Who Responds to What Incentives?

Speaking of alcohol...

Ordinarily, one would expect that lowering the blood alcohol content [BAC] limits that apply to automobile drivers would lead to reduced driving under the influence [DUI]. Maybe it does, but not always, according to a recent paper by Carpenter and Harris [abstract available at this link; guest registration available for the entire paper].

Over the past two decades, states have toughened their adult drunk driving laws by setting the legal blood alcohol content (BAC) threshold at .08, down from .10. Although several studies have shown that these laws have been effective at reducing alcohol-related traffic fatalities, there is very little evidence on the underlying behavioral mechanisms through which .08 BAC laws achieve the fatality reductions. We estimate reduced form models of the effects of .08 BAC laws on a wide range of self-reported alcohol-related risk behaviors using large samples from the Behavioral Risk Factor Surveillance System (BRFSS) and the National Surveys on Drug Use and Health (NSDUH) from 1999-2003 – a period when 32 states’ .08 BAC laws went into effect. Models with state and year fixed effects provide no evidence that .08 BAC laws reduced alcohol-involved driving, and we similarly find no effects on drinking participation or the likelihood of binge drinking.

These results are enough to make one wonder whether (a) economics isn't as powerful an explanation as we'd like to think, and/or (b) there is something wrong with the study. But reading on, we see

We do find robust evidence, however, that .08 BAC laws reduced past month alcohol consumption among moderate drinking males by about five percent. These reductions are larger for older, college educated, and married men. Taken together with results from previous research on other drunk driving interventions, our findings confirm a general deterrence effect and suggest that tougher drunk driving laws work primarily by reducing alcohol consumption.

I'm not quite sure how this conclusion follows from their results. There's no evidence that lowering the BAC limits led to reduced alcohol-involved driving, yet tougher drunk driving laws work primarily by reducing alcohol consumption? Perhaps I am missing something.

What I see from these results is that people who have a lot to lose [older, college-educated, married men], and who probably were less likely to engage in DUI in the first place, cut back on their drinking when the BAC limits were lowered. They probably did so as a form of risk reduction; for them, drinking less is analogous to buying some specific insurance. But if, overall, lower BAC limits had no impact on alcohol-involved driving, how could the authors have possibly come to the conclusion they did?