Obesity and Fast Foods Revisited

A recent post on the Marginal Revolution blog cites a study by economists Michael Anderson and David Matsa suggesting that greater access to fast food restaurants has little effect of body weight. This conclusion is drawn from two pieces of evidence. First, eating out increased total caloric consumption by only 24 calories a day. Second, the authors use the presence of highways as instruments for the local density of restaurants (to get around the fact that the placement of fast food restaurants may be correlated with demand for increased body weight and fatty foods, etc), with the resulting estimates showing little to no effect of an extra neighborhood restaurant.

Interestingly, another study, by economist Richard Dunn, finds a much larger effect of fast food restaurants on body weight despite using a very similar instrumental variable strategy. The differential results might be due to the time period under consideration (the former studies the mid 1990s while the later examines 2005) or, more intriguingly, the way fast food restaurant intensity is defined and coded in the data. The Anderson and Matsa study groups all full and limited services restaurants together while Dunn focuses specifically on McDonald's, Burger Kings, KFCs, and other establishments similar in spirit. Indeed, Dunn's results suggest that the type of restaurant matters in terms of the treatment effect estimates.

A short digression: I'm surprised that people think 24 extra calories a day is not a big deal. I recall a Stan Misler renal pathophysiology lecture where he commented that one extra can of soda a day, without concomitant increases in energy expenditure, over several months would greatly facilitate a transition from normal to high BMIs. The extra can of soda is about 100 calories. Given this, its not hard to imagine an extra 24 calories adding up (though over a longer period of time), as well. And this is saying nothing about the differential distribution of nutrients across home produced foods and fast foods.

Anyway, the results from both papers suggests that the jury on whether fast foods are a culprit in driving the obesity epidemic is still out. Furthermore, from the perspective of a graduate student, this debate provides two further take-home points. First, changes in measurement can drastically alter the substantive conclusions gleaned from any study. Second, it is important to follow up on seemingly closed research areas and/or definite studies, both to scrutinizing methodology and to revisit strong conclusions that may have broad policy impacts. Related to this latter notion, here is an interesting working paper by Daniel Hamermesh on replication in economics.