Monday, September 13, 2010

Beeronomics and Economist's Notebook: Does Drinking in College Hurt Performance?

I figured I'd put this in the Beeronomics sphere as well as the Economist's Notebook as beer is one of the beverages of choice among collegge students: A new NBER paper [HT: Freakonomics] reports that drinking in college impairs performance.  From the abstract:

This paper examines the effect of alcohol consumption on student achievement. To do so, we exploit the discontinuity in drinking at age 21 at a college in which the minimum legal drinking age is strictly enforced. We find that drinking causes significant reductions in academic performance, particularly for the highest-performing students. This suggests that the negative consequences of alcohol consumption extend beyond the narrow segment of the population at risk of more severe, low-frequency, outcomes.

Uh oh. This should come as no surprise to anyone who has ever been, or been exposed to, college students: students drinking alcohol is common and the effects of alcohol (as most college students will tell you) impedes their academic performance.

But look more closely at the abstract to understand just what it is the authors are doing in this paper, and the actual story becomes less less clear.  The authors took data from the US Air Force Academy where underage drinking is not tolerated and can lead to immediate expulsion and compared the performance of students just before their 21st birthdays and the performance of students just after their 21st birthdays.

So it is not the alcohol itself the authors are isolating, but the entire change in lifestyle and time management that happens when students turn 21. The authors claim that the USAFA data are a source of strength, but I actually see it as a major weakness.  With such a tightly controlled campus, 21st birthdays may traditionally be a moment of liberation, not just to drink alcohol but to become more social off campus and perhaps less serious on campus.  Now, I imagine that they have a control group for comparison of students who didn't drink prior to age 21 and who don't drink after turning 21, but this is not really a counterfactual because not drinking may also signal a rejection of the entire post-21 lifestyle of which I speak.  [I don't have access to NBER working papers through OSU]

This is just another problem of identifying causality that always happens with social data: is it the alcohol that impedes performance or the behaviors correlated with alcohol?  I suspect both.

Another point is that students are maximizing utility functions that include academic performance, social performance and just plain fun.  Many students recognize the tradeoffs inherent in drinking, and in socializing in general, with academic performance and make their own decisions about where their optimum point lies.  The question then is do they accurately estimate the effect of alcohol consumption on academic performance or do they not understand the deleterious effects?  If the former, than there is nothing to say unless there are significant social costs. If the latter, than we need to understand these effects and transmit the knowledge to students. Unfortunately, I don't think this paper does this.

Finally, the other important question is if there really are significant social costs.  Underage drinking has many by the way, which is why we prohibit it, but I am talking here only of the effect on academic performance.  Well, there are social returns to education so if alcohol impairs individual performance, it should also have a social impact as well, as a less well educated population is less productive.  But how significant are these costs is an important question and one we don't have the answer to.

So it is an important research agenda, but a tough causal link to uncover.  

1 comment:

Dann Cutter said...

I think you hit the nail on the head with the discussion of other confounding factors upon turning 21. Specifically, the authors claim the method used to analyze the data (regression discontinuity design) depends that "other determinants of performance such as motivation or maturity vary smoothly over age 21, access to and consumption of alcohol varies discontinuously at that point". However, as you point out, the social construct in an environment such as the USAFA changes once certain establishments are no longer off limits.

Regardless of consumption, the ability to now socialize with one's peers in an establishment which presumes a certain status (age in this case, which is similar to the military structure of hierarchal ranking) creates a new motivation to network as unavailable previously. This social pressure can be significantly confounding (for example, I am by necessity attending my company picnic this weekend instead of practicing my thesis presentation; though alcohol is thankfully involved, it is the social networking necessity which takes precedence.)

Test performance as well belies the social cost function - while performance may show degradation, in a socially networked organization such as the military, lowered performance on measured exams may weigh significantly less than the benefits of these social contacts. This is suggested as well by the results that those who were higher performers - these folks may have a better understanding of the value of the social networks, and thus willing to allow a greater material effect for the intangible social benefits.

Finally, the authors put a lot of stock in the self reporting survey of cadets which show a 30ish percent pre 21 drinking vs 90 percent drinking after turning 21. Having been in a similar environment, I can only anecdotally say this premise and assumption is flawed.