The Scottish Government

« Previous | Contents | Next »

Listen

CHAPTER SIX MEASURING THE MARGINAL COST OF CONGESTION

6.1 As identified in chapter 5, there are several economic terms that can be rightfully called the cost of congestion, the first of which is the marginal cost of congestion. Chapter 7 discusses the two other terms that appear in the literature, that of the Total Cost of Congestion and the Excess Burden of Congestion.

Marginal costs

6.2 Marginal cost is an economic and financial concept and refers to the change in total cost that occurs when the quantity produced changes by one unit. It is a very useful and important concept as it illustrates the manner that, in the case of a transport system, total transport network costs change as vehicle-kilometres or numbers of trips change.

6.3 The marginal cost often differs from average cost (total transport network cost divided by number of trips). This is because the cost of producing an additional unit of output ( e.g. a trip or vehicle-kilometre) may increase ( e.g. as capacity is approached) or may decrease due to economies of scale, scope or density in the supply of the transport service. Marginal cost of road travel typically increases with each additional unit of demand, as roads become more congested, whilst that for rail travel may decrease with demand due to economies of density ( e.g. longer trains) and scope ( e.g. more services).

6.4 There is also a distinction between short and long run marginal cost. Short run marginal costs are those associated with keeping capacity fixed, whilst long run marginal costs allow capacity to be expanded (the cost of the capacity expansion itself forms a component of the long run marginal cost).

6.5 Marginal external costs are items of marginal cost that are not borne by say the trip maker. With respect to trips made by road they include road wear and tear, delays to other users, increased accident risk and environmental costs. When these are added to those costs borne directly by the user ( e.g. fuel, their own time) the result is called marginal social cost. One of the marginal external cost items is delays to other users and this in fact is often referred to as the marginal external cost of congestion ( MECC). The MECC specifically refers to user costs and does not include other cost items that may also change with levels of congestion ( e.g. accident risk and environmental costs). A number of well known authors use MECC (Walters, 1961; Glaister, 1981; Newbery, 1988; Button, 1993) in this sense. It does however appear that some authors use the term marginal cost of congestion and marginal external cost of congestion inter-changeably ( e.g. Dodgson et al., 2002; Shires, 2006). It is important to note that the MECC is defined as the external costs that are borne by the users of the transport system ( e.g. delay and reliability costs).

Components of marginal external cost

6.6 Table 6.1 sets out a categorisation of the marginal costs of a change in road traffic vehicle kilometres. The congestion category within this tabulation would strictly speaking include all non-monetary related user impacts including reliability impacts and the impacts on the quality of the driving experience ( e.g. stop/start conditions).

Table 6.1 - Definition of marginal external cost for road traffic vehicle kilometres

Notes to table
In the presence of imperfect economic markets positive consumption externalities ( e.g. agglomeration effects and imperfect competition in transport using sectors of the economy) would be a further cost category
Source: Samson et al. (2001)

6.7 There is a substantial literature on the calculation of the marginal costs of each of the cost categories in Table 6.1. A review of all these categories is beyond the scope of this report. The reader is therefore referred to Bickel et al. (2005, 2006) for reviews on environmental and safety costs and Link et al. (1999) for infrastructure costs. With respect to the marginal external costs of congestion, these costs arise as a result of delay to other users of the system and reliability impacts on other users. This and the linkage between changes in congestion and the economy are reviewed below.

Marginal Value of Time

6.8 There are countless examples in everyday life of people's willingness-to-pay to save travel time - think of the premium fare a high speed train service attracts. Clearly therefore time savings have value. So why do people and businesses value time savings? This apparently simple question has to be answered using many areas of economic thought including that of labour supply, home production and transport. From the perspective of businesses time lost for production costs money. Staff are paid for the time they work, including the time spent travelling which if lost for production is a cost to the business. Money is also bound up in stock inventories including that in distribution warehouses. Therefore transport improvements that help increase staff productivity or reduce stock inventories help improve business efficiency. Businesses recognise this and are willing-to-pay for the time saving ( e.g. by paying a premium for a high speed rail fare or paying for air travel rather than train travel). Individuals value savings in their personal travel time for a variety of reasons. A primary reason, similar to that of businesses, is that the time individuals spend travelling is lost to production- but in this case production is leisure activities and household business activities (washing, cooking and shopping). The improvements in in-car entertainment systems, mobile phones, lap-tops, portable DVD players all, however, make time spent travelling more enjoyable (or productive in an economic sense) for the individual. Such improvements in the 'usefulness' of personal travel time are cited as one of the reasons why empirically the value of non-working time has been observed to increase at less than the rate of income growth. The other reason why individuals value travel time savings is that individuals operate within a time budget. There are only 24 hours in a day, some of which has to be spent asleep, at work and engaged in household production tasks. This leaves limited time for travelling to access locations for work, leisure and household related activities. Thus the choice set of possible workplaces, schools, swimming pools, cinemas, retail parks, etc. is limited by travel time, particularly when some of these activities have to be undertaken at or between set times. Reductions in travel time can therefore increase individuals' choice regarding the activities they undertake and this increased choice is of value.

6.9 There is a substantial volume of evidence on the marginal value of travel time. Wardman (2001) identified 143 value of travel time datasets in the UK, of which 2 relate to the 1986 and 1994 UK national value of time studies. The latter of which is the basis of the current appraisal values for the UK. The values set out in appraisal guidance ( DfT, 2005) range from £10.18 to £44.69 per hour for people travelling during the course of work, whilst the average values for commuting trips is £5.04 per hour and other non-working trips is £4.46 per hour. These 'average' values for commuting and other non-work trips belie a very large range. Such values vary systematically by income, distance, age, gender and household type (see for example Whelan and Bates, 2001). The main determinants of the variation are however income and distance (Mackie et al, 2003 p30).

6.10 Whilst it is fairly apparent that the value businesses place on travel time savings lead to business efficiency savings, it is less clear how such savings translate into increased profitability as companies re-structure, re-organise, expand output and change the size of their workforce (including reducing the size of the workforce as travel time savings can increase labour efficiency). The impact of savings in non-working travel time on the general economy ( e.g. through a reduction in congestion) are even more opaque. The retail and service sectors rely on customers accessing their premises to sell their products and all businesses rely on their workforce accessing their premises. Clearly therefore changes in non-work travel time affect the wider economy but the extent of this affect is not clearly understood. What, however, is understood is the social welfare ³ value that businesses and individuals place on changes in travel time. Travel time savings therefore form one of the inputs into a social cost-benefit analysis.

Marginal Value of Time spent in congested conditions

6.11 Time spent in congested conditions can be more onerous on the traveller than time spent travelling in freeflow conditions. This arises because of the increased burden placed on the driver of the vehicle and from the irritating effect of stop-start conditions. Reliability problems also increase in congested conditions. A number of studies have therefore set out to differentiate the value of travel time by whether the travelling is undertaken in congestion or not.

6.12 Wardman's meta analysis identified that travelling in congested conditions is valued 48% more highly on average than time spent driving in free flow traffic; Eliasson's Swedish study found similar values (about 1.5) for driving in queues (Eliasson, 2004), whilst Steer Davies Gleave (2004) found values ranging from 1.2 times in-vehicle-time (for busy conditions/light congestion) to almost twice in-vehicle-time for 'gridlock' conditions. The UK value of time study found that travel time in congested conditions was about 40% higher than in free-flow conditions for commuters though only just significant at the 95% level, whilst no significant effect was found for the 'other' non-work trip purpose (Mackie et al, 2003, p31). This led to a recommendation for further research in this area, rather than a recommendation that values of time in congested conditions should be increased. Outside of Europe the recent New Zealand value of time study and guidelines suggest that high levels of congestion may lead to values of time savings between 1 and 1.5 times in-vehicle-time depending on the degree of congestion and whether the congestion occurs on urban or rural roads.

6.13 It should be stressed that these aggregate values for time spent in congested conditions implicitly include the values for reliability that are discussed below. Including both the value for time spent in congested conditions and the value of reliability would double count the economic impact of reliability.

Marginal Value of Reliability

6.14 One of the impacts of congestion is reliability problems. Reliability, or lack of, is considered to impose a significant cost on business travellers and commercial goods traffic (see for example SACTRA, 1999; McQuaid et al., 2004). Travel time variability and large unexpected delays are two of the consequences of reliability problems. The distinction between them is that travel time variability is considered 'predictable' as it occurs from day to day, whilst it is not possible to attach a probability to the likelihood of an 'unexpected delay'. The distinction is therefore slightly blurred, as essentially they are both forms of uncertainty in travel time. In contrast to the value of travel time, the value journey time reliability is not well understood.

6.15 The main body of the literature on the value of reliability (VoR) relates it to the value of travel time (VoT) through a reliability ratio ( RR). The value of reliability (VoR) can be calculated by multiplying the value of travel time by the reliability ratio ( i.e. VoR = VoT x RR). The reliability ratio concept gives a relationship between one minute's standard deviation of travel time and one minute's travel time. A reliability ratio of 1 implies that a reduction of the standard deviation of travel time of 1 minute has equal value to a 1 minute travel time saving. A reliability ratio of one is recommended by the Department for Transport - though it is noted that the evidence on this matter is of variable quality ( DfT, 2003). Other studies have found a quite a range in the reliability ratio, from 0.35 to 2.4 (see literature reviews of Noland and Polak, 2000; Eliasson, 2004; De Jong et al., 2004a). In a workshop of international experts convened by AVV, the transport research centre of the Dutch Ministry of Transport, some consensus regarding reasonable reliability ratios for passenger transport was reached (Hamer et al., 2005) (see Table 6.2). No consensus on a reliability ratio for commercial goods traffic was reached. Kouwenhoven et al. (2005) have since derived a reliability ratio for commercial goods traffic. This has been derived from the Dutch guidelines on the value of change in the percentage of goods that arrive on time (see Table 6.3).

Table 6.2 - Reliability ratios

Notes to table
Source: Hamer et al. (2005), Kouwenhoven et al. (2005)

6.16 Research has found that the value of unexpected large delays is typically quite high, however, with the exception of one study, Eliasson (2004), this research relates to unexpected delays experienced on public transport and not by road. Eliasson in a large Swedish study found values around 3.5 times the value of in-vehicle-time (per minute of delay) for car drivers.

6.17 For commercial goods VTTS, reliability is treated explicitly by some of the most up-to-date studies, e.g. de Jong et al (2004b), Vandaele et al (2004), Bruzelius (2001). For example, the results of de Jong et al (2004), for the Netherlands indicate that a 10% change in reliability, measured as the percentage of deliveries not on time, can be valued as shown in Table 6.3.

Table 6.3 - Values of a 10% change in reliability (de Jong et al, 2004)

Notes to table
Converted to 2002 € at PPP factor prices by Bickel et al. (2005, p143)
Source: de Jong et al. (2004)

6.18 Another common approach is to recommend a multiplier on the value of expected travel time savings, to represent reductions in delay time. Typically factors of 2.0-2.5 appear in the literature. Bruzelius (2001) put forward a specific factor, 2.0, but also suggested that further research is required in order to validate it for use. Fowkes (2001,p7), cites evidence gathered on behalf of the Highways Agency in the UK, that the ratio of the value of delay time to expected goods travel time is in the region of 2 for chemicals, paints, food, drink and groceries, and 3 for other commodities. It seems that the commercial goods VTTS is sensitive to the nature and value of the goods being transported.

6.19 At this point in time there is still uncertainty as to what the value of reliability is for both personal and freight related travel. However, there can be no doubt, given the qualitative and increasing quantitative evidence, that these values can be significant and large. Unfortunately a still more significant challenge exists once values for reliability have been identified, that of forecasting how reliability will change as a consequence of a transport policy ( e.g. motorway widening). As evidenced by the UK work in this field (Ove Arup and Partners et al., 2004) this is a far from trivial task. Furthermore methods have yet to be developed for peri-urban and urban areas and for complex freight distribution chains.

Marginal Economic Impact

6.20 In the last decade there has been an increasing policy interest in the productivity impacts of transport. Through transport efficiency improvements the productivity of the economy can increase. In text book economics there is an equality between the economic benefits that occur in the transport market (time savings, reliability improvements, etc.) and the economic impacts that are felt in the general economy (including productivity gains from efficiency improvements). That is the marginal economic impact of reducing congestion would be the sum of the marginal values of the different congestion related impacts ( i.e. the sum of time savings, reliability benefits, etc.). Such an equality, however, relies on a number of technical economic conditions relating to perfect economic markets. The consequences of departing from these conditions are now the subject of some debate. If these conditions do not hold then for example agglomeration benefits may occur as may additional benefits in the labour and product markets. There is no direct evidence on the impact of congestion per se on agglomeration and other wider economic impacts. However, the fact that reduced levels of congestion imply quicker journey speeds it is possible to utilise the evidence base on the impact of journey speeds to understand the impact that congestion has on the wider economy. There is a small but growing evidence base that changes in regional density, through , increased journey speeds, can have a significant effect on regional productivity (Rosenthal and Strange, 2004; Rice and Venables, 2004; Graham, 2005). Rice and Venables estimate for the UK that the agglomeration economies from a 10% reduction in commuting time will lead to an increase of 1.12% in labour productivity. Graham estimates an average elasticity of productivity to effective employment density of 0.04, though this disguises significant variation by region and industrial sector. An elasticity of 0.04 implies that if employment density (number of people living within a certain journey time) increases by 10% productivity would increase by 0.4%.

6.21 In a review of the available evidence on the additional economic impact that imperfect markets might have on total economic impact, Laird et al. (2005) find a range of -15% to +147%. That is total economic impact is -15% to 147% higher than that measured using a conventional economic appraisal ( i.e. travel time savings and reliability improvements). It should be noted that the upper end of the range is only associated with projects that have a very significant impact on accessibility ( e.g. a new high speed rail network/line).

6.22 Table 6.4 identifies twelve studies that have considered the marginal external costs of congestion. In the main the driver for these studies has been the road pricing agenda and most of these studies report the marginal external cost of congestion in the presence of a road user charge. Because a road user charge will alter demand levels and therefore congestion the marginal external costs of congestion with a road user charge in place are not the same as without a road user charge in place. Only Samson et al. (2001) who estimates marginal external costs for roads in Great Britain (for 1998) and the DfT (2004) who updated Samson et al.'s figures to a 2000 price base and different forecast years, publish estimates of marginal external costs that relate to a situation without road user charges in place. These are re-produced in Table 6.5 and Table 6.6 respectively. Annex 2 reproduces the optimal congestion charges ( i.e.MECC at optimal demand levels) calculated by a set of studies, including those in Table 6.4, reviewed by Shires (2006). As can be seen from Table 6.6 congestion forms the largest proportion of quantifiable external costs - estimated to be around 77 per cent in 2000 increasing to around 88 per cent of external costs in 2010. Accident and emissions costs account for the remainder and, unlike congestion costs, are forecast to fall over time. Figures in Table 6.6 are averages, i.e. 7.3p represents the extra cost of the 'typical' additional vehicle anywhere on the road network. Marginal external costs will vary widely across the country, with time and place, in line with congestion and other externalities. The potential environmental costs such as biodiversity and landscape were excluded in the calculations due to lack of data.

6.23 As far as it is possible to tell from the study reports that are available it appears almost all of the studies have included monetary values for environmental impacts (noise, air pollution, climate change), accidents, vehicle operating costs and travel time delays due to congestion. None of the studies appear to have included reliability impacts in their estimates nor have they included benefits or dis-benefits associated with agglomeration and imperfect markets.

6.24 As Shires (2006) identifies the different transport modelling methods used to model congestion costs can give rise to differing results in the estimates of the marginal cost of congestion. One would expect the more aggregate modelling techniques (link speed/flow and area speed/flow) to be approximations to the techniques that explicitly account for junction delays ( e.g. network assignment and microsimulation). Where junction delays are important elements of congestion costs one might expect the largest divergence between these aggregate and disaggregate modelling methods. Similarly assumptions regarding behavioural responses to increased delay have a fundamental impact on the marginal cost of congestion. This is because the calculation of the marginal cost has to be calculated with the aid of a model from simulations of network user costs and different levels of demand. Shires (2006) also identifies that the marginal external costs of congestion can differ dramatically between similar sized cities and between countries, even when the same modelling methodology is applied (see for example Milne, 2002). In part this is due to the different levels of congestion in the cities, stemming from a mixture of topology, historical development of the network and economic development. These differences make it very difficult to transfer results from one city to another ( e.g. Edinburgh to Glasgow) or even to disaggregate results from a higher level down to a more disaggregate spatial level ( e.g. from Great Britain to Scotland).

Table 6.4 - Comparison of studies

Source: Shires (2006) and authors' research

Table 6.5 - Road sector marginal external costs Great Britain 1998

Source: Samson et al (2001)

Table 6.6 - Estimated marginal external costs and tax paid by road users (£b)

Source: DfT (Devereux et al, 2001)

« Previous | Contents | Next »