CHAPTER SEVEN MEASURING THE TOTAL COST OF CONGESTION AND EXCESS BURDEN OF CONGESTION
7.1 Aside from the marginal external cost of congestion, there are two other methods associated with calculating the cost of congestion. The first, the Total Cost of Congestion, has developed over the course of the last half century, whilst the second, identified here as the 'Excess Burden of Congestion', is more recent and whose development has occurred as a result of the increased interest in optimal transport investment decisions and road pricing. The principal difference between the two methods is that the Total Cost of Congestion approach has as its baseline a state of zero congestion; whilst the Excess Burden of Congestion has its baseline a situation in which the optimal amount of road capacity is provided. As will be drawn out in the discussion below it is not necessarily the case that the optimal level of road capacity is associated with a state of zero congestion - primarily because there are costs associated with providing capacity.
Total cost of congestion
7.2 The underlying approach associated with the Total Cost of Congestion ( TCC) method is that a visionary state of zero congestion is envisaged against which the current situation is compared. Table 7.1 summarises the estimates of the total cost of congestion and the methods employed as identified by our survey of the literature. As can be seen from this table a number of different methods have been used, though the age of some of these studies means that some of the specifics of the methods are slightly obscure.
7.3 The most frequently quoted estimate that congestion costs the economy £20 billion per year is an update of the £15 billion estimate calculated in a 1989 CBI study. The update reflects movements in prices over the intervening time period. It is however unclear as to exactly when and who undertook this update and no report has been identified. The 1989 CBI report uses data on the cost of congestion as a proportion of GDP (2.6% to 3.1%) taken from OECD analysis as its means of calculating the cost of congestion. We have not been able to trace the source of these OECD figures, though an OECD 1991 report (Bouladon, 1991 - cited in Quinet, 1994) identifies the cost of congestion as a proportion of GNP as 2.1% in France, 3.2% in the UK, 1.3% in the USA and 2% in Japan. Again the age of these studies means that it is unclear the exact methodology used to calculate the cost of congestion as a proportion of GDP or GNP.
7.4 The other studies set out in Table 7.1 use two broad methodologies. The first, adopted by Newbery (1995), Dodgson and Lane (1997) and Tweddle et al. (2003), is to use mathematical models to estimate costs in the current situation and in the uncongested situation. In all instances only link speed/flow based models are used, rather than the more sophisticated area speed/flow curve models, or network assignment models or microsimulation models (see chapter 5.3). The latter two model types can give a more accurate representation of junction delay. The final methodology adopted is one that uses actual measurements of vehicle speed to infer changes in journey time and was used by Trafficmaster (1996 and 1997) and the Scottish Executive (2005).
7.5 Whilst from the information that is available it is uncertain exactly how the OECD figures were calculated it does seem that in all the studies identified the marginal values of the different impacts of congestion ( e.g. values of time) were used. For example an estimate of the time lost due to congestion was made and then this was multiplied by the marginal value of time.
7.6 As can also be seen from Table 7.1 the different studies consider different impacts of congestion. Whilst saying that in all instances the costs of increased travel time are included. However some studies also include increases in fuel costs and other forms of vehicle operating costs, whilst the Trafficmaster study (cited by Santos (1999)) also includes the cost of missed deliveries and higher maintenance costs 4. As far as it is possible to tell no studies have included the reliability costs associated with congestion nor have they included the additional environmental or accident burdens that congestion can impose. There is significant variation between the estimates in the Total Cost of Congestion associated with the British road network. For example, the NERA study (Dodgson and Lane, 1997) estimate a figure of £7 billion whilst the Institute for Transport Studies study (Tweddle et al. 2003) estimate a figure of £15.2 billion. Both studies use similar modelling methodologies and both relate to 1996 traffic levels. Clearly small differences in modelling methods and assumptions can have a significant impact on the results. Interestingly the frequently quoted figure of £20 billion, with its suspect methodology ( i.e. it is not based on estimates of traffic delay), is comparable to the costs of congestion estimated in a more rigorous manner by the Institute for Transport Studies for 1998.
7.7 The Total Cost of Congestion approach to measuring the cost of congestion is not unique to the UK. The total cost of road traffic congestion in the 15 countries of the European Union is estimated at more than 120 billion euros a year ( EU, 2003) or by some estimates 0.5% of the EUGDP ( SUMMA, 2004). Every year the Texas Transportation Institute in the US estimates the cost of congestion in 85 of the largest urban areas in the US (Schrank and Lomax, 2005). Their latest estimate is that in 2003 the total cost of congestion was US$61.3 billion. This estimate includes delay costs and extra fuel costs only. Actual speeds are derived from reported traffic speeds in conurbations and compared to 'desired' speeds. Quinet (1994) in a survey also identifies similar studies associated with Japan (Osaka conurbation and Tokyo conurbation), France (Paris conurbation), Switzerland (Berne and Zurich) and the Netherlands. All of these studies compare some estimate of actual speeds/travel times to desirable or reasonable speeds/travel times.
Table 7.1 - Estimates of total cost of congestion in the UK and methods used
Methodology and comment
Glanville and Smeed (1958)
[cited in Goodwin (2004)]
£125M per year in urban areas,
£45M per year in rural areas,
£170M per year total
Delay only, but no allowance for non-working time to have a value
[cited in Goodwin (2004) and CBI email to authors]
£15 billion total per year for GB
(£5 per week per household per year)
The authors have not been able to obtain a copy of this report. However, the CBI indicated that the estimate is based on a report produced for the OECD which estimated the cost of congestion as a share of GDP, suggesting it lay in a range from 2.6% to 3.1%.
£20 billion per year (no date ascribed)
The CBI report that this often quoted £20 billion figure was produced "some years ago by updating the previous figure [the £15 billion CBI figure] to reflect movements in prices".
[cited in Goodwin (2004), Mumford (2000), Dodgson and Lane (1997)]
£19.1 billion per year for GB (1993 traffic levels and prices)
The authors have not been able to obtain a copy of this report. However, as reported by those who cite this study the method produces estimates of the cost of congestion for different road user types as well as a nationwide figure.
The approach adopted has been criticised ( e.g. by Dodgson and Lane) as providing an incorrect measure of the total cost of congestion as it "multiplies a marginal cost by a total volume".
[cited in Santos (2000)]
[sourced from internet press release]
£2.1 billion for 4 th quarter of 1996 (on motorways)
£1.5 billion for 1 st quarter of 1997 (on motorways)
Comparison of measured vehicle speeds in current year compared against measured vehicle speeds in the year in which the measuring devices became operational.
Costs reflect wasted time, extra fuel, missed deliveries and higher maintenance costs [as reported by Santos]
Dodgson and Lane (1997)
£7 billion per year for GB (1996 traffic levels and prices)
Comparison of costs at freeflow and estimated current speeds - modelled using link based methodology.
Time and vehicle operating costs (fuel and non-fuel).
£18 billion GB total (1999 prices)
A 'mid-point' of the CBI's estimate, Newbery's estimate and Dodgson and Lane's estimate updated to 1999 prices.
Tweddle et al. (2003)
£15.2 billion GB total (1996 traffic levels, 1998 prices)
£19.2 billion (1998 traffic levels, 1998 prices)
£24 billion (2005 traffic levels, 1998 prices)
Based on a comparison of estimated speeds and freeflow speeds. Traffic levels for 1998 and 2005 estimated by growing 1996 traffic levels.
Modelled using link based methodology.
Time costs only.
Scottish Executive (2005)
£71M per year over 10 areas of Scotland's trunk road network (2003 prices and traffic levels)
Measured speed compared to measured freeflow speed.
Time costs only
7.8 The Total Cost of Congestion approach, whilst being a reflection of the cost of congestion, has been criticised ( e.g. Goodwin, 2004) as not being particularly useful from a policy perspective. Primarily this is because the measure appears to imply that the British economy will be, say, £20 billion better off, or in the case of the Scottish trunk roads £70 million better off, from alleviating congestion. Clearly this will not be the case as any policy associated with alleviating congestion will have a cost associated with it. Additionally any reduction in congestion will reduce the impedance of travel and result in an increase in travel demand and average trip length - which will not only increase the environmental, accident and maintenance burden but may also lead to an increase in congestion above the zero congestion level. The Total Cost of Congestion measure is also criticised for the arbitrariness of its baseline. This is because the baseline reflects speed limits. As such transport policy changes in speed limits ( e.g. lowering speed limits in traffic management areas) can seemingly erase congestion, or correspondingly seemingly create congestion ( e.g. raising speed limits on motorways) when in fact there has been no change in operating conditions.
The excess burden of congestion
7.9 The final approach to measuring the cost of congestion can be termed the Excess Burden of Congestion. Such an approach has an important role in the road pricing debate as it reflects the benefits associated with a reform of road prices. It is also associated with the challenge of identifying the level of transport infrastructure capacity that maximises economic output. The Excess Burden of Congestion approach differs from that associated with the Total Cost of Congestion as at efficient prices and at an optimum level of capacity (the baseline) it is highly likely that congestion will be present on the transport network.
7.10 The excess burden of congestion arises because the prices faced by road users are not optimal and therefore demand and congestion levels are also non-optimal. For example, if prices are too low then demand will exceed economically efficient levels and there will be too much congestion. Technically the excess burden of congestion is what economists term the deadweight loss. It therefore relates to a situation where capacity is fixed. Clearly if capacity is also sub-optimal then even at efficient (optimal) prices there maybe too much congestion. Once prices are efficient ( i.e. reflect the full social costs of using the road) it is possible to develop simple investment rules to determine the optimal level of capacity: if the price for using the road is set above the cost of expanding capacity then this is a signal that capacity should be expanded (see for example Glaister and Graham, 2003; Dings et al, 2002). This is equivalent to the principle that the price for road use should be equivalent to short run marginal cost ( i.e. a charge equal to the marginal external cost of congestion should be levied on road users) and investment decisions should be based on social cost benefit analysis (Nash, forthcoming p2; Dings et al, 2002). Clearly there is an explicit trade off between the cost of investment in additional capacity and the benefits that that extra capacity will bring. If the benefits of reducing congestion are less than the costs of providing extra capacity some congestion will be present at the optimal level of capacity even at efficient prices. That is some congestion will be present at the level of capacity and set of prices that maximise economic output.
7.11 Table 7.2 sets out some of the studies that give estimates of the Excess Burden of Congestion at UK level. There are also numerous other UK studies that have looked at this problem at a city level ( e.g. academic related or government sponsored studies of London, York, Leeds, Edinburgh, Cambridge, Northampton, Hull, Lincoln, Norwich, Bedford, Hereford, Bristol, etc.) and there is also a substantial number of studies undertaken overseas. The primary difference between studies conducted at a national scale compared to those undertaken at a more local level is the nature of the modelling that underpins the study. The more tactical city wide studies typically use detailed network assignment models ( e.g. Santos, 2000) whilst the more strategic national studies use the simpler link based form of modelling ( e.g. Dodgson et al., 2002). Two things stand out from Table 7.2: the first is that the cost of congestion as measured by the Excess Burden of Congestion is substantial 5. Whilst substantial it is however significantly less than that measured using the Total Cost of Congestion approach. The second is that, in a manner similar to the results from the Total Cost of Congestion approach, there is a substantial variation in the estimates of the cost of congestion. Such a variation cannot be explained purely by the different years the estimates relate to. The results regarding the cost of congestion under both methods can therefore be seen to be heavily dependent on the values assumed for the external costs and the methodology used to model vehicle delay.
7.12 The studies outlined above are based on estimates of 'first-best' prices. That is the prices reflect the full marginal external costs of road travel. In practice such a charging structure would result in a myriad of different prices and for implementation reasons a more simple pricing structure would be required ( e.g. cordon charges as had been proposed for Edinburgh or zonal area charges as implemented in London). Such 'second-best' charges would not be expected to deliver the same level of benefit as first-best prices. Notwithstanding that it does appear that simplified charging structures if designed correctly can come close to delivering the benefits of a first best pricing scheme (see for example Shires (2006) for a discussion). There is also a substantial body of evidence that the manner that the revenue from a road taxation and pricing reform is used has strong implications for the efficiency, equity and acceptability impacts of the reform. Hypothecation of revenues to the transport sector appears to be one requirement for acceptability. A consequence of such hypothecation is that if there is a lack of good value for money transport projects, in which to invest revenue from road user charging, road prices may have to be set significantly lower than marginal external costs (to avoid generating surplus revenue) (see for example Tricker et al., 2006). The implication of these constraints on pricing reform imply that if the baseline for the Excess Burden of Congestion measure was defined to be a 'realistic' reform of transport prices rather than pure first-best prices, the cost of congestion estimates would be lower than those set out in Table 7.2.
7.13 None of the studies above have simultaneously considered transport pricing reform and investment in additional road capacity. The only study that has considered these issues simultaneously and within a rigorous framework at a national level is that undertaken by Dings et al. (2002) for the Netherlands. They demonstrate that for the Netherlands an optimal investment strategy would include a substantial investment in additional road capacity. Notwithstanding that they did find that the optimal level of capacity appeared to be lower than that set out in the Netherlands strategic plan (2010 to 2020). Their analysis also demonstrates that whilst capacity expansion does not increase welfare dramatically (once prices have been set to reflect the costs of congestion and on the environment) capacity expansion does bring about a substantial reduction in congestion (as measured by delays). Their results also demonstrate that some congestion would be present on the transport system at an optimal level of capacity.
Table 7.2 - Estimates of excess burden of congestion at the national level and methods used
Pricing reform only
Type of prices
Methodology to model delays
Costs of revenue collection included
Dodgson et al. (2002)
£2 billion per year (England)
(1998 prices and traffic levels)
Congestion charge additional to existing fuel and VED taxes.
Congestion charge varies by area, time of day, vehicle type and link type. Reflects delay costs only.
Delay, reliability (3)
Glaister and Graham (2003)
(a) £2.6 to £4.3 billion per year (1)
(b) £2.9 to £3.8 billion per year (1)(England)
(2003 Prices and 2000 traffic levels)
(a) Fuel tax replaced by congestion and environmental charge that varies by area, time of day, vehicle type and link type
(b) As (a) but mark-ups introduced to ensure revenue neutrality for the Exchequer
Delay, fuel, accidents, air pollution, climate change .
(a) £9 to £10.2 billion per year (2)
(b) £7.8 billion per year
(1998 Prices and 2010 traffic levels)
(a) Fuel tax replaced by congestion and environmental charge that varies by area, time of day, vehicle type and link type
(b) As (a) but revenue neutral for the Exchequer
Link speed/flow with variable demand modelling
Delay, fuel and non-fuel vehicle operating costs, accidents, air pollution, climate change, noise.
€17 billion (=£11.7 billion (4))
(2000 prices and traffic levels)
Fuel tax, VED, insurance tax replaced by congestion (including resource costs of parking) and environmental charge plus a charge that allows the government to recover lost VAT receipts. The charge varies by area, time of day, and vehicle type
Delay, fuel and non-fuel vehicle operating costs, accidents, air pollution and climate change.
Notes to Table
1: Range depends upon assumptions associated with environmental costs (low or high)
2: Range depends on number of different charges. The larger the range of charges the greater the benefit.
3: Reliability benefits assumed = 25% of delay benefits
4: June 2006 exchange prices 1 Euro = £0.687