110

TRAFFICCONGESTION

IN EUROPE

EU

RO

PE

AN

CO

NF

ER

EN

CE

OF

MI

NI

ST

ER

S O

F T

RA

NS

PO

RT ECONOMIC RESEARCH CENTRE

ROUNDT A B L E

OECD, 1999.

Software: 1987-1996, Acrobat is a trademark of ADOBE.

All rights reserved. OECD grants you the right to use one copy of this Program for your personal use only.Unauthorised reproduction, lending, hiring, transmission or distribution of any data or software is prohibited.You must treat the Program and associated materials and any elements thereof like any other copyrightedmaterial.

All requests should be made to:

Head of Publications Service,OECD Publications Service,2, rue Andre-Pascal, 75775 ParisCedex 16, France.

ECONOMIC RESEARCH CENTRE

REPORT OF THEHUNDRED AND TENTH ROUND TABLE

ON TRANSPORT ECONOMICS

held in Paris on 12th-13th March 1998on the following topic:

TRAFFIC CONGESTIONIN EUROPE

EUROPEAN CONFERENCE OF MINISTERS OF TRANSPORT

EUROPEAN CONFERENCE OF MINISTERSOF TRANSPORT (ECMT)

The European Conference of Ministers of Transport (ECMT) is an inter-governmentalorganisation established by a Protocol signed in Brussels on 17 October 1953. It is a forum inwhich Ministers responsible for transport, and more specifically the inland transport sector, canco-operate on policy. Within this forum, Ministers can openly discuss current problems and agreeupon joint approaches aimed at improving the utilisation and at ensuring the rational developmentof European transport systems of international importance.

At present, the ECMT’s role primarily consists of:– helping to create an integrated transport system throughout the enlarged Europe that is

economically and technically efficient, meets the highest possible safety and environmentalstandards and takes full account of the social dimension;

– helping also to build a bridge between the European Union and the rest of the continent at apolitical level.

The Council of the Conference comprises the Ministers of Transport of 39 full Membercountries: Albania, Austria, Azerbaijan, Belarus, Belgium, Bosnia-Herzegovina, Bulgaria, Croatia,the Czech Republic, Denmark, Estonia, Finland, France, FYR Macedonia, Georgia, Germany,Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, Moldova, Netherlands,Norway, Poland, Portugal, Romania, the Russian Federation, the Slovak Republic, Slovenia, Spain,Sweden, Switzerland, Turkey, Ukraine and the United Kingdom. There are five Associate membercountries (Australia, Canada, Japan, New Zealand and the United States) and three Observercountries (Armenia, Liechtenstein and Morocco).

A Committee of Deputies, composed of senior civil servants representing Ministers, preparesproposals for consideration by the Council of Ministers. The Committee is assisted by workinggroups, each of which has a specific mandate.

The issues currently being studied – on which policy decisions by Ministers will be required –include the development and implementation of a pan-European transport policy; the integration ofCentral and Eastern European Countries into the European transport market; specific issues relatingto transport by rail, road and waterway; combined transport; transport and the environment; thesocial costs of transport; trends in international transport and infrastructure needs; transport forpeople with mobility handicaps; road safety; traffic management; road traffic information and newcommunications technologies.

Statistical analyses of trends in traffic and investment are published regularly by the ECMTand provide a clear indication of the situation, on a trimestrial or annual basis, in the transportsector in different European countries.

As part of its research activities, the ECMT holds regular Symposia, Seminars and RoundTables on transport economics issues. Their conclusions are considered by the competent organs ofthe Conference under the authority of the Committee of Deputies and serve as a basis forformulating proposals for policy decisions to be submitted to Ministers.

The ECMT’s Documentation Service has extensive information available concerning thetransport sector. This information is accessible on the ECMT Internet site.

For administrative purposes the ECMT’s Secretariat is attached to the Organisation forEconomic Co-operation and Development (OECD).

Publie en francais sous le titre :LA CONGESTION ROUTIERE EN EUROPE

Further information about the ECMT is available on Internet at the following address:http://www.oecd.org/cem/

ECMT 1999 – ECMT Publications are distributed by: OECD Publications Service,2, rue Andre Pascal, 75775 PARIS CEDEX 16, France.

3

TABLE OF CONTENTS

INTRODUCTORY REPORTS

GERMANY

SCHALLABÖCK, K.-O. and PETERSEN, R......................................5

FRANCE

GERONDEAU, C...............................................................................45

NETHERLANDS

BOVY, P. and SALOMON, H. ..........................................................85

UNITED KINGDOM

DARGAY, J.M. and GOODWIN, P.B. ............................................155

OTHER CONTRIBUTIONS ........................................................................201

SUMMARY OF DISCUSSIONS

(Round Table debate on reports)......................................................................215

LIST OF PARTICIPANTS ...........................................................................231

4

5

GERMANY

Karl Otto SCHALLABÖCKRudolf PETERSEN

Wuppertal Institut für Klima, Umwelt, EnergieWuppertalGermany

6

7

SUMMARY

1. INTRODUCTION........................................................................................9

2. TRAFFIC CONGESTION AND ITS CAUSES ........................................13

2.1. Definition............................................................................................132.2. Kinds and causes of congestion..........................................................142.3. The role of vehicle speeds ..................................................................152.4. Road construction, accidents and other causes ..................................17

3. VOLUME OF TRAFFIC CONGESTION.................................................18

3.1. Introduction ........................................................................................183.2. Traffic flow on various road types .....................................................193.3. Further differentiation of traffic conditions .......................................243.4. Differentiation according to vehicle category ....................................263.5. Conclusion ..........................................................................................27

4. ASSESSMENT OF CONGESTION EFFECTS ........................................31

4.1. Overview.............................................................................................314.2. Environmental consequences .............................................................314.3. Economic consequences .....................................................................344.4. Social and other consequences ...........................................................364.5. Conclusion ..........................................................................................37

8

5. STRATEGY FOR THE REDUCTION OF CONGESTION ANDITS NEGATIVE IMPACTS ......................................................................38

BIBLIOGRAPHY ..............................................................................................41

Wuppertal, January 1998

9

1. INTRODUCTION

During the last decade, the European Union faced a rapid increase inmotorised transport, especially in road and air transport. This increase isexpected to continue in the near future. A recent study foresees the followingtrend development (see Table 1), all transport modes.

Table 1. Scenario of transport development in Europe (1)

Transport Mode TransportPerformance

1995

Growth1970-1995

Growth1995-2020

Passenger Billion PkmCars (2) 3590 125 % 50 %Aircraft 400 250 % 200 %Buses (2) 370 50 % 30 %Rail 290 40 % 20 %

Freight Billion tkmTrucks 1150 160 % 100 %Rail 240 – 5 % 0 %Inland Shipment 120 10 % 10 %

(1) EU 15 + Norway, Switzerland, Turkey;(2) Without Turkey.

Source: Dreborg, Karl et al.: Images of Future Transport in Europe; FinalDraft Sept. 1997; p. 15; project funded by the European Commissionunder the Transport RTD Programme of the 4th FrameworkProgramme.

10

One of the driving forces in road transport is the increase in the passengercar and truck fleets (see Table 2). The road infrastructure in the Europeancountries has not been extended in a comparable way (Table 3) which leads to areduction in available road space per passenger car and per truck in allEuropean countries (Table 4).

Table 2. Development of the vehicle stock in the European countries(in millions)

Area Cars Trucks1975 1985 1995 1975 1985 1995

Austria 1.720 2.469 3.480 0.146 0.233 0.290

Belgium 2.577 3.300 4.239 0.235 0.348 0.433

Denmark 1.297 1.440 1.685 0.225 0.253 0.275

Finland 0.996 1.474 1.901 0.128 0.194 0.272

France 15.520 20.800 25.100 2.150 3.310 4.926

Germany (1) 18.161 25.378 40.499 1.231 1.844 2.251

Greece 0.414 1.155 2.205 0.197 0.589 0.884

Ireland 0.511 0.711 0.955 0.052 0.092 0.121

Italy 15.061 20.888 31.700 1.128 1.792 5.050

Luxembourg 0.117 0.152 0.229 0.011 0.014 0.015

Netherlands 3.400 4.818 5.633 0.316 0.405 0.578

Norway 0.954 1.430 1.685 0.138 0.214 0.350

Portugal 0.873 1.136 2.560 0.168 0.346 0.866

Spain 4.807 8.874 14.212 1.001 1.610 2.937

Sweden 2.760 3.081 3.631 0.157 0.224 0.308

Switzerland 1.794 2.552 3.229 0.167 0.204 0.262

United Kingdom 13.747 19.088 24.307 1.776 2.718 3.157

Total 84.709 118.745 167.249 9.227 14.388 22.975

(1) 1975 and 1985 West Germany, 1995 East and West Germany.

Source: List of data by Holger Dalkmann, Wuppertal Institute, IRF 1978,IRF 1987, VDA 1990 and IRF 1997; and WI calculation.

11

Table 3. Development of road length in the European countries(in thousand km)

Area Highways Road Network1975 1985 1995 1975 1985 1995

Austria 0.651 1.261 1.596 102.787 104.461 130.023

Belgium 1.018 1.456 1.666 93.596 127.956 143.175

Denmark 0.345 0.593 0.830 66.137 70.093 71.420

Finland 0.180 0.205 0.394 73.552 76.105 77.723

France 3.401 6.438 9.140 794.690 805.038 892.700

Germany 6.200 8.350 11.190 464.000 491.250 641.860

Greece 0.091 0.092 0.420 96.573 108.092 116.440

Ireland 0.000 0.008 0.070 89.005 92.408 92.430

Italy 5.431 5.955 8.860 291.081 299.255 314.360

Luxembourg 0.025 0.058 0.123 4.465 5.258 5.136

Netherlands 1.530 1.975 2.300 87.582 112.775 120.800

Norway 0.165 0.074 0.105 77.101 85.774 90.366

Portugal 0.066 0.196 0.687 46.241 63.996 68.732

Spain 1.135 1.977 7.747 284.532 311.777 343.197

Sweden 0.692 0.939 1.231 125.490 130.639 137.464

Switzerland 0.662 1.054 1.540 61.635 70.654 71.055

United Kingdom 2.026 2.838 3.200 343.088 348.338 370.300

Total 23.618 33.469 51.099 3101.555 3303.869 3687.181

(1) 1975 and 1985 West Germany, 1995 East and West Germany.

Source: List of data by Holger Dalkmann, Wuppertal Institute, IRF 1978,IRF 1987, VDA 1990 and IRF 1997; and WI calculation.

12

Table 4. Development of the available road length per vehiclein the European countries (in metres)

Area Per car Per truck1975 1985 1995 1975 1985 1995

Austria 59.8 42.3 37.4 703.8 449.3 447.9

Belgium 36.3 38.8 33.8 397.7 368.2 331.0

Denmark 51.0 48.7 42.4 293.7 277.4 259.7

Finland 73.8 51.6 40.9 572.6 392.3 285.4

France 51.2 38.7 35.6 369.6 243.2 181.2

Germany 25.5 19.4 15.8 376.8 266.4 285.1

Greece 233.2 93.6 52.8 490.7 183.5 131.7

Ireland 174.3 130.0 96.8 1699.6 1009.9 763.9

Italy 19.3 14.3 9.9 258.1 167.0 62.2

Luxembourg 38.3 34.7 22.4 423.6 383.8 333.5

Netherlands 25.8 23.4 21.4 277.2 278.5 209.0

Norway 80.8 60.0 53.6 556.8 400.6 258.6

Portugal 52.9 56.3 26.8 276.1 185.0 79.4

Spain 59.2 35.1 24.1 284.2 193.6 116.9

Sweden 45.5 42.4 37.9 801.1 584.3 446.7

Switzerland 34.4 27.7 22.0 368.3 347.0 270.8

UnitedKingdom

25.0 18.2 15.2 193.2 128.2 117.3

Total 36.6 27.8 22.0 336.2 229.6 160.5

Source: Derived from Tables 2 and 3.

This leads to the question of how the growing traffic can be managed on anetwork with far less growth. The questions of interest are, how far does thetraffic today suffer from overload and what is the impact of the problems?Overload means more demand for road space in certain parts of the roadnetwork than can be provided. One of the consequences is congestion, which isa widely discussed phenomena in traffic today. It is necessary to develop anefficient strategy for the future to avoid an adverse impact of congestion.

This paper discusses congestion in general and analyses in particular somedata from Germany. Transport activity in Germany is extremely high due tohigh individual motorisation and a great deal of transit transport. It can beexpected that the findings are of relevance to other European countries, too.

13

It should be mentioned that although the problem of congestion is of greatimportance in the political debate, the empirical basis is rather poor. Theposition in this paper must be set with respect to this fact and must be seen astentative.

2. TRAFFIC CONGESTION AND ITS CAUSES

2.1. Definition

Congestion can be defined as a situation in which transport participantscannot move in a usual or desirable manner. Vehicles of all kinds andpedestrians can experience congestion. It is a general phenomenon when thecapacity of an infrastructure is exceeded. This capacity is defined by thenumber of traffic participants passing per time unit.

A chain of transport participants moving in the same direction can benamed a traffic flow. Comparable to the water flow in a river, congestionmeans that the usual or desired amount of water does not flow in a certaindirection. The decrease, compared to the standard situation, can be caused byseveral reasons: the diameter of a pipe can be reduced, or it can even beblocked completely. In an analogy to the traffic flow, this may be one closedlane or an accident with car wrecks blocking the whole road. The analogybetween traffic flow and water flow ends when we look at another reason forcongestion, which is a high demand from vehicles wishing to use the road.A large number of vehicles competing for road space leads to denser andslower traffic and can block the road totally in the end.

An exact definition of congestion is rather difficult to find – as is thecommon problem with the term "sustainability". For both, the public interest isextremely high but it can be seen that the topics are debated with ratherdifferent intentions. From an economist's point of view, congestion causeseconomic losses and should be avoided, especially by increasing the capacityof the infrastructure. On the other hand, transport planners point out that thispolicy has been followed in the last decade in nearly all countries and it hasn'tsolved the problem. Supplying more transport capacity would induce newdemand for transport services, instead of organising social and economicactivity in a region differently. Economists tend to argue that all time savingsreached by additional supply have to be counted as profit for the national

14

economy. According to this argument the cost-benefit analysis of transportprojects monitors the time savings of the road users between their origins andtheir destinations. Most models assume that the additional road space does notinduce additional traffic and the time savings reached by increased averagespeed are real benefits.

Congestion in general is discussed only with respect to motor vehicletraffic on roads. From the driver’s perspective, congestion begins when he isforced to slow down because of slower vehicles ahead of him. The speedreduction may only go below the normal or desired driving speed, or it mayeven come to a standstill. Constant change between standstill and driving atslow speed is called stop-and-go traffic. From the bird’s eye perspective, thelength of congested traffic can be observed, i.e. the length of the road where thetraffic moves slower than usual or where it even stops.

Apart from traffic congestion, there are other kinds of congestion whichare rarely discussed: for example, trams waiting because of a parking carblocking the lane, or aeroplanes flying in circles over an airfield that is not free.Even pedestrians may experience congestion when the size of crowdedpedestrian lanes reduces immediately. In all these cases, a slow-down of thetransport speed can be observed. Even on the so-called data highway,congestion occurs when too many internet users want to communicate. Thegeneral definition of congestion could read as follows: congestion means areduction of service quality in infrastructure due to excessive demand or toother reasons. The users suffer from speed reduction, i.e. time penalties.

2.2. Kinds and causes of congestion

In the public debate, congestion is mainly seen as one phenomenon,neglecting the different kinds and different causes of this loss in transportservice quality. But to find adequate and efficient solutions to the problemsassociated with congestion, it is necessary to differentiate between the varioustypes and the various reasons. This could be:

− A reduction in road capacity caused by an unplanned event, forexample, an accident with wrecks blocking a lane;

− A planned reduction in capacity due to construction or maintenanceof the lane;

− A traffic demand higher than the maximum flow capacity.

15

The last point seems to be the most interesting one because of its complexformation process. Also, this case is the most often-cited reason forcongestion, linked with political demands for more roads. Overload congestionmeans that the amount of vehicles moving in a certain direction is higher than acertain part of the road can carry. When, at a certain point of the road, fewervehicles can pass through than want to pass, a queue is formed, growing rapidlyin the direction of the origin of the vehicle flow. A typical case is when thenumber of lanes is reduced at a certain point or when several roads or rampslead to a part of the network with less capacity than the added vehicle flowmoving towards it.

The dynamics of congestion have been widely evaluated and alsosimulated in computer programs. When the traffic flow exceeds seventy oreighty per cent of the theoretical capacity, a pre-critical condition is reached.The traffic will flow with a somewhat reduced speed, but it is still of goodservice quality. In this situation, even small disturbances can lead to seriousproblems in traffic flow and vehicles travelling in the opposite direction to thetraffic flow can suffer from the worsening turbulence. The fascinatingdynamics of congestion show that small disturbances can lead to a total blockfar away from the place where the small event happened and also far awayfrom the vehicle which caused the problem. When the pre-critical conditionhas led to a critical condition and this has led to stop-and-go, it takes much timeuntil the phenomenon vanishes.

Several different phases have been mentioned above: in a critical situationthere is a significant drop in speed plus stand-still and slow moves (stop-and-go).In quantitative terms, these congestion phases can be described by the actualvehicle speeds. For traffic on highways, on other non-urban roads and forurban traffic, different criteria may be necessary. A general assumption coulddescribe stand-still with stop-and-go with speeds between zero and 10 km/h. Acritical phase could be with speeds between 10 and 25 km/h on highways. Apre-critical phase could be defined as speeds above 25 km/h but below thatspeed would be an optimum with respect to the maximum traffic flow of aroad.

2.3. The role of vehicle speeds

In theory, the capacity of a road depends on design parameters, namely,the number and width of lanes. In real traffic, the capacity (maximum flow)depends on the type of vehicle, the speed and the speed differences between the

16

vehicles. If we assume first that all vehicles drive at the same constant speed ina certain road section, the number of vehicles passing the section depends onthe distance between two vehicles following each other. To maintain a certainlevel of safety, the distances between vehicles following each other must beincreased disproportionately with the increase in speed. This means that thereis an optimum speed which allows a maximum number of vehicles per unit oftime to pass the section. When the vehicles exceed this speed, the safe distancenecessary to maintain a comparable standard would cause a larger stretching ofthe vehicle chain than could be compensated with respect to the flow by theincrease in speed.

The other way round, the physical laws lead to the consequence that withvehicles flowing at higher speeds, the optimum capacity cannot be maintained.This can be reached by forcing the vehicles to reduce their speed towards alevel that ensures optimum use of the capacity. Vice versa, for vehicles drivingtoo slow for this optimum capacity, these should be speeded up. For highways,normally, the maximum number of vehicles passing a section per unit can bereached at speeds between 60 and 80 km/h.

Speed differences between vehicles driving in one direction will not touch theflow when the flow is on a low demand level. When the traffic gets denser, speeddifferences cause reactions by drivers – decelerating or changing lanes – whichinfluence others. In the end, these reactions lead to a decrease of the averagedriving speed and to a decrease in capacity. Both vehicles driving significantlyfaster and significantly slower than the average flow cause problems and can lead tocongestion.

The theoretical basis of the relation between speeds and highway capacitywas researched during the thirties in the United States and has led to therecommendations of the Highway Capacity Manual, which remains a referencefor transport planners all over the world. The optimum traffic condition is arelatively low (compared to some European states) vehicle speed but with theleast possible speed differences between the vehicles. One consequence of thisfinding has been the implementation of a maximum speed limit of 55 mph forall types of vehicle.

In Europe, permitted speeds are higher on average, and in the US the55 mph limit has no longer been valid for several years. Due to lobbying ofmanufacturers and automobile associations, the States are now mandated to setspeed limits on their own responsibility.

17

In any case, to avoid congestion, an increasing number of trafficmanagement systems are installed on German highways which introduce speedlimits and lower the permitted level with an increase in traffic demand. Indoing so, the optimum traffic flow for maximum use of the capacity can bereached.

Optimising the traffic speed can only help to avoid congestion in certaincases. When traffic demand far exceeds design capacity, this inevitably leadsto congestion. This is often the case in large conurbations and also due tospecial events, e.g. at the beginning of the holiday season. Congestion and timelosses in these cases cannot be avoided even by advanced traffic managementsystems when the road space is simply insufficient. It would not be costeffective to provide enough road capacity for all cases. In most denselypopulated areas, this is also not desirable for environmental reasons.

2.4. Road construction, accidents and other causes

This section attempts to illustrate the cases of reduced road capacity due tospecial problems based on the infrastructure itself, or linked to the vehicles.Regarding the infrastructure reasons, we have to mention the construction andmaintenance activities necessary to keep the surface quality of the roads. Theactivities often demand stoppage of one or several lanes. Also roadsideactivities which do not cause a physical barrier demand speed reductions forthe safety of the workers, which may cause congestion in dense traffic.Construction activities aimed at widening capacity by adding lanes may belooked at differently because here the irony is that planned futureimprovements may cause severe problems today. There is no informationavailable on how much congestion is caused by regular maintenance and repairactivities on the one hand and by construction activities for higher capacity onthe other. In any case, the adverse impact on traffic flow is more serious intimes of high traffic demand than in times of a low demand level.

On the vehicles’ side, problems leading to congestion are, for instance,accidents, breakdowns and special vehicles with unusually large size and/orextremely low speeds. There is no information available on the amount ofcongestion caused by vehicle size.

Finally, some other reasons for congestion can be mentioned such astraffic control, border control, etc. The latter is a serious problem, especiallyfor trucks heading for eastern Europe. Severe weather can also cause trafficcongestion.

18

3. VOLUME OF TRAFFIC CONGESTION

3.1. Introduction

We will discuss the amount of congestion exemplified for the Germanroad network. A comprehensive survey of congestion, differentiating betweenthe amount, the type and the causes, is still missing. In the public debate, thereare very controversial positions concerning the amount of congestion and thecongestion-related environmental and economic consequences. The GermanAutomobile Manufacturers’ Association (VDA) has published, in its currentyearbook, a figure of 14 billion litres of fuel burnt in traffic due to congestion.It calculates the possible savings of up to about 23 per cent of the fuelconsumption if congestion were avoided. The Association argues that it wouldbe possible to realise these potential savings by additional investment in roadconstruction and transport telematics.

The Manufacturers’ Association does not mention its own studies relatedto this topic and instead cites a paper by the German car manufacturer, BMW.This study calculates a congestion-related fuel consumption increase of12 billion litres annually which could be avoided by certain measures and anunnecessary time penalty of 5 billion hours per year. But BMW does not haveany comprehensive surveys of its own with detailed analyses of the trafficsituation and a discussion of possible measures and chances for reductionstrategies. The paper only contains some simple assumptions and calculations.For instance, it assumes that in urban traffic in Germany the average speedwould be 20 km/h which could be increased up to 30 km/h by avoidingcongestion, resulting in time savings of 2.33 billion vehicle hours. This wouldbe already half of the total calculated time penalty.

The paper does not seriously discuss how this increase in average speedcould be realised. In a similar, rather simplified manner, the paper assumesincreases in the average travel speeds on non-urban roads and on highways, thecalculation resulting in the total saving of 5 billion hours mentioned above.

The effect of congestion on total fuel consumption in German road trafficis also evaluated rather roughly. The source is cited as follows: "experimentshave shown that the fuel consumption due to avoidable disturbances in traffic isabout 20 per cent higher than in free-flowing traffic (13)". Source No. 13 in theBMW study reads, "Dr. Ing. Metz, Technische Universität München,Vorlesungsreihe Auto und Umwelt" (lecture on the automobile and the

19

environment, Technical University, Munich). Dr. Metz is a BMW staffmember and seems to have mentioned the figure in his lecture. There is nobasis given for this figure.

The method of calculating congestion effects must be seen as rathersurprising, taking into consideration that very important manufacturers’associations and other associations in the transport business base their positionconcerning climate policy and transport policy – including a public statementto the Kyoto Climate Conference – on this weak basis.

As we have no reliable studies dealing with the amount and consequencesof congestion, we have tried to develop our own model calculation. This isintended to quantify at least the magnitude of congestion and the energy as wellas emission effects. This calculation is especially based on a transport modelnamed Tremod and on the so-called Handbook of Emission Factors. Thesemodels have been developed by federal authorities, in particular by the FederalEnvironmental Agency of Germany and can be taken as the most reliable basisavailable.

3.2. Traffic flow on various road types

First, we focus on the outer-urban traffic because conditions there andpossible disturbances are related to other conditions and influences than inurban traffic. About 30 per cent of the German motor vehicle traffic is relatedto the highway network, which is only about 2 per cent of the total roadnetwork length. The average daily traffic (DTV) on highways is, of course, farhigher than on the minor roads. We assume in the following that these hightraffic volumes on highways are the relevant traffic conditions for congestion.Besides the highway network itself, we have to include those federal roadsshowing similar design features as highways. About 10 per cent of the Germanfederal roads show a road width of more than 12 metres which allows morethan one lane per direction.

High traffic flow levels in the highway network are concentrated only on apart of the network which, in fact, can be assumed to be the relevant sectionsfor high congestion.

The Tremod traffic model and the emission factor system deliver verydifferentiated road types, traffic situations and vehicle types. We haveaggregated the situations to a system shown in Table 5. According to these

Table 5. Speed, consumption, emission and journey time by type of vehicle

Vehicle- Traffic Traffic performance Speed Consump. NOx-Em. HC-Em. Time spent Consump. NOx-Em. HC-Em.Category Situation (3) km kph g/km g/km g/km Mio h t T t

Two-Wheelers HW free w/oSL 1,091,735,494 123.3 42.982 0.502 2.072 8.854 46,925 548.087 2,261.902HW free, SL 528,447,357 105.1 38.022 0.361 2.513 5.027 20,093 190.685 1,328.019

HW bound 186,883,311 80.2 33.102 0.252 2.237 2.330 6,186 47.089 418.094

HW Stop+Go 28,940,174 19.0 56.708 0.110 4.786 1.523 1,641 3.183 138.508

RR 7,250,515,748 61.1 28.879 0.181 2.595 118.729 209,387 1,310.773 18,813.734

LS AS free 52,207,693 39.9 22.805 0.082 3.302 1.310 1,191 4.262 172.369

LS AS disturbed 2,328,010,644 31.5 24.481 0.069 3.334 73.864 56,992 159.649 7,762.267

LS SS 550,192,537 21.2 30.681 0.061 3.695 25.980 16,881 33.641 2,033.103

LS Stop+Go 46,132,869 19.5 33.737 0.056 4.199 2.368 1,556 2.562 193.698

Total 12,063,065,827 50.3 29.914 0.191 2.746 239.985 360,852 2,299.931 33,121.694

Cars HW free, w/oSL 86,242,580,920 130.0 64.045 1.344 0.233 663.404 5,523,383 115,896.669 20,074.905

HW free, SL 39,147,959,499 110.7 55.060 0.997 0.207 353.636 2,155,477 39,017.649 8,109.549

HW bound 14,571,803,407 84.9 44.218 0.625 0.174 171.577 644,336 9,108.178 2,536.394

HW Stop+Go 2,360,461,253 9.5 87.875 0.395 1.119 248.470 207,426 932.528 2,642.176

RR 217,009,435,120 75.4 45.511 0.819 0.354 2,879.472 9,876,274 177,739.308 76,760.008

LS AS free 2,933,863,717 58.4 43.603 0.645 0.430 50.237 127,926 1,891.588 1,261.471

LS AS disturbed 138,407,959,561 35.2 56.744 0.697 0.582 3,936.453 7,853,818 96,481.729 80,544.143

LS SS 33,488,407,248 18.6 84.302 0.753 0.941 1,800.452 2,823,128 25,202.424 31,501.801

LS Stop+Go 2,733,572,263 5.3 151.079 0.646 2.480 515.768 412,984 1,765.892 6,778.117

Total 536,896,042,987 50.6 55.178 0.872 0.429 10,619.471 29,624,751 468,035.966 230,208.566

Table 5 (continued)

Light Trucks (1) HW free, w/oSL 7,335,804,999 115.0 106.531 1.686 0.269 63.790 781,490 12,368.308 1,974.117

HW free, SL 3,348,115,992 108.8 99.881 1.609 0.259 30.773 334,413 5,385.674 868.158

HW bound 1,242,981,403 84.9 72.341 1.292 0.235 14.635 89,918 1,606.151 292.452

HW Stop+Go 199,590,579 9.5 100.738 1.015 1.223 21.010 20,106 202.553 244.056

RR 16,785,671,005 75.3 75.262 1.358 0.342 222.877 1,263,316 22,793.901 5,748.364

LS AS free 224,789,449 58.4 67.882 1.117 0.448 3.849 15,259 251.141 100.676

LS AS disturbed 10,493,498,169 35.2 83.510 1.199 0.667 298.448 876,315 12,580.830 7,003.503

LS SS 2,156,956,838 18.6 122.437 1.441 1.160 115.965 264,091 3,108.704 2,501.629

LS Stop+Go 205,617,261 5.3 168.952 1.535 2.487 38.796 34,740 315.615 511.396

Total 41,993,025,694 51.8 87.625 1.396 0.458 810.142 3,679,649 58,612.877 19,244.351

Trucks (2) HW free, w/oSL 14,260,517,222 85.8 258.306 8.394 0.821 166.139 3,683,581 119,704.667 11,701.185

HW free, SL 6,778,490,980 85.0 257.007 8.374 0.830 79.730 1,742,118 56,763.064 5,625.178

HW bound 668,352,590 77.7 250.280 8.161 0.899 8.602 167,275 5,454.697 600.711

HW Stop+Go 353,146,674 5.8 745.038 35.896 7.279 60.887 263,108 12,676.421 2,570.568

RR 14,495,790,545 70.0 205.807 6.775 1.002 207.133 2,983,336 98,204.951 14,521.545

LS AS free 195,013,665 52.5 201.274 6.681 1.299 3.716 39,251 1,302.805 253.326

LS AS disturbed 8,390,459,827 28.9 266.905 9.617 2.260 290.103 2,239,453 80,690.571 18,961.249

LS SS 903,271,957 15.1 341.956 12.411 4.249 59.819 308,879 11,210.585 3,837.973

LS Stop+Go 152,255,213 5.8 661.673 26.464 10.160 26.251 100,743 4,029.278 1,546.964

Total 46,197,298,673 51.2 249.533 8.443 1.291 902.379 11,527,744 390,037.040 59,618.699

Table 5 (continued)

Buses HW free, w/oSL 627,031,789 85.6 221.066 8.298 0.844 7.327 138,615 5,203.324 528.941

HW free, SL 312,465,293 83.6 219.585 8.248 0.870 3.738 68,613 2,577.280 271.712

HW bound 29,872,462 70.5 209.034 7.868 1.044 0.424 6,244 235.048 31.193

HW Stop+Go 15,690,215 5.8 829.677 33.730 11.190 2.705 13,018 529.227 175.578

RR 1,429,192,441 56.8 213.837 8.468 1.116 25.145 305,614 12,101.934 1,594.435

LS AS free 26,835,430 42.4 239.358 10.431 1.204 0.632 6,423 279.927 32.297

LS AS disturbed 1,276,473,870 22.7 324.829 14.563 2.236 56.264 414,636 18,588.994 2,854.166

LS SS 137,538,086 13.2 447.065 19.880 3.831 10.443 61,488 2,734.291 526.899

LS Stop+Go 25,198,500 5.8 767.511 33.279 9.117 4.345 19,340 838.575 229.747

Total 3,880,298,086 35.0 266.472 11.104 1.609 111.023 1,033,992 43,088.600 6,244.967

TOTAL HW free, w/oSL 109,557,670,423 120.5 92.864 2.316 0.334 909.513 10,173,994 253721.055 36541.050

HW free, SL 50,115,479,121 106.0 86.215 2.074 0.323 472.905 4,320,714 103934.352 16202.616

HW bound 16,699,893,173 84.5 54.729 0.985 0.232 197.568 913,960 16451.163 3878.844

HW Stop+Go 2,957,828,896 8.8 170.834 4.849 1.951 334.595 505,299 14343.913 5770.886

RR 256,970,604,858 74.4 56.963 1.215 0.457 3453.356 14,637,927 312150.867 117438.086

LS AS free 3,432,709,954 57.5 55.364 1.087 0.530 59.745 190,050 3729.724 1820.139

LS AS disturbed 160,896,402,070 34.6 71.109 1.296 0.728 4655.132 11,441,214 208501.773 117125.327

LS SS 37,236,366,667 18.5 93.308 1.136 1.085 2012.660 3,474,467 42289.645 40401.406

LS Stop+Go 3,162,776,104 5.4 180.020 2.198 2.928 587.527 569,364 6951.922 9259.922

Total 641,029,731,267 50.5 72.114 1.501 0.544 12683.001 46,226,988 962074.414 348438.277

(1) LT = Light Trucks; (2) T = Trucks; (3) HW = Highway, RR = Regional Road,LS = Local Street, w/oSL = without Speed-Limit, SL = Speed-Limit, AS = Arterial Street, SS = Side Street.(a) Share in respect to the vehicle category; (b) Share in respect to all vehicles.

23

statistics, of a total of 641 billion vehicle-kilometres in the German roadnetwork, there are 257 billion v-km on extra-urban roads, 205 billion v-km inurban traffic and about 180 billion v-km on highways.

Of the latter 180 billion v-km (reference here, 1995) only 3 billion km arededicated to stop-and-go congestion, showing an average speed of 8.8 km/h. Inthis congested situation on highways, a total of 505 000 tonnes of fuel areconsumed. This is to be compared to a total of 46.2 million tonnes of fuelconsumption in all the German road traffic. (It has been mentioned above thatthe figures for vehicle-kilometres and energy consumption in these statisticsdiffer somewhat compared to other statistics. It will be necessary to clarifythese differences in the future work.)

The total fuel consumption in urban traffic is given as about16 million tonnes, of which about 570 000 tonnes relate to stop-and-go traffic.For urban traffic, stop-and-go is dedicated to an average speed of 5.4 km/h.The share of stop-and-go in outer-urban traffic is even lower than on the otherroad types.

Together we can calculate that only 1.1 million tonnes of fuelconsumption in German road traffic can be related to congestion on highwaysand in urban traffic. Related to a total of 46.2 million tonnes of total fuelconsumption in road traffic, this is a share of about 2 per cent.

With respect to the time losses, we have based our calculation on theaverage speeds given for congestion. Summing up the travel time spent incongested situations, as has been discussed above, we calculate a total of334 million vehicle-hours on highways and 587 million vehicle-hours in urbantraffic in stop-and-go conditions. The total travel time in road traffic is givenas 12.7 billion hours in the statistics. Both congested situations on highwaysand in urban traffic make a total of 920 million hours, which is about 11 hoursper German citizen annually and 0.03 hours equals 1.8 minutes per day. Itseems very likely that even pedestrians exceed this congestion time per daywaiting on a traffic light or for a chance to cross a street.

The calculated amounts of fuel consumption and travel times related tostop-and-go traffic indicate the maximum possible potential which can bediscussed if congestion was avoided. This figure would only be reached ifstop-and-go traffic could be avoided totally. It can be seen that the potentialsavings are rather small compared to the total fuel consumption and traveltimes in German overall traffic.

24

3.3. Further differentiation of traffic conditions

As has been mentioned above, only a small share of the vehicle-kilometresand fuel consumption is related to congested traffic. For travel time, thestop-and-go share is somewhat higher, about 900 out of 12 700 millionvehicle-hours. A further analysis of traffic conditions leads to the followingfindings.

The largest share of the vehicle-kilometres driven on highways(110 billion of 180 billion) goes to the "free-flow without speed limit" trafficsituation. This condition is given an average traffic speed of 120.5 km/h for allvehicle types. For passenger cars, the average speed in this traffic condition isgiven as 130 km/h. The second most common traffic condition on highways isthe "free, with speed limit" condition, showing 50 billion vehicle kilometres.The average traffic speed is given as 106 km/h (for passenger cars:110.7 km/h). The intermediate condition between "free-flow" traffic on the onehand and "stop-and-go" traffic on the other is named "bound" traffic, whichaccounts for 16.7 billion vehicle kilometres; this is less than 10 per cent of thetotal highway vehicle-kilometres. The average speed given to this situation is84.5 km/h; for passenger cars this differs only a little, with 84.9 km/h. Theseslower speeds are a consequence of higher traffic densities. In this case, furtherincreases in demand may lead to critical conditions of congestion and finally tostop-and-go. The mentioned share of "bound" traffic cannot be identified as anunfavourable condition with respect to energy consumption. Given perkilometre for passenger cars, "bound" traffic with about 85 km/h only demandsa fuel consumption of 44.2 grams per kilometre, which is significantly betterthan for "free-flow with speed limit" (55 grams per kilometre) and especiallycompared to "free-flow without speed limit" (64 grams per kilometre). Withrespect to fuel consumption, "bound" traffic flow is the most fuel-efficientcondition on highways.

In urban traffic the statistics differentiate between "main road free" and"main road disturbed flow", also between traffic on minor feeder roads and thealready mentioned stop-and-go traffic. More than three-quarters of the urbanvehicle kilometres are within the category "main road disturbed flow", showingan average traffic speed of 34.6 km/h (passenger cars: 35.2 km/h). These arethe main urban traffic conditions. Undisturbed flow on main roads may onlyhappen at night when there are no stops at traffic lights and no cross traffic.

25

For free-flow conditions on main roads, the statistics give an averagespeed of 57.5 km/h (passenger cars: 58.4 km/h); this category only consists ofabout 3.4 of the total 200 billion urban vehicle-kilometres. In addition to bothcategories on main roads, we can mention the minor urban roads, with about20 per cent of urban vehicle traffic. We can assume that on minor roads thereis such a low traffic density that we do not have any significant share ofcongestion there.

This comprehensive analysis of the available data concerning thecongested vehicle kilometres, fuel consumption and time spent in congestion,in no way reveals a congestion-related share of energy consumption in themagnitude of 20 per cent, as cited by the automobile industry in section 3.1.According to our findings, we can calculate that about 2 per cent of total fuelconsumption in road traffic is related to stop-and-go conditions. The othertraffic flow conditions, both on highways and in urban traffic, cannot beidentified as congested. At least, there is no fuel penalty for high trafficvolume situations on highways which we have mentioned as "bound" trafficabove. Summing up our findings again: we do not see any reason for fuelconsumption penalties and time losses related to congestion in a relevantmagnitude. Of course, there is a rather high share of disturbed traffic on mainurban roads but with a rather high average speed. The disturbances are aconsequence of the fact that urban roads have traffic lights and cross traffic. Itis rather unlikely that there are any realistic measures which can avoid stoppingat traffic lights and which can eliminate cross traffic. The rather high averagetraffic speed in urban traffic in this category of 34.6 km/h indicates that thiscondition cannot be identified as congested.

Also, the time losses mentioned in the political debate seem to be at anunrealistically high level. Of course it would be possible theoretically to increasethe average speed on highways, e.g. from the "bound traffic flow" condition(84.5 km/h, all vehicles) by about 20 km/h to the "free-flowing traffic with speedlimits" category. This would theoretically save about 40 million hours in traveltime annually. But without a very comprehensive cost-benefit analysis, takinginto consideration the necessary investments and other effects, any of thosecalculations would be irresponsible. With that kind of calculation, no seriousconclusions can be drawn. It would be similarly naive to follow the idea thatthe average speed of free-flow with a speed limit of 106 km/h could beincreased to 120 km/h as is given for the condition "free-flow without speedlimit". In Germany, speed limits on highways are only implemented whenthere are rather good reasons with respect to traffic safety and other arguments.Time savings and cost reduction for the national economy cannot be argued by

26

simple ideas of increasing the average speed. As we have shown in ouranalysis, the main traffic conditions are not related to congestion and the shareof really congested stop-and-go traffic is far lower than in the studies supportedby the automobile industry.

3.4. Differentiation according to vehicle category

Some differences between the average figures for all vehicles and forpassenger cars have already been discussed above. The main differences intraffic speeds occur for the traffic condition "highway free, without speedlimits" between heavy trucks/buses on the one hand, with average speeds of86 km/h and passenger cars on the other, with 130 km/h. Light duty trucksreach an average speed of 115 km/h. The other traffic situations show smallerdifferences the more the average traffic speed is reduced, due to increasedtraffic demand. In urban situations, there is, of course, a very similar trafficspeed pattern among the different vehicles. One difference in speeds betweenbuses and other heavy-duty vehicles in outer-urban traffic is caused by busstops. Except for this case, the average outer-urban traffic speed is about75 km/h, which seems a reasonable level.

The emission factors given in the statistics specify the very generalfindings in section 4.2.: NOx emissions increase generally with increasingtraffic speed, emissions per kilometre are the lowest in stop-and-go traffic.This is not valid for heavy-duty vehicles, where the high energy demand foraccelerating the large vehicle mass leads to very high NOx emissions both inthe stop-and-go category on highways and stop-and-go in urban traffic.Summing up all NOx emissions in these cases, we only find 14 000 tonnes of NOxout of a total of 1 million tonnes which can be related to highway stop-and-go andonly 7 000 tonnes of NOx which can be related to urban stop-and-go. The majortraffic situations for high NOx emissions are both free-flow conditions on highwaysand, secondly, the extra-urban traffic contributions to the NOx balance. There areno significant influences of congestion with respect to the NOx emissions from roadtraffic.

Concerning HC emissions, we see the consequences of congestion in asimilar low magnitude: highway stop-and-go is only responsible for about6 000 tonnes of HC out of a total of 350 000 tonnes in road traffic in Germany;for urban stop-and-go we calculate about 9 300 tonnes. Even if it must be

27

admitted that specific emission factors are rather high in congestion situations,compared to free-flowing traffic, we do not see a general environmentalproblem for the total German traffic condition.

This also applies to the consumption of energy and the proportional CO2

climate emissions as derived above.

A presentation in percentages of data on transport performance,consumption and emissions, as well as the transport times in the individualtraffic situations, can be seen in Table 6.

3.5. Conclusion

The statistics show that congestion is only slight when compared to thewhole road network and the total transport volume in Germany. On the otherhand, the local and regional disturbances in road traffic which undoubtedlyoccur, especially in conurbations, can definitely present a serious problem on aregional scale. A first estimate of the ecological effects – in the form ofemissions of harmful substances and energy consumption – as well aseconomical/social consequences, estimated on the basis of travel time duringcongestion, suggest a rather cautious position with respect to the differentnotions that congestion must be eliminated immediately in order to reduce theconsiderable burden of costs in the whole economy. The analyses show thattravel times occurring during congestion are, however, considerably lower thanthose which are being put forward at this time in the political debate.

Table 6. Shares of traffic performance, consumption,emission and travel time

Vehicle TrafficTraffic

perform.Consump

tionNOx-Em. HC-Em. Travel

TimeCategory situation (3) (a) % (b) % (a) % (b) % (a) % (b) % (a) % (b) % (a) % (b) %

Two-Wheelers HW free w/oSL 9.1 0.2 13.0 0.1 23.8 0.1 6.8 0.6 3.7 0.1HW free, SL 4.4 0.1 5.6 0.0 8.3 0.0 4.0 0.4 2.1 0.0

HW bound 1.5 0.0 1.7 0.0 2.0 0.0 1.3 0.1 1.0 0.0

HW Stop+Go 0.2 0.0 0.5 0.0 0.1 0.0 0.4 0.0 0.6 0.0

RR 60.1 1.1 58.0 0.5 57.0 0.1 56.8 5.4 49.5 0.9

LS AS free 0.4 0.0 0.3 0.0 0.2 0.0 0.5 0.0 0.5 0.0

LS AS disturbed 19.3 0.4 15.8 0.1 6.9 0.0 23.4 2.2 30.8 0.6

LS SS 4.6 0.1 4.7 0.0 1.5 0.0 6.1 0.6 10.8 0.2

LS Stop+Go 0.4 0.0 0.4 0.0 0.1 0.0 0.6 0.1 1.0 0.0

Total 100.0 1.9 100.0 0.8 100.0 0.2 100.0 9.5 100.0 1.9

Cars HW free, w/oSL 16.1 13.5 18.6 11.9 24.8 12.0 8.7 5.8 6.2 5.2HW free, SL 7.3 6.1 7.3 4.7 8.3 4.1 3.5 2.3 3.3 2.8

HW bound 2.7 2.3 2.2 1.4 1.9 0.9 1.1 0.7 1.6 1.4

HW Stop+Go 0.4 0.4 0.7 0.4 0.2 0.1 1.1 0.8 2.3 2.0

RR 40.4 33.9 33.3 21.4 38.0 18.5 33.3 22.0 27.1 22.7

LS AS free 0.5 0.5 0.4 0.3 0.4 0.2 0.5 0.4 0.5 0.4

LS AS disturbed 25.8 21.6 26.5 17.0 20.6 10.0 35.0 23.1 37.1 31.0

LS SS 6.2 5.2 9.5 6.1 5.4 2.6 13.7 9.0 17.0 14.2

LS Stop+Go 0.5 0.4 1.4 0.9 0.4 0.2 2.9 1.9 4.9 4.1

Total 100.0 83.8 100.0 64.1 100.0 48.6 100.0 66.1 100.0 83.7

Table 6 (continued)

Light Trucks (1) HW free, w/oSL 17.5 1.1 21.2 1.7 21.1 1.3 10.3 0.6 7.9 0.5HW free, SL 8.0 0.5 9.1 0.7 9.2 0.6 4.5 0.2 3.8 0.2

HW bound 3.0 0.2 2.4 0.2 2.7 0.2 1.5 0.1 1.8 0.1

HW Stop+Go 0.5 0.0 0.5 0.0 0.3 0.0 1.3 0.1 2.6 0.2

RR 40.0 2.6 34.3 2.7 38.9 2.4 29.9 1.6 27.5 1.8

LS AS free 0.5 0.0 0.4 0.0 0.4 0.0 0.5 0.0 0.5 0.0

LS AS disturbed 25.0 1.6 23.8 1.9 21.5 1.3 36.4 2.0 36.8 2.4

LS SS 5.1 0.3 7.2 0.6 5.3 0.3 13.0 0.7 14.3 0.9

LS Stop+Go 0.5 0.0 0.9 0.1 0.5 0.0 2.7 0.1 4.8 0.3

Total 100.0 6.6 100.0 8.0 100.0 6.1 100.0 5.5 100.0 6.4

Trucks (2) HW free, w/oSL 30.9 2.2 32.0 8.0 30.7 12.4 19.6 3.4 18.4 1.3HW free, SL 14.7 1.1 15.1 3.8 14.6 5.9 9.4 1.6 8.8 0.6

HW bound 1.4 0.1 1.5 0.4 1.4 0.6 1.0 0.2 1.0 0.1

HW Stop+Go 0.8 0.1 2.3 0.6 3.3 1.3 4.3 0.7 6.7 0.5

RR 31.4 2.3 25.9 6.5 25.2 10.2 24.4 4.2 23.0 1.6

LS AS free 0.4 0.0 0.3 0.1 0.3 0.1 0.4 0.1 0.4 0.0

LS AS disturbed 18.2 1.3 19.4 4.8 20.7 8.4 31.8 5.4 32.1 2.3

LS SS 2.0 0.1 2.7 0.7 2.9 1.2 6.4 1.1 6.6 0.5

LS Stop+Go 0.3 0.0 0.9 0.2 1.0 0.4 2.6 0.4 2.9 0.2

Total 100.0 7.2 100.0 24.9 100.0 40.5 100.0 17.1 100.0 7.1

Table 6 (continued)

Buses HW free, w/oSL 16.2 0.1 13.4 0.3 12.1 0.5 8.5 0.2 6.6 0.1HW free, SL 8.1 0.0 6.6 0.1 6.0 0.3 4.4 0.1 3.4 0.0

HW bound 0.8 0.0 0.6 0.0 0.5 0.0 0.5 0.0 0.4 0.0

HW Stop+Go 0.4 0.0 1.3 0.0 1.2 0.1 2.8 0.1 2.4 0.0

RR 36.8 0.2 29.6 0.7 28.1 1.3 25.5 0.5 22.6 0.2

LS AS free 0.7 0.0 0.6 0.0 0.6 0.0 0.5 0.0 0.6 0.0

LS AS disturbed 32.9 0.2 40.1 0.9 43.1 1.9 45.7 0.8 50.7 0.4

LS SS 3.5 0.0 5.9 0.1 6.3 0.3 8.4 0.2 9.4 0.1

LS Stop+Go 0.6 0.0 1.9 0.0 1.9 0.1 3.7 0.1 3.9 0.0

Total 100.0 0.6 100.0 2.2 100.0 4.5 100.0 1.8 100.0 0.9

Total HW free, w/oSL 17.1 17.1 22.0 22.0 26.4 26.4 10.5 10.5 7.2 7.2HW free, SL 7.8 7.8 9.3 9.3 10.8 10.8 4.7 4.7 3.7 3.7

HW bound 2.6 2.6 2.0 2.0 1.7 1.7 1.1 1.1 1.6 1.6

HW Stop+Go 0.5 0.5 1.1 1.1 1.5 1.5 1.7 1.7 2.6 2.6

RR 40.1 40.1 31.7 31.7 32.4 32.4 33.7 33.7 27.2 27.2

LS AS free 0.5 0.5 0.4 0.4 0.4 0.4 0.5 0.5 0.5 0.5

LS AS disturbed 25.1 25.1 24.8 24.8 21.7 21.7 33.6 33.6 36.7 36.7

LS SS 5.8 5.8 7.5 7.5 4.4 4.4 11.6 11.6 15.9 15.9

LS Stop+Go 0.5 0.5 1.2 1.2 0.7 0.7 2.7 2.7 4.6 4.6

Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

(1) LT = Light Trucks; (2) T = Trucks; (3) HW = Highway, RR = Regional Road,LS = Local Street, w/oSL = without Speed-Limit, SL = Speed-Limit, AS = Arterial Street, SS = Side Street(a) Share in respect to the vehicle category; (b) Share in respect to all vehicles

31

4. ASSESSMENT OF CONGESTION EFFECTS

4.1. Overview

The effects of congestion can be broken down into three categories:

− Environmental consequences;− Economic consequences;− Social and other consequences.

All the effects are rather difficult to quantify. Just as the definition ofcongestion itself is difficult, the amount of congestion in the road network alsocannot be calculated exactly. This has to be left to future work based onprofound empirical data.

4.2. Environmental consequences

The environmental effects of traffic congestion can be differentiatedbetween the impact on the natural environment and the impact on humanhealth. The impact of exhaust emissions has to be looked at with respect to thedirect health effect from congested vehicles to persons and vegetation near theroad, e.g. the direct toxic effect of carbon monoxide on health and also theadverse effect of gases on a regional or international level. On a regional level,ozone formation out of nitrogen oxides and hydrocarbon is a severe problemwhich may worsen due to increased HC emission. Acid deposits also may bean effect of congested traffic and, finally, on a global scale, an increase in theconcentration of greenhouse gases in the atmosphere may be influenced bycongestion.

The environmental impact may also include traffic noise as well as the useof non-renewable resources. Some studies tend to exclude traffic noise fromthe environmental consequences and categorise it as a social effect, but thismay not be discussed here. Here, we also see this effect as an environmentallyadverse impact.

Comparing the environmental consequences of reduced traffic speeds anddisturbed traffic flows with normal traffic, as a first guess, the following assessmentcan be made:

32

− Disturbed traffic flow leads to an increase of carbon monoxide (CO)and hydrocarbon (HC) emissions per kilometre for passenger cars aswell as for trucks. This results from a higher share of the vehicleengines operating at low engine loads or even idling (standstill);these are the engine conditions favourable for the formation of bothproducts. The emissions can be characterised as not completelyburnt, which they would be at higher engine loads, which means athigh speeds. There, the higher engine temperature leads to loweremissions of these products.

− Nitrogen oxide (NOx) emissions per kilometre and per unit of timedecrease with decreasing traffic speeds. Also, non-stationaryconditions of stop-and-go traffic keep the NOx emissions on a lowerlevel than they would be at a higher reference speed. Only heavytrucks accelerating from standstill may favour the formation of higherNOx, but for passenger cars compared to the free traffic flow with theusual speeds, the specific emissions would be lower in congestedsituations.

− Energy consumption per kilometre increases in stop-and-goconditions compared to other speeds when the motor is idling and thecar is using second or third gear. These conditions mean operating atrather low engine loads with very poor efficiency. Basically, theenergy demand for moving a vehicle is as low as the speed, but theunfavourable engine efficiency at low loads overcompensates thereduced energy demand at low speeds.

− A comparison of gasoline and diesel engines shows an advantage ofthe latter in congestion because of its better efficiency at low engineload and lower idle consumption. A part of the disadvantage ofgasoline engines could be reduced by switching the engine off atstandstill, but this is only recommended for stops of more than30 seconds; otherwise the start procedure would cause higher fuelconsumption and CO and HC emissions than would be saved duringswitch-off time.

− For the inhabitants living by congested roads, the average noise levelwill be lower than at speeds of more than about 50 km/h. Above thisspeed level, tyre noise exceeds the engine noise normally. There maybe some special situations, e.g. a slope with lots of heavy trucks,when stop-and-go traffic noise may be higher than free traffic speeds.

33

These general assessments cannot be elaborated further in this study.With respect to the exhaust emissions, it would be desirable to differentiate notonly between gasoline and diesel engines, but also according to the emissioncontrol technology. Lower exhaust gas temperature in congested traffic maycause reduced catalytic efficiency: on the other hand, we find satisfyingcatalytic activity when warm engines are idling for a certain time. This may bedifferent for aged or damaged catalysts.

For a discussion of the emissions and the energy consumption ofcongestion, it is important to define the reference case. In congestion, theemission or energy consumption per kilometre may be higher than its constantor quasi-constant speeds of, for example, 100 kilometres for passenger cars, or80 kilometres for trucks, but if we compare this with free traffic with a lowvehicle density and significantly higher speeds, this may be seen differently. InGermany, on highways without any speed limit, the free traffic flow may causehigher per kilometre emissions and energy consumption than pre-critical oreven critical speeds. In general, we can assume that a flow around the speedsthat are optimal for maintaining the maximum capacity (60 to 80 km/h as hasbeen mentioned) also probably guarantees the best fuel efficiency perkilometre. If we relate the emission and the energy consumption to time units,lower speeds always show better values.

Basically, we can describe the relation between speed and exhaustemission as follows: CO and HC emissions are high at low speed, due to lowengine load, and at standstill and stop-and-go. With increasing speed we reacha minimum for CO and for HC. This may be somewhere between60 and 100 km/h. Above this speed, the CO and HC emissions increase. Forgasoline engines with rich mixtures at full load, the emissions may beexcessively high at high speeds. This is also the case with passenger carsequipped with 3-way catalysts and closed-loop control. Catalytic efficiencywill be reduced in the long run when vehicles go at very high speeds, whichmeans very high exhaust gas volume.

NOx emission increases with increasing speeds continuously, this is aconsequence of higher engine loads and higher engine temperatures at higherpower demand. Acceleration may cause higher NOx due to the higher powerdemand on the one hand, some types may use an enrichment of mixturesmitigating NOx but increasing CO and HC. The NOx increase with speed isbasically valid for gasoline and diesel engines f different design features

34

(with/without catalysts, direct or indirect injection diesels, turbo-charge engine,etc.). Apart from this general increase of NOx with speed, the values are quitedifferent.

Standstill on a road causes less emissions, especially NOx and less energyconsumption per time unit than any driving conditions. A passenger car engineconsumes about one litre of fuel per hour, this may be somewhat lower fordiesel engines. In normal traffic a passenger car would use 8 litres of fuel perhour (at 100 km/h this would equal 8 litres per 100 kilometres). If we assume aconstant travel time budget for certain activities (see section 4.4.), this wouldlead to reduced travel distances or reduced travel trips which, in the end, couldeven cause less emissions and less energy consumption due to congestion thanat undisturbed traffic conditions.

4.3. Economic consequences

Time losses due to congestion also cause direct economic losses for trafficparticipants. We will not discuss here if these time losses must be seen asexternal costs or not. In the literature, there are different positions. In anycase, in a scenario with constant origin-destination relations in passenger andgoods traffic, congestion leads to increased travel time compared with ascenario with undisturbed traffic flow. Following the calculation scheme thatis applied in cost-benefit analyses for German Federal transport infrastructureprojects, the following cost factors can be evaluated:

− Vehicle time costs in goods transport as well as for business trips inpassenger cars;

− Operation costs, broken down into labour costs and vehicle operation;for labour costs the calculation goes in direct relation to the time lostin congestion; for vehicle operation costs we have to take the fuelconsumption per kilometre into consideration, which normally ishigher in congestion than in the reference case.

For cost-benefit analyses of new road construction, the aspects of trafficsafety, of exhaust emissions and of traffic noise are evaluated and monetarised.

The economic evaluation of time losses for non-business trips is aproblem. Basically, time losses for commuters, for shopping trips and even forleisure trips can be monetarised. The methodological basis for an evaluation ofthese time losses is questionable. An equal evaluation of commuters’ travel

35

time with working times does not seem to be justified because, on the onehand, the employee will value the time differently and, on the other hand, manycommuters choose their places of residence and work in spite of congestedtraffic, that is, time losses. This means that they accept the low average travelspeed as it is. It would be very difficult to differentiate between additionalcongestion related losses. For leisure trips, we have to consider that, due to theresults of mobility research, there is a constant average daily travel time which,in the long run, would lead to higher travel distances without congestion thanwith congestion. This would make any evaluation of saved time or time lossesuseless.

Congestion-related additional costs for increased fuel consumption, highervehicle time costs and higher personnel costs must be allocated to theenterprises. In congested traffic, more vehicles and more drivers have to beprovided for the same transport activity. Looking at it the other way round: ifthere was no congestion, a proportion of the vehicles and drivers would not beused. This would lead to savings for the enterprises. Evaluation in monetaryterms for the national economy would be justified if, for the whole economy,productivity losses would occur because vehicles and drivers were notavailable for productive activities. This could only be argued if there was alack of employees which would only then mean a directly related loss increative value for the national economy.

It is highly questionable if this condition exists in all European regions.

Congestion makes the product more expensive for the customers due tothe higher cost for the enterprises. The amount of the consumer price increaseis extremely low. If we take the average transport cost share of 3-5 per cent ofthe market price of products, congestion could only cause an increase of about1 per cent of the transport costs, which would mean 1 per cent of 3-5 per cent.This could not be seen as an economic loss for the national economy when weassume a constant volume of the expenses of the consumers with and withoutcongestion. This cannot be seen. But indeed there is a problem for thecompetitiveness of regional enterprises when they have cost increases due tocongestion in the respective regions. We do not think that this influence issignificant, but this can only be assessed seriously with extended input-outputanalyses.

36

4.4. Social and other consequences

In addition to the environmental and economic consequences, we have tolook at various other effects resulting from a congestion-related reduction intransport speeds. It means a reduction in access (measured in number ofpersons that could be reached within a certain travel time). A reduction intravel speed would mean a reduction of the social contacts over certaindistances, for example, to relatives and friends living far away. This could bethe case when the individually tolerated travel times would be exceededregularly due to congestion. Also, in the tourist sector we could assumechanges due to congestion, when some destinations at greater distances wouldbe visited less often. Time losses due to congestion may basically influence thespatial orientation of all human activities. Settlement in the rural areas aroundlarge cities and conurbations has grown during the last decades because withthe individual automobile it has become possible to link working places,shopping areas and cultural events in the city with the rural settlements. Thesuburbanisation process was made possible because the automobile becameavailable to the masses. In the same process, the increase in motor vehicletraffic has caused infrastructure overload and congestion and thus worsenedaccess to various activity centres. Alternative transport modes, e.g. rail traffic,often cannot provide the same level of transport service. It cannot be expectedthat road capacity will be extended in the future, as it has been extended in thelast decades, due to financial and environmental constraints. If we assume aworsened access due to continued traffic demand increase in the next decades,a mitigation of the suburbanisation process or even a change of the trend couldbe possible. Such development would affect society significantly, it would hitthe poorer part of the population more than the richer part – these couldprobably afford houses in nice places near the city centres.

The theoretical consideration with respect to the social consequences isbased on the assumption that increasing congestion, which means decreasingtravel speed, would be compensated in the long run by decreasing tripdistances. This may sound logical and indeed the observed phenomenon ofconstant travel time is valid over decades and also in cross-comparison betweendifferent countries, but the discussed possible changes in spatial orientationhave not been observed in reality. Moreover, the average travel speeds onhighways in Germany have been monitored to increase from year to year.Obviously, the increased number of automobiles in the increased transportactivity in general has not caused reduced travel speeds in the highwaynetwork. A reason for this somewhat surprising observation is, of course, thelarge increase in capacity in the highway network during the last decade; we

37

have to look not only at the road length but also at the number of lanes.Another reason for the development towards higher travel speed in spite of theincreased number of vehicles may be based on changed time structure forworking and shopping. Part of the travel demand has been shifted to times withpreviously lower travel demand. In contrast to the typical daily demand curvesof some decades ago, where we had significant morning and evening peaks,today, we experience a broad daily curve with high demand levels over the day.

4.5. Conclusion

It is difficult to summarise the consequences of congestion. On the onehand, the emission increases and cost increases cause a negative impact on theenvironment as well as on the economy. On the other hand, we have to admitthat congestion is an indicator for the highly problematic cost structure of allroad transport. Because transport does not pay its real price, because it neglectsthe external cost, there is an increase in transport demand instead of making theenterprises and households organise their activities with less transport demand.If the transport costs were higher, enterprises would try to substitute transportfor other production factors in order to save money. This could either be theuse of alternative travel modes, or a spatial reorientation of production anddistribution chains. Also, private households would decide differently aboutlocations for housing and for other activities. Congestion increases the cost fortraffic participants and initiates research for alternative solutions.

Finally, these consequences could lead to a renaissance of local andregional workplaces which are often mentioned as very positive models forsustainable development in general. If we look at congestion, the directnegative environmental impact by emissions and fuel consumption is notdramatic. If society would decide in favour of measures to increase thecapacity of the infrastructure by very costly construction – not drawing on theconsequences of failures in this strategy during the last decades – this couldlead to even more transport demand, induced by improved conditions forcirculation. In the end, this could have far worse environmental consequencesand could also impose an economic burden on the public budgets which isgreater than the impact of congestion.

The current databases do not allow the specification of this discussion interms of calculation and prognosis; we have to acknowledge that theinteraction between the transport sector, the spatial development and social

38

behaviour cannot be computed. The question of newly-induced traffic, whencongestion conditions are removed, will depend on the future costs for mineraloil and for road use.

5. STRATEGY FOR THE REDUCTION OF CONGESTIONAND ITS NEGATIVE IMPACTS

Just as the causes of congestion are different, so are the possible measuresto reduce them. Here we focus on congestion caused by excessive trafficdemand on certain sections related to its designed capacity. Some of themeasures may be applicable to congestion caused by accidents and constructionwork, too.

To begin with, several decades of transport policy planning and financingthe extension of road networks have not seen any success in solving thecongestion problem. It seems to be impossible to avoid an increase of thenumber of vehicles on certain sections, as it seems to be impossible to preventcustomers from hurrying to buy an attractive, cheap product in a certain shop.When the customers want to buy it, they accept inconvenience and waitingtimes in the shop. In a free market economy, there may be a competitor withina short time providing better service, or the shops may hire more personnel toimprove the level of service. The shop may also increase the price of theproduct to increase profit up to an optimal level, given by the demand and theprofit per product.

The transport market does not show that simple market condition. If wesee congestion indicating very attractive conditions for customers, which wouldmean a large number of passenger cars and trucks in certain sections of theroad network, there are hardly any reactions on the side of the enterpriseproviding the supply (the State) nor from the side of competing enterprises (thismay be the railway). For a passenger car user and for a truck user, there ishardly any effective alternative available. At least not in the short run. Wehave to acknowledge that the advantages of passenger cars and trucks over itsalternatives exist even under congested conditions with reduced average travelspeeds. Goods transport via rail and passenger transport via publictransportation do not seem to be an attractive alternative to the motor vehicle,except in the heart of the cities where parking space is rare and very expensive.It should be accepted that these alternatives are only used when the restrictions

39

for road transport really hurt. It cannot be expected that congestion will ever bereduced by a massive shift towards the alternative transport modes, if suchadditional measures were taken. This leads to the conclusion that withinvestments in the other modes alone, congestion will not be reduced.

Although these experiences can be studied in many European cities, it isnecessary to develop alternative transport modes. One reason is that this wouldat least mitigate the increase in demand for passenger and truck traffic. Also,improvements in service quality of the alternative modes would be a politicalcondition to implement restrictive measures towards the passenger car and thetruck. Competitiveness of the alternatives must be seen relative to the car andthe truck; which means that a combination of improvements for the one modeand restrictions for the other modes may be an effective strategy. In urbantransport policy this is named push-and-pull strategy.

In general, we can differentiate between all measures to tackle congestioninto supply-side measures and demand-side measures. In abstract terms,demand would be reduced if the spatial resistance is increased. This couldeither be a cost increase – higher mineral oil taxes or road pricing – or increasein travel time. As we have discussed in another chapter, for part of the trips wecan directly relate time losses to cost increases; for other sectors of thetransport market this is not possible. In order to reduce the demand in motorvehicle traffic, cost increases or time losses would force the road users to lookfor strategies to reach the goals differently than by transport. Some of thepossible strategies of the market actors have already been mentioned. If weaccept transport to be a derived demand, then we must focus on the real socialand economic demand behind transport. An increase in transport costs, anincrease in travel time and similar measures would make all sorts ofalternatives very attractive. We can assume that a reduction in transportactivity would lead to less motor vehicle demand on congested road sections.

In order to use the cost instrument especially to avoid congestion onheavily trafficked road sections, road pricing can be applied very specifically.The price of road use can vary according to the time of day, or even start toincrease with increasing traffic demand. If the price per road use was adjustedcontinuously to avoid overload and critical flow conditions, this would be avery elegant way to let the traffic flow without any demand-related congestion.Car users would try to avoid peak hours because they are charged very heavily.They may even be charged excessively when there is a threat of congestion.

40

With modern transport telematics, this type of instrument can beimplemented without causing additional disturbances in traffic flow, astraditional toll road stations may do.

Congestion pricing seems to be the only realistic strategy to achieve areduction in congestion levels. This instrument has been recommended byeconomists for many years but has never been implemented consequently.There are some uncertainties about undesired side effects which may occur.Eventually, car drivers would drive other routes to avoid being charged, whichcould lead to an overall increase in fuel consumption and emissions. If onlypeak hours were charged, this would support a further shift towards using lowdemand hours, which could lead to more traffic noise in the evening and atnight, increasing the burden for the people living nearby.

With respect to the distributional consequences, a pricing scheme maycause disadvantages for the poorer part of the population, as has beenmentioned. It depends on the use made of the collected money, if this remainstrue. It is clear that due to the higher absolute amount of money spent fortransport services, the upper income classes would have higher costs inabsolute terms. Because low income groups use public transportation moreoften and spend less on passenger car use, the average cost increase due to roadpricing in general and congestion pricing especially, would be comparably low.For people depending on the automobile, the relative weight of the additionalburden would be higher than for the other income groups. This could bechanged if part of the collected money were directly channelled towards thelow income groups. The possibility of such compensation strategies cannot beelaborated further in this paper. In any case, we would not recommendspending the money on the construction of additional roads, because the supplystrategy has been proven to fail – see above.

41

BIBLIOGRAPHY

Acutt, M.Z., J.S. Dodgson (1997), Controlling The Environmental Impacts ofTransport: Matching Instruments To Objectives, TransportationResearch-D, Vol. 2, No. 1, pp. 17-33.

Baum, H., C. Maßmann, W.H. Schulz et al. (1992), Rationalisierungspotentialeim Straßenverkehr I. Forschungsvereinigung Automobiltechnik e.V.(FAT), Eds., FAT Issue No. 94, Frankfurt/Main.

Bayerische Motorenwerke AG (BMW) (1998), Umwelt und Verkehr:Abschätzung der volkswirtschaftlichen Verluste durch Stau imStraßenverkehr, without date (personal communication WI and D. Frankand J. Sumpf 1998)

Bundesverband Güterkraftverkehr und Logistik (BGL) e.V. (1997),Jahresbericht 1996/1997. Frankfurt/Main.

Bundesverkehrsministerium (Bonn) Eds. (1997), Verkehr in Zahlen 1997.Cologne. Deutsches Institut für Wirtschaftsforschung, Berlin.

Dreborg, Karl et al. (1997), Images of Future Transport in Europe, Final Draft,September, p. 15.

Fige GmbH (1994), Emissionsfaktoren für verschiedene Fahrzeugschichten,Straßenkategorien und Verkehrszustände (1). Second Interim-report,Research project No. 105 06 044 on behalf of Umweltbundesamtes,Herzogenrath.

Fige GmbH (1997), Maßnahmen orientiertes Berechnungsinstrumentarium fürdie lokalen Schadstoffemissionen des Kraftfahrzeugsverkehrs (Mobilev).Documentation and User-manual, Research project No. 105 06 044, on thebehalf of Umweltbundesamtes, Herzogenrath, Berlin.

42

Haag, M., C. Hupfer (1995), Wirkungen von Verkehrsmanagement- systematisch untersucht. Section Transportation, University ofKaiserslautern.

Hassel, D., P. Jost, F.-J. Weber, F. Dursbeck et al. (1993),Abgas-Emissionsfaktoren von PKW in der Bundesrepublik Deutschland.Abgasemissionen von Fahrzeugen der Baujahre 1986 bis 1990,Technischer Überwachungs-Verein Rheinland Sicherheit undUmweltschutz GmbH, Eds., on the behalf of Umweltbundesamtes,Cologne.

Heusch-Boesefeldt, Ed. (1995), Entwicklung von Strategien zur Vermeidungvon Verkehrsstaus auf BAB infolge des stark zunehmendenLkw-Verkehrs. Short-Report, Aachen, Berlin, Hamburg, München.

Hautzinger, H., D. Heidemann, Fahrleistungen und Unfallrisiko in derBundesrepublik Deutschland. Internationales Verkehrswesen, No. 12/97,pp. 634-641.

Kellermann, G., Geschwindigkeitsverhalten im Autobahnnetz 1992. Straße undAutobahn, No. 5/95, pp. 283-287.

Knörr, W., U. Höpfner, U. Lambrecht, H.-J. Nagel et al. (1998), Daten- undRechenmodell: Energieverbrauch und Schadstoffemissionen aus demmotorisierten Verkehr in Deutschland 1980 bis 2020. Institut fürEnergie-und Umweltforschung (IFEU), Heidelberg.

Landesanstalt für Umweltschutz Baden-Württemberg, Eds. (1996),Emissionsmindernde Maßnahmen im Straßenverkehr. Handbuch zurBeurteilung der Wirksamkeit, Karlsruhe.

Lensing, N. (1997), Straßenverkehrszählung 1995. Jahresfahrleistungen undmittlere DTV-Werte. Bergisch Gladbach. Report of Bundesanstalt fürStraßenwesen, Verkehrstechnik Issue V. 41.

MacKenzie, J.J., R.C. Dower, D.D.T. Chen (1992), The Going Rate: What itReally Costs to Drive. World Resources Institute, Eds., Washington.

Maennig, W.; M. Sames, K. Tullius, Verkehrsstaus im urbanen Raum -- Kostenund Lösungsmöglichkeiten am Beispiel Hamburgs. InternationalesVerkehrswesen (49) No. 11/97, pp. 561-568.

43

Plowden, S., M. Hillman (1996), Speed Control and Transport Policy. PolicyStudies Institute, Eds., London.

Teubel, U. Verteilungswirkungen von Straßenbenutzungsgebühren in einemstädtischen Ballungsraum, Internationales Verkehrswesen (49) No. 3/97,pp. 97-103.

Topp, H.H., Ed. (1995), Leistungsfähigkeit innerörtlicher Hauptverkehrsstraßenim motorisierten Individualverkehr bei verschiedenen Geschwindigkeiten.Section Transportation, University of Kaiserlautern.

U.S. Department of Transportation, Federal Highway Administration (1992),Examining Congestion Pricing -- Implementation Issues. Searching forSolutions -- A Policy Discussion Series, No. 6.

Umweltbundesamt, Eds. (1995), Handbuch der Emissionsfaktoren desStraßenverkehrs. Version 1.1, Berlin.

Umweltbundesamt, Eds. (1997), Jahresbericht 1996. Berlin.

Umweltbundesamt, Eds. (1996), Schadstoffemissionsberechnungen desVerkehrs mit dem Handbuch für Emissionsfaktoren. Möglichkeiten undGrenzen der Anwendung, speziell für Immissionsberechnungen nach § 40Abs. 2 BImSchG, Block I, 438. FGU-Seminar, Berlin.

Umweltbundesamt, Eds. (1996), Schadstoffemissionsberechnungen desVerkehrs mit dem Handbuch für Emissionsfaktoren. Möglichkeiten undGrenzen der Anwendung, speziell für Immissionsberechnungen nach § 40Abs. 2 BImSchG, Block II, 438. FGU-Seminar, Berlin.

Umweltbundesamt, Eds. (1996), Schadstoffemissionsberechnungen desVerkehrs mit dem Handbuch für Emissionsfaktoren. Möglichkeiten undGrenzen der Anwendung, speziell für Immissionsberechnungen nach § 40Abs. 2 BImSchG, Block III(1), 438. FGU-Seminar, Berlin.

Umweltbundesamt, Eds. (1996), Schadstoffemissionsberechnungen desVerkehrs mit dem Handbuch für Emissionsfaktoren. Möglichkeiten undGrenzen der Anwendung, speziell für Immissionsberechnungen nach § 40Abs. 2 BImSchG, Block III(2), 438. FGU-Seminar, Berlin.

44

Verband der Deutschen Automobilindustrie e. V. (VDA), Eds. (1995), Auto &Klima. Eine Präsentation des Verbandes der DeutschenAutomobilindustrie zur UN-Klimakonferenz Berlin 1995, Frankfurt/Main.

Verband der Deutschen Automobilindustrie e. V. (VDA), Eds. (1997), CO2:Automobilindustrie auf Reduktionskurs. VDA public-relation service,Frankfurt/Main.

Verband der Deutschen Automobilindustrie e.V. (VDA), Eds. (1997),Jahresbericht. Auto 1997. Frankfurt/Main.

Verband der Deutschen Automobilindustrie e. V. (VDA), Eds. (1997),VDA-Pressegespräch zu CO2 am 1. Dezember 1997. Ausführungen vonDr. Bernd Gottschalk, VDA public-relation service, Frankfurt/Main.

Verband der Deutschen Automobilindustrie e. V. (VDA), Eds. (1997),VDA-Pressegespräch zu CO2 am 1. Dezember 1997. Ausführungen vonProf. Dr. Gunter Zimmermeyer, VDA public-relation service,Frankfurt/Main.

Werkgroep ´2duizend, Delft University, Free University Amsterdam (1996),Time to Tame our Speed?. A study of the socio-economic cost and benefitsof speed reduction of passenger cars. Research Unit for IntegratedTransport Studies, Eds., Amersfoort.

45

FRANCE

Christian GERONDEAUChairman

Union Routière de FranceParis

France

46

47

ROAD CONGESTION IN WESTERN EUROPE

SUMMARY

INTRODUCTION..............................................................................................49

1. INTERURBAN ROADS............................................................................52

1.1. Technical characteristics: speeds and capacity .................................521.2. Volumes of traffic recorded and prospects.........................................53

2. URBAN NETWORKS...............................................................................59

2.1. Technical characteristics: travel speeds and capacity .......................592.2. Current traffic volumes and congestion .............................................62

3. THE CONGESTION TOLL: AN UNREALISTIC SOLUTION .............64

ANNEX: MAPS OF THE MAIN URBAN CENTRES IN EUROPE..............77

Paris, January 1998

48

49

INTRODUCTION

With the exception of three Mediterranean countries (Spain, Portugal andGreece) where the trend towards widespread use of the car has been slower, thevarious western European countries have very similar volumes of road traffic(all vehicles combined), amounting on average to about 7 500 to8 000 kilometres a year per head [Department of the Environment, Transportand the Regions (DETR), UK, see Figure 1].

A very interesting point is that the volume of traffic per head varies verylittle from one country to another and seems to be independent of the roadinfrastructure or the road transport taxation policies in use (DETR report,Table 18 – see Table 1 below).

The OECD estimate for the total volume of road traffic in western Europe(in 1995) -- generated by a total population of 380 million -- is 2 950 billionkm, with cars accounting for 2 500 billion, light commercial vehicles for295 billion and heavy commercial vehicles for 155 billion (reference: MotorVehicle Pollution, OECD, 1995).

About 30 per cent of this traffic is urban traffic and 70 per cent interurban,although the dividing line between the two is vague and the definitions varyfrom one country to another or even within the same country.

The traffic referred to above is carried on a road network which mainlycomprises five categories of roads, with two in the interurban and three in theurban sector.

50

Figure 1. Europe, annual kilometres per inhabitant in 1994(all motor vehicle traffic)

9

8

7

6

5

4

3

2

1

0

9

8

7

6

5

4

3

2

1

0

Source: National Road Traffic Forecasts 1997 ; Department of the Environment, Transport and the Regions, London.

Spa

in

Nor

way

Italy

Sw

eden

Net

herla

nds

Ger

man

y

Uni

ted

Kin

gdom

Sw

itzer

land

Den

mar

k

Irel

and

Fra

nce

Fin

land

Kilometres per yearKilometres per year

These categories are:

-- In the interurban sector:• conventional rural roads;• motorways and expressways;

-- in the urban sector:• conventional streets,• avenues and boulevards,• motorways and expressways.

The characteristics and utilisation levels of these roads must be describedbefore the scale of congestion problems can be defined.

Table 1. International statistics for national road traffic forecasts

GDP per head:$ 1994

Realchangein GDP

per head

Cars per 1 000population

Car traffic All motor vehicle trafficRatio of change intraffic per head to

GDP per head1984-94

At marketexchange

rates

Atpurchas-

ing powerparity

1984-1994*

1984 1994 Change1984-1994

Kms(000s)

per head1984

Kms(000s)

per head1994

Changeper head1984-94

Kms(000s)

per head1984

Kms(000s)

per head1994

Changeper head1984-94

Cartraffic

Alltraffic

UK 17 443 17 621 22% 297 372 25% 4.4 6.1 37% 5.5 7.4 35% 1.7 1.6Belgium 22 687 20 314 19% 335 416 24% 4.2 5.4 27% . . . 1.4 .Denmark 28 043 20 438 18% 282 309 10% 4.5 6.2 38% . 7.7 . 2.1 .France 22 987 19 232 16% 378 430 14% 4.7 6.1 28% 6.3 8.4 33% 1.8 2.0Germany 25 133 19 671 20% 365 488 34% . 6.2 . 5.4 7.2 32% . 1.6Greece 9 388 11 582 14% 116 188 61% . . . . . . . .Irish Rep. 15 099 15 794 53% 203 263 29% . 6.4 . . 7.9 . . .Italy 17 768 18 648 21% 369 532 44% 3.6 5.6 54% 4.9 6.8 39% 2.6 1.9Luxembourg 36 089 30 198 60% 399 567 42% 5.5 8.2 49% . . . 0.8 .Netherlands 21 896 18 723 22% 336 383 14% 4.9 5.8 19% 5.7 7.1 24% 0.9 1.1Portugal 8 575 12 027 39% 127 263 108% 2.1 3.7 74% . . . 1.9 .Spain 12 337 13 596 30% 232 343 48% 1.4 2.9 100% 2.0 3.8 91% 3.4 3.1Finland 19 186 16 274 8% 302 368 22% 5.1 7.0 36% 6.2 8.4 34% 4.4 4.1Norway 28 423 21 956 26% 345 381 10% 4.5 5.4 21% 5.2 6.5 24% 0.8 1.0Sweden 22 598 17 583 7% 370 409 11% 5.3 6.2 17% . 6.9 . 2.6 .Switzerland 36 669 23 860 10% 392 450 15% 5.1 6.3 23% 6.7 7.5 12% 2.3 1.2Australia 18 187 18 517 18% 439 460 5% 6.8 6.9 1% . . . 0.1 .Japan 37 509 21 171 33% 226 344 52% 2.2 3.3 50% . . . 1.5 .USA 25 512 25 512 18% 495 514 4% 8.3 9.8 18% 11.7 14.6 25% 1.0 1.4* Based on GDP at constant 1990 prices: OECD.Notes: United Kingdom: Norther Ireland traffic not available for 1984; traffic ratios based on GB figures.

Germany: Estimates for 1984 are problematic, as they involve imputation for former East Germany. Estimates of real change in GDP is for former WestGermany only.USA: Definition of “car” changed in 1991 and a consistent series is not available. Approximate adjustments have been made to allow reasonableestimates of change.

Source: Department of Transport: International Comparison of Transport Statistics, 1970-1994 and later information held by DETR.

52

1. INTERURBAN ROADS

1.1. Technical characteristics: speeds and capacity

The characteristics of the two categories of interurban roads are asfollows:

� Conventional rural roads:

Conventional two-lane rural roads provide limited capacity (usuallyless than 10 000 vehicles a day when traffic is normal). The averagespeeds possible on them are mostly quite low owing to their speedlimits which are imposed for safety reasons, their geometriccharacteristics and the frequent slowdowns through villages andbuilt-up areas. These average speeds usually range from 40 to70 km/h, and in some cases even less.

The characteristics of some interurban roads are half-way betweenthose of conventional rural roads and motorways, in which case theymust be given special attention from the safety viewpoint.

� Interurban motorways and expressways:

Interurban motorways and expressways present a completely differentpicture from conventional roads in terms of their capacity, theaverage speeds possible on them and safety.

The average daily capacity over a year for interurban motorways dependson their technical characteristics (gradients and horizontal curve radii), trafficstructure (in particular the proportion of lorries), seasonal variations in traffic,and what is considered to be the acceptable level of mutual difficulty created byvehicles. The following average values, however, can be used:

� 50 000 vehicles/day on two x 2 lanes;� 80 000 vehicles/day on two x 3 lanes;� 110 000 vehicles/day on two x 4 lanes.

Such levels involve dense flows resulting in occasional difficulties andslowdowns, even outside any peak periods in recreational travel (weekends andholidays).

53

Some countries have therefore opted for lower values in theirinfrastructure design policies and widened their interurban motorways beforesuch traffic levels are reached (for example, by going up to two x 3 lanes asfrom 30 000 vehicles/day), as in the case of France. The values quoted above,however, which correspond to about 13 000 vehicles/day per lane, are used bymost countries to estimate the capacity of interurban motorways.

Travel speeds depend on the regulations in force (usually 100 to130 km/h) for light vehicles when traffic is flowing smoothly.

1.2. Volumes of traffic recorded and prospects

Conventional interurban roads

Some conventional two-way interurban roads are close on or have alreadyexceeded the level of traffic consistent with satisfactory operation (about10 000 vehicles/day). The solution is then to convert them into two x 2 laneroads, or to build parallel expressways or motorways. In both cases thecorresponding capacity increase is very high and capable of absorbing a trafficvolume that has risen by a factor of 3 to 4. Such operations are required on agreat many sectors of the conventional interurban road network in Europe.

When these operations are properly carried out, they result in a radicalimprovement in safety as casualty rates are reduced by a factor of 3 to 4, andfuture generations will be shocked by the fact that heavy flows of vehicles weretolerated for so long on two-way single carriageways.

Interurban motorways

Interurban flows are usually far less dense than urban flows. For instance,the average volume of traffic on interurban motorways in western Europe canbe estimated at 30 000 vehicles/day, a level which can be easily be absorbed bya motorway with two x 2 lanes.

On some 38 000 kilometres of interurban motorways in western Europe,average daily traffic exceeds some 50 000 vehicles on only about 5 000 km, asagainst 80 000 vehicles on about 300 km (less than 1 per cent), of which ahundred or so are in Germany and slightly over a hundred in the UnitedKingdom. At present traffic does not exceed 90 000 vehicles/day on anyinterurban motorways in Europe.

54

Present situation

As we have seen previously, such traffic is substantially less than thecapacity of an interurban motorway with two x 4 lanes. It therefore seems thatthere should be no recurrent congestion problems at present on interurbanmotorways, provided that their width is adjusted where necessary to the volumeof traffic to be absorbed, since interurban motorway capacity can almostinvariably be improved either by increasing the number of lanes (for example,by going up from two x 2 lanes to two x 3, or two x 4 lanes), or in exceptionalcases by straddling them with two more carriageways, an operation which maybe less costly and easier than widening a carriageway as there is no disturbancefrom works on the existing facility.

The peak traffic flows involving weekend departure and return trips in thespring and summer or during the holidays must obviously be excluded, as therewould be no justification for using them as a basis for network design and theyare all the more acceptable since they are generated by recreational travel.Moreover, staggering policies can be used to reduce peaks, and the situationhas therefore been improved in this respect in a number of countries.

With the exception of these weekend migration and holiday departurepeaks for which it would not be reasonable to design motorways, there are veryfew cases of recurrent congestion on interurban motorways in western Europe(excluding the periods of road maintenance and widening works). Only a verysmall number of countries and a few hundred kilometres at most are affected.

It would be possible to put a stop to this limited congestion in the shortterm by means of appropriate design modifications, which seems to be all themore justified as the very high demand on the sections concerned guaranteeshigh economic and social returns from the works on the required capacityincrease.

Apart from these exceptional cases, congestion mainly occurs duringweekend and holiday migrations. It should be stressed that it therefore occursat times when lorry traffic is at minimum or even non-existent. Trying toreduce congestion on interurban motorways by reducing lorry traffic istherefore pointless, since almost all congestion occurs when there are veryfew or no lorries on the road network. We shall see subsequently that thesituation is similar, from this point of view, in the case of urban motorways.

55

On the whole, the traffic flow is smooth on European interurbanmotorways, which explains in particular the success of the just-in-timeprocedures that have developed so rapidly in industry in the past decade andhave transformed Europe into one vast production plant.

Prospects

Interurban motorway traffic is on the whole still rising in western Europe.The average increase rate is, however, now moderate, and usually rangesfrom 2 to 4 per cent a year. The most recent estimates have been made by theUK Ministry of the Environment, Transport and the Regions (National RoadTraffic Forecasts 1997). Increases of 50 per cent are forecast in the totalvolume of traffic on UK interurban motorways as a whole within 15 years and100 per cent within 30 years, which corresponds to overall growth in traffic, allnetworks combined, of 30 per cent and 60 per cent within these periods. Butthese estimates are challenged by other UK sources, which refer to totalincreases of 20 per cent and 40 per cent at the same dates (SMIT). The Frenchestimates are also lower than those of the UK Ministry.

It obviously cannot be said which of these estimates will materialise,given that it is also possible to see things in a different light, by relating theuncertainty not to the level of traffic at a given time but to the date at which agiven traffic level will be reached. The uncertainty is reduced by taking thisview, since it is not necessary to know today whether, for example, a 40 percent increase in traffic will be reached in 2025 or in 2030, as it does not affectthe decisions to be taken today.

In the remainder of this report the following estimates will be used foraverage traffic growth on western European interurban motorways comparedwith 1996: 50 per cent within 15 years (2011) and 100 per cent within 30 years(2026).

But this obviously does not mean that traffic will necessarily keep to thistrend on each of the interurban motorways concerned. This is particularly thecase when the motorway network can be enlarged, so that the traffic increase isspread over a greater length of motorway. Even if there is no change in thenetwork structure, it is to be expected that the increase will in most cases begreater on the motorways used least. It is therefore to be expected that theincrease rate will be in most cases lower than average on the motorways thatare now used most.

56

Even with similar traffic growth over the entire existing motorwaynetwork, which is assuming a great deal, the situation would be as followswithin 15 and 30 years on western Europe’s existing interurban motorways,taking into account the estimates selected above.

Table 2. Forecasts for traffic on the various categories ofinterurban motorways

Present situation(1996)

Year 2011 Year 2026

Category 1(3 300 kilometres)

fewer than50 000 vehicles/day

fewer than75 000 veh/day

fewer than100 000 vehicles/day

Category 2(5 000 kilometres)

from 50 000 to80 000 vehicles/day

from 75 000to 120 000 veh/day

from 100 000to 160 000 veh/day

Category 3(300 kilometres)

over 80 000vehicles/day

over 120 000veh/day

over 160 000veh/day

It emerges from this table that, in fifteen years’ time, only a few hundredkilometres of motorway might have a traffic exceeding the capacity of atheoretical two x 4 lane motorway. In thirty years’ time, the correspondinglength might rise to 2 500 km, or 6 per cent of the length of a western Europeaninterurban motorway network, which at that time will probably be about50 000 km long.

As we have seen, an accurate forecast for that date is extremely difficult tomake and is subject to considerable uncertainty.

In fact, only specific studies by route would make it possible to refinethese values, which are probably a maximum considering the factors statedabove, and in particular the fact that traffic naturally tends to spread betweencompeting routes, and therefore to increase less on the busiest roads.

In any case, the works for widening existing motorways to two x 3 ortwo x 4 lanes, or possibly for duplicating them over certain very limitedsections, cost very little compared with the economic and social benefitsprovided by the efficient operation of the network, which accounts for the bulkof European transport activity and plays a central role in the continent’seconomy.

57

It should be pointed out that the sums invested in road transport by thecommunity as a whole (government, firms, individuals) amount at present inwestern Europe to about Ecu 900 billion a year. But the works on increasingthe capacity of the interurban motorways mentioned above would cost, over aperiod of 30 years, about Ecu 150 billion, or 5 billion a year, i.e. less than1 per cent of the present level of community spending on road transport.

It must be added that even if it is not possible to carry out the desirableworks on some limited sections of the network -- which is by no meansnecessarily the case -- this would in no way prevent the operation of theinterurban motorway system. Recurrent congestion over a very smallproportion of its length would result unfortunately in time losses, but would inno way prevent the greater part of the system from continuing to workproperly.

As the section of this report on urban and suburban motorways will show,motorway traffic may at times greatly exceed the theoretical capacity referredto above without affecting what is, by and large, quite a smooth flow of traffic.

In the last analysis, there is no good reason why the operation of the entireor almost the entire western European interurban motorway network should notgive full satisfaction, provided that, where necessary, the works to increase itscapacity which are justified from the economic and social viewpoints arecarried out.

Only if no or extremely inadequate improvements were made to themotorway network would there be any risk of time losses and congestion(cf. DETR report, Table 3, see below).

Table 3. National Traffic Forecasts by Road Type (Central Estimates)

RURAL ROADS URBAN ROADS

Motorways Trunk andPrincipal1 Dual

Other2 Total Motorways Trunk andPrincipal3

Other4 Total

1996 traffic(bn veh kms)

57.5 49.3 149.1 255.9 15.9 74.3 92.1 182.3

1996=100 100 100 100 100 100 100 100 100

2001 116 110 107 110 110 106 110 108

2011 152 129 122 130 129 116 132 125

2021 188 146 136 150 142 125 153 141

2031 217 159 146 165 150 131 170 152

Note: Urban areas are those of continuous built development, while all others are rural. The resulting traffic figures differ from thepublished traffic statistics presented on the built-up and non-built-up basis, which are solely determined by speed limits.Thus roads in urban areas with 50 mph and higher speed limits are urban for the purposes of these forecasts, but non-built-upfor traffic statistics.

1 Dual carriageway “A” roads.2 Single carriageway “A” roads, B, C and unclassified roads.3 Dual and single carriageway “A” roads.4 B, C and unclassified roads.Source: Report of the Department of the Environment, Transport and the Regions, London.

59

2. URBAN NETWORKS

2.1. Technical characteristics: travel speeds and capacity

Conventional streets

The characteristics of urban roads differ greatly depending on whetherthey are conventional streets, avenues and boulevards or motorways andexpressways.

Old towns were not designed for today’s road vehicles, but mainly forpedestrians. This is why conventional streets are narrow, and often cover onlya very small part of the ground space in the areas served by them (10 per centto 15 per cent). Considering that their geometric pattern is also usuallyirregular, they can provide only very poor conditions for today’s vehicles,whether as regards capacity or the average speeds possible.

Conventional streets usually have a capacity of under 10 000 vehicles/dayand the average speed possible on them, except at night, is often 15 kilometresan hour at most, which has sometimes prompted the remark, albeit without dueregard for the other categories of roads, that the present situation showed noreal progress in terms of speeds over the days of horse-drawn vehicles -- whichis, for the most part, completely erroneous.

It must be added that conventional streets are found not only in old towns,and that some suburbs are served entirely by such streets which are not suitablefor car traffic and sometimes reproduce the irregular pattern of the formercountry lanes.

Avenues and boulevards

Prior to the 19th century, few urban roads were wider than ten or sometres. It was in 17th century Versailles that very wide, straight-runningavenues (almost 100 metres in breadth) were laid out for the first time, with theaim of opening up magnificent vistas onto the Sun King’s château, but with nofunctional purpose whatsoever in terms of transport.

In the course of the 18th century, a number of European monarchs wereinspired by Versailles and had wide avenues built in their capitals(Saint Petersburg, for example). But such avenues were few in number, and it

60

was not until the 19th century that large numbers of straight-running avenuesand boulevards about 30 metres wide or more between facades were built inmany European towns.

The first were built in London in the first half of the century, but it was inParis that construction work on the largest and most harmonious network ofboulevards and avenues was started in 1850, by command of the EmperorNapoleon III who was greatly impressed by the example of London where hehad lived for many years. In the wake of the French capital, many otherEuropean towns built similar kinds of roads which proved very suitable for cartraffic in the 20th century.

Avenues and boulevards can absorb a traffic exceeding 30 000 vehicles aday and in some cases much greater volumes with the use of modern operatingtechniques. The average speeds possible on them may be as high as30 kilometres an hour or more, except in periods of intense congestion.

The same may be said of suburban roads when they have physicalcharacteristics similar to those of urban avenues and boulevards.

Urban and suburban motorways and expressways

The appearance in the United States and then in western Europe of urbanand suburban motorways and expressways marked the opening of an entirelynew era for car traffic, for these roads were the first to be designed specificallyfor cars. They have almost nothing in common with conventional streets, andtheir potential is very much higher than that of avenues and boulevards in termsof both capacity and speed, not to mention safety.

Urban motorways have about twice the capacity of interurban motorways,since traffic is distributed more regularly throughout the year and day and,owing to the shorter trip distances, speeds on them can be lower, which giveshigher hourly flows.

More precisely, in the conditions prevailing in western Europe, capacityby lane ranges from 20 000 to 30 000 vehicles/day depending on thedistribution of traffic over the day. The level of 30 000 vehicles/day is,however, quite exceptional, and in practice is recorded only in the Paris andMadrid urban regions.

Such flows result in the following capacities:

61

Table 4. Daily capacity of European urban motorways

Capacity by lane 20 000 vehicles/day 25 000 vehicles/day 30 000 vehicles/day

two x 2 lanes 80 000 vehicles/day 100 000 vehicles/day 120 000 vehicles/daytwo x 3 lanes 120 000 vehicles/day 150 000 vehicles/day 180 000 vehicles/day

two x 4 lanes 160 000 vehicles/day 200 000 vehicles/day 240 000 vehicles/daytwo x 5 lanes 200 000 vehicles/day 250 000 vehicles/day4 x 3 lanes 240 000 vehicles/day 300 000 vehicles/day

All the capacity data in the table above apply to sections varying in lengthon the urban and suburban motorway networks of various cities in westernEurope. The values are therefore very high, even if the capacity of the vastmajority of European urban motorways corresponds to the boxes marked by aheavy line in the above table since they refer to two x 2 or two x 3 lanemotorways, where daily capacity per lane does not exceed 20 000 to25 000 vehicles a day. It should be added that the traffic actually recorded isoften less than this capacity, since demand is by no means up to the saturationlevel in all cases, except for a few brief periods.

The average speeds on these roads have nothing in common with thosepossible on conventional streets, or even on avenues and boulevards.

Recent studies have shown that, outside the holiday periods, the average weeklyspeed was 53 km/h on the busiest European urban motorway where average trafficdensity is 30 000 vehicles a day per lane, or 240 000 vehicles/day for two x 4 lanes.(The reference is to the Paris inner ring road, which has been wrongly named theBoulevard Périphérique.)

Studies on other motorways in the Paris region have resulted in estimatesfor average journey speeds of around 70 km/h or more, despite flows oftenamounting to 25 000 vehicles/day per lane. There seems to be no reason whythe situation should be all that different on most of the urban motorways inwestern Europe, especially as their daily traffic is substantially lower than thatrecorded in the central part of the Paris region.

Although marked congestion occurs at peak hours on working days andusually in a single direction, urban motorways absorb traffic without majordifficulties during most hours of the day. Urban traffic is almost light for up to12 hours a day or even more. In other words, only a very small proportion oftraffic is affected by congestion on urban and suburban motorways andexpressways. The fact must also be taken into account that working days

62

represent scarcely more than half the total number of days in the year, whichexplains why average speeds over the year mostly exceed 70 km/h on urbanand suburban motorways in western Europe.

2.2. Current traffic volumes and congestion

Almost all congestion in western Europe occurs in the urban and suburbanenvironment.

It is, however, difficult to quantify congestion owing to a definitionproblem, since congestion does not have the same meaning for everybody(cf. Is our transport system sustainable?, R. Prud’homme et al.). At least fivedefinitions can be given.

1. According to some authors, the term congestion is associated withdifficulty: congestion occurs as soon as traffic does not flowsmoothly, even if it is at 2 o’clock in the morning.

2. It is also possible to define arbitrarily a reference speed for traffic,and say that congestion occurs when speed falls below this level.

3. The economist considers that a congestion cost is incurred when thetraffic flow exceeds an “optimal” level that is located at the pointwhere the traffic supply and demand curves intersect, i.e. when everyuser has to pay the “external costs” for the time losses he causesamong other users. This optimal level varies with demand, andtherefore in relation to time and place.

4. Transport engineers consider that congestion occurs close to theroad’s capacity limit, i.e. at the flow level which blocks traffic, takinginto account the particular form of the speed-flow curve. Instabilityand blocking occur when the traffic volume is 10 to 15 per cent underthe road’s capacity limit.

5. The definition of congestion by the user will be mainly qualitativeand will vary from one individual to another. “Congestion” will thencome somewhere between the “difficulty” experienced by the driverand a completely blocked network.

It must be clearly understood that the first definition of “congestion”makes no economic sense in that it would be neither realistic nor reasonable todesign the road network in such a way that drivers would never encounter any

63

difficulties on it. In an urban environment, such an approach would result inobviously oversized infrastructure and would be very far removed from theeconomic optimum.

“Congestion cost” may therefore have quite different values depending onthe definition of congestion which is used, and this is a major source ofmisunderstandings.

Classification of European urban centres

The major European urban centres are quite heterogeneous in terms ofmotorways and similar roads. The seven maps in the annex show the networksfor motorways and similar roads in areas of the same size (63 x 89 km). Thefollowing seven regions are considered:

� Berlin (Germany), population 6 million;� the Ruhr (Germany), population 10 million;� the Randstad (Netherlands), population 7 million;� Milan (Italy), population 4 million;� Madrid (Spain), population 5 million;� Paris (France), population 9 million;� London (United Kingdom), population 7 million.

The lengths of motorways and similar roads as well as network designvary greatly across these regions, and it is understandable that the problemsencountered are quite different, whether as regards the geographicaldistribution of activities (homes, jobs, shops, etc.) and traffic flows.

To simplify the picture, it can be said that in areas with dense networks ofmotorways and similar roads serving almost the entire population, trafficconditions are on the whole satisfactory, even if peak hour congestion, whichmust not be exaggerated, necessarily occurs at different times and places. It ispossible to refer in this respect to the Randstad, Paris, Madrid and the Ruhrregions.

At the opposite extreme, an urban centre like London which has nomotorway network over most of its area means that its traffic flow capacity,expressed in vehicle-kilometres per km2 of land, is very low, which results invery poor traffic conditions and greatly reduces the use of land, particularly forhousing. There are therefore three times fewer inhabitants per 100 km2 in the

64

centre of London than in Paris where the high population density is explainedby the fact it is possible to own and use a car. (Car ownership for householdswith children in inner Paris amounts on average to 1.06.)

The great influence of road and motorway networks on land use, on thescale of congestion and finally on attitudes to the possibilities provided by useof the car can thus be clearly seen.

3. THE CONGESTION TOLL: AN UNREALISTIC SOLUTION

When the volume of traffic exceeds a certain threshold on a road or a roadnetwork, vehicles obstruct one another, so that their average speed declines. Aswill be seen subsequently, “congestion” in the usual sense of the word does notimmediately occur when this threshold is exceeded. It takes place only whenthe volume of traffic greatly exceeds that at which the first signs of mutualobstruction and a slowdown appear.

The more the traffic rises above the “obstruction threshold”, the morevehicle speed decreases to around a flow which cannot be exceeded and whichis the road’s capacity limit. At around this flow speeds become irregular andtraffic hold-ups occur as seen in the very special form of the speed-flow curve(Figure 2). This curve shows the existence of two quite distinct traffic patterns.In the upper part of the curve, the decrease in speeds is offset by an increase inthe flow. This is the “normal” or “primary” pattern. In the lower part of thecurve, the flow decreases at the same time as speed. This is an “enforced” or“secondary” pattern. But the most useful type of graph for theoretical studieson congestion and on the effects of introducing “congestion tolls” is one thatshows the relations between road traffic density and travel time (Figure 3).

65

Figure 2. Paris, Inner Ring Road, speed/flow per lane curve

0 2 000200 400 600 800 1 000 1 200 1 400 1 600 1 800

90

0

80

70

60

50

40

30

20

10

90

0

80

70

60

50

40

30

20

10

Speeds in km/hour Speeds in km/hour

Flow per lane in vehicles/hour

Figure 3. Paris, Inner Ring Road, travel time/flow per lane curve

0 2 000200 400 600 800 1 000 1 200 1 400 1 600 1 800

3.0

0

2.5

2.0

1.5

1.0

0.5

3.0

0

2.5

2.0

1.5

1.0

0.5

(Ci)

(D)

(Cc)

Q

P

RG

E(Ci)A

D

B

Y

Travel time in minutes per kilometre Travel time in minutes per kilometre

Flow per lane in vehicles/hour

66

The case represented here is that of an urban motorway but, as far as theprinciples are concerned, the findings are the same for any other type ofnetwork.

Traffic volume expressed in vehicles per hour on one lane appears on thex-axis. The cost per kilometre expressed here in minutes per kilometre appearson the y-axis.

The figure shows three curves, two of which relate to traffic supply andone to demand.

The first Ci (individual cost) represents the trend in travel time-- identified here with cost -- for a vehicle user depending on traffic density,since the time taken is the main component in the cost perceived by themajority of users, and no account will be taken here of the other costcomponents that would not affect the conclusions reached. When the trafficvolume is low, average speed is “freely” selected at a level estimated here at120 km/h, which corresponds to a travel time of 0.5 minute per kilometre. Upto a traffic of about 600 vehicles an hour in each lane, i.e. on average with aspacing between vehicles of 100 metres, there is no obstruction betweenvehicles, and average travel time is not affected. The traffic in a lane isprogressively slowed and the cost of the trip, expressed in travel time, rises foreach user (curve Ci).

But each vehicle is not only a victim in this situation. It is also partly aculprit, since it obstructs other vehicles and therefore contributes to theslowdown.

The cost to the community is therefore not only the cost met by thevehicle itself as a result of its slower speed, for the cost reflected in the losses itinflicts on other users must also be included. This results in a total collectivecost Cc = Ci + Cs representing the actual cost to the community when anadditional vehicle has moved into the lane in question. The additional cost Csis equal to the derivative of Ci multiplied by traffic volume.

It can be seen that the gap between the curves Ci and Cc becomesincreasingly wider the closer we are to the capacity limit which corresponds, inthe case of a motorway, to a density of about 2 000 vehicles an hour in eachlane (a vehicle every 40 metres) and to a speed of about 60 km/h.

67

In cases of serious congestion, the demand curve cuts the supply curve (Ci)beyond the point marking the road’s capacity limit, meaning that the trafficpattern is then of the “enforced” type.

The third curve shown in Figure 3 is the demand curve (D). The shorterthe travel times, the higher is the level of traffic demand, and vice versa. Thismeans that demand varies inversely with travel times, and its slope isnecessarily negative.

Two remarks have to be made concerning this demand curve.

Positioning of the demand curve

The first is due to the fact that, in the vast majority of the circumstancesoccurring on road networks in Europe, the demand curve will cut the supplycurves between points R and G, or at a point to the right of G but quite closeto it.

The reason is that, in the vast majority of cases, the traffic levels are belowthe obstruction threshold, and speeds are “freely” selected (provided of coursethat the regulations are respected), or the levels are slightly above this thresholdand speeds are then only slightly affected.

In this respect interurban roads should be distinguished from urban roads.

Interurban roads

1. Conventional roads

The traffic volume is low on most conventional interurban roads (a fewhundred or a few thousand vehicles a day) and the vehicles therefore do notobstruct one another. When traffic exceeds about 10 000 vehicles a day, thereis usually an economic justification for considering the possibility ofduplicating the infrastructure, either by increasing the number of lanes or bycreating a new motorway-type facility in order to improve available capacity.

2. Interurban motorways

Interurban motorway capacity is very high. As we have seen, it is usuallyestimated at 50 000 vehicles a day on average for two x 2 lanes; at 80 000 fortwo x 3 lanes and 110 000 for two x 4 lanes. But the average traffic recordedon European interurban motorways is about 30 000 vehicles a day. The

68

demand curve is then on the left of Figure 3. Apart from exceptional cases, forwhich a solution can be found by adding additional lanes, European interurbanmotorways are operated in normal traffic periods at well below their capacity,outside the exceptional peaks due to leisure migrations, which may also affectconventional interurban roads.

Urban roads

Virtually all cases of recurrent congestion occur on urban roads. But, evenin an urban environment, the vast majority of journeys in Europe are notaffected by serious time losses due to congestion. The reason is that manyvehicles are used at times and on days when traffic is slack. Many trips aremade in small or medium-sized towns, where congestion does not last verylong and is limited. Many take place in the outlying districts of major urbancentres where traffic flows are not so dense as in the central areas, consideringthe road capacity available. Many are in the opposite direction to the mainflows between centres and outlying districts.

The fact that the majority of urban trips are not seriously affected bycongestion is confirmed by the surprisingly short average duration of car trips.Door-to-door, they do not exceed on average about 15 minutes in Frenchprovincial towns, and 20 minutes for the United Kingdom as a whole, allreasons for travel combined.

Average home/work travel times for those who usually drive to work is17 minutes in the United Kingdom and Italy, 18 minutes in France and25 minutes in Germany, with an average in western Europe of less than20 minutes (cf. Christian Gerondeau: Transport in Europe). Most of thesetrips take place in urban or suburban areas. Even in the largest urban centres-- Paris and London -- average home/work trip times by car are about27 minutes. But, by definition, such trips mainly take place during peak hours,and in addition are above average in length. This finding is confirmed by thefact that 90 per cent of the Europeans who drive to work say that they areusually not held up by traffic jams (cf. Christian Gerondeau, op. cit.).

Contrary to the ideas often expressed, it therefore seems that by and largeserious congestion is mainly limited to the centre and near the centre of a fewmajor built-up areas, usually only for a small part of the time and, outside thecentral areas, to only one of the two traffic directions.

69

Even if there are obvious exceptions to which special attention should begiven, the demand curve (D) therefore mostly cuts the two supply curves(Ci and Cc) at points where the traffic density does not result inseriouscongestion and therefore does not justify any congestion toll, taking intoaccount the levels of charges that already exist in Europe. If this was not so,average door-to-door times by car would be much higher.

Demand elasticity

The second comment called for by the curve (D) is due to the fact thattraffic demand elasticity in relation to travel time -- which is usually the maincost component -- is low, especially at times when demand is at its highest,since most of the peak-hour trips cannot be avoided (home/work travel inparticular), however long they take.

In addition, apart from a minority of cases (trips to the centre of somemajor capitals served by a very dense and efficient public transport network),the time gained by driving is such that, even if the average traffic flow speed islow, the car is still much faster door-to-door than public transport. The averagetime gain by using the car instead of public transport has been estimated at aminimum of half an hour in today’s traffic conditions (cf. Christian Gerondeau,op. cit.). This is an essential although largely unacknowledged fact. In a greatmany cases, moreover, there is quite simply no credible alternative to the car.

It is therefore clear that traffic demand is finally not very sensitive tovariations in travel times (i.e. to speeds), which explains the steep slope of thecurve (D), with a slight decrease in traffic demand corresponding to a steepincrease in travel time.

The principle of the congestion toll

Even if the cases where demand is very high compared with road capacityare in a minority over time and in space, they should be given quite specialattention. In these cases the curve (D) is well to the right in Figure 3, and cutsthe curve (Ci) at a point where traffic is heavy, and where travel time is highcompared with the time taken in free-flowing traffic.

Two cases must be defined depending on whether the demand curve (D)cuts the lower part of curve (Ci), which corresponds to a “normal” trafficpattern, or the upper part of this curve, which corresponds to an “enforced”traffic pattern.

70

1. “Normal” traffic pattern

In the case illustrated by Figure 3, the travel time at the point concerned(A) is one minute per kilometre, or double that permitted by unrestricted speed(taking regulations into account). The corresponding speed is 60 km/h insteadof 80 or 90 km/h. Such a situation does not reflect the economic optimum,which is located at point B, where the community cost curve (Cc) cuts thedemand curve (D). For if all drivers had to meet the real cost generated bythem, some of them would decide not to use their cars.

Ideally it is the traffic level (Y) which should be recorded, since it is atthis level that the gains obtained by the community from the motorway aremaximised. To reduce traffic to this level, it can be seen that a toll equal to EBmust be added to the cost Ci directly incurred by the individual user.

This toll is known as the “congestion toll”.

A new balance is achieved with this approach. Average speed increasesand trip time per kilometre declines. In the example illustrated by Figure 3, thenew speed is 75 km/h, and the travel time per kilometre is reduced to0.80 minutes. In no case is there a return to completely free-flowing traffic:the economic optimum does not correspond to a situation in which speed isfreely selected. This essential point must be stressed.

The users who have remained on the motorway have therefore obtained acollective time gain which is proportional to the number of vehicles as well asto the time gained by each of them. It is represented in Figure 3 by therectangle PRED.

But the users whom the toll has forced off the motorway must be takeninto account. They have lost benefits represented by the triangle ABD.

The total gain from the introduction of the congestion toll EB is finallyequal to the difference between the two above-mentioned values, i.e. to PREDminus ABD.

But to obtain this result, all the users remaining on the motorway have hadto pay a toll equal to EB, which corresponds to the rectangle QREB.

Figure 3, which corresponds to a realistic situation, thus shows that, toobtain the optimal economic gain, tolls for which the grand total greatlyexceeds this gain must be charged in this particular case.

71

Although at first sight such a result appears very surprising, there is noreason why it should be. It is substantiated by the very high prices which haveto be charged in the central areas of large towns if parking fees are to beeffective. It confirms that users will do without their car only as a last resort,considering the very many advantages it provides in the vast majority of cases.Tolls must be very high if they are to be a deterrent to use of the car.

This situation has at least two kinds of consequences.

The first has an economic bearing. If the total amount of tolls to becharged to obtain a gain of 1 is about 6, collection costs of no more than 15 percent of the sums collected -- which is a low assumption -- will be sufficient toremove any justification for the operation. Even if this is not the case, it issufficient if the sums collected are not used on an optimal basis, therebyresulting in limited losses, for the results to be again negative. But there is noguarantee this will not happen.

The second consequence has a political bearing.

Owing to the high level of tolls needed to obtain the desired result, greatreluctance to introducing the congestion toll defined above must be expectedfrom the political authorities.

Admittedly, all in all, less time will be spent on the road, which iscertainly an improvement from the economic viewpoint. But if this is to bepossible, vast sums will have to be collected. From the economic viewpoint,these are simply “transfers”. But for those paying them, they are in factexpenditures. The body which has collected the tolls will therefore admittedlyobtain additional resources with which it can do as it pleases. If it transfersthem back to the users, the operation may be justified theoretically. But it willnot necessarily do so. Owing to the factors discussed above, it is doubtful thatthis is an optimal method of collecting money. Other solutions exist. It can beconsidered that, in almost every case, the economic gains which may beobtained by introducing congestion tolls will not outweigh their disadvantages,not to mention the practical difficulties involved, particularly with regard to theprocessing of offences in major urban centres, where they may amount tohundreds of thousands.

It must be accepted that the vast majority of road users will lose out insuch an operation. This obviously applies to motorists who will have to stopusing their cars. But it is also the case of most other users. On average, toachieve a time gain equivalent to ED, the toll to be paid will be much higher,

72

and equal to EB. The only lucky ones will be the minority group for whom thepersonal value of the time gain exceeds EB, either because they are rich orbecause the tolls are refunded by their firms as their time is so precious. This isa paradoxical aspect of the congestion toll. The community is supposed tobenefit, but the vast majority of users lose out. Only if the sums collected werereturned in one way or another to the users, which will not necessarily be thecase, could it be otherwise.

This finding does not mean that the principle of urban tolls is to berejected. There are facilities which can be partly or entirely financed bylevying tolls. In this case they are collected for strictly financial purposeswhich can be justified in practical terms when the aim is to create a newfacility. But such tolls have nothing to do with congestion tolls from which theproceeds should go at the very least just as much to existing roads as newroads. In particular, the tolls levied in a number of Norwegian towns cannot beseen as congestion tolls, since they are too low to have a substantial impact ontraffic volumes.

2. Enforced traffic pattern

The circumstances in which the demand curve (D) cuts the upper part ofthe supply curve (Ci) in the absence of a toll and in which an “enforced” trafficpattern then materialises must, however, also be considered.

It is only in such circumstances that the introduction of congestion tollsshould be envisaged.

The losses for the community may be much greater than in the precedingcase, and the congestion toll would be aimed this time at preventing an“enforced” traffic pattern (cf. Christian Gerondeau: Transport in Europe).

The higher demand is -- i.e. the further to the right the demand curve islocated -- the higher the theoretical congestion toll will have to be.

It should be pointed out that the flow corresponding to the “economic”optimum (defined above) will be practically equal to the road’s capacity. Theeconomist’s and transport engineer’s definitions then correspond to verysimilar values.

It has been concluded from studies on the London region that, if a toll is tobe effective, the level for a single traffic direction should be £4 for access toLondon (inner cordon) and £8 for access to the inner city (Central London).

73

But it must be added that the practical problems of collecting tolls are so greatin the largest urban centres that they seem insoluble in the short term,particularly because of the difficulties of prosecuting possibly as many as tensof thousands of offenders a day, and processing non-residents who are fromoutside the area and are not equipped with the appropriate systems.

From the theoretical viewpoint, this finding is obviously regrettable, for inthis case the number of people who might use their cars would be higher thanif there was no congestion toll (cf. Christian Gerondeau, op. cit.).

In order to regulate traffic in an urban environment and reduce it to itsoptimal economic level, it seems that other solutions will have to be used, suchas traffic and parking regulations, as well as the pricing system for parkingwhich, although not perfect, is a realistic way of adjusting demand to networkcapacity more effectively.

The impossibility at the present time of devising a congestion toll systemthat can be actually used in major urban centres should not, however, result inundue concern, since the losses attributable to the present traffic situationcompared with its theoretical optimum are usually by no means so high as aproportion of Gross Domestic Product as is claimed.

An initial remark should be made in this respect: the “time losses”compared with an ideal situation in terms of economic theory are very low onthe motorway and expressway network.

A study on the Paris Boulevard Périphérique (inner ring road motorway)has shown that, with a reference speed of 60 km/h, the value of time lossescame to FF 1.6 billion a year or Ecu 0.23 billion. But this ring road accountsfor two-fifths of congestion on all the Paris Region motorways, on which timelosses measured on the same basis therefore do not exceed FF 4 billion a year,or Ecu 0.6 billion.

Lastly, it has been shown that time losses on the other French motorwaysdid not exceed 20 per cent of those recorded in the Paris Region, which resultsin a national total of FF 4.8 billion, or Ecu 0.75 billion a year. This sumrepresents 0.07 per cent of France’s Gross Domestic Product.

Although the average population density is low in France, the ParisRegion has by far the highest density of motorway traffic in Europe, withaverage daily flows of 30 000 vehicles per lane, as against 20 000 to 25 000 atmost almost everywhere else.

74

According to estimates in the Netherlands, the value of time losses on themotorway network amounts to 0.2 per cent of the country’s GDP, although thefigure applies to the most densely populated region of Europe.

It is therefore legitimate to assume that the cost of the time losses onEuropean motorways must be around 0.1 per cent of the continent’s GDPwhich, all said and done, is very modest.

The conclusion must be spelt out: owing to the very high capacity ofmotorways, the time losses recorded on them and on similar roads are finally limited.

Where substantial losses attributable to poor operation of the roadnetwork exist, they are incurred on roads other than motorways. The lackof an adequate motorway network is therefore the cause of the problem.

In this respect, we have seen that the major European urban areas aredivided into two main categories.

Most of these areas are criss-crossed by an urban motorway networkwhich is so dense that most points within them are only a few kilometres froma motorway interchange.

This is the case, for example, of the Randstad in the Netherlands, theRuhr, the Paris and Madrid Regions, etc.

In such cases, congestion on the conventional urban network cannot resultin substantial losses as the distances on it to the motorway network are alwaysshort, especially if the urban network includes wide avenues. In this case theintroduction of congestion tolls is obviously unjustified. It is only in theopposite case, which is rather the exception than the rule, that the positionmight be different.

The situation of London in this respect is an extreme case, and it isunderstandable that the UK capital has been the subject of the most advancedstudies, the results of which have been recalled above.

To sum up, even if the losses attributable to congestion are not negligible,the European road system is operated much closer to the economic optimumthan is generally said.

There is in particular no foundation for the figure of 2 per cent of GDPwhich is frequently quoted as representing congestion costs.

75

The value of time losses in an optimal situation -- at any rate less than0.5 per cent of GDP -- can also be compared with the sums spent by Europeanson their road transport systems (vehicles, fuel, infrastructure, etc.) whichamount to about 15 per cent of GDP. The losses attributable to congestion aretherefore equivalent to an increase of about 3 per cent in the cost of the system.But this calculation does not reflect the full picture since it mainly takes intoaccount the monetary costs. If the value of the time spent by users on the roadduring their trips was taken into account, as it should be, it would appear thatthe Europeans spend considerably more than 15 per cent of GDP on their roadtransport system (meaning that it provides services of at least the same value),and the figure of 3 per cent would be proportionally reduced.

76

77

ANNEX

MAPS OF THE MAIN URBAN CENTRES IN EUROPE

(scale 1/400 000)

78

79

80

81

82

83

84

85

NETHERLANDS

Piet H.L. BOVYProfessor of Transportation Planning

Faculty of Civil Engineering & GeosciencesDelft University of Technology

Netherlands

Ilan SALOMONProfessor, Department of Geography

Hebrew University,Jerusalem

Israel

86

87

A PROSPECTIVE ASSESSMENT OF THE PROBLEM

SUMMARY

1. INTRODUCTION......................................................................................89

1.1. Scope of the report..............................................................................901.2. Are there “European” congestion patterns or problems? ...................91

2. THE NATURE AND EXTENT OF CONGESTION ................................92

2.1. Measures of congestion ......................................................................942.2. The causes of congestion..................................................................1002.3. Congestion patterns in Europe..........................................................108

3. BEHAVIOURAL RESPONSES TO CONGESTIONAND TO POLICIES ................................................................................118

3.1. Travellers’ response to changing congestion ...................................1193.2. Firms’ responses to congestion ........................................................126

4. ADDRESSING CONGESTION: POLICYMAKING ANDPOLICYTAKING ....................................................................................127

4.1. A desired level of congestion? An economic approach ..................1284.2. The gap between policymakers and policytakers.............................1334.3. Policy approaches ............................................................................135

88

5. CONCLUSIONS......................................................................................139

5.1. The notion and extent of congestion ................................................1395.2. The use of congestion measures as quality indicators......................1395.3. The spread of congestion levels .......................................................1395.4. The European dimension of congestion ...........................................1405.5. The true costs of congestion .............................................................1405.6. Variation in the distribution of congestion in Europe ......................1405.7. The responses to congestion.............................................................1415.8. The limits of congestion ...................................................................1415.9. Investments in congestion relief.......................................................141

6. RECOMMENDATIONS .........................................................................142

6.1. Statistics on congestion ....................................................................1426.2. Optimum level of congestion ...........................................................1426.3. Need for balanced spatial development............................................1426.4. Public transport is an ineffective congestion relief measure............1436.5. “Only the road can relieve the road” ................................................1436.6. Need for high-quality roads..............................................................143

NOTES.............................................................................................................144

ANNEX............................................................................................................145

REFERENCES.................................................................................................148

BIBLIOGRAPHY ............................................................................................149

Delft, December 1997

89

1. INTRODUCTION

Congestion has become an inseparable characteristic of manytransportation systems. Transportation systems are developed to support publicwelfare and facilitate economic growth by means of providing accessibility.More mobility is usually associated with greater welfare. However, theevolution of mobility, in both the qualitative and quantitative aspects, hasdeveloped to such levels that in many places and times it generates significantnegative impacts. These include externalities, such as congestion,environmental pollution of various kinds and safety costs. A congestedtransport system may fail to deliver sufficient economic benefits and may havenegative ramifications on the competitive position of a region in the Europeancommunity. It is also likely to increase the environmental costs in terms ofemissions.

Transportation policymaking is thus becoming an “art of balancing”between the desired improvements in mobility and the minimisation of thecosts to levels acceptable by society.

From the perspective of European policy analysis, the congestion issue isdirectly related to policy questions such as planning and investment in TERNs(Trans-European Road Networks), financial support to countries and regions indeveloping international and interregional road links and the question of fairand efficient pricing of transport in infrastructure and transport use (Kinnock,1995). These and other policy issues require a much deeper understanding ofthe congestion phenomenon and of the impacts of congestion

Congestion is experienced daily, not only by many road users but also byrail and airport travellers as well as by shippers of freight on these modes. Asroad congestion is probably the most common form of congestion experienceddaily by literally millions of travellers, the report focuses on this type ofcongestion. Another reason for this focus is that congestion in other modes oftransport may call for different solutions. Road congestion is sufficiently

90

complex that addressing it should not be jeopardised with an attempt to providea more generalised analysis. Road congestion is generally considered as a“public evil”, and much attention is paid by the public, planners andpolicymakers to mostly vain attempts to curb it.

The private and external costs of congestion are generally considered quitesubstantial. The spread of congestion across Europe, as well as many otherparts of the world, is worrying and consequently it has drawn a considerableeffort on the part of policymakers and researchers trying to identify policieswhich can mitigate its effects and reduce its costs. But there are some doubtsregarding the effectiveness of many such policies and even regarding therationale of congestion-related policies.

The antecedents of congestion have been widely studied (Cervero, 1991;Downs, 1992; Giuliano and Small, 1994). Road congestion is a result of amultitude of factors which culminate in the growing dependence of urbaniteson the private automobile and the temporary inability of the road network toaccommodate the consequent traffic flows. Major contributing factors are thetemporal and spatial structures of activities and the economics of car travel. Inmany cases the “transportation problem” is equated to the congestion problemand considerable policymaking efforts are directed at its reduction. Variousaspects of the underlying factors and the policy debate are addressed in thisreport.

1.1. Scope of the report

This report addresses the congestion issue from a western-Europeanperspective, by focusing on a number of key questions:

− What are the current patterns and trends of congestion?− Should congestion be eliminated altogether?− What can and should be done to mitigate the undesired level of

congestion?

The scope of the present study is focused on the trends in congestion andits underlying causes in Europe. The underlying causes of congestion are, inpart, directly related to transportation systems’ management and planning.Physical bottlenecks in the network or confusing network structure whichresults in inefficient weaving of traffic are two clear examples. But congestionis, to a great extent, determined by economic and social factors which lie

91

beyond the scope of transportation policy. The growth of the drivingpopulation due to the maturation of the “baby boom” generation, or due to thechanging role of women in society, illustrates this type of exogenous cause ofcongestion.

Chapter 2 focuses on the background factors. It first concentrates on themeasurement issues. Then, it describes the various external and internaldynamics which produce congestion. The last section of Chapter 2 describescongestion in Europe, relying on a variety of comparative sources of data.

In recent years, congestion seems to be expanding in its temporal andspatial distribution. This raises a number of questions which are to beaddressed in the following chapters:

� What would happen if it were left without any policy response?� What can be done to change the trends? and,� What are the relevant policy responses?

To reply to these questions, we discuss congestion from a behaviouralperspective in Chapter 3. The main point of this individual travellers’perspective is to demonstrate that the problem seems very different comparedwith the system-wide perspective. The implications of this gap are thendiscussed in Chapter 4 which focuses on coping with congestion. It opens witha discussion of whether or not it is desirable to have some level of congestion.It then presents a review of various approaches to address the congestionproblem, where and when it exceeds a desired level.

Chapter 5 presents the main conclusions, followed in Chapter 6 by somebrief recommendations.

1.2. Are there “European” congestion patterns or problems?

Congestion seems to be increasing in many parts of the world, from theNorth American megalopolis to the western-European conurbations and to therapidly growing metropolitan areas of Southeast Asia. The focus on Europeimplies that there are some unique attributes in European congestion which areabsent or different in other parts of the world.

While congestion may be viewed as a form of queue, wherein passengersare waiting to traverse a particular link or node where demand temporarilyexceeds supply, its antecedents and underlying causes differ across locations

92

and times. The focus on Europe is motivated by the fact that travel patternsand trends in Europe differ from those in North America, Japan or otherdeveloping metropolitan areas across the world. One of the explanations forsuch differences lies in the time factor. The timing of the introduction ofvarious technologies and social trends across different parts of the world mayexplain why present conditions and trends vary in different regions.

As will be shown, the private automobile, which lies at the basis of thecongestion problem, gained its popularity in Europe at a different time andagainst a different spatial, economic and social environment compared tocertain other parts of the world and consequently, congestion patterns evolve ina different way. The difference in the phasing of the growth in congestion alsoimplies, as will be shown below, that the range of relevant policy measuresappropriate for implementation in Europe probably differs from those relevantfor America or Asia. Two other important conditions for congestiondevelopment significantly differ between the continents, namely the spatialsettings and availability of travel options and transport alternatives to the car.

But one should notice that a European focus must be qualified as well.Within Europe, there are wide variations. In particular, it seems thatcongestion is not a continent-wide phenomenon and is not likely to becomeone. A continent is not the proper unit of analysis for such a study. Congestionis clearly a regional phenomenon, concentrated in the densely-populated areasof north-western Europe, as much as in highly urbanised areas in the rest of theworld. Consequently, it is also quite irrelevant to compare cross-nationalstatistics on congestion, as they conceal more than illustrate the differences.Regions seem to be the appropriate units, as suggested by Meyer (1990). Thus,in this report, we provide data from various regions in Europe, rather thannational travel patterns. But, for background information, it is worthexamining some of the basic differences between Europe and other developedeconomies (see Salomon et al., 1993 and Pucher and Lefevre, 1996).

2. THE NATURE AND EXTENT OF EUROPEAN CONGESTION

Ultimately, congestion is a temporary situation in which the demand forroad space exceeds the capacity, on a given section of the network. This is asimplistic view of congestion, as the nature of the gap between demand andsupply can be of very different character.

93

The colloquial “explanation” for congestion is that insufficient roadcapacity has been provided. This view is based on a widely accepted notionthat road space is a free public good that needs to be supplied by theauthorities, to accommodate any level of demand. However, alternative viewscan be suggested. First, as will be discussed in Chapter 3, the reasoning can beturned around and it can be suggested that congestion is a result of excess useof vehicles, rather than insufficient supply of road capacity.

But not all congestion is a result of insufficient supply. Some congestionis not recurrent and is a result of particular temporary conditions, such asaccidents, severe weather conditions and road maintenance work. Recurrentcongestion is caused by a structural lack of capacity (or, equivalently, excessdemand), whereas non-recurring congestion stems from incidental lack ofcapacity or excess demand. It is important to note that, even in cases ofrecurrent congestion, the travel characteristics can change from day to day: itis uncertain where queue building starts, when it starts, when it ends, how largewaiting time loss will be, etc. So, apart from the existence and travel timelosses of queues, the unreliability of queue location, queue duration and queuemoments is a major aspect of congestion having a great impact on travellers’behaviour.

Congestion is a double-faced phenomenon: on the one hand, it may beviewed as an attribute of the network; it can be described, for example, by thenumber and length of queues that have occurred in the network, or by theirduration. Similarly, the level of congestion may be described by the length ofthe network which was affected by queues. This network-related view is mostcommon in official statistics and in the public debate.

On the other hand, congestion is also an attribute of a trip. Thisperspective on congestion is of interest because it entails some importantattributes which influence the traveller's behaviour. Such trip-relatedcharacteristics are, e.g., whether a trip encounters congestion (percentage oftrips experiencing congestion along some segment), amount of time or distancetravelled under congested conditions and share of delay time to total trip time.Unfortunately, empirical data on trip-related congestion variables are very rare.

The extent of congestion can be demonstrated by some statistics of theDutch Randstad area (Ministry of Transport, 1997). On an average workingday in 1996, about fifty queues of a minimum length of 2 km build up mainlyat the fringes of the four big cities. As indicated before, congestion is highlyvariable, so the number, location and times of queues change from day to day.

94

Bridges and tunnels across major waterways are well-known queue locations.Also discontinuities in the freeway network (entries, exits, weaving sections,lane number alterations) are favourite queuing places. The typical location ofrecurrent queues around the bigger cities for a large part stems from changes inthe spatial orientation of travel demand in the last twenty years, such asreversed commuting and criss-cross travel demand between suburbs.

Concerning the timing of congestion, 80 per cent of the queues in theRandstad occur during the peaks, of which 45 per cent in the morning peak(from 7.00 to 9.00 a.m.) and 35 per cent in the afternoon peak (from 4.00 to6.00 p.m.). From the on-average 40 daily queues, about 10 queues take placeduring off-peak times. These are mainly queues caused by incidents.

2.1. Measures of congestion

2.1.1. Unit of analysis

Congestion is an important notion in transport decisionmaking. It is arelevant quantity in network design, facility dimensioning and pricingstrategies. It is therefore striking that policymakers are still struggling for aclearly defined, unambiguously measurable indicator of congestion. There areonly very few countries (e.g. the Netherlands) with reliable congestion statistics(see, e.g., Ministry of Transport, 1996 and 1997, and NEA, 1997).

A brief discussion of the problem of measuring the amount of congestionin transportation systems is presented below. It is intended to contribute to theongoing European efforts to produce a standardized approach (see Annex 3 inECMT, 1995, on Proposed method for harmonizing measurement of roadcongestion).

It is fundamental to distinguish two classes of congestion measures. Oneclass is related to flow conditions on the network and the other to parameters oftravel conditions between origins and destinations. In the first class ofmeasures, the unit of observation is a link in the network, and we may look atvolumes, speeds or link traversal times to derive values for the level of linkcongestion and, possibly, the related costs for that link (ECMT, 1995). Thismay be done using traffic assignment models (Transroute, 1992) or on the basisof observed link volumes (e.g. ECMT, 1995, Annex 3). Adding up the linkvalues results in a network-wide figure. This approach can also be applied tospecific network categories, such as non-urban motorways, urban radial

95

arteries, etc. Most of the data on traffic congestion belong to this class ofnetwork congestion measures (see, for example, Bukold, 1997 andECMT, 1995). These network-related congestion measures are indicators ofthe quality of the network performance.

The mirror image of the network performance attributes are measures ofcongestion as they apply to the individual’s travel behaviour. These measuresrefer to trip characteristics of travellers, from their respective origins to theirdestinations. In these type of measures, the trip or tour is the unit ofobservation. The quality of a transport system can thus be described by theeffects of congestion on travel conditions, on travel choices or on howcongestion pricing might affect travel behaviour. Such measures refer to thecongestion experienced by users and include the number of times a driverencounters a queue, the duration of waiting and the total excess trip time due tocongestion.

The rationale behind taking the trip as a unit of observation is thattrip-making behaviour, such as route, mode and departure time choices, isbased on the characteristics of the entire trip. In this respect it would be evenmore relevant to address a round trip (origin-destination-origin) as the relevantunit of analysis. Information on trip-related congestion and its derived costs isof utmost importance for policymaking, but is unfortunately rarely available.These measures, similarly to the classification of network measures, can beclassified by the type of network upon which various segments of the trip aremade. Congestion is likely to be experienced only on some of the segments ofa trip.

Presumably, correct accounting of congestion losses by both classes ofmeasures should lead to the same levels. However, the two classes of measuresdiffer in the information they convey. Some examples will highlight thedifferential sensitivity:

a) In certain bottlenecks of the road network, the total congestionduration may be long and last for hours. The bottleneck serves hightraffic volumes, so that the individual waiting times at the bottleneckare short, in the order of a few minutes. This is a negligible amountfrom the individual’s perspective and does not lead to adoption ofalternative travel choices.

96

b) The additional time spent in bottlenecks may be fully or partlycompensated by quick and easy progress because of high speeds atthe other segments of the trip, such that total trip time remains withinindividually acceptable values.

c) Network congestion measures only measure revealed congestion assuch. They do not measure the impact of congestion on users whoavoid the congestion by adapting their behaviour, such as drivers whotake a detour route or an earlier departure time with a longer trip timebut without congestion.

Thus, network congestion figures, on the one hand, may provide a picturewhich is too negative, as they are based on the accumulation of a large numberof small, behaviourally irrelevant queuing times. On the other hand, they failto account for the impacts of congestion on suppressing demand for travel.Therefore, trip-based congestion measures are to be preferred.

The distinction between network and trip-based congestion indicators canexplain the paradox identified by Gordon and Richardson (1991) which statesthat, while aggregate congestion figures appear to increase steadily from yearto year, average travel times, speeds and congestion losses per trip remain moreor less constant. Despite the increasing congestion on motorways (in aggregatefigures), such roads do not seem to lose their attractivity, as is evident fromgrowing volumes. The explanation is that congestion experienced byindividuals does not increase significantly.

In addition, it is hypothesised that drivers are willing to accept a certainwaiting time at bottlenecks. It seems that up to 10 to 15 minutes of queuing isacceptable, and only beyond this level do drivers engage in adapting theirbehaviour to alternative travel patterns. This hypothesis is derived from theobservation (in the Netherlands) that at a number of classic bottlenecks themaximum queue length is stable for many years (Westland, 1997).

2.1.2. Measurement of congestion: A critique

Having discussed the unit-of-analysis problem in measuring congestion,the question remains of the correct way of measuring congestion losses ineither case.

Congestion may be defined as a state of traffic flow on a transportationfacility characterised by high densities and low speeds, relative to some chosenreference state (with low densities and high speeds). It should be stressed that

97

high flows are not typical for congestion; in many instances, the congestionstate results in low flows and low speeds. Flow levels alone are thus not auseful indicator for congestion (see also NEA, 1996). What the reference state(zero congestion) should be depends, inter alia, on the purpose ofdecisionmaking (infrastructure decisions, traffic management decisions,congestion pricing decisions, etc.).

Both from a public policy making perspective and from an individualtravel decision making point of view, the congestion burden should betranslated into costs. For sake of clarity and comprehensive accounting, it isimportant to distinguish between congestion costs of the following four groupsof transportation system users who are affected by (rising) congestion:

a) Higher travel costs for road users who use bottlenecks and experiencecongestion;

b) Higher travel costs for road users who avoid congestion, e.g. bychanging route or departure time (suppressed bottleneck demand);

c) Higher travel costs incurred by other users of the transportationsystem, due to demand shifts caused by congestion, e.g. shifts of roadusers to public transport (suppressed road traffic demand);

d) Reduced benefits due to a change in activity and therefore derivedtravel pattern (suppressed travel demand).

Cases (a) to (c) imply a lower consumer surplus, given the same level ofactivity is maintained and the same benefits are accrued. In all four cases, bothprivate costs (borne by the congestion causing road user) and external costs(borne by others) may be involved.

We may now define congestion costs as the additional costs caused by theexistence of congestion, relative to some adequately chosen reference state.

In looking at the European congestion costs, published in officialEuropean documents (e.g. Green Paper, OECD, ECMT, DHV/Colquhoun,1991) a disturbing variety of cost figures emerges. This results, among others,from the fact that a variety of methods are applied, but also from the confusionof private, external and social costs of congestion. In Kinnock’s well-knownGreen Paper (Kinnock, 1995) the external congestion costs in Europe, namely,that part of the costs not borne by those who cause the traffic congestion, isstated to be about 2 per cent of GDP. This cost level was quoted from anOECD survey (Quinet, 1994) which, however, stated that social costs ofcongestion totalled to 2 per cent of GDP! In addition, the OECD survey

98

restricted its calculation of social costs to the additional travel costs oftravellers that experienced congestion. Thus, these costs include the privatecosts (which are matched by private benefits) and only part of the externalcosts. Generally, the calculated additional costs in the survey referred to thefree flow reference situation, which is a questionable approach, and in mostcases was calculated using static network assignment modelling which is aclearly deficient procedure to estimate congestion costs.

In recent years, there is a growing interest in estimating the full socialcosts of car use. This focuses on the environmental costs but also considerscongestion costs (e.g. Litman, 1997, Delucchi, 1997 and Kageson, 1993).However, these focus primarily on environmental costs and do not detail (atleast in the cited sources) how congestion costs were derived.

One may confidently state that the often quoted 2 per cent of the GDPcongestion cost figure lacks a clear empirical and methodological foundation,and is not more than a first rough guess unsuited for serious policy making.Calculations by Gerondeau (1997) show that a figure of 0.3 per cent might beequally plausible (this equals, for example, the congestion costs levelcalculated for Dutch motorway traffic based on an extensive congestionmonitoring system).

Almost all figures found on congestion costs in Europe are derived fromthe “additional time spent” travelling, relative to a chosen reference situation.Using appropriate value-of-time estimates for travellers and goods, and vehiclefuel consumption, this extra travel time then is transferred into a monetary costfigure.

A few comments can be made with respect to the choice of the adequatereference situation (see also Gerondeau, 1997).

a) Some congestion estimates have used as a reference a collection ofideal door-to-door trips, based on distance calculation, determined bya fixed detour factor relative to the airline distance (e.g. 1.2), and withcertain ideal travel speeds (e.g. 100 km/h outside and 50 km withinurban areas). In one of the European infrastructure studies(DHV/Colquhoun, 1991), zone-to-zone minimum speeds wereadopted (90 km/h for cars, 80 km/h for trucks) to determine the levelof inadequate performance of the network. When confronting theactual travel conditions, with such a reference, one is in factcalculating the costs of inadequate road network instead of the costsof congestion.

99

b) The most frequently applied approach is based on the existing roadnetwork as given and considers the “empty” network as referencepoint. This means that the actual traffic pattern is compared to tripson shortest routes at maximum speeds, even in peak periods. Clearly,such an empty network is an unrealistic yardstick; and moreimportantly, a network satisfying such conditions would look quitedifferent and would be very inefficient as well. Nevertheless, thestudies using this approach formed the basis for the OECD survey oncongestion costs (Kinnock, 1995).

c) A few studies have used the assumption that low speeds necessarilyimply that congestion exists, or that high volume/capacity ratios areunambiguous indicators of congestion (Transroute et al., 1992). Thelevel-of-service concept is often used to define the referenceconditions, e.g. Bukold, 1997. This is clearly a more reasonableapproach because one can choose the conditions in which the networkoptimally fulfils its transportation function. In the Netherlands,economic calculations have been performed to derive a social costoptimum for traffic flow (Stembord, 1991). It appeared that a 2 percent congestion probability (which means that on a yearly basis 2 percent of daily traffic of a road section will be experience a queue) isthe optimum level of congestion. This level serves as a reference forquantifying the costs of additional travel time.

2.1.3. Measures of congestion: a proposal for improvements

In summary, estimations of the economic costs of congestion exist inmany European countries and on the European level, but their outcomes are sodifferent and are based on such widely diverting assumptions and methods thattheir credibility is very poor. Policymaking in Europe with respect tocongestion needs to be based on valid and comparable facts, on measured andestimated congestion characteristics of the infrastructures and of the trips.On-going work on improving congestion measuring methods (see, e.g., WP5,1997) should be forcefully continued. The inventory of congestion figurescarried out for this report, clarified that a much more rigorous and systematicanalysis of congestion costs is needed in European countries, exhibiting, amongother things:

a) a clear distinction between private and external costs;b) a clear distinction between road users and non-users;c) a clear distinction between travel costs and other congestion costs;

100

d) a clear definition of the reference situation preferably based on astandardised, economically optimal network design;

e) a standardised and valid calculation procedure of the cost elements;f) congestion figures are needed for both network elements and for trips;g) flow and speed data should refer to hours.

Probably the best statistics on congestion nowadays available may befound in the Netherlands. In order to show the possibilities of informationprovision about congestion, the Dutch method of collecting and producingcongestion information is concisely explained below in the appendix (for moredetails, see Ministry of Transport, 1996 and 1997, and NEA, 1997).

2.2. The causes of congestion

Traffic congestion is the result of a multitude of factors. The importanceof each factor varies from one place to another and across time. Broadlydefined, the causes can be attributed to demand and to supply factors but theseare, of course, at some point, interrelated. In the following sections, twocomplementary explanations for the evolution of congestion are presented.First, we focus on the external forces that increase the car dependency of thepopulation in developed countries (growing car dependency in developingcountries is related to other factors and will not be discussed here). Second, theinternal dynamics of congestion will be described, to demonstrate the processesby which changes on a network occur in the presence of congestion.

Figure 1 presents a flow chart of the main effects that are at play on thedemand and supply sides. There are, as will be noted, many additional effectsand feedback mechanisms, but for the purpose of organising the description ofthe factors, only the main effects have been drawn.

101

Figure 1. The main external factors causing congestion(Feedback and minor effects were omitted for clarity.

Dashed lines represent negative effects)

Economic factors

Economic efficiency

Income

Energy costs

Socio-demographicPopulation size

Women’s roles

Household population

Driving population

Temporal structure

Work

Non-work

Transport supply

Network capacity

Competitive positionof public transport

Car availability

Spatial structure

Suburbanisation

Accessibility

Relocation

Car use

Congestion Improved speedPolicy

Transportsystem

performance

2.2.1. The sociodemographic factors

The driving population is growing and one of the consequences is agrowing demand for travel by car. The growth in the driving population is theresult of a number of background trends. First, there is a growth in thepopulation. While natural growth of the general population is small in Europe

102

since the maturation of the “baby boomer”, Western Europe is the target ofimmigration which contributes to its population growth. But, more importantlyfrom a travel demand perspective is the fact that the household population isgrowing faster than the general population, in particular in the more urbansegments of the European societies. This is, in part, attributed to the growth insmaller households (of single persons, single parents and smaller numbers ofchildren in the households). As households are independent units ofconsumption and production, more households imply more maintenance tripsand a greater demand for automobiles.

Another important factor contributing to the growth of the drivingpopulation is the changing roles of women in society. With the growingparticipation in the labour force, while in most cases still bearing theresponsibility for household chores, women often experience a greatertime-space pressure than men. This results in an increased demand forautomobile use and there is much evidence of a growth in licence availability inthe female population across Europe as elsewhere.

The ageing of the population, as a result of prolonged longevity, involvesyet another contribution to the size of the driving population. Older people oftoday, and increasingly so in the future, are more likely to own a drivers’licence than in the past. However, their contribution to congestion is limited.Assuming that most are retired, they are less likely to drive during the morningpeak, but are likely to contribute to the afternoon peak.

2.2.2. The economic factor

Growing income has brought about a general rise in the standards of livingand the automobile has become an integral part of these standards. Coupledwith the relatively low costs of automobiles and their operation (low energycosts), the availability of the car for a growing number of activities had becomethe norm. Generally, energy prices in Europe are significantly higher than inNorth America, but auto usage is still relatively cheap. Growing income has,of course, reduced some of the reliance on alternative modes of travel.

Growing income affects, inter alia, the changes in the residential location,as it facilitates the acquisition of private houses in suburban locations.

103

Another implication of the growing income, again coupled with relativelycheap cars and operating costs, is that automobile use has increased amongyoung people, who may form the first generation to have grown up in theprivate automobile. For the young generation, the use of the automobile seemsto be the norm, and alternative modes are decreasingly known and considered.

The growing car population is a major contributing factor to congestion.Cars are produced in response to demand which is growing steadily as peoplerealise the convenience of private transportation and the growing utility ofusing a car, given its costs and its advantages. The automobile industry inmany European countries is an important element in the national economy andis supported directly or indirectly by social causes. This in itself is acontributing factor to the growing popularity of the private automobile.

Costs of ownership and operation of private vehicles fall short ofreflecting the full social costs of using the automobile. In fact, the traditionalgap between personal costs and external costs of driving may have been one ofthe major mistakes of the twentieth century’s adoption of automotivetechnology. It can be speculated that if costs were internalised from the earlydays of the automobile age, many of the problems encountered today couldhave been avoided.

The costs of cars and their operation is taxed in all countries but thestructure of the various relevant taxes differs in the signals it generates withregard to auto usage patterns. In most cases, gasoline taxes (and parking taxes,as opposed to rates) are the only usage-based taxes. These seem to berelatively weak and have no bearing on congestion, as they do not reflectspatial and temporal variations. Only a few countries use road pricing as aninstrument to influence the use of the road infrastructure (e.g. France, Italy,Norway).

2.2.3. The spatial structure

The intricate relationship between urban structure and transportationtechnology has long been recognised. The role of the private automobile infacilitating suburbanisation of residences, and later employment, has been alsobeen acknowledged. But, in the present context, it is also important to note thatthe low density suburban setting has a very negative effect on the competitiveposition of public transport and also, to an extent, on non-motorised modes.Consequently, suburban accessibility is dependent upon the private automobile.

104

In an historical perspective, the relationship between suburbanisation andcongestion can be divided into two periods. Initially, with the suburbanisationof residences, congestion was primarily evident on radial links of the network.Later, with the growing suburbanisation of employment and commerce,congestion is becoming a problem of suburban regions, on both radial andcircumferential links of the network.

These changes also have implication on the temporal and spatialdistribution of flows and consequently on the likelihood of experiencing somelevels of congestion. When trips were primarily centre-oriented (and workschedules were quite fixed), flows followed a pattern of an inward movingwave, with congestion becoming more acute closer to the centre. The currentpattern of congestion are more complex, both in terms of the location andtiming.

Land use patterns in Europe are different from North America or Asia.European cities are more condensed, and many older town centres constrain thedevelopment of high-quality road infrastructure. Increasingly, European citiesare also experiencing suburbanisation, similar in quality to that observed inNorth America.

There are, however, some noticeable differences. European suburbs seemto exhibit greater densities than the American ones. Also, given that theincreased popularity of the private automobile in Europe lags behind theAmerican case, European conurbation’s have a more developed railinfrastructure which in many cases extends in the newly developed suburbs.Thus, while the spatial structure in Europe encourages car dependency, theintensity of the process is somewhat weaker than in America.

2.2.4. Activity-related factors

The demand for travel, except in some relatively rare situations, is derivedfrom the demand for activities performed at the trip ends. The structure ofactivities patterns is thus defined by time-space trajectories, which in turn aredetermined by the life styles individuals wish to exercise, the spatialdistribution of opportunities (land use pattern) and the temporal structure whichprevails in a given society.

105

The Temporal Structure lies at the heart of the congestion problem. Theprevailing temporal structure, which is very much a culture dependent factor,explains much of the activity patterns, especially with regard to the daily workschedules, and the daily and weekly shopping patterns.

As most workers begin their work day between 7 and 9 am, mostcommuting trips are made around this time. Assuming an eight hour workday,the home bound commute begins at 4 p.m., depending on variations with regardto the length of the workday and lunch break arrangements. In any case thehome-bound commute coincides with a peak in shopping related trips and thusthe daily peak period occurs between 4 and 6 p.m. This is clearly seen inFigure 2 below.

Figure 2. Typical temporal distribution of traffic congestion

Number of Queues ( 1993 )on Dutch Motorway Network

0

500

1000

1500

2000

2500

06.00 08.00 10.00 12.00 14.00 16.00 18.00

time

Source: Ministry of Transport, 1997.

The distribution of trips along time in each of the daily peaks depends onthe range of official work start times, the range of flexibility permitted in theworkplace and the level of congestion experienced by commuters.

106

Activity behaviour refers to work and non-work personal scheduling.Changes in work scheduling due to congestion include both daily workingtimes and working days, and chaining of daily activities. The same may beapplied for non-work, out-of-home activities where people change theirschedules. The main motive for work rescheduling will be decreasingreliability and travel time duration.

“Before” and “after” studies undertaken by Tacken and De Boer (1989and 1991) focused on the way employees use flexible working hours to avoidpeak hour traffic. The “after” study was done after an improvement in theurban beltway around Amsterdam, resulting in a decrease in congestion levelson the former bottlenecks.

2.2.5. A systems dynamics view of congestion causes

The road congestion problem as it developed in the last decades is atypical example of a self-reinforcement process with short- and long-termfeedback loops stimulating car use. Figure 3 describes, in a simplified andcondensed systems dynamics flow chart, the essentials of the mutual influences(on the level of individual households and firms) that endogenous factors in theeconomic and transportation system exert on each other.

For the sake of clarity, the exogenous factors (described in Figure 1above) that are simultaneously at work are omitted from Figure 3, despite theirclear influence on the levels of car use and traffic congestion.

Car availability and car use (in terms of distance travelled) are the basicengines of the process, fed by available incomes. Car use leads to higherdoor-to-door speeds which enable individuals and firms to cover a much largerrange for their activities with much higher utilities (including gaining a higherincome) achievable within the same travel time budget. An important part ofthese higher utilities are, for example, lower land and housing costs, leading inturn, to spatially dispersed settlement patterns. These increased traveldistances, combined with increased trip numbers due to demographic growthresulted in growing demand for road space and required extensions of the roadnetwork. Through the establishment of extensive motorway networks inEurope, medium and long distance door-to-door travel times were shorteneddramatically, thus speeding up the described spatial transitions. At the sametime, the improved roadway system contributed significantly to highereconomic performance and therefore higher income and lower car costs. Infact, car costs per travelled kilometre (for the same level of driving quality) is

107

continuously decreasing. This is another strong force for further increases incar ownership and use. A third feedback loop is the diminishingcompetitiveness of the alternatives to the car (walk, bike, public transport)mainly due to spatial dispersion and larger distances.

Figure 3. Systems dynamic model of factors contributing to congestion

Productionefficiency

Economicactivities

Household/firm income

Carownership

Caruse

Travel speeddoor-to-door

Traffic flowquality

Transportefficiency

Trafficmanagement

Road networkextension

Road trafficcongestion

Traffic flowstravel distances

Suburbanisationhousing/employment

Spatial relocationconcentration

production/services

Land access

Competitiveposition PTbike/walk

Carprices

+ +

+

+ + +

+

+

+ +

+

+ +

++ +

+ +

–– –

– –

–+

+

+

+

+

+ +

108

These circular processes have to be taken into account in the developmentof policies aimed at controlling road traffic congestion. A crucial factor isdoor-to-door speed. Further improvements in speeds lead to a furtherproliferation in the system. A critical element in any congestion policy(infrastructure extension, congestion pricing, etc.) is therefore to contain travelspeeds within economically tolerable limits.

Congestion policy measures should, therefore, be directed predominantlyat offering the required capacity at an economically sound level of service, butwithout further increasing travel speeds.

2.3. Congestion patterns in Europe

Congestion problems appear in particular locations and times, and hencethey are much more an urban or regional problem than a national or continentalone. Hence, the title of “congestion in Europe” must be qualified to addresspatterns that appear in some regions of the continent. In this section, somestatistics from different parts of Europe are provided and discussed.

A few studies have tried to estimate congestion characteristics ofEuropean roads to obtain an overall picture of the state of the network, in termsof spread, scope and costs of road traffic congestion in Europe as a whole, or atthe country level.

One of these studies (Transroute et al., 1992) shows that problems ofcongestion are being experienced on more than 5 000 kilometres (including3 800 kms of motorway) of the 54 000 kms of roads of internationalimportance within the European Community (half of which consists ofmotorways). This means that nearly 10 per cent of this high-level network isaffected. The capacity standards adopted to estimate the risk of congestion arethose used in Germany, namely, 50 000 vehicle equivalent units per day for a2x2 lane motorway, 80 000 for 2x3 lanes, and 110 000 for 2x4 lanes. As willbe demonstrated below, this figure of 10 per cent level is an average value withlarge variations between countries and even more between regions, and withstrong spatial concentration.

In an ECMT survey, Member countries (1991-92) were asked to reportabout traffic congestion on their main roads (ECMT, 1993; see also ECMT,1995). Figure 4 depicts the survey results in a scatterplot of congestion points.Unfortunately, the results from the different countries are not comparable at all.

109

Figure 4. Map of major congested links in European road network

Source : ECMT, 1995.

110

The quality and quantity of response were strongly correlated with the level oftraffic density in each country. In addition, there was a large variation inthresholds used to define congestion. Despite this, there is a clear spatialpattern in the congested spots, such as around high-density conurbations(London, Paris, Randstad, Ruhr area, Athens, etc.). The Scandinavian countriesclearly appear to suffer least from road congestion.

From this and other studies (Bukold, 1997), it can be clearly seen that thecontribution of international traffic to congestion is very limited. There arevirtually no cross-border links that suffer from congestion, which is notsurprising, given the relatively small international flows in absolute andrelative terms.

None of the studies reviewed is able to show the specific internationaldimension of congestion, such as the contribution of international traffic. Thesame holds for the specific contribution of road freight transport, with theexception of the DHV/Colquhoun study, which showed that, from the totalyearly cost of 350 million ECU due to inadequate level-of-service to roadtraffic, only 50 million can be attributed to freight traffic (1990).

2.3.1. Perceived quality of road infrastructure

An international comparison of European road traffic congestion suffers fromlack of readily available and recent data. Comparable data on congestion, based ona sound measurement methodology do not exist (with one exception discussedbelow). Nevertheless, some indicators of congestion in different countries can giverise to some hypotheses on differences in causes of congestion.

Road traffic congestion is related to the level of road infrastructure supplyrelative to the demand for trips. Table 1 provides some statistics on networksupply and congestion for some developed countries.

The data in Table 1 demonstrate that, even in countries with a similar levelof economic development, the supply of road infrastructure per capita differsby more than 100 per cent. The level of congestion, as measured by arelatively objective indicator (column 3), clearly appears to be related to thelevel of supply (columns 1 and 2). The relationship is corroborated by aperceived quality indicator reported by an international business panel(column 4).

111

Table 1: Road and traffic congestion parameters in different countries

Road network(km/1 000 inh.)

1993(1)

Motorways(km/million inh)

1993(2)

Congestion(% of links)

1993*(3)

Perceivedroad quality

1995**(4)

USA*** 14.5 331 -- 9.0Japan 6.2 37 (1987) -- 6.2United Kingdom 6.2 56 24.1 5.9Germany 7.6 136 7.9 8.3France 15.8 129 4.5 8.5the Netherlands 6.1 141 14.8 5.9Belgium 12.9 169 5.9 8.3Denmark 13.7 127 0.0 9.1

* Percent of motorway links with more than one hour of congestion per day(Bukold, 1997).

** Scale of 1 (low) to 10 (high) based on assessment of international business panel(IMD, 1996).

*** Data for the US includes the interstate system plus the urban freeways, for 1994.

The role of motorways in this context is noteworthy. Whereas motorwaysin most countries constitute only about 1 per cent of the total road network,they attract about 25 per cent of all travelled car kilometres (Brühning, 1997and Coughlin, 1994). In the Netherlands, the corresponding figures are 2 percent and 40 per cent. This is mainly due to the capacity of motorways whichcan reach more than ten times the capacity of an ordinary two-lane road. Thelatter is often considered to be saturated at about 10 000 vehicles per day.

Despite the significantly larger capacity, congestion problems arepresently predominant on (urban and suburban) motorways.

A comparison on a national scale certainly does not realistically reflect thetypical characteristics of traffic congestion. Some more location-specificcomparative analyses will therefore be added.

2.3.2. European congestion: some comparisons

While, from an American perspective, European travel patterns seem to behighly dependent upon public transport (Pucher and Lefevre, 1996), anexamination of European statistics highlights the crucial role that roadnetworks fulfil in Europe and the consequent congestion problems encountered

112

on some of these roads. Roads are the most versatile elements of transportationinfrastructure, as they serve both passengers and freight, both private andpublic transport, and they are available to individual operators (of cars) asopposed to restricted use of rail. Thus, some 85 per cent of all passengerkilometres in western European countries are made in cars and vans, and highshares of freight movement is on roads (Ministry of Transport, 1996). Roadsare the most fundamental element of the transport infrastructure and,consequently, a deterioration of its performance due to congestion is drawingmuch attention in Europe, as elsewhere.

Perhaps the first truly comparative study on the distribution of roadcongestion in Europe is the ECIS investigation on Bottlenecks in EuropeanInfrastructure (Bukold, 1997). The study provides a comparable description ofthe current conditions (1993-94) in the European major road network,consisting of some 13 000 links and including all motorways. To this end, allnational networks have been examined using a standard set of performanceindicators, namely the commonly used Level-of-Service (LOS) measure, asdefined by the US standards. A congested bottleneck is characterised by theLOS categories E (low speeds, unstable flow) or F (stop-and-go congestion).

Not surprisingly, the links with the highest traffic flows, exceeding50 000 vehicles per day, are found in the highly populated conurbations. Theseinclude London, the UK’s north-south corridors, the Rhine corridor (Randstad,Ruhr, Rhine-Main), Paris and the Rhone Valley, northern Italy, and also theMadrid and Barcelona regions. According to the conventional LOS criterion,most bottlenecks are found in the UK, Spain and the Rhine corridor, and to alesser extent also in Austria, Poland and the Czech Republic. By contrast,France’s road capacity appears to be sufficient, with the exception of a fewurban areas.

A more detailed insight into bottleneck situations is provided by observingthe duration of congestion on links in the road network. The number ofcongested hours provides a more accurate picture of the actual magnitude ofthe congestion problem in Europe. It appears that severe bottlenecks (linkswith more than three congested hours per day, on average) are quite limited innumber and that they mainly occur very close to major cities (see Table 2).

113

Table 2. Percentage of main network links exhibiting a congestionduration of a certain number of hours

Duration of daily bottlenecksCountry 0 hour 1 hour 2 hours 3 hours >3 hoursAustriaBelgiumDenmarkFinlandFranceGermanyGreeceIrelandItalyLuxembourgNetherlandsPortugalSpainSwedenSwitzerlandUnited Kingdom

95.594.1100100

95.592.198.886.290.6100

85.294.981.1100

93.675.9

02.3

000

0.60

3.500

3.80

0.900

3.7

00.9

00

0.50.81.3

00.8

02.8

01.8

00

6.5

3.00.9

00

0.51.2

03.52.4

03.1

00.9

00

2.8

1.51.8

00

3.65.3

06.96.3

05.25.1

15.30

6.411.1

Source: Bukold, 1997.

According to the ECIS study, European countries show strikingdifferences regarding the number and proportion of congested links.Exceptionally high proportions of congested links are found in Spain and theUK. Also, the Netherlands and Italy have comparatively high shares of linkswith severe bottlenecks in their road networks. By contrast, congestedbottlenecks hardly exist in Scandinavia. With a few exceptions, the Europeanroad bottleneck problem is mainly an urban problem rather than a problem forlong-distance connections or cross-border links.

The unique ECIS study allows the drawing of some conclusions:

a) Congestion in Europe is mainly within and close to urban areas.Improvements should be focused on urban infrastructure (including,for example, urban light rail, regional heavy rail, ring roads, tunnels,local by-passes) for single major cities and for conurbations such asthe Randstad and the Ruhr area.

114

b) The national situations differ widely. Though almost all countrieshave bottlenecks (except Scandinavia), only a few suffer from heavycongestion. The background factors which may account for thesedifferences are: 1) rapidly growing transport demand as a result ofeconomic development or population growth (e.g. Spain, Poland);2) persistent underinvestment (e.g. UK); and 3) environmentalconstraints or problems of physical limitations (e.g. the Netherlands,Germany).

c) Most bottlenecks and heavily congested roads coincide with areas ofhigh population density. Because of apparently severe spatial andenvironmental restrictions, new roads can be a solution only to alimited extent. Packages of road pricing, investments to divertthrough-traffic, and public transport (bus and rail) improvements arekey instruments in such cases.

2.3.3. A comparison of three conurbations’ road networks

In order to gain insight into the underlying factors of motorway congestionin the Randstad, the Dutch Ministry of Transport commissioned aninternational comparative study of three similar regions (Hilbers et al., 1996,1997). In that study the patterns of use of the main road networks in theRandstad area, the Ruhr area and the Antwerp-Brussels-Gent region werecompared and analysed using background factors such as network supply,spatial conditions, mobility patterns and socio-economic variables.

The regions belong to the high-density conurbations of Europe and aresimilar in size and structure. Interestingly, however, the level of motorwaycongestion (expressed in percentage of the network with more than three hoursof congestion per day on average, see Table 3) strongly differs. According tothe ECIS study (Bukold, 1997), the Randstad network by far shows the highestlevel, with the Ruhr area only half of the Randstad level, whereas the Flemishtriangle has a negligible number of links with this level of congestion hours.

In line with the ECIS congestion findings, motorway use in the Randstadis considerably higher than in the two other regions: a 25 per cent highervehicle density on a per-lane basis. A first important explanatory factor is thesupply of road infrastructure in relation to the number of inhabitants and thesize of the regions. Whereas the motorway supply in the Randstad is more orless similar to the other regions, the big difference is in the supply of theunderlying network which is much less cohesive and dense.

115

Table 3. Key roadway and mobility parameters for the Randstad,the Ruhr area and Flanders

Randstad Ruhr Area Flanders AreaShare (%) of network links with:

<1 hour of congestion> 3 hours of congestion

855.2

902.5

951.8

Daily volume per lane (veh/day)all main roadsmotorways

10 00016 800

8 10013 200

8 00013 600

Daily car/km per inhabitanton motorwayson other main roads

8.11.9

6.92.6

7.74.3

Network density (km/1 000 km²)Motorwaysother main roads

115105

120185

80180

Road capacity per capita(lane-km/million inh)

motorwaysother main roads

480320

523460

571690

Personal mobility (km/capita)all modes, all purposes 32.5 22.5 34.4

In addition, the Randstad motorway network is characterised by a higherlevel of accessibility in terms of entry/exit points; also its ring roads (aroundAmsterdam, Rotterdam, Utrecht) are much closer to the built-up areas than inthe other regions. Consequently, in the Randstad, more inhabitants have easyaccess to the motorway whereas at the same time more inhabitants have toshare the same scarce roadway space. These factors explain the relatively highpressure on the Randstad motorways which are characterised by a higher shareof short trips.

A second explanatory factor relates to the level of mobility of theinhabitants in the respective regions. Whereas the Randstad and the Flanderstriangle show very similar levels of daily kilometres travelled per capita, theresidents of he Ruhr area travel much less (22.5 km). This is partly due to thespatial distribution of activities, such as a greater concentration of employmentin the city centres in the Ruhr area and stronger local orientation of activitiesfor the Ruhr residents. In the Ruhr area, the housing demand is largelysatisfied in the local market, thus facilitating a much more spatially limitedactivity pattern. Conversely, the spatial distribution in the Randstad area ismuch more dispersed, with wider separation between residential areas andemployment.

116

It is interesting to examine the planned infrastructure investments in thethree regions (Hendriks et al., 1997). In all three regions, most investments arein rail infrastructure, for public transport. In the Randstad and Ruhr areas, thisshare is about 55 per cent, whereas in Flanders it is nearly 80 per cent. It isimportant to note that a large part of these rail investments are an improvementof interregional accessibility, especially as part of the European High SpeedRail Network. With respect to road infrastructure, it is notable that almost nonew links will be built in the coming decades. Instead, most of the investmentsare spent in extending existing links to 2x3 or 2x4 lanes. One may safely suggestthat the investments in rail, oriented to the service of long-distance travel, will notcontribute to congestion relief, whereas the road investments will.

It can be concluded that the high levels of congestion around majorurbanised areas are the result of two simultaneous factors, both associated withhigh population density:

− Less space available for road infrastructure; and− Large demand densities (more users per unit of road space).

2.3.4. Individually experienced congestion

Using the scarce readily available data on European congestion, a fewobservations can be made. In most European countries, the current share ofmotorway links with more than one hour of daily congestion is about 10 percent or less (Table 2).

Should this give reason for alarm? In official national and EC documentsand in the media, congestion losses are represented as hundreds of dailyrecurrent queues, millions of hours per day lost in queues and billions of ECUper year wasted in road congestion. There is no doubt that congestion is one ofthe most commonly cited problems of transportation systems. However, asCoughlin (1994) points out, problem definition is a political process. Interestedparties (industry, environmental groups, etc.) may have different definitions butmay all be interested in presenting doomsday futures. It is often suggested thatEuropean networks are close to collapsing. Is the situation really that bad?

It may be useful to reframe the question. For example, instead of providinghuge (and impressive) numbers of hours lost, one may pose the followingquestions: how well does the network fulfil its transportation function? Howmany travellers experience congestion regularly and for how long?

117

Before trying to answer this, it is useful to view the network-wide totalfigures in perspective. In the Netherlands, which, together with the UK, is thewestern European country most affected by congestion, total excess time due tomotorway congestion is about 2 per cent of total time spent travelling by car. Therelated congestion costs turn out to be about 0.25 per cent of GDP. Admittedly,this burden has to be carried by a relatively small part of the travelling populationbecause of the strong concentration of congestion in time and space.

Studies carried out on behalf of the French Road Federation corroboratedfindings reported in many other national sources, that those Europeans whotravel daily by car to work (that is 80 per cent of those who use motorisedmodes) need, on average, about 20 minutes to get to work (Gerondeau, 1997).No more than 10 per cent of them take more than 30 minutes in getting towork. Considering that most of these trips take place during peak hours, thishardly suggests a high level of overall road network congestion. Only 10 percent of these car commuters report that they usually encounter traffic jams ontheir way to work, thus, 90 per cent do not encounter congestion, even duringrush hours. Table 4 shows the percentage of daily car commuters in a numberof countries who declare that they usually encounter many traffic jams on theirway to work.

Table 4. Share of car commuters who experience congestion on their worktrip and average commuting time of car commuters

Country Per cent of commuters whoencounter much congestion

Average home-to-worktime of car commuters

(minutes)FranceGermanythe NetherlandsItalyUnited Kingdom

74

111219

1825231717

European average 10 19

Source: Gerondeau (1997).

These figures again show the favourable congestion situation in Franceand highlight the relatively poor conditions in the UK. If we consider thatcongestion outside rush hours is relatively rare, the figures do not support thesuggestion that congestion is very severe.

118

We may conclude that congestion is relatively rare in consideration of theoverall magnitude of the motorway network and the total amount of travelactivity on the road system, even in the urbanised regions of Europe.

3. BEHAVIOURAL RESPONSES TO CONGESTIONAND TO POLICIES

Congestion is a dynamic phenomenon. Its intensity is changing almostmomentarily, and certainly over longer periods of time. Generally, congestionis spreading in time and space, especially at the outskirts of large urban areas.However, these trends do not translate directly to the level of congestionexperienced by individual users.

The behaviour of three major classes of actors must be understood in orderto fully understand responses of actors to changing congestion and in order todesign effective ways of coping with it. We distinguish in this chapter betweenthe behaviour of individual travellers who make their decisions concerningtravel and its timing, modal choice, destination and so forth and firms, whohave a different set of choice variables. Each actor is concerned with differentattributes of congestion and has a different set of optional responses.Furthermore, in terms of sheer numbers, individual travellers are of greaterimportance, as firms are fewer in number of vehicles and in the number ofdecisionmakers who determine the travel patterns.

The third major actor is government, at all levels, which through itsactions (or inaction) influences the attributes of congestion, such as its size andtemporal and spatial distribution. Presumably, governments devise policies asa result of studies and evaluations. Unfortunately, policy measures are oftenadopted as responses to short-term political pressures, without the necessarygroundwork. Governments’ responses to congestion will be addressed inChapter 4, as part of the discussion of policy.

How do transport system users (travellers and shippers) and how dosuppliers of transport respond to a changing level of congestion? Addressingthis question is crucial for the development of policy measures designed to dealwith congestion. In this chapter we will discuss the response to changingcongestion of two major players: the users whose mobility is impaired bycongestion and the firms or shippers who incur various costs. The responses of

119

the third major player, namely, the authorities, which are concerned with thequality of service as it affects the social benefits and costs of transport systems,will be discussed in Chapter 4, which addresses the policy issues.

3.1. Travellers’ response to changing congestion1

Tripmaking involves a positive utility which is derived, in most cases,from the action performed at the trip end (and in relatively rare cases from theact of travelling per se) and a negative utility (or disutility) associated with theeffort of traversing distance. When congestion levels grow, the disutility of thetrip changes and the individual is likely to reconsider his or her utility of thetrip. Such a reconsideration may result in one or more responses over time,depending on a host of factors which are discussed below.

This section focuses on a number of behavioural attributes of theindividual’s response to congestion. First, it is necessary to discuss at aconceptual level, the structure of responses to changing travel conditions. Thedynamics of the process are described and, finally, the range of possibleresponses is identified. This range, which is the choice set from whichindividuals choose their preferred response, is wide and each alternative courseof action may fulfil various different functions.

As, in most cases, the issue is that of increasing congestion, the followingdiscussion assumes that, over time, travellers experience longer and lessreliable travel times. We will also address below the situation in whichcongestion is decreasing.

When facing increasing congestion, individuals experience growingdissatisfaction, which may reach some threshold level that triggers adeliberation or assessment of the situation. Salomon and Mokhtarian (1997)have suggested that the following process takes place: a search is initiatedwhen a certain level of dissatisfaction has been reached. Given the experienceone has gained, namely, prior adjustments to congestion, the individualidentifies the potential options for adjustment, evaluates them and chooses acourse of action which is likely to reduce dissatisfaction, at least temporarily.

Once a choice has been made and some action is taken, dissatisfactionmay be reduced, at least for a while, but in the context of increasing congestion,a threshold point of dissatisfaction may be reached again, triggering anothersearch for solutions. This time, previously adopted solutions may not be

120

feasible or desirable. However, it is also possible to choose an alternativerepeatedly, such as adjusting work-trip departure times or changing routes.Consider the case in which some low-cost strategies were selected and,subsequently, a high-cost strategy, such as a residential relocation, wasselected. In the following rounds, the low-cost strategies may again beconsidered.

The innovation of this model lies in three elements. First, it addresses theissue of search initiation, through an identification of the dynamics of theprocess. Second, also through the dynamic perspective, it focuses on theindividual’s limited choice set and, third, it articulates the implications of thelateral impacts as factors which affect the behavioural response.

Travel and commuting are not independent of other facets of theindividual’s life. They are just one element in a broad activity programmewhich is motivated by various drives and bound by various constraints. It isthis broader context of behaviour which should be viewed in the analysis ofbehavioural responses to congestion.

3.1.1. The dynamics of responses to congestion

The dynamics of the process deserve significant attention to improve thelikelihood of successful policy intervention. The timing of an adjustmentdecision, or deliberation about a decision, depends, among other things, on thehistory of such adjustments.

With growing dissatisfaction, a search trigger to ameliorate the costs ofincreasing congestion is assumed to be activated. The length of time to reachthe threshold depends on the time elapsed since the previous behaviouraladjustment, the nature of that previous change and the rate at which congestionincreases.

Understanding the issue of the time required for deliberation about changeis important for policymaking considerations. This is likely to be a function ofthe transaction costs. Residential relocation is not a decision made on the spurof the moment, while route change may be. Thus, when a situation changes, orwhen a policy is introduced, there is a span of time in which each potentialresponse may be employed. This is a very important point from a policyevaluation perspective. If a policy measure is evaluated before the range oflikely responses have been adopted, premature decisions may result. The caseof the Santa Monica (California) Diamond lane in 1976 may serve as an

121

example (Billheimer, 1978). There, under political and media pressure, aHigh-Occupancy Vehicle lane was discarded soon after it was inaugurated, notallowing sufficient time for travellers to make the necessary adjustments. (Thisdoes not mean that the project would have been a success if left intact, but thehaste of its removal did not allow for sufficient adjustments to be made.)Similarly, the Dutch experiment with a reversible HOV lane at theA1 motorway in 1994 which has been abolished before behaviouraladjustments were made.

Yet another attribute of the dynamics of the process is the fact that whilesome responses are reversible (e.g. change in departure time) others are notperceived as such. This difference implies that the amounts of informationacquisition and deliberation are greater for non-reversible responses andconsequently the response time is expected to be longer.

Thus, the complexity of the dynamics of the response mechanismunderscore the importance of identifying where the decisionmaker is “located”at a given time, so as to be able to assess her/his choice set and the attributes ofthose options, as perceived by the individual.

3.1.2. The choice set

The “universal set” includes a wide range of reasonable responses tochanging congestion. Each individual may not face this full set, but a subset ofthese. The individual choice set is determined by constraints as not allresponses will be available to a particular individual. The set of possibleadjustments can be classified along several dimensions.

Stern, Bovy and Tacken (1995) have proposed a hierarchical structure,based on increasing frequency of choice, which distinguishes betweenresponses made in different time horizons and decreasing scope. Such aframework facilitates the analysis of decisions that are made within differenttime and space contexts.

Bearing in mind the broad context of the response behaviour, which mayreflect more than just the direct reaction to increasing (or decreasing)congestion, Salomon and Mokhtarian (1997) have identified a wide range ofpotential responses, including some which are passive but, nevertheless, ofrelevance.

122

In reviewing the list of potential responses, it is useful to note that manyare not exclusively responses to congestion and may in fact be actions taken inresponse to other stimuli as well. While transportation policies usually aim tochange travel attributes, the above list shows that some responses are veryremote from travel attributes and affect other realms of life. Moreover, theresponses, as will be emphasized below, have lateral impacts on otherhousehold members, not only the commuter.

Goodwin et al. (1992) have suggested a very useful classification ofalternative adaptation options. They place coping strategies into a hierarchy,based primarily on the effort involved in the change, using this four-levelclassification:

1. Actions to increase the utility of existing behaviour;2. Actions which change travel behaviour while maintaining the same

activity set;3. Actions which modify the basic activity pattern; and4. Actions to modify the constraint and widen the choice of activities

and travel opportunities.

It is also possible to classify the responses on the basis of the objectivethey fulfil for the individual, as will be described below. Based on Mokhtarianand Salomon (1997), the following range of responses should be considered:

1. Accept travel costs: This “do-nothing” situation seems to be aprevailing response. It may indicate that despite the public andpolitical grievance about congestion, it may not be as severe aproblem as commonly believed. In economic terms, it implies thatthe costs of adopting any other response strategy are greater than thecosts of congestion to the individual.

2. Reduce travel costs: The automobile and car gadget manufacturersseem to cater to the frustrated driver by offering an increasinglypleasant and functional “commuting environment”: air-conditioning,a quality music system, a cellular telephone and other elements ofcomfort make the time spent travelling by auto more acceptable.

123

3. Adapt departure time: This strategy can reduce travel time, if thepeak period is relatively narrow. It will be less effective in thoseareas where congestion prevails for many hours continuously.Constraints, such as rigid work schedules or driving family memberslimit the ability to adopt this response.

4. Change route: By changing to a route with less stop-and-go traffic,the traveller may reduce commuting stress even though the new routemay be longer or slower.

5. Buy time: By paying a congestion toll one can buy travel time, whilepaying parking fees may reduce access time. A popular strategy tocompensate for time lost in travelling is buying the time of others, suchas baby-sitters, household help or support services at the work place.Investing in technologies for the home which increase productivity isanother way of buying time. So, extra travel time is compensated bytime gains elsewhere (activities at home) at a certain cost.

6. Temporal changes: (Flexitime, compressed work week and changesfrom full- to part-time jobs). Temporal changes allow diversion oftrips from peak periods to other periods, either by the adoption offlexitime or by adopting four ten-hour workdays.

7. Change mode: Switching to other, more efficient modes of travel isoften the solution suggested by transportation professionals,environmentalists and politicians. However, based on experience, thesuccess of that particular approach is limited to situations wherecongestion is very severe and shared-ride modes are competitive intime and cost to the automobile (e.g. in CBD-bound trips whereparking is limited and costly).

8. Telework (telecommute) from home or from a local work centre:Alternative work arrangements which allow flexibility not only intime but also in space, facilitate responses which allow the individualto avoid congestion.

9. Relocation of workplace or home: Avoiding congestion by locationaladjustments is an option for long-term response. It can either reducedistance or facilitate travel on routes which do not suffer fromcongestion.

124

10. Start a home-based business: This strategy entails costs for theindividual along with potential benefits like monetary gain, time,lower stress (greater control of one’s work), and convenience(schedule flexibility).

11. Quit work: This response carries a monetary cost even greater thanthat of strategy 10. If the motivation to quit work is predominantlythe stress of congestion, the result is likely to be deep frustration.Quitting work, which was mentioned earlier as a radical act, may infact be quite common. We suggest that many people who do notwork are those for whom the given (mostly time) costs of congestionhave exceeded the costs of other responses and the benefits of work.This may be more common among women compared to men.

The list is ranked on the likely frequency of responses but, in addition, italso identifies three types of relevant strategies: responses which maintain thecurrent level of travelling, by making travel cheaper or more convenient,responses which reduce travel, and life-style/locational changes. From a policyperspective, the latter two groups are of interest although the locational changesmay result in undesired effects on congestion.

As travellers are assumed to be utility maximisers and not costminimisers, they tend to explore the possible options for adjustment on thebasis of “what is good for them”, which may not coincide with a societalperspective. Thus, when conditions change, individuals are likely to exhibit“evasive” behaviour, namely, that they will try to identify and adopt thoseoptions which are least onerous. By contrast, when constraints are relaxed,such as in the case of highway expansion, individuals may exhibit “expansive”behaviour, thus improving their relative position. The “return to the peak”phenomenon observed in some cases (see below), is an example of expansivebehaviour. However, when road pricing is introduced, it may have a variety oflife-style and locational changes, rather than the often-expected modal shift topublic transport.

3.1.3. Some empirical evidence of adaptations to changing congestion

Few careful analyses of responses to changes in congestion have beenperformed. In most cases, research efforts focus on evaluating theeffectiveness of a particular policy measure and do not monitor the wide rangeof options which individuals may consider and adopt.

125

Tacken and De Boer (1991) performed a “before and after” evaluation ofchanges in the timing of trips, as a new roadway facility was opened. Theirstudy measured the 1989 and 1990 departure times by residents affected by theopening of the urban beltway around Amsterdam, with the completion of theZeeburger tunnel crossing the North Sea Canal. Congestion at existingbottlenecks was reduced or disappeared as the new tunnel was opened. Variousbehavioural changes were observed and reported by Kroes et al. (1996).Relevant in the context of this report is the way people changed their planned,“normal” departure times. The results (see Table 5) show the reaction ofemployees, who changed their activity patterns due to former congestion levels,in the new situation of 1990 where congestion levels have decreased. A clear"back to normal" reaction is observed with workers returning to moreconvenient departure times, usually “back to the peak”.

Table 5. Changes in work starting times after opening of tunnel

Start work in 1990Start work

1989Before: 7:00 7:00-7:30 7:30-8:00 8:00-8:30 8:30-9:00

After9:00

Total ≠Commuters

Before 7:007:00-7:307:30-8:008:00-8:308:30-9:00After 9:00

22215001

11110462230

543

1495783

02058

1794614

087

4012710

0458

1151

3820627030619579

Total ≠commuters

49 192 265 317 192 79 1 094

Source: Tacken and De Boer, 1991.

Based on data provided by Tacken and DeBoer (1991), Table 6 providesthe ranking that workers have given to the acceptability of various changes tocongestion (ranging from 1 = most acceptable to 5 = least acceptable).

126

Table 6. Average rank of alternative behavioural reactionsto changing congestion

Alternative Averaged rankChange working hours

Other routeMode choiceOther workMove home

1.92.62.93.23.8

Source : Tacken and De Boer, 1991.

3.2. Firms’ responses to congestion

Firms’ sensitivity to congestion results from three types of costs: labour,clients and freight-related. As all firms are dependent on labour, congestionexperienced by commuting employees is incurred, in part, as a cost to theemployer. Employees may demand higher wages to compensate for highertravel costs. But firms are also concerned with the time lost to labour travellingas part of their jobs. While such travel may not be in peak periods, it still mayinvolve significant time intervals in which the labour is not productive.

The second, and possibly more important, costs are the direct costsresulting from congestion which affect delivery of goods. Slow and congestedtraffic may imply, for a shipper, an increase in the size of the delivery fleet.This may be exacerbated with the transition to Just-In-Time production, wherethe reliability of delivery time is crucial.

Some evidence for these costs was found in a recent study conductedamong executives and employees of 15 firms in the Netherlands (Korver et al.,1994). Executives were asked about the responses of the firm to increasedinterurban road traffic congestion. The employees were asked about theirlikely responses to a further increase of congestion on their home-to-work trips.From a catalogue of potential responses, the most likely reactions of the firmsare changing working schedules and stimulating shorter home/work distances,whereas the least likely responses are locational change of the firm andencouraging employees to work more at home. It should, however, bementioned that responses differ strongly between types of firms. Industrial

127

firms see much less potential for adaptation to congestion than theservice-sector industries. The responses of individual trip-makers in order oflikeliness were the following:

− Earlier departure from home;− Changing working times;− Changing route to/from work;− Working more often at home;− Using public transportation more often;− Change job;− Work fewer days a week; and− Change home location.

Thus, the least attractive to workers is changing their home, whereasemployers -- and government -- see it as a realistic option.

Firms’ responses to growing congestion are somewhat different fromthose of individuals, but they too may consider a range of options includingshifts in time, in location, in modal use, etc. Again, it is necessary toemphasize that the range of adjustments may be wide -- and wider than thatconsidered by policymakers.

4. ADDRESSING CONGESTION: POLICYMAKINGAND POLICYTAKING

With congestion being an important item on both the public andpolicymakers’ agendas, it receives much attention in many countries. It mayeven receive too much attention. It also captures a significant share of theagenda of the transportation research community. A search for policymeasures which would curtail congestion is evident in the abundance ofprofessional literature and popular debates in most of the developed world.

Beyond the wide consensus that “something should be done”, a widerange of views are brought forward and a lively debate is evident. On the onehand, it refers to the definition of objectives. What is the desired level ofcongestion? The more popular position is that congestion should beminimised, namely nullified, while from a societal perspective, the question iswhat level of congestion is appropriate? On the other hand is the question of

128

the means: what types of policies can and should be pursued? Here, in anutshell, the dispute is between promoters of effective policies which may bepolitically too costly and policies which are attractive but less effective or evenineffective. Underlying these divergent views is a more philosophical questionon the degree of regulating peoples’ behaviour and the appropriateness of usingeconomic measures where there are no standard market mechanisms toencourage “correct” behaviour.

This chapter opens with a discussion of the objective, namely, questioningthe issue: is there a desired level of congestion and, if it exists, how is it to beidentified? Then, we discuss the interplay between policy and behaviour of thepolicy recipients. Finally, a brief review of policy approaches and anassessment of the potential policies in Europe is given.

4.1. A desired level of congestion? An economic approach

Congestion imposes costs in a number of different forms. First andforemost from a political perspective are the personal costs incurred by myriadsof individuals whose travel times are prolonged due to congestion. Second arethe costs incurred by society as a whole. This includes the uncompensated lossof time of individuals as well as the greater costs of friction in productionsystems. As congestion is also a very unstable situation, the travel timereliability during congestion is low. This is imposing additional costs onindividual users, who have to allocate sufficient time to account for uncertaintyin expected arrival times, as well as on firms which need to adjust to lowerreliability by adding, for example, to their inventories.

To the extent that individuals’ time has an economic value, these lossescan be translated into a social cost. In addition to the time losses, trafficmoving in congested regimes is likely to emit more air pollutants thanfree-flowing traffic and for some types of vehicles, noise emissions are alsolikely to increase. Safety may also become an issue in congestion. So,congestion entails significant costs on society. Estimates in the United Stateshave suggested a loss of $38 billion in 1988 (cited by Downs, 1992).In Europe, congestion costs are estimated at about 2 per cent of GDP(Kinnock, 1995).

If the costs are so high, something “ought to be done”. But what exactly isthe objective to be attained? The colloquial perception of the congestionproblem, as experienced daily by millions of Europeans (and others around the

129

urban world), is that capacity should be added by means of additional road andrail infrastructure. This view is also held by many policymakers. The questionof the desired level of congestion is only rarely addressed.

The suggestion that congestion should be “minimised” is dubious. It isnot possible to design and implement a transportation system which will notexperience congestion. That level of zero congestion can technically beaccomplished only through an unreasonable investment in construction. Forlong periods each day this system will be under-utilised, as will be the fundsinvested in its construction. From a societal perspective, there is a desired levelof social costs. Evans (1994) has pointed to the fact that “perfect safety” is nota socially reasonable policy objective, even though it involves such basicvalues as human lives. Similarly, zero pollution levels are not appropriate aspolicy objectives, as some non-zero level of pollution entails less social coststhan the social benefits accrued by its production (Arnott and Small, 1994).

The desired level of congestion is that level which optimises the socialcosts and benefits of the various elements involved, taking developments overtime into consideration.

The determination of the optimal congestion level is a complicated task, asthe quantification of some of the costs and benefits is subject to dispute byvarious interests such as industry, environmentalists and economists.Moreover, as infrastructure facilities have a very long life (a century or more),realistic assumptions about the present value of very long economic streams isliterally insignificant. Nevertheless, despite the methodological barriers,attempting to establish a desired level is a more productive approach thanresorting to irrelevant (zero level) objectives or hiding behind ambiguouslanguage for the definition of such objectives.

The costs and benefits which should be considered include capitalnecessary for construction, maintenance costs, environmental costs (airpollution, noise and the value of land), safety costs and benefits and travel timelosses and gains.

It is clear that there is a trade-off between capital investment andcongestion. Accepting higher levels of congestion will reduce construction andmaintenance costs (less roads, fewer lanes) but the costs of travel time losseswill increase.

130

For roughly ten years, the approach adopted in the Netherlands hasexplicitly considered congestion in the design procedure for trunk roads. It isrecognised that having congestion in the network is not all bad if it is containedat a desirable level.

The quality-of-flow criterion in use nowadays in the Netherlands for thedesign of motorways is the probability of congestion. This measure expresses,for a particular road section, the percentage of daily users of that section whichexperiences a queue. (It has been chosen as the measure to quantify the qualityof flow achievable with a given road layout vis-à-vis a predicted demand flowlevel.)

Compared to the classical criterion of speed, the probability of congestionadds two important criteria for quality of flow: travel time loss, namely, excesstravel time, and reliability of travel time.

This figure, as explained below, is based on the analysis of various costsassociated with the construction and use of roads.

Whereas the classical approaches rely on average values for capacity andflow, the estimation of congestion probability explicitly assumes that bothcapacity and flow are stochastic variables. Both fluctuate over time, partly in asystematic way due to hourly, daily and seasonal factors, but also to aconsiderable degree because of unpredictable influences such as road incidents,special events, weather conditions, and many more. Even if flows on averageare below capacity, the fluctuations in both quantities can give rise toconsiderable congestion. Based on historic observations, these variations arecaptured in a probabilistic model from which congestion probabilities aredetermined (Stembord, 1991).

So, if levels of congestion can be predicted, what is the maximum levelthat should be permitted by design, given economic considerations?

Economic analyses have shown that (in 1990) the optimal congestion levelfor the Dutch trunk road network is equal to a congestion probability of about2 per cent (Stembord, 1991). This optimal level means that, on average, over along period, 2 per cent of daily traffic on a road encounters congestion to somedegree. This needs some further clarification. If we assume that all congestiononly takes place in the two peak hours, one in the morning, the other in theafternoon, where each of which carries 10 per cent of daily traffic, then this2 per cent means that there is a 10 per cent chance of peak-hour travellersgetting in a queue on a working day (off-peak travellers will not meet

131

congestion at all). That is only a few minutes’ delay in a queue once afortnight. If we had, however, a road section with a congestion probability of,e.g. 20 per cent (such cases do occur), this would mean that we had a structuralbottleneck with recurrent congestion during peak hours with delays of20 minutes or more each working day.

The dimensioning of the roads and the capacity calculations are nowadaysbased on this 2 per cent congestion standard. It is considered an economicoptimum.

The determination of the optimal congestion level took into considerationthe following costs: construction, maintenance, safety, travel time losses andenvironmental damage. It is clear that accepting higher levels of congestionwill diminish construction and maintenance costs (fewer roads, fewer lanes) butthe costs of travel time losses will increase. The overall minimum appeared tobe at 2 per cent congestion probability (Figure 5); at that level, the total socialcost of the trunk road network is considered minimal.

Figure 5. Social costs of road infrastructure provision in relation toaccepted congestion levels

0 1 2 3 4 5 6

Optimumca. 2%

Congestion percentage

Social costs

The interesting question, of course, is how current traffic flow conditionsrelate to this standard?

132

Let us compare the 1995 situation in the Netherlands vis-à-vis thisoptimum level of 2 per cent (see Figure 6).

Figure 6. Current congestion levels in the Randstad motorway network

5 - 10 %

2 - 5 %

10 - 15 %

> 15 %

< 2 %

From data as shown in Figure 6, it can be derived that in 1995 nearly20 per cent of the national trunk network was characterised by highercongestion levels than the required maximum standard. In the Randstad area,only 50 per cent of the links satisfy the standard, and even on the hinterlandaxes only 65 per cent of the links were acceptable. These latter roads beartwo-thirds of the national congestion hours. There are many links with acongestion probability of more than 20 per cent, which means permanent dailycongestion during peak hours.

133

Considering that congestion indicators only show the visible part of theproblem, neglecting the unknown, suppressed latent demand, we may safelysay that, as in most large conurbations, the low countries suffer seriously fromroad traffic congestion.

4.2. The gap between policymakers and policytakers

Government authorities, at all geographical levels, seem to be sensitive tocongestion for a number of reasons. Congestion is not only an economicburden, but it is also a major political concern, as it negatively affects the livesof many citizens. Moreover, congestion is associated with environmental costs,yet another reason for its prominence on the public agenda. Hence, local,regional and national level governments are preoccupied with the question ofhow to curb congestion. Some European level agencies are also involved in theefforts, as can be seen from the large number of congestion-related researchpublished by various agencies, such as the ECMT, OECD and others, in recentyears.

Needless to say, underlying each policy measure is a set of assumptions bythe policymaker with regard to the potential impacts of the policy. Policyanalysis is a necessary prerequisite to any implementation of a policy, so thatits potential benefits, effectiveness and costs can be considered in thedecisionmaking process.

The discrepancy between individual and external costs as the underlyingcause of congestion must be borne in mind when policy measures to curbcongestion are devised. Very often, such policies assume that an individualwill respond to a policy in a manner congruent with the social objective.Assuming that people will change their commuting patterns so as to improvethe environment, however, may prove to be too optimistic. Very likely,individuals will respond in a manner which best suits them. As Arnott andSmall (1994) have stated:

“It is also clear that some of the common-sense solutions do not solve theproblem. Only by understanding the full nature of people's traveldecisions and how they interact can sensible policies be formulated(p. 455).”

134

Among the necessary criteria for policy evaluation, one must include ananalysis of the behavioural assumptions which underlie the policy underconsideration. Too often, policymakers assume that “policytakers” will behavein a certain manner, congruent with the policy objectives. Such assumptionsmay be valid for policies which expand the supply and allow individuals toadjust their behaviour to their convenience. However, with the growinginterest in demand management techniques, which tend to constrain theindividual’s behaviour, such simplistic assumptions may be misleading. Forexample, Athens (Greece) introduced, during the 1980s, a curb on auto use toreduce automotive emissions. A given vehicle was allowed to enter the centralbusiness district area only on odd or even days based on the last digit of thelicence plate. The outcome was an increase in car ownership, where the secondcar was commonly older, polluting more than the new car (Giaoutzi andDamianides, 1990). Such a response was not anticipated and rendered thepolicy useless.

Policytakers tend to evade restrictive policies and invent responses thatallow them to maintain their objectives at a minimum cost. This can partiallybe done by transferring the costs to others. Thus, it is suggested that thepotential gap between the policymakers’ perspective and that of thepolicytakers must be considered in the evaluation of congestion-mitigatingstrategies.

The assumptions that are (often implicitly) incorporated intocongestion-mitigating policies seem to be part of the reasons for the relativelack of success in reducing congestion. The following assumptions seem to beparticularly incongruent with current understanding of travel behaviour:

− Assuming fixed travel demand and ignoring the possiblematerialisation of latent demand;

− Assuming that travellers are cost minimisers rather than utilitymaximisers;

− Assuming that only a limited choice set is available to the individual,and consequently that the addition of an option is likely to have asignificant effect; and

− Assuming that responses to demand-management techniques aresimilar to those for supply-side measures.

135

4.3. Policy approaches

Historically, it is possible to identify at least three periods in which policymeasures to curb congestion have emerged from very different assumptionsabout the nature of the problem. Initially, and through the mid-1960s, theprincipal tool was expansion of infrastructure: more roads were built toaccommodate demand. Later, there was a shift toward improved managementof the available infrastructure. This was the Transportation SystemsManagement (TSM) period, which prevailed during the 1970s, and TSM is stilla relevant tool. However, TSM is also limited in its potential contribution, andin the early 1980s there was an increasing realisation that altering humanbehaviour is the next necessary step. This led to the development andimplementation of Transportation Demand Management (TDM) strategies,involving a wide range of policies to reduce dependence on the drive-aloneautomobile.

While the first two periods can be characterised as emphasizingsupply-side measures, the third is, by definition, designed to affect demand.Supply-side measures which cater to accommodating demand are likely to bepositively received by users (albeit not necessarily by non-users, who may bethe very same individuals when they are not behind the steering wheel).Politically, measures which infringe on constituents’ personal behaviour (andfreedom) are considered undesirable and, therefore, according to Altshuler(1979), policymakers refrain from implementing policies which have directnegative impacts on users, such as those directed at modifying demand.Rather, where possible, policymakers will prefer a policy that “looks good”even if its effectiveness may be limited.

The case of road pricing, widely advocated by transportation professionals as apromising congestion management policy, but so rarely applied, is a clear exampleof a policy which directly affects constituents' pockets (Emmerink et al., 1994;Jones, 1991; Grieco and Jones, 1994; Wachs, 1994).

Supply-side and demand-side interventions differ in another aspect whichis important in the current context. Generally, the direction of behaviouralresponse to supply-side measures can be expected to conform to thatanticipated by the policymakers, and the question is whether levels of adoptionwill be lower (as is often the case for ridership on a new transit service) orhigher (as when the release of latent demand triggers nearly immediatecongestion on a new facility) than forecast. However, in the case ofdemand-side measures, the individual is confronted with a situation which

136

imposes a constraint. In this case, new “outlets” are likely to be sought, andinnovation may generate new, possibly unexpected responses, as describedbelow.

From a public policy perspective, congestion mitigation strategies can beclassified into five groups, differing in the nature of intervention assumed:regulation, planning, economic, technological and educational.

Regulation includes a wide variety of measures, many of which are at thedisposal of local agencies, through which the policymaker exercises somepower to alter behaviour of consumers, i.e. transport system users. This mayinclude parking restrictions, changes in schedules of work and schools, etc.Regulations are not politically attractive and are likely to be economicallyinefficient, but their major advantage is that they are relatively easy toimplement and consequently, if they are effective, will deliver the benefits inthe short term.

Planning2 includes a variety of measures intended to change the physicalenvironment by changing the spatial relationship of opportunities. Land usechanges which alter the density or the mix of land use are often suggested toenable a greater reliance on non-motorised travel, and hence a reduction oftravel by car. On the other hand, the planning, and implementation, oftransport infrastructure facilities also change the relative positioning ofopportunities and consequently are likely to affect travel patterns. Usingplanning strategies to contain congestion is a long-term policy and its prospectsare widely disputed (Handy, 1997; Breheney, 1995).

Economic measures are widely considered to be effective policyinstruments as they send the users unambiguous signals as to the desiredchange of behaviour. Some European countries have begun to implement suchmeasures as congestion pricing, or more generally, road pricing. Pricing ofcentral city parking to reflect the externalities is also a possible measure. It isexpected that with the introduction of electronic toll collection systems, therewill be growing interest in economic measures, despite the widespreadopposition by the public and elected officials, who view it as another tax.

Technology-based approaches include a variety of measures which mayimprove the management of transport systems, including improvements in themanagement of road capacity. Intelligent Transport Systems (ITS) provide arange of options to act both upon the supply side and upon the demand side.

137

Education is the often forgotten policy measure. Many policymakers andthe public at large are often unaware of the nature of the congestion problemand, more so, have unrealistic expectations with regard to its “solution”. Henceeducation, which is geared to explain the nature of the problem and theimplications of various policies, is important. However, the effectiveness ofeducation, if any, is realised in very long time horizons and that is probably themain reason for its negligence.

Figure 7 repeats the structure presented in Figure 1 above, but focuses onthe policy side. It complements Figure 1 by showing how congestionmitigation policies can be enacted to affect various factors which causecongestion.

138

Figure 7. Influence of relationships of congestion mitigation policieson congestion factors

Economic factors

Economic efficiency

Income

Energy costs

Socio-demographicPopulation size

Women’s roles

Household population

Driving population

Temporal structure

Work

Non-work

Transport supply

Network capacity

Competitive positionof public transport

Car availability

Spatial structure

Suburbanisation

Accessibility

Relocation

Car use

Congestion Improved speedPolicy

Land use

Technology

Infrastructure expansion

Traffic management

Demand management

Road pricingTemporal policies

Transportsystem

performance

139

5. CONCLUSIONS

The analysis presented in this report gives rise to the followingconclusions.

5.1. The notion and extent of congestion

A clear, unambiguous and widely accepted definition of congestion andhow it should be measured, is not available. A systematic data collection onroad congestion is absent (with the Netherlands as an exception). Realistic andcomparable figures about the extent of congestion and its costs are thereforelacking (with the exception of the ECIS-study). This hampers a validinternational comparison of congestion conditions in European states andregions. Comparative studies can potentially provide insights into the differentcauses of congestion, under varying background conditions and, moreimportantly, lessons about the effectiveness of policy approaches. The figurespresently available on the extent of congestion in Europe, as those used inmany official European documents (EC, OECD, ECMT), lack a validfoundation and offer little sound basis for decisionmaking.

5.2. The use of congestion measures as quality indicators

Congestion figures, as such (e.g. queue length, congestion duration, excesstime, excess costs), are poor indicators of network quality or trip quality.Congestion is not necessarily a sign of a poorly designed network or of anunacceptable quality of flow. From an economic perspective, there exists anoptimal level of congestion in transportation systems which depends on localcircumstances, such as construction costs and travellers’ value-of-time as wellas the weight of transport and environmental considerations relative to othersocial problems.

5.3. The spread of congestion levels

If measured at an aggregate (network) level (e.g. total excess travel hours,total queue lengths), congestion has increased constantly over the last decade,at a rate similar to that of motorway usage (on average 4 per cent per annum).Much of this aggregate growth is a spread of peak levels in time and space.

140

This is a useful indicator from a public policy point of view. With respect tothe traveller’s choice behaviour, however, the congestion experienced by theindividual user, as measured directly, remains quite constant.

5.4. The European dimension of congestion

The scale of the congestion problem is not European, nor is it national. Itis also not a typical characteristic of major long-distance thoroughfares orborder crossings. Road traffic congestion is an urban and regional(metropolitan) problem. It manifests itself predominantly in and aroundhigh-density conurbations. International comparisons, therefore, should rely onregional data rather than national level data.

5.5. The true costs of congestion

Given the absence of unambiguous congestion figures and the spreadingof congestion in time and space, there is a widely accepted expectation of animminent catastrophe or breakdown of the system. This view seems to beexaggerated and the costs of congestion, as shown in various officialpublications may be overestimated. Congestion on main roads appears to be alocal problem which affects a limited number of travellers. Recalculated tonational level figures, the excess travel time, most probably, is less than 2 percent: excess travel time amounts to 0.5 per cent of the Gross DomesticProduct.

5.6. Variation in the distribution of congestion in Europe

Road traffic congestion is evident in many different regions of Europe.But congestion levels, growth rates and distributions differ widely betweencountries and regions. These differences can be attributed to variations in theunderlying factors which generate congestion, in particular, spatial patterns ofland use and network conditions. The most important common factor ispopulation density. In high-density areas, space for sufficient roadinfrastructure is scarce, whereas the density of potential users is high.

141

5.7. The responses to congestion

When facing changing congestion levels, users of the networks(individuals and firms) have a wide range of behavioural responses at theirdisposal. Some responses are short-term travel adjustments while others arelong-term, locational and life-style changes. A variety of behavioural changesare evident when policies to relieve congestion are implemented: shifts in time(return to the peak), route and mode diversion and a change of tripmakingbehaviour are among the common responses. They reflect mostly shifts inexisting demand. New transport demand as a result of congestion reliefappears to be very limited.

5.8. The limits of congestion

Total demand for road transport is not unlimited. Population growth andincreased speeds due to increased welfare were the predominant factors forroad transport consumption. The growth in total road traffic kilometrage islikely to diminish as populations grow at a slower pace and travel time budgetsof travellers are increasingly binding, giving rise to modification of activities inlieu of increased tripmaking.

5.9. Investments in congestion relief

Lack of investments in road infrastructure is a major factor in congestiongrowth in Europe. In most countries the lion’s share of national infrastructureinvestments is in rail, predominantly for long-distance connections. These willhardly contribute to road congestion relief because they serve thin,long-distance travel flows. A serious problem is that investments inlong-distance rail are at the expense of short-distance public transport networksand road investments as well. The contribution of urban and regional publictransport investments to congestion relief is limited because of inherent systemcharacteristics of public transport (such as spatial coverage and servicequalities).

142

6. RECOMMENDATIONS

Based on the analysis and conclusions presented above, a number ofpertinent recommendations are warranted.

6.1. Statistics on congestion

In view of the dominant role of congestion in public policymaking, thereis a need for development of clear, unambiguous operational definitions forcongestion measures and measurement methods suited for cross-regionalcomparison. A system of European-wide statistics on congestion should beestablished. A distinction should be made between aggregate (network-wide)measures and individual (traveller-related) indicators. Such statistics willimprove information provision about congestion conditions and their societalimplications, and will thus enhance transport policy decisionmaking, inparticular with regard to investments in infrastructure, pricing and publictransport.

6.2. Optimum level of congestion

The notion that there is an optimum, non-zero level of congestionshould be developed further and should be communicated to opinion leaders,politicians and interest groups. An approach should be developed to determinethe optimum congestion level that can be used in road network planning anddesign.

6.3. Need for balanced spatial development

Long-term solutions to congestion have to be directed at curbing thegrowing travel demand and travel distances by a balanced development ofinfrastructural networks and spatial distribution of activities. New concepts ofspatial configurations of settlements need to be developed and assessed. It isstill unclear how spatial structure affect changes in behaviour, but there is nodoubt that density is closely related to the efficiency of different modes oftravel. Thus, careful examination of these relationships is warranted.

143

6.4. Public transport is an ineffective congestion relief measure

Development and stimulation of alternative modes (such as publictransport) as a means to tackle congestion is not a very effective policyapproach in general, except for specific high-density corridors. This is due totheir inherent system characteristics. Rail transport can only serve very limitedsegments of the travel market. Public transport has to play an importantfunction in offering services in dense areas and corridors and giving mobilityopportunities to the careless.

6.5. “Only the road can relieve the road” [Gerondeau (1997)]

Addressing congestion, from the supply side, is most effective byintervening in the road system itself instead of suggesting alternative modes.Such an approach includes offering extra capacity by widening of roads,building buffers to reduce secondary congestion, increasing road capacity andcapacity utilisation by dynamic traffic management and including demandmanagement techniques, such as congestion pricing. This does not imply thatthe road system should be indiscriminately expanded. Other considerations,such as environmental quality, social and spatial impacts, must also be takeninto account. However, the hope that public transport can “solve” congestion,is probably an illusion.

6.6. Need for high-quality roads

Given the forthcoming demographic and economic developments thereseems no escape but to upgrade the road infrastructures significantly, both in aquantitative (additional capacity) and qualitative (environmental) respect. Newroads as well as upgrading of existing roads must meet high environmental andaesthetic standards. Because congestion-prone areas are characterised byscarcity of space, and a vulnerable natural and manmade environment (noise,aesthetics, etc.), costly solutions are inevitable. These may include, in extremecases, underground or deepened roads, double-stack roads, roads with coveringand special (double-layered) tunnels for private cars and trucks. In addition,investments are needed in high-quality transfer facilities at the fringes of citiesfor travellers going to city centres by public transport or other modes. In viewof the dynamics of the process of congestion building, such high quality roadsmay not need to increase speeds, but ensure flow at reasonable levels ofservice.

144

NOTES

1. This section draws heavily on Stern, Bovy and Tacken, 1995; Goodwinet al., 1992; Salomon and Mokhtarian, 1997 and Mokhtarian, Raney andSalomon, 1998.

2. From a public policy perspective, planning is a form of regulation.However, in view of its unique importance in transportation policy, weaddress it separately.

145

ANNEX

“STATISTICS ON ROAD TRAFFIC CONGESTION INTHE NETHERLANDS”

In order to show the possibilities of information provision on congestion,the Dutch method of collecting and producing congestion information isconcisely explained below (for more details, see Ministry of Transport, 1996and 1997 and NEA, 1997).

Three types of statistics on congestion are assembled and published, on:

a) Queues (locations, frequency, weight, etc.);b) Congestion levels (LOS, congestion probability, etc.);c) Travel time losses and costs.

The statistics relate to the main road network under national leveljurisdiction.

(a) Statistics on queues (Ministry of Transport, 1997)

These have been collected and published since 1983. The unit ofobservation is a traffic queue (standing or slowly moving platoons of vehicles).The national traffic police collects and processes reports on queues given bydrivers, police patrols and traffic service patrols underway who observe, or arein, a queue. Ninety per cent or more of the reports are given by drivers usingtheir car or mobile telephone.

After processing these messages, the following information on a particularqueue is available:

146

− Cause;− Location;− Starting time;− Length per period;− Ending time.

The following causes are distinguished, among others:

− Bottleneck in network (insufficient capacity);− Roadworks (reduced capacity);− Accidents;− Special events (exceptionally high demand);− Weather conditions;− Public actions, demonstrations.

From the messages, the following characteristics can be derived for eachseparate queue:

− Average length;− Maximum length;− Total duration;− Weight (product of queue length, time step length and number of

lanes summed over all congested time steps).

In 1996, nearly 16 000 queues were observed and processed.

The statistics on queues comprise tables, figures, maps, etc. on, inter alia,the following data:

− Rank ordering of queue locations according to frequency and weight;− Total queue frequency, duration and weight, classified by region, road

type, cause, day type, period of the day, month, etc.;− Road maps showing intensity of queue occurrence and weight

classified by cause.

(b) Statistics on congestion levels (Ministry of Transport, 1996)

A second statistic gives yearly congestion characteristics of all roadsections based on measured hourly flows (classified by vehicle type) and theroad section’s capacity (derived from the section’s dimensions). The unit ofobservation is the road section (by direction). A calibrated statistical

147

congestion estimation model (Van Toorenburg, 1991) calculates congestionprobabilities (yearly averaged percentage of daily flow which encounterscongestion) and excess travel times due to congestion.

The model takes into account the effect of stochastic variations in capacityand demand as a function of flow level, hourly flow pattern and road type.

The congestion statistic is presented in the form of road maps showingcongestion probability levels.

Tables show which roads do not meet the level of service standards (2 percent congestion probability for hinterland axes, 5 per cent at other main roads).

(c) Statistics on time loss and costs (NEA, 1997)

These statistics give time losses and costs caused by queues split bypassengers and goods, trip purposes, type of queue. The information basis isthe queue data base (a) and the trip characteristics at a sample of road sectionsare derived from regular roadside surveys.

Travel time loss is calculated using a queuing build-up and dissipation model.Queuing costs consist of time costs, calculated using standard value-of-time figures,and extra vehicle costs (such as extra fuel consumption).

The statistics consist of tables of excess travel times and costs,respectively, split by period of day, day type, trip purpose, vehicle type, etc.

148

REFERENCES

Ministry of Transport (1997), Verkeersgegevens Jaarrapport 1996 (in Dutch)(Traffic data yearly report 1996), Rotterdam, Adviesdienst Verkeer enVervoer, July.

Ministry of Transport (1996), Beleidseffectmeting Verkeer en Vervoer:beleidseffectrapportage 1995 (in Dutch), (Traffic and Transport Policy impactmeasurement: policy impact report 1995), The Hague, Ministry of Transport,Strategy and Programming Directorate, September.

NEA (1997), Filekosten op het Nederlandse hoofdwegennet in 1996 (in Dutch),(Congestion costs at Dutch main road network in 1996), Rijswijk NEA, 1997,on behalf of Dutch Ministry of Transport.

Van Toorenburg, J.A.C. (1991), Performance of motorways and trunk routes athigh traffic volumes, in: U. Brannolte (ed.), Highway capacity and level ofservice, Rotterdam, Balkema, 1991, pp. 413-418.

149

BIBLIOGRAPHY

Altshuler, A. (1979), The Urban Transportation System, MIT Press,Cambridge, Mass.

Arnott, R. and Small K. (1994), "The economics of traffic congestion",American Scientist, 82, 446-455.

Billheimer, J. (1978), "The Santa Monica Freeway Diamond Lanes: EvaluationOverview", Presentation at the Annual Transportation Research BoardMeeting, Washington, DC.

Breheny, M. (1995), “The compact city and transport energy consumption”,Transaction of the Institute of British Geographers, 20, 81-101.

Brühning, E., et al. (1997), Entwicklung der Verkehrsicherheit auf EuropäischeAutobahnen (in German) (Development of traffic safety on Europeanmotorways), Strassenverkehrstechnik, January.

Bukold, S. (1997), Bottlenecks in European Infrastructure, Rotterdam, ECIS,January.

Cervero, R. (1991), Congestion, growth and public choices, University ofCalifornia Transportation Center, Reprint No. 51, Berkeley, California.

DHV/Colquhoun (1991), The cost of inadequate transport infrastructure inEurope, Report to European Parliament.

Coughlin, J. (1994), “The tragedy of the concrete commons: Defining trafficcongestion as a public problem”, in: Rochefort, D. and R. Cobbs (eds.),The Politics of Problem Definition, Kansas.

150

Delucchi, M. (1997), “The social cost of motor vehicle use”, Annals of theAmerican Academy of Political and Social Sciences, 553, 130-142.

Downs, A. (1992), Stuck in Traffic: Coping with Peak-Hour Congestion,Brookings Institute, Washington DC.

ECMT (1993), Survey procedure and results by country, Paris, November.

ECMT (1995), European transport trends and infrastructural needs, Paris.

Emmerink, R.H.M, P. Nijkamp and P. Rietveld (1994), How feasible iscongestion pricing?, Tinbergen Institute, TI, 94-62.

Evans, A. (1994), “Evaluating public transport and road safety measures”,Accident Analysis and Prevention, 26, 4, 411-428.

Gerondeau, C. (1997), Transport in Europe, London, Artech House.

Giaoutzi, M. and L. Damianidias (1990), "The Greek transport system andenvironment", in: Brade, J. and K. Button.(eds.), Transport policy and theenvironment. Six case studies, Earthscan, London.

Giuliano, G. and K. Small (1994), “Alternative strategies for coping with trafficcongestion”, University of California Transportation Center, WorkingPaper No. 188, Berkeley, California.

Goodwin, P., P. Jones, J. Polak, P. Bonsall and J. Bates (1992), AdaptiveResponses to Congestion: Proposals for a Research Programme, TransportStudies Unit, Oxford University (TSU Ref: 736).

Gordon, P. and H. Richardson (1991), "The commuting paradox", Journal ofthe American Planning Association, 57, 4, 416-420.

Grieco, M. and P.M. Jones (1994), "A change in the policy climate? CurrentEuropean perspectives on road pricing", Urban Studies, 31, 9, 1517-1532.

Gutierrez, J. and P. Urbano (1996), "Accessibility in the European Union: Theimpact of the trans-European road network", Transport Geography, 4,1,15-26.

Handy, S. (1997), “Travel behaviour--land use interactions: An overview andassessment of the research”, presented at the IATBR Meeting, Austin, Tx.

151

Hendriks, F. et al. (1997), Infrastructureel investeringsbeleid in vergelijkendperspectief (in Dutch), (Infrastructure investment policy in a comparativeperspective: a transport policy analysis of the Randstad, Ruhr area andFlanders city ring), Tilburg, Catholic University/TNO-Inro.

Hilbers, H.D. and E.J. Verroen, (1996), “An international comparison ofaccessibility and congestion problems of urban areas: Can we stillcompete with our neighbours?”, PTRC.

Hilbers, H.J. et al. (1997), Infrastructuur en mobiliteit in de Randstad (inDutch), (Infrastructure and mobility in The Randstad, the Ruhr Area andthe Antwerp-Brussels-Gent region), Delft, TNO-Inro.

IMD/World Economic Forum (1996), Global Competitiveness Report, Genova,IMD, 1996.

IWW/NEA et al. (1996), Bottlenecks in the European transport infrastructure:final report. Karlsruhe/Rijswijk, study on behalf of ECIS.

Jones, P. (1991), "Gaining public support for road pricing through a packageapproach", Traffic Engineering and Control, April.

Kageson, P. (1993), Getting the prices right: A European scheme for makingtransport pay its true costs, The European Federation for Transport andEnvironment, Stockholm.

Kinnock, N. (1995), Towards fair and efficient pricing in transport, EC,Brussels.

Korver, W. et al. (1992), Gedragsveranderingen bij bedrijven als gevolg vanreistijdvertragingen op het wegennet: deel II het zakelijkepersonenverkeer (in Dutch), (Congestion delays in personal businesstravel: behavioural changes of companies), Delft, INRO-TNO.

Kroes, E., A. Daly, H. Gunn, T. Van der Hoorn (1996), “The opening of theAmsterdam ring road: A case study on short term effects of removingbottlenecks”, Transportation, 23, 71-82.

Litman, T. (1997), “Policy implications of full social costs”, Annals of theAmerican Academy of Political and Social Sciences, 553, 143- 156.

152

Meyer, M.D. (1990), Dealing with congestion from a regional perspective: thecase of Massachusetts, Transportation, Vol. 16 (1990), 197-220.

Ministry of Transport (1996), An international comparative study oninfrastructure, The Hague, SDU Publishers.

Ministry of Transport (1996a), Beleidseffectmeting Verkeer en Vervoer:beleidseffectrapportage 1995 (in Dutch), (Traffic and Transport Policyimpact measurement: policy impact report 1995), The Hague, Ministry ofTransport, Strategy and Programming Directorate, September.

Ministry of Transport (1997), Verkeersgegevens Jaarrapport 1996 (in Dutch),(Traffic data yearly report 1996), Rotterdam, Adviesdienst Verkeer enVervoer, July.

NEA (1996), ECIS Study: Bottlenecks in European transport networks; roadtransport, Rijswijk (Netherlands), study on behalf of ECIS.

NEA (1997), Filekosten op het Nederlandse hoofdwegennet in 1996 (in Dutch),(Congestion costs at Dutch main road network in 1996), Rijswijk NEA, onbehalf of Dutch Ministry of Transport.

Pucher, J. and C. Lefevre (1996), The Urban Transportation Crisis in Europeand North America, Macmillan, London.

Quinet, E. (1994), “The social costs of transport”, in: Internalising the socialcosts of transport, ECMT/OECD, Paris, 31-75.

Salomon, I., P.H.L. Bovy, and J.-P. Orfeuil (eds.) (1993), A billion trips a day:Tradition and Transition in European Travel Patterns, Dordrecht, KluwerAcademic Publishers, 1993.

Salomon I. and P. Mokhtarian (1997), “Coping with congestion: Reconcilingbehavioural responses and policy analysis”, Transportation Research,D, 2, 2, 107-123.

Stembord, H.L. (1991), “Quality of service on the main road network in theNetherlands”, in: U. Brannolte (ed.), Highway Capacity and Level ofService, Rotterdam, Balkema, pp. 357-365.

153

Stern, E., P. Bovy and M. Tacken (1995), "Traffic congestion and behaviouralreaction", European Research Conference on “European transport andcommunication networks: Policies on European networks”, Espinho,Portugal, April 17-23.

Tacken, M. and E. DeBoer (1991), “Flexitime and the spread of traffic peakhour: an analysis of conditions and behaviour”, Delft University ofTechnology, OSPA (in Dutch).

Tacken, M. and E. DeBoer (1991), “Change in spread of travel and workingtimes due to opening of the Amsterdam Orbital motorway”, DelftUniversity of Technology, OSPA (in Dutch).

Transroute ISIS, Heusch-Boesefeldt and A.T. Kearney (1992), EC MotorwayNetwork Perspectives, Study for CEC-DG VII.

Van Toorenburg, J.A.C. (1991), “Performance of motorways and trunk routesat high traffic volumes”, in: U. Brannolte (ed.), Highway capacity andlevel of service, Rotterdam, Balkema, 1991, pp. 413-418.

Wachs, M. (1994), "Will congestion pricing ever be adopted?", Access, 4,15-19.

Westland, D. (1997), “The Gattis hypothesis tested on Dutch motorwaybottlenecks”, Delft University of Technology, Faculty of CivilEngineering

WP5 (1994), Report on the methodological basis for the definition of commoncriteria regarding bottlenecks, missing links, and quality of service oftransport infrastructure networks, Bonn, June.

155

UNITED KINGDOM

J.M. DARGAYP.B. GOODWIN

ESRC Transport Studies UnitUniversity College London

United Kingdom

156

157

SUMMARY

1. EARLY GROWTH OF CONGESTION IN EUROPE............................159

2. DEFINITION OF CONGESTION...........................................................160

3. INDICATORS OF CONGESTION .........................................................166

4. INFLUENCES ON TRAFFIC GROWTH...............................................180

4.1. The effect of income on car ownership ............................................1814.2. The effects of costs ...........................................................................184

5. CONGESTION SCENARIOS .................................................................191

6. CONCLUSIONS......................................................................................194

NOTES.............................................................................................................197

REFERENCES.................................................................................................199

London, December 1997

158

159

1. EARLY GROWTH OF CONGESTION IN EUROPE

Lay (1993) observes that traffic congestion is not a new phenomenon.Two thousand years ago, a Roman edict declared that “the circulation of thepeople should not be hindered by numerous litters and noisy chariots”. AncientPompeii had parking restrictions, and Julius Caesar introduced the first knownoff-street parking laws. The centre of Rome was banned to vehicles between6 a.m. and 4 p.m., and in AD 125 Hadrian limited the number of vehiclesentering Rome. Around AD 180 Marcus Aurelius extended the bans inprinciple to all towns in the Roman Empire, which is probably the first exampleof a European agency seeking to go somewhat beyond what would be describednow as the principle of subsidiarity. It is not clear whether the ban wasuniformly implemented in practice.

Since then, many mediaeval and industrial cities have their local accountsof the occasional problems of excessive traffic, and attempts to deal with it.They also have found that their wealth and power depended in large measureon incomes from trade, and the Hansa towns in particular stand as examples ofthe prosperity which can be generated if a city can gain control over someproportion of the money spent on the movement of goods and people. Thisreminds us that in some senses congestion is simply a property of popularity.Cities are places where people wish to congregate in order to carry outexchange of goods and ideas, and it is natural that such congregation shouldresult in crowds.

This should not be taken to imply, however, that there is nothing new inthe scale, effects, speed of growth, environmental, economic and socialconsequences, and extensiveness of modern congestion. We have a problemwhich is similar in form, but different in content, from that of the ancient bustleof successful cities.

160

2. DEFINITION OF CONGESTION

The fundamental defining relationship of traffic engineering is thespeed-flow curve. This shows that the more traffic uses a road, the slower itgoes, the effect becoming more and more severe as the traffic flow approachesthe maximum capacity of the network, until finally overload is so extreme thatall vehicles are unable to move. If we extend the idea of travel time to a widerdefinition of cost including inconvenience and discomfort, the same processmay be taken to apply as a general rule for virtually all forms of transport, andindeed to some extent as a general property of all systems subject to capacityconstraints and some degree of random variation.

Thus congestion is a characteristic of all heavily used transport systems.Its general feature is that users impede each other’s freedom of movement. Thegeneral definition of congestion therefore most usefully relates to this generalproperty of transport systems, namely:

Congestion is defined as the impedance vehicles impose on each other,due to the speed-flow relationship, in conditions where the use of a transportsystem approaches its capacity.

This definition indicates that the underlying cause of congestion does notconsist of the transient and immediate triggers which drivers notice when theyare in a traffic queue, such as roadworks or taxis or accidents: the cause isbecause traffic flows are too close to capacity, when any of these transientincidents will have a disproportionate effect.

161

Measures of congestion

The classic method of measuring the amount of congestion, andconverting it into an economic value, was devised by Glanville and Smeed(1958). It is elegant, simple, and has had a pervasive influence in the last40 years. In spite of its provenance, it is based on a conceptual error.

Glanville and Smeed proposed:

The calculated total cost of delay depends on what is regarded as areasonable speed for traffic. Under light traffic conditions on good roadsthe average speed of traffic is about twenty-five mph (40 km/h) in built-upareas, and forty mph (65 km/h) in non built-up areas. Taking these asstandards, calculations give a cost of £125 million in urban areas and£45 million in rural areas, making a total of £170 million per annum.

Thirty years later, the British Road Federation (1988) did a similarcalculation, using essentially the same method, and concluded:

The additional cost over and above that experienced in free flowconditions is defined as the congestion cost. This amounts to £3 billionper year in the conurbations alone.

The Confederation of British Industry (1988) estimated £15 billion peryear for the UK. From time to time this figure has been updated, either simplyby inflation, or by new calculations. Newbury (1995) proposed £19.1 billionin 1993.

Allowing for inflation and changing values of time, such figures indicatethat the cost of congestion has increased in real terms by something like400-500 per cent in four decades.

Possibly the most careful example of this approach has been produced byDodgson and Lane (1997). They proposed a somewhat more rigorous methodof calculating the cost of congestion, summarised in Figure 1. They suggestthat the cost of congestion in this diagram is equal to the area gfg1de, beingnumber of vehicles multiplied by the difference between the averagegeneralised cost per vehicle in free flow conditions and in the actual conditions(as distinct from multiplying by the marginal costs of congestion which, theyclaim, is implicitly done by Newbury), and disaggregating for many road typesand vehicle types.

162

Figure 1. The cost of traffic congestion

(v)

vf

v1

(g)

gf

g1

0 To T1 Tc

0 To T1 Tc

d

e

Highway speed-flow curve

Speed

Generalised costTraffic flow

Generalised cost per km

= a + b/v + cv2

Generalised costper vehicle-kmwith no congestion

Traffic flow

Source: Dodgson and Lane (1997).

Their figure amounted to £7 billion -- less than half the previously quotedfigure, though of course still an appreciable cost.

The UK Department of the Environment, Transport and the Regions,DETR (1997) suggested that under current growth trends, traffic levels wouldincrease for various road types by 31 per cent to 117 per cent by 2031, andconsequently journey times would increase -- to double their present level forurban motorways in the peak hour, but only by 5 per cent or so for off-peaktravel on rural principal roads. Overall, this would imply a further increase inthe costs of congestion, as calculated, by about 100 per cent.

Such calculations are ubiquitous in national and international transportpolicy discussions. They all show that traffic congestion is a very large cost tothe economy, which has been increasing and is expected to increase in thefuture.

At this point in the argument, however, it is useful to pause to consider aproblem. Overall, travel has been getting faster, not slower.

163

How do we reconcile observations of increasing speeds of travel, withcalculations of increasing costs of congestion?

There are two elements to this reconciliation. First, the proposed methodof calculating the costs of congestion is wrong. Secondly, travel behaviouradapts to changing conditions.

The Erroneous Calculation

The method of calculating congestion costs due to Glanville and Smeed isbased on the following formula:

(Time at “target” speed) - (Time at actual speed)

multiplied by

(Volume of traffic)

equals

(Total Congestion Delays)

What this means is that the target speed changes, congestion costs ascalculated can increase even if nobody is actually worse off. Considerpeak-period traffic travelling at 20 km/h on a 30 km/h local road. Then weredesignate the road as 60 km/h, and implement improvements allowing anactual peak speed of 25 km/h. According to the formula, congestion costs arenow greater, though in fact every vehicle is travelling faster.

This occurs even if the volume of traffic does not change. However, iftraffic grows, then the congestion cost must grow with it, even if speeds do notfall -- or even if they increase, but by less than the volume of traffic.

So the calculation would say that a growing volume of traffic, using acontinually improved road system, at continually increasing speeds, could stillbe suffering an increased total cost of congestion.

164

And conversely, if we revise downwards our accepted “target” speed (as iswidely done in speed restrictions), or if we alter road design in such a way thatthe free-flow speed of traffic falls (as is widely done in traffic calming) then ineither case the calculated total cost of congestion to the economy would appearto fall.

It is difficult to persuade ourselves that such measures of the cost ofcongestion tell us something useful about the economy, or about transportpolicy.

Adaptive behaviour

The calculations above all rely on comparing the real travelling conditionswith hypothetical conditions that would apply if the same volume of trafficwere able to enjoy faster speeds. In the real world, the volume of traffic issubject both to external changes and also to those brought about by changes inthe ease of travel. This presents a great difficulty in calculating the costs ofaverage quantities over an economy as a whole.

The following example -- with invented (though not completelyunrealistic) figures -- illustrates that as traffic grows, the overall speed of travelcan increase even if speed-flow effects slow down the speed on everysingle road.

We consider the case of an economy where there are two classes of road(“fast” and “slow”, e.g. motorways and local streets), and two time periods (“peak”and “off-peak”). In the initial state, a total traffic level of 400 vehicle-kms isdivided equally among the four conditions of travel, as shown in Table 1. Thespeeds vary from 20 km/h on the slow roads in the peak, to 120 km/h on thefast roads, off-peak.

Table 1. Initial conditions

Fast roads Slow roadsPeak speed km/h Vehicle-km

60 100

20 100

Off-peak speed km/h Vehicle-km

120 100

30 100

165

Then we allow traffic to grow by 150 per cent, to 1000 vehicle-kms. Thisgrowth -- as is perfectly normal -- does not occur uniformly on all roads and atall times, but in such a way that there is very little growth on the alreadycongested slow roads at peak periods (though there, the effect of this smallgrowth on speeds is substantial), and most growth takes place outside thepeaks, and on the faster roads, with relatively little effect on speeds.

We note, however, that every single category of road suffers some degreeof speed reduction. In other words, everybody travelling both in the before andthe after situation notices that their travel has slowed down, both on the fast andthe slow roads, both in the peak and off-peak.

Table 2. After 150 per cent traffic growth

Fast roads Slow roadsPeak speed km/h Vehicle-km

45 280

15 120

Off-peak speed km/h Vehicle-km

115 400

25 200

Now, let us consider the total amount of time spent travelling, and theoverall average speed, implied by these figures. The results are shown inTable 3.

Table 3. Average travelling conditions

Before AfterTraffic, vehicle-kmsTotal time spent travelling, hrsOverall average speed, km/h

400 10.8 37

1000 25.7 39

It is clear that the total amount of time spent travelling has increased, dueto the increased numbers of vehicles travelling. It is clear that the speed oneach road type has gone down, due to the congestion effect of the speed-flowcurve. However, these two effects are, in the example, disconnected from eachother. The overall average speed of travel has gone up, due to the differentialgrowth on the road types and between peak and off-peak. The question then

166

arises, in what sense has congestion “increased” for the economy as a whole?We have the result that there are more people, all travelling faster, but allobserving their travelling conditions in decline everywhere they go.

In the example, the only form of behavioural adjustment allowed isbetween two road types and two times of day. In the real world, there is amuch wider range of choices that can be made -- many more different roadtypes, and also different modes of transport, frequency of travel, journeypurposes, and so on. It will be noted that the more different choices that areopen to the traveller, the more scope there is for deteriorating conditions ineach cell to be offset by movement among the cells, slowing down the pace ofdeterioration or improving measures of the overall average1.

This is what has been happening in Europe in recent decades.

In summary, statements of the form “congestion costs the economy£15 billion a year”, updated from time to time by inflation, imply an annualdividend of £1000 waiting to be distributed to each family. This is aconvenient, consensual fiction. It is calculated by comparing the time spent intraffic now, with the reduced time that would apply if the same volume oftraffic was all travelling at free flow speed, and then giving all these notionaltime savings the same cash value that we currently apply to the odd minutessaved by transport improvements. This is a pure, internally inconsistent, notionthat can never exist in the real world. (If all traffic travelled at free flow speed,we can be quite certain that there would be more of it, at least part of the timesaved would be spent on further travel, and further changes would be triggeredwhose value is an unexplored quantity.) It is a precise answer to a phantomequation.

We now proceed to make some quantitative estimates of the past andfuture growth of traffic levels, road capacity, and measures of congestion.

3. INDICATORS OF CONGESTION

In this chapter, we will look at some indicators of congestion for variousOECD countries. As discussed in the previous chapter, congestion is bothlocationally and temporally specific: it is determined by the traffic at a givenpoint in time on a particular road stretch. In order to arrive at a measure of

167

congestion on a national level, to examine its development over time or tocompare congestion in different countries, we would need to have detailed dataof the time distribution of volume-capacity ratios on particular roads and somemethod of aggregating them. This is well beyond the scope of this paper. Wecan, however, provide a rough indication of the growth in congestion and acomparison amongst countries by considering some aggregate relationships.

Simplistically, congestion is determined by the interplay of the demandfor road vehicle travel in a given unit of time and the supply of road space. Thedemand for road travel, in terms of vehicle-kms (per unit time), can further bebroken down into the number of cars or vehicles available and their average use(in terms of kms per unit time). We will begin by looking at how some of thesecomponents of congestion differ in various countries and how they havechanged over the past 25 years. The analysis is based on annual data2. Forsimplicity, we consider only cars and vehicle-kms driven by car. Road goodsand public transport vehicles and the traffic relating to them would of courseadd to the figures shown, but should not detract substantially from the validityof the comparison between countries and the development over time.

In order to have a meaningful comparison across counties of differentpopulations and land area, we need to normalise the data with respect to thisvariation. Figure 2 considers car ownership, measured by the number of carsper 1 000 inhabitants in a number of OECD countries for the years 1970, 1980,1990 and 1994, the last year for which data are available.

168

Figure 2. Cars per 1000 inhabitants in OECD countries 1970-94and GDP per capita 1994

0

100

200

300

400

500

600

700

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

Car

s pe

r 10

00 in

habi

tant

s

0

5

10

15

20

25

30

GD

P p

er c

apita

100

0 U

S$

1994

PP

P

1994

1990

1980

1970

GDP per Capita

The wide variation in car ownership amongst countries is apparent – witha range in 1994 of from about 0.31 cars per capita in Denmark to 0.58 cars percapita in the US. The variation in ownership level, however, is decreasing overtime. As expected, we see that car ownership has increased over the entireperiod in all countries. Only two countries – Finland and Sweden – show aslight decline during the 90s (marked by arrows and dotted lines showing 1994ownership levels), perhaps due to declining real incomes in these countriesduring these years. Growth in car ownership, however, has begun to slowdown in most countries, both in absolute and percentage terms. This slowdownin growth is particularly apparent in the highest income countries. The mostobvious exception is Portugal – the poorest of the countries considered – wheregrowth has accelerated.

Also shown in the figure is per capita GDP for the different countriesin 1994 (in US$ Purchasing Power Parity). In general, there seems to be aclose relationship between GDP and car ownership – as would be expected. A

169

few countries, however, have rather lower ownership rates than would beanticipated in consideration of their high income levels, most notably,Denmark, Japan and Norway.

Car use is illustrated in Figure 3, which shows the number of annualvehicle-kms (by car) per capita for the four years3. The overall pattern issimilar to that noted for car ownership. There are, however, a few markeddifferences. Firstly, the variation between countries is somewhat greater than itis for car ownership, with a range of from around 3 000 in Spain and Japan tonearly 10 000 in the US. Secondly, growth has been greater in absolute termsduring the 80s than during the 70s in the majority of countries, while theopposite was the case for car ownership. Again, the relationship to GDP isapparent, but a few countries – Japan, Norway and Austria – have lower car usethan would be expected considering their high incomes.

Thus Japan and Norway have both lower car ownership and lower usemade per car than the average for countries at their income levels. These twocountries are not notably alike in (for example) density of development,history, culture, road provision (as discussed below) or transport policy.

170

Figure 3. Vehicle-kms per inhabitant in OECD countries 1970-94and GDP per capita 1994

0

1

2

3

4

5

6

7

8

9

10

11

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

1000

veh

icle

km

s pe

r in

habi

tan

0

5

10

15

20

25

30

GD

P p

er c

apita

100

0 U

S$

PP

P 1

994

1994

1990

1980

1970GDP per Capita

A very simple measure of traffic density can be obtained by relating carownership and use to the physical size of the country, i.e. its area. This isshown in Figure 4 for cars and Figure 5 for vehicle-kilometres, along withpopulation density (population divided by land area) in the various countries in1994. Both car and kilometre “density” is most obviously related to populationdensity, as would be expected. An interesting exception is Japan, where trafficdensity is comparatively low considering its high population density. This may bedue to the high concentration of the population in densely populated urban areas.

171

Figure 4. Cars per square kilometre 1970-94 and population density 1994

0

20

40

60

80

100

120

140

160

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

Car

s pe

r sq

km

0

50

100

150

200

250

300

350

400

Inha

bita

nts

per

sq k

m

1994

1990

1980

1970

Population Density

Clearly, the most densely populated countries – the Netherlands, Belgiumand the UK – have a far higher average car and traffic density than sparselypopulated countries such as the Scandinavian countries and the US. Suchdensely populated countries will also generally have a greater propensity forcongestion, as road expansion will be limited by land availability and presentgreater problems for the physical environment. Although there is a clearrelationship between car ownership and use and population density, fewspecific conclusions can be drawn concerning congestion from these figures.The primary reason is that this measure of population density assumes auniform distribution of the population over the entire land area. A morerelevant measure would be the proportion of the population living in areas ofhigh population density -- i.e. cities and their surrounding areas. For suchurban areas we might expect car ownership and distance travelled to be lowerthan in more sparsely populated areas, because the need to travel long distancesis less and the availability and convenience of alternative public transport isgreater than in rural areas. Despite this, the demands placed by transport on thelimited land area available in densely populated areas creates the more obviouscongestion problems.

172

Figure 5. Vehicle-kms per square kilometre land area 1970-94and population density 1994

0

1

2

3

4

5

6

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

1000

veh

icle

km

s pe

r km

2 pe

r da

y

0

50

100

150

200

250

300

350

400

Inha

bita

nts

per

sq k

m 1

994

1994

1990

1980

1970

Population Density

It is clear from the figures that the amount of physical space taken over bycars and car travel is clearly growing in all countries, and only appears to belevelling off in a few.

Congestion is not only determined by the demand for road space, but alsoby its interaction with the supply of roads. The next two figures illustrate howthis supply differs amongst the various countries. Figure 6 shows the roadkilometres related to population. A few words of caution must be made beforeinterpreting these results and comparing the different countries. Firstly, themethod of measuring road length may be misleading, as no distinction is madebetween road of different capacity or quality. One kilometre of a small countrylane is treated the same as one kilometre of a six-lane motorway, which canobviously carry far more traffic at far greater speeds. Road expansion whichhas the character of increasing the number of lanes, is thus not captured in thedata. Secondly, the roads included are not the same in all countries.

173

Specifically, the figures for Denmark, Norway and Spain do not include urbanroads, and the exceptional growth noted for Italy since 1970 is an artefactcaused by the inclusion of urban roads in data later than this base year.

Figure 6. Road kilometres per inhabitant 1970-94 and GDP per capitaand population density 1994

0

5

10

15

20

25

30

35

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

Roa

d km

s pe

r th

ousa

nd in

habi

tant

sG

PP

per

cap

ita U

S$

1994

PP

P

0

50

100

150

200

250

300

350

400

Inha

bita

nts

per

sq k

m 1

994

19941970Population DensityGDP per Capita

Even after allowing for these problems of definition, there is aconsiderable, and real, difference in per capita road availability across countries-- from a low of around 7 kilometres per thousand inhabitants in Great Britain,to nearly 30 in the US. There also is a clear inverse relationship between roadsper capita and population density, also shown in the figure, with sparselypopulated countries such as the US and the Scandinavian counties generallyhaving a greater road space per inhabitant than densely populated countriessuch as Britain, Netherlands and Japan. In addition, we find that per capitaroad availability has not increased significantly in most of the countries since1970. In fact, it has declined slightly in a few countries: Austria, Finland,Sweden, Japan and the US (the figure for 1994 is indicated by the dotted lines).

174

Finally, we see that there is a vague relationship between road availabilityand GDP (also shown in the figure). High-income countries such as the US andNorway have high road-population ratios, and a low-income country likePortugal has low road availability. This relationship, however, is certainly notclear-cut, since other high-income countries such as Switzerland and Japanhave relatively low per capita road space.

Road density, in terms of road kilometres per square kilometre land area,is illustrated in Figure 7, along with population density. As expected, we findan obvious relationship between road density and population density. A fewcountries, however, do have rather lower road densities than the average fortheir population density, most significantly, Britain and the Netherlands, and toa lesser degree, Germany and Japan. We see, too, that the amount of land areagiven over to roads continues to increase in most countries, and particularly soin the more densely populated countries, although the increase noted for Italylargely reflects the definitional change noted earlier. It should also be stressedthat the growth in road kilometres illustrated in the figure most likelyunderestimates the actual growth in road capacity. The reason for this is thatincreases in the number of lanes per road are not accounted for, and much roadexpansion during the past decade has been to increase lane kilometres ratherthan the road kilometres as measured in our data.

175

Figure 7. Road kilometres per square kilometre land area 1970-94and population density 1994

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

GB

BELGIU

M

DENMARK

FRANCE

GERMANYITALY NL

PORTUGALSPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWITZERLAND

JAPAN

USA

Roa

d km

s pe

r sq

km

0

50

100

150

200

250

300

350

400

Inha

bita

nts

per s

q km

1 9 9 4

1 9 9 01 9 8 0

1 9 7 0P o p ula tio n D e ns i ty

Vehicles per road kilometre as an index of congestion

For reasons discussed above, any simple index of congestion will sufferfrom problems, especially when applied over the whole network, since an indexby definition cannot capture the transient, site-specific, demand-interactingnature of congestion, and in practice there will always be inadequacies ofdefinition.

Nevertheless, the discussion indicates that the most useful general formwill have the character of a ratio of an amount of traffic and the capacity of anetwork. This must always be treated with great caution, but it seemsdefensible to argue that a simple ratio of this type might be used as a roughindicator at least of changes in congestion levels.

Therefore the following sections take together the demand for road spaceas measured by the number of vehicles, and of the kilometres driven, and thesupply of road space measured by road kilometres. These allow us to make ageneral comparison of the different countries as well as to examine changes

176

over time in each of the countries. (We emphasize again, as shown above, thatan increase in such a ratio indicates a potential or likelihood of lower speeds,but does not of itself determine that, since the demand responses will becrucial.)

The following two figures show the number of cars and vehicle-kms perroad kilometre, along with GDP and population density. The number of carsper road kilometre for the various countries is given in Figure 8 for 1970 to1994. A considerable variation amongst countries is apparent – from about19 cars per road kilometre in Norway to nearly 80 in Spain. Neither of thesecountries data includes urban roads, however, so they are not strictlycomparable with most of the other countries. This definitional differencewould mean an overestimation for both these countries in comparison to theothers. If we ignore Spain, the variation between the Germany, next highestcountry where urban roads are included, and Norway, had urban roads beenincluded, is greater than that depicted in the figure. The highest car-per-roaddensities are found in Germany, Great Britain and the Netherlands – allrelatively densely populated countries – and the lowest in the US and northernScandinavian countries – which have a low population density. However, therelationship with population density does not hold for all countries. There arevast differences between countries with similar population density,e.g. between France and Portugal, and between Germany and Italy. As we haveseen in the earlier figures, although per capita car ownership in Portugal isabout 15 per cent lower than it is in France, it has only about ½ the total roadkilometres. Similarly, Italy has 6 per cent more cars per capita than Germany,but a road length twice as long.

177

Figure 8. Cars per road kilometre 1970-94 and GDP per capita andpopulation density 1994

0

10

20

30

40

50

60

70

80

90

GB

BELGIU

M

DENMARK

FRANCE

GERMANY

ITALY NL

PORTUGAL

SPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWIT

ZERLAND

JAPAN

USA

Car

s pe

r ro

ad k

m

0

100

200

300

400

500

600

GD

P p

er c

apita

per

wee

k U

S$

PP

P 1

994

Inha

bita

nts

per

sq k

m

1994199019801970GD P per C apitaPopulation D ensity

It is also apparent in the figure that the number of cars per road kilometreis generally increasing over time in most countries, so that road building hasnot kept up with increasing car ownership. On average, congestion wouldexpect to be increasing. The 90s, however, have seen a slight reduction in thecar-road ratio in a few countries (marked with arrows and dotted lines) – inSweden and Finland explained by the reduction in car ownership due to fallingincomes. For Germany, the figures may be misleading since the data are notconsistent before and after reunification. Finally, the figure indicates that thereis no relationship between the car-road ratio and GDP, at least betweencountries.

Congestion, of course, is not determined directly by the cars available, butby their use. A more fitting measure of aggregate congestion is thus given bythe number of vehicle-kms actually driven compared to available road space.This is illustrated in Figure 9, which shows the number of car-kms per roadkilometre per average day. The pattern mirrors that seen for cars -- GreatBritain, the Netherlands and Germany have the most densely travelled roads,while Norway has the least densely travelled. There are, however, a number ofsignificant differences in the rank order of the various countries.

178

Specifically, Great Britain, Sweden, Finland and Denmark show relativelyhigh vehicle-kms per road in comparison to their car-road ratios, suggesting thatinhabitants in these countries use their cars to travel greater distances than is thecase in the other countries. From the data we estimate an average number ofkilometres per car of between 16 and 20 thousand for these four countries,while for all others the values are around or below 14 thousand. It isconceivable that the somewhat similar distributions of population in the threeScandinavian countries could explain their higher figure, but this does notexplain the GB figure, which remains behaviourally (and politically)problematic.

Finally, we see that the number of vehicle-kms per road kilometre isincreasing over time in all countries and at a faster rate than car ownership.Cars are being used more intensively, i.e. for more trips and over greaterdistances. Again, we see that road supply has not kept pace with the demandfor road space, and that this problem is not becoming worse, rather than better.

Figure 9. Vehicle-kms per road kilometre per day 1970-94and GDP per capita and population density 1994

0

0.5

1

1.5

2

2.5

3

GB

BELGIU

M

DENMARK

FRANCE

GERMANYITALY NL

PORTUGALSPAIN

AUSTRIA

FINLA

ND

NORWAY

SWEDEN

SWITZERLAND

JAPAN

USA

1000

Veh

icle

km

s pe

r roa

d km

per

day

0

100

200

300

400

500

600

GD

P pe

r cap

ita p

er w

eek

US$

PPP

199

4In

habi

tant

s pe

r sq

km

199 4199 0198 0197 0GD P per C ap itaP o pula tio n D ensity

179

Using average daily vehicle-kms per road kilometre as a proxy measure ofcongestion, Figure 10 illustrates the development of congestion over the past25 years in a number of European countries. Not all countries are shown inorder to make the graph more legible. As earlier, we see a clear ranking incongestion levels, with Great Britain being the most congested and Norway theleast. The same general pattern holds for the entire time period. The mostmarked exception is the relative development in Great Britain and theNetherlands. During the 70s both countries had nearly identical vehicle-km toroad kilometre ratios. Subsequently, and particularly in the late 80s, the ratioincreased far more rapidly in Great Britain, so that by 1994 it was nearly 20 percent higher than in the Netherlands. As shown in the figures presented earlier,it appears that it is chiefly the result of a far lower increase in road capacity inBritain. From Figure 7 we see that road kilometres rose by nearly 40 per cent inthe Netherlands since 1970, but by less than half this in Britain, while thenumber of vehicle-kms doubled in both countries (Figure 5).

Figure 10. Vehicle-kms per road kilometre and day 1970-94

0

0.5

1

1.5

2

2.5

3

1970 1975 1980 1985 1990 1995 2000

Veh

icle

km

s pe

r ro

ad k

m p

er d

ay

GB

NL

Germany

Switzerland

Portugal

Denmark

France

Norway

The relative growth rates of this measure of congestion in the individualcountries is illustrated more clearly in Figure 11, which shows a “congestionindex” with 1970 set equal to 1 for each country. We see that for the period as awhole, “congestion” has more than doubled in Finland and Norway, both ofwhich have low absolute levels, while it has increased by only 50 per cent inDenmark. France, Britain and Germany have seen similar growth rates ofnearly 100 per cent while a slower growth of around 80 per cent is noted for the

180

Netherlands and Portugal. Despite the fact that these latter two countries haveincreased road kilometres by 40 per cent and 50 per cent respectively over theperiod, they were still unable to keep pace with increasing car use.

This is perhaps the most important conclusion of this analysis: the ratiovehicle-kms per road kilometre have increased, fairly consistently, throughoutEurope, for at least a quarter of a century. The reason for this is primarily theincrease in the volume of traffic, which increases at a pace which no realisticexpansion of the road network can match.

Figure 11. Congestion index (vehicle kms per road km per day) 1970 = 1

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

1970 1975 1980 1985 1990 1995 2000

onge

stio

n in

dex(

vkm

s/ro

ad k

m p

er d

ay)1

970=

1

Norway

Denmark

Portugal

NL

FranceGB

Germany

Finland

4. INFLUENCES ON TRAFFIC GROWTH

In the previous chapter we have examined some of the components ofcongestion, and have seen that some of the differences between countries couldbe related to variation in factors such as income and population density. In thischapter, we will explore in a more quantitative fashion the impact of incomeand transport costs on car ownership and use. The discussion will be basedlargely on evidence obtained from econometric models, and will drawprincipally on work carried out at TSU. Some of the studies have an

181

international character in that they are based on data for many countries, whileothers pertain solely to the UK.

4.1. The effect of income on car ownership

The relationship between per capita car ownership and per capita income isillustrated in Figure 12. The figure shows car ownership on the vertical axisand real GDP on the horizontal axis, both in per capita terms, for the countrieslisted for the years 1970, 1980 and 1994. Both variables are expressed aslogarithms. Graphed in this way, the slope of the line best fitting theobservations (the dotted line) can be taken as a measure of the income elasticity.For reference, a line representing the slope of an income elasticity equal to oneis also indicated in the diagram (the solid line). Clearly, for the data shown, theslope of the line, and hence the income elasticity, is in excess of 1. Of course, thissame relationship need not hold for every country or for all years into the future,and taking account of other influences, including prices, will modify this.

Figure 12. The relationship between car ownership and income

3.5

4

4.5

5

5.5

6

6.5

1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8

GDP per capita, log scale

Car

s pe

r ca

pita

, log

sca

le

Austria

Germany

Denmark

Spain

Finland

France

GB

Italy

NL

Norway

Portugal

Sweden

1970

1980

1994

The effects of income on car and vehicle ownership in an internationalperspective are investigated in a recent study by Dargay and Gately (1997). The

182

results and discussion in this chapter are largely based on that paper. Theestimates of car ownership are based on an econometric model relating percapita car ownership to per capita income. It is assumed that the long-runrelationship between per capita car ownership and per capita income can berepresented by a Gompertz curve. The Gompertz function is an s-shaped curve,which allows a slow growth in demand at the lowest income levels, followed byan increasing rate of growth as income rises which finally declines as saturationin car ownership is approached. The model is estimated on the basis oftime-series data for a sample of 26 OECD and non-OECD countries, in mostcases covering the time period 1973 to 1992. For the empirical implementation,the Gompertz function is set within a framework of a partial adjustment modelso that lags in the adjustment of the car stock to income changes could beaccounted for. In this way, both short and long-run income elasticities can beestimated, as well as the time required for adjustment to equilibrium. Suchstatistical calculations provide a better estimate of the income elasticity than canbe gleaned from diagrams as that above.

Figure 13. Estimated long-run car ownership functions.Dargay-Gately (1997)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0 5 10 15 20 25 30

GDP per capita 89 US$

Car

s pe

r ca

pita

Italy

Portugal

Austria

France

Finland

Norway

Sweden

NL

Germany

GB

Spain

Denmark

183

The estimated long-run relationship between car ownership and per-capitaincome is illustrated in Figure 13. The estimated saturation level of 0.62 carsper capita4 – common for all countries – is apparent. Saturation is reached atdifferent income levels for different countries. Car ownership saturation isreached at the lowest income level in Italy and at the highest income level inDenmark: at under $20 000 (in 85 US$) in Italy and at over $30 000 inDenmark. For the countries shown, saturation is reached on average at about$25 000.

Because of the nature of the functional form used in the estimation, theincome elasticities are not constant for all countries or over time. Instead, theelasticity varies for the individual countries and with income level, and henceindirectly with vehicle ownership itself. The resulting long-run5 elasticityestimates are shown in Figure 14, based on the projected per capita GDP growthfor the individual countries. In 1992, all countries, with the exception ofGermany, France, Italy, Norway and Britain, had elasticities in excess of 1. Thehighest elasticity – approaching 2 – is noted for the lowest income country,Portugal. We see that the elasticity is projected to decline in all countries, asincome and car ownership grow and saturation is approached. By 2005, theelasticity is projected to be reduced to well under 0.6 in all countries, andby 2015 to fall further to below 0.4. It should be noted that these elasticities arebased on the assumed income growth (an average real per capita GDP growth of2.4 per cent per annum, ranging from 1.9 per cent in Sweden to 3.5 per cent inPortugal) and resulting car ownership levels. Lower income growth orreductions in car ownership levels, resulting for example, from cost increases,would reduce the rate of decline of the elasticities.

184

Figure 14. Estimates of the long-run income elasticity

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Austria Germany Denmark Spain Finland France GB Italy NL Norway Portugal Sweden

GD

P e

last

icity

1992

2005

2015

4.2. The effects of costs

Although, historically, income growth has been the major driving force toincreasing car ownership, other factors are also of relevance, and may be ofgreater importance in the future, particularly if price-related policy measureswere adopted to order to control traffic growth. In most European countriesmotoring costs have remained comparatively low in real terms over the pastdecades. The only substantial increases were a result of the petrol price risesprecipitated by the oil price shocks of the seventies, but the price collapseof 1986, along with general inflation, reduced prices in real terms to thoseprevailing in the early seventies in most European countries. Although nominalpetrol prices have increased substantially over the past decade – and particularlyover the past few years -- real prices still remain at comparatively low levelsthroughout Europe. Although total car running costs have increased somewhatin many countries during the nineties, car purchase costs have generally fallen,so that total motoring costs have remained more-or-less constant. This has beencombined with rising public transport fares in a number of countries, so that therelative price development has strengthened the economic advantage of privatemotoring in comparison to public transport6. Given this price development, it isnot surprising that car ownership and use have been largely determined byincome growth.

185

Despite this long-term trend, petrol prices and motoring costs havefluctuated significantly over the past decades and there is a good deal ofempirical evidence regarding the effects of costs on car ownership and use.Although there have been numerous attempts to estimate the price elasticity ofcar travel, there is a wide degree of variation in the answers. A summary ofsome of these is presented below.

In the UK, particularly important studies were carried out by Tanner, of thethen Road Research Laboratory, who for many years was the main personresponsible for the UK forecasting practice (until his forecasting method wasabandoned due to inaccuracies which do not derive primarily from itselasticities. Mackett (1998) currently notes that his forecasts for the presentperiod have actually turned out to be more accurate than subsequent revisions.)Considerable attention was given to consistency between price elasticities andthe generalised cost elasticities within which they sat. Tanner (1974) gaveelasticities of car use with respect to generalised cost in the range - 0.9 to - 1.3.He took - 1.1 as a preferred value and corresponding elasticities with respect toall motoring costs of - 0.67, and fuel price only of - 0.17. At that time thedistinction between short and long run effects was not made. Later(Tanner, 1977) he confronted the very important problem of interaction betweenincome elasticities and price elasticities, such that if too much of the growth incar ownership and use were attributed to income growth, one would inevitablyunderestimate the effect of price or generalised cost effects. After consideringalternative ways of assessing the balance between these effects, he suggested a“middle value” of - 0.6 (range - 0.4 to - 0.8) for the elasticity of car ownershipwith respect to money cost (the report is not quite clear, but it seems to relate tototal motoring cost). Tanner (1981) estimated a car ownership elasticity withrespect to car purchase cost of - 0.87, and with respect to fuel price of - 0.31;also an elasticity of kilometres per car with respect to fuel price of - 0.26.He interpreted these as long term elasticities, and used other evidence to suggesta short-term elasticity of around - 0.15. He proposed a dynamic adjustmentprocess such that the long term effect on car use was achieved over aboutfour years (with the first year effect being about 40 per cent of the total), andabout ten years for car ownership (first year less than 20 per cent of the total).

Subsequent work along the same lines (Tanner, 1983) attempted an explicitlag estimation procedure, with coefficients that were not very robust, due toproblems of data. Tanner retired at about this time, and the line of work seemsto have been abandoned in UK official practice, until Virley (1993), who used asimilar approach for elasticities of motor fuel consumption with respect to price,using an explicit lagged model. The results were a short run elasticity of petrolconsumption with respect to price of - 0.09, and a long run elasticity of - 0.46,with about 20 per cent of the full adjustment in the first year.

186

This work may be seen in the context of several reviews of existingevidence which were revisited in the early 1990s. The evidence consists of twoseparate strands of work -- effects of fuel price on fuel consumption, and effectson traffic levels. There are well over 100 separate empirical studies, calculatingdemand elasticities from time series data, which are analysed with dynamicmodels incorporating an explicit lag structure. Goodwin (1992) cited 13 studiesin which the effect of fuel price on fuel consumption had been calculated,(giving a short-term elasticity of around - 0.25 to - 0.3 and a long-termelasticity of - 0.7 to - 0.8); eleven studies in which the effect of fuel price ontraffic-levels had been calculated (with results of - 0.16 for a short-term effectand about - 0.3 for a long-term effect). Sterner et al. (1992) estimated gasolinedemand elasticities for 21 countries separately. The mean result for thepreferred model form gave a short-run elasticity of - 0.24 and a long-runelasticity of - 0.79. A review of Australian evidence by Luk and Hepburn(1993) cited 28 studies, and came to the conclusion that the elasticity of trafficlevels with respect to fuel costs was - 0.1 in the short run and - 0.26 in the longrun.

Table 4 summarises the literature review carried out by Goodwin (1992)indicating that although petrol demand responds quite strongly to changes inpetrol prices, car traffic is relatively insensitive to the changes. In general,longer-term elasticities of demand with respect to travel cost are different, andgreater, than shorter-term responses. This is consistent with, and reinforces, theidea that behaviour is more constrained in the short run than in the longer run.This table is based on the results of twenty-four empirical studies, using avariety of different methods, not all consistent with each other.

Table 4. Summary of evidence on elasticity of demandwith respect to petrol price

Short run Long runTIME SERIES Petrol consumption -0.27 -0.71 Traffic -0.16 -0.33CROSS-SECTION Petrol consumption -0.28 -0.84 Traffic n.a. -0.29

187

A more recent study carried out by Dargay and Gately (1997a) investigatesthe question of the reversibility of the demand for transport fuels to upward anddownward price changes, based on data for eleven OECD countries/regions forthe 1961-1990 time period. The resulting elasticities are shown in Table 5. Thestatistical evidence suggests that demand has responded more strongly to certainprice rises than to others, particularly, to the price rises of the seventies, whichwere sudden and large. The impact of falling prices and subsequent pricerecoveries has been far smaller. The study also supports the hypothesis ofhysteresis: when prices rise above some previous maximum level, the long-rundemand relationship itself changes, so that subsequent price declines will nottotally undo the demand reductions caused by the initial price rise. Theinterpretation of these results is that much of the demand savings are explainedby improvements in vehicle technology and government policies concerningfuel-efficiency standards, which themselves were “induced” by higher prices.Although more fuel efficient vehicles in combination with cheaper fuel havelowered the per-kilometre cost of road transport, neither greater vehicle use nora return to larger, less fuel-efficient vehicles has fully reversed the demandreductions caused by the price increases. The study also suggests that demandmodels that do not take into account such asymmetries will be misspecified,leading to biased elasticity estimates. This should be held in mind wheninterpreting results such as those presented in Table 4 above.

Table 5. Price and income elasticities for motor fuelbased on an irreversible model

Short run Long runPrice elasticity Large price rises -0.18 -0.60 Small price rises -0.04 -0.13 Price cuts -0.04 -0.13Income elasticity 0.34 1.13

Of course, fuel prices are only one component of motoring costs – aboutone half of running costs and a third of total motoring costs (including carpurchase costs)7. In addition, fuel demand is not necessarily a goodapproximation to car use, since the fuel-efficiency of vehicles variesconsiderably and has not remained constant in the past, nor can it be expected toin the future.

Table 6. Estimated elasticities for kilometres travelled per capita

Trafficcar kms

per capita

Car ownershipCars per capita

Car useKms per car

Trafficcar kms

per capitaSR LR SR LR SR LR SR LR

Cost elasticity Car purchase costs -0.11 -0.20 -0.06 -0.38 - - -0.06 -0.38 Car running costs -0.46 -0.86 -0.10 -0.63 -0.27 -0.65 -0.37 -1.28

Public transport costs 0.37* 0.69* 0.11 0.69* - - 0.11 0.69*Income elasticity 0.45 0.84 0.13 0.81 0.09 0.22 0.22 1.03Adjustment coefficient 0.53 0.16 0.41

Note: The cross-elasticity of car demand with respect to bus fares is the right sign, but quite implausible in size, and it is likely that itis influenced by the omission of public transport service levels, which are correlated with price. Thus it may be morereasonable to interpret these cross-elasticities as relating to public transport generalised cost, rather than to fares as stated.(The same does not apply to the direct elasticities, since traffic speeds have been more closely correlated with income thanwith car costs.)

189

Because of this, increases in fuel prices will have a smaller impact on caruse and traffic levels than they will have on petrol consumption, i.e. theelasticity of traffic with respect to fuel prices will be smaller than the elasticityof petrol demand, as was suggested in Table 4. For this reason, it is moreappropriate to consider the demand for car travel directly.

The results of estimating such a model for the UK are shown in the firsttwo columns of Table 6. The model used is a fairly simple log-linearrelationship between per capita passenger kilometres by car and busrespectively, and explanatory factors such as income per capita, car purchaseand running costs and bus fares. Allowance is made for delayed responses tochanges by including a lagged value of the dependent variable. The model isestimated on the basis of annual data for the 1970 to 1993 period.

The results in the table are the long- and short-run price, cross-price andincome elasticities. We see that car use -- in terms of total car-kms perinhabitant -- is sensitive to both car purchase and running costs as well as to busfares. The long-run elasticity with respect to running costs, - 0.8, is muchgreater than with respect to purchase costs, - 0.2, and taken together indicate anelasticity with respect to total motoring costs of about – 1.0. The adjustmentcoefficient indicates that 53 per cent of total adjustment occurs within one year,and 90 per cent within three years. The estimated income elasticity for cartravel is 0.45 in the short run and 0.84 in the long run.

More insight into the actual mechanism behind the response of trafficlevels to travel costs can be obtained by separately analysing its components-- car ownership and use per car. From such estimates, the demand for cartravel in terms of vehicle-kilometres -- or traffic – per capita can be derived.The results of such a study, reported in Dargay and Goodwin (1994), also forthe UK, are summarised in the right-hand section of the table.

Here we see that car running costs influence traffic levels though theirimpact on both car ownership and car use, but in a rather different fashion. Inthe short run, changes in running costs affect car use much more than they docar ownership, whereas in the long run the impacts are more or less identical:half occurring through changes in use per car and half through changes in thenumber of cars. In all cases, the long-run elasticities are rather large, over - 0.6for both car ownership and car use, and - 1.3 for total car-kms. It should bepointed out, however, that our definition of running costs includes much morethan petrol prices, so we would expect the elasticity to be higher than thoseobtained for petrol prices alone. On average over the period petrol costsaccounted for around 50 per cent of total running costs. If the elasticities werethe same for petrol prices as for other running costs our results would imply along-run elasticity of car traffic to petrol prices of around - 0.6. Finally, the

190

effects of the prices of cars and public transport fares on traffic levels are rathersmaller than those of running costs and arise solely through their impact on carownership.

As in the previous analysis, the results concerning the effects of publictransport prices seem to be well outside the received wisdom. One clue is thatthe large figure only appears in the long run: if the money costs of publictransport are correlated to the quality of service, so that the elasticity is pickingup both effects, then the suggestion is that this could in the long run be a majorinfluence on car ownership and therefore on traffic levels, but more work needsto be done before relying on this relationship.

The estimated income elasticity for car traffic is about 0.2 in the short runand unity in the long run. In the short run, income affects both car ownershipand car use to a similar degree, so that both contribute about the sameproportion to changes in traffic. In the long run, however, the influence ofincome is primarily through car ownership, as one would expect.

A major difference between the estimates for car ownership and car useconcerns the speed of adjustment. Car ownership responds slowly to changes inincome and costs: the adjustment coefficient of 0.16 implies that only 16 percent of the total adjustment occurs within one year and that it takes aboutaround 13 years before most of the impact (90 per cent) is realised. On theother hand, car use responds comparatively quickly to changes in income andcosts: 41 per cent of the total impact occurs within one year and 90 per cent in4.4 years. Since car traffic is obtained as the product of car ownership and caruse, the adjustment process it follows depends on the combined effect of thefast car use response and the slow car ownership response, and its adjustmentpath lies somewhere between the two.

In general, the elasticities obtained from the “decomposed” model aresomewhat greater than those obtained from the single equation model and theadjustment coefficient is somewhat greater. They are, however, of similarorders of magnitude and display the same general pattern.

Using an entirely different method of analysis of cohorts treated as thoughthey were members of a panel (of interest because it takes into account thechanges in travel behaviour related to age and generation), Dargay (1997) foundelasticities of demand for car ownership with respect to the total purchase costsof cars of -0.35 for France, and -0.33 for the UK, and elasticities with respect torunning costs of -0.22 for France, and -0.51 for the UK. These were long termeffects, with estimated speed of response such that just over a third of the fulllong-term response took place in the first year, the full response being (nearly)completed in about ten years.

191

It is clear from the results to the empirical studies that both income andprice changes can have an important influence on car use -- and on traffic andcongestion. Concerning income effects, we can postulate -- on the evidenceprovided in Figure 14 -- that the income elasticity is currently less than one andwill decline as income rises and saturation is approached. Concerning the priceeffects, we can conclude from Table 6 that the short-run sensitivity to changesin total motoring costs is around – 0.5, increasing to – 1.0 or more in the longrun. The empirical evidence further suggests that the behavioural response tochanges in travel costs is far from instantaneous, and in general builds up overtime, so that some ten years might need to elapse before the full effects onbehaviour, traffic and congestion can be seen.

Finally, we note that the provision of road capacity itself, by initiallyincreasing speeds, generates or induces some additional traffic which in turn tosome extent reduces the speed again. SACTRA (1994) assessed a wide rangeof evidence on the effect of increasing road capacity on the total volume oftraffic, and ECMT (1998) collects evidence from other European countriesshowing that this is a widespread phenomenon. While the effect may not beperfectly symmetrical, it is approximately so: Cairns et al. (1998) carried outthe corresponding study on the effects of reducing capacity. The results in bothcases were that traffic levels are sensitive to the changes in speed or reliabilitybrought about by changes in capacity -- to an extent greater than had beenassumed in prevailing advice. In both studies, a wide range of results werefound, with increases or reductions in traffic levels which could be up to 60 percent or so, and with mean levels of the order of magnitude of 20 per cent,depending on circumstances and conditions. Both studies concluded that (aftertaking account of the fact that general trends for traffic growth would tend tooverestimate the traffic growth due to capacity increases, and underestimate thetraffic reduction due to capacity decreases) the elasticity effect itself tended togrow over time.

5. CONGESTION SCENARIOS

In this chapter, we will look at the possible development in congestion tothe year 2015. The definition of congestion is the simplistic one used earlier-- total car-kms over total road kilometres. The projections are for an“idealised” European country, having the average characteristics of those

Table 7. Assumptions used in the simulations

High LowGDP growth 2.4% per year1 53% to 2015 1.5% per year 16% to 2015Kilometres per car 12 700 per year2 - - -Population growth 0.13% per year3 2% to 2015 - -Road expansion 0.4% per year4 7% to 2015 None 0Motoring costs 1% per year 20% to 2015 2% per year 40% to 2015Cost elasticity Short run -0.4 Long run -1.0 Short run -0.2 Long run –0.5Income elasticity Average estimates for European countries based on Gompertz modelInduced traffic elasticity Short run 0.04 Long run 0.1 - -

1. GDP growth is an average of projections from the World Bank for the period up until 2005, which are assumed tobe the same to 2015.

2. Estimated as the average for the European countries over the past 5 years.3. Average projections for the European countries given by the UN.4. Average growth over the past ten years.

193

countries described in the previous chapter (Figure 12). The projections arebased on assumptions concerning population and real income (GDP) growth,road expansion and motoring costs. The projections are obtained by a dynamicsimulation procedure. The income effects are taken into account through theGompertz model for car ownership described in the previous chapter. Traffic isthen calculated on the basis of these predictions, assuming a constant pervehicle use. The impacts of increases in motoring costs on traffic are thencalculated on the basis of assumed cost elasticities, as is induced trafficresulting from road expansion. The congestion index is then derived as totalvehicle-kms per road kilometre. The assumptions used for the simulations aresummarised in Table 7.

The resulting projections are shown in Figure 15, normalising congestionto 1.0 in 1997. We find a considerable difference in the growth in congestion inthe various scenarios – from an increase of around 27 per cent to a reduction of7 per cent, depending on the assumptions made. In the Base Case (A), weassume only a real income growth of 2.4 per cent per year, with no roadexpansion or cost increases. In this “do nothing” scenario, the congestion indexwill increase by 27 per cent. Assuming a lower income growth (B) would leadto only a slightly lower growth in the index, about 24 per cent.

Figure 15. Projections of the growth in congestion to 2015

0.90

0.95

1.00

1.05

1.10

1.15

1.20

1.25

1.30

1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017 2019 2021 2023

Con

gest

ion

Inde

x

A: GDP: 2.4%

B: GDP: 1.5%

C: Costs: 1% e: -0.5

D: Roads: 0.4%

H: Costs: 2% e: -1.0

E: Costs: 1% e: -1.0

F: Costs: 1% e: -0.5 Roads: 0.4%

G: Costs: 1% e: -1.0 Roads: 0.4%

I: Costs: 2% e: -1.0 Roads: 0.4%

194

The remaining cases assume policy intervention, either in the form of roadexpansion or price-related measures. They are all based on the Base Caseincome growth of 2.4 per cent per year.

The next two cases show the effects of (C) an increase in total motoringcosts of 1 per cent in real terms per year, or 20 per cent over the entire periodand a low price elasticity of – 0.5 in the long run and (D) road expansionat 0.4 per cent per year, or 7 per cent over the period. We find that both of thesemeasures would have a similar effect on congestion, given a price elasticityof - 0.5. Congestion would still increase by 19 per cent from its present levelsby 2015, but would be 30 per cent lower than in the Base Case (A). Case Eshows the effects of the same price increase as in (C), but assuming a higherelasticity, - 1.0. Even with this higher elasticity, the increase in congestionwould still be around 12 per cent, but 60 per cent lower than in the Base Case.A road expansion programme of 0.4 per cent per year combined the same pricerise would result in a similar reduction in the rate of congestion growth, even ifthe cost elasticity were only - 0.5 (F). Case G shows the combined effect of thesame road expansion and cost increase with an elasticity of – 1.0. Assumingthis higher elasticity would reduce the growth of congestion by half comparedto Case F, that is to about 6 per cent over the period.

The final two cases assume a much higher cost increase: 2 per cent perannum in real terms, or 40 per cent by 2015. Both cases are based on the “high”cost elasticity of – 1.0. Case H assumes no road expansion, while (I) assumesthe growth rate of 0.4 per cent per year used earlier. We see that, given the“high” elasticity, the cost increase alone would be sufficient to maintaincongestion at today’s levels. Combined with road expansion (I), congestioncould be reduced by 7 per cent. The same price rise with a “low” elasticitywould result in a growth in congestion similar to (E) without road expansionand to (G) with road expansion.

Note that a “1 per cent per year increase in total car costs” is equivalent toabout 3 per cent per year in fuel costs if these are the only instrument used.

6. CONCLUSIONS

The two main influences on the spread of congestion are (a) increases inthe volume of traffic, and (b) increases in capacity of the road network. In thesimplest case, we can say that if traffic grows faster than capacity, thencongestion must increase. (At a detailed local level this may not always be true,depending on a careful definition of “congestion”, which is not only identified

195

by the heaviest jams, and “capacity”, which is influenced by road building butalso by more efficient forms of traffic management, information and driverbehaviour.)

Analyses of national and Europe-wide statistics show that in recentdecades traffic has almost always grown faster than capacity, both consideringthe road network as a whole and also when considering specific parts of thenetwork, e.g. long-distance motorways. Therefore it is not surprising thatcongestion has spread. However, this certainly has not meant that traffic speedsevery year get slower and slower: it is obvious that the possibilities for fastlong distance movement now are substantially greater than, for example,fifty years ago.

Closer examination shows that the spread of congestion is often measurednot by its intensity, but by its occurrence in space and time. Thus in urbanareas, what was once a “peak-hour” is often now a sustained peak period formuch of the working day. Similarly in rural areas which were once relativelyfree of traffic, now there is often a higher level of flow, with particular problemsin areas of tourist attraction, or en route between two cities. A characteristicfeature of this process is differential growth. Notably, those times and placeswhere congestion is most intense show the slowest growth rates. The greatestgrowth occurs where and when there is room for it to do so -- subject tounderlying economic conditions. The differential growth is also connected withthe tendency for different countries or regions to experience different phases offast or slow economic growth -- resulting at the present time in very rapidincreases in congestion in Eastern European countries.

The dominant trends revealed in most forecasting models suggest that thisprocess is likely to continue: traffic will continue to grow faster than capacity,and congestion will continue to spread, growing faster in the middle of the day,the evening, weekends, suburban and rural areas, and on parts of the roadnetwork that are expanded.

However, it is also important to take into account some “counter-trend”developments. Town centres, which have traditionally been the location of themost intense congestion, are also the places where traffic restraint is mostpopular and most successful. It is often argued that long-distance commutingmay be influenced by substitutes such as telecommunications. So there is somepossibility we may see (to some degree) a “spread of de-congestion”.

196

Consideration is given to two distinctly different sort of policy instruments-- increasing the cost of car use, which has the effect of reducing traffic levels ortheir growth, and increasing the capacity of the road network. Using availableinformation on demand and supply effects at the level of a notional “averageEuropean country”, we estimate:

a) If car costs are not increased, continuation of the scale and nature ofroad building typical of recent years can only slow down the increasein the congestion index, which would rise by 20 per cent by 2015.

b) If little or no substantial expansion of road capacity is carried out, a1 per cent to 2 per cent annual increase in total motoring costs(depending on whether the elasticity is - 1 or - 0.5) would also slowdown the congestion index, though it would still rise to some extent.

c) To reduce the congestion index without road building would requirean increase in total motoring costs of 2 per cent per year (on anelasticity of - 1). Continuation of recent road building in addition tosuch an increase in costs would lead to a reduction in the congestionindex.

d) These conclusions do not take into account other policies such asimprovement in public transport services, traffic restraint using meansother than motoring costs, or changes in the actual incidence ofcongestion brought about by spontaneous or encouraged managementof demand away from the most congested places and times.

197

NOTES

1. In the example, we do not distinguish between behavioural adjustmentwhich is the spontaneous result of changing social circumstances(e.g. suburbanisation), and adjustment which is the result of a demandresponse to the changing conditions (e.g. a cross-elasticity effect from peakto off-peak travel as peak conditions worsen). This distinction makes nodifference to the arithmetic of the example, but does make an importantdifference to the economic assessment of welfare changes, since there issome loss in welfare due to travelling at a “second-best” time of day, etc.We have argued that the welfare implications of such a search forimproved conditions may only be rigorously assessed using a dynamicdemand framework, rather than by comparing equilibrium conditions(Dargay and Goodwin, 1995), because the adjustment takes time. It seemslikely that a proper assessment of the “cost of congestion” which tookaccount of the welfare implications would similarly require a dynamicformulation, but that has not yet been demonstrated. Meanwhile, we notethat if (and only if) it is possible to identify the external and internaldemand responses (and distinguish between them with confidence), thenan equilibrium assessment of the effect of changing congestion onconsumer surplus may be calculated with the aid of an obsessive attentionto the rule of a half.

2. The data are taken from UK Department of Transport: InternationalComparisons of Transport Statistics 1970-94, ECMT: Statistical Trends inTransport 1965-1992 and UN: Annual Bulletin of Transport Statistics forEurope and North America 1996.

3. Definition of vehicle-kms pertains to cars and taxis, and for all countrieswith the exception of GB, Germany, NL, Spain, Finland, Switzerland,Japan and the US includes only national vehicles.

4. Dargay and Gately also estimate a saturation level of 0.85 total roadvehicles per capita. As expected, the saturation rate is higher for totalvehicles than for cars and suggests a saturation of “other vehicles” of0.23 per capita.

198

5. The estimates indicate that about 10 per cent of the total response of carownership to income changes occurs within one year, so the short-runelasticities are about one-tenth of the long-run values.

6. In the UK, total motoring costs declined by about 7 per cent since 1980,while public transport fares increased on average by 10 per cent. ForFrance, the comparable figures are an increase of motoring costs by 7 percent and an increase in public transport fares by 14 per cent.

7. On certain assumptions about rational behaviour, elasticities will bebroadly in the same proportion to each other as they are proportions oftotal cost -- i.e. if the fuel elasticity is -0.3 and is one-third of total cost, thetotal cost elasticity will be about -0.9. Similarly, using a generalised costframework, if travel time is twice as large a component of generalised costas money, then the elasticity will also be twice as great. Such identities aremore likely to be useful when using long-run elasticities than short run.

199

BIBLIOGRAPHY

Brindle, R. (1996), Putting the car in its place -- a historical perspective,Brisbane City Council and Queensland University of Technology,Brisbane.

Dargay, J. (1997), “Modelling car ownership in France and the UK:a pseudo-panel approach”, TSU Working paper.

Dargay, J. and D. Gately (1997a), “The demand for transportation fuels,imperfect price-reversibility?”, Transportation Research - B, 31B (1)pp 71-82.

Dargay, J. and D. Gately (1997b), “Income’s effect on car and vehicleownership, worldwide: 1960-2015”, TSU Working Paper Ref. 97/61,ESRC Transport Studies Unit, University College London, 1997.

Dargay, J.M. and P.B. Goodwin (1994), “Transport evaluation in adisequilibrium world, some problems in dynamics”, 11th AnnualConference on Transport Research, Linköping, Sweden.

Dargay, J.M. and P.B. Goodwin (1995), “Evaluation of consumer surplus withdynamic demand”, Journal of Transport Economics and Policy, XXIX, 2,179-193.

Dodgson, J. and B. Lane (1997), The costs of road congestion in Great Britain,National Economic Research Associates, London.

Glanville, W.H. and R.J. Smeed (1958), The basic requirements for the roads ofGreat Britain, Institution of Civil Engineers, London.

Goodwin, P.B. (1992), “A review of new demand elasticities with specialreference to short and long run effects of price changes”, Journal ofTransport Economics and Policy, XXVI (2).

200

Lay, M.G. (1993), Ways of the World: a history of the world’s roads and of thevehicles that used them, Primavera, Leichhardt. as cited by Brindle (1996).

Luk, J. and S. Hepburn (1993), New review of Australian travel demandelasticities, Victoria, Australian Road Research Board.

Mackett, R.L. (1998), “Why are travel demand forecasts so often wrong (anddoes it matter)?”, Universities Transport Studies Group Conference,Dublin, January.

Newbury, D.M. (1995), Reforming Road Taxation, The AutomobileAssociation, Basingstoke.

Tanner, J.C. (1977), Car ownership trends and forecasts, TRRL Report LR 799,Transport and Road Research Laboratory, Crowthorne.

Tanner, J.C. (1983), A lagged model for car ownership forecasting, TRRLLaboratory Report 1072, Transport and Road Research Laboratory,Crowthorne.

201

OTHER CONTRIBUTIONS

During the Round Table, several participants submitted writtencontributions. These contributions are reproduced below as complementaryinformation.

NETHERLANDS B. van Wee and R. van den Brink .....................203

UNITED KINGDOM D. Newbery........................................................209

202

203

NETHERLANDS

Bert VAN WEENational Institute of Public Health

and the Environment (RIVM)University of Utrecht -

Department of Geography

Robert VAN DEN BRINKNational Institute of Public Health

and the Environment (RIVM)

ENVIRONMENTAL IMPACT OF CONGESTIONAND POLICIES TO REDUCE IT

Introduction1

Growing congestion levels in western countries receive much attention,mainly because of accessibility and economic aspects. From an economicperspective, on many road networks congestion levels exceed the “optimumlevel” (which is not the same as no congestion at all: marginal costs to reducethe last bit of congestion are very often higher than marginal benefits). But howabout the environmental impact of congestion and measures to reduce it? Inthis paper, I will briefly discuss this subject, focusing on both direct effectscaused by differences in per kilometre emissions between traffic on congestedand on non-congested roads, as well as on indirect effects related to changes intraffic volumes, assuming constant time budgets. Only congestion on the mainnetwork is considered; urban congestion is excluded. Data for the Netherlandsare used to indicate quantitative levels of congestion and effects of congestionon emissions.

According to regular research in the Netherlands, in this paper I willdistinguish between two kinds of congestion: heavy congestion and othercongestion. Heavy congestion leads to relatively low speeds of roughly15 km/h on average, with cars very often standing still. Other congestion occurs

204

when speeds are lower than the “free flow speed”, but not as low as duringheavy congestion. Average speed might be about 60 km/h. It is very difficult todistinguish both kinds of congestion. Therefore statistics are not more thanindicative. According to the NEA (1997), in 1995 in the Netherlands, lostvehicle-hours due to heavy congestion were approximately 18 to 19 million anddue to other congestion, approximately 25 million hours. Assuming 15 km/hduring heavy congestion and 60 km/h during other congestion the share ofheavy congestion in kilometres is about 15 per cent, the rest is other congestion.Table 1 summarises these statistics.

Table 1. Congestion on the main road network in the Netherlands

Lost vehicle hours Share in vehicle kilometresduring congestion

Heavy congestion 18-19 mln 15%Other congestion 25-26 mln 85%Total 44 mln 100%

Source: NEA/RIVM.

Direct effects of congestion: emissions per kilometre

I will focus on emissions of CO2, Nox and VOC. CO2 is supposed to causeglobal warming and Nox causes acidification and health problems, both directlyand indirectly because of the contribution to ozone formation. The choice forthese pollutants is based on the relatively high share of transport in totalemission and on the fact that, in the Netherlands, current policy will not lead toreaching the targets (RIVM, 1997).

Research in the Netherlands and in Germany shows that per kilometre CO2

emission of cars and lorries in free flow is roughly 20 per cent lower thanduring heavy congestion. Reducing all congestion on Dutch main roads willreduce total energy use of road traffic by less than 0.1 per cent, assuming onlydirect effects. CO2 emissions of other congestion are approximately the same asunder free-flow conditions.

Nox emission per kilometre of both cars and lorries is lower duringcongestion (both heavy and other) than in free flow traffic. Emissions of VOCare higher during heavy congestion but not during other congestion.

205

Table 2 summarises the differences in emissions.

Table 2. Differences in emissions per kilometre between road trafficduring congestion, compared to free flow traffic

Heavy congestion Other congestionCO2 Higher than free flow About the same as free flowNOx Lower than free flow Lower than free flowVoc Higher than free flow About the same as free flow

indirect effects of congestion: traffic volumes

Indirect effects are caused by differences in travel volumes. Manyresearchers have concluded that people have a constant time budget fortransport (see Kraan, 1997, for an overview). Although there is some discussionabout the validity of the constant time budget theory, I will use it here.According to the theory, higher travel times (e.g. due to congestion) will lead tolower passenger kilometres. Most literature on long term travel time elasticitiesfor car use give values of about -0.5 to -1.0 (Goodwin, 1997). The (output)elasticity calculated with the Dutch National Model System (LMS; Bovy et al.,1992) is even -1.27, which means that 1 per cent longer travel times reduces carkilometres by more that 1 per cent. This might seem contra-intuitive but can becaused by two effects. Firstly, if travel times are lower, especially due to higherspeeds on motorways, routes might be changed: people might prefer longerroutes between given destinations because they are faster. Secondly, shortertravel times by car might result in a modal shift from other modes to car. In thispaper I will use -1.0, and therefore constant travel budgets for car use. Theimportance of the modal shift should not be overestimated. According tocalculations with the LMS, only a third of the reduction of car kilometres comesback in the form of other modes (public transport, car-pooling, slow modes).For environmental impact, only public transport is important. Assuming that10 per cent of the reduced car kilometres will come back as public transportkilometres and because per-passenger-km energy use of trains is about half ofcars (Van den Brink and Van Wee, 1997), and per-passenger-km emissions ofmost other pollutants are much lower for trains than for cars, I assume thatchanges in emissions due to congestion or congestion measures can beneglected.

206

Due to congestion, car use will be lower. Assuming 15 km/h during heavycongestion and 100 km/h during free flow, reducing congestion to zero will leadto about seven times as many kilometres in the same time span. Assuming60 km/h during other congestion, this increase will be about 65 per cent.Therefore, indirect effects of (a reduction of) congestion are much moreimportant than direct effects. Assuming the lost vehicle hours as mentionedbefore and a share of cars of 90 per cent (and 10 per cent lorries, vans andbuses), reducing congestion to zero will result in an increase in car use of about3 per cent.

These calculations are in line with model simulations. Results of ascenario study of TNO-INRO show that, in 2015 in the Netherlands, the levelof car use would be about 6 per cent lower if the main road network would notbe extended according to current policy (Verroen et al., 1995). This highereffect than calculated for 1995 is mainly the result of the higher congestionlevel in 2015 if the road network would not be extended.

For the use of lorries, no travel time elasticities were found. They probablywill be much lower than for car use.

Measures to reduce congestion

Assuming the constant time hypothesis reducing congestion by buildingmore roads will lead to relatively much higher emissions of all relevantcomponents.

Congestion pricing might lead to shorter travel times but also to highertravel costs. The overall influence on car use varies strongly according to thesystem of congestion pricing. Calculations with the LMS show that currentpolicy plans for the Netherlands (congestion pricing only during the peak hoursand only in the four main regions) will lead to an overall reduction of car use inthe Netherlands of about 1 per cent, whereas a system developed for the SecondTransport Structure Plan, assuming congestion pricing not only in the Randstad(the densely populated western part of Holland) but also outside the Randstad,high tariffs and not only during the peak hours, should lead to about 13 per centless car use.

207

Simulations with the LMS show that 1 per cent less car use due to higherlevies on fuels will lead to 2-5 per cent less lost vehicle hours. Higher fuelprices, although not the first best solution to solve congestion problems(Verhoef, 1996), have positive effects on congestion levels but also onemissions.

Conclusions

If congestion on main roads were to be reduced to zero, CO2 emissions ofroad traffic would be about 0.1 per cent lower, assuming only direct effects.Emissions of VOC will also be lower, but emissions of Nox will be higher.

Reducing congestion by building more roads has a strong indirect effectbecause car kilometres will increase. This indirect effect is much higher thanthe direct effect, due to changes in emissions per vehicle-kilometre. If allcongestion in the Netherlands were to disappea, car kilometres would increaseby about 3 per cent.

NOTE

1. This paper is based on Van Wee (1997)

208

BIBLIOGRAPHY

Bovy, P.H.L., J. Jager, H. Gunn (1992), The Dutch National and RegionalModel Systems: Principles, Scope and Applications, Selected Proceedingsof the Sixth World Conference on Transport Research (WCTR), Vol. II,Demand. Traffic and Network Modelling, WCTR Society, 1992,pp. 1197-1208.

Goodwin, P.B. (1996), Empirical evidence on induced traffic: A review andsynthesis, Transportation, Vol. 23, No. 1, .pp. 35-54.

Kraan, M. (1997), Time to travel? A model for the allocation of time andmoney, PhD thesis, University of Twente, Enschede.

NEA (1997), Congestion costs on the main road network, NEA, Rijswijk(in Dutch).

RIVM (1997), National Environmental Outlook 4 1995-2020, Alphen aan denRijn, Samson, H.D., Tjeenk Willink (in Dutch).

Van den Brink, R.M.M. and G.P. van Wee (1997), Energy use and emissionsper kilometre, RIVM Report No. 773002007, National Institute of PublicHealth and the Environment, Bilthoven (in Dutch).

Van Wee., B. (1997), Less congestion, better environment? Verkeerskunde, No.3, 1997, pp. 18-19 (in Dutch).

Verhoef, E. (1996), Economic efficiency and social feasibility in the regulationof road transport externalities, PhD Thesis, Thesis Publishers Amsterdam.

Verroen, E.J., H.D. Hilbers, C.A. Smits (1995), Model evaluation of the visionon the Randstad: the Results, INRO-TNO, Delft (in Dutch).

209

UNITED KINGDOM

David NEWBERYApplied Economics Department

University of Cambridge

MODELLING TRAFFIC CONGESTION IN CAMBRIDGE

To calculate the social costs of congestion in Cambridge we are using theSATURN (Simulation and Assignment of Traffic in urban Road Networks)model. SATURN is a software that simulates and assigns traffic in urban roadnetworks and iterates until the equilibrium is reached. The equilibrium is thesituation in which no tripmaker can reduce his or her generalised cost. Thegeneralised cost is the cost of time plus the cost of distance. The model stopsiterating when the cost of travel and all unused routes have equal or greater cost.

In Cambridge and the surrounding area, the average loads of traffic atdifferent times of the day during weekdays are the following:

Time of day Number of cars Veh-km Veh-km/vehMorning peak 48 119 965 400 20.06Evening peak 44 421 861 380 19.39Off peak 28 227 568 850 20.15

Source: W.S. Atkins on behalf of Cambridge County Council.

Using the following values of time and distance:

Peak Off-peakValue of time (PPM) 7.63 8.53Value of distance (PPK) 5.27 6.25

Source: Department of Transport, Environment and the Regions.

210

we arrived at the following conclusions:

1) The average and marginal costs at different times of the day are:

Morning peak Evening peak Off-peakAC in ppk 13.12 12.84 13.38MC in ppk 20.57 18.81 13.74MSC in ppk 18.77 17.01 11.94

2) Assuming a linear inverse demand function with a point elasticity at theprevailing level of traffic of 1 and 0.5, the deadweightloss (or differencebetween the marginal social cost and the price actually paid by tripmakers) percar, expressed in pence/km, is:

Morning peak Evening peak Off-peakη=1 0.4396 0.2589 0.0409η=0.5 0.3326 0.1907 0.0224

3) Total DWL expressed in pounds:

Morning peak Evening peakη=1 9 244 2 230η=0.5 3 211 1 643

These numbers were obtained by multiplying the DWL by the number ofveh-km driven in the system.

211

FIGURES

212

Figure 1. Average and marginal costs in ppk during the evening peak

0

40

0

30

25

20

15

10

1.50.5 1.0

35

5

40

0

30

25

20

15

10

35

5

Average and marginal costs in ppk Average and marginal costs in ppk

Traffic load (vehicles in the network per hour)

Figure 2. Average and marginal costs in ppk during the morning peak

0

45

0

35

30

25

20

15

10

5

1.50.5 1.0

40

45

0

35

30

25

20

15

10

5

40

Average and marginal costs in ppk Average and marginal costs in ppk

Traffic load (vehicles in the network per hour)

45

0

35

30

25

20

15

10

5

40

213

Figure 3. Average and marginal costs in ppk during the off peak

0

17

11

16

15

14

13

12

1.50.5 1.0

17

11

16

15

14

13

12

Average and marginal costs in ppk Average and marginal costs in ppk

Traffic load (vehicles in the network per hour)

215

SUMMARY OF DISCUSSIONS

216

217

SUMMARY

INTRODUCTION............................................................................................219

1. DEFINITION, SCALE AND SPREAD OF CONGESTION ..................220

1.1. Definition of congestion ...................................................................2201.2. The scale of congestion ....................................................................2201.3. The spread of congestion ..................................................................222

2. DEVELOPMENT AND IMPACT OF CONGESTION ..........................223

2.1. Recent trends ....................................................................................2232.2. Results ..............................................................................................2252.3. Future trends .....................................................................................225

3. SOLUTIONS TO CONGESTION...........................................................226

CONCLUSIONS..............................................................................................228

218

219

INTRODUCTION

The problems caused by road congestion are frequently reported in themedia and finding a solution to congestion ranks high on politicians’ agendas.Countless studies have been published detailing the number of working hourslost in traffic jams. The figures quoted in these studies are alarming and paintan apocalyptic picture of road congestion. The sensitivity of certain segmentsof public opinion to conditions on the roads may be explained by theimportance attached to the environment and the large share of responsibility thattransport bears in the lowering of the quality of living conditions.

However, beyond the every-day congestion faced by motorists, differencesof opinion are now starting to appear with regard to the scale of the problem andthe ways in which it should be tackled. Some analysts feel that, becausecongestion is restricted to certain periods and routes, the answer is to supply themissing capacity, arguing that road congestion can only be solved by buildingnew roads; others think that congestion can only worsen and become morewidespread and that the solution primarily lies in developing alternative meansof transport.

In order to clarify the issues involved, the Round Table proceeded in threestages:

� First, it attempted to define congestion and to determine the true scaleof the problem;

� Second, it addressed the trends in congestion and the consequences ofthose trends;

� Third, it considered possible solutions to alleviate this congestion.

220

1. DEFINITION, SCALE AND SPREAD OF CONGESTION

1.1. Definition of congestion

There is no universally recognised definition of congestion. Congestion isa traffic condition in which vehicles are constantly stopping and starting and inwhich vehicle concentration is high while flow speeds are low. A highconcentration of vehicles on the road is not in itself a characteristic ofcongestion; it would need to be accompanied by low flow speeds to create asituation in which capacity is saturated. It should also be noted that, conversely,low flow speeds are not sufficient to characterise a situation as being one ofcongestion. Flow is therefore not a relevant indicator of congestion. Lastly, itneeds to be said that while speed can readily be measured, it is far more difficultto determine the speed below which traffic flows start to become congested.

Perhaps the most distinctive characteristic of congestion is the fact that thetraffic load starts to approach the maximum capacity of the infrastructure andthat at such a level of traffic flow any additional vehicle on the carriageway willconsiderably slow down traffic. Furthermore, when traffic flow starts to reachmaximum capacity, any unforeseen event (accident, roadworks) can causesevere disruption. In view of these considerations, the Round Table adopted thedefinition of congestion proposed by P. Goodwin and J.M. Dargay in theirintroductory report, namely that “congestion is the impedance vehicles imposeon each other, due to the speed-flow relationship, in conditions where the use ofa transport system approaches its capacity.”

We obviously need a clear and unambiguous definition of congestionwhich can serve as a basis for a proposed measurement of the scale ofcongestion, since the precise definition of congestion will clearly depend uponthe way we wish to apply it and, in particular, on how we wish to measurecongestion.

1.2. The scale of congestion

The Round Table noted that congestion could be measured in threedifferent ways. First, free-flow speeds could be used as a reference to calculateand then cost the time lost through congestion. Under the above definition, thatis to say, on the basis of free-flow traffic conditions, congestion costs in Europecould amount to around 2 per cent of GDP. However, this method of measuringcongestion, despite being simple and easy to use, is unrealistic in that it takesthe speed under freely flowing traffic conditions as a reference. If traffic

221

conditions are fluid, then infrastructure capacity is under-utilised to the extentthat there is no longer any justification for the investment in that infrastructureand, in economic terms, the situation is sub-optimal. There therefore exists aneconomically optimal level of congestion, which is the level that needs to beachieved in order to ensure that capacity is not under-utilised. The Round Tabledrew attention to the fact that, in the Netherlands, the optimum level ofcongestion has been defined as one at which 2-3 per cent of motoristsencounter congestion on an average day. The figure of 2 per cent of GDPobtained for Europe, in terms of the cost of lost time, is undoubtedly also anestimate which relies on controversial hypotheses about the value of time. Itneeds to be borne in mind, however, that any definition or measurement ofcongestion will have its limits, as may clearly be seen from the range of otherpotential measurements.

Secondly, road congestion can be measured in terms of the revenue thatwould be raised if road pricing were introduced to internalise the costs ofcongestion. It is clear that an additional road user who uses a congestedinfrastructure generates costs for other road users. The introduction ofdifferentiated road pricing for off-peak and peak-hour driving would make itpossible to charge for infrastructure use on a cost basis and to generate revenue.

The third way of measuring congestion would be to estimate the benefits,that is to say, the efficiency gains, that could be realised though the levying ofsuitable charges on road infrastructure use. Excise duties on fuel, which do notadequately reflect the costs of infrastructure usage, could be replaced by ahighly sophisticated pricing system which would partly dissuade motorists fromusing congested infrastructure. Congestion costs, measured in terms of theefficiency gains achieved with regard to infrastructure use, are estimated toamount to less than the 2 per cent obtained through the first type ofmeasurement. The Round Table estimated that the figure obtained through theuse of this method would amount to around 0.75 per cent of GDP.

The last two of the above methods of measuring congestion are based oneconomic concepts and refer to an optimum for the assessment of congestioncosts. As a result, it is hard to assess this type of measurement in practice,which considerably reduces the operational scope of such estimates. It isundoubtedly for this reason that the most widely-used measurement ofcongestion has been to estimate time lost in comparison with fluid trafficconditions. Nonetheless, it is necessary to develop more sophisticated methodsof measuring congestion and its costs.

222

1.3. The spread of congestion

Time lost in comparison with fluid traffic exaggerates the scale ofcongestion, which the Round Table emphasized affects only a small portion ofthe road network in Europe as a whole.

The available capacity in terms of intercity road links is sufficient. Europehas a very high capacity motorway network. Only 300 kilometres of thisnetwork (mainly in Germany and the United Kingdom) experience traffic flowshigher than 80 000 vehicles a day, i.e. a flow rate which would requiremotorways with more than 3-lane carriageways. Congestion is not necessarilythe norm in the urban road network either. The average duration ofjourney-to-work trips in Europe is 20 minutes. In Paris, the average duration ofsuch trips is 27 minutes. In contrast, the average figure to emerge fromhousehold surveys is 1 hour 20 minutes, which shows how congestion isexaggerated subjectively. It should also be noted that 90 per cent of theinhabitants of Germany, France, the Netherlands and the United Kingdom claimnot to experience any congestion at all during trips. It is worth bearing in mind,too, that traffic speeds have been steadily rising for many years.Notwithstanding the above, the effects of congestion, in the places where it doesoccur, are spectacular and are largely attributable to a long-standing policy ofimposing restrictions on investment in road infrastructure. In Germany, despitethe lack of comprehensive data, it appears that only 2 per cent of the network iscongested. In the Netherlands, which is often cited as an example of congestedinfrastructure, barely 2 per cent of motorists report encountering congestion onthe roads during an average day. Congestion is therefore a minor, althoughadmittedly spectacular, phenomenon.

Why is it then that, despite the facts of the matter, public opinion andpoliticians remain so sensitive to congestion, focusing on it to the exclusion ofall other transport issues?

Congestion is not a widespread problem in the European road network as awhole but, when it does occur, it is critical. Studies have shown that individualmotorists perceive the time spent immobilised in traffic jams to be three timeslonger than the time actually spent waiting. Moreover, as noted earlier, thereference time which is commonly used is one in which traffic flow is fluid,even though in economic terms such an approach is questionable. Aggregatingvalues for traffic flows under such conditions provides an inaccurate reflectionof the actual situation. When considered in terms of the total volume of roadtrips, congestion is seen to affect only a limited number of trips. Moreover, in asystem that functions properly in overall terms, exceptions, that is to say,unstable flow conditions, can readily be identified and have a major impact. Inaddition, congestion occurs at a number of highly specific locations, which

223

makes their impact all the more perceptible. On the other hand, it should benoted that the construction of high-capacity roads has prevented any furtherworsening of the situation and that there are many other factors which canexplain why travel times have remained constant over a long period. Anyconsiderations that might tend to reduce the perceived scale of congestion mustbe seen in relative terms if heavy traffic flows are taken into account alongsidecongestion, since this would be a much larger-scale phenomenon. One of theimpacts of high-density traffic is that it accelerates urban decline. Whentransport conditions become difficult, the people living in the centres of townseventually move out, congestion being one of the deciding factors forrelocation.

2. DEVELOPMENT AND IMPACT OF CONGESTION

2.1. Recent trends

Over the past few years, growth in demand for transport in the UnitedStates has outpaced growth in capacity. The severity, duration and scale ofcongestion have therefore increased. However, the share of trips carried outunder congested conditions in the total number of trips has fallen. This showsthat motorists avoid making trips during peak traffic hours, thereby helping tospread the traffic load.

In Europe, even though the situation varies from one country to another aswell as within the same country, rather than intensifying, congestion is tendingto become more evenly spread both over time and spatially. The factors behindthis spread in congestion are rising rates of car ownership -- and the stable ordeclining costs of car use, linked among other things to falling petrolconsumption -- as well as the population density, which means that congestionis primarily a regional but by no means a national or international phenomenon.Demand for transport, particularly by car, tends to increase with rising standardsof living and there can be no doubt that infrastructure capacity has not been andcannot be adjusted accordingly. The outcome of this will be a trend towardsgreater congestion unless there is a change in people’s behaviour, namely achange in departure times, route, modal transfers, home-working, part-timeworking, place of residence, etc. As a result of these very real andunderestimated changes in behaviour, motorists tend to experience no change intravel times, even though congestion is becoming more widespread both overtime and spatially. Furthermore, infrastructure improvements have made itpossible to prevent congestion from worsening. In all, congestion remainsconstant in terms of how it is experienced by the individual, even if, globally, it

224

is spreading. In order to gain some insight into these trends, it would be veryhelpful to develop indicators of congestion with regard to individual trips. Atall events, the nature of the problem relates not to declining travel speeds but tothe increasingly severe environmental and economic impacts of the problem.

While travel times tend to remain stable, the same cannot be said of theirpredictability in that congestion is a phenomenon that is both recurrent andunpredictable, with the result that travel times can vary. This is a problemwhich affects all users but particularly road freight hauliers for whomcongestion means additional costs in terms of personnel and vehicles. Bothfirms and private cars users are obliged to anticipate the worst possible trafficconditions when planning trips. The growing uncertainty over travel times,which prompts both firms and individuals to take precautions, is one of the mostdamaging consequences of congestion in that the reliability of the transportsystem has declined considerably and this has undoubtedly had a significantimpact at the macroeconomic level. It is above all on this aspect that efforts tocombat congestion should be focused. Moreover, it is a fact that attention hasbeen concentrated on the visible aspects of congestion and not on the latentchanges in demand. By leaving at an earlier time, individual users can limit theimpact of congestion on their trips, but there is also a resultant loss ofwell-being that is not measured. What it is not possible to measure in practiceis the collective cost of changes in behaviour aimed at avoiding congestion.

Furthermore, congestion also weighs on the choice of transport policy.Plans to put in place transport systems offering an alternative to private car useare postponed out of fear of the impact they might have on already saturatedroad infrastructure. Politicians, for example, will be reluctant to dedicate part ofthe carriageway to public transport if it becomes apparent that such action mightbring road traffic to a standstill. At the same time, predictions are regularlymade of chaos on the roads unless new infrastructure is constructed. Theconclusion that can be drawn from this is that congestion is an issue that canweigh heavily on transport policy.

It has been shown that growth in road traffic is related not to an increase injourney-to-work trips but to growth in other areas. Journey-to-work tripsaccount for merely 20 per cent of trips, but it is precisely the congestionresulting from journey-to-work trips that needs to be addressed because theimpacts it has are those which are the most significant. And yet even withregard to this type of congestion, which occurs on a daily basis, it is impossibleto predict the exact time at which queues of traffic will start to build up.

225

2.2. Results

When vehicles are stationary -- in situations of intense congestion -- thepollution is considerable. However, in the case of dense moving traffic theenvironmental consequences cannot easily be determined. In these trafficconditions, vehicles travel slowly and therefore emit fewer pollutants. On theother hand, the concentration of such pollutants at certain locations raisespollution above the levels at which it is considered to have a harmful effect.The first to be exposed to the effects of pollution levels being exceeded are thedrivers of vehicles. In addition to which, account needs to be taken of the factthat in most cases vehicles travel over short distances with engines that remaincold, with the result that pollutant emission levels are higher. The flows ofvehicles on the secondary road network generated by drivers attempting toavoid congested areas also create environmental problems.

With regard to pollution and more generally all matters relating tocongestion, the individual situations observed varied considerably and make itdifficult to draw any general conclusions. In addition, not enough detailedmeasurement systems are available, prompting analysts to claim that congestionmight well be simply the tip of the iceberg. Account also needs to be taken ofthe impact of congestion on road accidents. While road accidents in congestedconditions are not as serious as they would be if vehicles were able to travelfaster, in contrast, they are higher in number and frequently involve pedestrianswhose movements are hampered by the omnipresent car traffic. In addition tothese costs arising from congestion, there is also the impact on public transport,whose operating speeds are reduced as a result of congestion, which has aconsiderable negative impact on the financial performance of public transportoperators.

2.3. Future trends

How will congestion evolve in the future? It will follow the trend in caruse. The dominant position of car use in the modal split is such that someexperts claim that households choose to live in locations where they can usetheir cars and cite this as the reason for which households are moving awayfrom the centres of cities, where car use is often problematic, to areas in thesuburbs.

It has also been observed that, as a result of rising living standards, thereare more and more women drivers and that the elderly remain dependent uponcar use for increasingly long periods of time. The populations of easternEuropean countries are only now starting to enjoy the benefits of car ownership.In general, cars are now used more than they were ten years ago, and there is a

226

direct link between higher incomes and car use. Even though growth in carownership rates is now declining, it is by no means certain that car ownershiphas reached saturation levels. The Round Table was unable to reach aunanimous conclusion with regard to the eventual saturation in car ownershiplevels, primarily because multiple-car ownership among households isbecoming increasingly widespread. In addition, although the growth in thenumber of trips made by car is low, the distances travelled, on the other hand,are rising. It is as though individual car users had taken advantage ofimprovements in the infrastructure to travel further in the same period of time,for example, by moving to the outskirts of urban areas. This would suggest thatpeople have adapted to congestion, while at the same time the environment isstarting to benefit from the use of catalytic converters and the performance ofcars has improved, thus making the time spent inside vehicles seem lessunpleasant. It is therefore difficult to predict future trends in the developmentand impacts of congestion.

With regard to town planning, i.e. the spatial distribution of activities, theconcentration and dispersal of activities both add to congestion but at differentlocations, namely, in the city centre if activities are concentrated or along accessroutes to the city centre if activities are dispersed. In addition, trips will bewidely induced as a result of the provision of infrastructure. Consequently, theprovision of new road infrastructure will generate induced traffic, even thoughat our present level of understanding it is difficult to forecast the scale of suchtraffic. It is perfectly possible that building more motorways will encouragegreater car use. In order to study these phenomena, we need to examine thedifferent time horizons at which drivers start to change their behaviour to adaptto changed conditions. It is therefore likely that in the long term there will be alarge volume of induced traffic. The distances travelled may also be influencedby the supply of infrastructure. Traffic flows to city centres appear to be givingway to traffic between suburbs. What we do know, on the other hand, is thattraffic has not increased on main roads but on alternative routes, which showsthat an increasing number of motorists are consciously trying to avoidcongestion.

3. SOLUTIONS TO CONGESTION

As congestion is an essentially urban phenomenon, the solutions arediverse and depend a great deal on circumstances. First, if congestion can beavoided then the road capacity of encumbered zones could be notably increased,given that traffic speeds would rise. Teleworking might possibly become morewidespread in Europe, thus reducing the need for journey-to-work trips.

227

However, the Round Table remained sceptical on this point, sincetelecommunications and transport seem more likely to be complementary.Telematics could play a major role in improving the fluidity of traffic flow,either by providing road users with more information regarding trafficconditions or by acting directly upon traffic flow. Telematics and itsapplications can redistribute transport demand within the network. However, itneeds to be borne in mind that improved traffic flow will encourage greater caruse, thereby creating a new phenomenon comparable to that of induced traffic.In the long term, individuals move and take advantage of better trafficconditions in order to live in an environment that is more to their pleasing. It isfor this reason that efforts to find lasting solutions to congestion should bedirected towards charging a suitable price for the use of infrastructure, which isthe only way to reduce the number of non-essential trips during peak traffichours.

Pricing and taxes are currently aimed at securing funding for infrastructureand not at directing the choices made by users. It is precisely in this area,however, that action is urgently needed. As a general rule, the cost of travel bycar in urban areas should be higher while the cost of intercity travel should belower. There are several ways of achieving this. Urban tolls are a theoreticalpossibility and charges could also be introduced for parking. In the case ofurban tolls, electronic technologies will soon make it possible to reconciletheory with practice through the use of modulated pricing (peak hours, off-peakhours), but the large-scale introduction of such systems using commonlyaccepted forms of payment (smart cards, badges) still poses problems. Anotherfactor which also needs to be taken into account is the crowding-out oflow-income users. Studies have shown that high tolls will have to beintroduced in order to achieve the desired results, given the inelasticity oftransport demand. In view of this and as a short-term measure, parking chargesremain a viable alternative. In order to discourage car use, traffic speeds canalso be reduced through traffic calming measures. The aim of these measures isto make car use less appealing. At the same time, a viable alternative needs tobe provided in the form of public transport. It should be noted that measuresaimed solely at developing public transport will be destined to failure. Trafficlevels in densely populated areas, therefore, can only be reduced through a mixof incentives and disincentives. The implementation of a combination ofpolicies may well produce the desired result. It is also worth noting thatinvestment in public transport cannot be justified solely as a means of reducingcongestion. Public transport services have a social utility which goes farbeyond the simple objective of reducing congestion.

The situation with regard to public transport in peripheral areas is whollydifferent. Public transport in such areas is problematic and, above all, costly todevelop to any degree of satisfaction. While access can be provided to the city

228

centre, the provision of public transport services between suburbs is seldomsatisfactory. There are therefore no real alternatives to the development of roadlinks, a conclusion that will mostly prove unpalatable to politicians who, to alarge extent, will prefer to develop public transport. One option might be tobuild roads meeting very high environmental standards, although theconstruction of new road capacity poses other problems in that it inducesdifferences in capacity at points where different networks intersect (e.g. urbanand intercity networks), with the result that the capacity gains at certain pointsare cancelled out by the shortcomings of the complementary networks. Theconstruction of new roads, despite being the only solution in certain cases, cantherefore be seen to pose its own problems.

It would seem that reducing road capacity also reduces the number of tripsthat are made. Here too, long-term solutions start to appear in the form ofchanges in destinations, for example, or the combining of trips, with the resultthat the overall volume of traffic is lower. Traffic volume is not a staticparameter in that it depends, to some extent, on the policies pursued. It is forthis reason that consideration might be given to reducing road capacity in citycentres and to the assignment of capacity to other uses. It is clear that roaddevelopment in city centres is not a solution to congestion, given that it simplyencourages businesses to relocate to peripheral areas, thereby increasing thenumber of trips made by car, in a shift towards a North American lifestyle.When access capacity is restricted, activities relocate elsewhere.

CONCLUSIONS

Congestion can be compared to the time spent in queues waiting to pay ina large department store during peak shopping hours; it is an intrinsic part ofthe system and one that cannot be wholly eliminated. Furthermore, people haveadapted to congestion, partly due to improvements in the road network but alsoas a result of improved vehicle performance and motorists’ behaviour. Theseare two parameters that cannot be measured, and because of this the cost ofcongestion remains partly hidden. One of the main reasons people complainabout congestion is that they think there is a solution to the problem. The truthof the matter is that congestion cannot be totally eliminated; it can, however, bealleviated.

At all events, investment in the road network is not aimed solely atreducing congestion. The primary objective is to ensure a high level ofaccessibility to all locations within a country, but that objective may also

229

include efforts to combat congestion. The danger lies in thinking that becausethere is no congestion there is no need for investment.

The situation in Europe varies widely from one country to another, whichmeans that there is no all-embracing solution to traffic problems. Considerationmight be given to adopting different approaches that take account of culturalfactors. Advanced telematics applications would make it possible to improvetraffic management and driver information; however, such technologies aresimilar in their effect to investment in road infrastructure, in that improvingdriving conditions will simply attract new road users and, as a result, theycannot be expected to have a major impact on congestion. The way to achievetangible results is to combine policies which encourage the use of alternativemodes of transport with others aimed at discouraging car use.

One of the priorities in the short term would be to set up a standardized andcoherent system for assessing both the volume and the cost of congestion. Itwould also be of particular interest to conduct additional research into thehidden costs of congestion.

230

231

LIST OF PARTICIPANTS

Professor Peter JONES ChairmanDirectorTransport Studies GroupUniversity of Westminster35 Marylebone RoadGB-LONDON NW1 5LS

Professor Dr. P.H.L. BOVY RapporteurTRAILDelft University of Technology/Erasmus University RotterdamP.O. Box 5048NL-2600 GA DELFT

Professor Ilan SALOMON Co-rapporteurDepartment of GeographyHebrew University of JerusalemJERUSALEM 91905Israel

Monsieur Christian GERONDEAU RapporteurPrésidentUnion Routière de France10, rue Clément MarotF-75008 PARIS

232

Professor Phil.B. GOODWIN RapporteurESRC Transport Studies UnitUniversity of London Centre for Transport StudiesUniversity College LondonGower StreetGB-LONDON WC1E 6BT

Dr. J.M. DARGAY Co-rapporteurESRC Transport Studies UnitUniversity College LondonGower StreetGB-LONDON WC1E 6BT

Dr. Karl Otto SCHALLABOCK RapporteurWuppertal Institute for Climate, Environment and EnergyDöppersberg, 19D-42103 WUPPERTAL

Dr. Rudolf PETERSEN Co-rapporteurWuppertal Institute for Climate, Environment and EnergyDöppersberg, 19D-42103 WUPPERTAL

Dr. Silvia BANFIINFRASGerechtigkeitsgasse, 20CH-8002 ZURICHSuisse

Dr. David BANISTERBartlett School of Architecture and PlanningUniversity College LondonWates House22 Gordon StreetGB-LONDON WC1H 0QBDr. Halina BRDULAK Observer

233

Head of European Integration SectionMotor Transport Institute80 Jagiellonska St.PL-03-301 WARSAW

Mr. Harry CALDWELLChief, Highway Needs & InvestmentOffice of PolicyFederal Highway Administration400 7th St. SWWASHINGTON D.C. 20590

Prof. Dr. A. DIEKMANNUniversität zu KölnHardtstr. 1D-61250 USINGEN

Mlle Maria-José GUERRERO GARCIACivil EngineerArea de Estudios y PlanificacionConsorcio de Transportes de MadridPza. Descubridor Diego de Ordás, 3E-28003 MADRID

Mr. Keith KEENCommission EuropéenneDirectorate General VII - TransportAvenue de Beaulieu, 31 4/52B-1160 BRUXELLES

Prof. Dr. Boris KERNERVerkehrstechnik (FT1/V) HPC E224Daimler-Benz AGD-70546 STUTTGART

234

Ing. Jiri LANDACityPlan Ltd.Spálená 5CZ-111 21 PRAGUE 1

Monsieur Marc LEMLINIngénieur civilDirecteur général des Routes et autoroutes de la Région WallonneMinistère de l’Equipement et des TransportsRégion WallonneAvenue Reine Astrid 39-43B-5000 NAMUR

Mr. Gunnar LINDBERGCTSDalarna UniversityS-78188 BORLÄNGE

Professor Dr. David NEWBERYDirector of Applied Economics DepartmentUniversity of CambridgeSidgwick AvenueGB-CAMBRIDGE CB3 9DE

Professor S. PROOSTDepartment of Economics - KU LeuvenCentre for Economic StudiesNaamse Straat, 69B-3000 LEUVEN

Monsieur le Professeur Emile QUINETChef du Département d’Economie et des Sciences SocialesEcole Nationale des Ponts et Chaussées28, rue des Saints PèresF-75007 PARIS

235

Dr. Farideh RAMJERDIDepartment of Infrastructure and PlanningRoyal Institute of TechnologyTeknikringen 72SE-100 44 STOCKHOLM

Dr. Aisling REYNOLDS-FEIGHANUniversity College DublinDepartment of EconomicsBelfieldIRL-DUBLIN 4

Professor Werner ROTHENGATTERInstitut für Wirtschaftspolitik und WirtschaftsforschungUniversität Karlsruhe (TH)Kollegium am Schloss, Bau IVD-76128 KARLSRUHE

Drs. Arjen `T HOEN ObserverTransport Research Centre AVVMinistry of Transport, Public Works andWater ManagementP.O. Box 1031NL-3000 BA ROTTERDAMPays Bas

Monsieur Jean-Pierre VAN DE WINCKELTraffic ManagementTouring Club de Belgiquerue de la Loi, 44B-1040 BRUXELLES

Dr. Jeremy VANKEHead of Public PolicyRoyal Automobile Club156A Upper Clapton RoadGB-LONDON E5 9JZDr. Attila VOROSHead of Department of Transport

236

System Research and Network PlanningInstitute for Transport Sciences LtdThán K. u. 3-5H-1119 BUDAPEST

Mr. Bert van WEENational Institute of Public Healthand the Environment (RIVM)P.O. Box 1NL-3720 BA BILTHOVEN

Herrn Dipl.-Volkswirt G. WEICHHead of Traffic Department (VEK)ADAC e.V.Am Westpark 8D-81373 MUNICH

ECMT SECRETARIAT

Mr. Gerhard AURBACH – Secretary-General

ECONOMIC RESEARCH, STATISTICS AND DOCUMENTATIONDIVISION

Mr. Alain RATHERY – Head of DivisionMr. Michel VIOLLAND - AdministratorMrs Julie PAILLIEZ - AssistantMs Françoise ROULLET - Assistant

TRANSPORT POLICY DIVISION

Mr. Stephen PERKINS - Principal Administrator

237

ALSO AVAILABLE

New Trends in Logistics in Europe - Round Table 104 (1997)(75 97 05 1 P) ISBN 92-821-1224-1 France FF215 £28 $US42 DM63

Infrastructure-Induced Mobility. Series ECMT - Round Table 105 (1998)(75 98 07 1 P) ISBN 92-821-1232-2 France FF400 £40 $US67 DM119

Intercity Transport markets in Countries in Transition. Series ECMT - Round Table 106(1998)(75 98 10 1 P) ISBN 92-821-1235-7 France FF400 £41 $US66 DM119

User charges for railway Infrastructure. Series ECMT – Round Table 107 (1998)(75 98 14 1 P) ISBN 92-821-1240-3 France FF290 £30 $US50 DM86

14th International Symposium on Theory and Practice in Transport Economics. WhichChanges for Transport in the Next Century? (1999)(75 1999 01 1 P) ISBN 92-821-1241-1 France FF590 £63 $US105 DM176

What Markets Are There For Transport by Inland Waterways? Series ECMT – RoundTable 108 (1999)(75 1999 06 1 P) ISBN 92-821-1246-2 France FF300 £32 $US53 DM89

Freight Transport and the City. Series ECMT – Round Table 109 (1999)(75 1999 08 1 P) ISBN 92-821-1247-0 France FF280 £29 $US47 DM84

Prices charged at the OECD Bookshop.

The OECD CATALOGUE OF PUBLICATIONS and supplements will be sent free of chargeon request addressed either to OECD Publications Service,

or to the OECD Distributor in your country.

OECD PUBLICATIONS, 2, rue Andre-Pascal, 75775 PARIS CEDEX 16PRINTED IN FRANCE

(75 1999 09 1 P) ISBN 92-821-1248-9 – No. 50861 1999