Background. Traffic accident costs are a significant component of total road transport costs.
I have been examining proposals for introducing distance-based car insurance charges as a way of addressing traffic accident issues as well as environmental externalities associated with road use. There are arguments for utilising per kilometre (km) insurance charges rather than lump-sum annual fees. This is consistent with the principal that prices should reflect costs. Such charges more accurately reflect the risks of traffic accidents than do risk-adjusted fixed charges since they directly limit distances travelled. Spinoff benefits of such measures include reduced congestion and local pollution.
Per km charges could be imposed, in states such as Victoria, by modifying the Transport Accident Charge (TAC) that is levied as a compulsory component of the annual vehicle registration charge to cover third-person damages. The overall registration charge could be based on a per km charge that is both driver and vehicle specific. Political objections to the scheme could be eliminated by offering motorists the chance to stick with a higher fixed charge or utilising the per km charging. Comprehensive insurance covering personal accident costs and property damage costs could be provided privately on the same basis. Increasingly insurance markets are offering such options.
Traffic accidents impose significant potential private costs on road users. Road users can die, suffer injuries or incur damage costs as a consequence of traffic accidents. As a response to such costs a motorist can refrain from driving, exercise greater care while driving or insure to cover themselves against possible damages that occur as a consequence of driving. Such costs are then internalised by the motorist and, provided they can choose, also internalised by passengers travelling in the motorist’s vehicle.
Some traffic accident costs however are not internalised. When an extra car joins a traffic stream the accident probabilities of all other motorists in that traffic stream change because of the extra possibility of colliding with the extra motorist. While, for the most part, we might expect the possibility of collisions would increase – the evidence supports this view – traffic might slow as densities increase and driver care may increase with increased densities thereby reducing the rate of serious traffic accidents and deaths.
There are other traffic accident externalities as well – a pedestrian struck by a vehicle is an instance of external cost.
To the extent that there are externalities these are analogous to standard traffic congestion externalities and need to be priced away. If negative traffic accident externalities are unpriced individuals will undertake a socially excessive number of journeys and will generate a socially excessive number of traffic accidents.
There are no careful estimates of traffic accident externalities for Australia. The estimates that have been made are typically arbitrarily assumed to simply be some fixed proportion of total accident costs – for example, 10-50 per cent (Martin, 2005). These total costs themselves are subject to controversy because, for example, of the intrinsic difficulties in valuing a human life. The lack of Australian evidence on the extent of traffic accident externalities is a key gap in Australian transport economics research and inhibits policy discussion. Inferences need to be drawn on the basis of international experience that may not reflect local conditions. Moreover, the international evidence on such costs is mixed although some US studies, discussed below, suggest they are very large.
Unlike congestion externalities when time and place of travel condition social costs accident externalities are driver-specific. For example, young men are disproportionately represented in traffic accident collisions.
One approach to pricing traffic accident externalities – and a much discussed potential policy reform – is to set a fixed charge for car registration that reflects the person-specific probability of an accident and then add to this a variable charge that depends on how far the motorist travels. This proposal is discussed below.
Empirics. Traffic accident costs are a significant cost of vehicle use everywhere. In Australia, in 2007, 1,616 persons were killed in 1,466 road crashes. This absolute number of deaths is less than half the figure that prevailed in 1981 so there has been a long-term absolute decline. This decline was concentrated in the late 1970s and early 1980s. Total deaths however have not changed much since 2003 although deaths as a fraction of population have decreased. The road crash death rate during 2007 was 7.7 deaths per 100,000 population compared to 22.3 deaths per 100,000 in 1980 (DITRELG, 2008). In 2005 comparable figures on deaths per 100,000 were 5.5 for the UK and 14.7 in the US (BITRE, 2008c) so that Australia’s accident mortality experience is somewhere between that of these two countries. Apart from a secular decline of about 65 per cent in Australian deaths per capita there has also been an even greater proportionate decline in deaths per hundred million vehicle km travelled. This has fallen from 3.55 in 1976 to 0.76 in 2006 (DITRELG, 2008). Thus, even though road users are driving further, they are experiencing much fewer fatal accidents per km of distance travelled.
Fatal traffic accidents in Australia peaked in the 1970s and have declined steadily since that time as have serious traffic accidents generally. The fall in accident rates is due to seatbelt use, to reduced drink-driving and improved vehicle and road technology. Pedestrian deaths have fallen partly because fewer people walk. In addition accident costs are a function of the frequency and severity of accidents with higher traffic densities increasing the number of accidents but plausibly reducing the severity of accidents due to decreased average speeds and increased driver caution.
Putting dollar values on the costs of traffic accidents is important if sensible road safety investments are to be made. The value of human lives lost can be quantified using ‘human capital’ and ‘willingness-to-pay’ methodologies.
There were also many injuries, property damages, net losses of production as well as pain and grief (BITRE, 2008c) which can be costed if, in some cases, only approximately. Connelly and Supangan (2006) provided the most up-to-date estimates of the total costs of road traffic crashes in Australia, in 2003 which they estimated as $17b ($47 million per day) or 2.3 per cent of that year’s GDP. They find substantial intra-national variations in accident costs by state with costs varying from between 0.62-3.63 per cent of Gross State/Territory Product. Death rates are particularly high in the Northern Territory where until 2006 speed limits were not imposed on highways outside metropolitan areas. The average Australian cost proportion they estimate is lower than US estimates which put costs there at 4.3 per cent of GDP (Parry et al., 2007). A new set of cost estimates for Australia will be provided by the Bureau of Transport and Regional Economics in 2009. These will provide the first comprehensive official estimates since 1996.
Traffic accident frequency is known to be related to the age and gender of drivers (as mentioned, young males have high accident rates), to alcohol consumption and other drug consumption per capita, use of mobile phones and other distracting devices and to other influences such as weather conditions. A somewhat surprising empirical regularity is the simple, strong positive dependence of accident costs on traffic densities. Indeed the work of numerous authors (e.g. Litman, 2008) suggests that accident rates increase with density at an increasing rate.
In principle one might expect congestion, the effects of density in encouraging greater care in driving and other such phenomena to potentially reduce accidents. In fact however such effects do not seem to be strong enough to offset the link between increased density and the accident rate. Note that the accident rate only partially determines aggregate accident costs because average accident severity also needs to be accounted for. It needs to be determined how accident severity is related to traffic density. The intuition might be that at higher densities since traffic speeds are reduced that accident severity might fall.
Accident externalities. For individuals injury risks in single vehicle crashes to the driver and car occupants are internalised if those travelling make rational transport choices. The bulk of crashes however are multi-vehicle – in the US multi-vehicle crashes as a fraction of the total exceed 70 per cent, see Edlin and Karaca-Mandic (2006, p. 933). Thus there is an issue of whether such costs are internalised or not. EM argue that when two cars crash, although only one party may be negligent in causing the accident, the accident would not have occurred had either driver had not travelled. In this sense both drivers cause the accident. Motorists however pay for and internalise only the average cost of the accident, not its marginal cost, so there is a substantial auto accident externality. Motorists pay too little for driving and hence, in the absence of a requirement to pay the accident externality cost, drive too far.
The Edlin and Karaca-Mandic (2006) study (hereafter EM) provocatively emphasised what it saw as the vast scale of automobile externality costs in the US economy.
EM estimate accident externalities using panel data on state average insurance premiums and loss costs. These external costs are substantial in traffic-dense states though negligible in states where densities are low. In traffic-dense California an increase in traffic density due to a single extra driver raises total state-wide insurance costs by $1725-$3239 depending on the car model. This adds to the $744 in premiums the average driver paid for insurance in 1996. A Pigovian tax to internalize this externality would raise $44 billion annually in California which is more than all state taxes combined. A national corrective tax would raise $113 billion (Edlin & Karaca-Mandic, 2007).
Moreover, the EM externality cost estimates are likely to be an underestimate since damages are ignored in accidents where parties are uninsured and costs of delays due to accidents are ignored. If uninsured costs behave as do insured costs then EM estimate that accident externalities could be 3.5 times those estimated or $10,000 per motorist annually in California.
While building more roads to reduce congestion would reduce the externality this is cost-ineffective because of the scale of the required expenditures involved to reduce densities to those prevailing in low traffic density states.
Insurance and other reforms. While various Pigovian tax bases (per mile, per driver, per litre of fuel) might deal with the accident externality issue identified by EM their preferred solution is to tax car insurance premiums and to take moves that will force motorists to be fully insured. This is a heterogeneous tax base reflecting the accident and skill experience of drivers. Implementing this proposal in California would require a 200-400 per cent levy on the premium EM recognize will probably not work politically. Such a premium also has the undesirable feature of encouraging drivers not to insure.
EM suggest a second-best compromise which leaves overall driving costs the same but increases the marginal cost of driving with insurance premiums rising with distance travelled as per-km premiums. People would then have an incentive to cut insurance costs by driving less which is the outcome sought. Per-km premiums are fair since they impose lower charges on those (like women and senior citizens) who drive less and who cause fewer accidents. The charges can be made driver-specific with high per-km premiums on high accident risk drivers who live in high traffic density areas. Edlin (2003) has estimated US gains from ‘per km’ premiums of $12.7b per year in the US.
One difficulty with this proposal is that the chance of having an accident depends on when one travels as well as how far one travels. If driving at certain times involves encountering few cars the probability of an accident involving another vehicle is low. The ideal solution would be to seek congestion pricing of travel and to add to congestion tolls a charge that reflects the contemporaneous accident externality. Such a charge would be much higher in peak periods than a simple congestion toll.
Parry et al. (2007) criticise EM on the grounds that the externality evidence is based on insurance data which mainly covers property damage. While this captures some of the effects of accident severity property damage costs are only 14 per cent of the Parry et al. (2007) estimated social costs of accidents. EM in fact only provide inconclusive evidence on whether fatal traffic accidents – which account for 33 per cent of the Parry et al. (2007) estimated social costs – increase with traffic density.
Parry et al. (2007) estimate marginal external costs inclusive of pedestrian and cyclist injuries and come up with a figure of 1.2-4.8 cents per km US for the year 2000 which turns out to be 13-44 per cent of average accident costs. Samson et al. (2001) provide estimates for 1998 UK marginal external costs of from 0.82-1.40 pence per km which is around 1.2-2.8 cents per km US. These figures may provide some clues as the likely scale of Australian costs.
Alternatively the 200-400 per cent premium figure provided by EM provides an alternative method of estimating the implied accident externality costs in Australia ignoring the issue of uninsured motorists and such things as time delay costs caused by accidents. Accepting the Connelly and Supangan (2006) estimate of the costs of accidents in Australia as 2.3 per cent of GDP this suggests a tax on insurance costs of 4.6-6.9 per cent of GDP, an enormous figure. This is obviously a rough ballpark figure only and ignores reductions in driving that will occur as a consequence of the tax. Indeed there may also be less high-density driving in Australia and fewer multi-vehicle accidents.
Distance-based Vehicle Insurance. This proposal for distance related insurance charges can be seen as part of a package of reforms that private insurance companies might seek to or be induced to seek as distance-based vehicle insurance policies (Litman (2008), Parry et al. (2007)). It is a practical ‘second-best’ policy that forgoes the ideal of seeking to charge the marginal accident cost by taxing distance travelled at relatively low transaction cost.
A charge levied on distance alone will be more effective than an increased excise on fuels since the behavioural response to it will involve less driving. Litman (2007) estimates the charge as around 6 cents US per mile for the average motorist which is about 7.5 cents Australian per km. The actual figure charged to a motorist would depend upon the motorist’s (and possibly their vehicle’s) relative risk factor which would be determined by an insurance company.
The basic idea is that reduced travel as a consequence of distance-related insurance charging implies reduced accident risks. Distance-based insurance options give motorists the incentive to reduce their travel and hence to reduce accident costs and insurance premia. This also provides side benefits of reducing fuel usage and congestion.
Moreover, while people who drive long-distances are likely to be relatively capable drivers who drive relatively safer vehicles still evidence supports the claim that a 10 per cent reduction in mileage will reduce crashes by 17 per cent. Thus those who drive only short distance pay too much with fixed insurance charges.
There are various pricing options such as using estimated mileage as a rating factor, paying a surcharge at the petrol pump to fund basic vehicle insurance, prepaying for a certain mileage cover, per minute premiums and using GPS-based pricing. There are implementation costs of such schemes although all seem to satisfy basic cost-benefit tests even are estimated transaction costs of $10-$150 annually (Litman, 2008). Private insurance firms have adopted specific schemes already and can be expected to increasingly offer such products as technologies for monitoring vehicle mileage improve.
EM claim that this solution might need to be regulated since individual insurers may not adopt per km premium policies on their own since gains are external (they accrue to other insurers) while the monitoring costs are internalised. In fact, however as Litman (2008) shows, some firms are already developing such policies and there is even limited endorsement of them in Australia by firms such as Apia. Apia has adopted what is effectively a distance-related premium by charging those over 50 less on the grounds that they ‘drive less’. Reader Conrad points out that the firm Hollard already provides pay-as-you-drive insurance services in Australia.
The possibility of linking third-party insurance with distances travelled by imposing a ‘second-best’ compulsory excise levy on petrol has the advantage of simplicity and helps bring the uninsured into the payment system. Disadvantages of this proposal are that it fails to distinguish driver heterogeneity. In addition fuel-efficient vehicles pay lower charges even if they impose the same accident costs. Fuel surcharges are most effective in targeting reduced petrol consumption while pay-as-you-drive insurance targets specifically target distance travelled. But Parry (2005) shows that distance-based charges outperform excises on fuels once congestion, local accident and accident reduction benefits are accounted for.
Bureau of Infrastructure, Transport and Regional Economics (BITRE), Australian Transport Statistics June 2008, June, 2008c.
L.B. Connelly & R. Supangan, ‘The Economic Costs of Road Traffic Crashes: Australia, states and Territories’, Accident Analysis and Prevention, 38, 6, 2006, 1087-93
A.S. Edlin, ‘Per Mile Premiums for Auto Insurance’ in R. Arnott, B. Greenwald, R. Kanbur & B. Nalebuff (eds), Economics for an Imperfect World: Essays in Honour of Joseph E. Stiglitz, MIT Press, MA, 2003.
A.S. Edlin & P. Karaca-Mandic, ‘The Accident Externality from Driving’, Journal of Political Economy, 114, 5, 2006, 931-955.
A.S. Edlin & P. Karaca-Mandic, ‘Erratum: “The Accident Externality from Driving”’, Journal of Political Economy, 115, 4, 2007, 704-705.
T. Litman, ‘Distance-Based Vehicle Insurance as a TDM Strategy’, Victoria Transport Institute, Victoria, BC, Canada, 18 November 2008.
L. Martin, ‘External Accident Costs of Motor Vehicles Revisited’, Staff Paper, Department of Transport and Regional Services, 28th Australasian Transport Research Forum, September 2005.
I. Parry, ‘Comparing Alternative Policies to Reduce Traffic Accidents’, Resources for the Future, 2005.
I. Parry, ‘Is Pay-As-You-Drive Insurance a Better Way to Reduce Gasoline than Gasoline Taxes’, American Economic Review, Papers and Proceedings, 95, 2, 2005, 288-293.
I. Parry, M. Wells & W. Harrington, ‘Automobile Externalities and Policies’, Journal of Economic Literature, 45, 2, 2007, 373-399.