Venue: The Fuqua School of Business, Duke University, 1 Towerview Drive, Durham, NC 27708-0120
Presentation
State Unemployment Rates and High School Student Drinking
Recent research has shown that alcohol consumption is one of many potentially unhealthy behaviors and outcomes that varies procyclically among adults, in the sense that more drinking occurs when economic conditions are good. This paper investigates whether a similar relationship prevails for alcohol use among high school students in biennial Youth Risk Behavior Survey data from 1991?2005. I analyze several measures of drinking in the past 30 days, including indicators of doing so at all, in binge quantities, on school property and prior to driving, as well as days of consumption among drinkers and binge use among binge drinkers.
Paralleling the literature, I use state unemployment rates to proxy for short-run economic conditions. An innovation compared to previous work, relevant specifically for this age group, is that I control for both age 16?19 and total unemployment rates. This potentially allows for distinguishing effects operating through teen labor supply from those operating more broadly through economic conditions that shift the household budget constraint. State fixed effects are included to hold constant spurious time-invariant state-level correlates of unemployment rates and alcohol consumption. Time-varying heterogeneity is further addressed using state-specific linear time trends in a sample of the seven populous states surveyed each wave (and 10 surveyed in all but one wave), along with region-by-year fixed effects in larger samples up to the 40 states surveyed at least twice over the period. Models also hold constant state-level beer taxes and per capita alcohol consumption as well as indicators for gender, age, grade, race/ethnicity, and survey year, along with height and indicators for metropolitan status and the education level of each parent in years that these latter measures are reported. Sample-weighted regressions are estimated using probit models for drinking indicators and interval models for the log of drinking days, with standard errors adjusted for heteroskedasticity and clustering at the state-by-year level. Results are also obtained using per capita income in place of the total unemployment rate and from samples stratified by age and grade, to further explore the distinction between effects occurring through teen employment and household income, as well as by gender and race. Preliminary estimates suggest that the prevalence of overall and binge drinking is strongly countercyclical with regard to teen unemployment rates. A procyclical pattern is evident with respect to per capita income, but this relationship is weak and not robust across samples or specifications.