Empirical Evidence on Economic Growth

Econ302, Fall 2004 Prof. Lutz Hendricks,

The Questions

Why are some countries (USA) so much richer than others (India)? Why do some countries grow fast (Japan) while others do not (India)? How important are education, investment, "institutions," ...?

Key Development Facts

2.1

Fact #1: Large income gaps

The richest countries are at least 20 times richer than the poorest countries. Compare real GDP per capita:

USA: $20,000 Uganda: $700 How to compare real GDP across countries?

Real GDP = [Nominal GDP] / [Price index] Use the same price index for all countries. Note: Do not use exchange rates to convert Japanese GDP into USD.

Should we compare GDP per capita or GDP per worker?

GDP per capita measures how rich countries are. GDP per worker measures how productive countries are.

The dierence is due to labor force participation.

2.1.1

The world income distribution

More than half of world population earns less than 10% of U.S. China and India account for 40% of world population.

2.1.2

Change in the world income distribution

The world income distribution has become somewhat more equal since 1960. Large disparities remain.

2.1.3

Poor countries do not grow faster

1960 real income per worker

2.1.4

Geographic patterns

All poor countries are in Africa and Asia. All of South/Latin America is middle-income. Almost all rich countries are in North America and Western Europe. Why?

2.2

Fact #2: Growth rate dierences

BS Figure 1.3 Average post-war growth rates range from below 0 to 8% p.a. Most rich countries grow at 1.5 to 2% per year.

It takes 40-50 years to double income per person.

10

2.2.1

Growth Miracles and Disasters

There are growth miracles with growth rates above 5%.

It takes 12 years to double income per person. All of the growth miracles were middle income countries in 1960.

There are growth disasters with negative growth rates.

All of these are in Africa and South America. Growth miracles are usually middle income. Growth disasters are in Africa and South America.

11

Source: Temple (1999)

12

2.3

Fact #3: Persistent growth is a modern phenomenon

Maddison (1991) estimates that per capita incomes roughly doubled in Western Europe between 1500 and 1800. That amounts to a growth rate of 25 percent per century, compared with 500 percent for the centuries after 1800. One reason: innovation is a modern phenomenon (recall Galileo).

13

2.3.1

The world was poor for nearly all history

Source: Jones (2003)

14

Large income dispersion is a modern phenomenon. Prior to about 1800 a more or less common standard of living characterized all major civilizations (Prescott 1998).

Fact #4: Country growth rates vary over time Implication: Some countries that were rich in the past are poor today.

Argentinas per capita income in 1900 was about the same as that of the U.S.

15

U.S. Data

Kaldors stylized facts:

1. The growth rate of output per capita is constant over time.

2. The real interest rate is constant over time.

3. The shares of income earned by capital (rK/Y = 0.3) and labor (wL/Y = 0.7) are constant over time.

These facts suggest a particular production function (CobbDouglas) of the form Y = K 0.3 (zL)0.7

16

U.S. GDP since 1870 . The U.S. growth rate has been constant for 130 years

Source: Jones (2003) .

17

Growth Accounting

How large are the contributions of capital and labor to output growth? Growth accounting provides and accounting answer, but does not identify the sources of growth. Assume an aggregate production function: Yt = F (Kt, Lt, zt). Then the growth rate of GDP is given by: g (Y ) = EF,z g (z ) + EF,K g (K ) + EF,L g (L) (1) where EF,z is the elasticity of Y w.r.to z (a parameter of the production function). In words: GDP growth is a weighted average of input growth rates and productivity growth.

18

L1. Cobb-Douglas example: Yt = zt Kt t

Take logs: ln Yt = ln zt + ln Kt + (1 ) ln Lt. Take the time derivative: d ln zt d ln Kt d ln Lt d ln Yt = + + (1 ) dt dt dt dt

(2)

Note that the growth rate is about the same as the time derivative of the log: d ln Yt g (Yt) dt Then g (Y ) = g (z ) + g (K ) + (1 ) g (L) Growth rate of GDP per worker: g (Y /L) = g (Y ) g (L) (4) (3)

= g (z ) + g (K/L)

(5)

19

4.0.2

Estimating the variables

Capital stock: Estimate from past accumulated investment (perpetual inventory method). Start from an arbitrary K0 way back in the past. Compute Kt+1 = (1 ) Kt + It by forward iteration. Quality adjustments can be made to account for the fact that newer vintages of K are more productive.

Labor input: Estimate from aggregate hours worked. Quality adjustment can be made to account for the fact that more educated workers are more productive (etc.)

Factor income shares: Should use social marginal products when calculating EF,K and EF,L. These are not observable. With Cobb-Douglas, exploit the fact that capital receives fraction of total income: rK/Y = and wL/Y = 1 .

20

TFP: z is unobservable. Estimate this as the residual: g (z ) = g (Y ) (rK/Y ) g (K ) (wL/Y ) g (L) Note: This means that TFP growth captures all omitted factors (everything other than K and L)!

21

4.1
4.1.1

Empirical Results
U.S data

Roughly half of U.S. growth is due to productivity. The rest is mostly due to capital accumulation. Note the productivity slowdown after 1973.

22

4.1.2

Other countries

The Newly Industrialized Countries (NICs) grew at spectacular rates. Why?

TFP growth is not unusually high.

23

4.1.3

TFP growth around the world

TFP growth in SGP is negative! OAN is below COL.

24

4.1.4

How did the NICs sustain high growth?

The key is high and rising investment. Recall g (Y /L) = g (z ) + g (K/L) (6)

In the long-run, K and Y must grow at the same rate (why?).

25

For any constant I/Y, eventually output growth is determined by TFP growth: g (z ) g (Y /L) = 1 (7)

Temporarily, countries can sustain faster growth by raising I/Y .

26

4.2

Summary

The share of growth not accounted for by factor accumulation is 30% to 40% on average. There is a slowdown in g (Y ) and in TFP growth after 1973. Latin America: Slower overall growth. Slower TFP growth. TFP accounts for a small fraction of g (Y ). East Asia: Very high g (Y ), mainly due to capital accumulation (not sustainable). .

27

4.3

Limitations of Growth Accounting

Growth accounting assumes that factors are paid their social marginal products. Externalities bias the calculations. Examples of externalities:

1. Learning by doing: investment improves TFP.

2. Human capital spillovers.

28

4.3.1

Learning by doing example

L1 = Firm i produces output according to: Yi = A K Ki i L . A K ki i

K reects learning by doing (an externality). Firms pay factors their private marginal products: R = Yi/Ki and w = (1 ) Yi/Li. In equilibrium: all rms choose the same ki = K/L = k. Therefore: Y =
P + L1. i Yi = A K k L = A K

The correct growth accounting equation is: g (Y ) = g (A) + ( + ) g (K ) + (1 ) g (L). The standard growth accounting approach weights g (K ) with and therefore attributes part of capitals eect to TFP. .

29

4.3.2

Growth accounting does not identify sources of growth

Factor accumulation responds to technical change. Example: Solow model with constant population. All growth is ultimately due to technical change. But in steady state, growth accounting attributes fraction of g (Y ) to g (K ). .

30

Growth Regressions

Why do some countries grow fast or slow? A large literature addresses that question by regressing a countrys growth rate on

initial conditions: y, k, h. control variables: investment, government spending, institutions, etc.

31

5.0.3

Cross-country growth regression

32

Poorer countries grow faster (conditional convergence).

33

Weak relationship between institutional measures and growth

34

35

5.1

Problems

1. Robustness (Levine and Renelt 1993):

Nearly all regressors become insignicant when some other regressors are added to the equation.

2. Interpretation.

Regressions show that growth rates are correlated with investment, but cannot resolve causality.

Regressors may proxy for other omitted variables. Example: Low government spending could reect low taxes.

Only quantitative theory can resolve the question why some countries grow faster than others.

3. Do long-run growth rates really dier across countries?

Long-run data cast doubt on that assertion.

36

Perhaps the regressions capture transitional dynamics.