The Effect of No-Fault Automobile Insurance on Driver Behavior and Automobile Accidents in the United States
Jan 1, 2001
New Research Controls for Bias that Tainted Previous Studies
A new RAND study refutes a common criticism of no-fault auto insurance—that it may increase the accident rate by reducing drivers' incentives to drive carefully. An analysis of accident trends in the United States between 1967 and 1989 found no statistically significant relationship between states' adoption of a no-fault system and the fatal accident rate, overall accident rates, and other measures of driver care.
Thirteen states currently either mandate no-fault auto insurance or allow drivers to choose between no-fault and tort insurance. No-fault auto insurance requires individuals to carry personal injury protection (PIP) insurance that compensates them, regardless of fault, for economic losses sustained in automobile accidents and prohibits individuals from suing for non-economic damages such as pain and suffering unless their injuries exceed an established threshold. In contrast, under a traditional tort-based auto insurance system, all drivers must carry liability insurance that compensates third parties for injuries caused by the insured driver. First-party insurance under tort is typically voluntary.
Proponents of no-fault auto insurance claim the system delivers speedier and more-equitable compensation at lower cost than traditional tort insurance. However, its opponents frequently counter that no-fault lowers the cost of driving negligently by limiting first-party liability for the injuries suffered by third parties in auto accidents. Under such a scenario, no-fault would increase the accident rate and also lead to higher auto insurance costs.
Previous empirical work in this area has yielded contradictory findings. Several studies have estimated that the adoption of no-fault auto insurance increased fatal accident rates; one study claimed rates would rise by as much as 10 percent annually. Others have drawn the opposite conclusion. The confusion surrounding this issue arises from differences in the years (the periods around or following the implementation of no-fault) and countries the studies analyzed (including Australia, New Zealand, and Canada, as well as the United States) and from important differences in their empirical methods.
None of the previous studies used data that span the years in which states first implemented no-fault laws (1971 to 1976). Consequently, these studies are open to the criticism that they failed to control adequately for characteristics correlated with both the adoption of no-fault and the fatal accident rate. States that adopted no-fault auto insurance in the 1970s were not a random subset of all states; they adopted no-fault for a reason, and that reason could be correlated with fatal accident rates in those states.
The RAND study compares differences in fatal accident rates between tort and no-fault states both before and after the 1971 to 1976 period of no-fault implementation. In both tort and no-fault states, fatal accidents per vehicle miles traveled (VMT) fell steadily over the study period 1967 to 1989 (see the figure). These fatal accident rates fell particularly sharply between 1967 and 1978, probably because of factors such as greater seat-belt use, declining rates of drinking and driving, and heightened vehicle and road safety.
The key question is whether fatal accident rates fell more in tort than in no-fault states before and after implementation of no-fault laws. If they did, no-fault laws may have caused fatal accident rates in no-fault states to decline less rapidly than they otherwise would have. Such a finding would indicate that no-fault laws caused fatal accident rates to be higher than they would have been under tort.
RAND's analysis controls for differences both across and within states and over time in the level of urbanicity, climate, and per capita income and, importantly, state-specific pre-implementation time trends in fatal accident rates. The analysis finds that the fatal accident rate fell 3 percent less in no-fault states than in tort states, a statistically insignificant difference. Therefore, it is reasonable to conclude that the implementation of no-fault laws did not influence the fatal accident rate in no-fault states. Another empirical approach using within-state variation in dollar thresholds for non-economic damages also rejects the hypothesis that no-fault auto insurance influences fatal accident rates.
Given how far removed fatal accident rates are from the immediate impact of no-fault on driver behavior and the idiosyncratic nature of many fatal accidents, perhaps it is not surprising that the data do not reveal an empirical relationship between no-fault and fatal accident rates. Consequently, the RAND study also examines whether no-fault affects the overall accident rate and whether it affects driver negligence more generally. Past researchers have not examined the overall accident rate because of poor data. Many accidents go unreported to the police and, more significantly, the incentive to report accidents to the police might vary between tort and no-fault states. This study circumvents this problem by employing a proxy measure for the overall accident rate—the ratio of property damage claims to property damage exposure. Whereas this measure no doubt underestimates the overall accident rate, it is measured consistently across tort and no-fault states and over time. Controlling for state characteristics, this analysis finds that accident rates are no higher in no-fault states than in tort states over the period 1976 to 1998.
Even though no pre-implementation data on accident rates are available, RAND argues that any bias introduced into these estimates should be in the direction of finding a positive relationship between no-fault and the accident rate. Thus, one can be reasonably confident that the finding here of no effect would hold were data on accident rates prior to 1971 available. In addition, an alternative empirical approach using within-state variation in dollar thresholds for non-economic damages also rejects the hypothesis that no-fault auto insurance influences accident rates.
A final empirical test of the no-fault hypothesis described here was conducted using data from the Department of Transportation's Fatal Accident Reporting System (FARS), a census of all fatal accidents in the United States between 1975 and 1998. Among the elements recorded in the FARS data are indicators for whether some traffic violation or other negligent behavior on the part of any driver precipitated the accident. These elements include speeding, improper lane changing, failure to stop or signal, unsafe passing, and other negligent actions. FARS also reports on the blood alcohol content of drivers involved in accidents when available.
RAND used the FARS information to determine whether the incidence of negligence as a contributory factor in fatal accident rates is higher in no-fault states than in tort states. Controlling for detailed characteristics of the accidents themselves, about 58 percent of drivers involved in fatal accidents in tort states were classified as having engaged in a negligent act, compared with 54 percent in no-fault states. Using a different test that controlled for additional accident and driver characteristics, RAND found that the overall negligence rates are about 2 percentage points lower in no-fault-states. Consequently, the hypothesis that no-fault increases rates of driver negligence in fatal accidents cannot be supported.
If no-fault lowers the level of care drivers exercise, it must be the case that no-fault lowers the expected cost of getting into an accident. Although a rigorous empirical test of this hypothesis is not possible with available data, there is little reason to believe that no-fault as implemented in the United States significantly affects expected accident costs.
Presumably, the overriding reason to drive carefully is self-preservation, an incentive that does not vary between no-fault and tort states. Moreover, no-fault insurance creates many of the same incentives to drive carefully as tort insurance. First, no evidence exists that insurance companies are less likely to penalize drivers for accidents they may cause under no-fault insurance. Getting into an accident in a no-fault state is just as likely to increase insurance premiums as getting into an accident in a tort state. Second, property damage is covered under third-party liability in all states but Michigan. Therefore, expected property damage costs should be the same in both tort and no-fault states.
Previous research frequently cites the limitation no-fault places on liability for non-economic damages as a major reason why no-fault lowers the incentive to drive carefully. Several problems undermine this argument, however. First, no-fault insurance systems in the United States do not shield all claims from non-economic damages (typically less than 75 percent of all claims). Furthermore, the expected cost of non-economic damages to the at-fault driver in tort states is quite low (median non-economic damages amounted to roughly $1,600 in 1997) and, in any case, these damages are typically insured. In comparison with self-preservation and the desire to avoid traffic citations, the margin over which no-fault might change the incentive to drive carefully seems small indeed.
Given the weakness of the theoretical link between no-fault and expected accident costs, it is somewhat surprising, then, that the question of the incentive effects of no-fault has played such a powerful role in the debate over no-fault in both state legislatures and in the federal arena.