such as Sweden, the rate of entry into the labor
force is slowing down compared to exit; see Blix (2013a,b)123. Moreover, firms report
a skills mismatch for available jobs. These developments will increase the incentive to
automate tasks but also make adjustment easier since there will be fewer entrants to
the labor market searching for fewer available jobs. At the same time, the matching
of skills and available jobs may continue to be an issue and is a challenge to make
theperiod of adjustment less rugged in terms of risks to social welfare.

Young people entering the labor market will have higher digital skills than their
immediate predecessors but this may not be enough as long as polarization continues
in the labor market. The middle-income earners have been reduced as a group and
many of the tasks they perform are increasingly amenable to automation. With some
oversimplification, one can envisage a risk that the labor market will continue to
polarize in the years to come, with entrants either to the right of the skills scale with
good prospects for a strong wage trajectory and others who risk experiencing stagnant
productivity and wage growth.

The key to welfare during this period of change is a flexible, equal opportunity labor
market. More overall agility would help smooth structural change and reduce risk of
higher structural unemployment during the transition. The sharing economy discussed
throughout this report can provide some of the agility needed if it is allowed to
flourish. It will allow people to shift between freelancing/education/regular employ-
ment; see Hall and Krueger(2015). And when regular employment is not available,
itwill also provide a source of income that reduces the risk of increasing inequality.

For the sharing economy to function well, the safety nets and social security systems
need to be adapted. The systems in Sweden and many other countries are designed
around full time employment at one institution or firm. People working freelance
asself-employed or in small companies do not fit easily into this box, a topic discussed
in several reports and governmental commissions.

123
This is discussed on page 133 in the Swedish text and page 139 in the English version.

178
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The issue is not that the risk of being an entrepreneur should be taken down to a
lower level, closer to that associated with being an employee. Self-employed people
take higher risk for a variety of reasons, such as being driven by an idea, preferring
more freedom, or seeking the rewards that can come with success. The issue is rather
that the asymmetry between self-employment and being an employee is particularly
large in Sweden. The self-employed pay into social security to qualify for the universal
safety net, but are not able to draw on those benefits to the same extent because
the entire structure is built around employee status, with all the concomitant rights
and benefits (parental leave, paid vacation, additional pension rights, job security).
According to one calculation, employees get 88 percent of the money back over a life-
time of the social security contributions paid by their employers; the corresponding
figure for the self-employed is about 49 percent; see Sandanji(2012); see also SOU
(2015b, page 962). Due to the nature of business commitments in small firms,
self-employed people are less able take sick days or parental leave since there may
beno one else to take up the slack. The effects on the bottom line might be lower
profits or indeed endanger the long-term viability of the firm.

The same holds true for parental leave. Self-employed people can, in theory, be on
parental leave on weekdays, perhaps with some limited work in the evenings, and
try to catch up on work during the weekends with some help, e.g. from a spouse, but
then they get lower benefits. To get full benefits, parents must use seven days of leave
per week, including the weekends, which is typically not sustainable for someone
trying to keep a business afloat; to work on a day for which parental leave benefits
are paid is a breach of the rules. Notably, the lower number of self-employed women
is probably largely explained by the asymmetry between employee vs self-employed
status in social security; see Sandanji(2012) and Fretagarna (2014).

There are also other differences that may be significant, especially for newly started
businesses. In the event of sickness, payment is based on the so called SGI measure of
income. For the self-employed, this measure is calculated on the basis of the equivalent
pay an employee would receive but there is an inevitable area of judgment involved
on the part of the authorities. To reduce this uncertainty, the boundaries between
self-employed, employee and small incorporated firms would need to be better defined
but this is difficult to do in practice and may create unintended side-effects; see SOU
(2015b, page 963966). The uncertainty may be the greatest in the transition from
employee to self-employed when the risk of failure of the firm may also be highest;
SOU (2015b, page 962). Overall, the benefits systems are thus geared towards being
an employee and people who are self-employed have to pay twice if the want to
reduce the level of risk by investing in supplementary private insurance.

Despite paying twice for insurance, private insurance has become increasingly popular
in Sweden; see Figure 5.2. The ability to get faster services is often important to small
business owners for whom absence is especially problematic.

179
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Figure 5.2. Private health care insurance increasing, number of persons

700 000

600 000

500 000

400 000

300 000

200 000

Source: Insurance Sweden.

100 000

0
2000-12-01

2001-08-01

2002-04-01

2002-12-01

2003-08-01

2004-04-01

2004-12-01

2005-08-01

2006-04-01

2006-12-01

2007-08-01

2008-04-01

2008-12-01

2009-08-01

2010-04-01

2010-12-01

2011-08-01

2012-04-01

2012-12-01

2013-08-01

2014-04-01

2014-12-01
Unionization and reducing risks, digital version
Organized labor, unions, originated from the desire to improve working conditions in
industry and services. The same forces are now at work in the digital era. A number of
organizations are developing to support and give guidance to freelancers in the sharing
economy, including www.faircrowdwork.org in Germany and www.peers.org in the
US. There are also various digital services that help freelancers improve their income,
pay taxes, smooth income and cash flow, such as www.sherpashare.com and even.
me. As freelancing is rising in importance, these organizations and the services they
provide are trying to meet a demand for help and security not met elsewhere; see
Giridharadas (2015), Singer and Isaac (2015).

Through the app Even, for example, freelancers can even out their income over time
for a fee, not unlike the idea behind risk-sharing provided by the public sector in
many countries. These developments may indicate a search for better risk-sharing
in the economy between people in regular employment and freelancers. In Sweden,
social security benefits are supplied through contributions in the tax system and
many of the other factors affecting work are decided in collective bargaining agree-
ments between the unions and employers organizations. The Swedish system will not
be immune to the need to reexamine how risk is shared in the economy and between
generations. The legitimacy of the system may hinge on large groups not being
excluded or less fully covered by the established social safety nets.

Surveillance and ethical issues at work

Also for employees, age-old issues about monitoring at work are coming into a new
light. It is now possible to monitor and survey employees to a degree hitherto impos-
sible. More specifically, digital tools can provide data about performance, health,
anxiety, and other issues, 24/7. For example, reports have highlighted working con-
ditions at Amazon where employee performance is monitored and assessed contin-
uously in real time; see for example OConnor(2013) and Kantor and Streitfeld

180
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

(2015). An editorial in the Financial Times (August 17, 2015) posits that white-collar
workers at Amazon choose to work in the competitive environment it embodies in
exchange for the potential monetary rewards for those who succeed or the experience
that might help open other job opportunities in the labor market. Sandbu (2015b),
also writing in the Financial Times, points out that the argument about being freeto
leave hinges on such practices not becoming the common standard elsewhere. Of course,
employee monitoring has always been a feature of competitive workplaces, including
banks, law firms, and other workplaces where high pay and benefits accrue to those
who make the grade. Sales forces, logistics professionals, and other sales-oriented pro-
fessions have a long tradition of being measured on performance, since measurement
has been easy and is directly related to performance.

Digital monitoring, however, can take surveillance to a new and potentially prob-
lematic level. Notably, when employees wear digital devices throughout the day,
employers can become privy to their levels of anxiety, sleep deprivation, and other
factors that may affect performance; see for example OConnor(2015c), Scheiber
(2015) and the Economist (2015h). In an experiment, 31 employees at the firm Pro-
fusion volunteered to wear fitbits for ten days. The data generated made it possible to
divide the employees into various categories (such as busy and coping and irritated
and unsettled); see OConnor(2015c). How much of this will be voluntary and what
invasions of privacy are we prepared to accept? While the practices mainly come from
American companies, the tools thus developed are beginning to be used in many places.
In Sweden, municipalities have used the Parag system in smart phones to detect fraud
or inefficiencies; see Eriksson(2015). For countries in Europe with other traditions in
law and privacy issues, the questions raised should concern everyone: employees, firms,
unions, and employers organizations alike. The boundaries between work and leisure
have become more blurred with the use of smart phones and email and digital tools
are now taking these issues to another level.

Strong pressure for automation in public sector administration due to aging populations
We have not discussed the outlook for employment in the public sector in this report,
nor the challenge of financing public welfare, which is a very broad topic and a challenge
for most OECD countries. While the health care sector is likely to need more labor in
the future, especially in providing elder care services, the overall cost pressures on sup-
plying those services will likely imply strong needs to cut costs elsewhere, especially in
administration where it is likely that much more automation is possible (see Chapter
3). To ease such a process, it will be especially important to provide opportunities for
the people affected to learn new skills throughout their lives so that they can remain
competitive in the labor market. Facing middle age with dwindling job prospects in a
shrinking public administrative sector is not a far-fetched scenario and action to avoid
and smooth the process would be desirable.

5.5 A future with jobs

The position taken in this report is that there will be work in the future, just as there
has been in the past. But the path may be more or less challenging for people in terms
of welfare and risk, depending on how our institutions react. The continued thinning
of the middle class and further job polarization are likely but not inevitable. Bad
policy and protectionist responses exacerbate the risks of bad outcomes for work and
welfare, and may even lead to the onset of wage polarization in Sweden in contrast
to the last two decades of high real wage growth.

181
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Our social welfare is affected by many factors. Digitalization is one of the major trends
affecting the economy, but not the only one. Industry has been working for a long time
to streamline productivity by using more and better machines, a trend that digitalization
continues. Rising incomes have implied demand for new services, all from household
sector areas, from more forms of entertainment to personal trainers. In some areas, such
as the financial sector, increased complexity of regulation may actually create work (for
lawyers) even as other segments of banking are shrinking. While each sector may be
affected by the different trends discussed in Chapter 1 (globalization, the shift from
goods to services, technology, and urbanization), the overall trend towards aging pop-
ulations will have gradual but significant impact on the entire labor market. In par-
ticular, it will not be long before fewer young people are entering the labor market
than the number of retirees leaving.

Together with the challenge of skills mismatch in the labor market, demography
will increase the need for automation in some areas over and above the incentive to
improve and streamline work. In some areas, demography may be the most important
factor, such as health care. In Chapter 3 we discussed the rapid pace at which auto-
mation is making it possible to replace labor even skilled labor in many areas.
But the speed of automation is likely to be slower than what is technically feasible
because the pace will be slowed by regulatory obstacles as well as habits and prefer-
ences (see Box 3.2. Demand-related issues with further automation). It would be
a serious mistake to view such obstacles as a way to rescue work; indeed, the situa-
tion is quite the opposite. With protectionist responses and slow regulatory overhaul
that prevent realizing the benefits of digitalization, the results would be worse on all
fronts: slower productivity growth, higher risks of technological unemployment, and
rising inequality (as illustrated in Scenario 2 in Box 2. Scenarios for Sweden in the
Extended Summary). The key to prosperity has always been to ensure good conditions
for new jobs to be created in the private sector. And that is not going to change.

A future of work without further rise in inequality is possible but needs help from
policy (Scenario 3 in Box 2 in the Extended Summary). Five general points stand out
that would make structural change more unrugged:
Reduce taxes on labor (Chapter 3). Expand tax reductions for household-related
services (the RUT and ROT programs in Sweden) to cover more areas than today,
including professional services such as IT support, in line with a proposal in SOU
(2015a)
Embrace the sharing economy and the flexibility it implies by reducing regulatory
uncertainty. Allow competition to flourish with emphasis on lowering unnecessary
standards for all rather than by raising the bar for entrants (Chapters 3 and 4)
Take more strategic steps towards lifelong learning and improving skills, especially
digital skills. Focus on quality rather than quantity (this chapter)
Reduce the asymmetry between employee/freelancer in terms of risk by making it
easier for freelancers to use and benefit from social security (this chapter)
Establish principles of regulatory overhaul that can be used to apply to each indi-
vidual area to speed the process (Chapter 4)

182
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The last point concerns productivity growth from new technology (IoT, 3D printers,
etc.), which may be held back unless regulatory obstacles are removed faster than the
usual slow pace. Against the background of macroeconomic headwinds discussed in
Chapter1, it is imperative that the microeconomic innovations from digitalization
are allowed to counter the effects of aging populations and rising public debt that
may check productivity growth; see for example Blix (2015).

The reasoning behind the other points is relatively straightforward, even though the
mechanisms at work may be complicated. The basic idea is to make people more
employable and earn higher incomes after tax while simultaneously reducing the risk
of rising inequality. The thrust of policies should be aimed at countering the risk that
the rapid pace of change will result in large groups falling behind and becoming less
productive than machines, as argued in Brynjolfsson and McAfee(2014). If skills
are improved throughout peoples working lives and taxes on labor are reduced, the
incentive to automate will be weaker. Higher skills then imply that more people will
complement rather than replace machines (see Section 3.3 in chapter 3).

The period ahead in the labor market may still be tough, but would be eased by more
flexibility. The greatest risk of rising inequality lies in increasing and lasting unemploy-
ment. In this regard, the rise of the sharing economy that allows people more freedom
to work and combine work with parenthood or studies will be a crucial element. While
the sharing economy has become significant in the US, it may be even more so in
European countries where there are strong protections for labor market insiders. For
the sharing economy to serve this function of increasing flexibility without worsening
inequality, the risks of freelance work should be addressed in the social security systems,
not by according employment status but by reducing the inherent asymmetries that
favor employees over the self-employed.

If all these policies are implemented, we are likely to see increased productivity growth
and continued high employment without increasing inequality. There will be jobs in the
future, but our institutions need to address the challenges to make it happen.

I robot, you employed, he or she freelance...

183
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

References

Acatech (2015). Recommendations for the Strategic Initiative Web-Based Services

for Businesses. Smart Service Welt and Acatech, final report. March 2015.

Acatech (2013). Recommendations for Implementing the Strategic Initiative

Industrie 4.0. Final report of the Industrie 4.0 working group. April 2013.

Acemoglu, Daron and James Robinson (2013). Why Nations Fail. London: Profile
Books.

Acemoglu, Daron, Simon Johnson and James Robinson (2005). Institutions as

aFundamental Cause of Long-Run Growth. In Philippe Aghion and Steven N.
Durlauf, Handbook of Economic Growth. Volume 1A. Amsterdam: Elsevier.

Acemoglu, Daron (2002). Technical Change, Inequality, and the Labor Market.
Journal of Economic Literature 40(1). 772.

Acemoglu, Daron (2003). Labor- and CapitalAugmenting Technical Change.

Journal of the European Economic Association 1(1). 137.

Acemoglu, Daron and David Autor (2011). Skills, Tasks and Technologies: Implica-
tions for Employment and Earnings. Handbook of Labor Economics 4. 10431171.

Aeppel, Timothy (2015). 50 Million Users: The Making of an Angry Birds Internet
Meme. <http://blogs.wsj.com/economics/2015/03/20/50-million-users-the-making-
of-an-angry-birds-internet-meme/>. March 20.

Adermon, Adrian and Magnus Gustavsson (2015). Teknisk utveckling och jobb-
polarisering. SNS Analys No. 28. March 2015. Stockholm: Centre for Business and
Policy Studies (SNS).

Aghion, Philippe and Peter Howitt (2006). Joseph Schumpeter Lecture Appropriate
Growth Policy: A Unifying Framework. Journal of the European Economic Association
4(23). 269314.

Agrawal, Ajay, John Horton, Nicola Lacetera, and Elizabeth Lyons (2015). Digitaliza-
tion and the Contract Labor market: A Research Agenda. In Avi Goldfarb, Shane M.
Greenstein, and Catherine E. Tucker, eds., Economic analysis of the digital economy.
National Bureau of Economic Research Conference Report, May 2015. Cambridge,
MA: National Bureau of Economic Research.

Ahmed, Murad (2015a). Facebook Understands You Better than Your Spouse.
Financial Times. January 12.

Ahmed, Murad (2015b). Swedens Unicorn Factory. Financial Times. March 31.

Akerman, Anders, Ingvil Gaarder and Magne Mogstad (2015). The Skill Comple-
mentarity of Broadband Internet. NBER Working Paper No. 20826. Cambridge,
MA: National Bureau of Economic Research.

Allen, Richard (2015). European Disintegration Threatens Business on the Internet.

Financial Times. April 29.

184
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Alloway, Tracy and Emma Dunkley (2015). Rebranded P2P Revolutionaries Form
Ties with Traditional Lenders. Financial Times. January 29.

Almega (2014). Kunskap fr konkurrenskraft bttre villkor fr nringslivets

investeringar i kompetensutveckling. <http://www.almega.se/politik-och-ekonomi/
rapporter/kunskap-for-konkurrenskraft >.

Ambrosini, Vronique and Cliff Bowman (2001). Tacit Knowledge: Some Suggestions
for Operationalization. Journal of Management Studies 38(6). 811829.

Andersson, Fredrik (1999). Job flows in Swedish manufacturing 19721996. Institute

for Evaluation of Labour Market and Education Policy. Working paper. No. 1999:4.

Apel, Mikael, Eric Frohm, Jyry Hokkanen, Christina Nyman and Stefan Palmqvist
(2014). Resultat frn en enkt om fretagens prissttning. Ekonomiska kommentarer
2014, nr 4. Stockholm: Swedish National Bank.

Arnold, Martin and Tom Braithwaite (2015). Banks ageing IT systems buckle under
the strain. Financial Times. June 19.

Arnold, Martin (2015). BBVA head says lenders must replace spaghetti. Financial
Times. June 19.

Ashenfelter, Orley C. (2012). Comparing Real Wages. NBER Working Paper

No.18006. Cambridge, MA: National Bureau of Economic Research.

Autor, David H. and David Dorn (2013). The Growth of Low-Skill Service Jobs
andthe Polarization of the US labor market. American Economic Review 103(5).
15531597.

Autor, David H. (2014). Polanyis Paradox and the Shape of Employment Growth.
Paper presented at the Federal Reserve Bank of Kansas City Jackson Hole Economic
Policy Symposium, August 2014.

Autor, David H., Frank Levy and Richard J. Murnane (2003). The Skill Content
ofRecent Technological Change: An Empirical Exploration. Quarterly Journal of
Economics 118(4). 12791333.

Baeck, Peter, Liam Collins and Bryan Zhang (2014). Understanding Alternative
Finance. Nesta and University of Cambridge, November 2014.

Bakhshi, Hasan, Carl B. Frey and M. Osborne (2015). Creativity Vs. Robots. Nesta,
April 2015.

Barroso Morin, Miguel (2014a). The Labor Market Consequences of Electricity

Adoption: Concrete Evidence from the Great Depression. Manuscript March, 2015,
University of Cambridge.

Barroso Morin, Miguel (2014b). General Purpose Technologies: Engines of Change?

PhD Dissertation, University of Columbia.

Bartelsman, Eric J. (2013). ICT, Reallocation and Productivity. European Economy,

Economic Papers 486, April 2013.

BCG (2014). Digital Infrastructure and Economic Development an Impact Assess-

ment of Facebooks Data Center in Northern Sweden, The Boston Consulting Group.

185
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Beaudry, Paul, David A. Green and Benjamin.M. Sand (2013). The Great Reversal
inthe Demand for Skill and Cognitive Tasks. NBER Working Paper No. 18901.
Cambridge, MA: National Bureau of Economic Research.

Belleflamme, Paul and Martin Peitz (2015). Industrial Organization: Markets and
Strategies. 2nd edition, Cambridge, UK: Cambridge University Press.

Benzell, Seth, Laurence J. Kotlikoff, Guillermo LaGarda, and Jeffrey D. Sachs (2015).
Robots Are Us: Some Economics of Human Replacement. NBER Working Paper
No. 20941. Cambridge, MA: National Bureau of Economic Research.

Bergh, Andreas and Christian Bjrnskov (2011). Historical Trust Levels Predict
Current Size of the Welfare State. Kyklos 64(1). 141169.

Bernstein, Amy and Anand Raman (2015). The Great Decoupling: An Interview With
Erik Brynjolfsson and Andrew McAfee. Harvard Business Review. June 2015. 6674.

Bessen, James (2015). Toil and Technology.Finance & Development 52(1). IMF
March 2015.

Blinder, Alan S. (2009). How Many US Jobs Might be Offshorable? World Economics
10(2). 41.

Blix, Mrten (2015). Produktivitetstillvxt hot och mjligheter. Ekonomisk Debatt

43(5). 5566. September 2015.

Blix, Mrten (2013a). Framtidens vlfrd och den ldrande befolkningen. Delutredning
frn Framtidskommissionen. Ds 2013:8. Stockholm: Fritzes.

Blix, Mrten (2013b). Future welfare and challenges from the ageing population.
Interim report from the Commission on the Future of Sweden. http://www.martenblix.
com/uploads/6/2/7/2/62723607/blix_future_welfare_and_the_ageing_population_
report_from_the_commission_on_the_future_of_sweden_2013.pdf

Bloom, Nicholas, Raffaella Sadun, and John Van Reenen (2012a). Americans Do IT
Better: US Multinationals and the Productivity Miracle. American Economic Review
102(1). 167201.

Bloom, Nicholas, Raffaella Sadun, and John Van Reenen (2012b). Does Management
Really Work? Harvard Business Review 90(11). 7682.

Bosker, Bianca (2013). Philip Parkers Trick For Authoring Over 1 Million Books: Dont
Write. <http://www.huffingtonpost.com/2013/02/11/philip-parker-books_n_2648820.
html.>. February 2013.

Bostrom, Nick (2014). Superintelligence Paths, Dangers and Strategies. Oxford, UK:
Oxford University Press.

Boumediene, Jimmy and Gran Grahn (2015). Produktivitetsutveckling, investeringar

och vlstnd. Samhllsekonomisk analys, The Confederation of Swedish Enterprise.
January.

Bower, Joseph L. and Clayton M. Christensen (1995). Disruptive Technologies:

Catching the Wave. Harvard Business Review, January-February issue 1995, 4353.

Bradshaw, Tim (2015). California Court Dents Uber Bodywork by Classifying its
Drivers as Employees. Financial Times. June 18.

186
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Breman, Anna and Anna Fellnder (2014). Diginomics. Ekonomisk Debatt 42(6).
2838. November 2014.

Bryant, Chris (2014). Volkswagen to Replace Germanys Retiring Baby Boomers

with Robots. Financial Times. October 6.

Brynjolfsson, Erik and Andrew McAfee (2014). The Second Machine Age. New York:
W.W. Norton & Company.

Buchanan, Matt (2014). Can you Uber a Burger? International New York Times.
October 25.

Bylund, Markus (2013). Personlig integritet p ntet. Stockholm: Fores

Byrne, David and Eugenio Pinto (2015). The Recent Slowdown in High-Tech Equip-
ment Price Declines and Some Implications for Business Investment and Labor Produc-
tivity. Board of Governors of the Federal Reserve System, FEDs Notes, March 26.

Cahuc, Pierre (2010). Det svenska anstllningsskyddet. Bilaga 6 till Lngtids

utredningen 2011. SOU 2010:93.

Cain Miller, Claire (2015). Can an Algorithm Hire Better Than a Human? New
York Times, < http://www.nytimes.com/2015/06/26/upshot/can-an-algorithm-hire-
better-than-a-human.html?_r=0>.

Calmfors, Lars (2013). Sweden From Macroeconomic Failure to Macroeconomic

Success, in Maguire, M. and Wilson, G. eds., Business and Government, Volume IV:
Challenges and Prospects. Routledge.

Chakravorti, Bhaskar, Christopher Tunnard, and Ravi Shankar Chaturvedi (2015).

Where the Digital Economy is Moving the Fastest. Harvard Business Review.
February 19.

Charles, Kerwin Kofi, Erik Hurst and Matthew J. Notowidigdo (2013). Manu-
facturing Decline, Housing Booms, and Non-Employment. NBER Working Paper
No.18949. Cambridge, MA: National Bureau of Economic Research.

Clerwall, Christer. (2014). Enter the Robot Journalist: Users Perceptions of Auto-
mated Content. Journalism Practice 8(5). <http://dx.doi.org/10.1080/17512786.201
4.883116>.

Cohen, Patricia (2015). Working, but Needing Public Assistance Anyway. New York
Times, April 12.

Cohen, Molly and Arun Sundararajan(2015). Self-regulation and Innovation in the

Peer-to-Peer Sharing Economy. The University of Chicago Law Review Dialogue,
82:116, P. 116133.

Colin, Nicolas and Bruno Palier (2015). The Next Safety Net. Foreign Affairs. June 16.

Comin, Diego and Mart Mestieri Ferrer (2013). If Technology Has Arrived Every-
where, Why has Income Diverged? NBER Working Paper No. 19010. Cambridge,
MA: National Bureau of Economic Research.

Corkery, Michael (2015). Pitfalls for the unwary borrower out on the frontiers of
banking. New York Times. September 13.

Cowen, Tyler. (2015). Average is over. New York: Penguin Group.

187
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Coyle, Diane (2015). Why GDP Statistics are Failing us. US Chamber of Commerce
Foundation, <http://www.uschamberfoundation.org/article/why-gdp-statistics-are-
failing-us.>

Croft, Jane (2015). The Internet of Things Keeps one Step Ahead of the Law.
Financial Times. January 28.

Crow, David (2015). Google and J&J Aim for Cutting Edge of Healthcare with
Robotic Surgeons. Financial Times. March 28.

Danish Enterprise (2010). The Value of Danish Address Data. Danish Enterprise
and Construction Authority. July 2010.

Dempsey, Michael (2014). Clever Tools Improve Stock Control. Financial Times.
October 22.

Davis, Steven J. and John Haltiwanger (2014). Labor Market Fluidity and Economic
Performance. NBER Working Paper No. 20479. Cambridge, MA: National Bureau
of Economic Research.

Davison, Nicola (2015). 3D-Printed Cities: Is This the Future? The Guardian.
February 26.

De la Merced, Michael J. (2015). Oscar, a Health Insurance Start-Up, Valued at $1.5

Billion. New York Times, April 20.

Dolan, Mike (2015). New Normal May Not Be Here to Stay. International New
York Times. 2015-05-28.

De Grauwe, Paul and Marianna Grimaldi (2006). The Exchange Rate in a Behavioral
Finance Framework, Princeton University Press.

Dellham, Daniel, Erik Lakomaa, and Richard Wahlund (2013). Konsumenterna

frlter inte dligt dataskydd. Aftonbladet. October 11.

Deloitte (2014a). Making a List, Clicking It Twice. Deloittes 2014 Annual Holiday
survey. Deloitte University Press.

Deloitte (2014b). The Changing Face of Retail Where Did All the Shops Go?
Retail Thought Leadership Series.

Dimon, Jamie (2015). Letter to shareholders in JP Morgan & Chase co, Annual
report, 2014.

The Economist (2015a). Workers on Tap. The Economist. January 3.

The Economist (2015b). Theres an App for that. The Economist. January 3.

The Economist (2015c). Made to Measure. The Economist. May 30.

The Economist (2015d). Crowdfunding. The Economist. May 9.

The Economist (2015e). Robochef gets cooking. The Economist. April 18.

The Economist (2015f). McJobs and UberJobs. The Economist. July 4.

The Economist (2015g). A question of utility. The Economist. August 8.

The Economist (2015h). Digital Taylorism. The Economist. September 12.

The Economist (2015i). A Bridge to the Future. The Economist. September 15.

188
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

The Economist (2015j). Banks v fintech. The Economist. May 9.

The Economist (2015k). Crowdfunding. The Economist. May 9.

The Economist (2015l). Peer-to-peer lending. The Economist. May 9.

The Economist (2015m). International banking. The Economist. May 9.

The Economist (2015n). Whom to trust. The Economist. August 15.

The Economist (2014a). Getting to Know You. Special Report, The Economist.
September 13.

The Economist (2014b). The Future of Jobs. The Economist. January 18.

The Economist (2014c). Digital Disruption on the Farm. The Economist. May 24.

The Economist (2014d). The Light Fantastic. The Economist. May 17.

The Economist (2014e). Arrested Development. The Economist. October 4.

The Economist (2014f). Tourists beware. The Economist. May 10.

The Economist (2014g). Unshackle the Entrepreneurs. The Economist. July 5.

The Economist (2012a). Big data. The Economist. May 19.

The Economist (2012b). Technology Quarterly. The Economist. September 1.

The Economist (2010). A Different Game Information is Transforming Traditional

Businesses. The Economist. February 25.

Edquist, Harald O. and Magnus Henrekson (2006). Technological Breakthroughs

and Productivity Growth Research in Economic History 24, 153.

Ek, Irene (2014). Digitaliseringens bidrag till tillvxt och konkurrenskraft i Sverige
Tillvxtanalys Dnr: 2014/002, Myndigheten fr tillvxtpolitiska utvrderingar och
analyser.

Ekholm, Anders (2015). Det biologiska samhllet. Om Sverige i Framtiden en

antologi om digitaliseringens mjlighter, Delbetnkande av Digitaliseringskommissionen,
SOU 2015:65, Stockholm: Fritzes.

Ericsson (2015). Ericson Mobility Report, June 2015.

Eriksson, Thord (2015). GPS-system avsljar fusk i hemtjnsten Dagens Samhlle.

September 24.

European Commission (2015). A Digital Single Market Strategy for Europe.

Communication from the Commission to the European Parliament, the Council,
theEuropean Economic and Social Committee and the committee of the regions.
SWD 100 final. May 2015.

Evans, Dave (2011). The Internet of Things How the Next Evolution of the Internet
of Things is Changing Everything. Cisco White Paper.

Feldstein, Martin (2015). The U.S. Underestimates Growth. Wall Street Journal.
May 2015.

189
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Fellnder, Anna, Claire Ingram and Robin Teigland (2015). Sharing Economy
Embracing Change with Caution. Nringspolitiskt Forum rapport, No. 11,
Entreprenrskapsforum.

Flood, Chris (2015). Virtual financial advice becomes profitable reality. Financial
Times. June 22.

Fleming, Sam (2015). Occupational red tape costs millions of jobs, warns White
House. International New York Times. 30 July.

Ford, Martin (2015a). Rise of the Robots Technology and the Threat of a Jobless
Future, New York: Basic Books.

Ford, Martin (2015b). Chinas Troubling Robot Revolution. International New

York Times. June 11.

Ford, Martin (2009). The Lights in the Tunnel: Automation, Accelerating Technology
and the Economy of the Future. U.S.: Acculant Publishing.

Foy, Henry (2014). Technology Start-ups in Central and Eastern Europe Blow Away
Stereotypes. Financial Times. October 1.

Foley, Stephen (2015). Unicorns Rarely Speak a European Language. Financial

Times. August 3.

Fraiberger, Samuel, and Arun Sundararajan (2015). Peer-to-Peer Rental Markets in

the Sharing Economy. manuscript, New York University, March 2015.

Frey, Carl Benedikt and Michael A. Osborne (2013). Future Employment: How
Susceptible Are Jobs to Computerisation? Oxford Martin School Working Paper.

Frick, Anders (2015). Sverige som IT-nation. Om Sverige i Framtiden en antologi

om digitaliseringens mjlighter, Delbetnkande av Digitaliseringskommissionen, SOU
2015:65, 135163, Stockholm: Fritzes.

Flster, Stefan (2014). Vartannat jobb automatiseras inom 20 r utmaningar fr

Sverige. Stiftelsen fr Strategisk Forskning, Stockholm.

Flster, Stefan (2015). De nya jobben i automatiseringens tidevarv. Stiftelsen fr

Strategisk Forskning, Stockholm.

Fretagarna (2014). Frldrafllan. Fretagarna.

Gapper, John (2015). The lenders of the revolution look familiar Financial Times.
June 18.

Gapper, John (2014). Technology will hurt the banks, not kill them Financial
Times. October 14.

Garrahan, Matthew and Hannah Kuchler (2015). The Future of News: Stop the
Presses! Financial Times. July 2.

Garrahan, Matthew and Shannon Bond (2015). A Very Modern Media Mogul.
Financial Times. January 22.

Gaspar, Vtor, Pierre Collin, Michael P. Devereux, Jim H. Snabe, Tea Varrak, Mary
Walsh, and Bjrn Westberg(2014). Commission Expert Group on Taxation of the
Digital Economy. European Commission, May 2014.

190
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Giridharadas, Anand (2015). Want a Steady Income? Theres an App for That. New
York Times, April 29.

GHK (2011). Revised Final Report Study Evaluation the Professional Qualifi-
cations Directive Against Recent Educational Reforms in the EU member states
A report submitted by GHK in association with the Danish Technological Institute
(DTI). Job Number J8428, DG Internal Market and Services, October 28.

Glaeser, Edward L. and Joshua D. Gottlieb (2009). The Wealth of Cities: Agglomer-
ation Economies and Spatial Equilibrium in the United States. Journal of Economic
Literature 47(4). 9831028.

Gleick, James (1993). Genius: The Life and Science of Richard Feynman. New York:
Vintage.

Goel, Vindu (2015). Survey Finds Americans Dont Trust Government or Companies
to Protect Privacy. New York Times, May 20.

Goldfarb, Avi, Shane M. Greenstein, and Catherine E. Tucker (2015). Introduction,

In: Economic Analysis of the Digital Economy, National Bureau of Economic Research.
Conference Report, May 2015.

Goos, Maarten, Alan Manning and Anna Salomons (2014). Explaining Job Polar
ization: Routine-Biased Technological Change and Offshoring. American Economic
Review 104(8). 25092526.

Gordon, Robert J. (2014). The Demise of U.S. Economic Growth: Restatement,

Rebuttal, and Reflections. NBER Working Paper No. 19895. Cambridge, MA:
National Bureau of Economic Research.

Graetz, Georg and Guy Michaels (2015). Robots at Work. CEP Discussion Paper
No. 1335.

Greengard, Samuel (2015). The Internet of Things. MIT Press: Cambridge,

Massachusetts.

Grene, Sophia (2015). Asset management stuck in the digital ice age. Financial
Times. July 20.

Gustafsson, Kjell (1983). Tekoindustrin och frndringar i den internationella

arbetsfrdelningen. Konsekvenser fr lokalisering och sysselsttning i Sverige.
Kulturgeografiska institutionen, Gteborgs Universitet.

Hall, Jonathan and Alan Krueger (2015). An Analysis of the Labor Market for
Ubers Driver-Partners in the United States. Working Paper No. 587, Princeton
University, Industrial Relations Section.

Halman, Loek, Ronald L. Inglehart, Jaime Dez-Medrano, Ruud Luijkx, Alejandro

Moreno and Miguel Basez (2008). Changing Values and Beliefs in 85 Countries.
Brill: Leiden, Boston.

Hannemyr, Gisle (2003). The Internet as Hyperbole: A Critical Examination of

Adoption Rates. The Information Society 19(2). 111121.

Harding, Robin (2015). Robots Fast Developing Brainpower Leaves Their Dexterity
Needing a Hand Financial Times. February 23.

Harding, Robin (2014). US Economy: The Productivity Puzzle. Financial Times.

June 29.

191
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Hargreaves, Ian (2011). Digital Opportunity-A Review of Intellectual Property and

Growth. An Independent Report, May 2011.

Heyman, Stephen (2015). Taxes on Digital Books are Called Unfair. International
New York Times. July 30.

Heyman, Fredrik, Pehr-Johan Norbck and Lars Persson (2015a). Digitaliseringen

och strukturomvandlingen i svenskt nringsliv. Draft, June 3. Stockholm.

Heyman, Fredrik, Pehr-Johan Norbck and Lars Persson (2015b). The Turnaround
of Swedish Industry: Reforms, Firm Diversity and Job and Productivity Dynamics.
IFN Working Paper No. 1079, Stockholm.

Heyman, Fredrik, Pehr-Johan Norbck and Lars Persson (2013). Var skapas jobben?
En ESO-rapport om dynamiken i svenskt nringsliv 19902009. Expertgruppen fr
studier i offentlig ekonomi, 2013:3. Stockholm.

Hill, Andrew (2015). How a Subscription Model Could Benefit Your Business.
Financial Times. July 2.

Hoffman, Reid (2015). Expertise in Scaling up is the Visible Secret of Silicon Valley.
Financial Times. September 15.

Hooghe, Liesbet, Marks Gary and Arjan H. Schakel (2010). The Rise of Regional
Authority A Comparative Study of 42 Democracies (19502006). Routledge.

Hosea, Devin (2015). Interacting with Humans on Human Terms. letters, Financial
Times. June 9.

International Publishers Association (2015). VAT/GST on Books & E-books. An

IPA/FEP Global Special Report. July 2015.

Jackson, Brian (2015). Meet Ross, the Watson-Powered Super Intelligent Attorney.
itbusiness.ca. January 23.

Jacob, Zoe (2015). Impact in the Collaborative Economy the Evidence so far.
Nesta blog, January 19.

Jacobs, Ken, Ian Perry and Jenifer MacGillvary (2015). The High Public Cost of
Low Wages. Labor Center, CA: University of Berkeley, April 13.

Jain, Neha, James OReilly and Nicholas Silk (2015). Driverless Cars: Insurers
cannot be Asleep at the Wheel. Bank of England Underground Blog. 19 June 2015.

Johnson, Steven (2015). The Creative Apocalypse that Wasnt. New York Times
Magazine, August 19.

Johnson, Steven (2014). How We Got to Now: Six Innovations That Made the
Modern World. Penguin: New York.

Jones, Charles I. (2015). The Facts of Economic Growth. NBER Working Paper
No. 21142. Cambridge, MA: National Bureau of Economic Research.

Jonsson, Magnus (2007). kad konkurrens och inflation. Penning-och Valutapolitik2,

Stockholm: Swedish National Bank.

Kahneman, Daniel (2011). Thinking Fast and Slow. Farrar Straus Giroux: New York.

Kaijser, Arne (2015). Vxlande frbindelser konkurrens och samverkan mellan

stat och nringsliv i utvecklingen av svensk teleindustri. Om Sverige i Framtiden en

192
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

antologi om digitaliseringens mjlighter, Delbetnkande av Digitaliseringskommissionen,

SOU 2015:65, 213239, Stockholm: Fritzes.

Kantor, Jodi and David Streitfeld (2015). Inside Amazon: Wrestling Big Ideas in
aBruising Workplace. New York Times, August 15.

Kantor, Jodi (2014). Unpredictable schedules keep low-income families in constant

turbulence. International New York Times. 16 August.

Katz, Lawrence F. and Robert A. Margo (2013). Technical Change and the Relative
Demand for Skilled Labor: The United States in Historical Perspective. NBER Working
Paper No. 18752. Cambridge, MA: National Bureau of Economic Research.

Keen, Andrew (2015). No: It Is a Great Deal for the Owners, but a Bad One for
Workers. Wall Street Journal. Blog debate with Rachel Botsman, May 10.

Kessler, Aaron M. (2015). Few Laws Deter Hands-Free Driving in U.S.; Regulatory
Scramble Precedes the Imminent Rollout of Automated Cars. International New
York Times. May 4.

Kjellberg, Anders (2015). Kollektivavtalens tckningsgrad samt organisationsgraden

hos arbetsgivarfrbund och fackfrbund. Studies in Social Policy, Industrial Relations,
Working Life and Mobility, Research Reports 2013:1. Updated May 8, Department
of Sociology, Lund University.

Kleiner, Morris (2015). Reforming Occupational Licensing Policies. The Hamilton

Project, Discussion Paper 2015-01, March 2015, Brookings.

Kleiner, Morris and Alan B. Kruger (2013). Analyzing the Extent and Influence of
Occupational Licensing on the Labor Market. Journal of Labor Economics 31(2).
173202.

Knowles-Cutler, Angus, Carl B. Frey, and, Michael A. Osborne (2014). Agiletown:

The Relentless March of Technology and Londons Response. London: Deloitte.

Koski, Heli (2011). Does Marginal Cost Pricing of Public Sector Information Spur
Firm Growth? The Research Institute of the Finnish Economy, Discussion Papers
No. 1260, September 2011.

Koumenta, Maria, Amy Humphris, Morris Kleiner, and Mario Pagliero (2014).
Occupational Regulation in the EU and UK: Prevalence and Labour Market
Impacts. Final Report, Department for Business, Innovation and Skills, School
ofBusiness and Management, Queen Mary University of London, London.

Kuchler, Hannah (2015). The Wealth of Opportunity Carries Enormous Risk.

Financial Times. January 27.

Lazer, David, Ryan Kennedy, Gary King and Alessandro Vespignani (2014). The
Parable of Google Flu: Traps in Big Data Analysis. Science 343(6176). 12031205.

Leubsdorf, Ben and Jon Hilsenrath (2015). US Workers Ask: Wheres My Raise?
Wall Street Journal, June 3.

Levay, Charlotta (2013). Framtida utmaningar fr sammanhllning och rttvisa,

Delutredning frn Framtidskommissionen, Ds 2013:3, Stockholm: Fritzes.

Lindbeck, Assar and Solveig Wikstrm (2003). E-Exchange and the Boundary
Between Households and Organizations. Kyklos 56(3). 315340.

193
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Lindbeck, Assar and Solveig Wikstrm (2000). The ICT Revolution in Consumer
Product Markets. Consumption, Markets & Culture 4(1). 7799.

Lohr, Steven (2015a). IBM Creates Watson Health to Analyze Medical Data. New
York Times, April 13.

Lohr, Steven (2015b). IBM Scores Weather Data Deal and Starts Internet of Things
Unit. New York Times, March 31.

Lohr, Steven (2015c). IBM Adds Medical Images to Watson, Buying Merge Health-
care for $1 Billion. New York Times, August 7.

Lohr, Steven (2015d). If Algorithms Know All, How Much Should Humans Help?
New York Times, April 6.

Lohr, Steven (2015e). The Internet of Things and the Future of Farming. New York
Times, August 4.

Luna, Taryn (2014). Webrooming Shoppers Research Online, Then Buy in Stores.
Boston Globe, November 2014.

MacCrory, Frank, George Westerman, Yousef Alhammadi and Erik Brynjolfsson

(2015). Racing with and Against the Machine: Changes in Occupational Skill and
Composition in an Era of Rapid Technological Advance. Thirty Fifth International
Conference on Information Systems, Auckland 2014.

Maddison-Project (2013). <http://www.ggdc.net/maddison/maddison-project/home.

htm>, 2013 version.

Magnusson, Lars (2014). Sveriges ekonomiska historia. Lund: Studentlitteratur.

Manjoo, Farhad (2015a). Unicorns Rise as a Tech Buzzword; Some Now Reject
Term, Which Denotes Start-Ups Valued Above $1 Billion. International New York
Times. July 6.

Manjoo, Farhad (2015b). Would You Let the IRS Prepare Your Taxes? International
New York Times. April 17.

Manjoo, Farhad (2015c). Tech Training for the Masses. International New York
Times. September 17.

Markoff, John (2015a). Reality check on the state of robotics. International New
York Times. 27 May.

Markoff, John (2015b). Korean roboticists walk off with $2 million prize in Pen-
tagon contest. International New York Times. 8 June.

Markoff, John (2012). Skilled Work, Without the Worker. International New York
Times. August 18.

Markoff, John (2011). Armies of Expensive Lawyers, Replaced by Cheaper Software.

New York Times. 5 March.

Mayer-Schnberger, Viktor and Kenneth Cukier (2013). Big Data: A Revolution That
Will Transform How We Live, Work and Think. London: John Murray.

McAfee, Andrew (2015). What Do the Rich Have Now That Will Soon Spread?
Financial Times. April 7.

194
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

McKinley, Jessey (2014). With Farm Robotics, the Cows Decide When Its Milking
Time. International New York Times. April 22.

McKinsey (2015). The Fight for the Customer. McKinsey Global Banking Annual
Review 2015, McKinsey & Company, September 2015.

Meyer, Julie (2015). Personal Data: Time to Argue About Money, Not Privacy.
Financial Times. April 28.

MGI (2015a). The Internet of Things: Mapping the Value Beyond the Hype.
McKinsey Global Institute. June 2015.

MGI (2015b). A Labor Market That Works: Connecting Talent With Opportunity
inthe Digital Age. McKinsey Global Institute. June 2015.

MGI (2011). Big Data: The Next Frontier for Innovation, Competition and
Productivity. McKinsey Global Institute. June 2011.

Milne, Richard (2014). The Spotify Effect. Financial Times. October 25.

Mims, Christopher (2015). Money Isnt Free, but Moving It Is Now Cheaper.
WallStreet Journal. June 7.

Mokyr, Joel (2013). Is Technological Progress a Thing of the Past? EU-Vox,

http://www.voxeu.org/article/technological-progressthing-past

Mokyr, Joel (2004). Accounting For the Industrial Revolution. in Floud, Roderick
and Paul Johnson: The Cambridge economic history of Britain, Vol.1 Industrialisation,
Cambridge University Press, 3.

Morozov, Evgeny (2015). The Worldwide Fight Over Personal Data has Barely
Begun. Financial Times. October 9.

Morse, Adair (2015). Peer-to-peer crowdfunding: information and the potential for
disruption in consumer lending. NBER Working Paper No. 20899. Cambridge, MA:
National Bureau of Economic Research.

Mrozfeb, Jaqueline (2015). Hand of a Superhero-3D Printing Prosthetic Hands That

Are Anything but Ordinary. New York Times, February 16.

Mullin, Benjamin (2015). Robot-Writing Increased APs Earnings Stories Tenfold.

<http://www.poynter.org/news/mediawire/315931/robot-writing-increased-aps-earnings-
stories-by-tenfold/>.

NBT (2015a). Online Trade, Offline Rules-a Review of Barriers to E-Commerce in

the EU. Kommerskollegium (Swedish National Board of Trade). 2015:7.

NBT (2015b). Servicification on the Internal Market a regulatory perspective. The

case of customisation by 3D printing. Kommerskollegium (Swedish National Board
of Trade). 2015:6.

Nordhaus, William D. (2007). Two Centuries of Productivity Growth in Computing.

Journal of Economic History 67(1).

OConnor, Sarah (2015a). Recovery Prompts Rise in Precarious Job Contracts.

Financial Times. August 5.

OConnor, Sarah (2015b). The Human Cloud: A New World of Work. Financial
Times. October 9.

195
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

OConnor, Sarah (2015c). The new frontier of staff surveillance. Financial Times.
June 9.

OConnor, Sarah (2013). Amazon unpacked. Financial Times. February 8.

OECD (2015a). The Future of Productivity. Manuscript Presented at Working

Group 1. 27 February.

OECD (2015b). OECD Economic Surveys: Sweden 2015. Paris: OECD Publishing,

OECD (2015c). OECD Digital Economy Outlook 2015. Paris: OECD Publishing.

OECD (2015d). Students, Computers and Learning: Making the Connection. Paris:
OECD Publishing. September 15.

OECD (2015e). In It Together: Why Less Inequality Benefits All. Paris: OECD
Publishing.

OECD (2015f). Enquiries into Intellectual Propertys Economic Impact. Paris:

OECDPublishing.

OECD (2015g). Addressing the Tax Challenges of the Digital Economy. Action 1
2015 Final Report. OECD/G20 Base Erosion and Profit Shifting Project. OECD
Publishing, Paris. http://dx.doi.org/10.1787/9789264241046-en.

OECD (2014a). Going for Growth: Interim Report. Paris: OECD Publishing.

OECD (2014b). Measuring the Digital Economy: A New Perspective. Paris: OECD
Publishing.

OECD (2013). New Sources of Growth: Knowledge-Based Capital, Key Analyses

and Policy Conclusions Synthesis Report. OECD, Paris

OECD (2012). Highlights of the OECD skills strategy in: Better Skills, Better Jobs,
Better lives: A Strategic Approach to Skills Policies. Paris: OECD Publishing.

Olsson, Kent (1995). Big Business in Sweden: The Golden Age of the Great Swedish
Shipyards, 19451974 Scandinavian Economic History Review 43(3). 301338.

Orphanides, Athanasios and Simon Van Norden (2002). The Reliability of Output
Gap Estimates in Real Time. The Review of Economics and Statistics 84(4). 569583.

Peterson, Hayley (2014). Amazon Changed the Price Of An Item 8 Times In A Single
Day. Business Insider, August 1.

Peysakhovich, Alex and Seth Stephens-Davidowitz (2015). How Not to Drown in

Numbers. International New York Times. May 4.

Poeter, Damon (2015). How Moores Law Changed History (and Your Smartphone).
PC Magazine, April 19.

Podolny, Shelley (2015). If an Algorithm Wrote This, How Would You Even Know?
New York Times, March 7.

Porter, Eduardo (2015a). How It is Not Working Right. International New York
Times. April 22.

Porter, Eduardo (2015b). Licenses as Barriers to Competition. International New

York Times. January 28.

196
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Powell, Naomi (2012). Nordic Countries Increasingly Attractive as Sites for Data
Centers. International New York Times. April 2012.

PwC (2014). The Sharing Economy Consumer intelligence Series. USA.

Rajan, Raghuram G. (2010). Fault Lines: How Hidden Fractures Still Threaten the
World Economy. Princeton University Press.

Reuters (2015). Another Legal Setback for Uber. International New York Times.
May 23.

Richtel, Matt (2012). Wasting Time is New Divide in Digital Era. New York Times,
May 29.

Robinson, Duncan, Richard Waters and Murad Ahmed (2015). US Tech Giants Race
to Overhaul Operations After EU Data Ruling. Financial Times. October 6.

Sachs, Jeffrey D. and Laurence J. Kotlikoff (2012). Smart Machines and Long-Term
Misery. NBER Working Paper No. 18629. Cambridge, MA: National Bureau of
Economic Research.

Sachs, Jeffrey D., Seth G. Benzell and G. LaGarda (2015). Robots: Curse or Blessing?
A Basic Framework. NBER Working Paper No. 21091. Cambridge, MA: National
Bureau of Economic Research.

Sanandaji, Nima (2012). Frskring ett lyft fr kvinnors fretagande. Svenskt

Nringsliv. March 2012.

Sandbu, Martin (2015a). Free Lunch: The Great and the Good on Secular Stagnation.
Financial Times. April 8.

Sandbu, Martin (2015b). Free Lunch: Amazon agonies. Financial Times. August 18.

Sandstrm, Christian (2015). The Non-Disruptive Emergence of an Ecosystem for

3D Printing- Insights from the Hearing Aid Industrys Transition 19892008. manu-
script, The Ratio Institute, Sweden.

Sandstrm, Christian (2013). Facit and the Displacement of Mechanical Calculators.

IEEE Annals of the History of Computing 35(3). 2031.

Sandstrm, Christian (2011). Hasselblad and the Shift to Digital Imaging. IEEE
Annals of the History of Computing 33(3). 5566.

Scheiber, Noam (2015). A kinder office is often sought, but rarely found. Interna-
tional New York Times. 19 August.

Schn, Lennart (2014). En modern svensk ekonomisk historia-Tillvxt och omvandling

under tv sekel. Lund: Studentlitteratur.

Shapiro, Jesse (2006). Smart Cities: Quality of Life, Productivity, and the Growth
Effects of Human Capital. Review of Economics and Statistics 88(2). 324335.

Sharman, Andy (2015). Ford Scraps Plan to Build Car Seat That Detects Heart Attack.
Financial Times. May 3.

SIND (1991). Vgval TEKO. Svensk tekoindustri p 1990-talet. Statens industriverk,

1991:3.

Singer, Natasha and Mike Isaac (2015). An App That Helps Drivers Earn the Most
From Their Trips. New York Times, May 9.

197
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Singer, Natasha (2015). Sharing Data, but Not Happily. International New York
Times. June 5.

Singer, Natasha (2014). Check app. Accept job. Repeat; Doing One-Time Tasks Brings
Workers Flexibility but Also Uncertainty. International New York Times. August 14.

Skypala, Pauline (2015). Alternative lenders have yet to be tested through

a downturn. Financial Times. May 12.

Sleigh, Andrew, Hannah Stewart and Kathleen Stokes (2015). Top Findings From
the Open Dataset of UK Makerspaces. Nesta blog, April 2015.

Smith, Randall (2015). $1 Billion Unicorns Rarer Than They Seem. International
New York Times. June 2015.

SOU (2015a). Gr Sverige i framtiden. Delbetnkande av Digitaliseringskommissionen,

SOU 2015:28. Stockholm: Fritzes.

SOU (2015b). Part 2 in: Mer trygghet och bttre frskring. Slutbetnkande av
Parlementariska Socialfrskringsutredningen, SOU 2015:21. Stockholm: Fritzes.

SOU (2014a). En digital agenda i mnniskans tjnst en ljusnande framtid kan bli vr.
Delbetnkande av Digitaliseringskommissionen, SOU 2014:13. Stockholm: Fritzes.

SOU (2014b). Ett steg vidare nya regler och tgrder fr att frmja vidareutnyttjande
av handlingar. Betnkande av PSI-utredningen, SOU 2014:10, Stockholm: Fritzes.

Svensk Digital Handel and HUI Research (2014). e-barometern, <http://www.hui.se/

statistik-rapporter/index-och-barometrar/e-barometern>.

Svensk Handel (2015). A European data protection regulation for the future.
letterto Swedish minister for justice, 2015-05-20.

Svensson, Thommy (1983). Frn ackord till mnadsln. En studie av lnepolitiken,

fackfreningarna och rationaliseringarna inom svensk varvsindustri under 1900-talet.
Svenska Varv AB, Kunglv.

Sveriges Konsumenter (2014). I agree men till vad? en rapport om integriteten

pntet. June 2014.

Stockholms Handelskammare (2014). Remissvar avseende betnkandet Ett steg

vidare nya regler och tgrder fr att frmja vidareuttnyttjande av handlingar
(dnrS2014/1992/SF), 15 August.

Stokes, Kathleen, Emma Clarence, Lauren Anderson and April Rinne (2014).
Making Sense of the UK Collaborative Economy. Nesta, Collaborative Lab,
September 2014.

Streitfeld, David (2015a). Data-Crunching Is Coming to Help Your Boss Manage

Your Time. New York Times, August 17.

Streitfeld, David (2015b). Mind your manners, or youll be waiting a long time for
aride. International New York Times. January 31.

Streitfeld, David (2014). Airbnb Listings Mostly Illegal, New York State Contends.
International New York Times. October 15.

Stross, Randall (2010). Closing digital divide may widen the achievement gap.
www.mydigitalfc.com, July 11.

198
 THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Summers, Lawrence H. (2014). U.S. Economic Prospects: Secular Stagnation,

Hysteresis, and the Zero Lower Bound. Business Economics 49(2).

Susskind, Richard and Daniel Susskind (2015). The Future of the Professions How
Technology will Transform the Work of Human Experts. Oxford University Press.

Syverson, Chad (2011). What Determines Productivity? Journal of Economic

Literature 49(2). 326365.

Tanner, Adam (2014). Different Customers, Different Prices. Forbes Magazine,

March 26.

Tett, Gillian (2015). The sharing economy is a playground for wall street. Financial
Times. September 25.

Thakur, Subhash, Michael Keen, Balzs Horvth and Valerie Cerra (2003). Swedens
Welfare State Can the Bumblebee Keep Flying? International Monetary Fund,
Washington D.C.

Thomas, Daniel (2014). Battle to shape internet of things hots up. Financial Times.
October 6.

Tillvxtanalys (2010). Regelbrdans ekonomiska effekter en teoretisk ochempirisk

analys. Rapport 2010:14, Dnr 2010/017, Myndigheten fr tillvxtpolitiska utvrde-
ringar och analyser.

TRR (2015). Astra Zeneca flyttar forskningen, Trygghetsrdet.

TRR (2014). Saab i konkurs: en extrem omstllning, Trygghetsrdet.

Tufekci, Zeynep (2015). Why Smart Objects May Be a Dumb Idea. New York Times,
August 10.

Twentyman, Jessica (2015a). Fear of Reputational Damage After Security Leaks

May be Overhyped. Financial Times. May 20.

Twentyman, Jessica (2015b). 3D Printing of Complex Organs Is Still Far Off.

Financial Times. June 17.

Vanek Smith, Stacey (2015). An NPR Reporter Raced a Machine to Write a News
Story- Who won? Planet money, NPR, May 20.

Van Gorp, Nicolai and Olga Batura (2015). Challenges for Competition Policy in
a Digitalised Economy. Directorate for Internal Policies, Economic and Monetary
Affairs, IP/A/ECON/2014-12. European Parliament. July 2015.Vasagar, Jeevan (2015a).
Berlin urged to ease its curbs on Google. Financial Times. June 2.

Vasagar, Jeevan (2015b). German crowdfunders fear new risk rules will stifle invest-
ment. Financial Times. June 23.

Vasagar, Jeevan, Duncan Robinson, Richard Waters and Tim Bradshaw (2014).
Uberhit with nationwide ban in Germany. Financial Times. September 2.

Vickery, Graham (2011). Review of Recent Studies on PSI-Use and Related Market
Developments. manuscript, Information Economics, Paris.

Vigdor, Jacob L. and Helen F. Ladd (2010). Scaling the Digital Divide: Home
Computer Technology and Student Achievement. NBER Working Paper No. 16078.
Cambridge, MA: National Bureau of Economic Research.

199
THE ECONOMY AND DIGITALIZATION OPPORTUNITIES AND CHALLENGES

Waters, Richard (2015a). Digital Concierges Face a Daunting To-Do List in the
E-commerce Fight. Financial Times. May 22.

Waters, Richard (2015b). Instacart Turns Contractors Into Staff. Financial Times.
June 23.

Waters, Richard (2015c). Setbacks Do Not Erase the Disruptive Potential of the
Sharing Economy. Financial Times. July 3.

Waters, Richard and Andy Sharman (2015). Google Aims to Eat Carmakers Lunch
Financial Times. February 14.

Weintraub, Karen (2015). Off the 3D Printer, Practice Parts for the Surgeon.
International New York Times. January 26.

White House (2015). Occupational Licensing: A Framework for Policymakers.

Report prepared by the Department of the Treasury Office of Economic Policy, the
Council of Economic Advisers and the Department of Labor, 2015-07.

Williamson, Oliver E. (1979). Transaction-Cost Economics: The Governance of

Contractual Relations. Journal of Law and Economics 22(2). 233261.

Wolf, Alison (2011). Review of Vocational Education The Wolf Report, March
2011.

Wolf, Alison (2002). Does Education Matter? Myths about Education and Economic
Growth. London: Penguin Books Ltd.

Wolf, Martin (2015). An Economic Future That May Never Brighten. Financial
Times. April 15.

Wolf-Watz, Sanna (2015). Invasion av peer to peer-lending i Sverige fortfarande lngt

borta. Breakit.http://breakit.se/artikel/780/invasion-av-peer-to-peer-lending-i-sverige-
fortfarande-langt-borta. June 11.

World Bank (2015a). Doing Business 2015 Comparing Business Regulations for
Domestic Firms in 189 Economies. The World Bank, Washington DC.

World Bank (2015b). Swedens Business Climate A Microeconomic Assessment.

The World Bank, Washington DC.

World Economic Forum (2011). Personal Data: The Emergence of a New Asset Class.
World Economic Forum.

Wosskow, Debbie (2014). Unlocking the Sharing Economy An Independent Review.

Report to the Department for Business, Innovation and Skills. BIS/14/1227, November
2014.

Wright, Robert (2014). Regulatory Concerns Hold Back 3D Printing on Safety.

Financial Times. November 24.

Yang, Stephanie (2015). Can You Tell the Difference Between a Robot and a Stock
Analyst? Wall Street Journal. July 9.

rstadius, Kristoffer. and Linus Larsson (2015). Mobilen sprar dig verallt.
Dagens Nyheter, August 31.

200
 DECEMBER 2015

MRTEN BLIX

The economy and digitalization

opportunities and challenges

Arkitektkopia AB, Bromma, 2015