Thursday, July 2, 2009

The Shell Game

A recent study found that "binge drinking"* has declined overall from 1979 to 2006 among young people, with one notable exception--college students. Oh yeah, and young females also.

During this period of nearly three decades, men and women experienced converging trends in reported "binge" drinking:


Age 15-17: -50%
Age 18-20: -20%
Age 21-23: -10%


Age 15-17: unchanged
Age 18-20: unchanged
Age 21-23: +40%

while college students and nonstudents the same age experienced diverging trends:


Age 18-20, males: unchanged
Age 18-20, females: unchanged
Age 21-23, males: unchanged
Age 21-23, females: +40%


Age 18-20, males: -30%
Age 18-20, females: unchanged
Age 21-23, males: -10%
Age 21-23, females: +20%

The authors took this as evidence that raising the drinking age to 21 was successful. But there are two problems with that theory. One is that they are assuming that correlation was causal without controlling for other variables, while in reality the downward trend predated the raising of the drinking age to 21. It was a secular trend, albeit one with some groups bucking it. And a higher drinking age may also increase underreporting as well. The other flaw is that it even if it were partly due to the drinking age being raised, it appears to have been a rather hollow victory upon closer examination. Females seeing no change or even an increase? Ditto for 21-23 year olds? Zero progress among collegians? If that's success, I'd hate to see what failure looks like.

Using an overly broad definition of "binge" drinking, like the one used above (5 drinks in an evening at least once in the past month)* may mask trends in truly dangerous drinking. Another study found that between 1998 to 2005, there was an increase in alcohol-related deaths, self-reported "binge" drinking, and self-reported driving after drinking among college-age 18-24 year olds (but in most cases a peak was reached in 2001-2002). Deaths were driven by a near tripling in the number of alcohol poisoning fatalities. That's a good index of extreme drinking, well above the 5 drinks threshold. But because the increase in self-reported "binge" drinking and driving after drinking was primarily among 21-24 year olds, while 18-20 year olds stayed roughly the same, some people have taken that as a sign of success. Again, this is specious reasoning. It seems that the 21 drinking age (and increased enforcement thereof) simply shifted such behavior a few years into the future, not at all unlike the drunk driving studies mentioned on this blog.

And remember, from 1993-2001 when "binge" drinking was (modestly) increasing for all age groups, 18-20 year olds ironically saw the largest increase of all (+56% in the number of "binge" episodes per person per year) before ostensibly declining again. So part of the increase in 21-24 year old binge drinking could in fact be a cohort effect a few years later. But for 18-20 year olds, it remains elevated in spite of (or perhaps even because of) tougher 21 law enforcement over time. It is also worth noting that the number of episodes per person per year increased faster than the proportion of "binge" drinkers in the population.

Further evidence of failure is that, according to MADD, about half (48%) of the of the alcohol consumed by college students (at four-year schools) is consumed by students under 21. This is even more striking when you consider that only 37% of four-year collegians are under 21.

In other words, at best, a drinking age of 21 appears to merely delay binge drinking, drunk driving, and alcohol-related injuries and deaths. At worst, it may even prolong such behavior. In other words, a shell game.

*A convention on this blog is to always use scare quotes when referring to "binge drinking" defined as an arbitrary number of drinks in an evening. Definitions very as there is no international consensus on what qualifies as a "binge," but the most common definition is 5 or more drinks in the same occasion (evening), or 4 for a woman in some definitions, regardless of drinking speed, weight, etc. When no scare quotes are used, we are talking about truly heavy, rapid, high-intensity drinking--often involving serious chugging or slamming many shots in a short period of time--which involves an unacceptably high risk of harm for anyone of any age. Our definition implies very high intoxication, at least 0.15 BAC if not 0.20+. That is where the lion's share of real problems occur. Fortunately, the vast majority of Americans, even college students, do not do this, regardless of what the media says.