This empirical study evaluates the policy responsiveness of teen drinking i
n models that can condition on the unobserved state-specific attributes tha
t may have biased conventional evaluations. The results demonstrate that cr
oss-state heterogeneity can be important and that beer taxes have relativel
y small and statistically insignificant effects on teen drinking. Models of
youth traffic fatalities also indicate that the conventional beer tax elas
ticities are not robust to additional controls for omitted variables. The i
mportance of these omitted variables is illustrated by a counterfactual whi
ch compares models of nighttime fatalities to those that occur in the dayti
me when the rate of alcohol involvement is substantially lower. (C) 1999 El
sevier Science S.A. All rights reserved.