Improving economic evaluation of urban transport projects in Australia
Improving economic evaluation of urban transport projects in
Australia
David Bray
1
Economic and Policy Services Pty Ltd, Adelaide, Australia
1 Introduction
This paper was prompted by observation of the use of data from computerised integrated
travel demand models in the evaluation of urban transport projects. Two principal issues
emerge. The first relates to matrices used in travel demand models. It may be expected that
many policy options and project proposals initiated by governments seek to alter travel
demand, either the mode, time or location of travel or, more fundamentally, land use. Others
may affect demand, irrespective of government intentions. Current levels of congestion
suggest a likely difficulty in accommodating future unconstrained travel demand on an
existing (or even ‘do minimum’) Base Case. Use of a fixed trip matrix for modelling travel
demand in future years to provide information for economic evaluations is inappropriate in
these circumstances because it does not fully reflect the potential effects of projects and
policies on travel demand. In general, use of fixed trip matrices is likely to result in over-
estimation of project benefits, though this is not certain, nor is the potential extent of over-
estimation.
The second key issue is the measurement of traveller benefits. The literature on this subject
goes back over three decades. It makes clear the limited extent to which link based
information from travel demand models can be used as input to economic evaluation. Rather,
evaluations of projects that result in a change in the travel demand matrix relative to the
Base Case need to draw on both origin-destination based information and link based data.
Some key guidelines for evaluation practice in Australia do not provide information on this
correct practice.
Finally, this paper raises several other issues related to the evaluation of urban transport
projects, in particular the unit of account used in valuing benefits, disaggregation of benefits,
quality of valuation data, definition of the Base Case, and sensitivity testing.
The objective of this paper is not to set out recommendations for future practice, though an
implication that emerges is that use of variable trip matrices should be the default
methodology for demand forecasting of major urban transport projects and fixed trip matrices
should be used only when inelastic demand is a realistic condition Nor does it present new
data or methodologies. Rather, it notes some key historic literature that remains pertinent to
current transport planning and evaluation practice.
2 Use of fixed and variable trip matrices
A fundamental theme of most urban transport strategies for Australian capital cities is travel
demand management, with a desire for increased use of public transport and reduced car
use. It may therefore be expected that many policy options and project proposals initiated by
governments seek to alter travel demand, either the mode or time of travel or, more
fundamentally, land use (ie the location of activities, and therefore travel). Others may affect
demand, irrespective of government intentions, whilst some may simply seek to
accommodate travel demand.
1
The author acknowledges the significant role of Philip Sayeg of Policy Appraisal Services Pty Ltd,
Brisbane, in stimulating this paper and discussions on many aspects of it. The paper has also
benefited from discussions with Dr Peter Tisato, Professor Derek Scrafton and Dr Dimitris Tsolakis.
Responsibility for the paper remains with the author.
Page 1
Print
Go Back
Next Page
Menu
Improving economic evaluation of urban transport projects in Australia
There will generally be no need to use an integrated network model where the effect of a
project on travel demand is small or localised. However, network-based travel demand
models are needed to identify the impacts of larger and/or more complex project and policy
initiatives. That is, by definition, the need to use a demand travel model implies that
significant travel change is expected.
Economic evaluations compare a Project Case with a Base Case. Current levels of
congestion suggest a likely difficulty in accommodating future unconstrained travel demand
on an existing (or even ‘do minimum’) Base Case. This requires that some future travel in the
Base Case change location, shift to another time or be suppressed, and thus increases the
likelihood that a future trip matrix for the Base Case and Project Case will differ.
It appears that demand modelling of urban transport proposals for Australian cities is
generally based on the number of trips generated and attracted for each zone being the
same for the time period being modelled for the Base Case and the Project Case (ie a fixed
trip matrix). Practice usually allows for alternative travel modes within this general constraint
of fixed origins and destinations. Some effects of this situation are:
Mode choice impacts may be partially taken into account, but not to the full extent if the
project or policy is also likely to influence the location or time of travel.
People can be expected to change the location of some of their trips in response to an
improvement in accessibility that results from a proposed project, and potentially all of
their trips if they make more fundamental changes such as moving to another place.
Holding origins and destinations fixed does not permit people to optimise their use of
the changed transport situation. This will lead to under-estimation of user benefits (eg
see Section 3.3 for an example).
The effect of capacity constraints in the Base Case is not taken into account. This will
generally result in over-estimation of project benefits due to excessive travel demand in
the Base Case relative to the available network capacity (or excessively high travel
costs if all demand is accommodated on a constrained network).
An increase in transport supply resulting from a project may allow some people who
changed the time of their travel due to capacity constraints in the Base Case to instead
journey at their preferred time with the project. These people gain a benefit from the
change. However, if they shifted to the peak they will increase travel time for continuing
peak users, though people who continue to travel in the off-peak time period gain some
benefit from reduced traffic congestion. With congestion in the Base Case and an
absence of marginal social cost pricing for travel, a fixed trip matrix approach is likely to
over-estimate user benefits by failing to take account of the external costs of temporal
changes in travel. Modelling a wider period about the peak that assumes temporal
diversion occurs within the period cannot take account of costs and benefits that result
from the changes in the time of travel.
A project that improves accessibility can be expected to result in some generated travel
relative to the Base Case. Ignoring generated travel will, in the absence of prices set
equal to marginal social cost, result in over-estimation of user benefits, as in the
previous example. (Bray and Sayeg (2002) illustrates the effect of generated road
traffic on project benefits in the case of vacated road capacity resulting from diversion
of car drivers to Bangkok’s elevated rail line being filled to varying degrees with
generated traffic.)
While transport agencies in Australia may sometimes use trip matrices that vary between the
Base Case and the Project Case, economic evaluation methodologies for projects presented
in guidelines are usually appropriate only for situations where the trip matrix for any given
Page 2
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
model year is the same for the Base Case and the Project Case (eg RTA 2004 and Main
Roads 1999). While the potential for variable trip matrices is noted in these references (and
also Austroads 2004a), the implications for evaluation methodology is not clearly addressed.
Australia is not alone in the general use of fixed trip matrices. While not seeking to provide a
comprehensive review of international practice, some features of UK practice are noted. The
standard practice for cost benefit analysis introduced in the United Kingdom in 1980
determined a limited set of circumstances where it was considered that variable trip matrices
should be used (DOT 1980). In recent years, the Department for Transport has enhanced its
appraisal methodology, including introduction in 2001 of new evaluation software that allows
multi-modal evaluation with variable trip matrices. The new model is called Transport User
Benefit Appraisal (TUBA) – see DfT 2004a for an overview of the model. Nevertheless,
limited guidance is provided on use of variable trip matrices, other than with regard to
indicating their use for multi-modal projects.
The Scottish Transport Appraisal Guidance (Scottish Executive 2003:B17) provides more,
though still methodologically ambiguous, direction. It indicates that “generally a variable trip
matrix assessment is appropriate for this type of scheme [see below], however, a fixed trip
matrix assessment should be undertaken for comparison purposes”, where the schemes for
which a variable trip matrix should be used is defined by:
“Are the existing roads in the study area operating close to capacity, or are they
expected to do so within the design life of the scheme? In these circumstances,
congestion is likely to lead to suppression of traffic effects, and schemes may result in
the release of some of the suppressed traffic.
Is the potential change in overall traffic flows high with respect to changes in travel
times or costs? This is likely to be the case where there are good alternatives available
for the movements affected by the proposed scheme, e.g. other routes or public
transport alternatives.
Will the implementation of the proposed scheme cause large changes in travel costs,
road capacity or both? These conditions are likely to occur where the scheme or
improvement bypasses extended lengths of low standard or congested network, or
where new road links or public transport systems cause major changes in accessibility
(e.g. estuarial crossings, LRT network).”
The Standing Advisory Committee on Trunk Road Assessment has been less ambiguous,
recommending that “variable matrix economic evaluations are undertaken for schemes as
the cornerstone of the economic appraisal in every case, except where it can be shown that
the trip matrix will not vary as a result of the scheme being appraised” (SACTRA
1994:15.24).
DOT (1980s) suggested that “In most cases the variable trip evaluation of benefits is unlikely
to yield more than about 10 percent extra benefits over the fixed trip evaluation, although this
will be scheme specific.” Williams and Moore (1990) used a simplified, uni-modal equilibrium
model to examine the effect of fixed and variable trip matrices on benefits. They found
circumstances in which use of fixed trip matrices could result in either severe over or under
estimation of benefits. In particular, they noted that “substantial over-estimation will tend to
be associated with congested conditions and moderate (to large) traveller response.
Predicably, such conditions will be found in or near urban areas.”
A generalisation of the analysis by Williams and Moore found that “Both the equilibrium
demand forecasts and the benefits measures have been shown to be sensitive to a range of
parameters and in particular the elasticity of demand. We have demonstrated that application
of inelastic, or ‘fixed demand’, methods which are commonly applied in highway appraisal
Page 3
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
might result in significant overestimation both of road traffic and the benefits of policies if
there is even a small propensity to respond to cost changes.” (Williams and Yamashita
1992:281).
With regard to peak shifting, Henderson (1992) notes that “ignoring peak shifting in cost-
benefit analysis leads to an overestimate of benefits and excessive capacity investment.”
It is concluded that use of a fixed trip matrix approach for many urban transport projects in
Australia is inconsistent with government objectives for urban transport, with future
circumstances likely to exist in Base Cases, and with the likely impacts of the types of
projects and policies that require use of a travel demand model. In these circumstances, a
fixed trip matrix will not provide data that accurately indicates the benefits of projects. Finally,
the direction and degree of the inaccuracy will not be uniform, and will depend on current
travel conditions and the extent and nature of generated and diverted travel demand, but the
general likelihood is that it leads to overestimation of benefits.
3 Deriving user benefits
3.1 Historic literature
The literature on the estimation of user benefits goes back over three decades. Two
essential references are Neuberger (1971) and McIntosh and Quarmby (1972). McIntosh and
Quarmby provide the general case for estimating user-related benefits, which comprise:
A. increase in user surplus: 0.5 * Σ
ijkt
(
kt
T
1
ij
+
kt
T
2
ij
) * (
kt
PC
1
ij
kt
PC
2
ij
) plus
B. increase in perceived user costs: Σ
ijkt
[(
kt
T
2
ij
*
kt
PC
2
ij
) - (
kt
T
1
ij
*
kt
PC
1
ij
)] less
C. increase in resource costs: Σ
ijkt
[(
kt
T
2
ij
*
kt
RC
2
ij
) - (
kt
T
1
ij
*
kt
RC
1
ij
)]
where:
kt
T
ij
= number of trips from zone i to zone j by mode k during time period t;
kt
PC
ij
= perceived generalised (ie behavioural) cost per trip from zone i to zone j by
mode k during time period t;
kt
RC
ij
= resource cost (ie excluding taxes that are transfer payments) per trip from
zone i to zone j by mode k during time period t; and
superscript 1 represents the Base Case and superscript 2 the Project Case.
As indicated in Table 1, component A is commonly known as the change in consumer
surplus, and the net effect of components B and C as the resource correction. Component B
reflects the increased willingness of users to pay (WTP) for the additional travel that they
undertake, though it needs to be offset against the actual change in travel cost indicated by
component C. This articulation of benefits differs from that presented in recent national
evaluation guidelines in which the combination of components A and B are described as the
“increase in WTP” and component C is the change in “total social cost” (ATC 2004: Vol 2, pp
49-50). While the terminology differs, both the ATC and McIntosh and Quarmby articulations
result in the same benefit.
Subsequent rows in Table 1 relate to the measurement of the benefits, and are discussed
later in this paper. The three components that make up the net user-related benefit are
shown diagrammatically in Figure 1.
Page 4
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Table 1 Features of components of user-related benefit
Component (as per McIntosh and Quarmby 1972)
A
(change in user surplus)
B
(change in user costs)
C (change
in resource cost)
Nature of benefit Change in the value of
travel in excess of its
perceived cost
Benefit as indicated by
demonstrated willingness to
pay for additional travel
Change in the real
(ie resource) cost
of travel
Common term Change in consumer
surplus
Net effect is often called a ‘resource correction’, ie an
allowance for resource costs not perceived by travellers
Based on Perceived costs, which are in turn related
to market prices
Resource costs
Derived using Number of trips and perceived cost
for each origin-destination pair
Sum of costs (eg VOCs) for
each link in the network
Situation with a
variable trip matrix
Takes account of both changes in the number of trips
and perceived trip cost for each origin-destination pair
Change in resource costs
estimated from link data
Situation with fixed
trip matrix
Value of components A and B offset each other
with a fixed trip matrix
User benefit is equal to the
change in resource costs
Figure 1 User-related benefits
Cost
with
Pro
j
ect
Demand
without
pro
j
ect
T
1
T
2
PC
1
PC
2
(A) Change in user surplus =
shaded area
Quantity
Cost
with
Pro
j
ect
Demand
without
pro
j
ect
T
1
T
2
PC
1
PC
2
(B) Incr. in perceived user
cost =
-
Quantity
Cost
with
Pro
j
ect
Demand
without
pro
j
ect
T
1
T
2
PC
1
PC
2
(A+ B) Incr. in perceived user
benefit = shaded area
Quantity
Cost
with
Pro
j
ect
Demand
without
pro
j
ect
T
1
T
2
Quantity
RC
1
RC
2
(C) Incr. in resource costs
=
-
Page 5
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Part of the difference between resource and behavioural costs is indirect taxation, which
finds its way into project benefits via its inclusion in perceived costs. However, taxation is a
transfer payment, and there will be a compensating effect elsewhere in the economy.
Accordingly, there is a need to take account of this effect of indirect taxation only to the
extent that taxation rates in the transport sector differ from those applying elsewhere in the
economy. This may occur, for example, as a result of GST exemptions and fuel excise. While
no reported analysis is evident, the difference may be sufficiently small that the effect can be
ignored.
It is notable that this general case provided by McIntosh and Quarmby (1972) applies both
where demand is fixed and variable. Neuberger (1971) describes a “Method 1” for the
specific case in which the trip matrix is fixed and a “Method 2” where this is not the case.
Under Method 1, the benefit will equal the change in resource costs, ie component C of the
formula above – this occurs because the change in user benefits is equal to the change in
perceived user costs but opposite in sign. Neuberger provides useful discussion of ways in
which user benefits may be miscalculated.
Neuberger (1971) also presents a Method 3 that avoids the implicit assumption in his Method
2 (and also that of McIntosh and Quarmby) that the demand curve is a straight line between
the levels of travel demand being considered. His Method 3 is also necessary for the
evaluation of land use plans. The issue of evaluating land use plans is addressed in more
detail in Neuberger and Wilcox (1976) and, for example in more recent times, Bates (2003).
3.2 The implications of variable trip matrices for benefit estimation
Changes in perceived travel costs from the Base Case to the Project Case, and
consequences of the changes for travel demand, must be determined on an origin-
destination basis. That is, the benefit to travellers for any given time, and mode (ie T
2
ij
- T
1
ij
)
is a function of travel conditions between the Base Case and the Project Case for each ij
zone pair. It is therefore not a function of any particular link (or set of links that can be
uniquely identified from aggregate link information).
By contrast, resource costs (eg resource vehicle operating costs - VOCs) are independent of
the origin and destination of a trip. They are a function of the length and travel conditions on
each link over which vehicles travel. It is therefore possible to estimate the change in
resource costs by summing costs for each link in the Base Case and in the Project Case and
to thus obtain the net benefit.
If the trip matrix is the same in the Base Case and Project Case in every respect, user-
related benefits will be equal to the difference in the resource cost of travel between the two
cases. It is in this case only that link based data alone can be used to determine user
benefits. This outcome occurs because any perceived user benefit will be offset by a
decrease in user costs (ie component A exactly offsets component B), leaving reduced
resource costs (component C) as the sole benefit. This is Neuberger’s Method 1. In
explaining the situation, Neuberger (1971:54) notes that it “arises from the fact that demand
is unaffected by cost changes. Thus it should not matter how users perceive costs, since
they do nothing about it.”
If it is accepted that use of the same trip matrix for the Base Case and the Project Case for a
given time period and mode for a particular model year is inappropriate for many urban
transport projects, link data cannot be the sole basis for user benefits. Rather, perceived
user benefits need to be estimated for each origin-destination pair. Implementation of this
approach would represent a substantial change in current Australian practice.
Page 6
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Page 7
3.3 An example
Figure 1 shows the situation with three zones and a single link (or weighted average of
multiple links) between each zone. A set of cases examines the potential effect of various
demand responses to an improvement in the link between zones A and B. A final example
illustrates the effect of a network wide change. It can be seen that:
Project Case 1a shows that, with a fixed trip matrix, the benefit is simply the change in
resource costs between the Base Case and the Project Case.
Project Case 1b shows a situation in which the total number of trips remains
unchanged but the distribution of them between zones changes (in Project Case 1b 30
trips made each way between zones A and C in the Base Case are made instead
between Zones A and B in the Project Case to take advantage of the improved link
between the latter pair of zones. It shows that user benefits increase as people change
their travel pattern to take advantage of the improved link (from 7,975 units in with a
fixed trip matrix in Project Case 1a to 8,422 units with a somewhat variable trip matrix
in Project Case 1b).
Project Case 1c tests the effect of no shift in travel between Zones A and C, but allows
for trip generation in response to the improved link between Zones A and B. It is
notable that the user benefit of 7,237 units is less than for Project Case 1a (ie the
generated trips result in a net disbenefit).
Project Case 1d shows a more fully variable trip matrix by combining both of the two
previous cases, ie the improved link between Zones A and B both results in a land use
change (ie a change in the distribution of trips) and in the generation of additional
travel. The user benefit in this case is 7,683 units, which is less than would have been
the case if the project has been modelled with a fixed trip matrix (ie Project Case 1a).
Even with a uniform change in the cost of travel across the network (Project Case 2),
there is a need to derive benefits on a trip basis.
As the extent of the error is not uniform or proportional, it is not possible to use a simple
approach using aggregate network data and make a correction to obtain the right answer.
Factors such as the elasticity of demand and the scale and location of the project’s impact
affect the extent of the error.
More detailed analysis indicates that an error still occurs if aggregated perceived benefits are
used, eg the number of trips in the Base Case multiplied by the average unit perceived
benefit, plus generated trips multiplied by half the unit perceived benefit, plus account for
changes in resource costs. The data shown can be used to illustrate other incorrect methods
for estimating benefits, eg change in the total perceived cost of travel with or without a
resource correction.
Hence, it should be evident that any diversion and generation of travel demand that might
occur in response to a project or policy needs to be taken into account if user benefits are to
be reasonably estimated, and where either or both of these effects occur, user benefits must
be derived on an origin-destination basis to be accurately calculated.
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Page 8
BC
BC
BC
BC
BC
BC
A
Figure 1 Estimating project benefits with fixed and variable trip matrices
300 600 700
400 800
500
A
300 600 700
A
300 630 670
A
300 630 700
A
300 660 670
A
292.5 585 682.5
B
600
B
600 400 800
B
630 400 800
B
630 400 800
B
660 400 800
B
585 390 780
C
700 800
C
700 800 500
C
670 800 500
C
700 800 500
C
670 800 500
C
682.5 780 487.5
BC
BC
BC
BC
BC
BC
A A
10 45 60
A
10 45 60
A
10 45 60
A
10 45 60
A
10.5 52.5 63
B
50
B
45 10 70
B
45 10 70
B
45 10 70
B
45 10 70
B
52.5 10.5 73.5
C
60 70
C
60 70 10
C
60 70 10
C
60 70 10
C
60 70 10
C
63 73.5 10.5
Input data Input data Input data Input data Input data
Change in perc. cost of A-B trip Change in perc. cost of A-B trip -10% Change in perc. cost of A-B trip -10% Change in perc. cost of A-B trip -10% Change in cost of travel
Resource cost (% of perc. cost) Elasticity of demand with respe
10 50 60
10 70
10
-10% 5%
133%
c
-0.5
Elasticity of demand with respe
c
-0.5 Elasticity of demand with respe
c
-0.5 Elasticity of demand with respe
c
-0.5
133%
D
Resource cost (% of perc. cost) Resource cost (% of perc. cost) 133% Resource cost (% of perc. cost) 133% Resource cost (% of perc. cost) 133%
A
gg
re
g
ate outcomes
(
Base Case
)
:A
gg
re
g
ate
(
Pro
j
ect Case 1a
)
:
Agg
re
g
ate
(
Pro
j
ect Case 1b
)
:
Agg
re
g
ate
(
Pro
j
ect Case 1c
)
:
Agg
re
g
ate
(
Pro
j
ect Case 1d
)
:
Agg
re
g
ate
(
Pro
j
ect Case 2
)
:
Number of trips 5,400 Number of trips 5,400 Number of trips 5,400 Number of trips 5,460 Number of trips 5,460 Number of trips 5,265
Total perceived cost of travel 268,000 Total perceived cost of travel 262,000 Total perceived cost of travel 261,100 Total perceived cost of travel 264,700 Total perceived cost of travel 263,800 Total perceived cost of travel 274,365
Average trip cost 49.63 Average perceived trip cost 48.52 Average perceived trip cost 48.35 Average perceived trip cost 48.48 Average perceived trip cost 48.32 Average perceived trip cost 52.11
Correct estimate of net benefits Correct estimate of net benefits Correct estimate of net benefits Correct estimate of net benefits Correct estimate of net benefits
McIntosh and Quarmby (component A) --> 6,000 6,150 6,150 6,300 -13,233
McIntosh and Quarmby (component B) -->
-6,000 -6,900 -3,300 -4,200 6,365
McIntosh and Quarmby (component C) -->
-7,975 -9,172 -4,387 -5,583 8,461
Total benefit 7,975 Total benefit 8,422 Total benefit 7,237 Total benefit 7,683 Total benefit -15,328
Derivation of perceived benefits Derivation of perceived benefits Derivation of perceived benefits Derivation of perceived benefits Derivation of perceived benefits
Total
perceiv-ed
cost
Differ-
ence from
BC
Total
perceiv-ed
cost
Differ-
ence from
BC
Total
perceiv-ed
cost
Differ-
ence from
BC
Total
perceiv-ed
cost
Differ-
ence from
BC
Total
perceiv-ed
cost
Differ-
ence from
BC
A-A 3,000 A-A 3,000 0 0 A-A 3,000 0 0 A-A 3,000 0 0 A-A 3,000 0 0 A-A 3,071 71 -148
A-B 30,000 A-B 27,000 -3,000 3,000 A-B 28,350 -1,650 3,075 A-B 28,350 -1,650 3,075 A-B 29,700 -300 3,150 A-B 30,713 713 -1,481
A-C 42,000 A-C 42,000 0 0 A-C 40,200 -1,800 0 A-C 42,000 0 0 A-C 40,200 -1,800 0 A-C 42,998 998 -2,074
B-A 30,000 B-A 27,000 -3,000 3,000 B-A 28,350 -1,650 3,075 B-A 28,350 -1,650 3,075 B-A 29,700 -300 3,150 B-A 30,713 713 -1,481
B-B 4,000 B-B 4,000 0 0 B-B 4,000 0 0 B-B 4,000 0 0 B-B 4,000 0 0 B-B 4,095 95 -198
B-C 56,000 B-C 56,000 0 0 B-C 56,000 0 0 B-C 56,000 0 0 B-C 56,000 0 0 B-C 57,330 1,330 -2,765
C-A 42,000 C-A 42,000 0 0 C-A 40,200 -1,800 0 C-A 42,000 0 0 C-A 40,200 -1,800 0 C-A 42,998 998 -2,074
C-B 56,000 C-B 56,000 0 0 C-B 56,000 0 0 C-B 56,000 0 0 C-B 56,000 0 0 C-B 57,330 1,330 -2,765
C-C 5,000 C-C 5,000 0 0 C-C 5,000 0 0 C-C 5,000 0 0 C-C 5,000 0 0 C-C 5,119 119 -247
Total
268,000
Total
262,000 -6,000 6,000
Total
261,100 -6,900 6,150
Total
264,700 -3,300 6,150
Total
263,800 -4,200 6,300
Total
274,365 6,365 -13,233
Notes: -->
Change in
con-sumer
surplus
Change in
con-sumer
surplus
Perc. travel cost Change in
con-sumer
surplus
Trips Trips (change A-B trips according to
elasticity but leave other trips
unchanged)
Trips (no change from Base Case) Trips (change number of trips
according to elasticity)
Trips (no change in total trips, but
redistribute trips to improved link)
Trips (change A-B trips according to
elasticity)
From zone
To zone
To zone
To zone
From zone
To zone
Perceived cost of travel (-10%
change in A-B trip cost)
From zone
Perceived cost of travel (-10%
change in A-B trip cost)
From zone
Perceived cost of travel
To zone
From zone
To zone
From zone
Perceived cost of travel (-10%
change in A-B trip cost)
Benefit increases compared with option 1c
because of trip diversion. But benefit is less
than for option 1b when resource cost of
generated trips is more than the perceived
cost.
An increase in travel costs results in a total
benefit that is greater than the consumer
surplus if resource costs are greter than
perceived costs.
Perceived cost of travel (-10%
change in A-B trip cost)
To zone
From zone
Perceived cost of travel (5% increase
in unit travel costs)
To zone
From zone
From zone
Meets Neuberger Method 1 conditions.
M&Q equations A and B cancel out each
other and are hence redundant, ie meets
Neuberger comment that no need to know
perceived costs in Method 1.
Benefit increases because more trips take
advantage of improved A-B link.
Benefit decreases compared with previous
case (option 1b) when resource cost is
greater than perceived cost.
Change in
con-sumer
surplus
Change in
con-sumer
surplus
Perc. travel cost Perc. travel cost Perc. travel costO-D
pair
Project Case 2 (with uniform increase in
cost of travel, eg change in fuel tax)
Enter resource/perceived
proportion, eg 125% (or use D
for default value of 133%) --->
Change in user benefits (ie
incr. in consumer surplus)
Change in perceived user costs
To zone
From zone
To zone
From zone
Project Case 1d (improve A-B route,
with additional trips on A-B from
generation and diversion)
To zone
Base Case Project Case 1a (improve A-B route, no
change in trips)
Project Case 1b (improve A-B route,
same no. of network trips but different
distribution)
Project Case 1c (improve A-B route, with
additional trips on A-B from generation)
To zone
Change in user benefits (ie
incr. in consumer surplus)
Change in user benefits (ie
incr. in consumer surplus)
From zone
Change in perceived user costs
Change in resource cost
Change in user benefits (ie
incr. in consumer surplus)
Change in perceived user costs
Change in resource cost
O-D
pair
Change in perceived user costs
Change in resource cost
Change in user benefits (ie
incr. in consumer surplus)
Change in perceived user costs
Change in resource cost Change in resource cost
Perc. travel costO-D
pair
O-D
pair
O-D
pair
O-D
pair
Total
perceiv-
ed travel
cost
Note: red cells contain variables that
can be altered
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
3.4 Implications for travel demand models
The previous discussion on the use of variable trip matrices and the associated calculation of
user benefits places particular needs on transport models. These are discussed below.
Obtaining necessary evaluation data from models
Data needed from travel demand models to properly estimate user benefits comprises: (a)
for the Base Case and the Project Case, the number of trips and average perceived cost of
travel for each origin-destination pair; and (b) the resource cost of travel on each link in the
network.
Given the large number of origin destination pairs in most models, it is preferable that the
change in perceived user surplus and perceived user costs be estimated within the model.
This should not be difficult because the necessary data on user costs for each origin-
destination pair is used in the internal processing of the model and user surplus will be based
on the combination of assigned demand and these user costs.
It is unlikely that the transport model will have the necessary data to estimate the resource
cost of travel on each link in the network as the data is not needed for any other model
functions. Data on link characteristics (ie physical features of the link and traffic using the
link) therefore need to be transferred from the model into a spreadsheet or similar associated
model and used in conjunction with unit resource costs to estimate the value of Component
C as indicated by McIntosh and Quarmby.
A possible short-cut
Deriving benefits for each origin-destination pair can be onerous. Neuberger (1971) notes a
practical short-cut for use with variable trip matrices that allows link based rather than origin-
destination based data to be used. The method loads both the Base Case and Project Case
trip matrices to each of the Base Case and Project Case networks (sometimes called “cross-
loading”). The approach is also described in Transfund (2004). In this approach, the change
in user surplus is derived as:
0.5 * [ Σ
l
(V
1
l
* PC
1
l
+ V
2
l
* PC
1
l
) – Σ
m
(V
1
m
* PC
2
m
+ V
2
m
* PC
2
m
) ]
where:
V = Traffic volume on a link
PC = Perceived cost of travel on a link
1 = Base Case network, and 2 = Project Case network
l = each link in the Base Case network, and m = each link in the Project Case
network
Model equilibrium
Using demand and perceived travel cost for each origin-destination pair requires that
demand for travel for each pair be matched to the related cost of travel. Two issues arise.
Travel demand models usually load demand in a number of steps (so that demand in a
subsequent step will choose routes that take account of the presence of other traffic). Thus,
the generalized cost of travel between each origin destination pair in one iteration of the
model is used to assign demand for travel between the same pair to particular links in the
Page 9
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
next iteration. As a result, a single model run contains mismatched demand and travel costs.
This should not be a problem if the iterations eventually converge.
The second issue is that the assignment process will result in final demand for each origin
destination pair using multiple routes. Hence, there is a need to derive the average travel
cost for travel between each ij pair taking account of the various routes used. Effective
evaluation using variable trip matrices requires a model that can provide the origin-
destination information needed for economic evaluation, and application of the model to
ensure equilibrium is achieved.
Model time periods
Modelling for a given time period cannot ascribe any benefits or costs that result if trips
change their travel schedule within the time period in response to the project or policy. It is
therefore necessary to ensure that projects that may affect temporal choices are modelled so
that these changes are taken into account.
4 Other Issues
Some other issues related to sound project evaluation are considered below. The list is not
exhaustive, but raises some issues considered worthy of debate and clarification.
4.1 Perceived and resource costs
It is likely that perceived costs used in models and unit resource costs used for economic
evaluations will be derived independently. As both are required to derive user benefits with a
variable trip matrix, it is essential that they be based on prices at the same date.
4.2 Unit of account
The conventional unit of account that underlies transport evaluation in Australia is social
costs valued at factor prices. However, implicit to demand modelling and associated project
evaluation is the use of perceived costs. Perceived costs are related to market prices
because people can be expected to base their decisions on the face value of prices they
encounter. For example, they will perceive the market price of fuel, not its untaxed or some
other value. Even travel time is derived relative to the value to some other thing that has a
financial (ie market) value. Hence, there is merit in the underlying basis for valuation of
benefits being willingness-to-pay based on market prices adjusted to reflect differences from
resource values.
This is the approach that has been adopted by the UK Department for Transport, which
describes it as the “willingness-to-pay calculus” (DfT 2003 and DfT 2004b). Its principal
advantage is that it starts with market prices, which are arguably more intuitive than
synthesised resource prices. More importantly, it allows the distribution of benefits between
those affected by a project to be more clearly identified. The choice of the commencing unit
of account does not affect the final result of an evaluation.
It also accords with derivation of benefits as described by McIntosh and Quarmby earlier,
wherein components A and B are based on perceived prices, ie market-related prices.
4.3 Disaggregation of benefits
An underlying assumption of economic evaluation is that a unit value of resource has the
same value irrespective of whether a person gains more or less of it, ie a reduction of say
five minutes of travel time has the same value as a five minute increase in travel time. A
second assumption is the potential compensation principle: that a potential welfare
improvement is achieved if the beneficiaries of a project could compensate those who lose to
Page 10
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
the extent of their loss and still be better off, ie it is net benefits that matters. It is not
necessary that this compensation be paid - that is a matter of equity rather than whether
society is, as a whole, better off. If compensation is not paid, as is generally the case other
than with respect to the purchase of land, a project will have distributional impacts.
It may be judged that politicians sense that neither of these assumptions is fully realistic – for
it is usual for those who will be disadvantaged by a project to be more vociferous than those
who will benefit. In other words, there is a need to understand distributional impacts.
The issue of weighting different components of costs and benefits to reflect distributional
issues was debated in the early 1970s (eg Little and Mirrlees 1974), but has generally been
found to be difficult to implement and thus not used in practice. Sugden and Williams
(1978:131) note that “in most cases dealt with in cost-benefit analysis, the actual magnitude
of income effects is likely to be small in relation to the margins of error present in all the
information used by the analyst.” A practical approach is to separately identify those who
gain and lose from a project, and the scale of their respective benefit and disbenefit, to aid a
better understanding the likely concerns of the community and potential remedial measures if
distributional issues are substantial. Some travel demand models allow benefits and
disbenefits to be separately identified.
The UK Department for Transport notes difficulties in attributing benefits to groups of users
where, for example, users change their mode of travel (DfT 2004c). It recommends a
methodology developed by Sugden (1999) to attribute user benefits to particular modes of
transport. This subject, amongst others, is also discussed in Bates (2003).
4.4 Quality of resource cost data
The quality of economic evaluations is dependent on reliable unit cost parameters (both
perceived and resource values). The most recently released general guidance on unit
resource costs for use in transport projects in Australia pertain to prices in June 2002 or
about that time (Austroads 2004b). However, the presentation of the data does not allow
users to update VOCs to take account of inflation other than by applying an average price
change index. In addition, the approach does not make fully clear the variables assumed to
influence VOCs, eg the road surface condition, terrain, congestion and vehicle deterioration,
and is constrained by the limited number of road and vehicle categories and traffic conditions
for which costs are derived.
The role of VOCs and other unit costs is to allow estimation of the value of user benefits over
the duration of the evaluation period. The current approach implies that prices that existed in
about June 2002 are representative of future prices and will be unchanged over time.
However, for example, the substantial fluctuation in the resource cost of fuel that has
occurred in the past suggests it would only be by chance that the cost at any point in time
was an average for the past or, more importantly, was a reasonable estimate of its future
value. The trend of improved fuel efficiency of vehicles and changes in vehicle size and
technology suggests that a historic average fuel consumption will again only by chance
represent future fuel consumption. This suggests a need for further consideration of
appropriate values for such parameters. Expressing values to be used in evaluations in
probabilistic terms would facilitate risk analysis (see Section 4.6).
In addition, a VOC model needs to include the range of link variables (ie road and traffic
conditions) that affect VOCs and should be capable of being applied to each link (including
intersections) in a transport model to accurately determine resource costs (ie Component C
described by McIntosh and Quarmby). This is best done with an integrated model such as
TRAMS (Austroads 2005) but can also be done externally by exporting data from the
demand model into a separate evaluation model (Rust PPK et al 1996).
Page 11
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
4.5 Definition of the Base Case
There is generally a focus on the Project Case in project evaluation studies. However, the
result of an economic evaluation is the difference between the Base Case and Project Case,
and hence is affected as much by definition of the Base Case as the proposed project.
In the past, the Base Case might reasonably have been a “do nothing” or “do minimum”
situation. However, with urban road networks already congested, such a Base Case will
result in travel costs that go up exponentially as traffic rises towards and beyond capacity. A
Base Case with excessively high and unrealistic travel costs may allow an evaluation to
show almost any project to be worthwhile – which could convey a flawed impression. That is,
an unrealistic Base Case will result in a misleading project evaluation.
Providing substantial additional capacity in the Base Case to accommodate rising future
demand is undesirable because it means that the evaluation considers only the incremental
cost of the project compared with the investment needed in the Base Case. It might be
assumed that if the incremental investment is worthwhile, the initial tranche of expenditure
was even more cost-effective, on the basis of a declining return from additional levels of
expenditure. This is a substantial assumption that is better avoided.
A better alternative may be to accommodate future traffic growth in the Base Case to the
extent that is realistic through changes in travel behaviour, and to assume that additional
travel is suppressed. These constraints on travel behaviour would be released by the
proposed project, at least to some extent. By definition, a change in quantity, time, location
and mode of travel in the Project Case relative to the Base Case represent a variable trip
matrix (and thus requires use of an origin-destination based approach to the estimation of
user benefits).
4.6 Sensitivity testing
All values in an economic evaluation are estimates, if not conjecture. An evaluation result
that is based on the best estimate of each input variable provides no guidance on the extent
of uncertainty and hence the level of confidence in the result.
Sensitivity testing, which is a form of risk analysis, is generally applied only simply in public
sector transport projects in Australia. Individual tests are generally made of the effect of
variations in input parameters in the evaluation, though these have little meaning without an
indication of the likelihood of their occurrence. The somewhat better methodology of applying
a probability to each of the variations and determining a weighted result (as described in
Austroads 1996) appears to be little used, and still represents a simple approach.
Recent guidelines (ATC 2004) describe four forms of risk analysis: the above two (ie
alternative point estimates and a weighted average of various point estimates); an adjusted
benefit-cost analysis (which factors the monetary value of individual components of cost or
benefit according to their perceived importance); and use of a probability distribution for input
values rather than point values. The first two approaches provide only a limited
understanding to decision-makers of the confidence that should be given to evaluation
results. The guidelines do not represent the adjusted benefit-cost analysis as a sensitivity
test, but its effect is the same, ie to illustrate the impact on the evaluation results of an
alternative set of judgements. However, the approach is not guided by a clear analytical
approach, and is thus open to arbitrariness. It is not considered by the guidelines to be an
essential form of analysis.
The methodology for undertaking the last of the approaches was detailed in Pouliquen
(1970). Software to make the approach computationally practicable was introduced in the
mid-1980s, and has been taken up by, for example, business schools and major companies.
The methodology has also been noted in other reports in Australia (eg RTA 2004 and
Page 12
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Austroads 2002). Yet it appears to have been rarely used in public sector transport projects.
Two reasons have generally been given to the author in the past for limited use of the
technique. Firstly, it adds complexity to decision making – for example, a result that a project
has a ‘95 percent probability of having a benefit cost ratio of more than 1’ may be considered
to be more difficult to interpret that one that simply states the project has a ‘benefit cost ratio
of 1.5.’ Overcoming this difficulty seems to be a matter of experience, and should not be
insurmountable. It can also be seen as a cost of providing sounder advice.
The second concern is determining probability distributions for input variables. However,
work such as that presented by UMTA (1989), Flyvbjerg et al (2002, 2004, 2005), time series
data on fuel prices, the range of estimates of the unit value of environmental externalities,
and the considerable data held by transport agencies on cost estimates at various stages of
project preparation and at construction, and on transport demand forecasts and outcomes,
provide a considerable database on which authorities could develop appropriate distributions
of parameter values.
Finally, the usefulness of risk analysis that is prepared as part of project appraisal will be
considerably enhanced if it is used to identify actions that need to be taken during ongoing
project preparation and implementation to ensure project objectives are achieved.
5 Conclusions
The discussion in this paper suggests that:
Use of a fixed trip matrix for modelling urban travel demand is inconsistent with
government objectives for transport, future conditions to be expected in the Base Case,
and the likely impact of large and/or complex projects and policies. Hence, there is a
general case to be made for use of variable trip matrices for any project or policy that
needs to be subject to a computerised travel demand model to determine their likely
impact.
Link-based data from a demand model can be the sole basis for deriving user benefits
only with a fixed trip matrix. In other cases, perceived user benefits need to be
estimated for each origin-destination pair rather than based on link data.
In these circumstances, a fixed trip matrix will not provide data that accurately indicates
the likely benefits of projects and policies being tested. The direction and degree of the
inaccuracy resulting from use of a fixed trip matrix is not uniform, and will depend on
the extent and nature of the changes in travel demand.
This suggests that if it is necessary to use a transport model to determine travel
demand effects, it is likely that a variable trip matrix should be used and perceived user
benefits derived on an origin-destination basis. This represents a substantial change in
current Australian practice.
The use of a variable trip matrix and the associated requirement for the calculation of
user benefits on an origin-destination basis places particular requirements on transport
modelling. It is necessary to ensure that the necessary data can be obtained from the
models; that assignment iterations lead to consistency between perceived travel costs
and travel assignment; consideration is given to sub-divisions of the day for which
modelling is undertaken; and consideration is given to testing the reliability of the
cross-loading short-cut that allows link-based data to be used to determine the
reliability with which it can be used.
Page 13
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Some other related matters for which practice could be improved include the unit of
account used in valuing benefits, disaggregation of benefits, quality of unit resource
cost data, definition of the Base Case and sensitivity testing.
These conclusions suggest a need for further deliberation and debate regarding the
appropriate methodologies for the planning, modelling and evaluation of urban transport
projects, and possible change to current practices.
6 References
Austroads (2002) “Economic Evaluation of Road Investment Proposals: Risk Analysis” AP-
R203, ARRB Transport Research Ltd
Austroads (2004a) “Economic Evaluation of Road Investment: Estimating Urban Inter-Modal
Benefits – A Literature Review." AP-R242, ARRB Transport Research Ltd
Austroads (2004b) “Economic Evaluation of Road Investment Proposals, Unit Values for
Road User Costs at June 2002”, AP-R241, ARRB Transport Research Ltd
Austroads (2005) “Guide to Project Evaluation Part 3: Models and Procedures”, AGPE03/05,
prepared by D. Tsolakis, S. Patrick and T. Thoresen
Australian Transport Council (2004) "National Guidelines for Transport System Management
in Australia" (3 volumes), prepared by the Guidelines Assessment Methodology Working
Group, Canberra
Bates, J. (2003) “Economic Evaluation and Transport Modelling: Theory and Practice”, in
10th International Conference on Travel Behaviour Research (10-15 August), Lucerne
Bray, D. J. and Sayeg, P. (2002) “Valuing externalities of Bangkok’s Mass Transit (Skytrain)
System”, Australasian Transport Research Forum, Canberra, October.
Bray, D. J. and Tisato, P. (1997) “Broadening the Debate on Road Pricing”, 21
st
Australasian
Transport Research Forum Proceedings (pp 599-616), Adelaide
Department of Transport (1980) “COBA 9 Cost Benefit Analysis Manual”, HMSO, London
Department for Transport (2003) “MSA: Cost Benefit Analysis TAG Unit 3.9.2”, UK, June
Department for Transport (2004a) "TUBA Guidance", prepared by Mott MacDonald, UK
Department for Transport (2004b) “Cost Benefit Analysis TAG Unit 3.5.4”, UK, April
Department for Transport (2004c) “Transport User Benefit Calculation TAG Unit 3.5.3”, UK,
April
Flyvbjerg, B. (2005) "Measuring inaccuracy in travel demand forecasting: methodological
considerations regarding ramp up and sampling." Transportation Research Part A: Policy
and Practice, 39(6), 522-530.
Flyvbjerg, B., Skramris Holm, M., and Buhl, S. L. (2005) "How (In)accurate Are Demand
Forecasts in Public Works Projects?" Journal of the American Planning Association, 71(2),
131-146.
Flyvbjerg, B., Skramris Holm, M., and Mette-Buhl, S. (2004) "What Causes Cost Overrun in
Transport Infrastructure Projects?" Transport Reviews, 24(1), 3-18.
Page 14
Next Page
Go Back
Menu
Improving economic evaluation of urban transport projects in Australia
Flyvbjerg, B., Skramris Holm, M., and Buhl, S. L. (2002) "Underestimating Costs in Public
Works Projects - Error or Lie?" Journal of the American Planning Association, 63(3), 279-
295.
Henderson, J. V. (1992) "Peak shifting and cost-benefit miscalculations", Regional Science
and Urban Economics, 22(1), 103-121
Little, I. M. D. and Mirrlees, J. A. (1974) “Project appraisal and planning for developing
countries”, Heinemann Educational Books, London
Main Roads (1999) “Cost benefit analysis manual for road infrastructure investment” 2nd
edition, Brisbane
McIntosh, P. T. and Quarmby D. A. (1972) “Generalised costs and the estimation of
movement costs and benefits in transport planning”, Highway Research Record, 383, 11-26
Neuberger, H. (1971) “User benefit in the evaluation of transport and land use plans”,
Journal of Transport Economics and Policy, vol 5, 52-75
Neuberger, H., and Wilcox, J. (1976) "The Economic Appraisal of Land-Use Plans" Journal
of Transport Economics and Policy, 10(3), 227-236
Pouliquen, L. Y. (1970) “Risk Analysis in Project Appraisal”, World Bank Staff Occasional
Paper No. 11, Washington DC.
Roads and Traffic Authority (2004) “Economic analysis manual”, Version 2, Sydney, April
RUST PPK in association with Symonds Travers Morgan and Highway Planning and Design
Institute (1996) “Study of Prioritisation of Highway Investments and Improving Feasibility
Study Methodologies [in China]”, prepared for the World Bank
Standing Advisory Committee on Trunk Road Assessment (1994) "Trunk Roads and the
Generation of Traffic", HMSO, London
Scottish Executive (2003) "Scottish Transport Appraisal Guidance", prepared with assistance
from Steer Davies Gleave, Atkins, SISA Limited, Ironside Farrar, TRL Limited and Derek
Halden Consultancy.
Sugden, R. (1999) “Review of cost/benefit analysis of transport projects”, UK Department for
Transport
Sugden, R. and Williams, A. (1978) “The principles of practical cost-benefit analysis”, Oxford
University Press
Transfund (2004) “Project evaluation manual”, PFM2, Amendment No. 8, October, New
Zealand
Urban Mass Transit Administration (1989) "Urban Rail Transit Projects: Forecast Versus
Actual Ridership and Costs", Department of Transportation, Washington DC.
Williams, H. C. W. L., and Moore L. A. R. (1990) "The Appraisal of Highway Investments
under Fixed and Variable Demand" Journal of Transport Economics and Policy, 24(1), 61-82.
Williams, H., and Yamashita, Y. (1992) "Travel Demand Forecasts and the Evaluation of
Highway Schemes under Congested Conditions" Journal of Transport Economics and Policy,
26(3), 261-282.
Page 15
Go Back
Menu
Chrome Web Store
It looks like you haven't installed the Fill Chrome Extension Add to Chrome