« Previous | Contents | Next »
Listen
CHAPTER FIVE BACKGROUND TO MEASURING COSTS OF CONGESTION
5.1 In measuring the costs of congestion, there are a number of issues to highlight which affect the approach that may be taken and the interpretation or use of the output as follows:
- A difference exists between the total costs of congestion, the marginal costs of congestion (the effect on congestion of one extra vehicle) and the costs of the 'excess burden of congestion'. In chapters 6 and 7 below these three types of congestion costs are defined in greater detail and the relevant literature reviewed.
- The methods used to measure costs of congestion can be typified as primarily static versus dynamic methods, with some approaches forming a hybrid between these. The broad principles are described below, with further detail on relevant studies which have used different methods given in chapters 6 and 7.
- The appropriate approach to measuring costs of congestion will vary according to the end use of the data. For example, in cases where the aim is to consider road pricing measures, the marginal cost of congestion has been calculated. To review the benefits of significant investment decisions, the total or excess burden of congestion may be calculated. The purpose of the research here is to provide objective evidence on each based on the existing literature. The work will inform subsequent stages of research to be conducted by the Scottish Executive and at this point it is not possible to propose recommended methodologies until the nature of that programme is defined.
5.2 Dynamic methods of calculating the costs of congestion essentially relate to an iterative process between supply, demand and the cost of travel. Some care is needed with the terminology in order to avoid confusion between a dynamic approach to calculating costs and a dynamic network model. The latter is termed dynamic in a traffic engineering sense - i.e. dynamic assignment techniques vs. static (steady state) techniques. In fact a dynamic method of modelling the cost of congestion can use either a static or dynamic traffic model. The advantage of using a dynamic model is that it attempts to represent detailed changes at the spatial level e.g. in route choice and also the temporal level e.g. departure time choice.
5.3 The estimation of marginal cost and marginal external costs ( defined and described in chapter 6) is a far from trivial task. Primarily this arises as it is necessary to model how user costs (travel time, reliability, etc.) change in response to an additional vehicle-kilometre or trip. Additionally it is a fundamental requirement that that marginal cost functions for each of the cost components (detailed in Table 6.1) are available. Shires (2006) identifies four principal methods for the calculation of congestion impacts on the users of the transport system. These methods are set out below.
5.4 Link speed-flow relationships. This method is relatively simple and assumes a single link speed-flow relationship for all links of a certain type (quality, time period, location) in the transport system. Diversion from one link type to another is not possible; however, trip suppression and generation can be modelled using simple elasticities.
5.5 Area speed-flow curves. This method uses a single speed/flow relationship to represent average travel times in a particular area of the network. That is a single relationship is taken to represent average travel times on all links within a particular area and at all junctions in that area. Different areas of the network have different relationships attributed to them. Diversion between areas is possible as is trip generation and suppression. An example of such a model in Scotland would be the TRAM/ DELTA model developed by MVA for the City of Edinburgh Council for the appraisal of congestion charging.
5.6 Network assignment models. This method utilises detailed transport network models which model link and junction delay. Diversion between different roads (links) is possible and depending on the complexity of the model diversion between modes is also possible. The Scottish Executive's Transport Model for Scotland is an example of a network assignment model.
5.7 Microsimulation models: Microsimulation models have a more recent history than the above three model types. They offer a detailed representation of the behaviour of individual vehicles in a system and can respond in real-time. Whereas the above three model types utilise relationships describing average behaviour, microsimulation models simulate the behaviour of vehicles in response to dynamic changes in the transport network ( e.g. incidents, vehicle actuated traffic signals, etc.). Microsimulation models are typically developed for smaller areas of the network than network assignment models. A significant number of these model types have been developed for parts of the Scottish road network over the last 10 years including: Edinburgh city centre, Edinburgh city bypass and Forth Bridge approaches, the corridor studies (M74, M8 and M80), M8 (through Glasgow), Perth, Stirling, Ayr and parts of Dundee, Inverness and Aberdeen.
5.8 Static methods are generally based upon the idea of an 'area' speed/flow relationship that can be simply inverted to give an estimate of travel times for different flows on a network. This can also be linked to an equilibrium traffic assignment model and assumes a stationary state of congestion and continuous demand - in practice this may be criticised as unrepresentative of the real life instances of congestion.
5.9 Within each approach, the economic total cost of congestion is generally given by Delay multiplied by (Volume of traffic) multiplied by (Value of Time). The variation between the different approaches relates generally to:
- how vehicle delay is measured or estimated ( i.e. the definition of the baseline level of delay)
- how the volume of traffic is measured or estimated
- how the value of time is incorporated
- whether values for the environment, reliability, accidents are included
5.10 These factors will, in turn, relate to the scale at which costs are required - driven by the overall objectives of the study. The total cost of congestion is, however, only one measure of costs and in chapters 6 and 7 further elaboration of alternative measures is given (specifically marginal costs and the costs of the excess burden of congestion).
5.11 An example of the modelling of the supply curve in higher scale national models is that adopted in the UKFORGE model ( DfT, 2005). The fundamental basis is a representation of the relationship between supply, demand and cost of travel. This is illustrated in a 'cobweb pattern' cycle of iterations as shown in figure 5.1 below. The costs are essentially fuel and vehicle operating costs plus a monetary valuation of time costs. It is outside the scope of this report to consider optimal means of modelling congested networks, but there are criticisms of the traffic flow based approach (see Hills and Gray, 1999) and there is a train of argument which suggests it should be trip based and have a temporal dimension to accommodate departure time changes (see DfT 2001a).
Figure 5.1 - Supply and Demand Cobweb (source: DfT 2005)
5.12 The FORGE model produces output according to a number of 'congestion bands', but these are not readily available in the documentation, however, the fundamental basis for the congestion outputs is that of the difference between free flow travel time and actual travel time.
5.13 For a focused geographical area, such as a particular city in Scotland, it is feasible to establish a dynamic network model and even a microsimulation model which can give detailed outputs on particular links and junctions and represent the time dimension of congestion through changes in driver behaviour. For a larger scale estimate of costs, a national model would be appropriate - in the case of Scotland potentially based upon extensions to the Transport Model for Scotland. As discussed above, this may lose some of the degree of sophistication in picking up the dynamic time dependent elements of congestion. In terms of data requirements, the methods utilise the standard sources of data that form inputs to micro or macro level models i.e. loop counters plus household survey data or interview data to give information on trip purpose. Proposed Values of time are available from a number of studies ( see chapter 4) and can be used at disaggregate level to reflect a number of trip purposes according to geographical location. A summary of data and modelling considerations in measuring cost is given in Annex 3, based on work by Nash and Sansom, 1999.
5.14 It lies outside the scope of the work to analyse National Transport Models, but a number already exist at European level and these are reported in DfT (2001) - notably models for the Netherlands, Norway, Sweden, Denmark, Switzerland, Austria, Germany and Italy. In general these have similar objectives of measuring the impacts of policy and infrastructure measures. The model for the Netherlands is particularly concerned with policy measures intended to reduce traffic congestion.
5.15 A considerable tranche of research has been carried out at EU level over a period of 10 years or more into the marginal social costs of travel and transport pricing more widely (including implementation, pricing principles and consequences for transport market imbalances). Whilst this research extends well beyond the scope of this project into a much wider set of issues, an overview of the relevant work is given below.
5.16 An initial set of research ( e.g.PETS, QUITS, TRENEN) concentrated primarily on developing pricing principles, on measuring the various elements of marginal social cost and on case studies to model implementation impacts. Research then turned towards implementation issues, with projects such as AFFORD and MC- ICAM (which identified the need for policy packages and phased approaches, as set out in the 1998 White paper). The REVENUE project specifically examined the use of revenue from transport pricing whilst PROGRESS, CUPID and DESIRE were concerned with the practical issues of implementing road pricing in urban and inter-urban areas respectively. The RECORDIT project specifically focusing on intermodal freight transport costs and on the identification of policies and measures to reduce the current market imbalances between intermodal and all-road transport services. More recently, SPECTRUM has been concerned with the potential to move towards a greater use of economic policy instruments either alone or as part of a package (with regulatory or physical measures) in managing the transport network. The project was concerned with a comprehensive socio-economic assessment of benefits, rather than efficiency alone. UNITE has had three main objectives aimed at supporting the introduction of a fair and efficient pricing policy for transport across Europe. Firstly, to develop pilot transport accounts for all modes, for the EU15 and additional countries, secondly to provide a comprehensive set of marginal cost estimates relevant to transport contexts around Europe; and finally to deliver a framework for integration of accounts and marginal costs, consistent with public finance economics and the role of transport charging in the European economy. On-going research includes the GRACE project, which is concerned with researching improvements in the accuracy and reliability of social cost calculations, with particular emphasis on the water and air modes and on generalization issues. Part of the work of GRACE is to consider the complexities of urban road congestion and the consequences for modelling and estimating the costs as a result. Case studies are being carried out as part of the research, including a network model of Edinburgh. The approach will be iterative, applying a single model ( SATURN) to a range of pricing structures of various levels of sophistication, estimating the optimal pattern of tolls, interviewing road-users then amending and re-running the model. The definition of congestion within this work is that proposed by the DfT formed from the difference between the free flow and actual travel times. At the time of writing, the project is yet to report case study outcomes. Other on-going relevant research includes the HEATCO project, which is seeking to promote the harmonisation of social cost calculations, particularly in the framework of EU transport infrastructure investment decisions. The DIFFERENT project, only recently underway, is investigating the scope, feasibility and effects of differentiated pricing schemes. This tranche of work is likely to continue with further projects funded by the EU, for example with the recent invitation to tender on 'the impact assessment of the internalization of the external costs of transport' (TREN/E1/395/2006). All these projects have included case studies and some have included Edinburgh as an illustration - if the future direction of subsequent research by the Scottish Executive is towards fair and efficient pricing schemes, these would be relevant sources for more detailed information on questions such as internalization approaches, pricing levels and wider impacts of introducing economic measures within the transport sector.
« Previous | Contents | Next »