Despite a well-established correlation between alcohol intake and various risk-taking sexual behaviors, the causality remains unknown. I model the effect of alcohol use on the likelihood of pregnancy among youth using a variety of estimation techniques. The preference is given to the semi-parametric model where the cumulative distribution of heterogeneity is approximated by a 4-point discrete distribution. Using data on 17-28 year-old women from the National Longitudinal Survey of Youth, I find that alcohol consumption increases the likelihood of pregnancy by 4.7 percentage points. Quantitatively similar but statistically weaker effects were found in the fully parametric models such as the two-stage least squares model and the bivariate probit model. Finally, the fully parametric models that ignore the effect of unobserved heterogeneity failed to establish this relationship.
