There has been a healthy debate about the origin of the obesity "epidemic" in the United States (and elsewhere). Principal culprits are: cheap food, availability of medication to counter the impact of obesity, corn syrup, restaurants and in particular fast-food restaurants. For the latter, the suspicion is that eating out provides a more calorific meals than at home, both because of the fatter food and because of the larger portions.
Michael Anderson and David Matsa challenge this statement on the basis that the causality could run both ways: it could be that the availability of more restaurants made people eat more unhealthily (the prevalent explanation) or it could be that changes in food preferences lead to more restaurants open. The latter would indicate that restaurants are a symptom rather than the cause, which is important from a policy perspective.
Anderson and Matsa use an interesting strategy to study the impact of restaurant supply on the body-mass index of neighboring communities, as travel distance is an important factor in restaurant demand. In rural areas, there is a very large variation of restaurant density due to interstate highways: restaurants agglomerate on highway exits to satisfy travelers. This exogenous variation in restaurant density should lead to higher obesity rates close to highways, if the conventional wisdom is to be believed. It turns out there is no difference.
But what about the positive correlation between restaurants and obesity? The authors conjecture this has to do with a selection bias: those eating greasy food in a restaurant would also eat that at home. And people compensate for the large restaurant portion by eating less otherwise. I can certainly confirm the latter: as any good economist would do, I stuff myself at the fix-price, all-you-can-eat buffet, and then eat nothing else for the day.
This study highlights that restaurants may not be at fault. If one were to restrict restaurants in some way (say with a tax), one would thus reduce social and individual welfare with no health benefit. And besides, how do the French do?