This paper presents an empirical investigation of the effect of a preliminary breath test law on drunk driving behavior. A preliminary breath test law reduces the procedural problems associated with obtaining evidence of drunk driving and thus increases the probability that a drunk driver will be arrested. In 1985, only 23 states had a preliminary breath test law. According to the theory of deterrence, increasing the probability of arrest for drunk driving will reduce the future occurrence of this behavior. The data set employed to test the theory is a time series from 1980 to 1985 of cross sections of the 48 contiguous states. Four highway mortality rates are used as measures of drunk driving. The effect of the breath test law was estimated using four independent variable models and 12 dummy variable models. The four independent variable models were also estimated using Leamer's specification test. The purpose of using these alternative specifications and Leamer's specification test was to examine the breath test coefficients for specification bias. The econometric results show that the passage of a breath test law has a significant deterrent effect on drunk driving. Simulations with these results suggest that if all states had a preliminary breath test law, highway mortality could be reduced by about 2000 deaths per year
