It turns out that there are in fact several reasons one should be skeptical of such an audacious claim. First of all, there was no noticeable effect of the drinking age on suicide and homicide rates among the general population exposed to the law change--only when the results were separated by gender was any sort of pattern noticed, and only among women born after 1960. The fact that men (who tend to drink more than women, and who also are more likely to kill themselves and others) were completely unaffected is very difficult to explain away assuming the effect is genuine. Secondly, the odds ratios were fairly small, 1.12 and 1.15, and any odds ratios less than 2.0 ought to be taken with a grain of salt (if not a whole pound). It could very likely be the result of chance, bias, or confounding factors. Thirdly, the study only looked at where the individuals were born, not where they lived at age 18. Fourthly, comparing the USA with countries with lower drinking ages does not appear to support the claim that allowing 18-20 year olds to drink results in higher homicide and suicide rates. For example, Canadians of both genders have lower homicide rates than Americans, and suicide rates that are intermediate between the lower 48 states and Alaska.
Unfortunately, we were unable to access the full text of the study, so we don't know what confounders (if any) the authors attempted to adjust for, except for state and birth-year fixed effects. However, since summaries of the study say that the effect was seen in 38 out of 39 states, that implies that the 12 states that did not change the drinking age at all (remained at 21 throughout) were not included. This is important since that would be a rudimentary way to test for secular trends, as we have done in this previous post. So many other things have changed during that time, making it difficult to tease out the impact of the drinking age change. And why weren't women (or men for that matter) born before 1960 affected? This study seems to leave the reader with more questions than answers.
The true believers in the 21 drinking age will need a lot more convincing in the error of their ways, however. That's why we took the initiative and looked up the mortality data ourselves in the publicly available CDC WONDER database. And here is what we found:
The above charts look at the female homicide and suicide rates of various age cohorts (15-19, 20-24, 25-34, 35-44, and 45-54) for the years 1979-1998. The study we are critiquing used the years 1990-2004 instead of 1979-1998, but we felt the latter would be more appropriate since a) the WONDER data are grouped into 1979-1998 and 1999-2007, each with somewhat different death codes, and b) more cohorts would be included. Voila--there is essentially no difference in the patterns of either rate over time between the various groups of states (all states, states that were always 21, and states that were 18 at some time) despite changes in the legal drinking age.