Most modern drunk driving legislation, such as an increased drinking age, lower blood alcohol thresholds, and zero tolerance laws, has been adopted using a quasi-federalist process, in which these laws are first implemented in a few states and then, after being evaluated favorably, diffuse nationwide with the encouragement of the federal government. However, the favorable conclusions of many such evaluations, which project large reductions in traffic fatalities as these laws are adopted in other states, have not borne out in practice.

Accordingly, this paper examines how the political and bureaucratic arms of the federal government incorporated the early evidence on the effects of key drunk driving laws into the formation of policy. Through a careful analysis of testimony in multiple Congressional committee and subcommittee hearings surrounding these laws’ passage, the paper examines how accurately these actors assessed the contemporaneous evidence and identified potential biases or flaws in the evidence. It then explains why these assessments were much more optimistic than the extant evidence could support.

Early studies of these laws, limited in the number of law-changing states and the amount of post-law data, used generally accepted, though simple, empirical methods whose limitations were recognized. Longstanding themes in social science, such as the “endogeneity of laws,” that were well-known to traffic safety officials articulated why any bias thereby generated was likely to be favorable. In contrast, position papers by safety advocates, testimony by government agencies, and debate in Congress selectively cited the available evidence, leading to an overall assessment that was more favorable than indicated by the body of work available at the time.

The root cause for this outcome is an intellectual bifurcation between a more policy-focused group of analysts estimating short run effects of new laws in the states that adopt them first, and a more academically-focused group that estimates laws’ long-run effects long after they have been widely adopted. The less-favorable findings of the second group of analysts do not strongly temper the findings or assessments of the first group, which are far more heavily represented in these Congressional hearings. (Meanwhile, the needs of the first group of analysts–effect estimates given limited data–do not influence the research agenda of the second group.)

This bifurcation is sustained and augmented by two supplementary factors. The first is the inherently adversarial nature of an inherently political process by which these laws are evaluated. This generates opposing camps that, for the laws we study, are not evenly matched in technical or political skill. As a result, the weaker side does not effectively check the excesses of the stronger side. The second is that the executive agency responsible for traffic safety has a small base of in-house expertise for assessing traffic safety legislation, contracting out most of its studies to the first group of analysts. This serves to reduce the inflow of dissenting judgements that would question the efficacy of these laws. We show that this arrangement furthers that agency’s interests.