The demand for public transport: The effects of fares, quality of service,
income and car ownership
Neil Paulley a,
*, Richard Balcombe a
, Roger Mackett b
, Helena Titheridge b
, John Preston c
,
Mark Wardman d
, Jeremy Shires d
, Peter White e
a
TRL, Crowthorne House, Nine Mile Ride, Wokingham, Berkshire RG40 3GA, UK
b
Centre for Transport Studies, University College London, London, UK
c
Transportation Research Group, University of Southampton, Southampton, UK
d
Institute for Transport Studies, University of Leeds, Leeds, UK
e
Transport Studies Group, University of Westminster, London, UK
Received 24 November 2005; accepted 1 December 2005
Available online 18 April 2006
Abstract
This paper reports on key findings from a collaborative study whose objective was to produce an up-to-date guidance manual on the factors
affecting the demand for public transport for use by public transport operators and planning authorities, and for academics and other researchers.
Whilst a wide range of factors was examined in the study, the paper concentrates on the findings regarding the influence of fares, quality of service
and income and car ownership. The results are a distillation and synthesis of identified published and unpublished information on the factors
affecting public transport demand. The context is principally that of urban surface transport in Great Britain, but extensive use was made in the
study of international sources and examples.
q 2006 N. Paulley. Published by Elsevier Ltd. All rights reserved.
Keywords: Public transport; Transport demand; Fares; Quality of service; Income; Car ownership
1. Introduction
This paper reports on the key findings of a collaborative
study undertaken by the Universities of Leeds, Oxford and
Westminster, University College London and TRL Limited
(Balcombe et al., 2004). The objective of the study was to
produce an up-to-date guidance manual for use by public
transport operators and planning authorities, and for academics
and other researchers. The context of the study was principally
that of urban surface transport in Great Britain, but extensive
use was made of international sources and examples.
While a wide range of factors was examined in the study, the
findings relating to fares, quality of service and car ownership
are the most significant and this paper concentrates on these.
However, as Balcombe et al. (2004) make clear, in practice the
factors cannot be treated either in isolation from each other or
in isolation from many other direct and indirect influences on
public transport demand. The main study also considered new
transport modes such as guided busways, the relationship
between land use and public transport supply and demand, and
the impacts of transport policies generally on public transport.
It also looked at the influence of developments in transport and
technology over the past two decades, such as innovations in
pricing, changes in vehicle size, environmental controls on
emissions, and developments in ticketing and information
provision facilitated by advances in computing.
1.1. Background
In 1980, the then Transport and Road Research Laboratory,
now the Transport Research Laboratory (TRL), published a
collaborative report: the Demand for Public Transport
(Webster and Bly, 1980). This report, which became widely
known as ‘The Black Book’, identified many factors which
influence demand and where possible, given the limitations of
the data that were available for analysis, quantified their
effects. The Black Book subsequently proved to be of great
value to public transport operators and transport planners and
Transport Policy 13 (2006) 295–306
www.elsevier.com/locate/tranpol
0967-070X/$ - see front matter q 2006 N. Paulley. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.tranpol.2005.12.004
* Corresponding author. Tel.: C44 134 477 3131; fax: C44 134 477 0356.
E-mail address: npaulley@trl.co.uk (N. Paulley).
policy makers. However, in the following 20 years there has
been a great deal of change in the organisation of the passenger
transport industry, the legislative framework under which it
operates, in technology, in the incomes, life-styles and aspirations
of thetravelling public, in car ownership levels, and in the
attitudes of policy makers. While these changes have not
invalidated the general conclusions of the Black Book, they will
have reduced the relevance to modern conditions of much of the
quantitative analysis. The new collaborative study, of which the
resultsinthispaperarea part,wastherefore setuptotakeaccount
of another 20 years’ worth of public transport information, and
more recent advances in transport research techniques. The
overall objectives of the study were therefore to:
† Undertake analysis and research by using primary and
secondary data sources on the factors influencing the
demand for public transport.
† Produce quantitative indications of how these factors
influence the demand for public transport.
† Provide accessible information on such factors for key
stakeholders such as public transport operators and central
and local government.
† Produce a document that assists in identifying cost-effective
schemes for improving services.
1.2. The scope of the paper
The results presented in this paper are a distillation and
synthesis of identified published and unpublished evidence on
the influencing factors drawn from three key areas:
† Fundamental principles relating to transport demand.
† Evidence from research carried out since publication of the
1980 report.
† Empirical results for a range of modes.
Where possible, this paper looks at changes in response
parameters since the 1980 study.
The data for the study mainly came from existing studies
and literature identified through searches for relevant literature
in publication databases, material supplied by public transport
operators and local authorities and contacts with researchers
engaged in analysis in the field. The information was collected,
assessed for relevance and, as far as was possible, quality, and
an analysis and synthesis made of implications of the overall
body of evidence; a meta-analysis of fares elasticities was also
conducted. In assessing the evidence it was recognised that
fares elasticities, for instance, can be derived in a number of
ways, for example: time trends, stated and revealed preference
surveys, before-and-after studies, time series analysis, cross
sectional analysis, and logit modelling. All of these approaches
have their advantages and disadvantages, depending on the
context in which the original research was conducted. The
various methodological approaches were noted during the
information gathering exercise to ensure that the outcomes did
not contain unwanted bias.
Most findings reviewed relate to the urban and regional
market, with some references to rural areas. The inter-city,
long-distance market as such is not covered, and hence ‘long’
distances refer to about 30 km, as in the original study of 1980.
2. The effects of fares
2.1. Summary of overall findings
Fares are fundamental to the operation of public transport
since they form a major source of income to operators. In
general, if fares are increased, patronage will decrease.
Whether revenue increases or decreases as a result of a fare
increase depends on the functional relationship between fares
and patronage as represented by the demand curve. Usually this
is expressed through the concept of ‘elasticity’. In its simplest
form the value of the fares elasticity is the ratio of the
proportional change in patronage to the proportional change in
fares. It has a negative value when, as is usually the case, fares
and patronage are inversely related: an increase in fares leads to
a decrease in patronage and vice versa. If the value of the
elasticity is in the range 0 to K1, then a fares increase will lead
to increased revenue. If the value exceeds K1, then a fare
increase will lead to a decrease in revenue.1
Fare elasticities are dynamic, varying over time for a
considerable period following fare changes. Therefore it is
increasingly common for analysts to distinguish between short
run, long run and sometimes medium-run elasticity values.
There are various definitions of short-, medium- and long run,
but most authors take short run to be 1 or 2 years, and long run
to be around 12–15 (although sometimes as many as 20) years,
while medium run is usually around 5–7 years.
As well as considering the direct effects of a change in fares,
it is often important to consider the effects of fare changes on
other modes. The usual method to take into account the effect
that other modes have on the demand for a particular mode of
public transport is to use cross-elasticities, estimating the
demand elasticity for a competing mode with respect to the
change in the given mode.
Fare elasticity varies significantly depending not only on the
mode, and the time period over which it is being examined, but
also on the specific circumstances in which a mode is
operating. In the study, elasticity values from many sources
were examined to provide an up-to-date overview of fares
elasticities and the effects of various factors on the values. The
principal results of this analysis are shown in Table 1 and
Fig. 1. It can be seen that, broadly speaking, bus fare elasticity
averages around K0.4 in the short run, K0.56 in the medium
run and K1.0 in the long run; metro fare elasticities average
around K0.3 in the short run and K0.6 in the long run, and
local suburban rail around K0.6 in the short run. There is
1
To avoid confusion in comparisons of elasticities, many of which are
negative, the terms ‘increase’ and ‘decrease’ will always in this paper refer to
the change in the magnitude (the numerical part) of the elasticity. Thus an
elasticity which changes from K0.5 to K0.7 is said to have increased.
N. Paulley et al. / Transport Policy 13 (2006) 295–306296
evidence for this in Dargay and Hanly (1999), Gilbert and
Jalilian (1991).
These results appear to indicate a significant change from
those reported by Webster and Bly (1980), which were based
on international aggregate measures of fares elasticity for all
journey purposes and passenger types across all trip lengths
and fares. This analysis led to the conclusion that overall fares
elasticities are low, so that increases in fare levels will almost
always lead to increases in revenue. The analysis resulted in the
then accepted ‘standard’ public transport fares elasticity value
of K0.3. Given the dominance of before-and-after studies in
the 1980 report, it is likely this value is what would now be
called a short run elasticity. In the current work, the short run
elasticity has been found to be about K0.4.
Two of the main reasons for this difference are as follows.
Firstly, given that fare elasticity is different for different
journey purposes, there may have been a shift in the
proportions of journeys of different types for which people
are using public transport (for example, more leisure travel).
Secondly, for the same journey purpose the elasticity may
actually have changed. This could be due a variety of factors,
some of which will interact with each other: one of these is
increased rate of market turnover, insofar as potential new
users may have different perceptions of using public transport.
Other factors include: rising incomes and car ownership and
the varying quality of public transport service over the last 20
years. Interestingly, suburban rail short run fare elasticity has
changed very little, remaining at about K0.5.
The 1980 report did not cover medium or long run
elasticities at all. Therefore the likely value of medium run
bus fare elasticity of around K0.56 cannot be compared with
earlier estimates.
The realisation that long-term elasticities can exceed K1
has serious implications for the public transport industry.
While the immediate effect of a fare rise might be a temporary
increase in revenue, the long-term effect is likely to be a
decrease, although if future cash flows are discounted,
operators may benefit from fare increases. Nevertheless,
attempts to counter falling revenue with fare increases alone
will eventually fail. Reversal of negative trends in public
transport patronage requires service improvements, and
possibly fare reductions.
The relatively wide ranges of elasticity values about the
means shown in Table 1 and Fig. 1 reflect variation in methods
of estimation, as well as variation between studies in a number
of other factors influencing demand and elasticity. A few of the
more significant disaggregations are considered below.
2.2. Effect of types of fare change
Fare elasticities may be affected by the magnitude of the
fare change. In general, greater fare increases produce higher
values of elasticity than lower increases. There is evidence of
this from modelling of rail fares in South-east England by
Mackett and Bird (1989). The differences are greatest for long
run elasticities. Fare elasticity is also affected by the current
Table 1
Comparison of fare elasticities from the current study and the 1980 Black Book
Current study 1980 Study
Mean Range of values reported Number of studies
From To
Public transport—UK and outside the UK—short run K0.41 K0.07 K1.02 99
Public transport—UK—short run K0.44 K0.07 K1.02 68
Public transport—outside the UK—short run K0.35 K0.09 K0.86 31
Bus—UK and outside the UK—short run K0.41 K0.07 K0.86 44
Bus—UK—short run K0.42 K0.07 K0.86 33 K0.30
Bus—outside the UK—short run K0.38 K0.23 K0.58 11
Metro—UK and outside the UK—short run K0.29 K0.13 K0.86 24
Metro—UK—short run K0.30 K0.15 K0.55 15 K0.15
Metro—outside the UK—short run K0.29 K0.13 K0.86 9
Suburban rail—UK and outside the UK—short run K0.50 K0.09 K1.02 31
Suburban rail—UK—short run K0.58 K0.10 K1.02 20 K0.50
Suburban rail—outside the UK—short run K0.37 K0.09 K0.78 11
Bus—UK—medium run K0.56 K0.51 K0.61 2
Bus—UK—long run K1.01 K0.85 K1.32 3
Metro—UK—long run K0.65 K0.61 K0.69 2
Bus—London—short run K0.43 K0.14 K0.84 15 K0.44
Bus—outside London—short run K0.44 K0.07 K0.86 14
Suburban rail—SE England—short run K0.61 K0.10 K0.95 13
Suburban rail—outside SE England—short run K0.55 K0.15 K1.02 11
Bus—UK—peak—short run K0.26 0.00 K0.42 9
Bus—UK—off-peak—short run K0.48 K0.14 K1.00 10
Metro—UK—peak—short run K0.26 K0.15 K0.35 6
Metro—UK—off-peak—short run K0.42 K0.23 K0.63 5
Suburban rail—UK—peak—short run K0.34 K0.27 K0.50 4
Suburban rail—UK—off-peak—short run K0.79 K0.58 K1.50 5
N. Paulley et al. / Transport Policy 13 (2006) 295–306 297
level of the fare relative to people’s income. This can be
illustrated by the results for London buses. When fares were
particularly low, from October 1981 to March 1982, the
elasticity was around K0.30 to K0.33, but at the higher
relative fare levels in 1983, it was over K0.40 (Collins, 1982).
Elasticity values have also been found to increase with fare
levels for short distance (%32 km) rail journeys outside
London (Association of Train Operating Companies, 2002).
The response to a fare increase may not be equal and
opposite to the response to a fare decrease; that is, they may not
be symmetrical. The evidence is however limited. Hensher and
Bullock (1979) found, for rail fares in Sydney, Australia, that
the fare elasticity was K0.21 when the fares were increased but
K0.19 when they were decreased. However, Wardman (2000),
in a review of stated preference studies, found no evidence of
elasticity asymmetry, although his study did not include very
many cases where the prices fell.
2.3. Variation of elasticity with type of area
There is enormous variation between different types of area
in the pattern, type and level of public transport services, and
the demand for them. Generally speaking, people in areas with
low population densities tend to rely more on cars and less on
public transport than their more urban counterparts, and are
therefore more likely to have the option of switching to car
travel if fares rise.
In Great Britain, elasticity values are much higher in the
shire counties than in metropolitan areas (Table 2) (Dargay and
Hanly, 1999), probably reflecting lower levels of captivity to
bus and the greater feasibility of using car as an alternative. The
greater difference between the long and short runs in the
metropolitan counties may reflect a greater turnover of
population in such areas, allowing a wider range of responses
in the long run relative to the short run compared with more
rural areas.
The same type of argument might lead to the expectation
that residents of large cities are likely to be more dependent on
public transport than those in smaller cities, with corresponding
differences in fare elasticities. However, the evidence is less
clear cut.
Fig. 1. Summary of mean values and ranges of fare elasticities.
Table 2
Bus fare elasticities in Great Britain by type of area
Metropolitan areas Shire counties
Short run K0.21 K0.51
Long run K0.43 K0.70
London Outside London
Short run K0.42 K0.43
N. Paulley et al. / Transport Policy 13 (2006) 295–306298
2.4. London as a special case for bus travel
London bus services may be regarded as a special case
within Great Britain, not least because of the size of the
conurbation, levels of congestion and the extent of public
transport networks, but also because of the degree of regulation
that still obtains in London. As shown in Table 2, in the short
run, at least, bus fare elasticity is marginally higher outside
London than inside London (based on 15 studies in London and
14 outside it). One might expect a higher elasticity value for
buses in London because of the availability of the underground
as an alternative. On the other hand the deregulation of buses
and the greater ease of use of cars outside London mean that the
elasticity might be expected to be higher there. It looks as if
these factors counterbalance one another.
2.5. Peak and off-peak demand
Trips made in the peak tend to be for work and education
purposes, and so tend to be relatively fixed in time and space.
Off-peak trips tend to include leisure, shopping and personal
business trips for which there is often greater flexibility in
terms of destination and time. Hence, one would expect offpeak
elasticities to be higher.
In the UK, off-peak elasticity values are about twice the
peak values, with slightly greater variation for suburban rail
than the other modes, with peak values for bus, metro and
suburban rail of K0.26, K0.26 and K0.34, respectively, and
equivalent values of K0.48, K0.42 and K0.79, respectively,
for the off-peak. This may reflect the greater use of off-peak
fare discounts on rail than on bus or metro. Outside the UK, the
mean peak elasticity for buses is calculated to be K0.24, while
the equivalent off-peak value is K0.51 suggesting a slightly
higher differential between the peak and off-peak.
2.6. Fares elasticities for different trip purposes
People travelling to work or to school generally have little
choice of trip ends or timing of journeys. Such trips are largely
the cause of the peak, which is when congestion tends to be at
its greatest, making car journeys slower. Hence, elasticities
tend to be lower than for other trip purposes. Evidence to
demonstrate this was found in London and, for suburban rail, in
South-east England. Business trips paid by employers have
very low elasticity values, because an employer is likely to
regard a fare increase as largely irrelevant if a local business
journey needs to be made.
2.7. Fares elasticities for different types of traveller
Because those with access to a car have more alternatives
than those without, they tend to have higher elasticity values,
particularly in the long run. Males tend to have higher elasticity
values than females. This may be partly because they are more
likely to have a car available. The evidence for this comes from
microsimulation modelling by Mackett (1990).
The evidence about age is not clear-cut, because there are
several effects at work here. Many of the trips by the elderly
will be discretionary, and so one would expect a high elasticity
value for these trips. However, they may have low car
ownership and difficulty walking which means that many of
them may be captive to public transport, and so they have a low
elasticity value. In many places in Britain they receive free
public transport, and so they will continue to travel whatever
the fare level.
Travellers with high incomes tend to have higher elasticity
values because their higher car ownership levels mean that they
have an alternative when fares increase.
2.8. Fares elasticities by distance travelled
For buses, there are two possible effects here: for very short
trips, walking is a feasible alternative for many people, so
elasticities tend to be higher (see Tyson, 1984 for example),
and for ‘long’ trips (within the range of distances considered
here) fares are a larger proportion of incomes and a wider range
of alternative destinations exists (for example, for shopping),
hence elasticities tend to be higher for these trips (see White,
1981 for example). Evidence was found to support the idea that
elasticities are higher for very short and very long trips, and
lower for medium-length ones.
For rail, fare elasticities decrease with distance (within the
range considered in this paper). This may be because fares are
often subject to a taper, that is the fare per unit distance
decreases with increasing distance. This effect may outweigh
the effects of fares for longer journeys being higher proportions
of income.
2.9. Effects of ticket types and fare systems
The effects of pre-paid ticketing systems (travelcards or
season tickets) are not clear and may depend on the level of
discount and the conditions of use. In such cases the purchase
decision relates to a ‘package’ of travel (for example, unlimited
journeys within given zone(s) for a whole month), rather than
individual single or return journeys. The user therefore may
wish to consider the ‘value for money’ offered by such a
package both relative to single and return fares on the same
system (i.e. the discount within the public transport system)
and other modes (notably car). When such a ticket is purchased
for the first time it will influence behaviour not only by
changing the average money cost per trip vis a vis previous
travel patterns, but also encouraging additional trips (and
interchanges) at zero money cost. This in turn may influence
the renewal purchase decision at a new price level (White,
2001).
2.10. Meta-analysis
Meta-analysis involves pooling together the results from
different empirical studies and developing a quantitative model
that explains variations in results across studies. A metaanalysis
of the British evidence on fare elasticities was
N. Paulley et al. / Transport Policy 13 (2006) 295–306 299
conducted as part of the study (Wardman and Shires, 2003).
The aim of the exercise was to corroborate the findings of the
more conventional review and to obtain insights into issues that
would not otherwise be possible—such as the estimation of
elasticities over a wide range of circumstances and the
influence of the methodological approaches used in the
individual studies reported.
The analysis took the form of a regression model, estimated
using 902 public transport fare elasticities obtained from 104
studies conducted in Britain between 1951 and 2002. The
markets covered were inter-urban rail travel, suburban rail
travel, urban bus travel and London underground. A number of
interesting findings emerged and the models can be used to
‘predict’ fare elasticities for a range of situations.
The elasticities predicted by the resulting model, for various
types of modes, journeys and travellers are shown in Table 3.
There is a good degree of consistency between these results and
those from the individual studies reported above, suggesting
that the model derived from the meta-analysis might prove a
useful tool for estimation of fare elasticities where it is not
possible to establish them by more direct methods.
2.11. Conclusions on fare elasticities
Fare elasticities tend to increase over time since the change
of fare. For example, bus fare elasticities are about K0.4 in the
short run, K0.55 in the medium run, and about K1.0 in the
long run. Similarly, metro fare elasticities tend to be about
K0.3 in the short run and K0.6 in the long run. Elasticities
seem to be slightly higher in the UK than elsewhere. Fare
elasticities have increased since the 1980 study.
Fare elasticities are affected by the time of day: off-peak
values are about twice those in the peak. This partly reflects the
nature of the trip purposes which dominate in each. Work and
education trips, which tend to be in the peak, have lower
elasticity values than discretionary trips such as shopping and
leisure which tend to be during the off-peak.
Elasticity values tend to be higher in rural areas than
in urban areas. This is probably because of the higher car
ownership levels and may also be because fares will be higher
because journeys will be longer on average. In addition, where
school journeys are made on separate services (especially for
pupils entitled to free travel) the market served by the ‘public’
network contains a very small proportion of ‘peak’ journeys
(such as adults to work), and is dominated to a greater extent
than urban services by purposes such as shopping.
3. The effects of quality of service
3.1. Introduction
Quality of service may be defined by a wide range of
attributes which can be influenced by planning authorities and
transport operators. Some of these attributes (access and egress
time, service intervals and in-vehicle time) directly involve
time, and can be quantified with relative ease and incorporated
in appropriate demand forecasting models, using relevant
elasticities. Others (vehicle or rolling stock characteristics,
interchanges between modes, service reliability, information
provision, marketing and promotion, and various bus specific
factors) are more problematical because changes in these
attributes are often accompanied by changes in other attributes,
particularly fare and journey time. Valuations of such attributes
are often derived from stated preference (SP) models, based on
hypothetical behaviour, as distinct from the revealed preference
(RP) methods, based on actual behaviour, reviewed
earlier in this paper to illustrate aggregate price elasticities, and
later in this section to derive aggregate service level
(frequency) elasticities. Although there is a body of evidence
that suggests RP and SP approaches are comparable in terms of
attribute valuation, there is also evidence to suggest that SP
approaches may give biased elasticity results (Louviere et al.,
2000). In practice emphasis continued to be placed on RP
estimated elasticities but where these were absent the results
from SP attribute valuation were used. For example, the
relative importance of quality of service characteristics is often
expressed in terms of an attribute weighting relative to another
journey component. This weighting may be in terms of
equivalent in-vehicle time minutes. For example, a real time
information system may equate to a 3 min reduction of invehicle
time per trip. Alternatively, service attributes may be
expressed in monetary terms, such as a minute of wait time
being worth the equivalent of 10 pence in fare. Where attribute
weightings are determined as monetary equivalents these may
be added to actual fares/journey times and used, together with
an appropriate fare/journey time elasticity, to estimate effects
on demand.
3.2. Access time to boarding point and egress time
from alighting point
The evidence for the impact of access and egress time is
dominated by attribute valuation studies. The majority of these
studies were based on the use of SP, rather than RP, techniques
(Wardman, 2001).
Table 3
Fare elasticities derived from the meta-analysis
Elasticities
Bus—UK—short run K0.36
Metro—UK—short run K0.37
Suburban rail—UK—short run K0.52
Bus—UK—long run K0.70
Metro—UK—long run K0.54
Bus—London—short run K0.37
Bus—outside London—short run K0.36
Suburban rail—SE England—short run K0.50
Suburban rail—outside SE England—short run K0.60
Bus—UK—peak—short run K0.30
Bus—UK—off-peak—short run K0.40
Metro—UK—peak—short run K0.30
Metro—UK—off-peak—short run K0.44
Suburban rail—UK—peak—short run K0.42
Suburban rail—UK—off-peak—short run K0.65
N. Paulley et al. / Transport Policy 13 (2006) 295–306300
Weightings for walking times to and from bus stops and
stations range between about 1.4 and 2.0 units of in-vehicle
time (based on 183 observations), with no obvious dependence
on trip type and main mode. The corresponding range for
access and egress journeys by all means (including driving and
cycling to stations, etc.) is similar (1.3–2.1—based on 52
observations).
3.3. Service intervals
The effect of service intervals can be measured in a number
of ways: total vehicle kilometres or hours, frequency, headway/service
interval, wait time and schedule delay. The
dominant indicator is the number vehicle kilometres operated.
This has an inverse, but generally inexact, relationship with
service headways.
A number of studies have estimated the elasticity of bus
demand with respect to vehicle kilometres. As shown in
Table 4, this is approximately 0.4 in the short run, and 0.7 in the
long run. For rail services, the short run elasticity is somewhat
greater (about 0.75), but this is based on only three
measurements and no long run elasticity appears to have
been estimated.
Service elasticities for buses have been found to be
considerably greater on Sundays and in the evenings, when
service levels are generally lower (Preston, 1998). Similarly,
elasticities tend to be higher in rural than in metropolitan areas,
where service levels are higher (Dargay and Hanly, 1999).
There is some evidence, however, that bus demand is shown to
be more service elastic in big cities (with populations of over
500,000) than small towns because of the competition from
other public transport modes. It is also suggested that service is
valued more highly in large cities due to higher income levels
(European Commission, 1997).
Elasticities for bus demand have also been estimated with
respect to passenger waiting times. The average value appears
to be K0.64, but values for off-peak journeys, and journeys to
non-central destinations, tend to be higher (Preston and James,
2000). Service levels may also be expressed in terms of vehicle
hours operated. Elasticities estimated from increases in bus
hours operated were found (in four studies) to be of the order of
C1.0 (Pratt et al., 2000).
It is also possible to consider the effects of service levels by
estimating attribute value of waiting time in terms of in-vehicle
times. For buses, wait time appears to be valued at about 1.6
times in-vehicle time, while the corresponding value for rail is
1.2 (both based on 11 observations—see Wardman, 2001).
3.4. Time spent on board the vehicle
There is limited evidence on bus elasticities with respect to
in-vehicle time (IVT), possibly because the options for
improving bus speeds are somewhat limited, especially in
urban areas. In addition, for short journeys, IVT may be only a
relatively small part of the total journey time.
The review suggests that IVT elasticities appear to be
roughly in the range K0.4 to K0.6 (based on three studies),
while those for urban or regional rail range between K0.4 and
K0.9 (based on five studies). Small and Winston (1999)
suggest greater values for longer interurban journeys (K2.1 for
bus, K1.6 for rail).
There is also some evidence on elasticities with respect to
generalised cost (GC) which brings together fare, in-vehicle
time, walk and wait times. Generalised cost elasticities lie in
the range K0.4 to K1.7 for buses, K0.4 to K1.85 for London
underground, and K0.6 to K2.0 for rail. These ranges
incorporate variations with journey purposes and income
(Halcrow Fox et al., 1993).
3.5. The waiting environment
Passengers who have to wait for buses or trains prefer to do
so in conditions of comfort, cleanliness, safety and protection
from the weather. Attribute values have been derived for
various aspects of bus shelters, seats, lighting, staff presence,
closed-circuit TV and bus service information. Estimates for
individual attributes of the waiting environment range up to 6 p
per trip (subject to a limiting cap of around 26 p on the total—
Steer Davies Gleave, 1996), or up to 2 min of in-vehicle time
per trip (Wardman et al., 2001).
3.6. Effect of vehicle or rolling stock characteristics
The attributes of public transport vehicles are largely
unquantifiable and they are too many and various for direct
analysis of their effects on demand. It is almost axiomatic that
passengers will prefer clean, comfortable vehicles that are easy
to get on and off, but the relative importance of such factors is
difficult to determine. SP techniques have therefore commonly
been used, sometimes in conjunction with RP approaches, to
obtain quantifiable measurements.
Studies using SP methods have suggested that a trip in a
low-floor bus may be perceived as being worth 5–14 pence
more than a trip in a conventional bus with high steps (Accent,
2002). Evidence from bus operators suggests a passenger
growth of about 5% on low-floor conversion, arising not only
from wheelchair users but also categories such as those with
heavy shopping or children in buggies. If considered as an
effective reduction in monetary fare, the SP valuation produces
a similar percentage growth when a typical short run fares
elasticity is applied.
Research for rail has estimated the effects of replacing old
with new rolling stock, using a combination of RP and SP
methods (Wardman and Whelan, 2001). Rolling stock
improvements are typically valued at around 1–2% of
Table 4
Bus and rail service elasticities
Bus No. of obs Rail No. of obs
Short run 0.38 27 0.75 3
Long run 0.66 23 – –
N. Paulley et al. / Transport Policy 13 (2006) 295–306 301
in-vehicle time. Refurbishment which changes the level of
train seating layout, ride quality, ventilation, ambience, noise
and seating comfort from levels associated with old ‘slam door’
stock to new air conditioned stock in South-east England was
worth around 2.5% of the fare. However, most refurbishments
would be worth somewhat less than this, with 1.5% being a
representative figure.
3.7. Public transport interchange
The ideal public transport service would carry the passenger
directly between origin and destination. In practice, given the
diversity of travel patterns, this is not an option for many
passengers who have to make interchanges between or within
modes. Studies in Great Britain have found that passengers
dislike interchange. The average equivalent penalty, including
walking and waiting times necessary to effect an interchange, is
21 min IVT on a bus trip (based on six observations), and
37 min IVT on a rail trip (based on 13 observations—
Wardman, 2001). There is however considerable variation
between journey purposes and from place to place. For
example, interchange penalties may be much smaller in
urban environments with high-frequency public transport
services.
3.8. Reliability
The main manifestations of public transport reliability are
excessive waiting times due to late arrival of buses or trains,
and excessive in-vehicle times, due to traffic or system
problems. It is common to express these forms of unreliability
in terms of standard deviations in waiting or in-vehicle times.
The limited available evidence suggests that the perceived
penalties are broadly equivalent to the standard deviation
multiplied by the corresponding value of waiting or in-vehicle
time (WS Atkins and Polak, 1997). For example if the mean
waiting time is 5 min, with a standard deviation of 2.5 min,
then the effective waiting time is 7.5 min.
3.9. Information provision
Some basic level of information about public transport
services is necessary for those who use or plan to use them. In
practice, regular travellers rarely make use of formal
information systems, and many occasional travellers rely on
informal sources such as advice from family and friends. While
it is relatively easy to discover who makes use of various
different information systems, there is little direct evidence of
their effect on demand.
The vast majority of evidence on information provision
takes the form of attribute valuation, using stated preference
and other attitudinal survey methods. There is considerable
variation between the results from different studies, partly
because of methodological differences, and partly because the
resulting attribute weightings are generally small compared
with other factors that vary between studies. Most recent
research has been on the effect of real time public transport
information systems, with digital displays at bus stops or metro
stations displaying the predicted arrival times of relevant buses
or trains. Evidence from Countdown in London (Steer Davies
Gleave, 1996) and similar systems elsewhere (Accent, 2002)
suggests a valuation somewhere between 4 p and 20 p per trip.
Service information available at home, through printed
timetables, bus maps, telephone enquiry services, etc. seems to
be valued at between 2 p and 6 p per trip (based on four
observations), and similar information at bus stops at between
about 4 p and 10 p per trip (based on 43 observations).
3.10. Conclusions on quality of service
There is generally less evidence on the demand impacts of
service quality variables than that of fares. The main body of
evidence on elasticities relates to bus service levels, although
there is also some elasticity evidence on the impact of IVT.
There is a large body of evidence on attribute values,
particularly for walk and wait time, IVT and information
provision, based largely on SP studies. In combination with a
knowledge of fare elasticity, the fare level and the level of the
service attribute (all derived from RP studies), the valuation of
the service attribute can be used to infer a service attribute
elasticity. For example, at 2000 prices it was found that the
mean value of time for commuting by urban bus was 4.2p/min
(based on 17 observations), whilst the value of leisure travel
was 2.6p/min (based on 17 observations—Wardman, 2001).
This in turn implies an elasticity of bus demand with respect to
IVT of around K0.4, which is consistent with the range given
in Section 3.4. There is likely to be further scope for combining
SP and RP evidence in this way. However, more evidence is
also needed on the demand impacts of service improvements,
particularly in terms of IVT, the waiting environment, vehicle
characteristics, interchange, reliability and pre-trip information.
There are other areas, such as personal security,
where there have been very few quantifiable results to date.
4. Demand interactions: effects of fare changes on
competing modes
Most evidence on public transport cross elasticities in Great
Britain has been collected in London, usually in research
undertaken by, or sponsored by Transport for London and its
predecessors (see Table 5).
In London the relatively high sensitivity of underground use
to bus fares (cross elasticityZ0.13) may reflect the overlap of
underground and bus networks which provide a choice of
public transport mode for many travellers. However, the
smaller sensitivity of bus use to underground fares conforms
less well with this observation, possibly because many
suburban areas served by bus are not accessible by the
underground. The relationships between rail and bus show a
similar asymmetry. The least interaction seems to be between
rail and underground, possibly reflecting the complementary,
rather than competitive roles of these modes. Car use is almost
independent of bus and underground fares.
N. Paulley et al. / Transport Policy 13 (2006) 295–306302
In other urban areas, public transport use is remarkably
sensitive to car costs, but car use is much less dependent on
public transport costs (Table 6). This reflects differences in
market shares of public and private transport: a small
percentage shift from car travel can amount to a large
percentage increase in public transport use. This observation
also applies to inter-urban travel (Table 7), where the relatively
high cross elasticities for inter-urban coach travel with respect
to rail fares (0.32), and vice versa (0.17), suggest a higher level
of interchangeability between these modes.
5. Effects of income and car ownership
5.1. Introduction
Traditionally income and car ownership have been deemed
‘background factors’, as compared to attributes of public
transport such as fares, service levels, journey times and
vehicle quality, which are directly under the control of the
operator. The broad relationships between income, car ownership
and the demand for public transport are well
documented. Despite this the exact relationships and the
correlation between all three factors, and in particular between
income and car ownership, would appear to be only marginally
clearer since the original demand for public transport
publication.
The last 23 years have seen marked increases in real income
and car ownership levels in the UK and across Europe. For
example, in this period GDP increased by around 68% in Great
Britain whilst the number of cars per household has increased
from 0.76 to 1.11. In that time, local bus journeys have fallen
by around a third. This is consistent with evidence from the UK
National Travel Survey that bus use (both in trips and personkilometre)
falls substantially as car ownership per household
rises. However, for rail the position is more mixed—while trips
per person decline with rising household car ownership,
person-kilometre shows little variation, as average trip length
becomes higher. The performance of rail at a local level
depends on congestion levels and, because of the perceived
higher quality of rail, is less sensitive to increases in car
ownership than bus. Indeed, Central London rail commuter
traffic has increased by 13% since 1980, associated with growth
in employment levels in that area.
Income is expected to increase the number of trips and their
average length. It is likely that this additional travel will be
split between increased public transport trips and increased car
trips, depending upon the level of car availability and assuming
that public transport is a normal good. Income is also a key
determinant of car ownership and hence there will be a
secondary and negative impact on the demand for public
transport via car ownership. Rising car and driving licence
ownership, income growth and the declining real cost of car
ownership have been identified as the key factors that have
shaped personal travel patterns in the last twenty years. Whilst
a host of other background factors can be cited, four key
relationships are outlined below:
† An increase in income will, depending upon the level of
income, lead to an increase in car ownership and so car
availability, or to an increase in public transport use.
† An increase in car ownership/availability will, other things
being equal, lead to a reduction in the demand for public
transport modes.
† The sign and magnitude of demand elasticities for public
transport with respect to car availability and income will
vary depending upon the income levels.
† Income growth can be expected to increase average trip
length.
Because of these relationships considerable care must be
taken when interpreting public transport demand elasticities
that have been estimated with respect to income and car
ownership. Income elasticities estimated using demand models
that do not have car ownership amongst their explanatory
variables will pick up the negative effect that car ownership has
on public transport and are not comparable with income
elasticities that are estimated alongside car ownership terms.
The problem with estimating models that include both
variables is the collinearity that exists between them. The
first demand for public transport book noted this in detail and
twenty years on the problem of collinearity still exists and is
particularly noticeable for models that have been calibrated
using time series data.
Table 6
Urban cross elasticities
Car use Rail use Bus use
Car cost – 0.59 0.55
Rail cost 0.054 – 0.08
Bus cost 0.057 0.24 –
Sources: Toner (1993), Wardman (1997b).
Table 7
Interurban cross elasticities
Car use Rail use Coach use
Car time – 0.33 0.60
Car cost – 0.25 0.34
Rail time 0.057 – 0.20
Rail cost 0.066 – 0.32
Coach time 0.054 0.17 –
Coach cost 0.014 0.17 –
Source: Wardman (1997a).
Table 5
Matrix of cross elasticities for London
Bus use Underground use Rail use Car use
Bus fare – 0.13 0.06 0.04
Underground fare 0.06 – 0.03 0.02
Rail fare 0.11 0.06 – N/A
Bus miles – 0.22 0.10 0.09
Underground miles 0.09 – 0.04 0.03
Bus journey time – 0.18 0.08 0.06
Source: Glaister (2001).
N. Paulley et al. / Transport Policy 13 (2006) 295–306 303
5.2. Effect of income on travel expenditure and distance
travelled
In almost all Western European countries total personkilometre
has risen at around 1–2% per annum, a little less than
the growth in real GDP. Table 8 illustrates the growth
experienced within Western Europe between 1990 and 1998,
with total person-km for motorised modes rising by 19%. The
greatest growth was experienced in air travel (65%), followed
by car (18%), bus and coach (9%), rail (8%), and tram and
metro (5%).
There can be no doubt that income has a positive impact
upon the total amount of travel. Further, the figures from the
Family Expenditure Survey for Great Britain show that the
percentage of household expenditure on transport and travel
has slowly increased over time, rising from 14.8% in 1981 to
16.9% in 1999/00. These figures exclude expenditure on air
travel, which has seen significant growth (nearly 50% more
passenger kilometres between 1989 and 1999) during the last
twenty years.
Given little change in the population, traffic growth comes
from two sources: people making additional trips and people
making longer journeys. There is clear evidence that trip
lengths are increasing with income, although the effects are not
particularly strong. In general, the elasticities of trip length
with respect to income lie in the range 0.09–0.21 but with
noticeably stronger growth for car commuting, business trips
by rail and business trips by bus. The latter is not a particularly
significant category, whilst the figures for rail business trips
will include longer distance journeys.
5.3. Effect of income on public transport demand
The empirical evidence from Britain clearly indicates that
the bus income elasticity, which includes the car ownership
effect, is negative. It appears to be quite substantial, in a range
between K0.5 and K1.0 in the long run although somewhat
smaller in the short run as is clear in Table 9. This would
explain the sustained reductions in bus demand over time.
However, as car ownership approaches saturation, the income
elasticity can be expected to become less negative.
In studies based on the volume of demand, there is strong
correlation between income and car ownership which means
that it is difficult to disentangle the separate effects of each. In
some instances, it has even resulted in coefficients of wrong
sign. Various studies have attempted to overcome this problem
using outside evidence and constrained estimates, whilst
analysis of trip patterns at the individual level, as is possible
with UK National Travel Survey (NTS) data, does not face
serious correlation problems.
There is some evidence to suggest that variations in the
demand for bus purely as a result of income growth are
negative, but in any event the overall effect after the
introduction of car ownership is negative.
Although car ownership has a negative impact on rail
demand, it is less than for bus and, although there are quite
large variations between market segments and across distance
bands, the overall effect of income on rail demand is quite
strongly positive. Unlike the bus market, there are many
segments in the rail market where car ownership has saturated
or where car availability is sufficiently high that the growth rate
and its negative impact on demand is low. Rail income
elasticities are generally found to be positive, and as high as 2
in some cases. As with the bus income elasticity, the rail
elasticity can also be expected to increase over time.
5.4. Effect of car ownership on public transport demand
There is some empirical evidence relating to the effect of car
ownership on public transport demand where income is not
entered into the model. However, there are fewer instances
where car ownership is the sole variable representing external
factors.
The evidence from studies which have concentrated solely
on car ownership as a predictor of the effects of external factors
on public transport demand indicate that the impact on bus
travel in Britain is negative (see also Section 5.1 above).
5.5. Conclusions on income and car ownership effects
Income and car ownership growth are fundamental to the
underlying demand for public transport. There has been almost
continual decline in the demand for bus travel over the past 25
years, although rail travel has recently experienced something
of a renaissance. To a lesser extent than for car travel, the
average trip length by public transport has increased with
income, with elasticities of trip length to income in the range
0.1–0.2. The income elasticity of bus demand, including the
indirect car ownership effect, is large and is in the range K0.5
to K1.0 in the long run. This can be expected to fall as the car
ownership growth induced by income growth slows as
saturation is approached. Rail income elasticities are somewhat
Table 8
Growth in public transport use: European Union countries 1990–1999
Mode Growth in passenger kilometres (%)
Passenger cars 18
Buses and coaches 9
Tram and metro 5
Railway 8
Air 65
All 19
Table 9
Bus income elasticities (Great Britain)
Short run Long run
National data (journeys) 0 K0.45 to K0.80
National data (pass-kilometres) 0 K0.15 to K0.63
Regional data (journeys) 0 to K0.29 K0.64 to K1.13
County data (journeys) K0.3 to K0.4 K0.6 to K0.7
PTE data (journeys) K0.7 K1.6
N. Paulley et al. / Transport Policy 13 (2006) 295–306304
larger, in part because car ownership levels are much higher in
the rail market, and can be as high as 2.
6. Concluding remarks
This paper reports on key findings from a collaborative
study whose objective was to produce an up-to-date guidance
manual on the factors affecting the demand for public transport
for use by public transport operators and planning authorities,
and for academics and other researchers. While a wide range of
factors was examined in the study, the paper concentrates on
the findings regarding the influence of fares, quality of service
and income and car ownership.
Fare elasticities tend to increase over time since the change
of fare, with bus fare elasticities being about K0.4 in the short
run, K0.55 in the medium run, and about K1.0 in the long run.
Similarly, metro fare elasticities tend to be about K0.3 in the
short run and K0.6 in the long run. For quality of service, the
mean value of time for commuting by urban bus was 4.2p/min,
whilst the value of leisure travel was 2.6p/min (at 2000 prices),
implying an an elasticity of bus demand with respect to invehicle
time of around K0.4. As incomes increase over time,
trip lengths increase. The impact varies across journey
purposes, but with elasticities in the range 0.1–0.2 the long
run impact on passenger kilometres, if maintained, will be
significant. Income has a positive impact on public transport
demand, but with an offsetting negative impact, particularly in
the bus market, through its effects on car ownership. As car
ownership growth slows and reaches saturation, these negative
effects will diminish.
As has been shown, a substantial body of evidence exists
with respect to fare elasticities and, to a lesser extent, service
and income elasticities, with important distinctions made
between the short run and the long run. There is also a sizeable
evidence base on the valuation of key attributes such as walk
time, wait time, IVT and some aspects of information
provision. However, there is more limited evidence on the
impacts of reliability, vehicle characteristics, the waiting
environment, interchange, personal security, and marketing
and awareness campaigns. Such attributes are increasingly
central elements of transport policy, and understanding their
impact is crucial if policies are to be properly formulated and
implemented.
Whilst there can be little doubt that a wide range of factors
influences the demand for public transport, and there is plenty
of empirical evidence as to what the relevant factors are, and
which of them may be more important than others, in different
circumstances, it must always be recognised that the results
may be subject to a considerable degree of uncertainty. One of
the problems encountered during the study was in determining
the context under which some of the reported experiments and
studies had been conducted. This was especially marked with
regard to separating short and long run effects. This whole issue
would benefit from further investigation, particularly to
ascertain whether attribute valuations refer to the short- or
the long run.
Acknowledgements
This document is the output of a project funded by the UK
Engineering and Physical Sciences Research Council under the
Future Integrated Transport (FIT) research programme (grant
numbers GR/R18574/01, GR/R18550/01 and GR/R18567/01)
and also supported by the Rees Jeffreys Road Fund, the
Confederation for Passenger Transport, the Association of
Train Operating Companies and the Passenger Transport
Executive Group; the authors are grateful for their support.
The information contained herein does not necessarily reflect
the views or policies of the supporting and funding
organisations.
References
The results reported in this paper, are drawn from a wide number of reports and
articles that had been published or were otherwise made available to the
authors. A complete bibliography is not included here, but may be found in
the full report of the study (Balcombe et al, 2004).
Accent, 2002. UK bus priorities modal shift. Report for the Commission for
Integrated Transport, London.
Association of Train Operating Companies, 2002. Passenger Demand
Forecasting Handbook, fourth ed. Passenger Demand Forecasting Council,
London.
Atkins, W.S., Polak, J., 1997. Bus Reliability—stated Preference Research.
Department of Transport, London.
Balcombe, R., Mackett, R., Paulley, N., Preston, J., Shires, J., Titheridge, H.,
Wardman, M., White, P., 2004. The demand for public transport: a practical
guide, TRL Report TRL 593, Crowthorne, UK.
Collins, P.H., 1982. The effects of recent London Transport fares changes.
Proceedings of Seminar M held at the PTRC 10th Summer Annual Meeting,
University of Warwick, 12–15 July, 1982, vol. P227, PTRC, London.
Dargay, J., Hanly, M., 1999. Bus fare elasticities. A report to the Department of
the Environment, Transport and the Regions. ESRC Transport Studies Unit
Working Paper 1999/26, University College London.
European Commission, 1997. ISOTOPE (Improved structure and organization
for urban transport operations of passengers in Europe). Final report. Office
for Official Publications of the European Communities, Luxembourg.
Gilbert, C.L., Jalilian, H., 1991. The demand for travel and for travelcards on
London regional transport. Journal of Transport Economics and Policy 25
(1), 3–29.
Glaister, S., 2001. The economic assessment of local transport subsidies in
large cities. In: Grayling, T. (Ed.), Any More Fares? Delivering Better Bus
Services. IPPR, London.
Halcrow Fox, Accent Marketing Research and the University of Leeds, 1993.
London congestion charging: review and specification of model elasticities.
Department of Transport, London.
Hensher, D.A., Bullock, R.G., 1979. Price elasticity of commuter mode choice:
effect of a 20 percent rail fare reduction. Transportation Research 13a (3),
193–202.
Louviere, L.L., Swait, J., Hensher, D.A., 2000. Stated Choice Methods:
Analysis and Applications. Cambridge University Press, Cambridge.
Mackett, R.L., 1990. Explanatory analysis of long-term travel demand and
policy impacts using micro-simulation. In: Jones, P. (Ed.), Developments in
Dynamic and Activity-Based Approaches to Travel Analysis. Avebury,
Aldershot, pp. 384–405.
Mackett, R.L., Bird, J., 1989. Commuting in South-East England. Report to
British Railways Board, Transport Studies Group, UCL, February.
Pratt R.H. Consultant, Inc., Texas Transportation Institute, Cambridge
Systematics, Inc., Parsons Brinkerhoff Qaade & Douglas, Inc., SG
Associates and McCollom Management Consulting, Inc., (2000). Traveler
response to transportation system changes. Interim handbook. TCRP Web
Document 12. TRB, Washington DC.
N. Paulley et al. / Transport Policy 13 (2006) 295–306 305
Preston, J., 1998. Public Transport Elasticities: Time for a Rethink? Working
Paper 856, Transport Studies Unit, University of Oxford.
Preston, J., James, T., 2000. Analysis of Demand for Bus Services. Final
Report. Transport Studies Unit, University of Oxford.
Small, K., Winston, C., 1999. The demand for transportation: models and
applications. In: Gomez-Ibanez, J. et al. (Eds), Essays in Transportation
Economics and Policy. Brookings Institute, Washington, DC.
Steer Davies Gleave, 1996. Bus passenger preferences. Report for London
Transport. Steer Davies Gleave, London.
Toner, J.P., 1993. Review of Urban Demand Elasticities, Project Note for Urban
Transport Market: Theoretical Analysis, Institute for Transport Studies,
University of Leeds.
Tyson, W.J., 1984. Evidence on Elasticities. Seminar Paper, Report 246,
Transport Studies Unit, University of Oxford.
Wardman, M., 1997. Disaggregate Inter-urban Choice Models: A Review of
British Evidence with Special Reference to Cross Elasticities, ITS Working
Paper 504, Institute for Transport Studies, University of Leeds, July 1997.
Wardman, M., 1997. Disaggregate Urban Choice Models: A Review of British
Evidence with Special Reference to Cross Elasticities, ITS Working Paper
505, Institute for Transport Studies, University of Leeds, July 1997.
Wardman, M., 2000. Using Existing Stated Preference Data to Analyse Bus
Performances. Working Paper 555. Institute for Transport Studies,
University of Leeds.
Wardman, M., 2001. Public Transport Values of Time. Working Paper 564.
Institute for Transport Studies, University of Leeds.
Wardman, M., Shires, J., 2003. Review of Fare Elasticities in Great Britain,
ITS Working Paper 573, Institute for Transport Studies, University of
Leeds.
Wardman, M., Whelan, G., 2001. Valuation of improved railway rolling
stock: a review of the literature and new evidence. Transport Reviews
21 (4), 415–447.
Wardman, M., Hine, J., Stradling, S., 2001. Interchange and travel choice.
Scottish Executive Central Research Unit, Edinburgh.
Webster, F.V., Bly, P.H. (Eds.), 1980. The Demand for Public Transport.
Report of an International Collaborative Study. Transport and Road
Research Laboratory, Crowthorne, Berkshire.
White, P., 1981. Recent developments in the pricing of local transportation
services. Transport Reviews 1 (2), 127–150.
White, P., 2001. Public transport: its planning, management and operation.
Spon, London (Chapter 7).
N. Paulley et al. / Transport Policy 13 (2006) 295–306306