Social Econ Blog

Lots of newsprint has been dedicated to stories suggesting that moderate drinking is good for your health. A new meta-analysis of 54 different studies finds that the studies that show positive effects of drinking code their data "incorrectly" by including former drinkers (e.g., recovering alcoholics) in the group of abstainers. That is, the studies which show positive effects of moderate drinking suffered from omitted variable bias. Individuals' drinking histories (and the lasting damage it might have done to their health) were not included in the analyses. As such, it is not surprising that comparing a group of moderate drinkers (which is unlikely to included very many people who've ever had serious drinking problems) to a group of abstainers (which includes people ranging from those who've never drank to recovering alcoholics) produces better health outcomes for the moderate drinkers.

Until we get some good, randomized study of this, don't believe anything that you read.