Working Paper: NBER ID: w11035
Authors: Jeff Desimone; Amy Wolaver
Abstract: We investigate the extent to which negative alcohol use coefficients in GPA regressions reflect unobserved heterogeneity rather than direct effects of drinking, using 2001 and 2003 Youth Risk Behavior Survey data on high school students. Results illustrate that omitted factors are quite important. Drinking coefficient magnitudes fall substantially in regressions that control for risk and time preference, mental health, self-esteem, and consumption of other substances. Moreover, the impact of binge drinking is negligible for students who are less risk averse, heavily discount the future, or use other drugs. However, effects that remain significant after accounting for unobserved heterogeneity and are relatively large for risk averse, future oriented and drug free students suggest that binge drinking might slightly worsen academic performance. Consistent with this, the relationship between grades and drinking without binging is small and insignificant on the extensive margin and positive on the intensive margin.
Keywords: alcohol use; academic performance; high school students; binge drinking; unobserved heterogeneity
JEL Codes: I1; I2
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| binge drinking (I12) | GPA (C00) |
| each additional binge drinking day (I12) | GPA (C00) |
| unobserved heterogeneity (C21) | drinking coefficients (C29) |
| risk preferences and future orientation (D15) | academic outcomes (I21) |
| binge drinking (I12) | omitted variable bias (C20) |
| proxies for unobserved factors (C59) | GPA (C00) |