CALCULATING TRANSPORT CONGESTION AND

SCARCITY COSTS
FINALREPORT OFTHE EXPERTADVISORS TOTHE
HIGH LEVELGROUPON INFRASTRUCTURE CHARGING
(WORKINGGROUP2)
MAY7 1999
This report was prepared by Professor Chris Nash and Mr Tom
Sansom, (ITS) and finalised in agreement with the other experts in
the group including:
Dr John Dodgson
NERA, UK
Professor Rainer
Friedrich
IER, Germany
Dr Lars Hansson
Research Leader
IIIEE , Lund University
Sweden
Professor Chris Nash
ITS, Leeds University
Leeds, UK
Dr Markus Pennekamp
Deutsche Bahn AG
Germany
Mr Stephen Perkins
ECMT, France
Professor Stef Proost
Center for Economic
Studies
Katolische Universiteit
Leuven, Belgium
Professor Rmy
Prudhomme
Universit Paris XII
France
Professor Emile Quinet
Ecole Nationale Ponts
et Chausses
France
Dr Andrea Ricci
ISIS, Italy
Professor Dr Werner Rothengatter
Inst. for Economic
Policy Research
(IWW), University of
Karlsruhe
Karlsruhe, Germany
Dr Rana Roy
Adviser
International Union of
Railways
UK
Professor Michel Savy
ENPC
France
Meetings were chaired by Professor Phil Goodwin, member of the High
Level Group.
3
TABLE OF CONTENTS
SUMMARY AND RECOMMENDATIONS............................................................................................................ 4
INTRODUCTION....................................................................................................................................................... 8
1. ROAD CONGESTION................................................................................................................................... 10
RECOMMENDATIONS................................................................................................................................ 15
2. CONGESTION AND SCARCITY COSTS OF RAIL....................................................................................... 15
RECOMMENDATIONS..................................................................................................................................... 17
3. MONETARY VALUATION................................................................................................................................ 17
RECOMMENDATIONS...................................................................................................................................... 20
4. CONCLUSIONS.............................................................................................................................................. 21
REFERENCES.......................................................................................................................................................... 21
ANNEX A: EXAMPLES OF SPEED FLOW RELATIONS AND CONGESTION COSTS.......................... 24
ANNEX B: TYPICAL VALUES OF TIME (PETS D7) .................................................................................... 29
4
SUMMARY AND RECOMMENDATIONS
Efficient pricing requires that prices reflect social marginal cost. To implement this, it is
necessary that estimates be made of all the elements of social marginal cost. The current report
aims to advise on the best approach to estimate external congestion and scarcity costs for road
and rail infrastructure. This requires a method of forecasting the increase in journey time and
unreliability for other traffic caused by an increase in traffic on the mode in question, and then
placing appropriate money values on them. These values vary with vehicle type, road type and
time of day, and thus fully reflecting them in prices requires price mechanisms which can
themselves vary in these dimensions. Although it is outside our remit to consider the issue of
implementation in detail, we consider that a combination of electronic road pricing in congested
cities and on congested trunk roads, fuel tax and annual licence duty (and possibly electronic
kilometre-based changes for certain vehicle categories) is capable of achieving a reasonable
approximation to marginal social cost.
In the case of rail, as with other types of transport infrastructure where specific slots are allocated
to particular users, the major issue is not so much congestion as the scarcity value of slots; when
the infrastructure approaches capacity, other users are unable to obtain the slot they want.
Whereas for road, keeping track of the use of the infrastructure by all individual users is a major
problem, for rail the information is readily available to the infrastructure manager. It is therefore
assumed that there is no great practical difficulty in implementing complex pricing structures for
rail infrastructure - including two part tariffs or individually negotiated contracts - if this is
desired.
It is not the task of this paper to review the arguments for and against marginal social cost
pricing. However, we acknowledge that a number of issues other than economic efficiency, such
as distributional effects, implementation costs and acceptability must be considered before such
an approach is implemented. The degree of accuracy to which it is worth reflecting marginal
social cost in price for roads must always be the subject of a cost-benefit analysis, given the
relatively high implementation costs of electronic road pricing. Recent studies have shown net
benefits from road pricing in London of 225m per annum after allowing for implementation
costs (MVA, 1995), and of 2.5 billion francs per year before implementation costs in Paris
(Prud'homme, 1999). Whilst these figures suggest that investments in these measures show very
high benefit-cost ratios relative to other transport investments in large cities, they are an order of
magnitude smaller than the often quoted but irrelevant figure of the total cost of congestion as
2% of GDP.
The environmental and safety implications of road congestion are not considered here as they are
covered in the reports of other working groups.
The recommendations are grouped into three areas:
1. Forecasting road congestion and unreliability.
The requirement here is not simply to be able to estimate the impact on other road users
of an additional vehicle at existing traffic levels. Rather it is necessary also to be able to
forecast how road users would adapt to being charged for these costs, in order to find an
5
equilibrium combination of charges and traffic levels. Bearing this in mind, we make the
following recommendations:
1.1 Wherever possible, external road congestion costs should be estimated from a model
which simulates the interaction of demand and supply on the road network. The model
can then be used to approximate the marginal external costs of congestion by rerunning it
with small changes in traffic volumes, and examining the effects on journey time for
existing traffic. This model would ideally incorporate a detailed network description, with
both speed/flow relationships and junction delays, and allow for user behaviour in terms
of rerouteing, retiming, changing destination or mode or changing frequency of travel, in
order to obtain a new set of flows and journey times following imposition of a charge.
Data is therefore required on the base origin/destination matrix, base generalised costs
and responses to changes in these values. The calculation of generalised cost requires
knowledge of operating costs, values of time and vehicle occupancy rates. Only when the
charge is equal to the marginal external cost in this new position has the optimal level of
charge and traffic been found.
1.2 Where this is not possible, we recommend that calculations are undertaken for typical
inter urban or rural roads at alternative traffic levels and mixes of vehicle types using link
speed/flow relationships. Separate calculations will be needed according to the type of
road (number of lanes; motorway or conventional road). Again, data on base traffic flows
and generalised costs are needed, and traffic volumes should again be adjusted for the
introduction of charges, if necessary by means of a simple price elasticity of demand, in
order to obtain an equilibrium value.
1.3 For urban areas, the degree of interaction between roads means that such an
approximation will be particularly crude. If a full network model is not available, the use
of area speed/flow relationships relating to the entire network for central, inner and outer
urban areas is likely to be preferable to link based speed/flow relationships.
1.4 Forecasting the impact of increased traffic on unreliability is more difficult, but given the
importance of the issue it should be attempted wherever possible. A variety of approaches
exists, including the use of micro-simulation models which model individual vehicles and
can thus estimate the spread of journey times, and purely empirical approaches, which
require data on unreliability and on traffic flows for a set of roads over time.
1.5 All the above relationships should relate to local conditions in the area concerned, and
relate to conditions such as driving styles and typical speeds in that location. It would be
counter-productive therefore to attempt to specify Europe-wide relationships, although
results may with care be transferred from comparable situations elsewhere in Europe if
local information is not available.
2. Forecasting rail delays and scarcity values
Fundamentally the approach we take to pricing on rail is consistent with our
recommendations on road. That is we seek to reflect in the price all the social costs
imposed by the operator on the rest of society by the use of a particular slot. However, in
practice there are significant differences. Given the fact that specific slots are allocated to
particular operators on rail infrastructure, the main effect of excess demand is not
congestion as such, but the inability of particular operators to obtain the slots they want.
The element of social cost to which this gives rise is the 'scarcity value' of the slot - i.e. its
6
value in the next best use. This cost is strictly only an externality however where it is
borne by another operator; when the next best use is by the same operator the cost should
be already internalised in that - provided they are efficient - they will already have
assessed the alternative uses of the slot. There is no general way of calculating this
'scarcity value' from information about the volume of traffic and the characteristics of the
route. This means that a rather different approach is needed to the estimation of the
marginal social cost of rail infrastructure from that taken in the case of roads.
2.1 Estimation of the scarcity value of specific slots on rail infrastructure requires a way of
revealing the value placed on the slots by alternative possible users, both in terms of
commercial rail operators and in terms of government bodies wishing to provide social
services. It may be possible in some cases to reveal these values by auctioning the slots,
but given the complexities involved in terms of the alternative ways in which the
infrastructure may be used, this is difficult. Some pre-packaging of slots is probably
necessary, in order to offer attractive combinations to alternative bidders. In general, a
process of negotiation appears the most practicable way forward. This might work in
terms of train operators first registering their wishes, the infrastructure manager using
these to produce packages of paths and charges and further negotiation then taking place
to determine whether operators would be prepared to pay more to improve their package,
or to surrender some of their paths in return for a reduced charge. Such negotiations
would also naturally encompass investment in expanded or enhanced capacity and the
sharing of the development costs.
2.2 Unscheduled delays imposed by one train operator on another may be measured ex post if
adequate monitoring is undertaken to measure both the extent and the cause of delays.
However, this will only measure the delays directly caused and not those where the
presence of the additional train has worsened the consequences of other delays by
absorbing part of the recovery margin. It is therefore more accurate to measure anticipated
delays by simulation modelling and charge these as part of the tariff. Of course these
additional delays will vary by route, type of train and time of day.
3. Monetary valuation
Changes in speed may lead to changes in operating costs in terms of fuel, tyres and
brakes. This may be estimated from appropriate formulae, and converted to resource cost
by deducting taxes. But the main effects of congestion and unreliability are in terms of the
time of people, poorer utilisation of vehicles and delays to the goods they carry. In all
cases, values for these must be converted to per-vehicle values (e.g. for passengers, by
weighting by the occupancy of vehicles).
3.1 For staff working in the transport industry, the usual approach of estimating the marginal
cost of their time as their wage rate plus an allowance for the overhead costs of employing
labour is generally appropriate. Similarly, the costs of poorer vehicle utilisation may be
estimated by calculating the impact on fleet size of the delay, and the additional interest
and depreciation costs of a larger fleet.
3.2 For other staff who travel in the course of their work, a more sophisticated approach,
which takes account of factors such as their ability to work on route, the fact that part of
their journey time may be at the expense of leisure time and the fact that the length of
7
their journey time may affect their productivity later in the day, is needed. Appropriate
formulae exist, and a number of studies have made estimates of the elements involved.
3.3 Values of commuting and leisure time should be based on empirical evidence, and
segmented by variables such as journey purpose, length, mode and income of travellers,
whenever evidence of significant variation by this variable exists. A large number of
studies, using revealed and stated preference methods, exists, and both methods appear
capable of producing reliable results when used with care.
3.4 The evidence that travelling in congested conditions produces higher values of time than
in uncongested conditions requires particularly careful examination because of its
importance in the current context.
3.5 Empirical estimates also exist, and should be used, for valuing time spent waiting for
public transport, late arrivals and the difference between desired departure time and the
time at which the service actually departs. All may be affected by congestion or scarcity
of slots.
3.6 As a first approximation, values of time for passengers may reasonably be assumed to
increase over time in proportion with income. The value of marginal external cost of
congestion will also need to be updated for changes in traffic volumes, infrastructure
capacity, technology and operating cost.
3.7 Valuations of time for freight consignments for which transit time is increased, or made
more unreliable, are an important component of social costs, and should be based on
empirical estimation (using revealed and/or stated preference methods) rather than the
alternative approach which is in use, which is to make estimates of the interest cost of
stock in transit.
8
INTRODUCTION
The background to this report is the conclusion of the Commission in its White Paper on
infrastructure pricing, and following the deliberations of the High Level Group on infrastructure
pricing, that social marginal cost pricing of infrastructure is the most efficient policy to follow.
However, in order to implement this conclusion, practical methods of estimating social marginal
costs are needed. The scope of this report is to consider alternative ways of estimating the costs
of congestion and scarcity that are relevant for the pricing of existing transport infrastructure.
Detailed consideration of the resulting price structure, and of the implications of this for equity,
or of considerations such as administrative costs of pricing systems and of the implications of
second best theory, are outside the scope of this report. However, it should be noted that for road
charging, because the value of marginal social cost varies with vehicle type, road type and time
of day, full implementation of marginal social cost pricing requires a pricing system that can also
vary in these dimensions. We consider that a combination of electronic road pricing in congested
cities and on congested trunk roads, fuel tax and annual licence duty (and possibly electronic
kilometre-based changes for certain vehicle categories) is capable of achieving a reasonable
approximation to marginal social cost.
In the case of rail, as with other types of transport infrastructure where specific slots are allocated
to particular users, the major issue is not so much congestion as the scarcity value of slots; when
the infrastructure approaches capacity, other users are unable to obtain the slot they want.
Whereas for road, keeping track of the use of the infrastructure by all individual users is a major
problem, for rail the information is readily available to the infrastructure manager. It is therefore
assumed that there is no great practical difficulty in implementing complex pricing structures for
rail infrastructure - including two part tariffs or individually negotiated contracts - if this is
desired.
It is not the task of this paper to review the arguments for and against marginal social cost
pricing. However, we acknowledge that a number of issues other than economic efficiency, such
as distributional effects, implementation costs and acceptability must be considered before such
an approach is implemented. The degree of accuracy to which it is worth reflecting marginal
social cost in price for roads must always be the subject of a cost-benefit analysis, given the
relatively high implementation costs of electronic road pricing. Recent studies have shown net
benefits from road pricing in London of 225m per annum after allowing for implementation
costs (MVA, 1995), and of 2.5 billion francs per year before implementation costs in Paris
(Prud'homme, 1999). Whilst these figures suggest that investments in these measures show very
high benefit-cost ratios relative to other transport investments in large cities, they are an order of
magnitude smaller than the often quoted but irrelevant figure of the total cost of congestion as
2% of GDP.
The environmental and safety implications of road congestion are not considered here as they are
covered in the reports of other working groups..
In contrast to the case of environmental externalities, congestion and scarcity costs are internal to
the transport sector as a whole. They are of concern in the current context not because they are
unpriced, but because existing ways of pricing for the use of transport infrastructure are
inadequate. The result is that additional users of transport infrastructure may well impose
externalities on other transport users, as well as experiencing congestion themselves.
9
It is very important to distinguish between this external element of the cost of congestion or
scarcity and total or average congestion cost. It is sometimes argued that the total cost of
congestion is borne by the sum of users themselves. The point, however, is that, in a situation of
congestion, each additional user inflicts costs on other users as well as suffering costs himself or
herself. It is this external element that inflicted by one user on other users that is relevant for
pricing purposes.
A further reason why estimates of total congestion costs are not helpful for pricing purposes is
that congestion costs vary enormously in time and space. It is thus necessary to prepare estimates
of the marginal external cost of congestion for various circumstances. For roads, we might think
in terms of a three way categorisation e.g.
- location: inner city, outer city, other urban area, inter urban, rural
- type of road: motorway, dual carriageway, single carriageway
- time of day: peak, inter peak, off peak
How many categories to adopt depends on a trade off between the accuracy with which costs are
reflected in prices and the cost and complexity of the pricing instruments.
For railway lines, and other infrastructure on which specific slots are prebooked, no such simple
categorisation is likely to be helpful, as discussed in section 2 below.
A key distinction must be made between infrastructure where access is open to anyone without
prebooked slots, as is the case with roads, and infrastructure where pre-booking of specific slots
is required, such as railways, airports and ports. For roads, increased demand materialises as
more traffic coming on to the road. Beyond a certain level, the result is reduced speeds and
queuing at junctions and other bottlenecks; in other words congestion. In congested conditions,
there is a clear externality involved, as additional road users delay others.
For systems where access takes the form of allocation of specific slots, the situation is very
different. It may still be the case that additional users impose costs on existing users by causing
delays, and indeed delays do become more common the closer the system is to capacity. But the
key point on a scheduled system is that as it approaches capacity some users are unable to get the
slots or run to the schedules they want. As a result, slots acquire scarcity value. Strictly, this is
only an externality where the competition for slots is between different train operating
companies; otherwise it is already internalised. However, with separation of infrastructure from
operations now a part of European Rail Policy, this situation is becoming more common, and
pricing of rail infrastructure therefore needs to take it into account. Of course, congestion may
also take place on trains themselves, leading to a case for taking this into account in prices to
final users, but that is outside the scope of this note.
For both road and rail systems one obvious reaction to shortages of capacity is to invest in new
capacity. One approach to marginal cost pricing would be to try to estimate the costs of capacity
expansion caused by additional traffic. However, this is inclined to vary enormously with the
precise circumstances and the extent of the capacity expansion required. Charging this amount
will also lead to misallocation of the existing infrastructure capacity during the (often long) time
period before the capacity is adjusted.
10
The approach adopted by the Commission, correctly in our view, is to examine the costs of
adding additional traffic to the existing infrastructure. This is consistent with the short run
marginal cost approach taken by Working Group 1, on infrastructure costs, which does not
include the capital costs of infrastructure expansion. It should be stressed however that an
optimal outcome requires that such capacity expansion takes place whenever the benefits of
doing so exceed the costs. The result is that charges will adjust according to the level of capacity;
as investment takes place on a congested road other things being equal, optimal charges will fall,
whereas traffic growth on currently uncongested roads may cause the optimal charge to rise.
This note considers first road congestion, then rail congestion and scarcity values and finally the
crucial issue of value of time. Ports and airports are not considered further, although many of the
same issues arise, and indeed there is a much greater literature on slot allocation at airports than
exists for rail systems.
1. ROAD CONGESTION
External congestion costs occur when the presence of one vehicle increases the journey time of
another. This phenomenon may happen for two distinct reasons as traffic builds up. Firstly,
increased traffic density obliges drivers to drive more slowly simply because the gap between
vehicles is reduced. Secondly, queuing may occur at junctions or other bottlenecks. The standard
way of estimating levels of congestion for inter-urban roads is via the use of speed-flow curves.
Estimated speed/flow curves differ with the characteristics of the road (number of lanes, lane
width, urban or rural etc). Frequently as in the COBA manual (UK DOT, 1996) used for cost-
benefit analysis of road schemes in Britain - they are estimated as a series of linear segments
according to the traffic level on the road as well. The COBA manual suggests speed-flow curves
which consist of two linear segments, with a steeper slope of the segment at higher flow levels
indicating that at higher densities additional vehicles have a greater impact on reducing speeds.
Some difficulties with this relationship are that:
the relationship breaks down near capacity (e.g. for COBA it is undefined below 40
kph);
this is the most important segment for charging purposes it is where charges rise
steeply and the role of charging is most important and effective
1
.
Thus, a more appropriate speed-flow relationship will be non-linear, or have a large number of
linear segments to enable the approximation of non-linear relationships.
Most countries will have estimated speed-flow relationships to suit their own situation in terms
of road characteristics, topography and driving styles, and we do not consider that it would be
sensible to try to adopt a single set of speed flow curves throughout Europe (see Annex A).
Particularly in urban areas, however, heavily congested roads are almost always the result of
bottlenecks. These bottlenecks are frequently due to junctions but also sections with steep
1
For more complex non-linear functions, external congestion pricing will become relevant at around 90% of
capacity; for simpler non-linear functions at around 75% of capacity. Thus, the choice of speed-flow function which
is appropriate to the observed road characteristics is a fundamental issue, particularly in the region where an
equilibrium congestion charge is likely to lie. As an example, HGV speeds begin to be affected only in the highest
peaks.
11
gradients, reduced carriageway width, reduced number of lanes or roadworks or accidents. (The
costs of congestion caused by roadworks or accidents should of course be attributed to the
vehicles that create that condition.)
Bottlenecks require an alternative form of model based on queuing (Small, 1992). Basically
whenever demand exceeds the capacity of the bottleneck, queues build up, and these will only
dissipate when demand falls below capacity. The result is that in a queuing model, vehicles that
arrive when the queue first starts to form impose delays on all subsequent vehicles for the
duration of the queue; the external cost of congestion in this situation declines steadily through
the duration of the queue. The implied pattern of charges is very different from those implied by
the standard speed/flow relationship.
In practice, in a road network, delays occur for a mixture of reasons, and traffic between different
origin/destination pairs interacts at junctions, where the capacity in one direction depends also on
the flows on the other arms of the junction. In addition, traffic may reassign itself to different
routes, so that the delay occurs because of a longer journey distance rather than a reduced speed.
Ideally, the additional congestion created by extra traffic is best measured by running a detailed
traffic simulation model, which assigns traffic to specific roads. An example of such a model is
SATURN, which has been used to estimate the costs of congestion in Cambridge (Newbery,
1998).
Another approach is to approximate the speed-flow relationship for a discrete area of the road
network. The aggregate relationship derived is known as an area speed/flow curve. This
estimates a relationship between the time per kilometre and the overall volume of traffic in an
area of the road network, allowing for the re-routing possibilities in an urban network, and for the
crucial issue of delays at junctions. Such a relationship may be estimated by repeated running of
a traffic simulation model (May et al, 1998); it will of course need modification if the
characteristics of the network change. It is the approach taken to the urban studies in the
TRENEN project.
For the development of a practical charging system for an urban area it may be appropriate to
estimate a number of area speed-flow relationships, for example, working out from the central
core, inner city area and suburban area.
The two main components of delay costs for road users (leaving aside accident and
environmental costs, which may also vary with the level of congestion
2
), are time costs, and
vehicle operating costs represented in terms of fuel, tyres, etc. The derivation of the external
element of time costs from a speed-flow relationship is given below, following Newbery (1990).
External operating costs may be derived in a similar fashion. Formulae are available showing
how vehicle operating costs for various types of vehicles vary with speed (see for example the
COBA manual in the UK). It should be noted that if elements of vehicle operating cost, such as
fuel, are subject to tax in excess of the standard level of Value Added Tax, this should be
deducted from any additional costs, as it does not constitute a true resource cost; rather it is a
transfer between vehicle users and the government concerned. Since they are typically very
small compared with time costs, changes in vehicle operating costs are not considered in detail
here. For public transport and freight vehicles, the time costs should be expanded to consider
2
These are not the focus of this paper, although they are highly relevant for pricing purposes and should be
considered in conjunction with the environmental and accident papers.
12
vehicle utilisation, and the resulting increase in interest and depreciation costs if the fleet needs
to be expanded. Increases in journey time and reliability for the consignments themselves may
also have a value. Other elements of increased operating cost are again small.
The time cost per km of an average vehicle is simply the time per kilometre times the appropriate
value of time. Since the time per kilometre is simply the reciprocal of the speed, this may be
written:
Equation 1
v
b
= t
where v is speed in km/h and b is the value of time for the average vehicle (i.e. it takes into
account factors such as the value of time for the driver and occupants).
The total time costs per kilometre of a flow level q, measured in passenger car units (PCU) per
hour is simply the time cost per kilometre times the flow: T = t * q.
Note that, because traffic flow in this equation is always measured in PCU values, the varying
composition of traffic is automatically allowed for. Similarly, the differing effect of adding
different type of vehicle to the traffic flow is given by multiplying the value of the marginal
external cost of congestion as derived below by the PCU value of the vehicle in question. For
example, the speed/size characteristics of a goods vehicle may mean that it has a PCU factor of 2,
i.e. to convert to charges per vehicle the per PCU charge is doubled. Of course, this requires
knowledge of the relevant set of PCU values to use, and substantial empirical work has derived
factors for use in road traffic modelling.
The marginal cost of an additional vehicle is obtained by differentiating this expression by q,
Equation 2
dq
dt
* q
dq
dT
+ =t
In words, the marginal time cost of an extra vehicle is the cost it incurs itself (t), plus the increase
in the average time cost, multiplied by the number of vehicles incurring that increase.
Differentiating equation 1 gives that the increase in time cost per vehicle is equal to the
proportionate change in speed multiplied by the time (1/v) multiplied by the value of time:
Equation 3
dt
dq
b
v
=
2
dv
dq
and substituting equation 3 into equation 2 gives:
Equation 4
dq
dv
v
b
q t
dq
dT
2
=
Clearly the first element in this equation, t, is the time cost (including congestion) borne by each
user, including the marginal user, themselves (this is what we defined it as in equation 1).
Therefore the second part of Equation 4 represents costs over and above those borne by the
individual (i.e. by other road users), so the marginal external congestion cost (MCT), in terms of
the value of time of other drivers and occupants is:
13
Equation 5
dq
dv
v
b
q MCT
2
=
It is clear that the marginal external cost of congestion will vary with:-
dv
dq
- the slope of the speed/flow relationship, which varies with the type of
road and volume of traffic
q - the volume of traffic (in PCUs)
v - the resulting speed, which varies with the type of road and volume of
traffic
b - the value of time, which varies with the mix of journey purpose and
income of the users
It is therefore necessary to think in terms of varying charges whenever these characteristics of the
road or its traffic vary. DIW et al (1998) survey speed/flow relationships and congestion costs
throughout Europe, and produce estimates of marginal congestion costs for some countries
including the UK (Annex A).
Clearly, weather conditions have a direct impact on the speed-flow characteristics of roads. The
practical implications of this impact raise the possibility of varying charges to take account of
weather conditions; indeed, this currently occurs in San Diego, California, where dynamic tolls
based on traffic levels may be doubled during adverse weather conditions.
Apart from congestion due to the volume of traffic, other causes of congestion can include road
maintenance, vehicle breakdowns and accidents. The delays due to such factors may be
calculated by means of a simple traffic model which takes account of road capacity and
(essentially random) events such as vehicle breakdowns and accidents. In the UK, the QUADRO
model is used to calculate time delays due to road works and incidents in road works leading to
delays. This model is also used to calculate charges to road maintenance companies, lane rental
charges, to build in incentives for timely completion of maintenance. A similar model could be
used in order to calculate appropriate ex-ante vehicle charges.
It is important to note that the optimal congestion charge will be at the point where the marginal
social cost equals the marginal social benefit of travel. Since congestion is currently unpriced,
this will occur at a traffic flow lower than the current flow. Thus a model showing the reaction of
users to alternative levels of congestion and to prices for the use of the road is needed to compute
the optimal congestion charge. The need to calculate congestion charges at this point of
equilibrium implies the need for an iterative process in determination of the optimal charge.
As well as the expected delay, congestion can lead to variability in travel time, or in other words
unreliability. Valuation of unreliability is discussed below, but the bigger difficulty is forecasting
its increase as congestion builds up. Ideally this requires a micro simulation model, which
simulates the journey times of individual vehicles and which can be run for a large number of
14
days, so that distributions of travel time for different levels of traffic may be estimated, or
extensive observations to estimate such a relationship from real data.
15
RECOMMENDATIONS
1.1 Wherever possible, external road congestion costs should be estimated from a model
which simulates the interaction of demand and supply on the road network. The model
can then be used to approximate the marginal external costs of congestion by rerunning it
with small changes in traffic volumes, and examining the effects on journey time for
existing traffic. This model would ideally incorporate a detailed network description, with
both speed/flow relationships and junction delays, and allow for user behaviour in terms
of rerouteing, retiming, changing destination or mode or changing frequency of travel, in
order to obtain a new set of flows and journey times following imposition of a charge.
Data is therefore required on the base O/D matrix, base generalised costs and responses to
changes in these values. The calculation of generalised cost requires knowledge of
operating costs, values of time and vehicle occupancy rates. Only when the charge is
equal to the marginal external cost in this new position has the optimal level of charge
and traffic been found
1.2 Where this is not possible, we recommend that calculations are undertaken for typical
inter urban or rural roads at alternative traffic levels and mixes of types of vehicle using
link speed/flow relationships. Separate calculations will be needed according to the type
of road (number of lanes; motorway or conventional road). Again, data on base traffic
flows and generalised costs are needed, and traffic volumes should again be adjusted for
the introduction of charges, if necessary by means of a simple price elasticity of demand,
in order to obtain an equilibrium value.
1.3 For urban areas, the degree of interaction between roads means that such an
approximation will be particularly crude. If a full network model is not available, the use
of area speed/flow relationships relating to the entire network for central, inner and outer
urban areas is likely to be preferable to link based speed/flow relationships.
1.4 Forecasting the impact of increased traffic on unreliability is more difficult, but given the
importance of the issue it should be attempted wherever possible. A variety of approaches
exists, including the use of micro-simulation models which model individual vehicles and
can thus estimate the spread of journey times, and purely empirical approaches, which
require data on unreliability and on traffic flows for a set of roads over time.
1.5 All the above relationships should relate to local conditions in the area concerned, and
relate to conditions such as driving styles and typical speeds in that location. It would be
counter-productive therefore to attempt to specify Europe-wide relationships, although
results may with care be transferred from comparable situations elsewhere in Europe if
local information is not available.
2. CONGESTION AND SCARCITY COSTS OF RAIL
In principle, the approach we take to estimating the social marginal cost of rail traffic is
consistent with that for road - namely, we try to estimate the additional costs imposed on society
by the use of the infrastructure by an additional train. However the methodology for rail is quite
different from that for road transport. The reason is that for rail the volume of traffic is directly
controlled by allocation of slots, so capacity should never be exceeded. Nevertheless, as traffic
approaches capacity, so delays become more frequent. Where one operator delays trains of
another through unscheduled departure from the timetable, compensation may be paid directly
16
for increased costs and passengers time, provided that adequate records are kept of amounts and
causes of delay. This is a feature of the performance regime embodied in track access
agreements in Great Britain. However, this is only likely to measure the delays directly caused
by the train in question. Simulation modelling, using a model such as the MERIT model used by
Railtrack, which simulates a large number of days operations using probability distributions of
the various causes of delay, will estimate the full effect of running additional trains, including the
worsening of delays from other causes by the reduction of the recovery margin in the system.
But the main consequence of full utilisation of capacity is that users simply cannot get the
capacity they want when they want it; they have to run their trains at times and possibly speeds
different to their preferred alternative, or to give up the journey.
The carrying capacity of a railway link is the maximum number of physical transport units which
can use the link, and can be expressed as a function of the number of tracks in a section, average
train speeds, geometry, signalling and safety systems, section lengths, length of trains, etc.
(Rothengatter et al, 1996). However, over and above all these factors, the mix of train speeds and
the precise order in which trains are run is crucial. For instance, on a predominantly high speed
line an additional slow freight train may remove the paths of several high speed passenger trains;
on a heavy freight route the reverse may be true. Capacity is also maximised by grouping trains
of like speeds, so that a 'flight' of fast passenger trains is followed by a 'flight' of slow freights
and vice versa. However, this conflicts with providing a good service of well spaced trains at
regular intervals for the public. More complicated still is the interaction of trains on different
routes or between different origins and destinations; as with roads, junctions and other
bottlenecks (e.g. speed restrictions) are key factors determining capacity.
The result of all these considerations is that it is impossible to come to a ready definition of the
capacity of a rail route corresponding to that for roads. More seriously, the impact of an
additional train of a particular type on the paths available to other trains will differ enormously
according to the precise mix of traffic on the line. At the same time, the value of a slot to other
commercial operators or to government bodies providing social services will also differ
enormously in time and space. It does not therefore seem possible to come to a general
methodology to estimate scarcity values for rail slots in a variety of typical circumstances, in the
way in which we have for road.
There are other ways of seeking to derive scarcity values. One is by competitive bidding for the
slots. However, in rail systems capacity can be used in such a wide variety of ways to produce
different mixes of trains of different types, origins and destinations that any bidding exercise is
likely to be very complex. Moreover the value of a particular slot depends very much on what
other slots are obtained, in order to put together a commercially attractive service. We see some
scope for bidding processes for alternative packages of slots in a pre-planned timetable, but in
general we do not consider bidding processes as a practical way of revealing scarcity values.
We see it as inevitable that the use of rail infrastructure will be planned by the infrastructure
manager rather than being determined by a purely market process. The most efficient mechanism
for revealing scarcity values is probably to allow the infrastructure manager to negotiate with the
potential users about their willingness to pay for alternative slots in determining that plan. This
allows for negotiation over desired packages of access rights and iteration to improve upon the
initial solution. It might work in terms of train operators first registering their wishes, the
infrastructure manager using these to produce packages of paths and charges and further
negotiation then taking place to determine whether operators would be prepared to pay more to
17
improve their package, or to surrender some of their paths in return for a reduced charge. Such
negotiations would also naturally encompass investment in expanded or enhanced capacity and
the sharing of the development costs.
It does raise fears that the infrastructure manager or the larger train operating companies may
exert undue monopoly power over the process (particularly when the two are part of the same
organisation), and calls for an independent regulator to intervene where that happens. However,
in a situation in which there is no ideal solution, it does appear to be the best way forward.
RECOMMENDATIONS
2.1 Estimation of the scarcity value of specific slots on rail infrastructure requires a way of
revealing the value placed on the slots by alternative possible users, both in terms of
commercial rail operators and in terms of government bodies wishing to provide social
services. It may be possible in some cases to reveal these values by auctioning the slots,
but given the complexities involved in terms of the alternative ways in which the
infrastructure may be used, this is difficult. Some pre-packaging of slots is probably
necessary, in order to offer attractive combinations to alternative bidders. In general, a
process of negotiation appears the most practicable way forward. This might work in
terms of train operators first registering their wishes, the infrastructure manager using
these to produce packages of paths and charges and further negotiation then taking place
to determine whether operators would be prepared to pay more to improve their package,
or to surrender some of their paths in return for a reduced charge. Such negotiations
would also naturally encompass investment in expanded or enhanced capacity and the
sharing of the development costs.
2.2 Unscheduled delays imposed by one train operator on another may be measured ex post if
adequate monitoring is undertaken to measure both the extent and the cause of delays.
However, this will only measure the delays directly caused and not those where the
presence of the additional train has worsened the consequences of other delays by
absorbing part of the recovery margin. It is therefore more accurate to measure anticipated
delays by simulation modelling and charge these as part of the tariff. Of course these
additional delays will vary by route, type of train and time of day.
3. MONETARY VALUATION
In general it is found that the external costs of congestion are dominated by time losses (rather
than effects on operating costs) and thus are extremely sensitive to the value of time. In order to
estimate the marginal cost of congestion, it is necessary to choose monetary valuations to be used
in the evaluation of increased journey time and congestion in both the passenger and freight
markets. Valuations are required for private car, bus, van, HGV, train, and aircraft.
In addition, disaggregations by journey purpose for passenger travel would be desirable, so that
the differing mix of journey purposes by time and location can be allowed for. A distinction
according to the duration of trip may also be made since it could well have a bearing on time
constraints and disutility of the journey.
18
The value of time represents the maximum amount an individual is prepared to pay for a time
saving or the minimum acceptable amount to compensate for an increase in journey time. The
marginal value of time is made up of two components:
the marginal utility of time; and,
the marginal utility of money.
Hence variations in the value of time can arise from variations in the marginal utility of time or
in the marginal utility of money. Variations in the former will depend on the conditions in which
travel time is spent and on the opportunity cost of travel time. Variations in the latter depend on
personal characteristics and particularly on income. Congestion may impact on the conditions of
travel and thus may affect the value of time.
For car, van and HGV users, the impacts of increased congestion can be expected to be:
increases in the average journey time, with the additional time being spent in congested
traffic which can be expected to be more highly valued than time spent in free flow
traffic because of the greater stress, frustration and unpleasantness involved; and,
increases in the variability of travel time, with a wider distribution around average travel
time. This is measured by the standard deviation of travel times or a proportion of
vehicles arriving late.
For public transport users, the situation is a little different because of the fixed schedules
involved. Congestion may again lead to increases in both the mean and the variability of in
vehicle time, but it will also lead to more waiting time, as vehicles became bunched and fail to
arrive at stops on schedule.
For values of time it is desirable to use values selected for the specific context concerned as far
as possible, since values of time will vary with income and other characteristics of the travellers
concerned. For many purposes this means that values estimated for the country concerned will
be the best choice. On international routes, it will be necessary to use a weighted average of the
values for the countries from which the travellers come.
There are established values of working time for passenger travel, based on the cost of employing
workers, (which reflects the wage rate and an overhead). These are appropriate for workers
whose job is actually driving. However, this approach may not be appropriate where passengers
may work en route, as is the case of many business travellers when travelling by rail, or where
much of the journey takes place outside normal working hours. Hensher (1977) outlines an
approach which may be used to estimate values empirically in such circumstances, but few
countries have applied this, perhaps because many of the elements are difficult to value in
practice. However, a number of studies are available which have taken this approach, using
surveys to quantify elements such as the gain in subsequent productivity as a result of faster
journeys leading to the traveller arriving at a meeting feeling more refreshed. The equation to be
estimated is the following:
Equation 6 V
B
= (1 r p.q)MP + (1 r)v
w
+ rv
l
+ MP
F
where:
V
B
= value of business time savings
19
r is share of saved time used for leisure
p is share of saved time used productively
q is relative productivity of time saved that was used for work
MP is marginal productivity of labour
v
w
is employee value of saved time otherwise spent in work
v
l
is employee value of saved time otherwise spent in leisure
MP
F
is the value of increased productivity from reduced fatigue
For passenger travel in non-working time, the general approach is to use behavioural values
estimated from either revealed preference or stated preference data. In general, revealed
preference models have the benefit of being based on actual data, but stated preference models
allow much more precise estimation with a given sample size. There is some evidence that a well
designed stated preference survey is relatively free from bias, but this issue remains
controversial. A fair amount of evidence suggest that the value of time for leisure purposes is
related to income, and a figure of the order of 25% of the wage rate is often used.
There is evidence that inter-urban travel is valued higher than urban travel, on which most values
of time are based. Wardman (1998) provides this evidence on the basis of regression analysis of
the inter-urban and urban values of time of over 100 revealed preference and stated preference
studies. The principal reason for this distinction appears to be the higher relative income levels of
inter-urban travellers. Time spent in congested road conditions is also valued more highly,
because of the discomfort of driving in stop-start conditions and the uncertainty associated with
the journey.
For rail, the appropriate values for in-vehicle time should be used for scheduled travel time, for
late time for delays and for wait time for any additional waiting time involved. Where scarcity
of capacity requires a change in departure time, the value per minute of switch in departure time
is also needed. All of these values may in principal be estimated by either revealed or stated
preference methods, although in practice their estimation by stated preference methods is much
more straightforward, as appropriate trade-offs may be postulated without having to find them in
practice.
For freight transport, values of time need to take account not just of the additional operating costs
and drivers wages caused by congestion, but also of the value to the consignor of receiving the
goods more quickly and - perhaps even more important - more reliably. A traditional approach is
to look at the value of the goods, and then consider the interest charges on the additional time the
stock is in transit. Except for very long journeys (e.g. inter continental sea transits) this gives very
low valuations. Empirical evidence suggests that consignors typically value journey time more
highly, presumably because slow transits make it more difficult to cope with short term
fluctuations in demand, and thus increase the amount of buffer stock held in the system. But
much more important still is reliability. Given the spread of just in time distribution and
production systems, failures in reliability may be very expensive, in again leading to additional
buffer stock or to failures to supply goods at the promised time. Obviously all these values will
vary with the commodity but also with other factors. For instance the values of time and
reliability for transits to depots holding buffer stock are typically below those applying to the
journey to the final customer.
Again values of time for freight may be estimated by either revealed or stated preference
analysis. There is a growing trend towards the use of stated preference analysis, because
20
appropriate revealed preference data is hard to obtain (both due to issues of confidentiality and
because of a lack of real alternatives for a lot of freight - road is dominant on all criteria).
Current evidence is not conclusive but suggests that values of time may rise over time in
proportion to income. Of course, this is not the only factor that needs updating over time;
changes in traffic levels, infrastructure capacity and operating costs all also need to be taken into
account.
RECOMMENDATIONS
3.1 For staff working in the transport industry, the usual approach of estimating the marginal
cost of their time as their wage rate plus an allowance for the overhead costs of
employing labour is generally appropriate. Similarly, the costs of poorer vehicle
utilisation may be estimated by calculating the impact on fleet size of the delay, and the
additional interest and depreciation costs of a larger fleet.
3.2 For other staff who travel in the course of their work, a more sophisticated approach,
which takes account of factors such as their ability to work on route, the fact that part of
their journey time may be at the expense of leisure time and the fact that the length of
their journey time may affect their productivity later in the day, is needed. Appropriate
formulae exist, and a number of studies have made estimates of the elements involved.
3.3 Values of commuting and leisure time should be based on empirical evidence, and
segmented by variables such as journey purpose, length, mode and income of travellers,
whenever evidence of significant variation by this variable exists. A large number of
studies, using revealed and stated preference methods, exists, and both methods appear
capable of producing reliable results when used with care.
3.4 The evidence that travelling in congested conditions produces higher values of time than
in uncongested conditions requires particularly careful examination because of its
importance in the current context.
3.5 Empirical estimates also exist, and should be used, for valuing time spent waiting for
public transport, late arrivals and the difference between desired departure time and the
time at which the service actually departs. All may be affected by congestion or scarcity
of slots.
3.6 As a first approximation, values of time for passengers may reasonably be assumed to
increase over time in proportion with income. The value of marginal external cost of
congestion will also need to be updated for changes in traffic volumes, infrastructure
capacity, technology and operating cost.
3.7 Valuations of time for freight consignments for which transit time is increased, or made
more unreliable, are an important component of social costs, and should be based on
empirical estimation (using revealed and/or stated preference methods) rather than the
alternative approach which is in use, which is to make estimates of the interest cost of
stock in transit.
21
4. CONCLUSIONS
Externalities occur between vehicles when the presence of one vehicle increases the journey time
of another. The marginal external congestion cost of additional traffic comprises the additional
time costs and vehicle operating costs imposed on others by an additional unit of travel. For
roads the calculation of additional travel time can be done using speed flow relationships, for the
country or corridor in question. However, the way in which traffic interacts in a network of roads
means that really a network simulation model is needed, particularly in urban areas; the results of
this may be approximated by an area/speed flow relationship. It is also necessary to remember
that the optimal charge represents the marginal external congestion cost at the optimum; it is thus
necessary to model the reactions of traffic to changes in road user charges and levels of
congestion to find the appropriate charge.
For rail, there are important differences. Use of the system is controlled through the allocation of
slots and the main consequence of full capacity is the inability to run the train when desired. The
costs of congestion and scarcity are only external when imposed by one train operator on another;
costs imposed on other trains of the same operator are already internalised. The complexities of
rail systems are such that no simple formula can be found to estimate scarcity values of slots for a
variety of typical circumstances. We recommend that negotiation between infrastructure
managers and train operating companies is the best way to reveal scarcity values of rail slots.
For values of time we recommend the use of values related to local conditions as far as possible.
There are established values of working time for passenger travel, based on the cost of employing
workers, (which reflects the wage rate and an overhead.), but for business travel the alternative of
empirical estimation is to be preferred. For passenger travel in non-working time, values based
on analysis of revealed or stated preference data and disaggregated in a number of dimensions
such as purpose, trip length and degree of congestion, are needed. There is evidence that inter-
urban travel is valued more highly than urban travel, on which most of the values of time are
based. Time spent in road congestion is also valued more highly. For rail and air public transport
the appropriate values for delay/late and wait time are also required.
For freight transport the value of time should cover not just drivers wages and operating cost
savings, but also the value to the consignor of receiving the goods more quickly and/or reliably.
Freight values of time vary by commodity and other relevant factors, although a simple mean will
often suffice.
A comprehensive survey of the evidence on values of time for passenger and freight traffic,
including all the separate values discussed above, is given in PETS deliverable D7 (See Annex
B).
REFERENCES
Department of Transport:
The COBA Manual. HMSO, 1996
DIW et al (1998):
22
Infrastructure, Capital, Maintenance and Road Damage Costs for Different Heavy Goods
Vehicles in the EU.
EURET:
Cost-Benefit and Multi-Criteria Analysis for New Road Construction Phase II. Draft Report on
Review Measurement Methods for Existing Criteria. Prepared for Commission of the European
Communities, 1993.
EURET:
Cost-Benefit and Multi-Criteria Analysis for New Road Construction: Final Report. Report to the
Commission of the European Communities, Directorate General for Transport, 1994. Doc
EURET/385/94.
EVA Consortium:
Evaluation Process for Road Transport Informatics: EVA-Manual. Prepared for the Commission
of the European Communities, 1991.
Fowkes, AS, Nash, C.A. and Tweddle G.:
Valuing the Attributes of Freight Transport Quality: Results from the Stated Preference Surveys.
Working Paper 276, Institute for Transport Studies, University of Leeds, 1989.
Hensher DA (1977):
Value of Business Travel Time. Pergamon Press, Oxford.
May AD, Shepherd SP, Bates JJ (1998, forthcoming)
Supply Curves for Urban Road Networks
The MVA Consultancy (1995) the London Congestion Charging Research Programme:
Principal Findings, Government Office for London, HMSO.
Newbery, D (1990):
Pricing and Congestion: Economic Principles Relevant to Pricing Roads", Oxford Review of
Economic Policy, vol 6, 22-38, 1990
Newbery, D. (1998)
PETS D7 (1998)
Internalisation of Externalities, University of Leeds, 1998.
Prudhomme, R (1999) Potential Welfare Gains of Road Congestion Pricing (unpublished note).
Rothengatter W et al (1996)
Bottlenecks in the European Transport Infrastructure. Technical Report. Study on behalf of the
European Centre for Infrastructure Studies (ECIS). April, 1996
Small KA (1992)
Urban Transportation Economy, Harwood Academic Publishers, Chur.
23
TransPrice:
Evaluation Framework Proposal. Annex A.5, Deliverable 2 - Common Analytical Framework.
Prepared for European Commission, 1996.
Tweddle G, Fowkes AS and Nash CA
Impact of the Channel Tunnel: A Survey of Anglo-European Unitised Freight. Working Paper
443, Institute for Transport Studies, University of Leeds, 1995.
Wardman M
Route Choice and the Value of Motorists' Travel Time: Empirical Findings. Working Paper 224,
Institute for Transport Studies, University of Leeds, 1986.
Wardman M
A Review of Evidence on the Value of Travel Time in Great Britain. Working Paper 495.
Institute for Transport Studies, University of Leeds, 1997.
24
ANNEX A: EXAMPLES OF SPEED FLOW RELATIONS AND
CONGESTION COSTS
A.1 Speed-Flow Relationships
This annex gives examples of speed-flow relationships for inter-urban motorways. These relate
to Austria, Germany, the UK and the USA.
A1.1 Austrian Example (from DIW et al, 1998)
The following formula was referenced in DIW et al. 1998:
V = v
k
+ (v
G
- v
k
)*(1-a)

with: a traffic volume / capacity

v
k
speed, if a --> 1:
dual carriage way 55 km/h
normal road 50 km/h if > 3.00 m lane width
45 km/h if < 3.00 m lane width
v
G
speed, if a --> 0
A1.2 Example of Speed-Flow Relationships fromthe German EWS Manual
3
The function that relates to car speeds in free-flow conditions can be expressed as:
V
P
= c
0
+ c
1
*exp(c
2
*(Q
P
+ 2*Q
G
)))
where:
V
P
: Speed passenger car
Q
P
: Flow passenger car
Q
G
: Flow heavy goods vehicle
c
0
: Constant, depending on road type, gradient and curvature
c
1
,c
2
: Constant, depending on road type (c
1
<0, c
2
>0)
3
EWS (1997) Forschungsgesellschaft fr Straen- und Verkehrswesen, 1997:
Empfehlungen fr Wirtschaftlichkeitsuntersuchungen an Strassen (EWS). Aktualisierung der RAS-W '86, Kln,
1997. This text provided by Michael Schoch, IWW, Karlsruhe University).
25
The slope of this function, a fundamental part of the marginal external congestion cost function is
then given by:
dV
P
/dQ
P
= c
1
* c2 * exp(c
2
* (Q
P
+2*Q
G
))
A1.3 UK COBA Curves
This was the type of relationship adopted in the DIW et al (1998) study. It is referenced in PETS
D7 (1998), from which the following text comes. The following examples relate to light vehicles
only.
Speed of a light vehicle on a two-lane motorway.
The speed-flow relationship provided by COBA is of the form v = a - q. When the flow level is
less than 1200 veh/hour/lane the following expression provides speed:
v =107 - 0.006 Q
The discontinuity point is given at Q=1200 veh/hour/lane. When flow levels are higher than 1200
vehs/hour/lane then
v =139 - 0.033 Q
Note that flow Q in this relationship is measured in vehs/hour/lane, and speed v in km/hour`.
Table A1.1: Parameters of the Speed-Flow Relationships for Light vehicles
Type of Road Q
B
(veh/hour/lane)
Constant
a before Q
B
Constant
a after Q
B
before
Q
B

after
Q
B
Rural single carriageways 80% of capacity
1
77 - 0.015
2
0.050
2
All-purpose dual carriageways 1080 Two lane: 103
Three lane: 110
Two lane: 132
Three lane: 139
0.006 0.033
Motorways 1200 Two lane: 107
Three lane: 114
Four lane: 114
Two lane: 139
Three lane: 147
Four lane: 147
0.006 0.033
1 The capacity of rural single carriageways depends on road width and the percentage of heavy vehicles.
2 Assuming a 15% of heavy vehicles.
A1.4 USA Freeway Estimates (Highway Capacity Manual, 1994)
For basic freeway sections, the manual gives speed-flow curves which are characterised by:
flows from zero to around 75% of capacity - a very flat, almost linear;
flows from around 75% of capacity non-linear, with slope increasing;
26
A2 Examples of Marginal External Congestion Charge Estimates
A2.1 Estimates for different European Countries (DIW et al, 1998)
The DIW et al (1998) study made use of the UK COBA speed-flow relationship, and in
estimating the marginal external congestion costs took account of variation by country in the
value of time and other key attributes such as road capacity.
0,00
0,02
0,04
0,06
0,08
0,10
0,12
0,14
0,16
0 1200 1400 1600 1800 2000
Marginal congestion costs 1994: Interurban motorways
- Passenger cars -
ECU
CH
F, L
B, DK,
I, E,
SF,
A
D, GR, NL,
Flow rate in vehicles per lane
27
Marginal congestion costs 1994: Interurban motorways
- Goods vehicles -
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0 1200 1400 1600 1800 2000
ECU
CH
F, L
B, DK,
I, E,
SF,
A
D, GR, NL,
Flow rate in vehicles per lane
:
28
A2.2 USA Estimates, for 2000 (HCAS, 1997
4
)
The following table summarises values estimated for 2000 (cents per mile). It provides a clear
indication of the range of variation in charges with:
type of vehicle;
type of road; and,
high, medium and low levels of estimates (according to traffic volume).
Rural Highways Urban Highways All Highways
High Middl
e
Low High Middle Low High Middle Low
Automobiles 3.76 1.28 0.34 18.27 6.21 1.64 13.17 4.48 1.19
Pickups and
Vans
3.80 1.29 0.34 17.78 6.04 1.60 11.75 4.00 1.06
Buses 6.96 2.37 0.63 37.59 12.78 3.38 24.79 8.43 2.23
Single Unit
Trucks
7.43 2.53 0.67 42.65 14.50 3.84 26.81 9.11 2.41
Combination
Trucks
10.87 3.70 0.98 49.34 16.78 4.44 25.81 8.78 2.32
All Vehicles 4.40 1.50 0.40 19.72 6.71 1.78 13.81 4.70 1.24
4
Federal Highway Cost Allocation Study (1997)
29
ANNEX B: TYPICAL VALUES OF TIME (PETS D7)
PETS D7 quotes the following values of working time. However, these are based on the wage
rate, and should not be used for business travel for the reasons discussed above.
Country Value Country Value Country Value
Belgium 23.06 Ireland 18.69 Spain 11.15
Denmark 20.81 Italy 22.60 UK 17.63
France 26.44 Luxembourg 21.40 Finland 20.36
Germany 26.44 Netherlands 21.45 Sweden 22.92
Greece 6.90 Portugal 5.54 Average 21.02
Norway 18.4 ECU (152.50 Kr): Source Handbook 140, Public Roads Administration
Norway
Figures were updated taking into consideration changes in GDP in the 15 countries of the European Union and the rates of inflation published
by the OECD
For non working time, most empirical evidence suggests values of the order of 25% of the wage
rate. There is evidence that inter-urban travel is valued more highly than urban travel, on which
most of the values of time are based., with typical evidence suggesting that the value of time
should be increased by 60%.
Time spent in road congestion is also valued more highly, with evidence suggesting that
congested time for passengers only is valued at 150% of normal time. For rail and air public
transport estimates of the appropriate values for wait time, departure time shifts and late time are
also recommended.
For freight transport the value of time should cover not just drivers wages and operating cost
savings, but also the value to the consignor of receiving the goods more quickly and/or reliably.
Typical studies suggest that a value of around 37 ECU's per hour should be used for LGV's. The
value of time for HGV's appears to be 10% more, ECU per hour, whilst the value of time for rail
should be only 25% that for LGV. If the proportion of the driver costs of these recommended
values can be established, then different values across countries can be used according to
variations in drivers' wage rates. Freight values of time by commodity are also given.
Reliability is generally highly valued by freight operators. Evidence was found that a value of
reliability (defined as a percentage of on time arrivals.) of 5% of the freight rate for a 1%
improvement in reliability index used in Tweddle et al. (1995) was appropriate. The indices
which most closely relate to the situations prevailing before and after the increased congestion
would be used.
d:\ts\capri\hlg\can0507.doc