1

Table of Contents

Table of Contents.......................................................................................................................... 1
Acknowledgements ....................................................................................................................... 2
Foreword ....................................................................................................................................... 3
1. Introduction.......................................................................................................................... 3
2. Identifying the Potential Set of Instruments to Modify Travel Behaviour ............................ 4
3. Decision Blocks and Decision Linkages ............................................................................. 7
4. Equilibration in the Travel, Location and Vehicle Markets................................................ 13
4.1 The Travel Market ................................................................................................... 13
4.2 The Residential Location and Dwelling Type Market ............................................. 14
4.3 The Automobile Market ........................................................................................... 16
5. Application Issues ............................................................................................................. 19
6. Specifying the Critical Paths and Components in the Model System............................... 21
7. Illustrative Applications of the Strategy Simulator............................................................. 22
7.1 Evaluation of Four Strategies.................................................................................. 26
8. Conclusions....................................................................................................................... 35
9. References........................................................................................................................ 36
Appendix A: Linkages Between Vehicle Use and the Set of Worker, Household and
Vehicle Choices ................................................................................................................ 39
Appendix B: Details of Variables and their Use in Models.......................................................... 41
Appendix C: Stated Preference Design: Commuter Mode Choice and Vehicle Type
Choice ............................................................................................................................... 46
C1: Commuter Mode Choice ............................................................................................ 46
C2 Vehicle Type Choice.................................................................................................... 55
Appendix D: The Behavioural Model System.............................................................................. 66
D1. Commuter Behavioural Choice Models .................................................................. 66
D2. Household Automobile Composition and Fleet Size Choice Models ..................... 90
D3. Residential Location Choice Models....................................................................... 97
D4. Total Annual Vehicle Use and Composition ......................................................... 103

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
2
Acknowledgements

The work described in this paper is based on a program of research and application
performed between 1993 and 1996 under contract for the Bureau of Transport and
Communications Economics (BTCE). The support of Dr Leo Dobes and Dr David
Gargett of BTCE is greatly appreciated as is the contribution of members of the ITS
team from 93-96, especially Frank Milthorpe, Michael Lowe, Helen Battellino, J ordan
Louviere and Don Anderson. ITS is continuing to develop an urban passenger transport
model capability with the release in 2000 of TRESIS (1.2) A Transportation and
Environment Strategy Impact Simulator. The latter has been developed by David
Hensher, Tu Ton, Carlos Funes, Kirk Bendall, Cam Ngo and Freddy Susanto.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
3
Foreword

This monograph presents an integrated urban passenger transport model system capable
of evaluating the impact of policy instruments on urban travel behaviour and the
environment. The model system has four integrated modules defining household
location and automobile choices, commuter workplace and commuting travel choices,
non-commuting travel activity, and worker distributed work practices. The demand
model system, estimated as a set of discrete and continuous choice models, is combined
with a set of equilibrating criteria in each of the location, automobile and commuting
markets to predict overall demand for passenger travel in various socio-economic
segments, automobile classes and geographic locations. A software package has been
developed to evaluate the relationship between changes in the mixture of policy
instruments and changes in total end-user costs up to the year 2017. Six Australian
Capital cities define the application context - Canberra, Sydney, Melbourne, Brisbane,
Adelaide and Perth. We illustrate the diversity of the system by applying a number of
policies to Perth, in the context of assessing their impacts on greenhouse gas emissions.
The model system forms the basis of ongoing develpments, extending the behavioural
basae to incorporate non-commuter travel choices as well as embedding the entire
system within a GIS-architecture.



1. I nt r oduc t i on

There is a strong interest internationally in strategic level transport model systems
which not only emphasise travel decisions but also the interrelated decisions on the
location of activities and automobile purchases. The combined set of travel, location
and vehicle decisions of individuals and households reflect the growing interest in the
need to evaluate policies from an environmental perspective. This paper presents an
urban passenger transport model system with a strategic focus, and a capability of
evaluating a large number of policy instruments. A software package has been
developed to implement the model system in Australian capital cities.

The monograph is organised as follows. We commence with a discussion of the policy-
linked issues that have motivated the development of the urban passenger model
system. This is followed by an overview of the structure of the model system as a series
of interlinked decision blocks. We then discuss the equilibration process within the
travel, location and vehicle markets. The data requirements for model estimation and
system application are then set out, with details of how the synthetic households used in
applications are constructed and aged, followed by an illustration of how the model
system operates in the presence of an exogenous policy shock.

A number of evaluative applications are presented to highlight how the strategy
simulator can be applied to a set of policy instruments to assess the relationship
between changes in greenhouse gas emissions and changes in total end-user cost over a
ten year period, 1993-2003. The illustrative policy instruments include a fuel excise
change, improvements in the fuel efficiency of passenger vehicles, increase in sales tax
on new automobiles and changes in public transport fares. The simulator is capable of
evaluating many more policies as detailed herein.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
4
2. I dent i f yi ng t he Pot ent i al Set of I nst r ument s t o
Modi f y Tr avel Behavi our

Identifying an expansive set of policy instruments is a critical first step in the
development of a strategic urban passenger transport model to ensure its relevance. The
impact of some policy instruments designed to achieve change within the urban
passenger transport sector will spillover to the non-urban passenger transport sector
(e.g. a carbon tax, emissions trading), and hence achieve greater impact than might a
policy instrument contained to urban activity (e.g. ride sharing). This makes for
potentially more attractive instruments than those limited to urban activity, although it
should not dismiss a mix of aspatial and spatial strategies. Many of the strategies should
be able to contribute to the achievement of the broader goals of urban management and
the performance of urban areas.

The discussion above suggests two primary ways in which the full impact of policy
instruments imposed on urban activity can be classified: (i) spatial vs aspatial
specificity and (ii) urban vs spillover impacts beyond the urban area. Potentially
interesting policy instruments within each of the four combinations are presented in
Table 1. For example, regulations designed to change engine technology or fuel source
have an impact on all contexts of vehicle use, are aspatial and have spillover benefits to
non-urban activity. Congestion pricing within an urban area is spatially specific and has
no (direct) spillover effects on non-urban activity (i.e. it is a contained urban policy).
Aspatial instruments are much easier to evaluate than spatial instruments because the
latter require much more information on an individuals current spatial context in which
a particular instrument requires evaluation. Within each class there are instruments that
are market-based, regulatory, standards, planning and infrastructure initiatives, and
those which involve institutional change.

The literature on travel behaviour provides disparate evidence on the impact of
particular instruments in changing travel behaviour. While some broad conclusions can
be stated, there are often some spatially-specific considerations which can have a strong
influence on behavioural response and hence the success of particular instruments.
Furthermore, any inferences are complicated by a general lack of assessment of the
system-wide impacts of particular policies, especially policies which involve
instruments not currently observed in real markets (e.g. congestion pricing). For
example, any incentive scheme to encourage a compressed work week in order to
reduce the levels of traffic congestion on each weekday should, from an energy and
emissions perspective, identify the use made of the time while not at work. On the non-
work day it is possible that extra non-work travel occurs, which may be by car,
compared to public transport use for the commuting trip. It may be a shorter or longer
trip length depending on the nature of the non-work activity. Anecdotal evidence
suggests that if a Monday or a Friday is part of the extended non-work period, then trip
length by car may increase substantially. Furthermore, increased leisure time tends to
increase the demand for less fuel-efficient vehicles such as 4WD and light commercial
vehicles. The net effect on emissions could be an increase rather than a decrease in
carbon dioxide (CO
2
). This ambiguity should be taken into account and accommodated
as a system wide response through an appropriate specification of the set of behavioural
choice models.


Table 1. Targeted Instruments and Main Class Impact
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
5
Potential Instruments to Evaluate spatial (sp) vs
aspatial(asp)
impact beyond urban area
( )
Congestion pricing (Mix of Charges and Taxes) sp
Parking charges (primarily CBD, regional centres) sp
Parking rationing/restrictions in CBD sp
Toll road charges sp
Restrictive automobile access to locations (CBD) sp

Sales tax on new autos (skewed and elimination) asp
Vehicle registration charges (by age, weight, fuel) asp
Company car provision and use asp
Maximum age of vehicles in the vehicle park asp

Carbon tax (linked to alternative fuels) * asp
Fuel excise by fuel type: change and exemptions* asp
Tradeable permits asp
Fee-based compulsory emissions checks asp
Price rebates/discounts on alternative fuelled vehs asp
Govt purchase and scrap high emitters asp

Alternative fuels - electric vehicles * asp
Alternative fuels - LPG , CNG, dieselhol* asp
Reformulated petrol * asp

Automobile engine/transmission technology * asp
Automobile vehicle design (weight, drag) * asp
On-board IVHS equipment asp

Urban form and density (physical planning,
dwelling mix)
sp

Ride sharing and employer incentives sp
Telecommuting sp
Compressed work week (time use) asp

Non-motorised options - bicycle, walk sp
New public transport - light rail sp
New public transport - bus priority systems sp
Public transport - park-n-ride/kiss-n-ride sp
Existing public transport - fares policies sp
Existing public transport - levels of service sp
Instruments starred (*) are associated with parameters of automobiles, fuel, vehicle operating
conditions and environmental conditions.

A time-differentiated congestion charge, for example, may not assist in reducing
greenhouse gas emissions if the substitution between time of day of travel (i.e. peak
spreading) changes travel speeds throughout the day in an adverse way. We might be
better off containing the congestion during a limited period of time if a uniform
congestion charge regime is not acceptable. Now that congestion pricing is
technologically feasible (Hau 1992, Martin and Michaelis 1993, Hensher 1999), the
emphasis has to be placed on studies designed to understand the response of drivers to
congestion pricing in the peak period, and at all times of the day. What fraction of users
will shift their travel to off-peak times? To what extent will time-sensitive drivers be
attracted to a less-congested road? What fraction of individuals will opt for ride-sharing
or public transport, and how will this vary by income level and other equity-based
criteria? What will be the degree of emissions reduction (if any) brought about by
congestion pricing? The evidence is currently limited (Poole 1992).
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
6

The issue of technological solutions versus behavioural/market-based solutions
pervades the literature on policies to achieve noticeable improvements in the
environment contributed by the automobile (e.g. Martin and Michaelis 1993, Small
1991, Hensher 1998). There is a substantial body of literature which supports the
technological fix: although difficult to achieve any form of change concomitant with
reducing greenhouse gas emissions or other air quality determining pollutants, it
appears easier to achieve progress via an initial stimulation of technology (Hensher et
al. 1992). Consumers will respond if the price is right and there are enough incentives.
As societies pursue greater mobility, the idea of sustainable mobility centred around
clean-fuelled automobiles becomes even more attractive: the desire for personal
mobility seems to be unstoppable (Lave 1992).

Market forces have had a substantial impact on the shape and density of our
metropolitan areas, with a very strong trend towards suburbanisation of workplaces,
self-employment and highly variable work hours and work days (Brotchie 1991,
Hensher 1993, Giuliano and Small 1993). We are beginning to see a long term trend to
shorter standard weekly, annual and lifelong working times (Dawkins and Barker 1987)
with a mix of polarised and redistributed reductions in working time. The reduction in
working time is not a uniform trend. Polarisation involves continuation of long work
hours for some and no work hours for others; redistribution involves shorter working
hours which are widespread on a sufficient scale to counteract job loss, leading to a
society enriched by the spread of liberated time (Tracy and Lever-Tracy 1991). This
increasing heterogeneity away from a classical work schedule has produced a
significant change in the composition of the workforce with respect to age, gender and
education. The most dramatic change is in the increase in the number of females
entering the workforce under different work schedules (number of hours, number of
days per week). A consequence of this development is the increase in the number of
multi-worker households producing residential location choice behaviour that may be
very different from that of traditional single-worker families. With multiple workers and
diversified spatial workplaces the residential location choice set is expanded.

With a trend towards a diminishing number of travel corridors of sufficient density over
reasonable time periods of each day to justify rail-based public transport (be they light
or heavy rail systems), the move away from monocentric to policentric urban activity
reinforces the mobility benefits of the automobile (Giuliano and Small 1991). Reversal
of this trend will be difficult but not impossible (Newman et al. 1993). It may also be
desirable although evidence is patchy. Under this evolving scenario, bus-based public
transport may be a serious alternative to rail. Bus priority systems such as those
introduced in Curitiba (Brazil) and Ottawa (Canada) can contribute significantly to
accommodating mobility needs in the growing number of corridors where the somewhat
lower densities of movement make rail systems less economic (Hensher and Waters
1995, Rutherford 1989, Hensher 1999, Richmond 1999, Smith and Hensher 1998).
Together with the physical zoning of urban activity type and density codes, Curitiba and
Ottawa have encouraged the linear development of medium to high density residential
activity around a bus priority system, without the need to build the more expensive rail
systems. The support in many countries for rail systems however is very strong,
promoted by a belief that the permanence of public transport systems is an essential
element of a physical planning approach to the production of better cities. Light rail
and heavy politics might best describe the ongoing debate. Newman and his colleagues
promote the physical planning paradigm as a major dimension of the urban reform
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
7
process centred on ... the transition to a more compact, transit-oriented city which is
more vital, sustainable, equitable and lively (Newman et al. 1993, 1998).

Sperling (1993) outlines an alternative suburbanisation paradigm centred on clean-
fuel automobiles in which he describes an optimistic yet achievable (in his view)
chronology of some of the key events along a potential pathway to sustainability up to
the year 2010. In 1996 revenue-neutral rebate programs that tax polluting vehicles and
provide rebates to clean vehicles should be introduced in the US, giving substantial
reductions to the cost of an electric vehicle ($US4000). In 1998 tradeable greenhouse
gas emission standards will be adopted for all cars and light trucks followed closely by
a more sophisticated air pollutant and greenhouse gas emission trading program for
automobile manufacturers and fuel suppliers. By the end of the century, developers
build a new town and purchase 20,000 very small neighbourhood electric cars, to be
given free with the purchase of a home. Streets are specifically designed for
neighbourhood cars. In the early years of the new century (up to 2003) the key
automobile manufacturers offer eight free days of car rentals each year for four years
with every new purchase of an electric vehicle. A number of major local government
areas ban all full size vehicles from 9am to 4pm in selected neighbourhoods. Speed
limits of 35 kph are set on many streets. In the year 2003 the first fuel-cell bus enters
commercial operation. The first mass-produced fuel-cell cars come on stream in 2008.
In 2010 construction begins on the first solar hydrogen energy farm to supply fuel-cell
vehicles. Although the timing is clearly invalid (as of 1999), the ambition is worthy of
support.

The importance of an objective assessment of alternative paradigms of efficient,
effective and ecologically sustainable transport strategies is critical to the acceptance
and hence success of particular strategies. We are reminded however of the conclusion
by Small (1991) who, in evaluating what might happen rather that what should happen
in terms of the type of adjustment that our societies will make to diminish the potential
adverse environmental impacts of transportation activities, concludes that People need
not and will not choose solutions that reverse the trends toward increased mobility via
personal vehicles. It is within the context of assessing the role of the expansive set of
instruments synthesised above as contributors to reducing environmental externalities
that the strategy simulator has evolved. There are a number of existing land
use/transport packages; however not one of them has the richness required in its
behavioural specification to evaluate such a large number of diverse and potentially
important influences on reducing the negative externalities on our environment.



3. Dec i si on Bl oc k s and Dec i si on Li nk ages

Urban passenger transport demand is multi dimensional. It encompasses the location of
activities, the alternative travel opportunities available, and the availability of types of
motorised and non-motorised transport. A potential user of the transport system faces
choice opportunities with varying degrees of availability. In the long run, individuals
have increasing opportunities to review all key transport-related choices - where to live,
where to work, the number and types of automobiles in the household, the choice of
means of transport and time of departure for the journey to work, and even negotiation
of the temporal and spatial nature of working hours (i.e. flexitime, a compressed work
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
8
week and telecommuting). In the short run, some of these choices are not available, and
hence condition the choices that can be evaluated and changed.

The behavioural models in the current version of the strategy simulator are presented in
sub-modules representing the four natural divisions of:

(i) commuter choice: spatial and temporal choice of working hours, departure
time choice, mode choice, and workplace location choice

(ii) automobile choice: vehicle type choice and household fleet size choice

(iii) residential choice: location and dwelling type choice, and

(iv) automobile use: total annual vehicle and household kilometres and the
spatial composition of kilometres.

The decision blocks for location decisions, vehicle decisions, and travel decisions and
their major linkages are summarised in Figure 1. Each of the blocks has a set of internal
linkages; the blocks are themselves linked by a set of external linkages. Three
instruments (land rents, used vehicle prices and commuting travel times) are used to
equilibrate within three of the decision blocks, with the option to bypass the vehicle
market equilibration. The non-commuting car use decision block does not have a
market clearing facility in the current specification. The estimated models are a set of
nested logit specifications for all of the models except the vehicle use models. The set
of variables and associated models used throughout the entire model system are
summarised in Appendices B, C and D.

There is an assumed decision hierarchy in which residential location is the uppermost
decision of a household, and as we move down the decision tree we condition each of
the worker-related choices on the higher order decisions. The choice of workplace
location for each worker in a household is conditional on the households choice of
residential location. Likewise the choice of commuter mode is conditional on the choice
of residential and workplace location. The presence of more than one worker in a
household is allowed for by having a separate choice for each worker, together with
additional exogenous variables to account for the influence of the number of workers on
each workers choice of mode, workplace and household residential location. In any
future development an endogeneity link should be evaluated. The modal opportunities
include the set of available alternatives (car as driver and passenger, train, bus) and
possible future investments in new modes in specific spatial contexts such as light rail
and bus priority systems. Stated choice experiments are combined with revealed
preference data in the estimation of the departure time choice and commuter mode
choice models (e.g. Louviere et al in press, Hensher and Bradley 1993, Bradley and
Daly 1993, Morikawa 1991). A summary of the strengths of revealed preference and
stated preference data paradigms is given in Louviere et al (in press). The design and
contents of the SP experiments are presented in appendix C.





TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
9
Residential
DwellingType Choice
Workplace
Location
Choice
Residential
Location
Choice
Vehicle Type
Choice
Vehicle Fleet
Size Choice
AutoKms
Per Annum
Commuter
Mode
Choice
Commuter
Departure
Time Choice
where to live
where to work
auto ownership
mode for
commuter trip
when to depart
for work
Household level
Worker Level
SC
SC
SC
IV
IV
IV
SAI |rl IV =
|wl
IV
SC
Choice of temporal and
spatial working hours
(yes, no)
IV
Given yes, choice of
regular, flexi, compressed
work week and telecommuting
IV
IV
IV
PTIU
. To/from work
. Part of work
. Other urban
. Non-urban
= linkages evaluated but rejected as statistically non-significant
CARUI
PTpub
PTpub
by 4 contexts:


Figure 1. The Linked Model System
(IV =inclusive value index, SC =selectivity correction)

Equilibration within the commuter car travel market is necessary. Given a
predetermined volume of non-commuting passenger travel (expressed in vehicle units)
and commercial vehicles, the commuter market is equilibrated on each origin-
destination pair by using a relationship between travel time, volume and capacity for the
synthetic road system. The presence of a departure time choice model suggests that
equilibration should be determined for each time of day. This requires a separate travel
time matrix for each time period and the ability to have travel substitution between
times of day; that is a dynamic traffic assignment. In the current model system we have
imposed a simplification (relaxed in ongoing development post 1999) - that
equilibration is undertaken as if all travel occurs in the one time period (however
defined). The resulting traffic however is allocated to the choice set of departure times
in accordance with the choice probabilities, to give a temporal profile of the traffic in
equilibrium.

The two lowest decisions in the hierarchy endogenise spatial and temporal work
practices defined by compressed work week choice and telecommuting choice. They are
qualitative enhancements, which have important roles to play in influencing the mix of
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
10
commuting and non-commuting kilometres and in changing the nature of the
commuting peak. Given the complexity of the overall model system we have assumed
initially that the choices of telecommuting and compressed work week behaviour are
not spatially dependent. The primary influences are assumed to be socio-economic and
employment related. However to have an impact on the levels of commuter traffic we
have to adjust the set of probabilities for each worker associated with living in a zone
and working in a zone. We derive the average number of one-way commuting trips per
day over a regular work cycle, assumed to be five days per week. The adjustments for
each of the telecommuting and compressed work week options are summarised in Table
2, based on three options for each type of work practice. Each probability is weighted
by the appropriate scale adjustment and, after the application of the probability of
choosing a car to travel to work, summed to give the overall number of one-way car
commuting trips for each O-D pair.

Table 2. Adjustments in Weekly Commuting Behaviour
CWW CWW CWW
TC zero once per month once per fortnight
zero 1.0 0.95 0.90
once per fortnight 0.9 0.85 0.80
once per week 0.8 0.75 0.70

Two important household-based decisions are meshed into the core hierarchy by
selective linking. These are the choice of dwelling type which links into the residential
location choice, and the choice of automobile number and composition which is linked
with the residential and workplace location choice models. Fleet size (and commuter
mode choice) varies according to residential and workplace location. The dwelling type
choice model provides the interface with the notion of living density - each spatial
location (i.e. zone) contains a mix of three dwelling types (detached houses, town
houses and flats). Within each zone, each dwelling type is given a plot size or density
index that can vary between zones for the same dwelling type. Changes in residential
density are triggered by an exogenous shock to the number of dwellings by type in each
zone. The current version of the behavioural model system is static, hence a number of
exogenously determined impact conditions (or temporal allocation rules - see Table 5
below) have to be imposed to ensure that impacts on each endogenous choice of change
associated with a strategy are not instantaneous and hence misleading. Equilibration in
the residential location market is achieved by the use of dwelling prices as the market-
clearing agent. The possibility of disequilibrium in the dwelling market is allowed for
when a new stock of dwellings is injected into the system since the possibility exists for
excess supply when the number of households is not sufficient in a particular year to
take up the total available stock of dwellings. To fully equilibrate would result in
artificially reducing dwelling prices and would require the scrapping of dwellings.

The automobile market at the household level is represented by a fleet size choice
model and a vehicle type choice model system. Given the interest in environmental
impact, the estimation of the vehicle type choice (and vehicle use) models for multi-
vehicle households is undertaken at the vehicle level, with suitable procedures
implemented to condition the choice of each vehicle type and their use on the set of
vehicle types in the mix. The fleet size choice and type choice models are embedded
within a nested logit framework, with a further nesting within type choice for multi-
vehicle households. Automobiles in the vehicle type choice set are defined by size class,
vintage and fuel type. The size classes for conventional fuelled automobiles are given
below in Table 3. There are six additional electric and alternative fuel vehicle types
(small, medium and large).
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
11

Table 3. Automobile Size Classes for Conventional-Fuelled Automobiles
C1. Micro ( 4 cylinders, < 1400 cc)
C2. Small (4 cylinders, 1400 - 1900 cc)
C3. Medium (4 cylinders, > 1900 cc)
C4. Upper Medium 1 (6 cylinders, < 3000 cc)
C5. Upper Medium 2 (6 cylinders, 3000 cc)
C6. Large ( 8 cylinders)
C7. Luxury (specific makes and engine capacities). All of:
Mercedes, BMW, Rolls Royce, Jaguar, Audi, Bentley, Lexus, Daimler and Eunos
Plus: Honda Legend / NSX (> 3000 cc), Volvo 2300 cc, Saab > 2100 cc
C8. Light Commercial (ABS body types 30-39)
C9. Four Wheel Drive (treated separately outside of ABS registrations)
C10. Light Trucks ( 3.5 tonne, ABS body types 40-49)

The multiple-vehicle branches of the vehicle type choice system are represented
hierarchically according to the chronological entry of each vehicle into the household.
For example, if a household has two vehicles and one was acquired in 1989 and the
other (through replacement or addition) in 1992, then the 1989 acquisition would be in
the upper level (level 1) of the tree; the choice of the vehicle obtained in 1992 is
conditional on the vehicle already held and which is not to be transacted at that time.
The vehicle type choice models allow for the introduction of electric and alternative
fuel vehicles (principally cng), with parameter estimates derived from both revealed
preference and stated preference data (see Appendix C for details of the design and
content of the SP experiment).

The final module is automobile use. The emphasis is on both the total annual use of
each automobile in the households fleet and its composition. The latter is disaggregated
into commuting travel, travel as part of work, other urban travel and non-urban
kilometres. There are three equations in the model system for annual vehicle kilometres
associated with (i) travel as part of work plus travel to/from work, (ii) other urban
travel, and (iii) non-urban travel. Separate generalised least squares regression
equations (correcting for heteroskedasticity) were estimated for each of 1, 2 and 3-plus
vehicle households.

Annual vehicle use is a function of the cost of operating an automobile, which embodies
the fuel performance of an automobile as well as the mix of city and highway driving
cycle, the extent to which public transport is used for commuting, and socioeconomic
characteristics of a household, especially household income and life cycle
characteristics (adults with no children, single parent family, number of children in the
household, number of workers).

Changes in the operating costs of automobiles have a direct impact on a number of
travel choices such as commuter mode choice, automobile type choice and automobile
use. The structure of the strategy simulator allows for the multiple impacts, which filter
through the entire model system. Selectivity variables (SC
i
) from the workplace
location choice, commuter mode choice, residential location choice and vehicle type
choice, although candidates for linking, were not statistically significant in the current
application context. Typically we might expect non-significant selectivity variables if
the vehicle use model is well specified since the key correlates have already been
identified by explicit exogenous variables of instrumented endogenous variables. The
concept of selectivity is summarised in Appendix A. It is worthy of noting that
typically, any statistically significant selectivity corrections often have a minimal
impact on the predicted annual vehicle kilometres.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
12

Operating cost (opcost) for automobiles fully decomposed contains information on fuel
efficiency, fuel prices and excise duties, and is the link into price-related applications
such as a carbon tax and a congestion charge, and the introduction of electric vehicles
and/or alternative fuelled vehicles:

opcost = [{cityFuel*propCityF + hwyFuel*(1-propCityF)}*0.01] *
[tPricePetrol*(1-propnDiesel) + tPriceDiesel*propnDiesel +
carbonTax * {carbLitD*propnDiesel+carbLitP*(1-propnDiesel)}] +
(cTank +carbonTax*carbTAlt) / rangealf +(cCharge +carbonTax*
carbTElc) / rangeelc

where:

cityFuel =city cycle fuel efficiency (litres/100 km)
hwyFuel =highway cycle fuel efficiency (litres/100 km)
propCityF =proportion of use which is in the city fuel cycle (default =0.7)
propnDiesel =proportion of conventional-fuelled vehicles using diesel
tPricePetrol =wpricepetrol +expricepetrol (cents per litre)
wpricepetrol =wholesale price of petrol (cents per litre)
expricepetrol =excise component of price of petrol (cents per litre)
tpricediesel =wpricediesel +expricediesel (cents per litre)
wpricediesel =wholesale price of diesel (cents per litre)
expricediesel =excise component of price of diesel (cents per litre)
carbonTax =carbon tax (cents per kg)
carbLitD

=carbon per litre of diesel (kg/litre)
carbLitP =carbon per litre of petrol (kg/litre)
carbTElc =carbon per full electric recharge (kg)
carbTAlt

=carbon per tank of alternative fuel (kg)
cTank =wctank +extank (cents)
wctank =wholesale cost of a tank of alternative fuel (cents)
extank =excise component of cost of a tank of alternative fuel (cents)
cCharge =wccharge +excharge (cents)
wccharge =wholesale cost of a full electric recharge (cents)
excharge =excise component of cost of a full electric recharge (cents)
rangeelc =range of electric vehicle on a fully charged battery (km)
rangealf =range of alternative fuelled vehicle on a full tank (km)

The electric and alternative-fuelled vehicles can be introduced onto the market and have
an impact on operating cost via the equation above. In the absence of these fuel-types
the operating cost component of the formula relating to such fuels is not operative in the
strategy simulator. The goods and service tax (GST) to be introduced in J uly 2000 will
replace a number of existing taxes such as excise and its impact can be evaluated within
the model system.

Throughout the model system a large number of choice probabilities are generated per
worker and per household. The summation of these probabilities, after base year
calibration, in various dimensions such as socioeconomic segment (income by life
cycle), vehicle class (vintage by size class), and residential location enables the model
system to produce a rich set of useful planning and policy outputs. Within each of these
presentation summations we can identify distributions of emissions, energy consumed,
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
13
annual vehicle kilometres, total end-user cost (decomposed into money and time costs),
consumer surplus, government revenue, and relevant absolute numbers and market
shares of automobile types owned by households and commuter modal shares.

A useful planning linkage index associated with the commuter mode choice model is a
modal accessibility index. It defines the opportunity to access a particular workplace
from a particular residential location given all the available commuting modal
alternatives. The index is conditional of each pair of origin (O) - destination (D) zones.
Each commuter has an associated probability of choosing a workplace location
conditional on a residential location, after due influence of commuter mode choice,
departure time choice, and work practice opportunities.

The link between workplace location choice and residential location choice produces a
spatial accessibility index for the commuting activity. This adds a further dimension to
the modal accessibility index. It identifies the accessibility of a residential location to
all (in contrast to one) workplace locations by all available modes.



4. Equi l i br at i on i n t he Tr avel , Loc at i on and Vehi c l e
Mar k et s

4.1 The Travel Market

Handling equilibration in the commuter trip market requires special comment in the
absence of a detailed micro spatial - network model. At the strategic level the interest is
not on travel times and congestion of particular links in the highway network, but on the
overall adjustment in commuting travel times consequent on particular strategies. It is
necessary, however, to identify the way that a behavioural response to a strategy is
traced through the spatial network to establish any impact on travel times. Changes in
travel times will operate iteratively through feedback from the trip to the market to the
trip etc. until a convergence limit is reached, as exogenously defined by a change which
is less than a fixed percentage change. As an alternative to a network-based approach,
the pre-2000 version uses an empirical equation calibrated on the sample of commuting
trips to represent the travel time profile of commuters (by time of day). The exogenous
variables must be capable of predicting changes in travel time due to the accumulated
behavioural responses throughout the model system (Figure 2).

A travel time model for each time of day would have the following exogenous variables
for each origin-destination pair: the number of trips, the road capacity, and base free
flow conditions represented by distance/speed. This equation is calibrated on observed
travel times and thus represents the base equilibrium conditions. If desired, this linking
equation can be replaced with a full-blown network for users who wish to link the
model system to a traffic assignment package such as TRANSCAD, TRANSTEP or
EMME/2. This is not part of the initial specification of the simulator, although the
integration is feasible (and has been integrated in a new prototype to be released in
2000). Our approach, however, is richer in that we are not limited to a 2-hour am peak
network, which currently drives urban areawide transport planning models in Australia
and in most urban contexts throughout the world. The possibility of time of day
commuter switching is very real in our model system. To provide a reliable mapping
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
14
between travel times and the three major exogenous variables, we use a network model
to generate average travel times under a large number of mixtures of trip volumes, road
capacities and distances. The predicted base travel times and flows on each link in all
six cities have been compared to the output of EMME/2 - the results are almost
identical.

The resulting database is used to obtain parameter estimates to represent the role of the
volume capacity ratio in determination of predicted travel times. This enables us to
impose an endogeneity condition on travel time at the aggregate level. That is, whereas
each individual commuter cannot influence their own travel time once a time of day is
chosen, the aggregation of individual choices (i.e. total trips) within a given network
defined by capacity and the spatial network will influence average travel times. This
gives an empirical relationship to revise travel times within the location-to-location
matrix in arriving at revisions in the probabilities of household commuters choosing
particular modes between particular residential and workplace locations, which get
aggregated iteratively to adjust total trips and hence travel times, given distance and
capacity. This procedure also provides a capability for evaluating the impact of changes
in location-specific road infrastructure (e.g. a new toll road, a bus priority system). The
introduction of rail infrastructure is handled via the commuter mode choice model
where we can exogenously adjust the attribute levels of existing rail public transport or
add in a new rail alternative (the latter by the inclusion of the light rail utility
expression).

Figure 2. Equilbration in the travel market


4.2 The Residential Location and Dwelling Type Market

Network Capacity
Free Flow Travel Time
Model
Probabilities
Total Network
Trips
Origin-Destination
Trips
Travel Time
Equilibration
Travel
Time
T = T
0
[ 1 + a(V/C)b ]
where T = travel time
T0 = travel time under zero trip conditions (ie distance/free flow speed)
V = number of trips (OD demand)
C = synthetic capacity from a zone to another zone
{
= volume difference
between iterations on each
OD link

100 converged if
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
15
Households adjust their residential location in response to changes in the transport
system and for other reasons. Consequently any one of a number of strategies can
influence the probability of a household both living in a particular location and the type
of dwelling they choose to occupy. At any point in time there will be a total demand for
dwelling types in each residential location. Excess demand will result in an increase in
location rents and dwelling prices; excess supply will result in a reduction in the
respective rents and prices. In the simulator, dwelling prices are used to clear the
markets for dwelling types and location, in the absence of data on location rents (Figure
3). The market clearing mechanism is linked into a set of impact indices that allocate
heuristically the impact of a strategy on the choice of residential location and dwelling
type across time, so that, in the absence of a dynamically specified adjustment process
within the behavioural model set, the temporal response profile is realistic.
Equilibration is secured for both the dwelling type market and the residential location
market. Disequilibrium is allowed for when an injection of new dwellings creates
excess supply given the number of households. Under this strategy the simulator needs
only to ensure that the demand for dwellings by type in a residential zone does not
exceed supply for the zone. Any additional dwellings will be left vacant in the particular
year as an indication that property developers may have created too much stock at that
time. In future years as households grow the take up rate increases without creating
increases in dwelling prices until the market is cleared.

It is important to observe the process of equilibration or disequilibration under a
temporal allocation rule applied to a static model system as a proxy for a dynamic
model specification. At the first iteration of equilibration, a set of choice probabilities
are obtained and scaled according to the temporal allocation rule (see Table 5 below).
The summed probabilities are used to identify the aggregate relationship between
demand and supply for each type of dwelling in each residential location. A set of
directional dwelling price adjustments are created as input into the second iteration
prediction of dwelling prices; they reflect the partial adjustment of the market to the
initial exogenous shock (i.e. strategy). A resulting set of new probabilities based on the
adjusted prices are obtained. These second-round choice probabilities are assumed to
represent further adjustments in the probabilities associated with the one-period
temporally adjusted annual impact probability outcomes; however since the choice
model still has the property of instantaneous response, a further temporal adjustment is
undertaken in each subsequent iteration in the annual equilibration. Another way of
expressing this is that iterations after the first iteration fine tune the adjustments
applicable to a years choice response. Where the adjustment is complete in one year
(i.e. temporal allocation is 100% in one year), then the static model is essentially a
dynamic model and the rules for each iteration are identical. This same logic applies to
equilibration in all three markets - travel, location and automobiles.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
16
Model
Probabilities
Dwelling Demand
Equilibration
Prices
{

= (D - S)/S
converged if
Current Housing Stock
n location zones
3 dwelling types
Total Demand
If D > S

1%
Households
pricei =pricei +(

) * pricei


Figure 3 Equilibration in the residential location market


4.3 The Automobile Market

Identification of automobile scrappage rates and expected future prices of used vehicles
are important features of the simulator. In the base year (1993) we begin with an
observed set of used and new vehicle registrations in each class (and vintage). For
classes in subsequent application years we identify the number of vehicles on register in
the existing and the new classes, the latter added over time at the annual rate of 10
conventional fuel classes (see Table 3) and six non-conventional fuel classes (if
applicable - i.e. two fuels and three vehicle sizes). New vehicles should be introduced
ideally in accordance with manufacturers release plans; however such information is
not readily available.

Two approaches have been implemented to determine the demand for new vehicles
each year. The first approach is fully implemented in the current version of the
simulator, which we call the vehicle price relativity approach; the second
(equilibration) approach is coded in the software but not available outside of ITS. The
vehicle price relativity approach controls the relativities of vehicle prices by vintage via
given exogenous new vehicle prices. The scrappage model is used only to identify the
loss of used vehicles consequent on vintage and used vehicle prices, where the latter are
fixed by new vehicle prices in a given year. The supply of new vehicles is determined
as the difference between the total household demand for vehicles and the supply of
used vehicles after application of the scappage model based on used vehicle prices
derived from a non-linear empirical equation which predicts used vehicle prices, given
exogenously provided new vehicle prices.

This approach ensures a predetermined relativity of prices of vehicles over all vintages
within a class. Used vehicle prices in the model are set as depreciated new vehicle
prices and reset each year for each vintage of a class, so that if the prices of new cars in
all classes (or just one) rose, then the used car market would rise in price also. The used
car prices of each age within a class are set as a constant function of new car prices to
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
17
give a price decay to establish the relativity of used to new prices each year. These
prices are then used in all demand calculations in type choice and fleet size models, as
well as in the scrappage functions.

The alternative equilibration approach treats new vehicle prices exogenously but
allows a freely determined set of used vehicle prices that are arrived by equilibration in
the vehicle market (Figure 4). To determine the total number of new vehicles to be
released on the market each year the calculations are as follows: given exogenously
defined new vehicle prices (cost-based), total demand for vehicles by class is
determined through the application of the vehicle type choice and fleet size models;
scrappage of used vehicles is also calculated using cost-based prices. A percentage of
used vehicles leave the market for various reasons, typically associated with age and
value. The difference between demand and scrappage gives the number of new vehicles
by class. These new vehicles are then fed into the equilbration process for the base
situation and for a policy application.

Thesimulator requires an empirical scrappage model as well as a used price model that
can be used to identify future stocks of passenger vehicles (by class) as at December
1995, December 2000, December 2005 and December 2010. A scrappage rate model of
the following form is parameterised

NR p-1
91
- NR p
92
NR p-1
91
=

0
+

1
Price p
92
+

a=1
A

2a
ageap
92
(1)

where

NRp-1
91
-NRp
92
NR
p-1
91
is the scrappage rate ( as a percentage) over a period p, NR
z
y
is the number of
vehicles on register in a class in year y and period z, Price
z
y
is the (expected future) price
in year y and period p for a vehicle class, and the other exogenous variables are, for
each class of vehicle, a series of dummy variables (1,0) representing ages of vehicles. In
the current version of the simulator (Version 1.2), the scrappage rate model is estimated
on annual data in 1991 and 1992, the latter based on a linear interpolation from
registration data for 1991 and 1993.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
18
Probabilities
Vehicle Demand
Equilibration
{
= (D - S)/S
converged if
Vehicle Fleet
1971 - current year
10-16 size classes
Model
Vehicle Fleet
6 Vintages
10-16 size classes
Total Vehicle Demand
6 vintages
10-16 size classes
Revised set
of prices
Used
vehicles
Existing
Stock
Scrappage
Model
New
vehicles
Predicted
prices
pricei =pricei +(

) * pricei

5%

Figure 4 Equilibration in the automobile market

The expected price equation was estimated as a lagged dependent variable model using
two stage least squares, with allowance for serial correlation. The loss rate equation was
estimated as an ordinary least squares model with correction for heteroskedasticity. The
expected price in the loss rate model is a prediction from the expected price equation.
The combination of the two equations enables us to predict vehicle loss rates for each
forecast year and to equilibrate on vehicle prices taking into account the role that
vehicle prices have on loss rates. We have built in a recognition that retailers have been
prepared to discount new vehicle prices in a particular class where sales are sluggish.
The empirical equations are embedded within the equilibration subroutines of the
simulator decision support system.

The final empirical scrappage equation is given below together with the profile of the
scrappage model (Figure 5) across each age/vintage for each vehicle class.

Percent scrappage =
0.65904-
0.36383*LN(price)+4*age1+4.3*age2+4.6*age3+4.8*age4+5.1*age5+5.5*age6+6*age7+6.8*age8+7.7*
age9+8.5*age10+10*age11+10.7*age12+11.6*age13+12.7*age14+13.5*age15+14.9*age16+15.6*age1
7+16.6*age18+17.1*age19+17.6*age20+18*age21+18.3*age2+18.45*age23plus
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
19
Scrappage Rate (ln of price)
Vintage
0.000
2.000
4.000
6.000
8.000
10.000
12.000
14.000
16.000
18.000
1
9
7
1
1
9
7
2
1
9
7
3
1
9
7
4
1
9
7
5
1
9
7
6
1
9
7
7
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3

Figure 5 Scrappage rates for each vintage



5. Appl i c at i on I ssues

The suite of behavioural choice and vehicle use models together with the conditions for
equilibration defines one part of thesimulator. The application of the model system to
evaluate a wide range of strategies and to derive useful empirical outputs such as
changes in consumer surplus, government revenue, greenhouse gas emissions, total end
user costs, total annual automobile kilometres, and energy consumed, requires a
specification of a number of contextual dimensions. In particular thesimulator requires
as inputs the following data:

The population of households
The population of automobiles (number by type)
The population of dwelling stock by location
The population of employment opportunities (i.e. jobs) by location
The attributes of automobiles
The socio-economic characteristics of individuals and households
The network characteristics of each form of transport
The future time profile of exogenous variables in a status quo scenario (e.g.
fuel prices, income, population growth, dwelling prices, public transport fares
and service levels, new vehicle releases, automobile prices, and attributes of
new vehicles)

The sample of travellers and households used in model estimation are not used in model
application. Synthetic households define the application units. There are 725 such
households in each city. Each synthetic household is defined by two core
socioeconomic variables number of vehicles (0,1,2,3, 4 plus) and lifecycle stage
(defined in terms of the number of adults by age and number of dependent children). A
weight is attached to each synthetic household to indicate its incidence in the
population. The set of socioeconomic characteristics that exist in the set of travel,
vehicle and location models are broader than the two core socioeconomic variables.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
20
To ensure that the richness of the fuller set of socioeconomic variables contributing to
the explanation of each choice are captured in the definition of synthetic households, so
that the diversity of household responses is captured throughout the model system, we
draw additional samples off of each core synthetic household. The approach involves
taking a random sample of households from the one percent unit record sample of
households from the 1991 census, conditional on each core synthetic household. Since
each of these households is a random sample from a 1% random sample, we capture the
distribution of household types within each core synthetic household type. The data
associated with each of these sampled households is sufficiently rich in socioeconomic
characteristics of the household and its members. The variables available at the
household level from the 1% sample are: household income, dwelling type, number of
vehicles; the variables available at the person level are: age, industry sector, hours
worked, industry, occupation, labour force status, relationship in the household, income,
sex, education qualifications and mode for the journey to work.

In application, each synthetic household is introduced into an urban area, carrying
only a bundle of socioeconomic descriptors for each household member and the
household as a whole. Through the application of the behavioural model system and
given the specification of an urban areas transport network, location attributes (e.g.
number of jobs, number and types of dwellings, dwelling prices), and automobile stock
and attributes, the simulator calculates a full set of choice probabilities and vehicle use
predictions associated with each of the alternatives in each of the models in Figure 1.
The probabilities and predictions of use are expanded up for each synthetic household
to represent the demand by all households in the population represented by a synthetic
household. The calculations are repeated for each synthetic household and then
equilibration in the three markets (travel, location and vehicle) is undertaken to arrive at
a final set of demand estimates. The set of outputs listed above are also accumulated
throughout the simulator calculations so that a comparison can be made for each
application year of each output before and after the simulation of one or more policy
instruments (from Table 1) which define a strategy.

Complementing the synthetic households are data specifications for new and used
automobiles by class and fuel type, the transport network for existing and new modes
(e.g. light rail, busways), spatial and dwelling attributes for residential locations, and
employment attributes for workplace locations. Forecasting the set of exogenous factors
through time (up to the 25 years in the current software) relies on external benchmarks
for population growth, household size growth, price changes for dwellings, fuel,
vehicles, fares etc., and the release of new vehicles by type. In addition the simulator
requires descriptions of electric and alternative fuel vehicles (in terms of the attributes
driving the vehicle type choice model), new public transport modes, and tollroads.

The base year for model development and implementation in the current version of the
simulator software package is 1993, with December the actual time point in which we
measure all activities and external data such as vehicle registrations and population.
(Updates to 1998 are to be released in 2000). The simulator is calibrated for the current
population profiles in December 1993, and then applied annually with summaries of
outputs for each year over the range of specified years. Each of the behavioural models
is calibrated to reproduce the base 1993 shares and total on each alternative. ITS-
CALIB (embedded within the strategy simulator) calibrates the models through a
complex iterative search process including feedback between the linked models. Once
the models are calibrated, the parameter set remains unchanged in all applications. New
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
21
calibration is required when base input data are changed. We have selected the
following data items for calibration in 1993 (Table 4).

Table 4. Base Year Calibration Criteria
Decision Block Data Criteria
Location (per location) . dwelling type share
. total number of households
. total number of workers
. household fleet size distribution (0,1,2,3+)
Vehicle (per vehicle class) . vehicle class shares
. total registered passenger vehicles
. total passenger vehicle kilometres
. household fleet size composition
Travel . commuter mode share
. travel time (origin-destination)
. commuter departure time profile
. sample spatial and temporal work practice
composition



6. Spec i f yi ng t he Cr i t i c al Pat hs and Component s i n
t he Model Syst em

To understand how thesimulator can identify the impact of a strategy, we have selected
one illustrative instrument - a fuel tax increase, which involves an initial exogenous
shock imposed on the vehicle and travel choice blocks. The imposition of an increase in
the tax on automobile fuel, via its impact on unit operating cost (c/vkm) has an
immediate and direct influence on (i) the use of each vehicle for particular trips such as
the commuter trip (i.e. mode choice, which includes both a switch to public transport
and vehicle-substitution from within the households vehicle park), (ii) a change in the
timing of the commuter journey to reduce the increased costs associated with traffic
congestion, and hence (iii) a change in the overall and non-commuting use of each
automobile available to a household. It also directly affects the households choice of
types of automobiles from the set of conventional fuel, electric and alternative-fuel
vehicles (the last two vehicle fuel types introduced in any year, under a reasonable
expectation of availability). The indirect impacts include a change in residential
location via the change in modal and spatial accessibility to work opportunities and a
change in the number of vehicles in a household (given the increased operating costs).
Changes in residential location may further affect the total use of each automobile, as
well as the mix of urban (commuting and non-commuting) and non-urban kilometres.
The adjustment in commuter travel may also affect non-commuting car use if a vehicle
previously used for commuting is released for use by another non-working member of
the household. Some adjustment in the loss rate of automobiles will also occur. The
response paths are summarised in Figure 6.

The adjustments in vehicle, travel and location choices at the household level translate
at the aggregate level into a new set of equilibrium levels for traffic congestion (broadly
measured by the ratio of travel time to distance travelled), residential densities, total
kilometres of travel by automobiles and various forms of public transport, fuel
consumed and greenhouse gas emissions. To establish the equilibration in the market,
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
22
we allow for a series of feedbacks between individual traveller and household
responses, and what this means in terms of the key market clearing variables which
lead, in an iterative manner, to further adjustments in behaviour as the market settles
down to a steady state from which we can extract summaries of behavioural outcomes.

Tracing the Impact of a Fuel Tax
Household Residential
Location Choice
Household Automobile
Fleet Size Choice
Household Automobile
Type Choice
Use of Each Household
Automobile (vkm)
Commuter Mode
Choice
Commuter Departure
Time Choice
Workplace Location
Choice
Location
Rents
(House
Prices)
Used
Automobile
Prices
Commuting
Travel
Times
Household
Non-Commuting
Automobile Use
by Each Vehicle
commuting
accessibility
index
Unit Fuel Cost
(c/vkm)
auto type choice
inclusivevalue
Exogenous Shock:
fuel tax increase
1
1
1
Codes: 1 = direct influence via the exogenous variable defining the shock (strategy)
2 = indirect traceable impact
1
3 = intervening calculation
3
2
2
3
1 3
2
Other linkages in the overall model system
3
2
3
dwelling type
choice inclusive
value
2
IV
2
2
2


Figure 6 Tracing the impact of a fuel tax



7. I l l ust r at i ve Appl i c at i ons of t he St r at egy Si mul at or

The strategy simulator is a complex software package designed to evaluate the
influence that one or more policy instruments in Table 1 will have on changes in a
number of performance criteria, including greenhouse gas emissions, total end user
costs, vehicle kilometres, vehicle type market share, commuter modal shares, end-use
consumer surplus, end-use government revenue, and energy consumed by
socioeconomic market, residential location and vehicle class.

To demonstrate the policy-value of the simulator, we have selected a number of
application scenarios and evaluated them for the city of Perth (Western Australia). We
have focussed on the impact on greenhouse gas emissions and total end user cost over
the period 1993-2003. Through the evaluation of each policy instrument, each varied to
such an extent that an impact profile is identified in emissions-cost space, we are able to
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
23
identify the minimum end-user cost solution over a predefined evaluation period, for a
given percentage reduction of greenhouse gas emissions in a target year.

Each policy instrument is varied in its level of implementation in order to trace out the
change in the total end-user cost (TEUC) curve (in $93 present value terms) associated
with percentage changes in greenhouse gas emissions. The minimum TEUC envelope
traces the minimum end-user cost solution over the range of change in greenhouse gas
emissions. Total end use cost includes all money cost dimensions plus time costs, the
latter associated only with any changes in commuting travel time for all modes.
Behavioural values of travel time savings derived from the commuter mode choice
models for each city are used to convert travel time in minutes to travel time cost. The
Simulator provides actual travel time totals, enabling the analyst to change the
behavioural values of travel time savings. A schematic overview of the simulators
architecture and the file structure is presented in Figures 7 and 8.

The current version of the software (Version 1.2)
1
is able to evaluate one or more policy
instruments (i.e. a strategy) simultaneously; starting or terminatimg policy instruments
in any year (e.g. introduce electric vehicles in year 2000). The level of an instrument
can be varied through time. This flexibility enables the analyst to undertake very
powerful investigations of the impact of instruments in combination or separately with
any variation in the magnitude of the policy instrument.

An important caveat is required in assessing each of the policy instruments. Sensible
specification of policy applications is required. For example, an across-the board
change in urban residential density is not a meaningful policy since version 1.2 of the
software is only redistributing the population and creating changes in prices to clear
each market of dwellings. A more realistic approach would entail increasing the number
of low-density dwellings in the inner zones and holding constant the amount of high
density dwellings in the outer zones.

The model system underlying the strategy simulator is particularly robust for aspatial
instruments such as a carbon tax and improvements to vehicle technology. The spatial
instruments (eg public transport improvements), while capable of being evaluated
within the current version of the simulator, are less robust for a number of reasons. For
example, on public transport changes, version 1.2 of the simulator is aimed at broad
based changes in levels of service (times and frequencies) and fares for existing public
transport. Whereas the ability to introduce new public transport exists, we feel less
comfortable with these applications given the current level of spatial aggregation. The
version for release in 2000 resolves these spatial limitations, fully integrating the
simulator with an assignment procedure
2
.

Temporal allocation rules are introduced as heuristics given the absence of a dynamic
process in the set of behavioural choice models. The simulator allows for the transition

1
The simulator (Version 1.2) is written in C++using Borland C++4.5. The source code for the
simulation core is completely generic and can be compiled and executed on any computing platform with
a C++compiler which supports C++templates and exceptions. The Simulator's interface is written using
a Borland OWL 2.0 class library and is dependant on both the Borland C++compiler and a Windows or
Windows/NT computing platform.
2
We have linked the simulator to TRANSCAD, although other platforms such as EMME/2 could have
been selected. The simulator has purposely been mad generic to avoid any dedpendence on third party
ancillary software.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
24
period in the adjustment period, simulating behavioural inertia, with the user defining
the rate of adjustment between adjacent years. The temporal allocation works as
follows. The user specifies the proportion of the change in a choice probability
associated with each and every choice model that occurs between period t and period
t+1. This proportion can vary between each of the choice models but can be the same if
the analyst wishes to treat it as a constant proportion. A set of default proportions is
available (Table 5). The changes in vehicle use occur fully in one year. This temporal
allocation applies to both the base case and the application case.

Table 5. The Rules for Temporal Allocation of Policy Impacts (defaults in column 2)
Behavioural Choice Proportion changed annually
Residential location choice (slow adjustment) 0.5
Workplace location choice (slow adjustment) 0.5
Fleet size choice (slow adjustment) 0.5
Work practices choice (slow adjustment) 0.5

Commuter mode choice (medium adjustment) 0.5
Vehicle type choice (medium adjustment) 0.5

Commuter departure time (fast adjustment) 0.5


Strategy
Base results
Strategy results
Strategy data
Base data
Policy Temporal Impact
Policy Temporal Scaling
Strategy Scenario
Base Scenario
Strategy
User Interface
SIM
SIM

Figure 7. The Architecture of the Strategy Simulator
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
25
City Scenario
Simulator
Load ( )
Run (Year)
SimSpec File
SimSpec File SimSpec File
UtilityParams
File
UtilityParams
File
UtilityParams
File
Params Year
Results File
Results Year
Calibration File
Calibration File
Vehicle File Vehicle File
Vehs Grouped
Household File Household File
Zone File
Zone Year
Network File
Network Year
City Year
Simulator
City Simulator
Run(simSpecFile)
Parameter File
HhldWeights File


Figure 8. The File Structure of the Strategy Simulator

To illustrate how the simulator handles the temporal response heuristics, suppose we
introduce a carbon tax in 1995. The simulator commences in 1993 with no change until
1995. The status quo or base (business as usual) situation for total end use cost and
GGEs in 1995 and each subsequent year over the forecast period is initially
determined, with appropriate temporal adjustment proportions. The base results for each
year are stored since they only need to be obtained once. Suppose the probability of
choosing a particular mode in year t in the presence of a strategy is 0.5 and in year t+1 it
is 0.6. Given the temporal adjustment constant of 0.5, the relevant choice probability in
year t+1 is 0.5 +0.5 *(0.6 - 0.5) =0.55. A full set of adjusted probabilities is calculated
for each alternative in each choice model for each synthetic household. The simulator
uses 0.55 and all the other adjusted probabilities, and then equilibrates in year t+1.
Temporal allocation can be turned off by assigning a default value of 1.0.

In year t+1 we have to adjust the with policy base at the end of year t to allow for
changes in items such as the stock of dwellings by type and by location, and the number
of people using each commuter mode by O-D pair etc, which arise through equilibration
under the policy. The simulator then sums the end user cost changes each year (and
converts them into $93 present value terms) and calculates the greenhouse gas emission
changes between the base year and the target year to obtain the total change in end use
cost and GGEs. This produces one point on a graph in cost-emissions space. The same
strategy can be repeated with different levels of the same policy instrument(s). Other
strategies can also be evaluated in a sequence, each always relative to the do-nothing or
business-as-usual situation, to obtain a trace of the impact of each policy or policy set.
A minimum cost envelope can then be identified, as illustrated in Figure 9. The strategy
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
26
can continue to have an influence up to the year 2017 if required. A comparison of the
base and application results for each year is automatically provided in the outputs.

C
o
s
t

-

E
m
i
s
s
i
o
n
s

S
p
a
c
e
Change in
Total End-Use Cost
(TEUC
B-A
)
(1 + d)
t
t=1
T
Percentage Change in Greenhouse Gas Emissions
Base in base year compared to Application in target year
B = before strategy
t
A = after strategy
GGE
T
A
- GGE
to
B
GGE
to
B
$93PV

Figure 9. Illustrative Minimum Cost Strategy given GGE target


7.1 Evaluation of Four Strategies

Table 1 synthesised a large number of potential policy instruments. Within the full set
are a number of broad generic types of instruments. We have identified three
instruments for assessment herein - sales tax on new automobiles, a fuel excise on
particular fuels and improvements in automobile technology.

In our illustrative applications the year of introduction (i.e. the exogenous shock) starts
in J anuary. A range of levels for each instrument are introduced. The combined impact
of multiple instruments is not additive, given the complex non-linearities in the model
system. Some instruments such as a carbon tax and a congestion charge have the same
impact on automobile operating costs in commuter mode choice; however their overall
impact varies because of a different treatment in other models such as annual
automobile kilometres, where some classes of travel (e.g. non-urban kilometres) are not
subject to a congestion charge - only a carbon tax, and automobile vehicle type choice
where only the carbon tax applies.

The simulator evaluates each policy annually, summing the impacts over time and
reporting the findings for each year. Table 6 summarises the three illustrative policy
instrument changes to be evaluated separately. All changes commence in 1996. The
cost items are calculated in the simulator in constant dollars ($93), but are converted
into present values at a real discount rate of 8% per annum for all dollar-based outputs
(i.e. TEUC, consumer surplus, and government revenue). 8% is the assumed long-term
value for gilt-edged treasury bonds.



Table 6. Illustrative Policy Instrument Evaluative Strategies
Policy Instrument Units Range of Assessment
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
27
Change in fuel efficiency (litres/100km) % decrease - 5, - 15, - 25
Sales tax on new automobiles % of wholesale price 10, 20, 30
Increase in fuel excise on fossil fuels cents/litre 60, 80, 100
* The conversion of a carbon tax from cents/kg to cents/litre is as follows: for petrol =5,10,15,20 and 25
cents/kg is equivalent to 3.1, 6.3, 9.2, 12.3, and 15.4 cents per litre in $94 assuming a retail price of 70
cents a litre. The equivalent diesel prices are 3.7, 7.3, 11.0, 14.7 and 18.5 cent per litre.

The definition of all output items available from the simulator is given in Table 8. Some
outputs are also presented as a ratio such as vehicles per 1000 population, change in
consumer surplus per capita, vehicle kilometres per annum per vehicle, energy
consumed per 100 kms, and CO
2
per vkm. Base or default forecasts of all exogenous
variables in the model system have been developed up the year 2017. Some items such
as public transport fares and fuel prices are assumed to change over time in line with the
consumer price index; other variables such as population and household growth are
assumed to follow the medium projection reported by each States demographic and/or
planning agency. All the exogenously specified projection criteria used to generate base
profiles of the exogenous variables can be changed in the strategy simulator by the user.
A series of excel spreadsheets embedded within the simulator define all the exogenous
data for each year.

The simulator has been estimated and calibrated for six capital cities in Australia.
Although we only present illustrative results for Perth herein, for each city the base year
aggregate statistics for the vehicle stock, the number of households and the total
population are summarised in Table 7. Our study area in total captures 57.6% of the
population of residents of Australia. The average number of passenger vehicles per
household for each city is: 1.56 in Canberra, 1.35 in Sydney, 1.55 in Melbourne, 1.49 in
Brisbane, 1.43 in Adelaide and 1.59 in Perth. The number of passenger vehicles per
1000 people varies from a low of 465 in Sydney to a high of 560 in Perth. It may be
possible to extrapolate the results to the rest of Australia provided that reasonable
adjustments are made to account for differences in average household size, vehicle type
mix, traffic profiles and annual vehicle use. Exploratory work to investigate the
geographical transferability potential is worth pursuing, using a small data set from a
regional locality.

Table 7. Base Year Aggregate Statistics for each Capital City
City No. of Hhlds No. of Vehicles Population No. Vehs/1000
people
Canberra 101,508 158,918 303,837 523
Sydney 1,220,050 1,643,965 3,538,970 465
Melbourne 1,048,636 1,620,305 3,022,447 536
Brisbane 456,695 681,481 1,321,625 516
Adelaide 379,833 546,890 1,023,623 534
Perth 403,413 639,746 1,143,265 560
Total 3,610,135 5,291,305 10,353,767
Sources: Vehicles: Derived from ABS Table E39, Households: ABS Census Table B52, Population:
From ABS Census tables of population. All data relate to the 1991 census with an assumed flat 2%
growth over the period 1991-93. From 1994 on, up to 2017, growth is assumed to be 1% per annum




Table 8. Explanation of the Range of Outputs Produced by the Strategy Simulator for Base,
Application and Comparison
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
28
(note: many of these outputs can be summarised by zonal location, vehicle class,
and socioecomic group -household income and lifecycle)
Output Dimension Acronym Units
Total annual carbon dioxide TCO2 kilograms (kg)
Total annual end-use money cost TEUC.MC dollars ($)
Total end-use money cost in present value terms TEUCPV.MC dollars 93 ($93)
Total annual end-use commuter travel time cost TEUC.TC dollars ($)
Total end-use time cost in present value terms TEUCPV.TC dollars 93 ($93)
Total annual end-use commuter travel time TEUC.Time minutes (min)
Total annual expected maximum utility from each model
system for each of the model components defined by
the upper level xx of the linkage: residential location
(RL) links, fleet size (FS) links, work practices (WPU)
links, and vehivle use (VKM) links. Disc = the set of
discrete choice models (ie all excluding vkm system)
TEMUxx


dollars ($)
Total annual passenger vehicle kilometres TVKM kilometres (km)
Total annual passenger vehicle kilometres: to/from work
and as part of work
TVKMTwAw kilometres (km)
Total annual passenger vehicle kilometres: other urban TVKMME kilometres (km)
Total annual passenger vehicle kilometres: non urban TVKMCou kilometres (km)
Average operating cost of autos AvOpCost c/km
Annualised automobile capital cost VehAnnCost dollars ($)
Total annual auto operating costs VehOpCost dollars ($)
Total passenger vehicles TVehicles number
Total energy consumed by passenger vehicles TEnergy litres
Total government revenue from auto. ownership TGovtAutoReg dollars ($)
Total government revenue from fuel excise TGovtExcise dollars ($)
Total government revenue from carbon tax TGovtCarbT dollars ($
Total government revenue from sales tax TGovtSales dollars ($
Total government revenue from public transport use TGovtPT dollars ($)
Total revenue from parking strategy TGovtPark dollars ($)
Total revenue from toll roads TRT oll dollars ($)
Total revenue from congestion pricing TR CongPrice dollars ($)
Total number of households THhld number
Total number of people resident in each city TPop number
Total number of workers (p/t and f/t) in each residential
location
TWrkrRes number
Total number of workers (p/t and f/t) in each workplace TWrkrWork number
Commuter modal share for car drive alone TDA percentage (%)
Commuter modal share for car ride share TRS percentage (%)
Commuter modal share for train travel TTrain percentage (%)
Commuter modal share for bus travel TBus percentage (%)
Commuter modal share for light rail travel TLrL percentage (%)
Commuter modal share for busway use TBwy percentage (%)
Total number of annual car drive-alone commuter trips TDA number
Total number of annual car ride-share commuter trips TRS number
Total number of annual train commuter trips TTrain number
Total number of annual bus commuter trips TBus number
Total number of annual light rail commuter trips TLrL number
Total number of annual busway commuter trips TBwy number
Vehicle kilometres per passenger automobile RVKMPvehicle vkm per wuto
No. of vehicles per household RVehiclePHhld vehs per hhld
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
29
Table 8 continued
Emissions per automobile kilometre RCO2PVKM carbon dioxide per vkm
Fuel efficiency of automobile fleet REnergyP100VKM litres per 100 km
Passenger automobiles per person RVehPCapita ratio
Total end user cost per vehicle kilometre RTEUCPVKM $ per km
Total end user cost per vehicle RTEUCPVeh $ per vehicle
Expected maximum utility from discrete choice models
per capita
REMUDiscPCapita $ per person
Expected maximum utility from residential locn linked
discrete choice models per capita
REMURLCPCapita $ per person
Expected maximum utility from fleet size linked discrete
choice models per capita
REMUFSPCapita $ per person
Expected maximum utility from work practices linked
discrete choice models per capita
REMUWPUPCapita $ per person
Change in consumer surplus associated with vkm
models per capita
RDeltaCSVKMPCapita $ per person
Chnage in total consumer surplus per capita RDeltaCSPCapita $ per person
Notes: Toll, parking and congestion charge revenue is separated from government revenue since the
collector of such revenue may be a private organisation.

A number of illustrative applications have been implemented for Perth over the period
1993 to 2003. The policy instruments are introduced in 1994. The findings serve to
demonstrate the power of the simulator (Version 1.2) in evaluating a very broad range
of policy instruments. One or more policy instruments can be introduced in any year.
Table 9 summarises the major results for changes in greenhouse gas emissions between
the base in 1993 and the application in 2003, comparing it with the total change in
discounted total end user cost over the entire 11 year period.

Within the ranges evaluated, improvements in vehicle fuel efficiency have the greatest
impact on reducing CO
2
, as well as reducing the cost to the end user. In contrast a fuel
excise also reduces CO
2
but has the effect of increasing the cost to the end user. The
sales tax policy has been evaluated at 10%, a level lower than the current 20% for the
majority of new passenger vehicles (excluding luxury vehicles) as well as at a higher
level (30%). We see that CO
2
never improves above the 1993 level even though it
improves compared to the do nothing effect in each year up to 2003. The cost to the
end user decreases for the lower sales tax, as is expected but increases at a
disproprtionately higher rate for the larger sales tax. A public transport fare decrease of
50% (not reporte herein) has a small but noticeable impact on changes in CO
2
,
decreasing it by 3.06%, while reducing total end user cost by $340m (in $93). There is a
consequent 2.65% reduction in the cars modal share for commuting and a reduction in
total vkm of less than 1 percent overall.
Table 10 provides a fuller set of outputs for each of the six capital cities associated with
a doubling of automobile fuel prices. Some of the intermediate calculations including
the status quo levels are provided to illustrate the detail available. We see that a
doubling of fuel prices reduces emissions of CO
2
of between 28.2% (for Melbourne)
and 35.34% for Brisbane with an average for all six capital cities of 30.69%. This is
equivalent to a percentage increase in total end user cost between 1991 and 2013 of
25.42%. These comparisons provide important inputs into a broader policy debate on
which mix of instruments can achieve desired improvements in greenhouse gas
emissions at cost increaes acceptable to the community.

In evaluating these policies one has to consider some of the broader implications. For
example, the fuel efficiency policy adds traffic to the road system by encouraging an
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
30
increase in vehicle kilometres travelled. While this may be attractive for individuals
with latent travel demand, it nevertheless may be undesirable from an environmental
point of view. Possibly a congestion charge might accompany this policy. This can be
evaluated within the strategy simulator. For example a 15% fuel efficiency gain
produces an increase in total vkm in 1994 of 460 million (or 4.8%) growing to over 7%
by the year 2003.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
31
Table 9. Summary of Selective Outputs for Perth, 1993 - 2003 for three Policy Applications

(a) Improvements in Fuel Efficiency
Perth TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($)
base fuel
efficiency
fuel
efficiency
fuel
efficiency
fuel
efficiency
fuel
efficiency
fuel
efficiency
base fuel efficiency fuel efficiency fuel efficiency fuel efficiency fuel efficiency fuel efficiency
5% 5% 15% 15% 25% 25% 5% 5% 15% 15% 25% 25%
%change %change %change %change %change %change
1993 2.93E+09 2.93E+09 0.00% 2.93E+09 0.00% 2.93E+09 0.00% 2.91E+09 2.91E+09 -0.07% 2.91E+09 -0.07% 2.91E+09 -0.07%
1994 2.95E+09 2.86E+09 -3.05% 2.64E+09 -10.51% 2.42E+09 -17.97% 2.58E+09 2.56E+09 -0.66% 2.52E+09 -2.17% 2.46E+09 -4.50%
1995 2.97E+09 2.79E+09 -6.06% 2.41E+09 -18.86% 2.03E+09 -31.65% 2.47E+09 2.43E+09 -1.54% 2.36E+09 -4.33% 2.27E+09 -7.98%
1996 2.98E+09 2.73E+09 -8.39% 2.22E+09 -25.50% 1.73E+09 -41.95% 2.29E+09 2.24E+09 -2.05% 2.15E+09 -5.94% 2.06E+09 -10.26%
1997 3.00E+09 2.68E+09 -10.67% 2.06E+09 -31.33% 1.52E+09 -49.33% 2.09E+09 2.04E+09 -2.54% 1.94E+09 -7.27% 1.85E+09 -11.53%
1998 3.01E+09 2.63E+09 -12.63% 1.93E+09 -35.88% 1.37E+09 -54.49% 1.93E+09 1.87E+09 -2.95% 1.75E+09 -9.12% 1.66E+09 -14.25%
1999 3.02E+09 2.59E+09 -14.24% 1.83E+09 -39.40% 1.26E+09 -58.28% 1.79E+09 1.72E+09 -3.86% 1.61E+09 -9.89% 1.52E+09 -14.86%
2000 3.03E+09 2.55E+09 -15.84% 1.75E+09 -42.24% 1.18E+09 -61.06% 1.68E+09 1.61E+09 -4.11% 1.50E+09 -10.54% 1.42E+09 -15.60%
2001 3.04E+09 2.52E+09 -17.11% 1.68E+09 -44.74% 1.13E+09 -62.83% 1.58E+09 1.51E+09 -4.24% 1.41E+09 -11.08% 1.33E+09 -15.70%
2002 3.04E+09 2.49E+09 -18.09% 1.63E+09 -46.38% 1.09E+09 -64.15% 1.50E+09 1.44E+09 -4.20% 1.33E+09 -11.33% 1.27E+09 -15.67%
2003 3.04E+09 2.47E+09 -18.75% 1.59E+09 -47.70% 1.07E+09 -64.80% 1.41E+09 1.35E+09 -4.11% 1.26E+09 -10.64% 1.20E+09 -14.89%
gge2003(a)-gge93(b)/gge93(b) -15.70% -45.73% -63.48% 2.22E+10 2.17E+10 2.08E+10 1.99E+10
difference= -5.40E+08 -1.48E+09 -2.29E+09

(b) Imposition of Increased Sales Tax on new Automobiles
Perth TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($)
base sales tax sales tax sales tax sales tax sales tax sales tax base sales tax sales tax sales tax sales tax sales tax sales tax
10% 10% 20% 20% 30% 30% 10% 10% 20% 20% 30% 30%
%change %change %change %change %change %change
1993 2.93E+09 2.93E+09 0.00% 2.93E+09 0.00% 2.93E+09 0.00% 2.91E+09 2.91E+09 -0.07% 2.91E+09 -0.07% 2.91E+09 -0.07%
1994 2.95E+09 2.96E+09 0.34% 2.95E+09 0.00% 2.95E+09 0.00% 2.58E+09 2.57E+09 -0.27% 2.59E+09 0.50% 2.61E+09 1.28%
1995 2.97E+09 2.97E+09 0.00% 2.97E+09 0.00% 2.96E+09 -0.34% 2.47E+09 2.44E+09 -1.13% 2.47E+09 0.08% 2.51E+09 1.70%
1996 2.98E+09 2.99E+09 0.34% 2.98E+09 0.00% 2.97E+09 -0.34% 2.29E+09 2.25E+09 -1.62% 2.29E+09 0.13% 2.33E+09 1.88%
1997 3.00E+09 3.00E+09 0.00% 3.00E+09 0.00% 2.99E+09 -0.33% 2.09E+09 2.06E+09 -1.63% 2.10E+09 0.29% 2.14E+09 2.20%
1998 3.01E+09 3.02E+09 0.33% 3.01E+09 0.00% 3.00E+09 -0.33% 1.93E+09 1.89E+09 -1.97% 1.93E+09 0.10% 1.97E+09 2.18%
1999 3.02E+09 3.03E+09 0.33% 3.02E+09 0.00% 3.01E+09 -0.33% 1.79E+09 1.75E+09 -2.23% 1.79E+09 0.00% 1.83E+09 2.23%
2000 3.03E+09 3.03E+09 0.00% 3.03E+09 0.00% 3.02E+09 -0.33% 1.68E+09 1.64E+09 -2.38% 1.68E+09 0.00% 1.72E+09 2.38%
2001 3.04E+09 3.04E+09 0.00% 3.03E+09 -0.33% 3.03E+09 -0.33% 1.58E+09 1.54E+09 -2.41% 1.58E+09 0.13% 1.61E+09 2.03%
2002 3.04E+09 3.04E+09 0.00% 3.04E+09 0.00% 3.03E+09 -0.33% 1.50E+09 1.47E+09 -2.27% 1.50E+09 -0.27% 1.54E+09 2.40%
2003 3.04E+09 3.05E+09 0.33% 3.04E+09 0.00% 3.03E+09 -0.33% 1.41E+09 1.39E+09 -1.35% 1.42E+09 0.78% 1.45E+09 2.91%
gge2003(a)-gge93(b)/gge93(b) 4.10% 3.75% 3.41% 2.22E+10 2.19E+10 2.23E+10 2.26E+10
difference= $-317 million $33 million $393 million
note: 20% is current sales tax for most new vehicles

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
32
(c) Imposition of a Fuel Excise

Perth TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TCO2(kg) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($) TEUCPV ($) TEUCPV($)
base fuel excise fuel excise fuel excise fuel excise fuel excise fuel excise base fuel excise fuel excise fuel excise fuel excise fuel excise fuel excise
current = 60c/litre 60c/litre 80c/litre 80c/litre 100c/litre 100c/litre 60c/litre 60c/litre 80c/litre 80c/litre 100c/litre 100c/litre
40 cents/litre %change %change %change %change %change %change
1993 2.93E+09 2.93E+09 0.00% 2.93E+09 0.00% 2.93E+09 0.00% 2.91E+09 2.91E+09 -0.07% 2.91E+09 -0.07% 2.91E+09 -0.07%
1994 2.95E+09 2.68E+09 -9.15% 2.43E+09 -17.63% 2.20E+09 -25.42% 2.58E+09 2.67E+09 3.61% 2.75E+09 6.67% 2.81E+09 8.99%
1995 2.97E+09 2.69E+09 -9.43% 2.44E+09 -17.85% 2.21E+09 -25.59% 2.47E+09 2.56E+09 3.68% 2.64E+09 6.88% 2.70E+09 9.31%
1996 2.98E+09 2.70E+09 -9.40% 2.45E+09 -17.79% 2.23E+09 -25.17% 2.29E+09 2.38E+09 4.02% 2.43E+09 6.16% 2.48E+09 8.34%
1997 3.00E+09 2.72E+09 -9.33% 2.47E+09 -17.67% 2.24E+09 -25.33% 2.09E+09 2.18E+09 4.07% 2.24E+09 6.94% 2.28E+09 9.28%
1998 3.01E+09 2.73E+09 -9.30% 2.48E+09 -17.61% 2.26E+09 -24.92% 1.93E+09 2.01E+09 4.20% 2.05E+09 6.22% 2.10E+09 8.81%
1999 3.02E+09 2.75E+09 -8.94% 2.50E+09 -17.22% 2.27E+09 -24.83% 1.79E+09 1.85E+09 3.30% 1.92E+09 7.21% 1.95E+09 8.83%
2000 3.03E+09 2.76E+09 -8.91% 2.50E+09 -17.49% 2.28E+09 -24.75% 1.68E+09 1.74E+09 3.51% 1.81E+09 7.68% 1.85E+09 10.00%
2001 3.04E+09 2.76E+09 -9.21% 2.51E+09 -17.43% 2.29E+09 -24.67% 1.58E+09 1.65E+09 4.49% 1.71E+09 8.23% 1.75E+09 10.76%
2002 3.04E+09 2.77E+09 -8.88% 2.52E+09 -17.11% 2.29E+09 -24.67% 1.50E+09 1.57E+09 4.33% 1.62E+09 8.27% 1.66E+09 10.93%
2003 3.04E+09 2.77E+09 -8.88% 2.52E+09 -17.11% 2.30E+09 -24.34% 1.41E+09 1.48E+09 4.97% 1.54E+09 9.22% 1.58E+09 11.99%
gge2003(a)-gge93(b)/gge93(b) -5.46% -13.99% -21.50% 2.22E+10 2.30E+10 2.36E+10 2.41E+10
difference= $764 million $1.39 billion $1.84 billion



TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
33
Table 10. Application of Critical Scenarios for GGE Study: Doubling Fuel prices
City Canberra Sydney Melbourne Brisbane Adelaide Perth All Cities
Base 1995 Average levels of Variables:
cityfuel (litres per 100 km) 11.51 11.51 11.51 11.51 11.51 11.51
hwy fuel (litres per 100 km) 8.42 8.42 8.42 8.42 8.42 8.42
proportion city 0.7 0.7 0.7 0.7 0.7 0.7
proportion highway 0.3 0.3 0.3 0.3 0.3 0.3
price of fuel (cents per litre) Doubled 120 120 120 120 120 120
wholesale price of fuel (cents per litre) 20 20 20 20 20 20
Increase in excise duty (ratio of increase) 2 2 2 2 2 2
excise component of fuel price (cents per litre) 100 100 100 100 100 100
carbon tax (cents per litre) 0 0 0 0 0 0
congestion charge (c/km) 0 0 0 0 0 0
propn of fleet petrol 0.97 0.97 0.97 0.97 0.97 0.97
propn of fleet diesel 0.03 0.03 0.03 0.03 0.03 0.03
carbon charge of petrol (cents per litre) 0 0 0 0 0 0
carbon charge of diesel (cents per litre) 0 0 0 0 0 0
carbon charge electric vehicle (cents per litre) 0 0 0 0 0 0
carbon charge cng vehicle (cents per litre) 0 0 0 0 0 0
opcost (c/km) 12.70 12.70 12.70 12.70 12.70 12.70
Household income ($000s) 50 50 50 50 50 50
Distance to CBD of each zone (use average) 10 20 20 16 14 14
Average age of fleet (years) 10.16 10.16 10.16 10.16 10.16 10.16
Average weight of vehicles (kg) 1163 1163 1163 1163 1163 1163
Propn of single parent families 0.2 0.2 0.2 0.2 0.2 0.2
Average no. of children per family 1.3 1.3 1.3 1.3 1.3 1.3
Propn of families with no children 0.2 0.2 0.2 0.2 0.2 0.2
Average no. of workers per family 1.6 1.6 1.6 1.6 1.6 1.6
Average household size 2.8 2.8 2.8 2.8 2.8 2.8
Perth dummy 0 0 0 0 0 1
Adelaide dummy 0 0 0 0 1 0


VKM for 1-vehicle household 8,341 8,676 8,676 8,540 8,473 9,731
Utility of KM travel to/from work 1 veh hhld 0.166 0.166 0.166 0.166 0.166 0.166
Utility of KM travel as part of work 1 veh hld 0.764 0.764 0.764 0.764 0.764 0.764
Utility of KM non-work related urban 1 veh hhld 1.350 1.350 1.350 1.350 1.350 1.350
Sumall 3.281 3.281 3.281 3.281 3.281 3.281
Propn of KM travel to/from work 1 veh hhld 0.051 0.051 0.051 0.051 0.051 0.051
Propn of KM travel as part of work 1 veh hld 0.233 0.233 0.233 0.233 0.233 0.233
Propn of KM non-work related urban 1 veh hhld 0.412 0.412 0.412 0.412 0.412 0.412
Propn of KM non-urban 1 veh hhld 0.305 0.305 0.305 0.305 0.305 0.305
Sum check 1 1 1 1 1 1

VKM for 2-vehicle household per vehicle 10,229 10,535 10,535 10,412 10,351 11,547
Utility of KM travel to/from work 2 veh hhld 1.741 1.741 1.741 1.741 1.741 1.741
Utility of KM travel as part of work 2 veh hld 0.892 0.892 0.892 0.892 0.892 0.892
Utility of KM non-work related urban 2 veh hhld 0.879 0.879 0.879 0.879 0.879 0.879
Sumall 4.511 4.511 4.511 4.511 4.511 4.511
Propn of KM travel to/from work 2 veh hhld 0.386 0.386 0.386 0.386 0.386 0.386
Propn of KM travel as part of work 2 veh hld 0.198 0.198 0.198 0.198 0.198 0.198
Propn of KM non-work related urban 2 veh hhld 0.195 0.195 0.195 0.195 0.195 0.195
Propn of KM non-urban 2 veh hhld 0.222 0.222 0.222 0.222 0.222 0.222
Sum check 1 1 1 1 1 1
Total VKM of all vehicles in 2-veh hhld 20,459 21,070 21,070 20,823 20,701 23,094

VKM for 3-vehicle household per vehicle 6,844 6,844 6,844 6,844 6,844 6,844
Utility of KM travel to/from work 3 veh hhld 0.852 0.852 0.852 0.852 0.852 0.852
Utility of KM travel as part of work 3 veh hld 0.720 0.720 0.720 0.720 0.720 0.720
Utility of KM non-work related urban 3 veh hhld 0.769 0.802 0.802 0.789 0.782 0.782
Sumall 3.341 3.374 3.374 3.361 3.354 3.354
Propn of KM travel to/from work 3 veh hhld 0.255 0.252 0.252 0.253 0.254 0.254
Propn of KM travel as part of work 3 veh hld 0.216 0.213 0.213 0.214 0.215 0.215
Propn of KM non-work related urban 3 veh hhld 0.230 0.238 0.238 0.235 0.233 0.233
Propn of KM non-urban 3 veh hhld 0.299 0.296 0.296 0.298 0.298 0.298
Sum check 1 1 1 1 1 1
Total VKM of all vehicles in 3-veh hhld 20,533 20,533 20,533 20,533 20,533 20,533

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
34
Table 10. continued
Fleet Size Weightings:
Propn of single person hhlds 0.1 0.1 0.1 0.1 0.1 0.1
Propn of 2 parent families 0.2 0.2 0.2 0.2 0.2 0.2
Average age of household head (years) 34 34 34 34 34 34
Inclusive value from mode choice 3.43 3.43 3.43 3.43 3.43 3.43
Inclusive value from workplace location choice 9.3 9.3 9.3 9.3 9.3 9.3
Utility of zero vehicles 1.310876503 1.31087
6503
1.3108765
03
1.31087
6503
1.31087
6503
1.31087
6503

Utility of 1 vehicle 6.377374549 8.25175
991
8.2517599
1
7.44362
9009
7.06974
5627
7.06974
5627

Utility of 2 vehicles 2.854047208 4.01275
5089
4.0127550
89
3.50146
7128
3.27079
8223
3.27079
8223

Utility of 3 plus vehicles 0.284727926 0.45408
9393
0.4540893
93
0.37675
3105
0.34317
4162
0.34317
4162

Sum all utilities 10.82702619 14.0294
8089
14.029480
89
12.6327
2575
11.9945
9451
11.9945
9451

Proportion choosing 0 vehicles 0.121 0.093 0.093 0.104 0.109 0.109
Proportion choosing 1 vehicles 0.589 0.588 0.588 0.589 0.589 0.589
Proportion choosing 2 vehicles 0.264 0.286 0.286 0.277 0.273 0.273
Proportion choosing 3+ vehicles 0.026 0.032 0.032 0.030 0.029 0.029
sumcheck 1 1 1 1 1 1

Weighted household annual VKM 10,846 11,794 11,794 11,416 11,227 12,621
STATUS QUO SITUATION
Number of households in 1991 87553 996911 910921 374786 340166 343614
Total VKM for all households in 1991 in millions 950 11,758 10,744 4,279 3,819 4,337 35,885
Total end use cost in $'millions ($91) 12,059 149,320 136,440 54,338 48,499 55,074 455,729
Total VKM for all households in 1993 in millions 969 11,994 10,960 4,365 3,896 4,424 36,607
Total end use cost in 1993 in $'millions ($91) 12,302 152,321 139,183 55,430 49,474 56,181 464,890
Total VKM for all households in 2013 in millions 1,182 14,635 13,373 5,326 4,753 5,398 44,667
Total end use cost in 2013 in $'millions ($91) 15,010 185,861 169,829 67,635 60,367 68,551 567,254

Energy consumed (million litres of petrol) in
1991
97 1,207 1,103 439 392 445 3,684
Energy consumed (million litres of diesel) in
1991
3 37 34 14 12 14 114
Total emissions of C02 in 1991 million kg's 3,450 42,720 39,036 15,546 13,876 15,757 130,384
Energy consumed (million litres of petrol) in
1993
99 1,231 1,125 448 400 454 3,758
Energy consumed (million litres of diesel) in
1993
3 38 35 14 12 14 116
Total emissions of C02 in 1993 million kg's 3,519 43,579 39,820 15,858 14,154 16,073 133,005
Energy consumed (million litres of petrol) in
2013
119 1,473 1,346 536 478 543 4,495
Energy consumed (million litres of diesel) in
2013
4 46 42 17 15 17 139
Impact of a fuel tax: double excise duty cents per litre ($93)
Total emissions of C02 in 2013 million kg's 4,210 52,127 47,631 18,969 16,931 19,226 159,093
Base Total emissions of C02 in 2013 mill.kg's 6,140 77,556 66,341 29,338 22,500 27,665 229,540
Total end use cost in 2013 in $'millions ($91) 15,010 185,861 169,829 67,635 60,367 68,551 567,254
Base Total end use cost in 2013 in $'m ($91) 12,481 153,584 134,082 51,937 45,804 54,381 452,268

Percent Change in emissions 1991-2013 -31.43 -32.79 -28.20 -35.34 -24.75 -30.50 -30.69
Percent Change in total end user cost1991-2013 20.27 21.02 26.66 30.23 31.80 26.06 25.42

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
35
8. Conc l usi ons

This monograph provides an overview of the essential elements of a new micro
simulation model of the urban passenger transport sector, which recognises and
accommodates the important linkages between location decisions, travel decisions and
automobile ownership and use decisions.

The simulator combines a set of behavioural choice models, equilibration conditions
and a profile of the base transport system, base automobile market and base location
market, into a unified architecture to give the user a powerful decision support system
for investigating the implications of a large number of evaluative strategies. Output
dimensions are extensive - they include consumer surplus, government revenue, total
end user (generalised) cost, energy consumed, emissions of greenhouse gases, the
composition of the automobile fleet, commuter modal shares and automobile use. Each
output can be summarised by geographical location within each city, socioeconomic
class (income and life cycle) and vehicle type (by fuel class).

The behavioural model system, the equilibration procedures and the strategy simulator
represent an advanced management tool for evaluating the potential role of many policy
instruments. However there remain important enhancements in order to ensure that the
investment to date remains current and relevant. Some areas of priority in progressing
version 1.2 of the strategy simulator include:

1. extension of the model system to capture the dynamics of behavioural choice
(this will require further data collection each year)

2. the development of a model capability to represent the new vehicle release
response (by vehicle size class) of manufacturers in the presence of each set of
policy instruments

3. a revisiting of the vehicle scrappage model and the ABS registration data to
improve on the robustness of the equilibration procedures in the vehicle market

4. the development of procedures to jointly equilibrate in each of the three markets

5. the extension of the travel choice models to include non-commuting travel
choices

6. the development of time-of-day network so that the equilibration in the travel
market more adequately represents the opportunities for time-of-day travel
substitution

7. an improved departure time choice model to incorporate level of service
attributes, and

8. the extension of the entire system down to a more detailed zonal and network
system.

Now that we have a system-wide framework within which individual components can
be enhanced, the value of the further activity should be clear.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
36
The simulator emphasises end-use; additional costs such as those associated with
automobile manufacture, energy generation and infrastructure provision (roads and
public transport) have to be added into a final calculation of the full costs of the
strategies evaluated by the simulator.



9. Ref er enc es

Bradley, M. A. and Daly, A. J . (1991) Estimation of logit choice models using mixed
stated preference and revealed preference information, paper presented to the 6th
International Conference on Travel Behavior, Quebec, May 22-24, 1991

Bradley, M. A. and Daly, A. J . (1992) Uses of the logit scaling approach in stated
preference analysis, paper presented at the 7th World Conference on Transport
Research, Lyon, J uly

Brotchie, J . (1991) The changing structure of cities, Urban Futures Special Issue5, 13-
26

Dawkins, P. and Baker, M. (1987) Working time in Australia: research evidence,
Report to the International Symposium on Working Time, Brussels, April

Domencich, T. and McFadden, D. (1975) Urban Travel Demand, North-Holand,
Amsterdam

Giulliano, G. and Small, K.A. (1991) Subcenters in the Los Angeles region, Regional
Science and Urban Economics, 21, 163-182

Giulliano, G. and Small, K. A. (1993) Alternative strategies for coping with traffic
congestion, paper presented to the Egon-Sohmen Foundation Symposium on Urban
Agglomeration and Economic Growth, Zurich, August

Hanson, W. and Martin, R.K. (1990) Optimal price bundling, Management Science, 36
(2), 155-171

Hau, T.D. (1992) Congestion charging mechanisms: an evaluation of current practice,
World Bank, Infrastructure and Urban Development Department, Washington D.C.,
March (mimeo)

Henderson, J .V. (1992) Peak shifting and cost-benefit miscalculations, Regional
Science and Urban Economics, 22 (1), 103-122

Hensher, D.A. (1993), Socially and environmentally appropriate urban futures for the
motor car, Transportation, 20 (1), 1-20

Hensher, D.A. (1998) The Balance between Car and Public Transport Use in Urban
Areas: What Can we do About it? Transport Policy, 5(4), 193-204

Hensher, D.A. (1999) A Bus-based Transitway or Light Rail? Continuing the Saga on
Choice versus Blind Commitment, Roads and Transport Research, 8 (3), 3-21
Hensher, D. A. (1999) Road Pricing and Charging: A Commentary, Traffic
Engineering, 5(1),
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
37

Hensher, D.A. and Bradley, M. (1993) Using Stated Response Data to Enrich
Revealed Preference Discrete Choice Models, Marketing Letters, 4(2), 139-152

Hensher, D.A. and Waters, W.G. II (1995) Light Rail and Bus Priority Systems: Choice
or Blind Commitment? in Starr, B. M. (ed) Readings in Transport Economics, Vol III,
J AI press, Connecticut, 139-162

Hensher, D.A., Smith, N.C., Milthorpe, F.W. and Barnard, P.O. (1992) Dimensions of
Automobile Demand: A Longitudinal Study of Household Automobile Ownership and
Use, Studies in Regional Science and Urban Economics Vol. 22, North-Holland,
Amsterdam

Hensher, D.A., Battellino, H.C. and Mackay, A. (1994-95) On the road to flexible
working times Policy, 10 (4), Summer, 22-27

Lave, C. A. (1992) Cars and demographics, Access, 1, Fall, 4-11

Louviere, J .J ., Hensher, D.A. and Swait, J . (in press 2000) Stated Choice Methods:
Analysis and Applications in Marketing, Transportation and Environmental Valuation,
Cambridge University Press, Cambridge, 697pp

Martin, D. and Michaelis, L. (1993) Road Transport and the Environment: Policy,
Technology and Market Forces, Financial Times Management Report, London

McCarthy, M. (1995) Greenhouse Gas Emissions and the Demand for Urban Passenger
Transport: Graphical Presentation of Community Views on Greenhouse Gas Emissions
and other environmental Topics, Report 6, Institute of Transport Studies, The
University of Sydney, February

McCarthy, M. and Hensher, D.A. (1995) Greenhouse Gas Emissions and the Demand
for Urban Passenger Transport: Descriptive Profile of Travel Survey Data,
Supplementary Report 6A, Institute of Transport Studies, The University of Sydney,
May

Morikawa, T (1989) Incorporating Stated Preference Data in Travel Demand Analysis,
PhD Dissertation, Department of Civil Engineering, M.I.T.

Newman, P., Kenworthy, J . and Vintala, P. (1993) Can we build better cities?: physical
planning in an age of urban cynicism, Urban Futures, 3(2), September, 17-24

Newman, P. and Kenworthy, J . (1999) Sustainability and Cities Overcoming
Automobile Dependence, Island Press, Washington D.C.

Poole, R.W. J r (1992) Introducing congestion pricing on a new toll road,
Transportation, 19 (4), 383-396

Richmond, J . (forthcoming) New rail investments - a review, Transport Reviews

Rutherford, G.S. (1989) Light rail, heavy politics, Pacific Northwest Executive,
October, 17-22
Small, K.A. (1991) Transportation and the environment, in Thord, R. (ed.) The Future
of Transportation and Communication, Swedish National Road Administration,
Borlange, 217-230
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
38

Sperling, D.L. (1993) Clean cars and sustainable transportation, Institute of
Transportation Studies, University of California, Davis, August

Tracy, N. and Lever-Tracy, C. (1991) The longer working week in Australia - working
time experiences and preferences of men, Labour and Industry, 4 (1), 71-94

Waddell, P. (1993) Exogenous workplace choice in residential location models: is the
assumption valid?, Geographical Analysis, 25 (1), J anuary, 65-82

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
39
Appendi x A: Li nk ages Bet w een Vehi c l e Use and t he
Set of Wor k er , Househol d and Vehi c l e Choi c es

Selectivity correction indices from the residential location choice, workplace location
choice, commuter mode choice and automobile type choice models can be carried into
the annual vehicle kilometres models. These linkages enable us to obtain predictions of
vehicle use which take into account the class of vehicle, the role of alternative modes in
the commuting task, and the location of residence and workplace(s). Whereas the
inclusion of the selectivity index in model estimation is manageable, in application it
can involve hundreds of empirical indices per household to allow for the influence on
vehicle use of residential and workplace location, commuter mode choice and auto type
choice. Although the selectivity indices can be calculated and applied at various levels
of detail, some simplifying assumptions or aggregation rules are introduced to
accommodate the complexity of the process. The gains from maximal disaggregation in
application are likely to be minimal.

Let us define a set of selectivity aggregation rules, and then investigate the implications
of aggregation. Each equation represents a summation across the particular choice set
conditional on the position of the choice in the decision hierarchy (Figure 1). The first
aggregation is a sum across all commuter modes for a given residential location and
workplace location. The second equation is an aggregation across all workplace
locations for a given residential location. The third aggregation relates to a summation
across all residential locations. The final equation represents a summation across all
vehicle classes for a given fleet size.


SC
cmc|wl,rl = SC
m|wl,rl
m=1
M
* Prob
m|wl,rl
; rl =1,...,RL; wl =1,...,WL
(A1)


SC
wl|rl = SC
wl|rl
wl=1
WL
*
Prob
wl|rl
; rl =1,...,RL; wl =1,...,WL
(A2)


SC
rl = SC
rl
rl=1
RL
* Prob
rl
; rl =1,...,RL
(A3)


SC
c|fs = SC
c|fs
c=1
C
* Prob
c|fs
;fs =0,...,3+
(A4)

Notation: cmc =commuter mode choice, wl =workplace location, rl =residential location,
m =1,..,M commuter modes, fs =fleet size

For each pair of residential and workplace locations, there will be a separate selectivity
index for each commuter mode for each worker in a household. If, for example, we
assume 10 residential (RL) and 10 workplace (WL) locations with 100 RL-WL paired
locations, 6 modes and say 2 workers in a household, we will generate 1200 selectivity
correction (SC) commuting mode indices per household. For each workplace location,
there will be 10 selectivity indices per worker, one for each residential location. There
will be 100 SCs per worker for all RL-WL pairs. There will be 10 SCs for each
residential location. The size of the calculation is highlighted when we introduce the SC
for auto type choice. With 60 vehicle classes (see Hensher et al. 1994), 3 non-zero
vehicle fleet sizes and a nested structure within the multi-vehicle branches, we will
generate a large number of SC indices: 60 in the 1-vehicle branch, 60*60 in the 2-
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
40
vehicle branch, and 60*60*60 in the 3+vehicle branch. This becomes quite unwieldy,
despite the power of the computer.

The better specified are the set of exogenous variables in the vehicle use models,
however, the less likely that the SC indices will have a statistically significant influence
on vehicle use. If all SC variables are not statistically significant then the application
process is considerably simplified. The SCs were found to statistically non-significant
in the current model system. Some of the exogenous variables, however, will provide
the necessary leverage to account for differences in vehicle kilometres according to the
location of activities (residential, workplace) and modes used for the work trip, for
example the inclusion of exogenous variables to represent the location of a residence
relative to the Central Business District. Since the locations are fixed, they can be
readily used in applications to condition the vehicle kilometre predictions. Vehicle
kilometre predictions by vehicle class and residential-workplace location are obtained
by weighting the predicted kilometres by the probability of selecting each vehicle class
given fleet size, the probability of a household having a particular fleet size and the
locations of activities, even if the SC indices are rejected on statistical criteria. Indeed
Hensher et al. (1992) found that the incorporation of location dummy variables
contributed substantially to the statistical non-significance of the selectivity index.


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
41
Appendi x B: Det ai l s of Var i abl es and t hei r Use i n Model s
Source Model
Variable Description Units Param SyHld Vhcle Zone Ntwk WP DTC CMC WLC ATC FSC DwT
C
RLC VKm Used
acBus Access Time Bus min z 1 1
acBwy Access Time Busway min z 1 1
acc Acceleration 0-100 kms / hr sec v 1 1
acLrl Access Time Lightrail min z 1 1
acTrn Access Time Train min z 1 1
ageHead Age of head of household years s 1 1
ageVeh Age of vehicle years cv 0
agHZ ageVeh/hSize cvs 1 1
altWrkPrc Discount on alt work practice
commuting trips
p 0
boot Bootspace cubic metres v 1 1
bootHSize boot/hSize cvs 1 1
brs Brisbane Dummy (1,0) s 1 1
can Canberra Dummy (1,0) s 1 1 2
capacity Road network capacity n 0
carbLitD Carbon per litre diesel kg / litre p 0
carbLitP Carbon per litre petrol kg / litre p 0
carbonTax Carbon Charge cents / kg p 0
carbTAlf Carbon per tank of alt fuel kg / tank p 0
carbTElc Carbon per tank of electric fuel kg / tank p 0
cCharge Cost of Charge (Elc vehicles) c p 0
cityFuel City Cycle Fuel Efficiency Litres/100kms v 0
clerk Clerical Occupation (1,0) s 1 1 2
cNoKids Couples without children (0,1) s 1 1 1 3
CO2LitD CO2 emmisions for diesel kg/litre p 0
CO2LitP CO2 emisions for petrol kg/litre p 0
CO2TAlf CO2 emissions for tank of
alternative fuel
kg/tank p 0
CO2TElc CO2 emissions for electric charge kg/charge p 0
comDayYear Number of commuting days per
year
days p 0
commTPct Percent of traffic that commuting % p 0
congCost Congestion cost cents / km n 0
costAuto Distance * (opCostM + congCost) cents cnv 1 1
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
42
costBus Cost Bus $ n 1 1
costBwy Cost Busway $ n 1 1
costLrl Cost Lightrail $ n 1 1
costTrn Cost Train $ n 1 1
cTank Cost of tank of fuel (Alf) c p 0
disdWCbd Distance to CBD kms z 1 1 1 1 4
distance Distance Home - Work kms n 0
egBus Egress Time Bus min z 1 1
egBwy Egress Time Busway min z 1 1
egLrl Egress Time Lightrail min z 1 1
egTrn Egress Time Train min z 1 1
exPriceDiese
l
Exercise on diesel cents / litre p 0
exPricePetrol Exercise on petrol cents / litre p 0
freqBus Frequency Bus min z 1 1
freqBwy Frequency Busway min z 1 1
freqLrl Frequency Lightrail min z 1 1
freqTrn Frequency Train min z 1 1
fuelBeta Consumer surplus fuel price beta p 0
fullT Full Time Worker (1,0) s 1 1 2
hHigh Dwelling price high density $000's z 0
hIncome Household Income $000's s 1 1 2
hLow Dwelling price low density $000's z 0
hMed Dwelling price medium density $000's z 0
hPHInc hPrice/hIncome $000's/$000's czs 1 1
hPrice Weighted house price $000's cz 0
hsePBeta Consumer surplus house price
beta
p 0
hSize Household size individuals s 1 1 2
hWeight Household weight s 0
hwyFuel Highway Cycle Fuel Efficiency Litres /
100kms
v 1 1
ivCMC IV Commuter Mode Choice c 1 1 2
ivDTC IV Departure Time Choice c 1 1
ivDwTC IV Dwelling Type Choice c 1 1
ivWLC IV Workplace Location Choice c 1 1 2
ivWPL IV Work Practices Lower Level c 1 1
jTot Total jobs in Zone z 0
lab Labourer Profession (1,0) s 1 1 1 3
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
43
lnAgeVeh Natural log of ageVeh cv 1 1 2
lnDisdWCbd Natural log of disdWCbd cz 1 1
lnHHigh Natural log of hHigh cz 1 1
lnHIncome Natural log of hIncome cs 1 1 2
lnHLow Natural log of hLow cz 1 1
lnHMed Natural log of hMed cz 1 1
lnHPrice Natural log of hPrice cz 1 1
lnJTot Natural log of jTot cz 1 1
lnPIncome Natural log of pIncome cs 1 1 2
lnTotDwl Natural log of totDwl cz 1 1
lnVMass Natural log of vMass cv 1 1
male Male dummy (1,0) s 1 1 2
mel Melbourne Dummy (1,0) s 1 1
mgtAdmin Managment & Admin occupation (1,0) s 0
mgtPrf Prof or MgtAdmin or Paraprof (1,0) s 1 1
mTimeBus InVehicle Time Bus min n 1 1
mTimeBwy InVehicle Time Busway min n 1 1
mTimeLrl InVehicle Time Lightrail min n 1 1
mTimeTrn InVehicle Time Train min n 1 1
nChild Number of children s 1 1
nCylind Number of Cylinders v 1 1
nHigh Number unit dwellings z 0
nLow Number detached dwellings z 0
nMed Number semi-detached dwellings z 0
nWorkers Number of workers (1,0) s 1 1 1 3
opCost Vehicle Operating cost cents / km cv 1 1
opCostC Vehicle Operating Cost - City cents / km cv 0
opCostM Weighted Avg Veh Opcost - City cents / km cv 0
paraProf Para Professional (1,0) s 1 1 2
partTime Part Time Worker (1,0) s 1 1
pDet Proportion detached dwellings % cz 0
per Perth (1,0) s 1 1
perAge Person Age Years s 1 1
pInc1 pincome 0 -9 $000's s 1 1 2
pInc2 pincome 9 - 25 $000's s 1 1
pInc3 pincome 25 - 35 $000's s 1 1
pInc4 pincome 35 - 45 $000's s 1 1 1 3
pInc5 pincome 45 - 55 $000's s 1 1 2
pInc6 pincome 55 - 65 $000's s 1 1 1 3
pInc7 pincome > 65 $000's s 1 1 2
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
44
pIncome Personal income $000's s 1 1
plant Plant Operator Profession (1,0) s 1 1
prkCtAuto Car parking cost $ z 1 1
PTProb public transport probability index 0 to 1 range 1
prof Professional Occupation (1,0) s 1 1 2
propCityF Proportional of City Cycle driving % p 1 1
propnDiesel Proportion of diesel vehicles % v 0
pSemi Proportion detached dwellings % cz 0
pUnit Proportion unit dwellings % cz 0
range Range before Refueling kms v 1 1
reg Number of vehicle registrations v 0
rg Registration cost $ v 0
rideSOcc Commuting ride share occupancy individuals p 0
rtailPct Retail Vehicle Price Markup % p 0
sales Sales occupation (1,0) s 1 1
salesTax Vehicle sales tax % v 0
self Self employed (1,0) (1,0) s 1 1
small Small Car dummy (1,0) v 1 1
sPers Single person household (1,0) s 1 1 1 3
sPFam Sole parent household (1,0) s 1 1 1 3
tech Trades or Plant occupation (1,0) s 1 1
timeAuto Auto travel time mins cn 1 1
timeAuto0 Free flow auto travel time mins n 0
tolCtAuto Auto toll cost $ n 1 1 2
totDwl nLow + nMed + nHigh cz 0
tPrice Price of vehicle $ v,cv 0
tPriceDiesel wPriceDiesel + exPriceDiesel cents / litre cp 0
tPriceHInc tPrice93*hIncome/1000 $/1000,000 1 1
tPricePetrol wPricePetrol + exPricePetrol cents / litre cp 0
trade Trade occuptation (1,0) (1,0) s 1 1
twoPFam Two parent family (1,0) s 1 1 1 3
veh3Plus Average numbers of vehicles in
FSC 3+ alternative
p 0
vehPBeta Consumer surplus vehicle price
beta
p 0
vMass Mass of vehicle kg v 1 1
VOTCar Value of time for car commuters p 0
VOTPT Value of time for public transport
commuters
p 0
wPrice Wholesale Price of New Vehicles $ v 0
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
45
wPriceDiesel Wholesale price of diesel cents / litre p 0
wPricePetrol Wholesale price of petrol cents / litre p 0
y10 >8 Year old vehicle (1,0) cv 1 1
y2 2-3 year old vehicle (1,0) cv 1 1
y5 4-8 year old vehicle (1,0) cv 1 1



TOTAL 8 16 27 21 14 11 8 11 13 128


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
46
Appendi x C: St at ed Pr ef er enc e Desi gn: Commut er
Mode Choi c e and Vehi c l e Type Choi c e

C1: Commuter Mode Choice

The set of travel and vehicle choice experiments evolved out of extensive review,
testing and re-design. In participating in the choice experiments, each respondent was
asked to consider a context in which the offered set of attributes and levels represent the
only available ways of undertaking a commuter trip from the current residential location
to the current workplace location. It was made clear that the purpose is to establish their
coping strategies under these circumstances.

Four alternatives appear in each travel choice scenario, CAR NO TOLL, CAR TOLL
ROAD, BUS OR BUSWAY, and TRAIN OR LIGHT RAIL. The five attributes for the
public transport alternatives are total in vehicle time, frequency of service, closest stop
to home, closest stop to destination, and fare. The attributes for the car alternatives are
travel time, fuel cost, parking cost, travel time variability, and for the toll road departure
time and toll charge. Three levels were selected for each attribute. The design allows for
six alternative specific main effect models for car no toll, car toll road, bus, busway,
train, and light rail. Linear by linear interactions are estimable for both car models, and
generically for the bus/busway and train/light rail models. While cross effects have been
assumed negligible, the four alternative design is perfectly balanced across all
attributes. The design has been blocked into 27 versions of size three also balanced for
every attribute, that is each person sees each level of each attribute exactly one time.

The base design for the travel choice task was a 27 x 3
27
orthogonal fractional factorial
in 81 runs. The 27 level factor was used for creating the versions, and the three level
factors selected and arranged to meet the design requirements. Two factors each at two
levels were constructed for the bus/busway and train/light rail modes. In the design, bus
and train appear in 36 scenarios while busway and light rail appear in 45.


Description of Variables in the Travel Choice Experiment

In this experiment the respondent is asked to consider their daily commute trip. The aim
of the experiment is to test a range of instruments for altering mode and departure time
choice. There are 12 types of showcards, with three trip lengths and four combinations
of public transport. The public transport combinations are: bus vs. LRT, bus vs. train
(heavy rail), busway vs. LRT, and busway vs. train. The trip length showcards are
printed in different colours for ease of use by interviewers when matching respondents
current trip lengths with those on the showcards. The four public transport combinations
appear in each showcard with incidence determined by the statistical design. The range
of attribute levels are summarised in Table C1, the design in Table C2, and the
contextual statement is given in Figure C1. Illustrative showcards are given in Figure
C2.





TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
47
Travel Time to Work

There are three different sets of showcards representing short (under 30 minutes),
medium (30-45 minutes) and long (45 minutes and over) commutes. They are matched
to the commute times currently experienced by respondents. Within each set of
showcards, there are three levels of travel times. All public transport options have the
same levels as each other, allowing for different combinations across the public
transport pairs in each replication. Having the levels the same enables us to investigate
the influence that image (through the mode-specific constants) plays in determining
preferences within the public transport modes after allowing for the influence of the
balanced set of attributes and levels in the design. Travel times on the tolled road have
been selected so that it is never worse than the time on the non-tolled route.


Pay Toll if You Leave at this Time (otherwise free)

The tolled route option only has a toll at peak congestion times. The peak over which
the toll applies is varied to see what impact a short, medium and long toll period have
on mode and departure time decisions.


Toll (one-way)

The toll only applies to the tolled routes when the respondents commute trip
commences within the times specified by the previous variable. There are three levels
of toll for each travel time set, with toll levels increasing for the longer travel time sets.
Tolls in excess of current tolls in Sydney have been included to assess the impact of
increases beyond the current levels in one City. The toll on the M4 in Sydney is $1.50
for a car; the M5 toll is $2 on the section, currently in place, but is likely to increase up
to $4 when the second section is open. The M2 is currently up to $2.50. Tolls in the
experiment vary from $1 to $6.


Fuel Cost (per day)

Fuel cost is varied from current levels to a tripling of current levels to assess possible
changes commuters will make as a response to large increases in fuel prices. The daily
fuel cost for the commute trip on the tolled road is assumed to be equal to or lower than
that experienced on the non-tolled route. Fuel costs can be as high as $15 per day for
trips in excess of 45 minutes on a free route. This represents a tripling of fuel prices.


Parking Cost (per day)

Another method for reducing the attractiveness of private vehicle use, particularly in
central city areas, is to increase parking charges. Three levels of parking charge are
used in the experiment to see how sensitive respondents are to parking costs. A fixed set
of charges ranging from free to $20 in evaluated.


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
48
Travel Time Variability

This variable is calculated for private vehicle modes only, with levels based on 0, 20%
and 30% of the average trip time on no toll roads, and 0, 5%, and 10% of the
average trip time for tolled routes. Toll roads will always be equal to or better than
non-tolled roads on trip reliability.


Total Time in the Vehicle (one-way)

For public transport only, this variable refers to the time spent travelling on a train, bus,
light rail (LRT) or busway. There are three travel time sets to match those of private
vehicles. Only two public transport systems are compared or traded off at once to make
the experiment more realistic for the respondent. Thus, there are four sets of public
transport combinations, listed above. Any other combinations are not meaningful. All
public transport options share the same experimental levels enabling the investigation of
the role of image in respondents preferences.


Frequency of Service

This variable gives the number of minutes between each service, and has three levels.
The frequency for all modes has a range from a low of 5 minutes to a high of 25
minutes.


Time from Home to Your Closest Stop

The distance from the respondents home to the public transport stop, in minutes, is
measured in both walk time and time travelling by a motorised form of transport. The
respondent was asked to indicate which means of access they would us if they were to
use public transport. There are three levels: 5,15 and 25 minutes walk time, and 4,6 and
8 minutes by a motorised mode. This same logic is applied to the Time to Your
Destination from the Closest Stop except that the only motorised mode available is bus.
It is very rare that a commuter will use a car to complete a trip after alighting from
public transport. The taxi option is excluded.


Return Fare

This variable gives the return fare in dollars. This has three levels, with the same fare
sets being used for all public transport modes for each trip length.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
49
Table C1 The Set of Attributes and Attribute Levels in the Travel Choice Experiment
SHORT (< 30
mins.)
Car no
toll
Car toll
road
PUBLIC
TRANSPORT
Bus Train Busway Light
Rail
Tr avel t i me t o
wor k
15, 20, 25 10, 12, 15 Tot al t i me i n
t he vehi cl e
( one- way)
10, 15, 20 10, 15, 20 10, 15, 20 10, 15, 20
Pay t ol l i f
you l eave at
t hi s t i me
( ot her wi se
f r ee)
None 6- 10, 6: 30-
8: 30, 6: 30-
9
Fr equency of
ser vi ce
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Tol l ( one- way) None 1, 1. 5, 2 Ti me f r om
home t o your
cl osest st op
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Fuel cost ( per
day)
3, 4, 5 1, 2, 3 Ti me t o your
dest i nat i on
f r om t he
cl osest st op
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Par ki ng cost
( per day)
Fr ee, $10, $
20
Fr ee, $10, $2
0
Ret ur n f ar e 1, 3, 5 1, 3, 5 1, 3, 5 1, 3, 5
Ti me
var i abi l i t y
0, 4, 6 0, 1, 2


MEDIUM (30-45
mins.)

Tr avel t i me t o
wor k
30, 37, 45 20, 25, 30 Tot al t i me i n
t he vehi cl e
( one- way)
20, 25, 30 20, 25, 30 20, 25, 30 20, 25, 30
Pay t ol l i f
you l eave at
t hi s t i me
( ot her wi se
f r ee)
None 6- 10, 6: 30-
8: 30, 6: 30-
9
Fr equency of
ser vi ce
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Tol l ( one- way) None 2, 3, 4 Ti me f r om
home t o your
cl osest st op
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Fuel cost ( per
day)
6, 8, 10 2, 4, 6 Ti me t o your
dest i nat i on
f r om t he
cl osest st op
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Par ki ng cost
( per day)
Fr ee, $10, $
20
Fr ee, $10, $2
0
Ret ur n f ar e 2, 4, 6 2, 4, 6 2, 4, 6 2, 4, 6
Ti me
var i abi l i t y
0, 7, 11 0, 2, 4


LONG (>45
mins.)

Tr avel t i me t o
wor k
45, 55, 70 30, 37, 45 Tot al t i me i n
t he vehi cl e
( one- way)
30, 35, 40 30, 35, 40 30, 35, 40 30, 35, 40
Pay t ol l i f
you l eave at
t hi s t i me
( ot her wi se
f r ee)
None 6- 10, 6: 30-
8: 30, 6: 30-
9
Fr equency of
ser vi ce
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Ever y
5, 15, 25
Tol l ( one- way) None 3, 4. 5, 6 Ti me f r om
home t o your
cl osest st op
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Wal k
5, 15, 25
Car / Bus
4, 6, 8
Fuel cost ( per
day)
9, 12, 15 3, 6, 9 Ti me t o your
dest i nat i on
f r om t he
cl osest st op
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Wal k
5, 15, 25
Bus 4, 6, 8
Par ki ng cost
( per day)
Fr ee, $10, $
20
Fr ee, $10, $2
0
Ret ur n f ar e 3, 5, 7 3, 5, 7 3, 5, 7 3, 5, 7
Ti me
var i abi l i t y
0, 11,
17
0, 7, 11


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
50
Table C2 The Statistical Design for the Final Travel Choice Experiment

1 2 3 4 1 2 3 4 5 6 1 2 3 4 5 1 2 3 4 5 1 2 BLK

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 0
1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 1 0 0

2 1 1 2 2 2 2 1 2 2 0 2 2 2 0 0 0 0 0 0 1 0 1
0 2 2 0 0 0 0 2 0 0 2 1 1 1 2 2 2 2 2 2 1 1 1
1 0 0 1 1 1 1 0 1 1 1 0 0 0 1 1 1 1 1 1 0 0 1

2 0 0 2 2 2 2 0 2 2 2 0 0 0 2 2 2 2 2 2 0 1 2
1 2 2 1 1 1 1 2 1 1 0 1 1 1 0 0 0 0 0 0 1 1 2
0 1 1 0 0 0 0 1 0 0 1 2 2 2 1 1 1 1 1 1 0 1 2

0 1 0 1 1 2 0 1 1 2 1 1 0 1 0 1 0 0 2 2 0 1 3
2 0 2 0 0 1 2 0 0 1 2 2 1 2 1 2 1 1 0 0 0 1 3
1 2 1 2 2 0 1 2 2 0 0 0 2 0 2 0 2 2 1 1 1 1 3

0 0 2 1 1 2 0 0 1 2 0 2 1 2 2 0 2 2 1 1 0 0 4
2 2 1 0 0 1 2 2 0 1 1 0 2 0 0 1 0 0 2 2 1 1 4
1 1 0 2 2 0 1 1 2 0 2 1 0 1 1 2 1 1 0 0 1 0 4

2 1 0 0 0 1 2 1 0 1 0 1 0 1 2 0 2 2 1 1 1 0 5
1 0 2 2 2 0 1 0 2 0 1 2 1 2 0 1 0 0 2 2 0 0 5
0 2 1 1 1 2 0 2 1 2 2 0 2 0 1 2 1 1 0 0 1 1 5

1 0 1 0 0 2 1 0 0 2 1 1 2 1 2 1 2 2 0 0 0 0 6
0 2 0 2 2 1 0 2 2 1 2 2 0 2 0 2 0 0 1 1 1 1 6
2 1 2 1 1 0 2 1 1 0 0 0 1 0 1 0 1 1 2 2 1 0 6

2 0 1 1 1 0 2 0 1 0 2 1 2 1 0 2 0 0 1 1 0 1 7
1 2 0 0 0 2 1 2 0 2 0 2 0 2 1 0 1 1 2 2 1 1 7
0 1 2 2 2 1 0 1 2 1 1 0 1 0 2 1 2 2 0 0 0 1 7

0 0 1 2 2 1 0 0 2 1 0 1 2 1 1 0 1 1 2 2 0 0 8
1 1 2 0 0 2 1 1 0 2 2 0 1 0 0 2 0 0 1 1 1 0 8
2 2 0 1 1 0 2 2 1 0 1 2 0 2 2 1 2 2 0 0 1 1 8

2 2 0 0 2 2 0 0 1 0 0 0 2 1 1 1 2 0 2 0 1 0 9
0 0 1 1 0 0 1 1 2 1 2 2 1 0 0 0 1 2 1 2 1 1 9
1 1 2 2 1 1 2 2 0 2 1 1 0 2 2 2 0 1 0 1 0 0 9

0 2 0 1 0 0 1 0 2 1 1 0 2 1 2 2 0 1 0 1 1 0 10
1 0 1 2 1 1 2 1 0 2 0 2 1 0 1 1 2 0 2 0 1 1 10
2 1 2 0 2 2 0 2 1 0 2 1 0 2 0 0 1 2 1 2 0 0 10

0 1 2 1 0 0 1 2 2 1 0 1 0 2 1 1 2 0 2 0 0 1 11
2 0 1 0 2 2 0 1 1 0 1 2 1 0 2 2 0 1 0 1 1 1 11
1 2 0 2 1 1 2 0 0 2 2 0 2 1 0 0 1 2 1 2 0 1 11

2 0 0 1 0 1 0 1 2 2 1 1 2 2 1 2 2 0 1 2 1 1 12
0 1 1 2 1 2 1 2 0 0 0 0 1 1 0 1 1 2 0 1 0 1 12
1 2 2 0 2 0 2 0 1 1 2 2 0 0 2 0 0 1 2 0 0 1 12

2 2 2 1 0 1 0 0 2 2 0 2 0 0 0 1 1 2 0 1 1 0 13
1 1 1 0 2 0 2 2 1 1 1 0 1 1 1 2 2 0 1 2 0 0 13
0 0 0 2 1 2 1 1 0 0 2 1 2 2 2 0 0 1 2 0 1 1 13

1 0 0 0 2 0 2 1 1 1 0 1 2 2 0 1 1 2 0 1 1 1 14
0 2 2 2 1 2 1 0 0 0 1 2 0 0 1 2 2 0 1 2 1 0 14
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
51
2 1 1 1 0 1 0 2 2 2 2 0 1 1 2 0 0 1 2 0 0 0 14

0 2 1 0 2 1 1 0 1 2 1 1 1 2 0 2 1 2 2 0 1 0 15
2 1 0 2 1 0 0 2 0 1 2 2 2 0 1 0 2 0 0 1 0 0 15
1 0 2 1 0 2 2 1 2 0 0 0 0 1 2 1 0 1 1 2 1 1 15

1 2 1 1 0 2 2 0 2 0 2 1 1 2 1 0 2 0 0 1 0 1 16
0 1 0 0 2 1 1 2 1 2 0 2 2 0 2 1 0 1 1 2 0 1 16
2 0 2 2 1 0 0 1 0 1 1 0 0 1 0 2 1 2 2 0 1 1 16

0 0 2 0 2 1 1 1 1 2 2 0 0 1 1 0 2 0 0 1 1 1 17
1 1 0 1 0 2 2 2 2 0 1 2 2 0 0 2 1 2 2 0 0 0 17
2 2 1 2 1 0 0 0 0 1 0 1 1 2 2 1 0 1 1 2 1 0 17

1 1 0 0 1 1 0 0 2 0 0 0 1 2 2 2 1 0 1 0 1 1 18
0 0 2 2 0 0 2 2 1 2 1 1 2 0 0 0 2 1 2 1 1 0 18
2 2 1 1 2 2 1 1 0 1 2 2 0 1 1 1 0 2 0 2 0 0 18

2 1 0 1 2 2 1 0 0 1 1 0 1 2 0 0 2 1 2 1 1 1 19
1 0 2 0 1 1 0 2 2 0 2 1 2 0 1 1 0 2 0 2 1 0 19
0 2 1 2 0 0 2 1 1 2 0 2 0 1 2 2 1 0 1 0 0 0 19

1 2 1 0 1 1 0 1 2 0 1 2 0 1 0 0 2 1 2 1 0 1 20
0 1 0 2 0 0 2 0 1 2 2 0 1 2 1 1 0 2 0 2 1 1 20
2 0 2 1 2 2 1 2 0 1 0 1 2 0 2 2 1 0 1 0 0 1 20

0 1 2 0 1 2 2 0 2 1 2 2 2 1 0 1 2 1 1 0 1 1 21
1 2 0 1 2 0 0 1 0 2 1 1 1 0 2 0 1 0 0 2 0 1 21
2 0 1 2 0 1 1 2 1 0 0 0 0 2 1 2 0 2 2 1 0 1 21

0 0 1 0 1 2 2 2 2 1 1 0 0 2 2 0 1 0 0 2 1 0 22
2 2 0 2 0 1 1 1 1 0 2 1 1 0 0 1 2 1 1 0 0 0 22
1 1 2 1 2 0 0 0 0 2 0 2 2 1 1 2 0 2 2 1 1 1 22

2 1 2 2 0 1 1 0 1 0 1 2 2 1 2 0 1 0 0 2 1 1 23
1 0 1 1 2 0 0 2 0 2 2 0 0 2 0 1 2 1 1 0 1 0 23
0 2 0 0 1 2 2 1 2 1 0 1 1 0 1 2 0 2 2 1 0 0 23

2 1 1 0 1 0 1 0 2 2 1 1 0 0 1 0 0 2 1 0 1 1 24
1 0 0 2 0 2 0 2 1 1 2 2 1 1 2 1 1 0 2 1 1 0 24
0 2 2 1 2 1 2 1 0 0 0 0 2 2 0 2 2 1 0 2 0 0 24

0 1 1 1 2 1 2 0 0 0 2 1 0 0 2 1 1 0 2 1 1 1 25
2 0 0 0 1 0 1 2 2 2 0 2 1 1 0 2 2 1 0 2 0 1 25
1 2 2 2 0 2 0 1 1 1 1 0 2 2 1 0 0 2 1 0 0 1 25

2 2 2 0 1 0 1 1 2 2 2 0 2 2 2 1 1 0 2 1 0 0 26
0 0 0 1 2 1 2 2 0 0 1 2 1 1 1 0 0 2 1 0 1 0 26
1 1 1 2 0 2 0 0 1 1 0 1 0 0 0 2 2 1 0 2 1 1 26

Notes: Each block of columns relates to the four alternatives on each showcard. The non-tolled car route
has 4 attributes because the toll and departure time do not apply. The fifth set of two columns indicates
the pair of public modes to be included in each replication.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
52
Figure C1 The Contextual Statement Associated with the Travel Choice Experiment

13. We would now like to ask some questions about your trip TO work. We need to speak to
the person in the household who completed the Commuter Questionnaire.

IF THERE IS NO COMMUTER IN THE HOUSEHOLD GO TO QUESTION 14

How long does it take you to travel to work, door to door, on a normal day (ie. without any
abnormal traffic or public transport delays)
READ OPTIONS
Less than 30 minutes ....... ....................1
30 to 45 minutes ......... ..........................2
Over 45 minutes ....... ............................3

SELECT THE RELEVANT SET OF CHOICE CARDS FOR THE
RESPONDENTS TRAVEL TIME

We are going to show you 3 possible choices of transport options in your area. We
are not suggesting that these changes will happen in your area, we are just using
these options to understand how individuals and households choose to cope with
possible changes to transport. We need your help to try to understand how
transport facilities can best service your needs under a variety of possible
conditions.

We would like you to consider each choice with reference to your current trip TO
work.


TRAVEL CHOICE 1.


CHOOSE A SET OF THREE CARDS AT RANDOM FROM THE TRAVEL TIME
SET WHICH IS RELEVANT FOR THE RESPONDENT. TAKE ONE OF THOSE
CARDS.

WHAT IS THE NUMBER OF THE CARD ___________

This is the first choice. (SHOW THE RESPONDENT THE CARD AND EXPLAIN
THE FEATURES OF THE OPTIONS)
If these were the options available for your trip to work, which one would you choose?

Car toll route........................ .......................1
Car no toll route........................ ..................2
Bus ..................... ........................................3
Train ...................... .....................................4
Light rail (Tram) ........................ ..................5
Busway....................... ................................6
(A busway is a dedicated lane for buses
which extends for most of your trip)
Which set of times did you consider when you were thinking about getting to/from the
public transport options (regardless of whether you chose public transport)?
From home: To your destination:
Walk ................... ................1 Walk ..............................1
Car/bus.................. .............2 Bus ...............................2
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
53

If you were to travel by private vehicle on either a toll or a no toll route (regardless of
whether you chose these options), would you
Drive alone ... ..............................................1
Carpool or share a ride as driver.. ..............2
Carpool or share a ride as passenger. .......3

If these were the set of travel choices that were available for your trip to work, do you find
them so unattractive that you would seriously consider
Yes No
Changing your work place .............................. 1 ..................2
Changing where you live................................. 1 ................2

IF THE RESPONDENT CHOSE EITHER OF THE CAR OPTIONS CONTINUE
WITH THE FOLLOWING QUESTION, IF NOT GO TO CHOICE 2.

Given the choice that you have made to travel by private vehicle on a (TOLL/NO TOLL
ROUTE) how would this affect the time that you leave home compared with now. Would
you leave
Earlier, if so by how many minutes ............1 ________mins
Later, if so by how many minutes ..............2 ________mins
At the same time..........................................3





TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
54
Figure C2 Example of the Format of a Travel Choice Experiment Showcard

Card Number S01A


1. CAR
TOLL ROAD
2. CAR
NO TOLL

Tr avel t i me t o
wor k
10 min. 15 min.
Ti me var i abi l i t y None None

Tol l ( one way) $1.00 free
Pay t ol l i f you
l eave at t hi s
t i me ( ot her wi se
f r ee)
6-10 am

Fuel cost ( per
day)
$1.00 $3.00
Par ki ng cost ( per
day)
Free Free



3. BUS 4. TRAIN

Tot al t i me i n t he
vehi cl e ( one way)
10 min. 10 min.
Ti me f r omhome t o
your cl osest st op
Wal k Car / Bus
5 min. 4 min.
Wal k Car / Bus
5 min. 4 min.
Ti me t o your
dest i nat i on f r om
t he cl osest st op
Wal k Bus
5 min. 4 min.
Wal k Bus
5 min. 4 min.

Fr equency of
ser vi ce
Every 5 min. Every 5 min.

Ret ur n f ar e $1.00 $1.00
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
55
C2 Vehicle Type Choice

The aim of the vehicle choice experiment is twofold firstly to see where
respondents preferences lie with regard to conventional vehicles given a range of price
and running cost attributes, and secondly to see whether respondents are willing to
consider alternate fuel or electric vehicles as substitutes for conventional vehicles given
a price, running cost, and some physical differences.

Each vehicle attribute has three levels with different levels employed to see which
variables are most important in a consumers product evaluation. The vehicle choice
design involved nine alternatives. These alternatives were three vehicle sizes crossed
with three fuel types, petrol, electric, and alternate fuel. Across the nine alternatives
there are 37 attributes at three levels and 6 at two levels for a total of 80 degrees of
freedom to estimate. The orthogonal fractional factorial for 40 factors each at three
levels in 81 runs was used as a base. A nine level factor was created from which the six
two level and one three level attribute were generated. With the exception of this one
three level attribute having small correlation with the two level attributes, the resultant
design is fully orthogonal.

The vehicle choice design allows (nine) alternative specific models including all cross
effects. Since cross effects are assumed to be negligible, factors have been arranged so
that linear by linear interactions of the price, registration fee, and fuel cost are estimable
in each model. No blocking factor was available, hence the 27 versions of size three
were randomly generated.

The experiments consider the following influences on choice.


Class of Vehicle
Vehicles have been grouped into three classes small, medium and large (Table b).
These classes reflect the differences in price, weight, and engine capacity, which are
suitable representative dimensions for classifying vehicles from a vehicle purchasers
perspective. Small vehicles are those under two litres in capacity, and range from 1 litre
micro cars like Daihatsu Charade to 1.6 and 1.8 litre small cars like a Ford Laser and
Toyota Corolla. Medium sized vehicles are those with engine capacities from two to
three litres inclusive, and consist of cars such as the Toyota Camry and Mitsubishi
Magna. Large cars are those over three litres, such as Holden Commodore and Ford
Falcon.

Within these classes are vehicles which may range from base sedans and wagons to
luxury vehicles. In order to capture some of the variation within each class respondents
are asked initially to identify the body type of car they prefer. This will distinguish
sedans, wagons, sports cars, light commercials etc. The experiment is then conducted
within the context of the selected body type. Although initially we planned to ask a
respondent to indicate (circle) from a list (see Table C3) the type of vehicles they are
thinking of when faced with a choice of small, medium or large vehicles, we changed
this to a description of each size class in terms of engine capacity, vehicle weight and
number of cylinders. Reference to a particular make and model of vehicle would have
created problems when the price range was shown. The main concern was at the low
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
56
end of the price range. For example, would a respondent accept that a Range Rover
could cost $30,000?

Table C3 A summary of Vehicle Makes and Models in each Vehicle Class
SMALL MEDIUM LARGE

Basic
Dai hat su Appl ause
Dai hat su Char ade
Dai hat su Mi r a
For d Fest i va
For d Laser
Hol den Bar i na
Hol den Nova
Hyundai Excel
Mazda 121
Mi t subi shi Lancer
Ni ki
Ni ssan Pul sar
Suzuki Swi f t
Toyot a Cor ol l a

Luxury
BMW3 ser i es
Eunos 300
Honda Ci vi c
Mazda 323
Peugeot 205
Subar u I mpr eza

Sports
Hol den Cal i br a
Honda I nt egr a
Mazda MX- 5
Ni ssan NX
Toyot a MR2
Toyot a Paseo

Other
Dai hat su Fer oza
Lada Ni va
Suzuki Vi t ar a
Suzuki Si er r a

Basic
For d Tel st ar
Hol den Apol l o
Hyundai Sonat a
Mazda 626
Mi t subi shi Magna
Ni ssan Bl uebi r d
Subar u Li ber t y
Toyot a Camr y

Luxury
Audi 80, 100
BMW5 ser i es
Ci t r oen BX
Eunos 500
For d Tel st ar V6
Honda Accor d
Lexus ES300
Mazda 626 V6
Mer cedes 170, 180
( C cl ass)
Mi t subi shi Ver ada
V6
Ni ssan Maxi ma
Peugeot 405
Renaul t 19, 21
Saab 900, 9000
Toyot a Vi ent a V6
Vol vo 440

Sports
Honda Pr el ude
Mazda MX- 6
Mi t subi shi 3000
GT
Subar u SVX
Toyot a Cel i ca

Other
Dai hat su Rocky
For d Rai der
Mazda MPV
Ni ssan Pat hf i nder
Ni ssan Ser ena
Toyot a 4- Runner
Toyot a
Spaci aToyot a
Tar ago
Vol kswagen
Car avel l e

Basic
For d Fal con
Hol den Commodor e
Toyot a Lexcen

Luxury
BMW 7 and 8
ser i es
Eunos 800
For d LTD
For d Fai r l ane
Hol den Capr i ce
Hol den St at esman
Honda Legend
J aguar
Lexus LS400
Mazda 929
Mer cedes E cl ass
Mer cedes S cl ass
Ni ssan I nf i ni t i
Rover
Vol vo 740, 850,
940, 960

Sports
Ni ssan 300 ZX
Por sche

Other
Hol den J acker oo
Land Rover
Di scover y
Mi t subi shi Paj er o
Ni ssan Pat r ol
Range Rover
Toyot a
Landcr ui ser


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
57
Price of Vehicles

New cars typically range in price from $15,000 to $100,000. Thus, a fairly wide range
of new prices is needed in each size class of vehicle. The prices used have been
generated using approximations gained from studying Glasses Dealers Guide to new
and used vehicle prices. Table C4 provides an indication of the types of cars that could
fit into each price category. Since age is important in determining the emissions
produced by a vehicle, it is necessary to give people the choice of choosing a new car or
a range of different aged used cars. Again for used car prices, Glasses Dealers Guide
has been used to generate the price ranges in each class to cover the full range of basic
to luxury models. Within the range one can interpolate quite substantial increases or
decreases in sales tax, as an instrument available to government to adjust to fuel
consumption of the fleet.

Table C4 The Classification of Automobiles (Not Shown to the Household)
SMALL CAR
$15,000 $25,000 $35,000
Daihatsu Charade Daihatsu Applause Eunos 300
Ford Festiva Ford Laser Holden Calibra
Holden Barina Honda Civic Honda Integra
Hyundai Excel Nissan Pulsar Mazda MX-5
Suzuki Swift Toyota Corolla Nissan NX

MEDIUM CAR
$30,000 $40,000 $50,000
Hyundai Sonata Honda Accord Audi 80
Mitsubishi Magna Mazda 626 Honda Prelude
Nissan Bluebird Nissan Maxima Mazda MX-6
Subaru Liberty Toyota Celica Saab 900
Toyota Camry Toyota Tarago Volvo 440

LARGE CAR
$30,000 $55,000 $80,000
Holden Commodore Holden Statesman Honda Legend
Ford Falcon Ford Fairlane Mercedes E Class
Toyota Lexcen Mazda 929 Nissan 300 ZX
Mitsubushi Pajero Toyota Landcruiser Range Rover
Volvo 940

In general, conventional and alternative fuel vehicles are assumed to have depreciated
by 25% after 2 years, 50% after 5 years, and 70% after 10 years. Electric vehicle are
assumed to retain their value better than conventional and alternate fuel vehicles since
the expectation is that they will have fewer moving parts, reducing maintenance costs as
the vehicle ages. They are assumed to depreciate by 20% after 2 years, 40% after 5
years, and 55% after 10 years. The new price levels of electric vehicles are set slightly
higher than other vehicles because it appears they will cost more to produce.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
58


Registration Fee

A possible policy to help reduce greenhouse gas emissions is to place a higher
registration fee on conventionally fuelled cars, with the charge linked to the level of
emissions produced. To test the influence of such a charge, a variable registration
charge is used in the experiment which increases with the size of vehicle, and is
considerably lower for electric and alternate fuelled vehicles. Compulsory insurance is
excluded from this charge


Fuel Cost to Travel 500kms

This variable is used to test respondents sensitivity to changes in fuel prices (increasing
fuel prices is one of the simpler policy measures available to attempt to reduce
greenhouse gas emissions) and also to demonstrate the fuel cost savings that can be
made with electric and alternative fuelled vehicles (be this real or legislated). The
variable is described as it is because generally conventional vehicles can travel 500kms
on a tank of petrol, so respondents are familiar with how much it currently costs to fill
their car up, and could use this as a basis to understand the levels of the variables. A
cost per litre of fuel would be meaningless for electric vehicles since they run on
batteries.

The base values for this variable have been calculated based on a fuel price of 70 a
litre and average fuel efficiency (sourced from the 1993-94 Fuel Consumption Guide
(Department of Primary Industries and Energy) 7 litres per 100kms for small
vehicles, 10 litres per 100kms for medium vehicles, and 13 litres per 100 kms for large
vehicles). Approximations have been used for non-conventional vehicles based on
expected future levels of performance. Fuel price increases of 100% and 200% are
included.


Fully Fuelled Range

This is given only for electric and alternative fuel vehicles because it is envisaged that
the range of conventional vehicles will not change (manufacturers have provided
similar ranges for several decades now), but that a major limitation of electric and
alternate fuel vehicles may be the distance they can travel before refuelling is required.
The variable is expressed as a percentage of the range of a conventional vehicle of the
same size. The actual range of a conventional vehicle is not given, firstly because it
generally does not vary greatly and is thus not important for choice, and secondly
because of the difficulties that may be caused in respondents trying to calculate
percentages. Thus the percentages are given to provide an indication of the reduced
range of non-conventional vehicles.



Acceleration

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
59
Instead of a percentage as used for the variable, fully fuelled range, descriptive terms
have been used for the different levels of performance of electric and alternate fuel
vehicles as compared to conventional vehicles. One of the major differences between
conventional and other fuelled vehicles is likely to be the rate at which they accelerate.
Thus, acceleration is set at three levels: the same as conventional vehicles, slightly
slower, and considerably slower than conventional vehicles. Such descriptions were
thought to be more meaningful for a respondent than percentages or seconds to reach
100kms/hr.


Boot Size

Luggage capacity is another of the possible problems with non-conventional vehicles.
More room for fuel or batteries may mean reduced boot space for luggage. This
variable, again expressed as a percentage of the boot size of a conventional vehicle in
the same range, is aimed to examine the importance of smaller boot sizes in
respondents future purchase decisions.

Each household with vehicles will be asked to participate in the vehicle choice
experiment by evaluating the set of possible replacements as if they would be replacing
a vehicle. At the completion of each replication we will ask the respondent to indicate
which current vehicle they would replace with the most preferred vehicle identified in
the experiment. The response could include the decision to not replace any vehicle. If
the household currently has no vehicles, we will ask them to assume that they were
acquiring a vehicle. The context treats the choice process as one faced today. The set of
attributes and their levels are summarised in Table C5. The statistical design associated
with the vehicle choice experiment is given in Table C6. The context of the experiment
is set out in Figure C3.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
60
Table C5 The Attributes and Attribute Ranges in the Final Vehicle Choice Experiment
Conventional Vehicle OPTION 1 OPTION 2 OPTION 3
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,25,35 30,40,50 30,55,80
2 yr old vehicle ($000) 11,18,26 22,30,37 22,42,62
5 yr old vehicle ($000) 7,12,17 15,20,25 15,30,45
10 yr old vehicle ($000) 5,7,10 9,12,16 10,20,30
Other costs: Regn. fee (excl. any ins.) $ 150,250,350 200,350,500 250,400,550
Fuel cost to travel 500kms ($) 20,40,60 30,60,90 40,80,120

Electric Vehicle OPTION 4 OPTION 5 OPTION 6
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,28,40 25,40,55 30,70,110
2 yr old vehicle ($000) 12,22,32 20,32,44 24,55,90
5 yr old vehicle ($000) 9,17,24 15,24,33 18,42,65
10 yr old vehicle ($000) 6,12,17 11,17,25 13,30,50
Other costs: Regn. fee (excl. any ins.) $ 50,100,150 75,125,175 100,150,200
Fuel cost to travel 500kms ($) 10,20,30 15,30,45 20,40,60
Features:
(compared to
Fully fuelled range (% of) 90,70,50 90,70,50 90,70,50
conventional
vehicles)
Acceleration s,ss,cs s,ss,cs s,ss,cs
Boot size (% of) 90,60 90,60 90,60

Alternate Fuel Vehicle OPTION 7 OPTION 8 OPTION 9
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,25,35 30,40,50 30,55,80
2 yr old vehicle ($000) 11,18,26 22,30,37 22,42,62
5 yr old vehicle ($000) 7,12,17 15,20,25 15,30,45
10 yr old vehicle ($000) 5,7,10 9,12,16 10,20,30
Other costs: Regn. fee (excl. any ins.) $ 50,100,150 75,125,175 100,150,200
Fuel cost to travel 500kms ($) 15, 30, 45 20,40,60 25,55, 85
Features:
(compared to
Fully fuelled range (% of) 100,85,70 100,85,70 100,85,70
conventional
vehicles)
Acceleration s,ss,cs s,ss,cs s,ss,cs
Boot size (% of) 90,70 90,70 90,70
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
61
Table C6 The Statistical Design for the Final Vehicle Choice Experiment

Conventional Electric Vehicle Alternate Fuel Vehicle
S M L S M L S M L
1 2 3 1 2 3 1 2 3 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 BLK

1 1 0 1 2 0 1 1 0 0 0 1 1 1 0 0 1 1 2 2 0 0 2 2 1 2 0 0 0 1 2 2 0 0 1 1 1 2 0 0 2 2 2 0 0 1
0 1 2 0 1 2 1 2 2 2 0 2 0 2 1 0 2 0 0 0 1 1 2 0 1 1 1 1 2 1 0 1 1 2 1 2 0 2 0 0 1 2 2 2 0 1
0 0 1 1 0 1 2 2 1 2 0 1 1 2 1 1 2 0 1 1 1 0 2 0 0 2 0 0 1 2 1 0 0 2 0 1 1 2 0 1 0 1 2 1 1 1

0 2 0 1 2 0 1 0 0 1 1 0 2 2 0 2 1 1 2 1 0 0 1 1 0 1 1 1 1 0 0 0 1 2 1 1 0 0 0 0 1 1 2 2 1 2
1 2 0 0 0 0 0 1 0 2 2 1 2 2 1 1 0 0 0 1 1 0 1 1 1 1 0 2 2 1 2 2 0 1 0 0 1 2 1 0 1 1 0 2 1 2
2 0 0 1 2 0 1 2 0 2 2 2 0 0 1 1 1 1 2 0 1 0 0 0 2 0 0 2 2 2 1 1 0 1 1 1 2 1 1 0 0 0 2 2 1 2

0 2 0 0 2 0 2 1 0 0 2 0 2 2 1 2 0 2 2 0 0 1 0 2 1 2 0 0 2 0 1 1 1 2 0 2 1 1 1 1 0 2 2 1 0 3
0 2 2 0 2 2 2 1 2 2 1 2 1 0 1 1 2 1 1 0 0 0 2 1 0 2 0 1 0 1 2 0 1 0 1 0 2 1 1 2 1 0 0 1 0 3
2 1 0 0 0 0 0 2 0 1 1 2 1 1 0 2 0 0 0 2 0 0 2 2 2 2 1 1 1 2 1 1 1 2 0 0 2 1 0 0 2 2 0 2 1 3

0 0 1 0 0 1 0 0 1 1 1 1 1 2 0 1 1 1 1 0 0 1 1 1 1 0 0 2 2 2 2 1 0 2 2 2 2 0 0 2 2 2 2 0 0 4
1 2 2 1 0 2 2 0 2 2 0 0 1 0 1 0 0 1 2 2 1 1 1 2 2 0 1 1 2 2 2 0 1 2 2 0 1 1 0 0 0 1 1 2 0 4
2 0 2 2 2 2 0 1 2 2 0 1 2 1 1 0 1 2 1 1 1 1 0 1 0 2 1 1 2 0 1 2 1 2 0 1 2 0 0 0 2 0 0 2 0 4

0 1 0 0 1 0 1 2 0 0 1 0 1 1 1 1 0 1 1 0 1 2 0 1 2 1 1 0 1 0 2 2 1 1 0 1 2 2 0 2 0 1 1 2 0 5
2 1 0 2 0 0 1 0 0 0 2 2 1 1 1 2 2 1 0 1 0 1 1 0 0 0 0 0 2 2 2 2 1 2 2 1 0 2 1 1 1 0 0 1 0 5
0 2 2 2 2 2 0 2 2 1 2 2 1 0 1 1 1 2 1 2 0 1 1 2 1 0 1 0 1 1 0 1 0 0 0 1 0 2 1 0 0 1 0 0 1 5

0 2 1 1 2 1 1 0 1 2 2 1 0 1 0 0 2 2 0 1 0 1 2 2 1 1 1 0 0 2 2 1 1 1 0 0 2 0 0 2 0 0 1 2 1 6
2 1 2 1 0 2 2 1 2 1 2 1 0 2 1 1 0 1 2 0 0 1 2 0 0 1 1 0 1 0 1 2 0 0 2 0 2 0 1 0 1 2 1 0 1 6
2 2 2 1 1 2 0 0 2 1 0 1 1 0 0 2 0 2 0 0 1 0 2 1 2 2 1 0 2 0 0 1 0 1 2 1 1 2 1 2 1 0 2 1 0 6

2 2 2 2 1 2 2 2 2 2 2 1 1 0 0 2 1 1 0 1 0 2 0 0 1 1 0 1 1 0 2 0 0 1 0 0 0 1 0 1 2 2 2 0 0 7
0 1 1 0 1 1 1 2 1 1 2 1 2 0 1 2 1 2 2 0 1 0 1 2 0 1 1 2 0 2 1 0 1 0 2 0 1 2 0 1 2 0 0 2 0 7
2 2 0 0 1 0 1 1 0 1 2 2 2 2 1 0 0 1 1 2 1 2 2 0 1 0 0 1 2 2 0 0 0 0 0 1 1 0 0 2 2 0 1 1 1 7

0 2 1 0 2 1 2 1 1 1 0 1 0 1 1 0 1 0 0 0 0 2 1 0 2 2 0 2 1 2 0 2 1 1 2 1 0 1 1 0 2 1 1 1 0 8
2 1 2 0 0 2 0 2 2 0 0 1 0 2 0 1 2 2 2 2 0 2 1 1 1 2 1 2 2 0 2 0 1 0 1 1 0 1 0 1 0 0 1 2 1 8
0 0 0 1 0 0 2 2 0 1 2 0 0 0 1 0 1 2 0 1 1 2 1 2 2 2 0 1 2 0 2 2 0 0 1 2 2 2 0 2 1 2 0 1 1 8

1 0 1 2 1 1 2 1 1 2 2 2 1 2 0 0 0 0 2 0 0 1 1 1 2 0 1 0 0 0 1 0 1 1 1 1 0 2 0 2 2 2 0 2 1 9
2 1 1 0 0 1 0 2 1 2 2 0 2 0 0 0 1 1 1 2 0 1 0 0 0 2 1 0 0 1 0 2 1 1 2 2 1 1 0 2 1 1 2 2 1 9
1 2 1 0 0 1 0 1 1 0 0 2 0 1 1 2 1 1 1 1 1 1 2 2 2 1 0 1 1 0 1 0 0 0 2 2 0 2 1 2 0 0 2 2 1 9

2 1 1 2 0 1 1 0 1 1 0 0 2 0 1 0 0 2 1 1 0 2 2 1 1 0 0 2 1 1 1 0 1 1 1 0 2 2 1 0 0 2 2 1 0 10
2 2 2 0 1 2 1 1 2 0 1 1 1 0 1 2 2 0 0 2 1 1 1 2 0 0 0 2 0 0 1 2 0 1 1 2 2 0 0 0 0 1 2 1 1 10
0 1 1 2 1 1 2 0 1 0 0 1 2 0 1 2 0 0 2 2 1 1 0 0 1 2 0 1 1 2 2 1 0 0 1 1 2 0 1 2 1 1 0 2 1 10

2 2 0 2 1 0 2 2 0 0 0 2 2 2 0 0 2 2 1 1 0 0 1 1 2 1 0 0 0 2 1 1 0 0 2 2 2 1 0 0 1 1 1 0 0 11
0 1 0 1 1 0 0 1 0 1 0 0 1 1 0 1 1 0 1 1 1 1 1 0 1 0 0 1 0 0 1 1 1 1 1 0 1 1 1 1 1 0 1 0 1 11
0 0 2 2 0 2 1 1 2 1 0 2 2 1 0 2 1 0 2 2 1 0 1 0 0 1 1 0 2 1 2 0 0 1 0 2 2 1 1 2 0 2 1 1 0 11

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12
1 0 2 1 1 2 0 2 2 2 1 0 2 1 1 1 0 2 0 2 0 0 1 0 1 1 0 1 0 2 1 2 1 0 2 1 0 0 1 2 0 2 2 1 0 12
1 0 2 2 1 2 2 1 2 0 0 0 2 1 0 1 1 1 0 0 0 2 2 2 0 0 1 2 2 2 0 1 1 0 0 0 2 2 0 1 1 1 2 2 1 12

1 2 1 2 0 1 1 2 1 2 1 2 0 1 0 2 0 2 1 0 1 2 1 0 0 2 0 0 2 0 2 1 1 0 1 0 1 0 1 0 2 1 2 0 1 13
1 0 0 2 1 0 2 1 0 1 1 1 0 0 0 2 2 2 1 0 0 0 0 0 1 0 1 1 1 1 2 2 1 2 2 2 1 2 0 0 0 0 1 2 1 13
0 2 0 2 2 0 0 2 0 2 0 0 2 2 1 2 2 0 2 2 0 2 2 0 2 0 1 2 0 0 2 2 0 2 2 0 2 2 1 2 2 0 2 0 1 13

0 2 1 2 2 1 0 2 1 0 1 1 0 1 1 0 0 1 0 2 0 0 0 1 0 0 1 1 2 2 1 0 0 1 1 2 1 2 1 1 1 2 1 0 1 14
0 0 0 2 0 0 1 1 0 2 1 0 0 0 0 0 2 1 0 2 1 1 2 1 1 1 1 2 1 0 1 1 0 0 2 1 1 1 1 1 2 1 0 1 0 14
1 1 0 0 2 0 2 2 0 2 1 1 1 1 0 0 0 2 2 1 1 1 1 0 2 0 1 2 1 1 0 0 0 0 0 2 2 0 1 1 1 0 2 1 0 14

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
62
1 2 1 1 0 1 2 0 1 1 2 2 0 1 1 2 2 0 1 2 1 0 0 1 1 0 1 2 0 0 0 2 1 0 0 1 2 1 0 1 1 2 2 2 0 15
1 0 1 1 1 1 0 2 1 1 0 2 1 2 1 0 2 1 2 2 0 2 0 2 0 1 0 2 1 0 2 1 1 1 0 2 1 0 1 0 1 0 0 1 0 15
2 1 2 2 0 2 1 0 2 2 1 1 0 2 1 1 1 0 2 1 0 0 0 2 2 0 0 1 0 0 0 1 1 0 0 2 1 2 1 2 2 1 1 1 0 15

0 1 2 1 1 2 0 1 2 0 2 2 0 2 0 0 0 2 0 1 1 0 0 2 0 0 0 2 1 1 2 0 1 2 2 1 2 1 1 2 2 1 2 0 1 16
0 0 1 2 0 1 1 1 1 0 2 1 1 2 0 1 0 2 1 2 1 2 0 2 2 1 1 1 0 2 0 2 0 2 1 0 0 1 1 0 1 0 2 1 0 16
1 2 2 0 0 2 0 1 2 1 1 0 1 0 1 0 2 2 2 1 1 2 0 0 0 1 0 0 0 2 0 1 0 2 1 1 2 2 1 1 2 2 1 2 1 16

1 2 0 2 0 0 1 2 0 1 0 1 2 2 0 1 2 1 0 0 1 1 0 2 2 2 0 1 0 1 0 0 1 1 2 1 2 0 1 1 0 2 0 0 1 17
2 2 0 1 1 0 0 0 0 2 1 2 2 2 0 0 1 0 1 0 1 1 0 2 0 2 1 2 1 2 2 2 0 0 1 0 0 2 1 1 0 2 1 1 0 17
0 1 2 2 1 2 2 0 2 1 1 2 0 2 1 0 1 1 0 2 1 2 1 1 2 2 0 0 0 1 1 2 0 2 0 0 1 0 1 1 0 0 2 2 1 17

2 0 0 2 2 0 0 1 0 0 1 2 0 0 1 1 2 0 2 1 1 2 1 2 1 2 1 0 1 2 0 0 1 1 2 0 1 0 0 2 1 2 2 2 0 18
2 0 1 0 2 1 2 0 1 2 1 0 1 2 0 2 1 0 0 2 1 2 0 2 1 1 0 0 2 1 1 0 1 0 2 1 2 2 1 0 1 0 1 0 1 18
2 0 2 0 2 2 2 0 2 0 2 1 2 1 0 0 2 1 1 2 1 0 1 0 2 1 0 2 1 0 0 1 1 2 1 0 1 2 1 2 0 2 0 0 1 18

1 1 2 1 2 2 1 1 2 2 2 0 0 2 0 2 0 0 1 2 0 2 1 1 0 2 0 1 1 2 0 1 0 1 2 2 2 2 0 1 0 0 0 0 0 19
2 2 1 2 1 1 2 2 1 1 1 0 0 1 0 1 0 0 2 1 0 1 2 2 0 1 0 2 2 1 0 2 0 2 1 1 1 1 0 2 0 0 0 0 0 19
0 1 0 2 1 0 2 0 0 2 2 0 1 1 1 1 2 2 1 2 1 0 2 2 0 2 0 2 2 0 0 0 0 1 2 2 0 0 1 0 2 2 1 2 1 19

0 0 2 1 0 2 2 2 2 0 1 2 2 1 1 2 0 1 2 1 1 1 0 1 1 2 0 2 0 1 0 1 0 1 2 0 0 2 0 0 2 0 1 1 1 20
2 2 1 1 1 1 0 0 1 0 2 0 0 1 0 1 2 1 2 0 1 2 1 0 1 2 1 1 0 1 1 0 0 2 0 2 2 2 1 0 2 1 0 1 0 20
2 0 1 1 2 1 1 2 1 0 0 0 1 2 1 2 2 2 0 0 1 1 1 1 0 0 0 1 1 1 0 2 0 0 0 0 1 1 1 2 2 2 1 2 1 20

1 1 1 1 2 1 1 1 1 1 1 2 2 0 0 1 2 2 0 2 0 1 0 0 2 2 0 2 2 0 1 0 0 2 0 0 0 2 0 2 1 1 1 0 0 21
1 0 0 0 1 0 1 0 0 2 0 1 0 0 1 2 0 1 1 1 0 2 1 2 0 2 1 2 0 1 1 1 0 2 0 1 0 1 1 2 1 2 1 0 1 21
1 1 2 0 2 2 2 2 2 1 0 0 0 2 0 2 2 1 1 1 1 0 0 2 1 0 1 0 2 2 1 2 0 1 1 0 0 0 1 2 2 1 0 1 0 21

1 0 0 1 1 0 0 2 0 0 2 1 0 0 1 2 1 0 1 2 0 1 2 1 2 1 0 0 2 1 0 0 1 2 1 0 2 0 1 1 2 1 1 1 0 22
1 0 1 0 1 1 1 0 1 0 1 2 1 2 1 0 1 2 2 1 0 0 2 0 1 2 1 1 2 0 0 2 0 1 2 0 2 1 1 1 0 1 0 0 1 22
1 1 1 0 2 1 2 2 1 0 2 2 2 0 0 1 1 0 0 1 1 2 2 1 0 0 1 1 0 0 2 1 0 2 2 1 1 0 1 0 0 2 1 1 0 22

1 2 2 2 0 2 1 2 2 0 2 0 1 0 0 0 1 0 2 0 1 0 2 1 1 2 0 2 1 2 1 2 1 2 0 2 0 0 1 2 1 0 1 0 1 23
1 1 1 2 2 1 0 0 1 2 0 2 2 0 1 1 0 1 0 0 1 0 1 2 1 1 0 0 1 0 0 2 0 2 1 2 2 1 0 1 2 0 1 1 1 23
2 2 1 0 1 1 1 1 1 2 0 0 0 1 1 1 1 2 2 2 1 0 0 1 2 0 0 0 1 1 2 1 0 2 2 0 0 0 0 1 1 2 0 1 1 23

1 2 0 1 0 0 2 0 0 0 1 1 2 2 1 1 1 2 0 2 1 2 2 0 0 0 1 0 1 1 1 1 1 1 1 2 0 1 0 2 2 0 0 2 0 24
2 0 1 2 2 1 0 1 1 1 2 0 1 2 1 2 0 1 0 1 1 0 2 0 2 2 1 2 0 1 2 1 1 0 1 2 0 0 0 1 0 1 1 2 0 24
2 0 0 0 2 0 2 0 0 1 0 2 0 0 0 1 0 2 2 2 1 1 2 1 0 1 0 1 0 2 2 2 1 1 0 2 0 2 1 1 2 1 2 0 1 24

0 0 2 0 0 2 0 0 2 2 2 2 2 1 0 2 2 2 2 0 0 2 2 2 2 0 0 1 1 1 1 2 0 1 1 1 1 0 0 1 1 1 1 0 0 25
1 1 2 2 2 2 0 0 2 0 1 0 0 2 1 2 1 2 1 0 1 1 2 0 2 1 0 2 0 2 2 0 0 1 0 1 1 1 0 0 1 2 0 1 1 25
2 1 0 1 0 0 2 1 0 2 0 2 1 1 1 2 1 2 0 0 0 2 0 1 1 1 1 2 0 2 0 0 0 2 1 2 1 0 1 2 0 1 0 0 1 25

0 1 1 1 1 1 0 1 1 2 1 1 2 0 0 2 2 1 2 1 1 2 2 1 2 0 0 0 2 2 0 2 1 0 0 2 0 1 1 0 0 2 0 0 1 26
1 0 2 0 1 2 1 0 2 1 2 0 2 1 1 1 2 0 0 1 0 1 0 1 2 2 1 0 1 2 2 0 0 0 1 2 1 1 1 0 2 0 2 0 1 26
2 1 1 1 0 1 2 1 1 0 1 0 2 0 1 0 2 0 1 0 0 0 1 2 2 1 1 1 2 1 2 1 0 1 0 1 0 0 1 1 2 0 2 0 1 26

0 2 2 1 2 2 1 0 2 0 0 2 1 0 0 1 0 0 1 1 0 2 0 0 2 1 1 2 2 1 1 2 1 0 2 2 1 0 0 1 2 2 0 2 1 27
1 1 0 2 2 0 0 0 0 1 2 1 1 1 1 0 2 0 2 0 1 2 0 1 0 1 0 1 2 1 1 1 0 0 2 0 0 1 0 2 0 1 2 1 1 27
2 0 2 1 2 2 1 2 2 1 1 1 2 1 1 0 0 0 1 0 1 2 2 2 1 0 0 0 0 0 2 0 0 2 2 2 0 1 1 1 1 1 0 2 1 27

Notes: The columns are organised as follows: Conventional vehicles for options 1,2 and 3 take up the
first 9 columns. There are three attributes per option (noting that all prices are treated as one attribute).
The next 18 columns are associated with the three electric vehicles, each with 6 attributes. The last set of
18 columns relate to the alternative fuel options. The block number are at the far right hand side.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
63
Figure C3 The Contextual Statement for the Final Vehicle Choice Experiment

14. We would now like to ask you some questions about choosing to buy a vehicle for your
household.
If you currently dont own a vehicle, assume you are buying one.
If you currently own one vehicle, assume you are replacing it.
If you currently own more than one vehicle, assume you are replacing one of
them.

We will show you 3 sets of choices of vehicles which we will assume are on the market.
We would like you to consider each choice in relation to your current households situation
and needs.

Before we look at the choices, can you tell me, if you were going to buy a vehicle for your
household which type are you most likely to consider. CIRCLE ONE ONLY
SHOWCARD 5
Sedan or hatch back....................................1
Station wagon..............................................2
Four wheel drive (eg. Landcruiser) ............3
Sports car ....................................................4
Family passenger van.................................5

We are going to show you choices of small, medium and large vehicles. We are assuming that
each of these vehicles has the engine capacity and weight as listed on this card.
SHOWCARD 6


VEHICLE CHOICE 1


CHOOSE A SET OF THREE CARDS AT RANDOM FROM THE VEHICLE
CHOICE CARDS. TAKE ONE OF THOSE CARDS.

WHAT IS THE CARD NUMBER ___________

This is the first choice.

SHOW THE RESPONDENT ONLY THE CONVENTIONAL PETROLEUM
FUELLED VEHICLES AND EXPLAIN THE FEATURES OF THE CHOICES

If these were the vehicles available in the type which you prefer which size and age of
vehicle would you choose, assuming that your household was going to buy a vehicle.
SIZE AGE
Small vehicle.. .........................1 New vehicle .................................1
Medium vehicle.. .....................2 2 year old vehicle .........................2
Large vehicle.. .........................3 5 year old vehicle ........................3
10 year old vehicle ......................4

We would now like to add to your choice by showing you some electric powered vehicles
and some vehicles which use alternative fuels such as Liquid Petroleum Gas (LPG) and
Compressed Natural Gas (CNG). Although it is true that these vehicles are not currently
readily available on the Australian market, we are going to assume here that they are
available now with the characteristics as shown. We will also assume that the recharging
facilities for electric vehicles and alternative fuel supplies are also readily available.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
64
If these vehicles were now also available, which one of all these vehicles shown would
you prefer to buy for your household?

LET THE RESPONDENT SEE THE ELECTRIC AND ALTERNATIVE
FUELLED VEHICLES ON THE SAME CARD

The same conventional fuelled vehicle as above....1
An electric powered vehicle....................................2 What size and age?
An alternative fuelled vehicle...................................3 What size and age?
SIZE AGE
Small vehicle.. .........................1 New vehicle .................................1
Medium vehicle.. .....................2 2 year old vehicle .........................2
Large vehicle.. .........................3 5 year old vehicle ........................3
10 year old vehicle ......................4
FOR HOUSEHOLDS WITH NO VEHICLES OR ONE VEHICLE
GO TO CHOICE 2

If you already have more than one vehicle in the household, which one are you assuming
that you would replace with the vehicle that you have just chosen.?

Replace vehicle number ...................... .1 .............2 .............3 ..............4 ______
(TAKE VEHICLE NUMBER FROM HOUSEHOLD QUESTIONNAIRE)

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
65
Table C7 Illustrative Showcards for the Final Vehicle Choice Experiment
Card Number V01A
Conventional Vehicle OPTION 1 OPTION 2 OPTION 3
Attributes Small Car Medium Car Large Car
Price: New vehicle $25,000 $40,000 $55,000
2 yr old vehicle $18,000 $30,000 $42,000
5 yr old vehicle $12,000 $20,000 $30,000
10 yr old vehicle $ 7,000 $12,000 $20,000

Other costs: Registration fee (not including
compulsory insurance)
$250 $500 $400
Fuel cost to travel 500kms $20 $30 $40

Electric Vehicle OPTION 4 OPTION 5 OPTION 6
Attributes Small Car Medium Car Large Car
Price: New vehicle $15,000 $25,000 $30,000
2 yr old vehicle $12,000 $20,000 $24,000
5 yr old vehicle $ 9,000 $15,000 $18,000
10 yr old vehicle $ 6,000 $11,000 $13,000

Other costs: Registration fee (not including
compulsory insurance)
$50 $125 $200
Fuel cost to travel 500kms $20 $30 $60
Features:
( d t
Fully fuelled range (% of) 70% 50% 70%
conventional
vehicles)
Acceleration slightly slower considerably slower considerably
slower
Boot size (% of) 90% 90% 90%

LPG/CNG Fuelled Vehicle OPTION 7 OPTION 8 OPTION 9
Attributes Small Car Medium Car Large Car
Price: New vehicle $15,000 $30,000 $30,000
2 yr old vehicle $11,000 $22,000 $22,000
5 yr old vehicle $ 7,000 $15,000 $15,000
10 yr old vehicle $ 5,000 $ 9,000 $10,000

Other costs: Registration fee (not including
compulsory insurance)
$50 $125 $200
Fuel cost to travel 500kms $30 $40 $85
Features:
(compared to
Fully fuelled range (% of) 70% 85% 70%
conventional
vehi cl es)
Acceleration considerably
slower
considerably slower same
Boot size (% of) 90% 90% 90%

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
66
Appendi x D: The Behavi our al Model Syst em

D1. Commuter Behavioural Choice Models

We considered two critical structures where mode choice and departure time choice are
reversed in an hierarchy which is itself conditioned on spatial and temporal work
practices and workplace location. The two structures are summarised in Figure D1. The
left hand hierarchy conditions departure time choice on mode choice; the right hand
hierarchy conditions mode choice on departure time choice. Both structures were
estimated and the left hand specification adopted for the current study. The link
between mode choice and departure time choice is very weak; the inclusive value
parameter estimates were close to zero (slightly negative) and statistically non-
significant for both hierarchies. We did find however that the linkage between departure
time choice and spatial and temporal work practices was statistically significant, with an
inclusive value in the 0-1 range, a condition required for consistency of the model
system with global random utility maximisation.

The final hierarchy selected treats commuter mode choice as an independent decision
from departure time choice and the spatial/temporal choice of work practices. Departure
time choice is conditional on the choice of spatial and temporal work practices and
access of the employee to a choice between regular and other work practices. The
commuter mode choice model however is conditioned on workplace location, providing
a link to the other components of the model system via workplace location. Thus the
joint probability of commuter mode choice and departure time choice is the product of
two independent choice outcomes.

choice of
work hours
no choice
of work hours
regular flexi
compressed
work week
TC
CWW =
TC = telecommute
CWW
Commuter Mode Choice
Commuter Departure
Time Choice
choice of
work hours
no choice
of work hours
regular flexi TC CWW
Commuter Mode Choice
Commuter Departure
Time Choice
DT1 DT2 DT3 DT5 DT6
DA RS TN BS LR BK WK BW
DT1 DT2 DT3 DT4
DT4
DT5 DT6 DA RS TN BS WK BK LR BW
DA=drive alone, RS=ride share, TN=train, BS=bus, WK=walk, BK=bicycle, LR=light rail, BW=busway
DT1=<6.31, DT2=6.31-7, DT3=7.01-7.30, DT4=7.31-8, DT5=8.01-8.30, DT6=>8.30

Note: (i) the mode choice (left hand structure) and departure time choice (right hand structure) link
into flexi, cww and telecommuting alternatives. (ii) At the bottom levels, the departure time choice
links into all of the commuter modes and vice versa. (iii) Walk and bicycle were eliminated from the
models due to small sample sizes choosing these modes.
Figure D1 Alternative Hierarchies within the Commuter Choice Module

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
67
Figure D2 illustrates how each of the choice probabilities associated with each discrete
choice model is conditioned on other choices. Even where there is no inclusive value
linking a particular set of models, the set of probabilities from one model can still vary
according to the alternatives in the choice set of another model. For example, fleet size
choice probabilities vary by residential location as do workplace location choice
probabilities. Commuter mode choice probabilities vary by residential and workplace
location. Departure time choice probabilities vary by residential and workplace
location. Vehicle type choice probabilities vary by fleet size and the chronological
location in the set of vehicles in a household. Each part of the vehicle type choice model
(eg a probability of choosing a vehicle class (defined by size and vintage) given fleet
size and residential location) are multipled by the probability of residential location and
probability of fleet size prior to aggregating to get total demand for vehicle class (size
and vintage). The vehicle type choice models at each level (upper, middle and lower if 3
vehicles, upper and lower if two vehicles) were found to be independent. That is, there
were no statistically significant inclusive values and hence the joint probability of a mix
of vehicle types in a multi-vehicle household is the product of a set of marginal
probabilities. There is an inclusive value feeding up from vehicle type choice to fleet
size with the probability of choosing a particular vehicle type varying according to
whether there are 1, 2 or 3-plus vehicles in the household (since parameter estimates in
all vehicle type choice models are different). The structure of linkages are summarised
in Table D1 in a more intuitive manner.

Table D1 Linkage profile
Model Probabilities vary by Linkages
RLC (residential location
choice)
residential location uses ivWLC, ivDwTC
(iv =inclusive value)
WLC (workplace location
choice)
residential and workplace
location
uses ivCMC
CMC (commuter mode
choice)
residential location
workplace location,mode

DTC (departure time
choice)
residential and workplace
location, dwelling type

WPU (work practices
choice upper level)
work practices upper level
WPL (work practices
choice lower level)
work practices lower level uses average of ivDTC across
residential and workplace locns
FSC (fleet size choice) residential location, fleet
size
uses average of ptProb across all
workplace locns and workers, sum
of ivWLC across all workers
(ptprob =probability of choosing
public transport), and ivATC from
auto type choice constrained to
equality across each vehicle level in
multi-vehicle branches
AT (vehicle type choice) fleet size,, vehicle level,
size, vintage

DwTC (dwelling type
choice)
residential location
VKM(vehicle kilometres) residential location, fleet
size, vehicle level, size,
vintage
uses average of ptProb across
all workplace locns and workers

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
68

Residential Location
0 1 2 3+
Fleet Size
1 2 10 ......
Workplace Locn
RS CDA TN BUS LRT BPS
Commuter Mode Choice
Departure
Time Choice
DT1 DT6 DT2.... DT1 DT2... DT6
0 1 2 3+
Fleet Size
1 2 10 ......
Workplace Locn
RS CDA TN BUS LRT BPS
Commuter Mode Choice
Departure
Time Choice
DT1 DT6 DT2.. DT1 DT2.. DT6
.............
.............
.............
Workplace location choice
commuter mode choice
Departure time choice
}
per worker
1
10
Figure D2 Hierarchical Linkages within the Core of the Discrete Choice Model System


D1.1 Spatial and Temporal Choice of Working Hours

An important linkage in the model system is between working hour opportunities,
commuter mode choice and commuter departure time choice. Although the last two
choices are conditioned on workplace location choice, the temporal choice of working
hours is hypothesised to be a separate linkage and not conditioned on or conditioning
the choice of workplace location (Figure 1 in text). An inclusive value from departure
time choice is fed into the choice of work hours model and a set of five probabilities are
obtained per sampled worker. These probabilities are then used to modify the actual
amount of traffic on the road by the procedure described in the next paragraph.

Given the complexity of the overall model system we have assumed initially that the
choices of working hours are not spatially dependent. The primary influences are
assumed to be socio-economic and employment related. However in order to have an
impact on the levels of commuter traffic we have to adjust the set of probabilities for
each worker associated with living in a zone and working in a zone. We need to derive
the average number of one-way commuting trips per day over a regular work cycle,
assumed to be five days per week. The adjustments for each of the telecommuting and
compressed work week options are summarised in Table 2 of the text.

The choice between having a choice to participate in alternative spatial and temporal
working hours including regular working hours and having no choice (i.e. fixed regular
hours per week) is explained by one city-specific effect and the inclusive value from the
conditional choice of alternative temporal and spatial working hours, which is itself
influenced by a set of socioeconomic characteristics determining departure time choice
as embodied in an inclusive value index (Table D2). A pooled model for all cities was
selected after preliminary analysis for each city showed that the differences were not
significant, with the exception of Canberra which is accounted for by a Canberra-
specific dummy variable. Living in Canberra, relative to the other five capital cities
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
69
increases the probability of a commuter not having a choice of spatial and temporal
working hours. The strong positive and statistically significant inclusive value
(satisfying the 0-1 bounds for global consistency with random utility maximisation)
suggests that the expected maximum utility associated with the 4-choice opportunities is
an important influence on the work practices choice.

Conditional on having a choice of spatial and temporal working hours, the lower level
of the model displays some significant influences on choice between regular working
hours, flexi-time (flexi), compressed work week (cww) and telecommuting (tc). There
are no statistically significant differences between cities which can be accounted for by
city-specific dummy variables. There is a strong gender effect on compressed work
weeks, with males tending to have a higher probability of choosing a compressed work
week than females. Middle income earners ($36,000 - $55,000 per annum) tend, ceteris
paribus, to have a lower probability of flexible working hours than workers on very low
and very high incomes. Labourers relative to other occupations tend to have regular
working hours, but full time workers in all occupations tend to have a lower probability
of regular working hours than part time and self-employed workers. The overall
statistical goodness-of-fit of the lower branch model is excellent, with a pseudo-R
2
of
nearly 0.5. For a non-linear function this translates into a linear r-squared of nearly 0.9
(Domencich and McFadden, 1975, 124).

This hierarchical set of models provide a set of probabilities to adjust the amount of
weekly trips per commuter by car, and hence is important in determining the impact on
traffic levels, vehicle kilometres and hence greenhouse gas emissions of policies which
influence the temporal and spatial profile of working hours.

Table D2 Results for the Spatial and Temporal Choice of Work Practices
Upper level: Yes (40.6%)=have a choice of alternative work practices, no (59.4%) =do not have a
choice Lower level: Conditional on yes, alternatives = regular hours (20.7%), flexitime (51.5%),
compressed work week (21.0%) and telecommuting (6.8%). No choice based weights due to absence of
data on the popn distribution
Upper Level: Alts Parameter estimate t-statistic
Alternative-specific constant yes - .032674 - .10
Inclusive value yes .31009 1.79
Canberra (1,0) no .82651 4.16

Sample size 1113
Log-likelihood at zero -785.34
Log-likelihood at convergence -754.59
Pseudo R-squared .039

Lower Level:
Full time work (1,0) regular - .89181 - 2.92
Labour occupation (1,0) regular 1.3360 3.05
Flexitime specific constant flexi .27361 0.49
Personal income ('000's) 36-45 flexi - .011516 - 1.68
Personal income ('000's) 46-55 flexi - .012886 - 1.75
Compressed work week specific const. cww - .93202 1.61
Gender status (male =1) (1,0) cww .75884 2.08
Telecommute specific constant tc - .85354 - 1.41
Inclusive Value departure time choice) all .63521 1.76

Sample size 451
Log-likelihood at zero -634.923
Log-likelihood at convergence -320.615
Pseudo R-squared .495
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
70
D1.2 Departure Time Choice Model

The commuter departure choice model, limited to trips by car on a tolled or non-tolled
road, can be estimated as a revealed preference (RP) and a stated preference (SP)
model; however, the general non-availability of level of service data by time of day
prevents us from currently estimating a full RP model. Inferring for the non-chosen
departure times, for each sampled commuter, the travel time and fuel costs by time of
day by using the averages applying to the sampled travellers observed to travel at each
time of day is one way of securing such data, but our sample size for each urban area is
too small for this procedure. We can estimate an RP departure time choice model in
which the key influences are socio-economic, with a toll cost being the only temporally
varying attribute of the travel choice design being imported from the SP model, with
rescaling.

There are a number of alternative approaches to modelling departure time choice. The
need to consider more than one approach is due to the way in which we apply the
models using synthetic households. We outline the options and the criteria used to
select an operational approach.

Approach I

The departure choice model, based on the SP data context, is estimated as a three-
alternative choice set of unordered and unranked alternatives - earlier, current and later.
For each SP replication, a commuter who chooses a car alternative in the mode choice
SP experiment is asked to indicate whether they would depart earlier, later or at the
current time. The precise variation from the known current time is also indicated. The
average time of an earlier or later departure is between 12 and 13 minutes for all three
SP replications, with a range of 1 to 90 minutes. In some cases a respondent is
switching departure time to secure the time benefits of a tolled route while avoiding the
toll. There is a good spread of responses across the three times of day (Table D3). For
our model estimation we have assumed that the current time includes any variation
which is five minutes earlier or later.

Table D3 Temporal responses to departure time choice in the SP travel choice experiment
Choice 1 Choice 2 Choice 3 Total
Depart Earlier 161 144 134 439
Depart Later 88 131 111 330
Current Time 391 385 364 1140
Total 640 660 609 1909

The departure time choice model of three alternatives would be estimated conditional
on the four extended car alternatives in the SP mode choice model (see below) - car toll
drive alone, car no toll drive alone, car toll ride share, and car no toll ride share.
Separate models can be estimated for each of the four car segments or three dummy
variables introduced into a single model to identify the mode segment. In application,
these dummy variables are replaced by the mode choice probabilities of being in each of
the four modal states.

Once the model is estimated, we can identify the time of day traffic profile and the
extent of peak spreading associated with particular strategies. Given the known specific
time of day of the current trip, we can assign choice probabilities per respondent of
travelling at very specific times of the day. Identifying the probability of departing at
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
71
the current time (plus or minus five minutes) is straightforward; identifying the
probabilities at earlier and later times is not so straightforward.

Approach II

An alternative approach to estimation of the departure time choice model is to define a
choice set of earlier and later time bands centred around the current departure time. For
example: earlier E1 (>30 mins), earlier E2 (16-30 mins), earlier E3 (6-15 mins), current
(plus or minus 5 mins.), later L3 (6-15 mins), later L2 (16-30 mins), and later L1 (>30
mins). A 7-alternative choice set is defined and driven by socioeconomic
characteristics. The departure time is embedded in the definition of each alternative
relative to the known current departure time. Application by specific time of day is
secured by obtaining 7 time of day probabilities and mapping them to each of the actual
time periods given the current time of day.

Approach III

While approaches I and II are appealing for model estimation and application on the
sample used in model estimation, the ability to call on a current time of day is absent in
the synthetic household sample unless we randomised the time of day of travel. Another
alternative is to define a ranked set of departure times such as the 7 alternatives: <6 am,
6-7, 7-7.30, 7.31-8, 8.01-8.30, 8.31-9, >9am. This approach must define the choice set
in such a way that the current departure time and the chosen time from the SP
experiment are contained in the range. It is not important if the current departure time
and the selected time are in the same band (i.e. alternative) - this simply tells us that the
adjustment is within the predetermined range. Selecting the ranges is important to
ensure that any changes in the departure time profile are identified in the context of the
strategies to be evaluated.

We support Approach III because of the ease in application. The observed sample
distribution of current departure times is shown in Figure D3 for each city. An RP
departure time choice model based on six times of the day is estimated. It is driven
primarily by socioeconomic characteristics because of the absence of time of day level
of service attributes. In future work we need a time of day network. We only know
levels of service for the time of day an individual travelled to work. The six alternatives
are: up to 6.30 am, 6.31-7, 7.01-7.30, 7.31-8, 8.01-8.30, >8.30 am.

An SP time of day model is not estimated. Although a time of day response was
obtained from the travel choice experiment, it was essentially a mode choice experiment
with the possibility of changing the time of day of travel if an individual chose a car as
a result of a time of day toll. However within the times of day individuals currently
travel, the variations in departure time across the six times are such that most
individuals either continued to pay a toll or continued to avoid a toll. Having a toll
which is constant across departure times within each replication is not informative.

We adopted an alternative strategy to secure parameterisation of toll costs for car trips,
so that some level of service differentiation across departure times was identified in the
RP departure time choice model. We estimated a separate four-alternative SP model of
the choice between a tolled and a non-tolled route for the sample of individuals who
chose a car as a drive alone or ride share mode on the tolled or non-tolled facility. A
parameter estimate for toll cost was identified. The toll cost parameter was rescaled and
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
72
added into the RP departure time choice model. The results are summarised in Table
D4.

Figure D3 The traffic profile (departure time) over the morning period

There are many socioeconomic impacts on the timing of travel. Departure time choice is
associated with personal income, gender, location of residence relative to the central
business district (CBD), occupation and employment status. The distance effect
suggests that workers who live further away from the CBD tend to have an increased
probability of departing for work prior to 6.30 am and a decreased probability of
departing between 8 and 8.30, relative to the period 6.30-8 and after 8.30. The data
suggest that trips commencing after 8.30 are by workers who are not employed fulltime
and who have relatively low incomes, predominantly part-time workers as distinct from
the self-employed. Labourers and plant workers depart early, with the probability of
departing prior to 6.30 am increasing for these occupation groups. Professionals and
para-professionals have a higher probability of departing in the mid-period of 6.30 am -
8 am. Ceteris paribus, clerks have a higher probability of departing between 8 and 8.30
am. Male commuters relative to females have a higher probability of commencing their
work trip early, especially up to 7.30 am, with the contribution to the outcome being
higher the earlier the departure time.

Finally there are three city-specific dummy variables which represent some effects
which are unique to a city and common amongst the commuters within that city.
Canberra commuters tend, ceteris paribus, to have a higher probability of departing
between 7.30 and 8 am after allowing for the significant socioeconomic discriminators,
than the commuters of the other cities; although Brisbane and Melbourne workers tend
to leave later relative to the other cities commuters. the statistically significant time of
day specific constants suggests that there are some unobserved effects which have a
major influence on the departure time. Traffic congestion maybe one of these
influences, although we are unable to evaluate this proposition in the current study.

0
10
20
30
40
50
60
Time Period
CANBERRA
SYDNEY
MELBOURNE
BRISBANE
ADELAIDE
PERTH
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
73
In future studies it is useful to design an experiment in which all of the attributes are
varied by time of day. Once we have a level of service network by time of day, the
combined mode choice and departure time choice models can be used to evaluate the
joint model switching and departure time choice switching. Our current data does not
allow this. We are however still able to identify departure time switching based on
socioeconomic variables and imposed tolls.

Table D4 RP Departure Time Choice Results
DT1=<6.31am, DT2=6.31-7 am, DT3=7.01-7.30am, DT4=7.31-8am, DT5=8.01-8.30am,
DT6=>8.30am. Shares: DT1=18.4%, DT2=10.8%, DT3=17.1%, DT4=19.7%, DT5=14.9%, DT6=19.1%
Exogenous Variables Alts Parameter estimate t-statistic
DT1= <6.31am:
DT1 specific constant DT1 - 2.1811 -8.75
Personal income (36-45)(000's) DT1 - .00739 -1.43
Male dummy (1,0) DT1 1.0799 4.76
Dist of dwelling from CBD (kms) DT1 .0092002 2.53
Trade occupation dummy (1,0) DT1 1.4225 5.67
Labourer occupation dummy (1,0) DT1 1.5891 5.98
Plant work occupation dummy (1,0) DT1 2.0645 7.04
DT2=6.31-7 am:
DT2 specific-constant DT2 -1.8851 -8.01
Paraprof occupation dummy (1,0) DT2 .56337 1.70
Male dummy (1,0) DT2 1.0180 4.41
DT3=7.01-7.30am:
DT3 specific dummy DT3 -1.1463 -5.92
Personal income (56-65) (000's) DT3 .008391 1.18
Male dummy (1,0) DT3 .65032 3.54
DT4=7.31-8am:
DT4 specific constant DT4 -.80536 -4.62
Professional occupation (1,0) DT4 .46125 2.62
Canberra specific dummy (1,0) DT4 .32076 1.44
DT5=8.01-8.30am:
DT5 specific constant DT5 -.56638 -2.77
Dist of dwelling from CBD (kms) DT5 -.012344 -1.91
Clerk occupation dummy DT5 .30815 1.33
Brisbane specific dummy (1,0) DT5 -1.1298 -3.82
DT6=>8.30am. Shares:
Melbourne specific dummy (1,0) DT6 .57876 2.83
Fulltime employment dummy (1,0) DT6 -1.3502 -7.45
Personal income <10 (000's) DT6 .09889 2.93

Toll cost ($) ALL -.13709 *0.01 -2.54

Sample size 731 (4396 cases)
Log-likelihood at zero -1802.51
Log-likelihood at
convergence
-1615.33
Pseudo R-squared .108


D1.3 The Commuter Mode Choice Models

The choice of mode for the commuting trip is estimated using both revealed preference
(RP) and stated preference (SP) data. The RP datas strengths lie in reflecting the
current state of market behaviour, whereas the SP datas strengths are that it mirrors a
more robust and less restricted decision environment and presents a well-conditioned
design matrix. Combined stated preference and revealed preference data take advantage
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
74
of the strengths of both types of data while recognising that some of the weaknesses of
one data type are a strength of the other data type.

RP data are best described as:

1. depicting the world as it is now (current market equilibrium)
2. having built-in relationships between attributes (technological relationships are
fixed)
3. only have existing alternatives as observable
4. embodying market and personal constraints on the decision-maker
5. having high reliability and face validity
6. yielding one observation per respondent at each observation point.

SP data are best described as:

1. depicting virtual decision contexts (flexibility)
2. having controlled relationships between attributes (permitting mapping of utility
functions with technologies from existing ones)
3. including existing and/or proposed and/or generic choice alternatives
4. having difficulty (if not impossibility) to effectively represent changes in market
and personal constraints
5. being reliable to the extent that respondents understand the task, are committed
to the task, and can really respond to the task
6. yielding multiple observations per respondent at each observation point.

The distinguishing appeal of the two types of data can be illustrated diagrammatically in
the context of mode choice (Figure D4). RP data provides information on the current
market equilibrium for the behaviour of interest and is useful for short term forecasting
of departures from the current equilibrium. In contrast SP data is especially rich in
attribute tradeoff information, but is to some extent affected by the degree of
contextual realism that we can establish for the respondents. SP is useful for
forecasting changes in behaviour.

1
cost
1
cost
speed
speed
walk
bus
LRT
heavy rail
car
walk
bus
LRT
heavy rail
car
frontier of existing
alternatives
RP data SP data

Figure D4 Attribute Space of RP and SP Data

The more strategic our objectives (i.e. a longer time horizon), the greater the need and
opportunity to use SP data, along with RP data. The benefits include an ability to map
trade-offs over a wider range of attribute levels than currently exists (adding robustness
for forecasting), and an ability to introduce new choice alternatives (accommodating
technological change in an expanded attribute space). The joint use of RP and SP data
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
75
as shown in Figure D5 provides two paradigms under which the two types of data
should be used in model estimation and application. Ideally SP and RP data can be
collected on a regular basis from ongoing travel surveys, with a mix of choice
experiments to both evaluate different issues and to monitor responses to the same
issue. While such a longitudinal strategy is highly desirable, much can be learnt from a
single cross-section study with an expectation that an ongoing monitoring activity might
be promoted to enable a better understanding of the timing on response and the ultimate
nature of the set of behavioural responses which have contributed to the new
equilibrium.

respondent
RP data SP data
current
equilibrium
SP trade-offs
conversion into
RP units (Scaling)
choice prediction
model
RP reality check
Paradigm 1
respondent
RP data SP data
current
equilibrium
+
RP trade-off
SP task equilibriu
+ SP trade-offs
conversion into
common units
(Scaling)
choice prediction
model
Paradigm 2


Figure D5 Two Paradigms for Integrating RP and SP Choice Response Data

A key role for SP data is to assist through enrichment in obtaining more robust
parameter estimates for a particular RP-based choice model so that in application one
can increase the confidence in the predictions as the analyst stretches the attribute space
and the choice set under policy assessment. Given that each replication of a choice
experiment is a rich individual observation, with typically 3 to 4 replications per
respondent plus an RP observation, one can generate 3 - 4 pairs of SP-RP data or up to
8 observations per respondent. A sample of 300 respondents for a market segment is
more than sufficient to obtain consistent and efficient parameter estimates. The SP
mode choice data set for all capital cities is 3599 observations after replication, with a
minimum of 365 in Canberra and a maximum of 727 in Sydney (see Table D5).



Table D5 Profile of Modal Share (%) for the Stated Preference Experiment
CAN SYD MEL BRS ADL PER TOTAL
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
76
Drive alone toll road 16.2 17.1 13.4 11.1 12.4 13.4 13.8
Drive alone no tolled road 14.5 11.9 18.5 13.5 13.2 16.5 14.7
Ride share toll road 9.0 9.7 6.0 9.4 9.8 7.2 8.4
Ride share no tolled road 6.8 8.1 11.2 8.4 9.3 6.9 8.6
Bus 10.4 11.4 9.6 13.3 14.6 12.7 12.0
Train 13.4 11.5 13.1 11.0 11.2 11.6 11.9
Light rail 18.6 15.0 14.8 17.6 12.4 18.1 16.0
Busway 11.0 15.2 13.4 15.8 17.1 13.7 14.6
Total (number) 365 729 717 727 492 569 3599

The SP and RP mode choice models contain overlapping and separate modes (Figure
D6). The walk and bicycle modes are not in the SP choice set; light rail and busway are
not in the RP choice set. Eight modal alternatives define the full choice set for the SP
model, with each sampled commuter considering the car alternatives, 1 of the rail
alternatives and 1 of the bus alternatives. The rail and bus alternatives are
systematically varied in and out of the choice set across the sampled population. The RP
model has a universal choice set of 4 alternatives with each commuter having a choice
set of their chosen mode and 1 alternative. Walk and Bicycle were excluded for sample
size reasons only. Differential choice sets for both SP and RP model estimation are
readily accommodated in the estimation method. The aim is to obtain an SP-enriched
RP model in which the SP model provides rescaled parameters for attributes of the
observed modes which are present in the SP experiment (i.e. the car alternatives, train
and bus), and entirely rescaled utility expressions for the modes which are present in the
SP experiments but absent from the observed choice set (i.e. light rail and busways).

The two data specifications can be combined in the following way:

1. Use the SP data to establish robust estimates of the travellers tradeoffs of the
attributes in the stated choice experiment through the vector
sp
corresponding to X
sp

The SP travel choice experiment choice set is car (toll), car (no toll), train, bus, light rail
and busway.

2. Use the RP data to ground the model in reality by estimating the mode-specific
constants for the alternatives which are observed in the market. This is the six modes:
car drive alone (no toll), ride share (no toll), train, and bus. This ensures that the
predicted aggregate modal shares equal the observed RP shares. The RP model is
estimated with choice based weights. In estimating the mode-specific constants, we
make them conditional on the
rp
being constrained to equal
sp
, but allowing for an
errors-in-variables correction to X
rp
, through the estimation of a multiplicative scale
factor to rescale X
rp
into the same units as X
sp
. The value of is selected so as to
minimise the log-likelihood of the overall model.

The SP choice set has a minimum of six alternatives - car on a tolled road, car on a non-
tolled road, bus, train, light rail and busway (Table D5). In addition, given the interest
in distinguishing drive alone and ride share for the car trip, we can duplicate the car trip
on the tolled and non-tolled road to distinguish car occupancy. This will increase the
universal SP choice set to 8 alternatives. Each individual trip involves the commuter
selecting an alternative from 4 of the original 6 alternatives (bus and busway being
varied, train and light rail being varied). In the presence of 8 alternatives, there are still
only 4 in each choice set out of the universal set of 8. One important refinement of the
attributes is in order. If an individual chooses the car tolled route, we have to check to
see whether they choose it during the period in which a toll is operating or at the toll-
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
77
free period. This check is undertaken by comparing the time of departure (calculated
from the current observed trip departure time plus or minus the time given as an earlier
or later departure) with the pay toll time period. If the departure time is outside of the
tolled period, then the toll is set to zero for the car toll road.

DAcar toll RScar toll
DAcar
notoll
RScar
notoll
Bus Train Light rail Busway
StatedPreference
Commuter Mode Choice
DA= drive alone
RS= rideshare
Revealed PreferenceCommuter Mode Choice
DAcar
RScar Bus
Train


Figure D6 The Linkage between the RP-SP mode choice models
(walk and bicycle are excluded because sample sizes are too small)

For applications based on the rescaled RP model, we do not need to use the mode-
specific SP constants associated with the observed RP modes. In the RP model, when
we position the new modes - light rail and busway - in the utility space associated
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
78
with the RP model, we only have to rescale parameters which are unique to the SP data.
The parameters which have to be rescaled are all of those associated with light rail and
a busway; plus the level of service attributes associated with car drive alone, car ride
share, train and bus. The mode-specific constants for car drive alone, car ride share,
train and bus and parameters of included socioeconomic characteristics in all of the RP
alternatives do not have to be rescaled.

In summary, the estimation strategy involves the following steps:

1. estimate the SP mode choice model on a universal choice set of 8 alternatives
(6 per respondent SP replication).

2. estimate the RP mode choice model on a universal set of 4 alternatives (2 per
respondent). The car drive alone and car ride share alternatives are assumed
to use a non-tolled road. Choice-based weights for each city are implemented
(Table D8 compared to Table D7).

3. enrich the RP model parameter set by adding in any attributes and
alternatives which do not exist in the RP model, and rescale the attributes
which are common to alternatives in the RP and SP models and which are
associated with the new alternatives which exist in the SP choice set but not
the RP choice set.

4. Calculate an inclusive value for input into other choice models.

The empirical models from the estimation using the SP travel choice set data are
summarised in Table D9. The RP models with optimal scaling are summarised in Table
D10. The full RP mode choice model for application has 6 alternatives: car drive alone,
car ride share, train, bus, light rail and busway. The light rail and busway utility
expressions are derived from Table D11 with each parameter estimate scaled by 0.4.
We evaluated scale parameter values over a large range from close to zero to 1.0. The
optimal scale parameter is 0.4.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
79
Table D6 The Set of Attributes and Attribute Levels in the Commuter Mode Choice
Experiment
(all cost items are in $s, all time items are in minutes)Note: in model
estimation, fuel cost is converted to cents.

SHORT (< 30 mins.)

Car no toll

Car toll road

PUBLIC TRANSPORT

Bus

Train

Busway

Light Rail
Travel time to work 15,20,25 10,12,15 Total time in the vehicle
(one-way)
10,15,20 10,15,20 10,15,20 10,15,20
Pay toll if you leave at
this time (otherwise
free)
None 6-10, 6:30-
8:30, 6:30-9
Frequency of service Every 5,15,25 Every 5,15,25 Every 5,15,25 Every 5,15,25
Toll (one-way) $ None 1,1.5,2 Time from home to your
closest stop
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Fuel cost (per day) $ 3,4,5 1,2,3 Time to your destination
from the closest stop
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Parking cost (per day) Free,$10,$20 Free,$10,$20 Return fare $ 1,3,5 1,3,5 1,3,5 1,3,5
Time variability 0, 4,6 0,1,2


MEDIUM (30-45 mins.)

Travel time to work 30,37,45 20,25,30 Total time in the vehicle
(one-way)
20,25,30 20,25,30 20,25,30 20,25,30
Pay toll if you leave at
this time (otherwise
free)
None 6-10, 6:30-
8:30, 6:30-9
Frequency of service Every 5,15,25 Every 5,15,25 Every 5,15,25 Every 5,15,25
Toll (one-way) 4 None 2,3,4 Time from home to your
closest stop
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Fuel cost (per day) $ 6,8,10 2,4,6 Time to your destination
from the closest stop
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Parking cost (per day) Free,$10,$20 Free,$10,$20 Return fare 4 2,4,6 2,4,6 2,4,6 2,4,6
Time variability 0, 7, 11 0, 2, 4


LONG (>45 mins.)

Travel time to work 45,55,70 30,37,45 Total time in the vehicle
(one-way)
30,35,40 30,35,40 30,35,40 30,35,40
Pay toll if you leave at
this time (otherwise
free)
None 6-10, 6:30-
8:30, 6:30-9
Frequency of service Every 5,15,25 Every 5,15,25 Every 5,15,25 Every 5,15,25
Toll (one-way) $ None 3,4.5,6 Time from home to your
closest stop
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Walk 5,15,25
Car/Bus 4,6,8
Fuel cost (per day) 4 9,12,15 3,6,9 Time to your destination
from the closest stop
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Walk 5,15,25
Bus 4,6,8
Parking cost (per day) Free,$10,$20 Free,$10,$20 Return fare $ 3,5,7 3,5,7 3,5,7 3,5,7
Time variability 0, 11, 17 0, 7, 11


Table D7 Profile of Revealed Preference Modal Share (%)
(a) Chosen Main Mode
CAN SYD MEL BRS ADL PER TOTAL
Drive alone 52.0 51.0 62.4 50.0 54.3 61.4 55.3
Ride share 22.0 16.0 14.8 26.3 21.3 17.4 19.4
Bus 19.5 7.4 1.3 7.2 11.6 9.2 8.3
Train 0.0 19.8 13.5 11.0 3.0 7.1 10.4
Walk 1.6 3.3 3.0 1.7 1.2 0.5 2.0
Other 4.9 2.5 5.1 3.8 8.5 4.3 4.6
Total N umber 123 243 237 236 164 184 1187

(b) Alternative Main Mode
CAN SYD MEL BRS ADL PER TOTAL
Drive alone 10.6 15.6 15.6 13.1 9.8 7.6 12.6
Ride share 22.8 21.8 21.5 24.6 25.0 24.5 23.3
Bus 41.5 23.0 15.6 24.6 28.0 28.8 25.4
Train 0.0 18.1 22.8 18.6 9.8 17.9 16.1
Walk 4.9 6.6 6.3 5.1 9.1 8.2 6.7
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
80
Other 20.3 14.8 18.1 14.0 18.3 13.0 16.1
Total Number 123 243 237 236 164 184 1187
note: (other is taxi, ferry, (SYD, BRS), motorbike and bicycle)
Table D8 Commuter Mode Share Population Weights
CAN SYD MEL BRS ADL PER
Drive alone 57.58 48.99 60.12 56.25 60.67 65.02
Ride share 24.18 18.12 17.8 21.0 20.06 18.62
Bus 9.82 9.55 3.29 6.21 8.51 7.61
Train/Tram 0.0 14.74 12.12 7.78 2.53 1.89
Walk 4.14 4.78 3.57 3.63 3.31 2.53
Other 4.28 3.82 3.10 5.13 4.93 4.33
Total Number 131955 1557288 1348859 553697 418507 455024
Source: CDATA91 Census Table: J ourney to Work by Sex
The sample modal shares are relatively similar to the population modal shares. In model estimation, we
use the population shares based on the first 4 alternatives only. These 4 alternatives represent respectively
91.58%, 91.4%, 93.33%, 91.28%, 91.7% and 93.14% of all modal trips to work in the 6 cities.

Table D9 Commuter Mode Choice SP Results for Each City and All Cities Combined
Car =drive alone (toll), drive alone (no toll), ride share (toll), ride share (no toll).
Public transport =train, bus, light rail, busway
Variable NAME ALL CAN SYD MEL BRS ADL PER
Drive Alone toll
constant
DATASC -.33988
(-2.97)
.77649
(2.05)
0.21911
(0.97)
-.49338
(-2.28)
-.035999
(-.16)
-.21555
(-.76)
-.34509
(-1.24)
Time Car TIME -.048638
(-9.68)
-.033535
(-1.79)
-.04440
(-4.35)
-.048412
(-4.44)
-.04708
(-4.21)
-.055867
(-3.64)
-.055183
(-3.87)
Time variation Car TIMEVAR -.023094
(-1.63)

Fuel cost Car FUEL -.23049
(-5.64)
-.33035
(-1.94)
-.17118
(-2.15)
-.30278
(-3.44)
-.13289
(-1.48)
-.16208
(-1.24)
-.40836
(-3.33)
Toll Car drive alone TOLLDA -.38877
(-2.19)
-.090551
(-1.38)

Toll Car rideshare TOLLRS -.030141
(-.39)

Toll Car TOLL -.10698
(-3.32)
-.18306
(-2.42)
-.23368
(-3.10)
-.15437
(-1.57)
.0305
(.34)
Parking cost Car PARKING -.092237
(-22.97)
-.12178
(-8.90)
-.081747
(-9.18)
-.08064
(-9.31)
-.086515
(-9.48)
-.10726
(-9.36)
-.11419
(-10.34)
No cars/No.workers
(car)
CARAV 1.4006
(8.63)
.62560
(4.40)
.92187
(5.45)
.52024
(3.49)
.46608
(3.24)
Personal Income ('000)
(car)
PINCOME .004502
(2.06)
.018420
(2.48)
.018935
(3.02)
.0087051
(1.45)
Self employed (1,0)
(car)
SELF .65920
(4.14)

Professional (1,0)
(car)
PROF .23483
(2.65)

Canberra (1,0) Car CAN -.22517
(-1.05)

Melbourne (1,0) Car MEL -.60135
(-3.50)

Brisbane (1,0) Car BRS -.29616
(-1.65)

Adelaide (1,0) Car ADL -.17123
(-.86)

Perth (1,0) Car PER -.54082
(-2.90)

Drive alone no toll
constant
DANTAS
C
.21561
(1.58)
.75142
(3.09)
.38837
(2.30)
.50832
(3.59)
.47409
(2.91)
.34575
(1.79)
.87973
(4.62)

Main mode time
Public Transport
TIMEPT -.039253
(-2.06)
-.025029
(-2.23)
-.024079
(-2.03)
-.026215
(-2.32)
-.033821
(-2.17)

Main mode time
Bus/Bwy
TIMEB -.041137
(-6.87)
-.040763
(-2.63)
Access+egress
time Bus/Bwy
ACCEGG
T
-.016064
(-4.35)
-.018393
(-1.37)
-.022043
(-2.74)
-.023851
(-2.76)
-.009792
(-1.21)
-.023320
(-2.21)
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
81
Access+egress
public transport
ACCEGG
PT
-.027043
(-3.56)

Fare public transport FAREPT -.44632
(-13.06)
-.27195
(-2.49)
-.33682
(-4.45)
-.56269
(-6.85)
-.49263
(-6.53)
-.65954
(-6.78)
-.38717
(-4.51)

Table D9 continued
Variable NAME ALL CAN SYD MEL BRS ADL PER
Frequency public
transport
FREQ -.01994
(-5.88)
-.029954
(-2.66)
-.011306
(-1.48)
-.021719
(-2.77)
-.025211
(-3.35)
-.021382
(-2.26)
-.019692
(-2.30)
Canberra (1,0) PT CAN .22315
(1.08)

Melbourne (1,0) PT MEL -.36763
(-2.22)

Brisbane (1,0) PT BRS .13890
(.81)

Adelaide (1,0) PT ADL .16981
(.88)

Perth (1,0) PT PER -.03165
(-.18)

Person age Bus/Bwy PERAGE -.005283
(-1.86)

Main time train/Lrail TMT -.018929
(-3.18)
-.047157
(-3.13)
Access +egress time
train/Lrail
ACCEGG
T
-.03160
(-8.88_
-.024611
(-2.35)
-.04174
(-5.03)
-.035862
(-4.44)
-.027988
(-3.57)
-.023820
(-2.70)
Train specific
constant
TASC .25193
(1.18)
1.7162
(2.77)
2.1639
(5.40)
1.2848
(3.15)
2.5307
(5.78)
1.2766
(2.73)
1.1124
(2.12)
Light rail specific
constant
LRASC .33381
(1.59)
1.9097
(3.13)
2.2150
(5.63)
1.2291
(3.11)
2.6894
(6.22)
1.2658
(2.70)
1.3490
(2.58)
Bus specific constant BASC .44102
(1.86)
1.2693
(2.02)
1.5844
(4.03)
.65394
(1.62)
2.1155
(4.85)
1.5853
(3.48)
.95471
(1.79)
Busway specific
constant
BWASC .3923
(1.67)
1.0819
(1.69)
1.6823
(4.38)
.65149
(1.68)
2.1329
(4.95)
1.4032
(3.08)
.73568
(1.39)
Rideshare toll road
specific constant
RSTASC -.83144
(-6.94)
-.31953
(-1.00)
-.46026
(1.80)
-1.2965
(-5.31)
-.19852
(-.86)
-.45522
(-1.57)
-.96225
(-3.24)

Sample size 3599 365 729 717 727 492 569
Log-likelihood at zero -7483.91 -758.996 -1515.91 -1490.96 -1511.75 -1023.08 -1183.20
Log-likelihood at
convergence
-5446.99 -536.934 -1091.89 -1077.35 -1078.69 -720.206 -816.22
Pseudo R-squared 0.272 0.293 0.279 0.277 0.286 0.296 0.310
Note: the unit of time is minutes, the unit of cost is dollars. Fuel cost and fare have both been divided by
2.0 to give the one-way trip cost.

The models capture the most important attributes of travel time and cost together with a
number of socioeconomic characteristics. The rich set of cost dimensions - fuel cost,
parking cost, toll cost, and public transport fares enables us to evaluate the impact of
tolls, congestion charges, parking pricing, public transport fares, carbon tax and excise
duties on commuter mode choice and the flow-through impact on automobile use and
greenhouse gas emissions. The distinction between light rail and conventional (heavy-
rail) train and between conventional bus and busway is interesting. The mode-specific
constants in Tables D9 and D10 provide important information on the average impact of
the set of unobserved influences on commuter mode choice. Of particular interest is the
comparison between the constants for each form of public transport. There is a
similarity within each city between the mode-specific constants for bus and busway and
likewise for train and light rail. The differences between cities however are quite
noticeable suggesting that transferability of models between cities without rescaling of
the constants to reflect the local context is not recommended. In the revealed preference
model (Table D10) the bus-specific constant is the base (set to zero). There is clear
evidence in the RP model that mode-specific constants vary between modes and cities.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
82
Similarities exist however between alternative types of rail systems and between
alternative types of bus systems, which might be expected and is an encouraging result.


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
83
Table D10 Revealed Preference Model Results with Scaled SP Parameter Estimates and
Choice Based Weights
(note: each parameter starred (*) has been scaled by 0.4)
Variable NAME CAN SYD MEL BRS ADL PER
Drive alone specific constant DAASC 2.7116
(-.93)
-1.7200
(-1.19)
.75315
(1.39)
-.89621
(2.16)
.22581
(1.75)
4.3525
(3.38)
Fuel cost (cents) (DA,RS) FUELS -.003303
(fixed)
-.001712
(fixed)
-.003027
(fixed)
-.001329
(fixed)
-.001621
(fixed)
-.004083
(fixed)
In-vehicle time (mins) (DA,RS) AUTOTIMS -.033535
(fixed)
-.04440
(fixed)
-.048412
(fixed)
-.04708
(fixed)
-.055867
(fixed)
-.055183
(fixed)
Parking cost ($) (DA,RS) VEHPRKCS -.12178
(fixed)
-.081747
(fixed)
-.08064
(fixed)
-.086515
(fixed)
-.10726
(fixed)
-.11419
(fixed)
Toll cost ($) (DA, RS) VEHTOLCS -.38877
(fixed)
-.18306
(fixed)
-.23368
(fixed)
-.03050
(fixed)
Toll cost ($) (DA) VEHTOLCS -.090551
(fixed)
-.15437
(fixed)

Toll cost ($) (RS) VEHTOLCS -.030141
(fixed)
-.14337
(fixed)

Personal income ($000's) (DA) PINCOME 0.0421
(1.67)
.02824
(1.74)
.02088
(1.89)
.05555
(1.75)
.04418
(1.64)
.00927
(1.67)
Personal income ($000's) (RS) PINCOME .02824
(1.74)
.03504
(1.77)
.00214
(1.45)
.00927
(1.67)
Ride share specific constant RSASC 0.70959
(1.60)
-2.6614
(-1.87)
-.86344
(-1.79)
-2.6329
(-1.45)
.84522
(1.29)
1.9855
(0.57)
Train specific constant TNASC -.42549
(-1.11)
-1.6897
(-2.61)
.21124
(0.48)
2.4230
(2.43)
1.5911
(2.04)
Main mode time (mins) (PT) MPTRTIMS -.039253
(fixed)
-.025029
(fixed)
-.024079
(fixed)
-.026215
(fixed)
-.033821
(fixed)

Main mode time (mins) (Train) MPTRTIMS -.047157
(fixed)
Main mode time (mins) (Bus) MPTRTIMS -.040763
(fixed)
Fare ($) (PT) PTCOSS -.27195
(fixed)
-.33682
(fixed)
-.49263
(fixed)
-.65954
(fixed)
-.38717
(fixed)
Fare ($) (Train) PTCOSS -.33682
(fixed)

Fare ($) (Bus) PTCOSS -.56269
(fixed)

Access-egress time (mins) (PT) ACCEGCTS -.018393
(fixed)
-.027043
(fixed)
-.023820
(fixed)
Access-egress time (mins) (Train) ACCEGCTS -.04174
(fixed)
-.035862
(fixed)
-.027988
(fixed)

Access-egress time (mins) Bus) ACCEGCTS -.02204
(fixed)
-.023851
(fixed)
-.009723
(fixed)


Sample size 170 274 308 314 216 260
Log-likelihood at zero -79.10 -188.54 -176.06 -185.76 -120.61 -158.04
Log-likelihood at
convergence
-26.40 -66.75 -39.89 -45.64 -24.58 -30.53
Pseudo R-squared .67 .65 .77 .75 .79 ..81

The key modal attributes of the observed modes are rescaled from the SP model and
imported into the revealed preference model. The final RP models produce extremely
good statistical fits. The implied commuter behavioural values of travel time savings for
automobile use, based on a trade-off between fuel cost and in-vehicle travel time for
each city are $6.09/person hour (Canberra), $15.56/person hour (Sydney), $9.59/person
hour (Melbourne), $21.24/person hour (Brisbane), $20.68/person hour (Adelaide) and
$8.10 (Perth). The weighted average time value across all cities is $12.66/person hour,
approximately 68% of the gross average wage rate of car users. The public transport
values, based on a trade-off between travel time and fare, are $8.66/person hour
(Canberra), $4.45/person hour (Sydney), $4.28/person hour for train (Melbourne),
$2.56/person hour for bus (Melbourne), $3.19/person hour (Brisbane), $3.08/person
hour (Adelaide), $7.31/person hour for train (Perth) and $6.31/person hour for bus
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
84
(Perth). The weighted average public transport behavioural value of travel time savings
for all cities is $4.50/person hour, equivalent to 30% of the gross average wage rate of
public transport users. Behavioural values of travel time savings can be derived by
trading-off time with the toll. There are clearly a number of trade-off dimensions.

We have deliberately distinguished the main sources of costs associated with
automobile use so that we can evaluate the differential sensitivity of commuters to
changes in fuel prices, tolls and parking costs. Given the lumpiness of tolls, it is not
valid to simply sum all sources of financial commitment since behavioural response is
different - a toll is often a fee at the time of use; whereas a fuel expense is a fee already
committed prior to travel. The ability to identify the sensitivity of automobile users
driving alone and ride-sharing is accommodated in these two modal alternatives. We
might anticipate that ride-sharers have a different sensitivity to automobile costs than do
drive alone commuters. Tables D9 and D10 show that cost sensitivities between single
and multiple-occupant automobile commuting relate to the toll cost, in Sydney and
Adelaide only. Sydney has the most extensive tollroad system of all cities (M4, M5 and
the M2 to commence construction next year), and thus one might expect this experience
to provide greater discrimination. This is borne out by the parameter estimates of -
.090551 for drive alone and -.030141 for ride share. The Adelaide estimates are much
closer (-.15437 and -.14337 respectively). Adelaide does not have any toll roads -
indeed we could have constrained the Adelaide parameters to be equal without any loss
of accuracy.

There are many interesting findings in the joint SP-RP commuter mode choice models.
Importantly, we should note that the set of city-specific models represent the first ever
development of a set of consistent and directly comparable models for the six capital
cities. Although the parameter estimates on each explanatory variable are often
significantly different between cities (with some notable similarities), the encouraging
observation is that there is a pattern of differences which are bounded to a range which
is intuitively appealing. For the first time in Australia we have a comprehensive choice
set of 6 commuting alternatives. The missing modes - walk, bicycle, taxi and motor bike
were not represented well enough in the data base to complete the full 10 alternatives.


D1.4 The Workplace Location Choice Model

Each metropolitan area has been subdivided into geographic areas. Sydney is
represented by 14 statistical sub-divisions, Melbourne by 18, Brisbane by 11, Adelaide
by 10, Perth by 9 and Canberra by 6 areas. The workplace location choice model has a
spatial dimension. The objective is to identify, for each sampled worker in the travel
survey, the influences on the choice of workplace location, given a knowledge of the
current workplace location and the presence of a finite but universal choice set of
workplace locations.

The workplace location choice model applies to the subset of workers who have a fixed
workplace location. Workers who have varying workplace locations are excluded from
the choice model; however to recognise the importance of the latter workforce we
estimate a binary choice model to identify the probability that a worker has a fixed
versus non-fixed workplace location. The presence of varying locations for work
activity is allowed for in the application of the model system since all work activity
(except working at home) directly impacts on the traffic. Since we are not able to
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
85
identify the locations of the non-fixed workplaces, for each worker, we allocate the
probabilities spatially according to the type of work undertaken and the incidence of
jobs of that nature in each geographical zone. For example, if 5% of all trades persons
jobs are in the CBD zone then 5% of all non-fixed work location trips are assumed to go
to the CBD.

The potential influences on choice of fixed workplace location are:

(i) supply of jobs by occupation category (managers, professional, para-professionals,
trades persons, sales and personal service workers, plant and machine operators,
labourers and related workers and other). This is a supply measure of the opportunities
for work.

(ii) worker-specific characteristics - income, employment status, occupation.

(iii) modal accessibility (inclusive value from commuter mode choice), distance of the
workplace location zone to the CBD, and distance of the workplace location zone to the
residential location zone of the sampled worker in the model estimation.

The modal accessibility index is derived from the parameter estimates of the RP-SP
scaled commuter mode choice model (excluding modes currently not available), and an
O-D matrix of travel times and costs for each mode for each city. The application of
these times and costs will produce a commuter modal accessibility index associated
with a given origin and a given workplace location. A simple summation across all
workplaces will produce a global modal accessibility index for a given residential (i.e.
origin) location.

The workplace location choice model is a ranked unordered logit model, with J -l
workplace specific constants to pick up the mean impact of the unobserved influences
on workplace location choice. Thus for Sydney with 14 alternatives, we would have 13
constants and one alternative is the location where the worker is currently employed.

Separate workplace location choice models are estimated for each city. The sample
sizes for each city however were not large enough to accommodate a model estimated
on the full zonal system. Aggregating the zones for estimation, however, still preserves
the individual characteristics of the fuller zonal system. The grouping summarised in
Table D11 below reduces the number of alternatives in the choice set and means that
the zone-specific constants will be reduced to one less than the number of grouped
zones. However the other parameter estimates associated with the attributes of each
zone (such as number of jobs) and the characteristics of each worker (e.g. occupation,
income) are consistent estimates associated with the original zonal system. This is
ensured by selecting a zone at random from within the grouped set and assigning it to
the grouped alternative, except that the chosen workplace location is the mandatory
selection for the group zone in which it is located. The estimated model is applied in the
context of the fuller zonal system in the application simulator. If it were necessary to
distinguish the zone-specific constants between the zones contained in each grouped
zone, then a simple model using only the fuller-zonal system can be estimated. This has
been undertaken to enable us to use this within-group zone variation in the zone-
specific constants in the simulator.


Table D11 Profile of Job Incidence for the Full and Grouped Zonal System
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
86
(percentages associated with each zone within the grouped zone are shown in parenthesis,
and total jobs are given in the first column for each city).
Grouped Zonal System Full Zonal System Total J ob Share (%)
CANBERRA (114,781):
1 101 52.0
2 102 14.7
3 103-106 33.4 (12.3, 2.4, 7.8, 10.9)
SYDNEY (1,381,071):
1 201,202 30.4 (26.1,4.3)
2 203 6.6
3 204,205,210 17.5 (6.2, 5.5, 5.8)
4 206,209 7.4 (2.8, 4.6)
5 207,208 14.2 (4.1, 10.0)
6 211 11.6
7 212-214 12.4 (4.0, 4.3, 4.0)
MELBOURNE (1,175,729):
1 301 30.6
2 302, 305, 309, 313, 315 19.0 (3.6, 2.7, 3.9, 3.9, 4.9)
3 303, 308, 311, 314, 316 27.3 (5.2, 3.4, 8.0, 7.7, 3.0)
4 304, 307, 312 7.6 (1.8, 3.6, 2.3)
5 306, 310 11.4 (5.9, 6.4)
6 317,318 4.1 (2.2, 2.0)
BRISBANE (497,802):
1 401,403,404 17.3 (7.7, 5.4, 4.2)
2 405,406 13.5 (4.2, 9.4)
3 407,410 41.8 (30.7, 11.0)
4 408,409 16.9 (5.9, 10.9)
5 402,411 10.5 (3.0, 7.6)
ADELAIDE (382,780):
1 501, 502, 503 15.9 (6.5, 5.2, 4.3)
2 504, 508 21.3 (13.6, 7.7)
3 505, 506, 507 52.3 (6.5, 31.9, 13.9)
4 509, 510 10.5 (5.4, 5.1)
PERTH (412,539):
1 601, 602 15.5 (5.6, 9.9)
2 603, 608 21.0 (10.9, 10.1)
3 604, 605 12.9 (7.2, 5.8)
4 606, 607 20.8 (7.8, 13.0)
5 609 29.8

The results for each city are interesting (Table D12). The Canberra model has very
significant income effects using the highest income as the base within the two largest
job zones, Central Canberra (101) and Belconnen (102). The mean estimate decreases
in magnitude as income increases suggesting a strong nonlinear diminishing impact of
location choice as income increases. That is, the probability of working in central
Canberra or Belconnen relative to elsewhere in the Australian Capital Territory (ACT)
decreases as personal income increases. Lower income workers appear to have less
mobility and are more confined to job opportunities in the central employment
locations. However when considering specific occupations, we find that, ceteris paribus,
if a person is a clerk, they have a lower probability of working in the central area and
Belconnen compared to elsewhere in the ACT, but that part-time workers have a higher
probability of working in the central area of Canberra.

The impact of job opportunities as measured by the total number of jobs is positive and
statistically significant, as might be expected. This variable, LNJ T, the natural
logarithm of no. of jobs is a size variable representing the physical size of the job
market within each zone. All other things being equal, the opportunities for jobs in each
industry class increases as the number of jobs in total increases. The high correlation
between the number of jobs in each occupation group (in excess of 0.8 and often over
0.95) is evidence of this phenomenon. The job categories considered are the ABS
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
87
categories manager, professional, para-professional, trades, clerks, sales, drivers,
labourers and other.

Table D12 Workplace Location Choice Models
(using choice based weights for the proportion of total jobs in each zone)
Variable Acronym Units Alts Estimated
parameters
t-values
CANBERRA:
pincome <10 PINC1 ($000's) 101-102 1.6558 7.02
pincome 10-25 PINC2 ($000's) 101-102 .087888 1.86
pincome 26-35 PINC3 ($000's) 101-102 .090117 2.66
pincome 36-45 PINC4 ($000's) 101-102 .051153 1.94
pincome 46-55 PINC5 ($000's) 101-102 .27585 14.07
pincome 56-65 PINC6 ($000's) 101-102 .21540 9.98
Mgt/admin/prof/paraprof MGTPRF 1,0 101 1.2110 2.21
Part time worker PARTIME 1,0 101-102 13.057 21.01
Inclusive value mode choice IVMC 101-102 .10535 1.35
Ln of total jobs in zone LNJ T no.s 101-106 .46683 1.09
Zone-specific constant 101 5.0973 -2.50
Zone-specific constant 102 -5.000 -2.68

Sample size 83
Log-likelihood
at zero
--91.184
Log-likelihood at
convergence
-67.079
Pseudo R-squared .264
SYDNEY:
personal income PINCOME $'000's 201, 202 .027414 2.04
personal income PINCOME $'000's 207,208 .033604 2.19
trade and plant worker TECH 1,0 203 -13.347 -26.35
sales worker SALES 1,0 203 -13.315 -29.34
sales worker SALES 1,0 206,209 -12.632 -24.30
self-employed SELF 1,0 203,207,208,211 13.794 17.78
self-employed SELF 1,0 204,205,210 -1.5796 -1.76
labourer LAB 1,0 204,205,210,206,
209,207,208
1.7418 2.20
para-professional PARAPROF 1,0 204,205,210 -13.916 -27.40
para-professional PARAPROF 1,0 206,209 -12.469 -17.403
part time worker PARTIME 1,0 206,209 -13.332 -26.85
part time worker PARTIME 1,0 207,208 -1.5116 -1.69
professional PROF 1,0 206,209 1.1272 1.58
professional PROF 1,0 207,208 1.0177 1.75
zone 204 dummy Z204 1,0 204 1.0505 1.63
zzone 205 dummy Z205 1,0 205 .94681 1.37
zone 206 dummy Z206 1,0 206 1.1623 1.55
zone 207 dummy Z207 1,0 207 .42834 .68
zone 212 dummy Z212 1,0 212 -.95824 -1.46
zone 213 dummy Z213 1,0 213 -1.3902 -1.88
Inclusive value mode choice IVMC 203,204,205,210 .11026 1.49
Ln of total jobs in zone LNJ T no.s all zones 1.1049 3.67
Zone-specific constant 201,202 -2.1660 -2.94
Zone-specific constant 203 -2.7335 -2.78
Zone-specific constant 204,205,210 -2.7257 -2.45
Zone-specific constant 206,209 -1.5157 -2.09
Zone-specific constant 207,208 -2.6789 -2.65
Zone-specific constant 211 -3.0871 -3.77

Sample size 135
Log-likelihood at zero -262.69
Log-likelihood at convergence -207.06
Pseudo R-squared .212
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
88
Table D12 continued
MELBOURNE:
log of personal income LPINCOME $'000's 301,302,305,309,
313,315
.87712 2.21
clerical worker CLERK 1,0 304,307,312 -11.833 -26.46
trade and plant worker TECH 1,0 301 -1.6777 -2.28
labourer LAB 1,0 301,302,305,309,
313,315,303,308,
311,314,316,306,
310
11.332 20.28
labourer LAB 1,0 304,307,312 -.98407 -1.78
para-professional PARAPROF 1,0 301,303,308,311,
314,316,306,310
12.024 28.27
para-professional PARAPROF 1,0 302,305,309,313,
315
-1.0291 -2.43
Inclusive value mode choice IVMC 302,305,309,313,
315,303,308,311,
314,316,306,310
.20596 1.67
Ln of total jobs in zone LNJ T no.s all zones 1.8212 4.45
Zone-specific constant 301 -5.8733 -3.26
Zone-specific constant 302,305,309,313,
315
-2.4329 -1.62
Zone-specific constant 303,308,311,314,
316
.27814 .42
Zone-specific constant 304,307,312 .38966 .76
Zone-specific constant 306,310 -.70126 -1.03

Sample size 149
Log-likelihood at zero -261.60
Log-likelihood at convergence -209.01
Pseudo R-squared .201
BRISBANE:
log of personal income LPINCOME $'000's 407,410,408,409 1.0116 3.12
sales worker SALES 1,0 408,409 -11.449 -24.55
trade and plant worker TECH 1,0 401,403,404,405,
406
.79304 1.57
labourer LAB 1,0 405,406 1.1913 1.76
professional PROF 1,0 405,406 1.0840 1.64
para-professional PARAPROF 1,0 407,410 -1.1572 -1.95
para-professional PARAPROF 1,0 402,411 -1.3372 1.36
age of worker PERAGE yrs 401,403,404 -.045543 -1.88
age of worker PERAGE yrs 408,409 -.064133 -3.38
Inclusive value mode choice IVMC 401-406,408,409,
411
.19605 2.83
Ln of total jobs in zone LNJ T no.s all 1.2876 3.37
Zone-specific constant 401,403,404 1.7548 1.87
Zone-specific constant 405,406 -.84027 -1.84
Zone-specific constant 407,410 -3.8972 -3.45
Zone-specific constant 408,409 -1.2102 -.94

Sample size 149
Log-likelihood at zero -239.81
Log-likelihood at convergence -184.37
Pseudo R-squared .231
ADELAIDE:
log of personal income LPINCOME $'000's 501-508 1.1781 2.36
clerical worker CLERK 1,0 501-504,508 -13.283 -27.13
labourer LAB 1,0 509,510 1.3741 1.53
para-professional PARAPROF 1,0 501-503 -12.153 -25.77
age of worker PERAGE yrs 501-504,508 -.037305 -1.84
part time worker PARTIME 1,0 501-503,505-507 1.2730 1.79
Inclusive value mode choice IVMC 501-503,505-
507,
509,510
.37641 2.95
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
89
Table D12 continued
Ln of total jobs in zone LNJ T no.s all 2.1277 4.37
Zone-specific constant 501,502,503 -2.0069 -1.09
Zone-specific constant 504,508 -1.9720 -1.09
Zone-specific constant 505, 506,507 -5.3068 -2.87

Sample size 100
Log-likelihood at zero -138.63
Log-likelihood at
convergence
-86.67
Pseudo R-squared .375
PERTH:
log of personal income LPINCOME $'000's 609 .93998 1.94
clerical worker CLERK 1,0 601,602,604,605 12.764 22.62
trade and plant worker TECH 1,0 601,602,604,605 12.990 18.91
trade and plant worker TECH 1,0 603,608 1.9797 2.92
labourer LAB 1,0 604,605 13.129 20.19
labourer LAB 1,0 606,607 -1.6498 -1.50
para-professional PARAPROF 1,0 601,602 12.835 12.89
professional PROF 1,0 601,602 13.314 19.41
professional PROF 1,0 604,605 11.844 12.97
professional PROF 1,0 609 1.0725 1.81
sales worker SALES 1,0 601,602 13.154 21.95
sales worker SALES 1,0 604,605 11.485 12.44
Inclusive value mode choice IVMC 601-605,608 .35520 2.88
Inclusive value mode choice IVMC 606,607 .25002 1.52
Ln of total jobs in zone LNJ T no.s all .45162 .71
Zone-specific constant 601,602 -13.727 -5.38
Zone-specific constant 603,608 -1.3924 -.57
Zone-specific constant 604,605 -13.444 -5.33
Zone-specific constant 606,607 .53081 .17

Sample size 126
Log-likelihood at zero -162.26
Log-likelihood at
convergence
-81.05
Pseudo R-squared .200

Comparing the full set of cities, the inclusive value from commuter mode choice varies
in its statistical significance between cities, but it lies between 0 and 1, a condition
required for consistency with global random utility maximisation. All models have
good overall fits. The linkage between relative modal accessibility and workplace
location is strongest in Brisbane, Adelaide and Perth, although the linkages in the other
three cities are important. The choice of workplace location is not independent of
commuter modal accessibility.

The importance of personal income in workplace location exists in all cities. We might
expect, for example, that individuals on higher incomes might on average work in the
central areas of each city or in the inner locations. This is supported by the set of
locations in which personal income is defined as alternative-specific for all cities.
Occupation-specific dummy variables have an important role in workplace location
decisions. Combined with the total number of job opportunities in each location, we are
able to differentiate location opportunities and observed choices according to the skill
base of a worker and the supply of jobs. For example, in Adelaide, labourers tend,
ceteris paribus, to have a higher probability of working in southern and outer southern
suburbs than elsewhere. In contrast, para professionals have a lower probability of
working in Salisbury, the outer north and the outer east than elsewhere in Adelaide;
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
90
while clerical workers tend to work in central Adelaide or in adjoining inner locations.
Part-time workers tend to work in areas closer in, in contrast to the southern locations.

Each cities workplace location choice model is different in terms of the set of socio-
economic descriptors. This might be expected given the spatial differences. There are
however some common trends associated with workplace opportunities in central
locations relative to middle to outer locations. Importantly we have identified a
significant number of occupation-based choices of workplace location and explicitly
allowed for the role of total job opportunities in each location. Together with the modal
choice inclusive value index, this enables us to account for the influence of modal
accessibility, occupation, income, and total job opportunities on workplace location
choice, which then flows through as an influence on other choices such as residential
location, household fleet size choice, and the utilisation rate of household automobiles.
In calibration, the mode choice inclusive values have been transferred across to all
workplace locations.



D2. Household Automobile Composition and Fleet Size Choice
Models

The Automobile choice module is a household level system of equations. The sequence
of estimation is as follows:

1. Estimate SP vehicle type choice models for each fleet size (0, 1, 2 and 3 plus). This
includes the zero vehicle households since we are interested in their propensity to
choose particular types of automobiles under alternative fuel, performance, price and
physical attribute scenarios. The SP experiment, fully documented in Appendix C is a
two-stage choice experiment. The first stage required a household member to consider
three conventional-fuelled automobiles, one from each size class, and to choose one.
The second stage added in three electric vehicles and three alternative-fuelled vehicles
(described as compressed natural gas (cng) vehicles). The household member was asked
to consider the 9 alternatives and choose one. This experiment was repeated three times.

The estimated SP vehicle type choice model has a maximum of 36 alternatives, defined
by 3 size classes, 3 fuels and 4 vehicle ages (see Table D13). Although a 36-alternative
choice model is empirically feasible, tests of reduced choice sets showed conclusively
that a reduced choice set of 12 alternatives would give statistically indistinguishable
parameter estimates. We adopted a rule of randomly selecting 4 alternatives from each
fuel class except that three were selected in the fuel class containing the chosen vehicle.
The age profile was randomised within each size class. The total sample size is 3996,
comprising 419 zero-vehicle households, 1520 1-vehicle households, 1375 2-vehicle
households and 682 3-plus vehicle households. The SP models are estimated for each
fleet size but for only one vehicle. The final models are summarised in Table D14.


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
91
Table D13 The Attributes and Attribute Ranges in the Final Vehicle Choice Experiment
Conventional Vehicle OPTION 1 OPTION 2 OPTION 3
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,25,35 30,40,50 30,55,80
2 yr old vehicle ($000) 11,18,26 22,30,37 22,42,62
5 yr old vehicle ($000) 7,12,17 15,20,25 15,30,45
10 yr old vehicle ($000) 5,7,10 9,12,16 10,20,30
Other costs: Regn. fee (excl. any ins.) $ 150,250,350 200,350,500 250,400,550
Fuel cost to travel 500kms
($)
20,40,60 30,60,90 40,80,120

Electric Vehicle OPTION 4 OPTION 5 OPTION 6
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,28,40 25,40,55 30,70,110
2 yr old vehicle ($000) 12,22,32 20,32,44 24,55,90
5 yr old vehicle ($000) 9,17,24 15,24,33 18,42,65
10 yr old vehicle ($000) 6,12,17 11,17,25 13,30,50
Other costs: Regn. fee (excl. any ins.) $ 50,100,150 75,125,175 100,150,200
Fuel cost to travel 500kms
($)
10,20,30 15,30,45 20,40,60
Features: (compared
t
Fully fuelled range (% of) 90,70,50 90,70,50 90,70,50
conventional
hi l )
Acceleration s,ss,cs s,ss,cs s,ss,cs
Boot size (% of) 90,60 90,60 90,60

Alternate Fuel Vehicle OPTION 7 OPTION 8 OPTION 9
Attributes Small Car Medium Car Large Car
Price: New vehicle ($000) 15,25,35 30,40,50 30,55,80
2 yr old vehicle ($000) 11,18,26 22,30,37 22,42,62
5 yr old vehicle ($000) 7,12,17 15,20,25 15,30,45
10 yr old vehicle ($000) 5,7,10 9,12,16 10,20,30
Other costs: Regn. fee (excl. any ins.) $ 50,100,150 75,125,175 100,150,200
Fuel cost to travel 500kms
($)
15, 30, 45 20,40,60 25,55, 85
Features: (compared
t
Fully fuelled range (% of) 100,85,70 100,85,70 100,85,70
conventional
hi l )
Acceleration s,ss,cs s,ss,cs s,ss,cs
Boot size (% of) 90,70 90,70 90,70

2. The parameter estimates for vehicle price, vehicle price interacted with household
income, vehicle operating cost (c/km) and vehicle annual registration (excluding
compulsory insurance) derived from the SP models are implemented in the RP model as
fixed estimates. The RP model is defined by a 10-alternative choice set, using a random
sampling procedure within each class size to assign vehicles of each vintage to the 10
alternatives given their size class. That is, the 60 alternative universal choice set (10
sizes by 6 vintages) for conventional-fuelled vehicles is reduced to 10 ranked
alternatives. The sample size for the RP data is not large enough to allow a 60-
alternative ranked model; an unranked choice set would have to be constructed.

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
92
The advantage of a ranked model is that we can introduce class-specific constants and
apply choice-based weights to reproduce the base market shares for the 10 size classes.
This is a major advantage, greatly simplifying the otherwise arduous task of calibrating
a vehicle type choice model ex post through simulating combinations of class-specific
constants. The RP model was estimated using a set of SP-derived parameters,
appropriately scaled in accordance with minimisation of the models overall log-
likelihood, together with a set of vehicle attributes and household socioeconomic
characteristics. The final models are summarised in Table D15.

Table D14 Vehicle Type Choice (Stated Preference) Models
(12 ranked alternatives)
Alternatives: conventional fuel (1-4 ), electric ( 5-8), alternative fuel (9-12)
Variable Name Alts 0 Vehicles 1-Vehicle 2-Vehicles 3+Vehicles
Vehicle price ($) PRICE 1-12 -.000088669
(-9.23)
-.000075621
(-15.08)
-.000074409
(-13.83)
-.000079582
(-10.00)
Price*hhld income
(000*000)
PRINC 1-12 .0015397
(6.63)
.00056203
(9.06)
.00049685
(7.55)
.00051543
(5.89)
Registration
cost ($)
REG 1-12 (0V) -.002628
(-2.99)
-.076529
(-1.78)

Registration
cost ($)
REGC 1-4 -.0012743
(-2.59)

Registration
cost ($)
REGE 5-12 -.0013863
(-1.34)
Operating
Cost (c/km)
OPCOST 1-12 -.034606
(-1.97)
-.022995
(-2.98)
-.023298
(-2.87)
-.037121
(-3.34)
Small vehicle
dummy (1,0)
SMC 1-4 (0,1 V)
1-8 (2 V)
1-12 (3 V)
.88965
(3.32)
-.90779
(-6.84)
-.79501
(-8.39)
-.75823
(-6.93)
2 year old vehicle YR2 1-8 (0 veh)
1-12(1,2,3 v)
-1.0802
(-4.83)
-.50558
(-6.86)
-.54445
(-7.02)
-.52093
(-4.86)
5 year old vehicle YR5 1-8 (0 veh)
1-12(1,2,3 v)
-.76195
(-3.92)
-.85851
(-10.70)
-.89402
(-10.56)
-1.0085
(-8.35)
10 year old vehicle YR10 1-8 (0 veh)
1-12(1,2,3 v)
-1.1845
(-5.62)
-1.6654
(-16.88)
-1.7529
(-16.64)
-1.9679
(-12.73)
Electric veh.
constant (1,0)
ELEC 5-8 -.549087
(-1.32)
-.76133
(-2.54)
-.26709
(-1.0)

Electric vehicle
range (kms)
RANGE 5-8 .00087039
(1.61)
.0010215
(1.78)
.00055278
(1.53)
Acceleration
(considerably slow -
50%) secs
ACCCS
E
5-8 -.016508
(-3.92)
-.014073
(-3.12)

Perth city
dummy (1,0)
CTY6 5-12 .67851
(1.79)

Small electric
vehicle (1,0)
SMALL
E
5-8 .94940
(2.52)

Medium electric
vehicle (1,0)
MEDE 5-8 .65054
(1.67)

Alt. fuel
constant (1,0)
ALTF 9-12 -.077117
(-.26)
-.66031
(-2.65)
-.51825
(-2.43)

Acceleration
(considerably slow -
50%) secs
ACCCS
A
9-12 -.010906
(-2.94)
-.012057
(-3.08)

Boot size (m
3
)
9-12 1.0278
(2.95)
1.5754
(4.15)
1.1797
(2.44)
Sample size 419 1520 1375 682
Log-likelihood
at zero
-1041.18 -3777.06 -3416.75 -1694.71
Log-likelihood at
convergence
-861.23 -3475.20 -3148.17 -1558.12
Pseudo R-squared .173 .080 .079 .081
Note: In order to quantify acceleration, boot size and range, all of which were expressed relative to
conventional fuelled vehicles in the same size class, we had to determine some weighted average
estimates. These estimates were obtained from the vehicle attribute data base for existing vehicles. The
figures used are:
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
93
Small Medium Large
Size Classes micro, small medium, upper med A Upper med B, Large
Acceleration
(0-100 kph in secs)
14.5 13.0 12.0
Luggage capacity (m
3
)
0.35 0.50 0.75
Range (kms) 580 580 580
Note: Considerably slower acceleration is 50% of conventional vehicle (ie twice as slow), slightly slower
acceleration is 80% of conventional vehicle (ie 1.25 times slower).

The RP models are defined by individual vehicles. Thus for multiple-vehicle
households we have to estimate a nested structure. The vehicles in the nest are ordered
according to the date that each vehicle enters the household. This behavioural rule
recognises that a vehicle acquisition (as part of a replacement or as an addition) is in
some sense influenced by the set of vehicles already present. Thus a two-vehicle
household has a two-level nested logit model. The bottom level is estimated first and
then an inclusive value derived for inclusion in the upper level model. After extensive
testing we were unable to find any evidence to support a statistically significant link
between the vehicles at each level. We suspect this can be explained by (i) the presence,
at each level, of socioeconomic attributes which already represent the major
determinants of the vehicle mix and (ii) the possible arbitrariness of vehicle chronology
in influencing the households required vehicle mix, such that the profile of vehicle
types at each level of the nested structure are very similar. Consequently each level was
treated independently such that the joint probability of each pair of vehicles (100 in a 2-
vehicle household) or triples (1000 in a three-vehicle household) is the product of
independent probabilities. The electric and alternative-fuelled vehicles are not included
in the estimation of the RP vehicle type choice models, but are added in with rescaling
of all parameter estimates in applications years where such vehicles are to be introduced
in a strategy. All the parameter estimates for the electric and alternative-fuelled vehicles
are set out below Table D15.

Table D15 Vehicle Type Choice (Revealed Preference) Models with Rescaled SP Parameters
Variable Name Alts 1 Veh 2 Vehs
Upper
2 Vehs
Lower
3+Vehs
Upper
3+Vehs
Middle
3+Vehs
Lower
Optimal Scale SC 1.4 2.1 1.2 2.5 0.95 0.5
Micro constant MC 1 -5.7104
(-2.40)
-9.0562
(-2.83)
-16.309
(-3.62)
-8.0141
(-1.89)
-15.711
(-2.86)
-12.627
(-2.620
Small constant SM 2 -6.0693
(-2.32)
-9.5719
(-2.91)
-16.450
(-3.62)
-8.0141
(-1.89)
-15.711
(-2.86)
-12.177
(-2.53)
Medium constant MD 3 -5.0894
(-1.95)
-7.8757
(-2.53)
-14.798
(-3.40)
-7.1034
(-1.78)
-14.637
(-2.80)
-11.177
(-2.44)
Upper Med A
constant
UA 4 -.56526
(-.79)
-2.7666
(-3.35)
-.43270
(-.49)
-1.3581
(-1.32)
-.42021
(-.36)
.92162
(1.96)
Upper Med B
constant
UB 5 .033155
(.04)
-2.7666
(-3.35)
.37080
(.41)
-.54792
(-.52)
.59964
(.52)
2.0368
(4.81)
Large constant LG 6 -1.8919
(-2.30)
-4.5644
(-4.90)
-.74563
(-.88)
-2.1268
(-1.89)
-.85571
(-.70)
.48219
(0.75)
Luxury constant LX 7 -.42323
(-.50)
-2.5777
(-2.86)
-.27397
(-.28)
-.44963
(-.41)
-.47021
(-.37)
.46543
(0.75)
Light Comm.
constant
LC 8 1.5603
(4.57)
1.6823
(4.30)
1.1359
(3.16)
1.2975
(2.37)
1.2079
(2.51)
1.1079
(2.12)
4 wheel drive
constant
FD 9 .37997
(.96)
0.56520
(1.28)
.70493
(1.52)
.69137
(1.10)
.70232
(1.10)
.63034
(1.07)
Vehicle price ($) PRCE 1-10 -.00007562
(fixed)
-.0000744
(fixed)
-.0000744
(fixed)
-.00007958
(fixed)
-.00007958
(fixed)
-.00007958
(fixed)
Price*hhld income
($000's*$000's)
PRIC 1-10 .00056203
(fixed)
.00049685
(fixed)
.00049685
(fixed)
.00051543
(fixed)
.00051543
(fixed)
.00051543
(fixed)
Registration
cost ($)
REG 1-10 -.0012743
(fixed)
-.0007653
(fixed)
-.0007653
(fixed)

Operating
Cost (c/km)
OPC 1-10 -.022995
(fixed)
-.023298
(fixed)
-.023298
(fixed)
-.037121
(fixed)
-.037121
(fixed)
-.037121
(fixed)
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
94
Luggage capacity
(m
3
) * hhld size
LUG 1-2 1.6723
(3.25)
3.9601
(3.64)
2.2659
(2.28)

Age of vehicle
/hhldsize
AGHZ 1-10 -.0071811
(-1.31)

ln (Age of vehicle) AGE 1-10 -.45643
(-6.51)
-.24547
(-2.89)
-.70684
(-8.85)
-.27683
(-2.47)
-.37557
(-4.63)
Acceleration
(0-100 kph) secs
ACC 1-10 -.23346
(-2.94)
-.52583
(-5.91)
-.17599
(-1.96)
-.29539
(-2.52)
-.15571
(-1.43)

4 cylinder (1,0) 4CYL 1-3 1.4637
(2.30)
1.5690
(2.17)
3.8170
(3.58)
1.7180
(1.87)
3.8038
(2.99)
3.2181
(2.91)
Ln(Vehicle
weight)(kg)
VM 1-10 -2.4176
(-1.67)
-3.9211
(-1.96)
-2.3584
(-1.460)
-3.9111
(-1.91)
-3.5310
(-1.41)

Sample size 494 441 441 211 211 211
Log-likelihood
at zero
-1137.48 -1015.44 -1015.44 -485.85 -485.85 -605.58
Log-likelihood at
convergence
-956.01 -834.64 -836.84 -394.87 -406.77 -511.10
Pseudo R-squared .159 .178 .176 .187 .163 .156

plus the two non-conventional fuel utility expressions (rescaled entirely):

1-vehicle household:

U(ELC)=1.4(-.76133-.000075621*PRICE+.00056203*PRICE*HINCOME-.022995*OPS-.012743*RGS
+.00087039*RANGE -.016508*ACCCONSL-.50558*Y2-.85851*Y5-1.6654*Y10)
U(ALF)=1.4(-.66031-.000075621*PRICE+.00056203*PRICE*HINCOME-.022995*OPS-
.012743*RGS+1.0278*BOOT -.010906*ACCCONSL-50558*Y2-.85851*Y5-1.6654*Y10)

2-vehicle household (SC=1.2, 2.1 receptively for lower and upper levels):

U(ELC)=SC(-.26709*SC-.000074409*PRICE+.00049685*PRICE*HINCOME-.023298*OPS-
.00076529*RGS+.0010215*RANGE-.014073*ACCCONSL-.54445*Y2-.89402*Y5-1.7529*Y10)
U(ALF)=SC(-.51825-.000074409*PRICE+.00049685*PRICE*HINCOME-.023298*OPS-
.00076529*RGS+1.5754*BOOT-.012057*ACCCONSL-.54445*Y2-.89402*Y5-1.7529*Y10)

3+vehicle household (SC=0.5, 0.95 AND 2.5 respectively for lower, middle and upper levels):

U(ELC)=SC(-.000079582*PRICE+.00051543*PRICE*HINCOME-.037121*OPS-.0013863*RGS
+.00055278*RANGE-.75823*SMALL-.52093*Y2-1.0085*Y5-1.9679*Y10)
U(ALF)=SC(-.000079582*PRICE+.00051543*PRICE*HINCOME-
.037121]*OPS.0013863*RGS+1.1797*BOOT-.75823*SMALL-.52093*Y2-1.0085*Y5-1.9679*SY10)

Table D16 Vehicle Class Sample Shares (percent)
Variable 1 Veh 2 Vehs
Upper
2 Vehs
Lower
3+Vehs
Upper
3+Vehs
Middle
3+Vehs
Lower
Total
Micro 7.7 9.1 8.6 6.2 8.5 8.7 7.9
Small 26.9 30.2 24.7 23.2 21.3 27.0 25.6
Medium 23.7 21.1 21.3 21.8 15.6 17.9 21.1
Upper Med A 6.7 3.2 4.5 3.8 4.7 3.0 4.6
Upper Med B 21.1 17.2 21.1 20.9 22.3 17.5 20.0
Large 3.4 3.2 2.9 8.1 4.7 3.4 3.8
Luxury 1.6 3.6 2.3 3.3 1.4 2.7 2.6
Light
Commercial
4.9 5.0 7.3 7.1 7.1 6.8 6.1
4 wheel drive 2.6 3.2 3.9 1.9 7.6 4.9 3.8
Truck/Utility 1.4 4.3 3.4 3.8 6.6 8.0 4.5

The strategy simulator can be used to evaluate the influence of the introduction of a
carbon tax (distinguishing the carbon content of each fuel type - petrol, diesel, cng etc.),
evaluate the impact of changes in excise duties, the cost per recharge of an electric
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
95
vehicle battery, the cost of a full tank of cng, and sales tax on each automobile class.
Noteworthy in the results is the importance of vehicle range before recharge for electric
vehicles, boot size for alternative-fuelled vehicles (usually reduced in size by a large
gas tank), and the acceleration performance of both electric and alternative-fuelled
automobiles.

The stated preference model has enabled us to identify parameter estimates for the fixed
registration charge (which is usually not statistically significant in a cross-sectionally
based RP model). Since the possibility exists for skewed registration charges based on
the emission characteristics of vehicles in particular classes to be used as a mechanism
for increasing the rate of removal of high-polluting vehicles from the stock, it is
necessary to parameterise registration charges.

3. For each sampled household, we use each vehicle type choice model, conditional on
fleet size, to derive an inclusive value associated with each fleet size. In the current
application, the vehicle type choice inclusive values have a weak albeit statistically
significant influence on household fleet size, the latter being more strongly linked to
commuter mode choice decisions and workplace location choice. Since this is the first
study, to our knowledge, to have multiple sources of linkages through to household fleet
size choice, we might postulate that the linkages between vehicle type choice and fleet
size choice in previous studies were in part representing some elements of modal
accessibility and location of employment. That is, after allowing for the expected
maximum utility associated with workplace and commuter mode choice decisions in the
decision on how many automobiles a household requires, it diminishes the importance
of the expected maximum utility associated with the choice of vehicle type, given fleet
size.

The fleet size choice model is a household-level model with four alternatives - zero,
one, two and three-plus automobiles. The choice of fleet size is influenced by commuter
mode choice, workplace location choice, vehicle type choice, distance of the
households residence from the central business district of their respective city, and a
number of household-level socioeconomic variables (Table D17). Given that two modal
alternatives in the commuter mode choice model define the use of household
automobiles, it is problematic to link the mode choice and fleet size models by the
conventional inclusive value index. An alternative specification is required. We have to
treat the choice of public versus private transport as a dummy endogenous variable and
estimate the fleet size choice model taking this endogeneity into account. In
applications the (1,0) dummy variable is replaced with the probability of choosing
public transport. Any changes in the levels of attributes in the mode choice model
(including the introduction of new modes of transport - light rail and busway) are
allowed for in their impact on household fleet size choice by adjustments in the public
transport probability (as a sum of the probabilities of choosing each form of public
transport in the mode choice set).

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
96
Table D17 Fleet Size Choice Model. Alternatives: 0 (8.6%), 1 (36.9%), 2 (36.5%) and 3 plus
(18.0%), choice-based weights of 1.6310, 1.1379, 0.9046 and 0.6102
Variable Acronym 0 Vehs 1 Veh 2 vehs 3+Vehs

One parent family (1,0) SPFAM 0.80082
(1.86)

Single person household (1,0) SPERS 1.4275
(4.94)

Fleet size specific constant FS 0.65834
(1.28)
-1.6503
(-2.94)
-6.5693
(-8.94)
Couples without children (1,0) CNOKIDS -.18719
(-1.56)

Distance to CBD (kms) DIST .0093012
(1.23)
.018601
(1.89)
.030062
(2.81)
Two parent family (1,0) TWOPFM .27166
(1.92)

Number of workers NW .76961
(5.60)
1.9373
(11.14)
Household income (000's) HINC .030474
(3.37)
.047238
(5.12)
.049683
(5.18)
Age of head of household AGEHD .040041
(4.52)
Public transport probability index PTPROB -.91624
(-1.65))
-2.1490
(-1.90)
Inclusive value workplace
location choice
IVWL .013577
(1.80)
.013577
(1.80)
.013577
(1.80)
Inclusive value vehicle type choice IVATC .045
(1.32)
.02
(1.20)
.04
(1.10)
Sample size 699
Log-likelihood at zero -1282.32
Log-likelihood at convergence -962.23
Pseudo R-squared .26

The set of eligible socioeconomic characteristics is limited by those available in the
description of each synthetic household in the strategy simulator. There is no explicit
allowance for the availability of company-provided automobiles in the household fleet.
That is we did not estimate a model to identify the probability that a household has
access to a company car; rather company cars in the household are counted as part of
the total fleet size in the fleet size choice model. Exogenous variables such as household
income in part account for the presence or absence of such vehicle registration
categories. However the presence of an inclusive value from the workplace location
choice model and the commuter mode choice probability conditions the model on the
exogenous role of occupation and use of an automobile for the commuter trip, which are
additional indicators of access to a business-registered automobile.

The inclusive value parameter estimate for workplace location in the non-zero vehicle
utility expressions is statistically significant although relatively close to zero. The
negative public transport probability index tells us that improvements in public
transport services will reduce the probability of owning 2 or 3 vehicles, with the impact
being greatest for 3 vehicles. Likewise an improvement in the operating cost and levels
of travel times for car commuting will, via a reduction in the probability of choosing
public transport, increase the probability of a household having 2 or 3 automobiles. The
probability of more vehicles in the household increases as the number of workers
increases, and as household income increases. Specifically, the probability of a
household with more than two vehicles increases as the age of the head increases which
might represent the presence on average of a greater number of households members of
driving licence age (and independent financial means). Two parent families, ceteris
paribus, have a higher probability of choosing two vehicles, and couples without
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
97
children have a lower probability of choosing 1 vehicle relative to none and more than 1
vehicle.

The empirical relationships are plausible and rich in socioeconomic, locational and
modal explanation. Within the strategy simulator, automobile ownership levels will be
conditioned on where household live, where workers work, the choice of commuter
mode and the lifecycle of a household.



D3. Residential Location Choice Models

The final set of choice models are associated with the households choice of residential
location and type of dwelling to reside in. This module is important for its general role
in determining total use of automobiles and public transport, as well as in evaluating the
impact of strategies associated with the residential density profile of the urban area.
Through exogenous shocks to the total number of dwellings and the mix of dwellings in
each geographic zone, we can evaluate the implications for automobile use and hence
greenhouse gas emissions.

The residential location choice model conditions a number of prior models such as
workplace location choice, dwelling type choice, commuter mode choice, fleet size and
departure time choice. The last two choices are not directly linked by inclusive values,
but contain a distance attribute in the utility expressions which requires that the
probability of a household having a particular fleet size and the probability of a
commuter departing for work at a particular time varies according to the location of
their residence relative to the central business district. The commuter mode choice is
conditioned on workplace and residential location, producing differentiation in the
origin-destination probabilities of choosing each mode for the commuter trip. The zonal
system for residential location choice is the same as that for workplace location choice.

The residential dwelling type choice model results are summarised in Table D18. There
are three dwelling types, each representing a level of residential density. The low
density choice is a detached dwelling, the medium density choice is a semi-detached
dwelling, and the high density choice is a unit or apartment. The actual profile of
dwellings by city is summarised in Table D19, derived from CDATA91. The most
important variable in this model is dwelling price, which as expected has a negative
parameter estimate and is statistically significant. This variable is an important link with
residential location choice and is the prime financial indicator of residential location
choice, after allowing for the households ability to finance their dwelling, represented
by household gross income. Actual median dwelling prices by location for each city are
summarised in Table D19.

The utility expression for a detached dwelling contains variables which suggest that,
ceteris paribus, the greater the income of the household, the number of workers, the
further out the household lives from the central business district and the presence of a
lifecycle described as a two-parent family, the higher probability of choosing to live in a
detached house. These are all plausible explanatory effects in a model with an
impressive goodness-of-fit.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
98
Table D18 Residential Dwelling Type Choice
Alternatives: 1(low density) =detached dwelling, 2 (medium density) =town house/semi-detached,
3(high density) =unit/flat. Sample shares: low density =81.5%, medium density =9.1%, high density =
9.5%
Variable Alt Acronym Parameter Estimate T-statistic
Detached specific constant 1 DET 2.1416 4.08
Two-parent family (1,0) 1 TPF 1.0074 2.64
Distance to CBD (kms) 1 DIST .040123 2.56
Household income (000's) 1 HINC .016784 2.36
No. of workers 1 NW .35220 1.63
Ln of median dwelling price ($) 1-3 LPHOUSE -.77857 -2.05
Semi-detached constant (1,0) 2 SDET .83480 2.48
Single-person household (1,0) 3 SPH .1.0061 2.29
Sample size 1269
Log-likelihood at zero -1368.871
Log-likelihood at convergence -256.42
Pseudo R-squared .813

The inclusive value from the dwelling type choice model enters the residential location
choice model carrying important determinants of location choice (Table D20). The
absence of data on location rents required us to use dwelling prices as a proxy. A
weighted average dwelling price based on the current composition of the dwelling stock
in each location was derived. Throughout the set of models for each city we had to
evaluate the implications of having common socioeconomic and financial attributes in
both the dwelling type choice and residential location choice models. The key attributes
of dwelling prices, household income and total stock of dwellings had to be carefully
assessed. Our preferred measure is the ratio of weighted dwelling prices to household
income (HPHINC) which encompasses all three key variables. This specification was
statistically significant only in Sydney; for the other cities household income appears in
Melbourne, Adelaide and Perth, and the total stock of dwellings is in the utility
expressions for Melbourne and Brisbane. Thus we have, to varying degrees, the direct
and indirect impact of dwelling prices, household income and the stock of dwellings in
all residential location choice models. The other influences on location choice are the
inclusive value associated with workplace location choice and socioeconomic
characteristics of each household, primarily life cycle status.

The residential location choice model completes the suite of discrete choice models. A
selectivity variable representing the potential influence of unobserved effects
influencing residential location choice is calculated and assessed in the vehicle use
model system to identify any possible link between residential location choice and
automobile use. Together with selectivity variables from fleet size choice, workplace
location choice and commuter mode choice, we have four potential influences on
automobile use dues to locational and modal decisions.
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
99
Table D19 Population and house prices by geographical location
CANBERRA LOW
DENSITY
MED
DENSITY
HIGH
DENSITY
AREA NAME POPULATION PRICE PRICE PRICE
101 Central Canberra 61,000 220000 196000 184000
102 Belconnen 85,508 161700 144060 135240
103 Woden Valley 32,617 185900 165620 155480
104 Weston Creek 25,950 163900 146020 137080
105 Tuggeranong 71,602 151800 135240 126960
106 Queanbeyan 27,160 143000 127400 119600


SYDNEY LOW
DENSITY
MEDIUM HIGH
AREA NAME POPULATION PRICE DENSITY DENSITY
201 Inner Sydney 262,261 $190,000 160650 145350
202 Eastern Suburbs 224,348 $335,000 203700 184300
203 St George/Sutherland 378,791 $217,000 166950 151050
204 Canterbury Bankstown 283,170 $157,000 114450 103550
205 Fairfield/Liverpool 273,322 $131,000 98700 89300
206 Outer South Western Sydney 190,640 $133,000 97650 88350
207 Inner Western Sydney 148,070 $223,000 195300 176700
208 Central Western Sydney 260,439 $161,000 157500 142500
209 Outer Western Sydney 270,461 $146,000 93450 84550
210 Blacktown/Baulkham Hills 325,804 $165,000 158550 143450
211 Lower Northern Sydney 258,463 $318,000 194250 175750
212 Hornsby/Ku-ring-gai 226,887 $306,000 213150 192850
213 Manly/Warringah 206,907 $297,000 189000 171000
214 Gosford/Wyong 229,407 $141,000 155400 140600


MELBOURNE LOW
DENSITY
MEDIUM HIGH
AREA NAME POPULATION PRICE DENSITY DENSITY
301 Central Melbourne 227,420 $192,000 126000 114000
302 Western Inner Melbourne 121,665 $138,000 94500 85500
303 Western Outer Melbourne 234,588 $110,000 94500 85500
304 Western Fringe Melbourne 106,966 $95,000 105000 95000
305 Northern Inner Melbourne 90,511 $116,000 90300 81700
306 Northern Middle Melbourne 184,011 $130,000 115500 104500
307 Northern Fringe Melbourne 142,297 $105,000 105000 95000
308 Northern Outer Melbourne 199,088 $127,000 114450 103550
309 Eastern Inner Melbourne 141,096 $210,000 157500 142500
310 Eastern Middle Melbourne 266,302 $152,000 140700 127300
311 Eastern Outer Melbourne 299,565 $126,000 101850 92150
312 Eastern Fringe Melbourne 141,275 $110,000 71400 64600
313 Southern Inner Melbourne 171,665 $197,000 136500 123500
314 Southern Outer Melbourne 175,637 $130,000 94500 85500
315 South Eastern Inner Melbourne 146,753 $104,000 94500 85500
316 South Eastern Outer Melbourne 167,584 $104,000 94500 85500
317 Mornington Peninsula Inner 110,808 $106,000 99750 90250
318 Mornington Peninsula Outer 95,216 $105,000 89250 80750


TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
100
Table D19 continued
BRISBANE LOW
DENSITY
MEDIUM HIGH
AREA NAME POPULATION PRICE DENSITY DENSITY
401 Far South 192564 160600 143080 134320
402 Redland Shire 80695 161700 144060 135240
403 Ipswich-Moreton Shire Part A 108035 154000 137200 128800
404 Outer Northern 114301 143000 127400 119600
405 Pine Rivers Region 113705 159500 142100 133400
406 North East 111561 231000 205800 193200
407 Metropolitan North 94128 157300 140140 131560
408 Western Suburbs 130434 211200 188160 176640
409 Outer South West 123849 170500 151900 142600
410 Metropolitan South 119089 192500 171500 161000
411 Eastern Suburbs 133264 167200 148960 139840


ADELAIDE LOW
DENSITY
MED
DENSITY
HIGH
DENSITY
AREA NAME POPULATION PRICE PRICE PRICE
501 Salisbury 106,011 $84,000 $53,000 $73,000
502 Outer North 75,924 $78,667 $61,000 $77,333
503 Outer East 106,657 $140,295 - $99,437
504 North West 133,271 $111,750 $99,115 $94,091
505 North East 88,795 $269,900 $92,014 $83,722
506 Inner Adelaide 88,294 $152,869 $135,207 $111,221
507 South Metropolitan 130,399 $162,076 $139,791 $97,240
508 South West 105,324 $126,063 $123,919 $102,258
509 South 95,642 $133,292 $145,054 $90,912
510 Outer South 93,306 $83,000 $103,000 $74,000


PERTH LOW
DENSITY
MED
DENSITY
HIGH
DENSITY
AREA NAME POPULATION PRICE PRICE PRICE
601 Wanneroo $167,889 $129,331 $115,223 $108,168
602 Stirling $152,435 $164,242 $146,325 $137,336
603 Inner North East $104,297 $139,408 $124,200 $116,595
604 Outer North East $128,287 $138,946 $123,788 $116,209
605 Outer South $114,095 $107,645 $95,902 $90,030
606 South Metropolitan $135,219 $174,022 $155,038 $145,546
607 Inner Coastal $110,769 $287,670 $256,288 $240,597
608 Inner South East $135,525 $127,531 $113,618 $106,662
609 Perth City $94,749 $178,011 $158,591 $148,882
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
101
Table D20 Residential location choice model
Variable Acronym Units Alts Estimated
parameters
t-values
CANBERRA:
Log house prices LHPRICE $ '000's all -3.1369 -.81
Log of Distance to CBD
from zone
LDISDWCBD kms all -2.8172 -1.49
Two-person family TWOPFAM 1,0 102 .97170 1.77
Couple with no children CNOKIDS 1,0 103-106 -.78507 -1.50
Inclusive value workplace
location
IVWL index 101, 102 .039442 1.06
Inclusive value dwelling type IVDT index 102, 103-106 .55220 1.20
Zone specific constant 101 -8.9468 -1.86
Zone-specific constant 102 -2.1857 -3.38

Sample size 83
Log-likelihood at zero -91.185
Log-likelihood at
convergence
-79.637
Pseudo R-squared .13
SYDNEY:
House price/hhld income HPHINC $000/$00
0
all -.40553 -1.96
Single person household SPERS 1, 0 210-205,210 1.4995 1.91
Single person household SPERS 1, 0 211 -9.6836 -12.30
Single parent household SPFAM 1,0 203-205,210,211 -11.991 -19.66
Inclusive value workplace
location
IVWL index 203-214 .09000 1.36
Inclusive value dwelling type IVDT index 201-203, 207,208,
211-214
.34195 1.65
zone 204 dummy Z204 1,0 204 -.16553 -.29
zzone 205 dummy Z205 1,0 205 -.77633 -1.24
zone 206 dummy Z206 1,0 206 .23985 .41
zone 207 dummy Z207 1,0 207 -.92131 -1.28
zone 212 dummy Z212 1,0 212 2.4166 2.97
zone 213 dummy Z213 1,0 213 1.9036 2.27
Zone-specific constant 201,202 2.5413 1.70
Zone-specific constant 203 .94748 1.24
Zone-specific constant 204,205,210 -.35557 -.32
Zone-specific constant 206,209 -1.3773 1.30
Zone-specific constant 207,208 1.1538 1.54
Zone-specific constant 211 1.2863 1.53

Sample size 135
Log-likelihood at zero -227.67
Log-likelihood at
convergence
-205.58
Pseudo R-squared .10
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
102
Table D20 continued
MELBOURNE: Acronym Units Alts Estimated
parameters
t-values
Log household income LHINCOME '000's 301,302,305,309,
313,315,304,307,
312, 306,310
.83051 2.21
Log of Distance to CBD
from zone
LDISDWCBD kms all -.90112 -2.23
Log of total no. of dwellings LTOTDWL no. all 1.5792 2.72
Single parent household SPFAM 1,0 303,308,311,314,
316
10.911 13.82
Single parent household SPFAM 1,0 306,310 12.059 16.58
Two person household TWOPFAM 1,0 303,308,311,314,
316, 317,318
.69783 1.67
Two person household TWOPFAM 1,0 304, 307, 312 2.2759 3.07
Inclusive value dwelling type IVDT index 301,302, 305, 309,
313,315
.23015 1.32

Zone-specific constant 301 -5.0310 -1.86
Zone-specific constant 302,305,309,313,
315
-2.0041 -.77
Zone-specific constant 303,308,311,314,
316
.11076 .20
Zone-specific constant 304,307,312 -4.1152 -2.58
Zone-specific constant 306,310 -4.4442 -2.94

Sample size 130
Log-likelihood at zero -232.93
Log-likelihood at
convergence
-194.13
Pseudo R-squared .17
BRISBANE:
Log of total no. of dwellings LTOTDWL no. all 1.8175 3.68
Single parent household SPFAM 1,0 401,403,404 .94173 1.77
Single parent household SPFAM 1,0 402, 411 1.3268 2.45
Two person household TWOPFAM 1,0 401,403,404 .84346 2.64
Two person household TWOPFAM 1,0 405,406 1.2342 3.25
Inclusive value dwelling type IVDT index 401-406, 408, 409,
411
.20823 1.54
Zone-specific constant 401,403,404 .0065957 .024
Zone-specific constant 405,406 -.40339 01.29
Zone-specific constant 407,410 -1.0047 -1.313
Zone-specific constant 408,409 .12912 .56

Sample size 250
Log-likelihood
at zero
-402.36
Log-likelihood at
convergence
-376.19
Pseudo R-squared .07
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
103
Table D20 continued
ADELAIDE: Acronym Units Alts Estimated
parameters
t-values
Log of household income LHINCOME $'000's 509,510 2.2303 3.37
Single parent household SPFAM 1,0 505-507 2.5394 2.28
Couple without children CNOKIDS 1,0 505-507 1.7722 2.59
Single person household SPERS 1,0 505-507 1.9814 2.29
Inclusive value workplace
location
IVWL index 505-507 .14615 1.16
Inclusive value dwelling type IVDT index 505-507 .81358 1.33
Zone-specific constant 501,502,503 8.7718 3.43
Zone-specific constant 504,508 8.7068 3.41
Zone-specific constant 505, 506,507 8.4042 1.42

Sample size 92
Log-likelihood at zero -127.54
Log-likelihood at
convergence
-114.76
Pseudo R-squared .11
PERTH:
Log of Distance to CBD
from zone
LDISDWCBD kms all -.71504 -1.95
Log of household income LHINCOME $'000's 609 .62723 1.43
Couple without children CNOKIDS 1,0 603,608 .51764 1.31
Single person household SPERS 1,0 609 1.5182 2.06
Inclusive value workplace
location
IVWL index 601-608 .66770 1.13
Inclusive value dwelling type IVDT index 601-608 .64810 1.32
Zone-specific constant 603,608 -.79830 -1.86
Zone-specific constant 604,605 -.25558 -.91
Zone-specific constant 606,607 -.050043 -.18
Zone-specific constant 609 -2.1555 -.29

Sample size 149
Log-likelihood at zero -239.81
Log-likelihood at
convergence
-223.52
Pseudo R-squared .07



D4. Total Annual Vehicle Use and Composition

The final module is automobile use. The emphasis is on both the total annual use of
each automobile in the households fleet and its composition. The latter is disaggregated
into commuting travel, travel as part of work, other urban travel and non-urban
kilometres. There are three equations in the model system for annual vehicle kilometres
associated with (i) travel as part of work plus travel to/from work, (ii) other other travel,
and (iii) non-urban travel. Separate generalised least squares regression equations
(correcting for heteroskedasticity) were estimated for each of 1, 2 and 3-plus vehicle
households.

Annual vehicle use is a function of the cost of operating an automobile, which embodies
the fuel performance of an automobile as well as the mix of city and highway driving
cycle, the extent to which public transport is used for commuting, and socioeconomic
characteristics of a household. Selectivity variables (SC
i
) from the workplace location
choice, commuter mode choice, residential location choice and vehicle type choice,
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
104
although candidates for linking, were not statistically significant in the current
application context. Typically we might expect non-significant selectivity variables if
the vehicle use model is well specified since the key correlates have already been
identified by explicit exogenous variables of instrumented endogenous variables. The
concept of selectivity is set out in Hensher et al. (1992). It is worthy of noting that
typically, any statistically significant selectivity corrections often have a minimal
impact on the predicted annual vehicle kilometres.

Important policy variables are included in the final set of vehicle use models (Table
D21). These include vehicle operating cost fully decomposed to contain information on
fuel efficiency, fuel prices and excise duties. OPCOST is the link into the application of
a carbon tax and a congestion charge:

OPCOST = [{cityFuel*propCityF + hwyFuel*(1-propCityF)}*0.01] *
[tPricePetrol*(1-propnDiesel) + tPriceDiesel*propnDiesel +
carbonTax * {carbLitD*propnDiesel+carbLitP*(1-propnDiesel)}] +
(cTank +carbonTax*carbTAlt) / rangealf +(cCharge +carbonTax*
carbTElc) / rangeelc

where:

cityFuel =city cycle fuel efficiency (litres/100 km)
hwyFuel =highway cycle fuel efficiency (litres/100 km)
propCityF =proportion of use which is in the city fuel cycle (default =0.7)
propnDiesel =proportion of conventional-fuelled vehicles using diesel
tPricePetrol =wpricepetrol +expricepetrol (cents per litre)
wpricepetrol =wholesale price of petrol (cents per litre)
expricepetrol =excise component of price of petrol (cents per litre)
tpricediesel =wpricediesel +expricediesel (cents per litre)
wpricediesel =wholesale price of diesel (cents per litre)
expricediesel =excise component of price of diesel (cents per litre)
carbonTax =carbon tax (cents per kg)
carbLitD

=carbon per litre of diesel (kg/litre)
carbLitP =carbon per litre of petrol (kg/litre)
carbTElc =carbon per full electric recharge (kg)
carbTAlt

=carbon per tank of alternative fuel (kg)
cTank =wctank +extank (cents)
wctank =wholesale cost of a tank of alternative fuel (cents)
extank =excise component of cost of a tank of alternative fuel (cents)
cCharge =wccharge +excharge (cents)
wccharge =wholesale cost of a full electric recharge (cents)
excharge =excise component of cost of a full electric recharge (cents)
rangeelc =range of electric vehicle on a fully charged battery (km)
rangealf =range of alternative fuelled vehicle on a full tank (km)

TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
105
Table D21 Vehicle Use and Composition Results (Estimation by generalised least squares
(GLS)) (T-statistics in parenthesis), LVKMW = natural log of kms which are to and
from work and urban travel as part of work, LVKMEL = natural log of kms which are
urban non work-related, and LVKMCO = natural log of kms which are non-urban. *
OthHld is life cycle category remaining after removing SPERS, SPFAM, CNOKIDS,
and TWOPFAM.
Variable Acronym Units 1-Vehicle Hhld 2-Vehicle Hhld 3-Vehicle Hhld
Sample Size 271 770 292
LVKM W Model:
Constant 9.0300 (29.26) 9.1369 (12.03) 9.0886 (22.32)
Ln (fuel cost) OPCOST c/km -.088441 (1.75) -.20172 (-1.56) -.28224 (-1.87)
household income

HINCOME income in
'000's
.038288 (1.75) .0020113 (1.58)
hhldincome/no. of
workers
HINCPWRK INCO .00212314 (5.27)
Prob using public
transport
PTPROB 0 to 1 -..18383 (-2.01)

-.62963 (-1.59) -.024785 (-1.54)
Adelaide City ADL 1,0 -.26959 (3.25)
Single person
family
SPFAM 1,0 .66966 (2.76)
Two parent family TWOPFAM 1,0 .37943 (2.12)
No children CNOKIDS 1,0 .22241 (1.64 )
Single person SPERS 1,0 .53802 (2.37 )

R-squared .1 .1 .11
LVKMEL:
Constant .8.5486 (24.92) 9.1156 (37.7) 8.1498 (19.32)
household income HINCOME income
'000's
-.0052351 (2.75)
hhldincome/no. of
workers
HINCPWRK INCO .00039696 (1.6)
Ln(household
income)
LHINCOME log ofincome
in '000's
.074312 (1.78)
Ln (fuel Cost ) OPCOST c/km -.16499 (-2.21) -.16052 (-1.55) -..15264 (-2.01)
Single person
family
SPFAM 1,0 -.45680 (-3.23) .16777 (1.45) .28983 (1.95)
No children CNOKIDS 1,0 -.12454 (-1.63) -.11945 (-2.34) -.38300 (-2.63)
Single person SPERS 1,0 -.135191 (-1.6) .34217 (1.54)

r-squared .12 .23 .21
Variable Acronym Units 1-Vehicle Hhld 2-Vehicle Hhld 3-Vehicle Hhld
LVKMCO:
constant 7.0951 (5.31) 8.7013 (36.5) 7.0715 (18.08)

Ln (fuel Cost) OPCOST -.12757 (-1.78) -.056266 (-1.89) -.044317 (1.99)
household income HINCOME income in
'000's
.010919 (1.92)
hhldincome/no. of
workers
HINCPWRK INCO .0033521 (2.51) .0033522 (2.63)
Single-person
family
SPFAM 1,0 -.14893 (-1.43) -.14893 (
Single person SPERS 1,0 -.58612 (-1.61)
Other hhld life
cycle (*)
.18614 (1.86) .19245 (1.65)
Adelaide City ADL -.99708 (-2.36)
Canberra City CAN .14879 (2.13) .12876 (2.10)
Perth City PER 1,0 -.91783 (-1.98) -.069182 (-1.43) -.065234 (-1.57)

r-squared .15 .14 .15
TRESIS: A Transportation and Environmental Strategy Impact Simulator
David A. Hensher
106
The equations are driven by socioeconomic characteristics, vehicle operating cost, use
of public transport for commuting and the city-specific dummy variables. Operating
costs have the expected negative relationship with vehicle use with implied elasticities
of vehicle use ranging from -.04 to -.28. These are short run elasticities, holding
location and vehicle ownership decisions fixed. 1-vehicle households appear to be the
least sensitive to operating costs for commuting but the most sensitive for non-urban
vehicle use. There is very little difference in sensitivity to operating costs across all
three fleet sizes for other urban vehicle use. 1-vehicle households who use a car for
commuting appear to be constrained in terms of the opportunities for modal switching
or use adjustment after allowing for the household income effect. In contrast, multi-
vehicle households appear to have more opportunities including automobile switching
as well modal and time of day switching.

There are some interesting influences of life cycle on annual vehicle use. Households
with no children tend to have lower vehicle kilometres for non-work related urban
travel for all fleet sizes, although the influence is greater in 3-plus vehicle households.
Single parent families tend to travel less non-work related urban kilometres in 1 vehicle
households, but do more kilometres in multi-vehicle households. The fleet size effect
may be telling us a lot about lifestyle and possibly less opportunities for such travel in
the single parent household.

We evaluated the implications of modelling each city separately but found that a pooled
model with city-specific dummy variables for Perth (in total kilometres) and Adelaide
(in proportion of commuting kilometres) in particular was appropriate. Adelaide
commuters with 1 vehicle generally travel less kilometres to and from work and as part
of work and out of Adelaide than do those in the other 5 cities. In multi-vehicle
households, we see a lower average annual use for Perth households of vehicles for
non-urban travel, possibly reflecting the very long distance to major interstate locations,
better served by air travel.

In the application of the strategy simulator, any policy instrument which changes
vehicle operating costs will influence total kilometres and its composition. Fleet size is
influenced by operating costs via vehicle type choice and the commuter mode choice
model and vehicle type choice is influenced by operating costs. Furthermore the
location of residence and workplace conditions vehicle use. The probability weighted
vehicle kilometres model (for each fleet size, vehicle type, commuter mode choice,
given work and residential location) produces predicted annual kilometres and their
composition for each household.