There's an interesting new paper by Jeff DeSimone in Economic Inquiry. The abstract:
Given my interests in the role of social identities in decision making, I wonder how much joining a fraternity changes one's identity (an identity with the potential behavioral prescription involving heavy (or heavier) drinking). I find it interesting that the results still hold given a full accounting of selection effects (i.e., potential drinkers may be more likely to join a frat). Is it that joining a frat changes fundamental behavior? Is it that a frat just provides more opportunities to drink?
This paper estimates the impact of fraternity and sorority membership on a wide array of drinking outcomes among respondents to four Harvard College Alcohol Study surveys from 1993 to 2001. Identification is achieved by including proxies for specific types of unobserved heterogeneity expected to influence the relationship. These include high school and parental drinking behaviors to account for time-invariant omitted factors and assessed importance of drinking-related activities and reasons for drinking to control for changes in preferences since starting college. Because self-selection is quantitatively important, I further hold constant variables plausibly affected by fraternity membership, such as current alcohol use categorization, ranging from abstainer to heavy drinker, and time spent socializing. Even in the fully saturated model, fraternity membership significantly increases drinking intensity, frequency, and recency, as well as the prevalence of many deleterious drinking consequences that potentially carry negative externalities.