External Accident Costs of Motor Vehicles Revisited

Staff paper given by Lyn Martin to the 28th Australasian Transport Research Forum, 28-30 September 2005, Sofitel Wentworth Hotel, Sydney

Road accidents impose a huge burden on the Australian community, estimated to be over $40 million per day, representing a major cause of death and disability (BTE 200 p. xi). It is widely argued road users do not face a sufficient share of this cost burden to ensure that they will act in the community interest. This paper explores the source of the possible disparities between the interests of individuals and those of society—disparities that would generate external accident costs.

This paper does not estimate the external costs of accidents. Rather, it reviews the range of definitions of external accident costs—from the popular to the theoretical—and examines the policy implications of each. The paper is aimed at promoting more critical discussion of this potentially significant component of road-user charges.

Accident cost is a function of accident rate and accident severity. Severity is influenced by speed and rate by the number of ‘vehicle passings’. The impact of these two factors tend to work in opposite directs, as increased traffic levels are often associated with a reduction in average speed. This will be explored more fully later as it has implications for determining the external component of accident costs.

An ‘economically correct’ approach to road safety would involve some mix of safer roads, safer cars and safer driving with the level of expenditure on each option stopping at the point where the marginal returns were equalised across the options. As the system currently stands, jurisdictional boundaries limit the scope for achieving an appropriate balance.

The focus of this paper, the external accident costs of road use, starts from this position.

High-end estimates for Australia seem to be around $5 billion (Stanley 2002) and

$7 billion per annum (Laird 2003), while the Australian Automobile Association put forward

a figure of $2.5 billion per annum (AAA 1998).

1 Bureau of Transport and Regional Economics (BTRE). Views expressed in this paper do not necessarily reflect those of the BTRE.

2 Given current technology, it is unlikely to be in the community interest to completely eliminate road accidents.

The higher the external accident costs and the greater the share of transport externalities

accounted for, the more significant accident externalities become as policy issue. Much of

the European analysis tends to conclude the external accident costs and a major cost

transport external accident costs accounted for almost 30 per cent of the (non-congestion)

SOURCE: INFRAS/IWW 2000

While there are widely disparate approaches to estimating external accident costs, most

analysts agree that they vary significantly with location/road type, vehicle type and traffic

flows. In one study, the European Commission observed that ‘the total accident externality

charge is likely to be over ten times higher in a central city area than on interurban roads

(EC 1999 p. 11). The estimates from the INFRAS/IWW 2004 study, presented in Table 1

cars range from €3.9/1000 vehicle km (Sweden, motorways, low estimate) to €161.3

(Netherlands, urban roads, high estimate)—41 times the lower estimate.

Perhaps surprisingly, the study suggests that passenger cars have higher external

accident costs than trucks (‘heavy duty vehicles’). Both the upper and lower estimate of

external marginal accident costs for passenger cars is higher than for trucks: €2.5

Switzerland)

(Sweden, motorways, low estimate) compared with €3.9 at the lower end and €73.3

SOURCE: INFRAS/IWW (2004) P. 92

Wide variations in external accident costs have major policy implications simply because

a tax on fuel or on distance traveled is not likely to provide an accurate mechanism for

internalising such costs.

Ultimately, it is the definition of external accident costs that is adopted that determines

both the total value and the variation. Commonly-used definitions of external accidents

range widely and still appear to be in ‘the state of flux’ that Cox identified more than a

decade ago (Cox 1994 p. 189).The various definitions and their implications are explored

in more detail below.

The following definitions are not mutually exclusive and some can be regarded as separately

identifiable components of total external accident costs. The order of presentation roughly

reflects popularity of the definition in general usage.

External accident costs are commonly defined as a fixed proportion of total costs,

generally ranging from 10—50 per cent (e.g. BTRE WP 40 p. 63 and Cox 1994, p. 274).

Externalities Study 1994, acknowledged that the ‘proportion of accident costs which is

borne externally is not known’ (EPA 1994, p. 13) but adopted 30 per cent as the external

component of accident costs. Parry (2001 p. 11), also acknowledging that it was ‘difficult

to pin down the split between private and external costs’, used a range between

10 per cent and 50 per cent

When adopting such a definition of external accident costs the final figure will be heavily

dependent on the estimate of total accident costs. There is a great deal of uncertainty

surrounding these total cost estimates since costing accidents is not an exact science—

estimates can vary widely depending on the two key parameters: the value of life and the

discount rate.7 Since human costs generally account for over half the accident costs8, a,

say, doubling of the value of life would increase any estimate of accident costs by 60 per

cent—and, accordingly, the external component if it were defined as a fixed proportion of

total costs.

4.1.1 Determining value of life

There are two major methodological approaches to valuing human life:

earnings; and

the risk of death or injury.

While initially providing some rationale for developing early valuations, more recently the

human capital approach has generally been criticised on a number of grounds, including

the unfortunate fact that it means that the life of individuals beyond the working age may

be accorded zero value. Small (1999) also argued that the human capital technique is

discredited because ‘it attempts to value the full transition between life and death rather

than the small changes in risk that people actually face’.

Feldman added his criticism of the human capital approach noting that:

…while it [human capital method] measures earning capacity, it does not measure how much

the deceased valued his own life. To measure how much people actually value their own lives,

it is said, we should see how much they are willing to pay to avoid fatal risks (Feldman 1997,

p. 1).

The perceived shortcomings of the human capital method have caused many policy

makers to embrace the willingness-to-pay methodology, with a concomitant increase in

the value of life—sometimes by an order of magnitude.

(1997 p. 2) noted that ‘the studies get wildly differing estimates for the value of life, and

many absurd results’. Professor Ross Parish (n.d) argued that, for a number of reasons,

the willingness-to-pay estimates have very little credibility. In particular, that they assume

that workers and consumers respond rationally to very small variations in the risk of

death, despite the fact that ‘people concerned have no way of knowing these

probabilities’. Parish concluded that ‘putting a credible economic value on human life is

not possible’ (p. 17) and that the main choices facing policy makers are to adopt some

political valuation or to restrict analysis to cost-effectiveness studies.

For the sake of the current exercise, it is sufficient to recognise both the almost-arbitrary

element in the value-of-life estimates and the influence that the estimates have on the

external costs of accidents if, as often happens, the external component is assumed to be

a fixed proportion of the total costs of accidents.

This definition of external accident costs as the difference between insurance premiums

and total accident costs draws on the logic that if insurance serves as the mechanism to

internalise the external costs of accidents and if the total cost of accidents exceeds the

total insurance premiums, the difference must, by definition, be external. This definition

was adopted by the Inter-State Commission (ISC 1990 vol. 1 p. 163) and is implicit in the

work of other researchers.

However, insurance is only one mechanism for internalising costs. Insurance represents the

market response to the legal liability that has its basis in Tort law—through which victims of

accidents may recover their damage costs from those that cause accidents.

The fact that an individual may not be insured does not automatically mean that such

damages cannot be recovered. Road users may opt to carry some risk themselves—to

‘coinsure. This is common for many types of insurance reflecting an aversion to

catastrophic costs but a willingness to bear moderate costs if insurance premiums are

reduced to reflect the share risk. To count only the insurance premiums would completely

ignore the residual costs borne by the road users.

The presence of heavy vehicles on the road (whether trucks or other passenger vehicles)

increases the severity of accidents for lighter vehicles and non-motorised road users.

Whether or not they are at fault in the accident, a heavy vehicle will cause significantly

more damage than a light vehicle in the same collision. This has given rise to a definition

of external costs based on the damages suffered by ‘less protected users’, in particular by

pedestrians and cyclists, regardless of fault. Some countries, such as the Netherlands,

have enshrined this principle in legislation. In the event of an accident between motorised

and non-motorised road users, the legal presumption is that motorised vehicles are at

fault. In 2002, the European Commission attempted to introduce similar legislation across

the European Union. The Commission argued that motor vehicles cause most accidents

and, regardless who is responsible, pedestrians and cyclists usually suffer more.

While it is easy to be sympathetic with the argument, some practical obstacles exist. If the

attribution of fault is the combination of actual presence in the vicinity of the road and the

larger size, then consistency would demand that roadside buildings and other ‘street

furniture’ were also labeled as ‘generators of external accident costs’.

There would also appear to be a theoretical inconsistency with this allocation of

responsibility. If the aim in ensuring that motorised vehicles face their marginal external

accident costs is to encourage more socially responsible behaviour, then it could be

argued that non-motorised road users should also face their marginal external cost. If one

group of road users were absolved of responsibility for their behaviour because they were

less protected than the other party involved in the accident, then the pressure for them to

act in a socially responsible way would be diminished.

Also, there are other mechanisms in place that take account of differential vehicle

weights—in the tort and the criminal law system. Where an insurance market is working

effectively, the fact that a truck can do considerable damage to lighter vehicles (and cars

to unprotected road users) should be reflected in the insurance charges facing the

particular accident, there are still specific penalties that can result from each accident. For

conscience and reluctant to risk their own lives in an accident. Proponents of this ‘differential

vehicle weight’ argument could still argue that the driver does not face the cost of the life he or she

might take. However, consider the case of the pilots of a Boeing 747-300. Is it necessary that they

face the real cost of all the 470 deaths if the plane crashes?

instance, criminal charges could apply if the driver had been responsible for a death. A

lesser offence could involve fines and penalties that could accumulate to the loss of a

licence, putting a driver’s livelihood at risk. There will also be insurance penalties through

In general, the mix of traffic of different weights is a wider problem than the truck and

other road-user interface. The disparity of vehicle weights on a particular road reflects the

benefits and costs of providing specific roads for different weight classes. Truck-only

lanes have been introduced on some high-density routes in the US and Europe. However,

due to our lower traffic densities, that is unlikely to be an option for the near future in

Australia.

There are two aspects of the costs imposed on non-road users: the personal and the

‘system costs’. The personal costs include grief to victims’ loved ones, property damages,

lost income, medical costs and reduced quality of life to non-users.

System costs cover subsidised hospital, medical and policing costs associated with road

externalities (ISC 1990 p. 89). Elvik (1994) defined system externalities as one of three

components of external accident costs estimating that they account for around 30 per cent

included in costs imposed on non-road users.

The logic behind defining costs imposed on non-road users as externalities is

straightforward. It is reasoned that participation in the road-use activity implies a full

understanding of the risks and related crash costs. Road users are regarded as

‘consenting adults’ who, as a group, bear the costs of accidents. As a consequence, there

not consented that are external costs. Two of the five studies examined by Gómez-Ibáñez

for automobiles of around one (US) cent per mile.

However, there are some stumbling blocks to this interpretation of external accident costs.

First, as with congestion, while road users may bear the costs as a group, it is still

important to ensure that the individual road users face incentives that reflect the costs

they impose on others, whether the ‘others’ are motorists or non-road users.

With regard to property damage, there are avenues (although imperfect) to recover these

costs through insurance and the legal system. The residual costs are likely to be

civil action is the only recourse, and this is generally not very satisfactory. However, with modern

technology allowing low-cost detection of unregistered vehicles, this should become a declining

problem.

the-art cost estimates to encourage the development of a sustainable transport policy through

the adoption of fair and efficient pricing of transport infrastructure use. For more details see

14 Physical injury externalities and traffic volume externalities (discussion following) were the other components.

15 This is analogous to an argument often made with regard to congestion.

reduced quality of life to non-uses can be accommodated within the legal framework and

the resultant insurance system.

Grief to victim’s loved ones is a clearly an important issue. Lindberg (2001 p. 41) noted

that although only a few studies were aimed at estimating relatives’ and friends’

valuations, a figure of around 40 per cent of the value of statistical life seemed to be

justified. However, it is not clear whether Lindberg had taken into account the

mechanisms that currently exist to internalise those losses: scope for friends and family to

separately insure the drive and insurance taken out by the driver that may already

The question of whether the cost of emergency response and crash prevention

expenditures should be regarded as an external accident cost is slightly more complex. It

is not made clear why road users should be singled out to cost recover these costs when

a community choice appears to have been made to fund them through the tax system. In

the words of one analyst:

…unwanted distributional consequences of trying to internalize all costs of traffic injury must be

weighed against the loss in efficiency entailed by a system that protects people from paying the

full costs of an injury (Elvik 1994 p. 729)

A secondary consideration is mechanisms currently operate the recover some of the

system costs from road users. In the event of accidents, there is often provision for the

costs to be reimbursed from compensation payments that may follow from a legal

settlement. Another method is the imposition of a tax on motor vehicle insurance. In some

countries, such as France, car insurance tax is specifically designed to cover some social

security expenses linked to road injuries (INFRAS 2000, p. 144).

From an economic perspective, if a vehicle entering a traffic stream changes the expected

prices signals facing the new entrant, then an accident externality will be generated. The

accident cost of the entering vehicle has two distinct components: the costs faced by the

driver and those imposed on other road users. The U.S. Transportation Research Board

described this as follows:

The expected accident loss from the added trip is the risk of accident involvement that the

If the accident cost of the added trip (the marginal accident cost) increases or decreases

the average accident cost, then an externality would be generated—which could be

positive or negative. An increase in average accident cost arising from the entrant of an

extra vehicle would imply an external cost while a reduction would imply an external

16 While this would be market solution, it is not to say that it would the most efficient solution. The

Coase Theorem (Coase 1960) suggests that if property rights were clearly defined and

enforceable and bargaining costs low, then the market outcome (whether the property owners or

the road users bear the roadside property risk of an accident) would be efficient. However, in the

absence of such conditions, societies generally resort to laws that assign liability to road-users.

17 At least one major insurance company reports that it is possible to take out life insurance on

anyone that is an Australian citizen and over the age of 16.

18 The term ‘risk’ appears to have been used as a shorthand for ‘expected cost’ through abstracting from variations in the severity of accidents.

4.5.1 Functional relationship between accident costs and traffic volume

If accidents are a random event and drivers take no adjusting behaviour, then one would expect

the number of rod accidents to increase as the square of the traffic flow (Dickerson, Peirson

and Vickerman in Lindberg 1999 p. 13).

There has been some modelling of the relationship between accident rates and vehicle

estimated that the marginal accident rate was 1.5 times the average rate, indicating an

served by exacting from every individual involved in an accident, a payment equal to the

full amount of the damage so inflicted, regardless of fault or relative contributory

Elvik (1994) found that the presence of an accident externality was not so predictable

concluding that the ‘relationship between traffic volume and the number of injuries is not

known in sufficient detail to determine whether the marginal external costs of traffic injury

reviewed the empirical modelling of the relationship between road traffic accidents and

traffic flows, noting Newbury’s conclusion that the number of road accidents is

proportional to flow, based on the 1982 US Federal Highway Cost Allocation Study. From

their own analysis they concluded that the accident-flow relationship is non-linear and

varies significantly between different road classifications and traffic volumes—while the

externality is typically close to zero for low to moderate traffic flows, it increases

substantially at high traffic flows. Also, at some point of increasing traffic flows, the

Lindberg 1999 p. 13).

Recent research by Edlin and Karaca-Mandic (2003) examined the extent to which an

increase in traffic density in a given (U.S.) state increases (or decreases) insurance

premiums. By using insurance premiums as a proxy for accident costs the researchers

concluded that the marginal external accident cost could be twice the average although in

some low density states there may be no accident externality (p. 19).

The functional relationship between traffic volume and accident costs determines the level

and nature of traffic volume accident externalities. If the expected accident costs are

proportional to traffic flow, then there would be no externality. An external cost is

accident rates and severity are sensitive to traffic volumes. For elaboration, see Gómez-Ibáñez (in

inconsequential.

(Vickrey 1968 p. 4).

generated if expected accident costs increase faster than traffic volume—an external

benefit if accident costs increase slower than traffic volume.

There are two broad institutional mechanisms for internalising the external accident costs

of road use: insurance and the legal system. Enforcement of the road rules represents an

Presumably the penalties associated with behaviour reflect the higher accident risks

associated with the behaviour. Stricter enforcement of road rules is technically possible. It

is possible that the greatest gains (in terms of internalising accident costs) would be

achieved from basic measures, such as ensuring that all drivers were licensed and

vehicles were covered by minimum insurance. With modern vehicle identification

techniques this should soon (if not already) be both technically and economically feasible,

with only the political will being the final requirement.

The second institutional mechanism for internalising the accident costs of road use is the

legal system, principally the law of tort. This provides the basic framework for ensuring

that road users bear the cost of the damage they impose on others. Whether a system of

An aversion to risk and catastrophic costs encourages many road users to voluntarily

seek insurance to deal with these legally-enforced responsibilities, on top of the

compulsory third-party insurance. However, insurance has been criticised because it

protects transport users against accident risks, thereby changing their behaviour (e.g.

Litman 2003) Elvik (1994 p. 279) found that while the evidence was conflicting, it generally

held that measures that reduced accident costs borne by individual road users tended to

increase the number of accidents.

It is also argued that the up-front nature of the insurance payments, only vaguely related

to accident risk, fails to send accurate signals about the day-to-day cost of road use. As

one commentator observed:

…once the lump sum insurance cost of admission to car registration is paid insurance cannot

put cost pressure on decisions about whether or not to take a trip…auto insurance provides

‘‘unlimited free mileage’ (Butler 1996)

It is generally accepted that there is significant scope for more finely differentiated and

more actuarially-correct insurance premiums. Much of Vickrey’s influential 1968 paper

focussed on the potential contribution of improvements in the structure of insurance

payments. While there have been some improvements since 1968 in that a few insurance

However, the lump sum payment of insurance does not completely abrogate the variable

accident cost of road use. There still remain considerable incentives to avoid accidents,

such as payment of excess, loss of no-claim-bonus and the personal risk and nuisance

factor associated with accidents. Also, there are many goods and services where

consumers can choose between fixed or variable costs or a combination of both. The

behaviour or the actual cost.

literature reveals little concern over the fact that many consumers choose the fixed cost

options for homes, cars, gym membership and mobile phones, thereby reducing their

variable cost and risking the ‘over consumption’.

There is appeal in the argument that improving the insurance system represents an

efficient option for internalising the external cost of accidents. The scope for more

actuarially-correct insurance charges will be increase significantly when global positioning

systems become standard equipment. The Final Report of the Expert Advisors to the High

Level Group on Infrastructure Charging concluded that:

…a sophisticated insurance structure based on detailed risk information and reflecting the

different variable components of the costs of accidents is the most satisfactory available means

of internalising the external costs of transport accidents (Lindberg 1999).

A pertinent question for policymakers is whether there are currently any administrative or

The legal system represents an important mechanism to internalise external accident

costs. The rigorous enforcement of the road rules, particularly the apprehension of

unregistered and unlicensed drivers, would reduce the external component of accident

costs. However, such enforcement is not costless—either in a budget or political terms.

Accident costs are an important component of road transport costs. If, as some analysts estimate, a significant proportion of these costs are external, then the level of road use would be higher than is in the community’s interest. Even if motoring taxes offset these external costs, it is unlikely that the structure of the taxes would correct the signals facing road users.

The estimation of external accident costs is very complex, raising both conceptual and measurement problems. In the absence of clear direction from the theory, it is understandable that simple definitions are often utilized. The range of popular measures include a share of total accident costs, the difference between total accident costs and total insurance premiums, costs imposed on non-road users (system costs being the main ones) and costs arising from the disparate vulnerability of road users. Most of these measures incorporate the concept of fault.

There is a three part message for policy makers. Facilitating improvements in law enforcement and in the insurance system would encourage the internalisation of external accident costs. The recovery of system costs is a separate issue that needs to be approached with a wider framework in mind. The most difficult area is the traffic volume externalities. This is an area requiring more research into the functional relationship between traffic flow and accident costs and into the link between congestion externalities and traffic volume externalities.

27 It was necessary to introduce enabling legislation in Texas before Progressive’s Pay-As-You-Drive insurance system could be introduced.

http://www.atsb.gov.au/road/mgraph/mgraph15/index.cfm

http://www.aaa.asn.au/submis/pre-2000/tax.htm

http://www.btre.gov.au/docs/r102/r102.pdf

Implications for Women, National Organisation for Women, Proceedings, Women's Travel

Issues Second National Conference, Chapter 39, 737-751. Federal Highway Administration

www.fhwa.dot.gov/ohim/womens/chap39.pdf

Edlin, A (2002) Per-Mile Premiums for Auto Insurance, Economics Department, University of

Economics Department, University of California, Berkeley, Working Paper E03-332,

for Internalization, Accident Analysis and Prevention, vol. 26, No. 6, pp. 719-732.

http://www.tc.gc.ca/programs/environment/economicanalysis/cost/docs/SoccostRpt.pdf

a common transport infrastructure charging framework in the EU, Brussels, July 1998,

http://europa.eu.int/comm/energy_transport/en/lb_en.html

art, in Greene, David L., Donald W. Jones, and Mark A. Delucchi (eds.), The Full Costs and

Benefits of Transportation: Contributions to Theory, Method and Measurement. (Berlin) Springer-Verlag, pp. 149-172.

Gómez-Ibáñez, J, Tye, W. and Clifford Winston, eds. (1999) Essays in Transportation

Economics and Policy: A Handbook in Honor of John R. Meyer. Washington, DC: Brookings Institution. (GTW)

Greene, D, Jones, D and. Delucchi, M (eds.) (1997) The Full Costs and Benefits of

Transportation: Contributions to Theory, Method and Measurement (Berlin) Springer-Verlag,

Costs of Transport in Western Europe. Zurich/ Karlsruhe.

http://www2.uic.asso.fr/d_environnement/envissues/full-report.pdf

Scheme, Volumes 1 and 2.

Advisors to the High Level Group on Infrastructure Charging (Working Group 3) April 27,

http://europa.eu.int/comm/transport/infr-charging/library/crash-cost.pdf

Road and Transport Research Institute (VTI), S-781 27 BORLÄNGE, SWEDEN paper

presented at the UNITE seminar Paris September 2001

www.its.leeds.ac.uk/projects/unite/paris/lindberg.pdf

http://www.ingentaconnect.com/content/lse/jtep/2001/00000035/00000003/art00004

http://www.rff.org/rff/Documents/RFF-DP-99-14-REV.pdf

Mayeres, I, Proost, S, Vandercruyssen, D, De Nocker, L, Int Panis, L. Wouters, G and. De

http://www.belspo.be/belspo/home/publ/pub_ostc/mobil/rapp04syn_en.pdf

University of California Energy Institute. Policy & Economics. Paper EPE-003.

http://repositories.cdlib.org/ucei/policy/EPE-003

Importance of Other Distortions with the Transport System,

http://www.rff.org/~parry/Transport/cong%20tax.pdf

Handbook in Honor of John R. Meyer. Washington, DC: Brookings Institution. (GTW), pp 137180.

presentation to Metropolitan Transport Forum Seminar, 20th July,