VI. Studies Supporting the Risk Compensation Hypothesis for Motorcycle Helmet Laws Introduction
In its crude form (Wilde (1982)), the risk compensation hypothesis states that individuals have a target (equilibrium) level of risk that they try to maintain. Thus, the implementation of helmet laws will lower the actual level of risk for a group of individuals (non helmeted riders). It is hypothesized that these riders will respond behaviorally by increasing their risk level to its target through other types of risky activity (i.e. higher driving speeds, alcohol consumption, more risky driving patterns, etc.). It is argued that this behavioral response can offset the positive affects of helmet laws on motorcycle safety.
The more sophisticated variant (Peltzman (1975) recognizes that the equilibrium level of risk is variable. In particular, regulatory legislation that lowers risk typically has two competing effects -- an income and substitution effect -- on an individual's response. Thus, total risk can increase or decrease in response to a regulatory act. In the case of motorcycle helmet laws, the law reduces the probability of a bad state -- injury and lost productivity -- and thus increases the expected income of individuals. The law also reduces the cost or price associated with driving intensity (i.e. high speeds) because of expected reductions in injury. In response, the individual uses the extra expected income (income effect) to buy more of all goods including more safety and more driving intensity but, because the price of driving intensity declines the individual will "buy" an additional amount of driving intensity because it is cheaper (substitution effect). The overall effect on safety depends on the size of the competing safety and driving intensity "purchases".
Thus the risk compensation effect of helmet laws becomes an empirical question. The measurement of such effects is a difficult empirical problem. Graham and Lee (1986) and Adams (1983) attempt to measure this effect. Both studies suggest that a risk compensation effect exists. The former study shows that a 2.5% increase per year in the fatality rate follows the initial 12% decline in the fatality rate from enactment of a helmet law. Thus within 5 years the fatality reducing benefits of helmet laws are eradicated. Adams (1983) casually argues that risk compensation responses could explain some of the stylized facts about motorcycle accidents and he cites other studies on automobile safety equipment and driver response as supporting evidence.
The risk compensation hypothesis is different but not inconsistent with the more risky behavior hypothesis (of non helmeted riders) discussed in Section I. The former argues that individuals have target levels of risk, the latter argues that different groups of riders (helmeted and non helmeted) have different target levels of and thus behave differently. While risk compensation is an interesting subject, good estimates of this effect are hard to come by.
1. Graham, J. D., and Lee, Y. (1986). "Behavioral Response to Safety Regulation: the Case of Motorcycle Helmet-Wearing Legislation." Policy Sciences, 19: 253-273.
As argued above the finding that the initial decline in the fatality rate of 12% after passage of a helmet law is eroded at a of 2. 5% per year is generated from a misspecified regression equation and thus is potentially biased. In this case the bias could be upward or downward. Thus, a reliable estimate of the risk compensation effect has not be generated.
2. Adams, J. G. U. (1983). "Public Safety Legislation and Risk Compensation Hypothesis: The Example of Motorcycle Helmet Legislation," Environment and Planning C, 1: 193-203.
Adams' paper has two parts. First he critiques the NHTSA and the Watson et al. (1980) studies reviewed above. Second he advances the risk compensation hypothesis, but offers only the we of empirical support for it.
The criticism of existing studies is noteworthy. The critic analysis focuses on the treatment of the data used by these studies rather than on the statistical techniques employed to analyze the data. NHTSA studies are criticized for the level of aggregation data and Watson et al. are criticized for the smoothing technique they apply. In both cases , Adams claims that the manner of data manipulation leads to overstated estimates of helmet effectiveness.