Session: Regulation
Room: Hollister 306
Time: Tue 08:30-10:00

Presenter: Darren Grant (Sam Houston State University. Economics)

Discussant: Andrew BarnesUniversity of California, Los Angeles

Abstract

This paper has three aims. First, it describes the dynamics of drunk driving in the U.S. over the past generation, as revealed in fatal crash data. Second, it develops an alternative method for estimating the effects of drunk driving laws on fatalities, via their effects on drunk driving, and uses it to evaluate four major drunk driving laws. Third, it explores the relative importance of law and social forces in reducing drunk driving.

Among drivers involved in fatal accidents in the United States, blood alcohol concentration conditional on drinking has remained essentially unchanged. Thus the evolution of drunk driving in the United States over the past generation can be described using simply the fraction of accident-involved drivers who have been drinking (called HBD). Nationally, this variable fell “from the outside in,” initially declining most rapidly among the youngest and oldest drivers, with middle-aged drivers catching up later.

At the state level, HBD is highly serially correlated, as are fatalities, and declines at different rates in different states, as do fatalities. But the variation in fatalities is far higher than that in HBD, and a statistical decomposition indicates that innovations in HBD are correlated with innovations to fatalities that are not drinking-related. These properties suggest that analyses relating drunk driving laws directly to fatalities are likely to be biased and more variable than those relating laws directly to HBD.

A simple nonlinear relationship between HBD and fatalities is developed in a basic model of traffic fatalities. Panel regressions of HBD on four major drunk driving laws imply that these laws' effects on fatalities are at the low end of the range of estimates reported in the literature, suggesting an upward bias in direct fatality regressions. Further investigation reveals that this bias is particularly large in the states that adopt these laws first; this bias does not appear in the HBD regressions. This is particularly important because evidence from early-adopting states is utilized by other states’ legislatures in deciding whether to adopt these laws themselves, and by Congress in deciding whether to encourage their adoption.

The net effect of the four laws considered, along with the dynamic properties of HBD, suggest that the role of law in reducing drunk driving over this period has been relatively modest, and leaves most of the variation in HBD unexplained. It may be that drunk driving responds more to "social forces" than to the legal disincentives. Various evidence on this point is surveyed and discussed.

Key Terms

drunk driving, dynamics, traffic safety legislation

Authors:

Darren Grant (Sam Houston State University. Economics)