Further Proof of White Noise Syndrome

We have commented in previous posts about a phenomenon we like to call "White Noise Syndrome", or the shifting of deaths from one age group to another when the drinking age is changed.  Sociologist Mike Males refers to this as a "seesaw effect".  This has been documented in several peer-reviewed studies, namely Males (1986), Asch and Levy (1987 and 1990), and most recently in Dee and Evans (2001).  Studies like Carpenter and Dobkin (2010) which find increases in deaths upon turning 21 (today, with a 21 drinking age) seem to dovetail with these results.  Of course, as with anything not every study has agreed with these results, and estimates of net effects varied, but they have been pretty tough to refute. 

Now a new study by economics student Dan Dirscherl (2010) at Notre Dame University further confirms the previous studies mentioned.  His award-winning paper, which uses a fixed-effects, difference-in-differences model on data from 1972-1994 finds that a drinking age of 21 (compared with a lower one) not only shifts deaths from 18-20 year olds to 21-24 year olds, but appears to result in a net increase in fatalities among 18-24 olds.  The effect was driven by males, while females saw no effect either way.  Interestingly, the effect on 18-20 year olds was statistically insignificant, while on 21-24 it was highly significant.  This is the dirty (not to mention deadly) secret of the supposed lifesaving effect of the 21 drinking age that the pro-21 crowd would rather sweep under the rug. 

One potential flaw to this study was that legal drinking ages of 18 and 19 (and even 20 for two states) were lumped together as states with an age limit below 21, compared with states at 21.  This could increase the noise in the data.  However, Miron and Tetelbaum (2009) found no statistically significant difference between drinking ages of 18, 19, or 20 on traffic fatalities in any of their models, even those models that found significant effects for a drinking age of 21 relative to 18.  (Of course, the best model found no lifesaving effects in the long run for MLDA 21 either.)  Another criticism one could make is that Dirscherl did not include data for those under 18.  Again, however, Miron and Tetelbaum found during a robustness check of their own study that a drinking age of 19, 20, or especially 21 increases fatalities of drivers under 18, contrary to what the "trickle-down" or "low-hanging fruit" theory predicts.  Still another criticism is that Dirscherl (like Miron and Tetelbaum) used total traffic fatalities and not "alcohol-related" ones.  But there is no reliable data for "alcohol-related" fatalities before 1982, and even after that the data can be biased due to differences in testing rates.  When using data going that far back, when about 2/3 of teenage fatalites were alcohol related, total fatalities would likely be sufficient, but surrogates like single-vehicle nighttime fatalities are a good robustness check.  And both studies by Asch and Levy do exactly that, with similar results. 

Of course, alcohol-related deaths among young people are by no means limited to the highways.  There is virtually zero evidence that non-traffic alcohol-related fatalities as a whole are significantly reduced by the 21 drinking age, and effects on specific causes of non-traffic death have been inconsistent at best.  In fact, from 1998 to 2005 (years during which enforcement of the 21 drinking age generally increased), these fatalities among 18-24 year olds increased significantly, including a near-tripling of alcohol poisoning deaths for this age group.  It is entirely plausible that forcing alcohol use underground for 18-20 year olds would make it more dangerous than it has to be, most notably for alcohol poisoning deaths (which are really just the tip of the iceberg of high-risk drinking).  An honest cost-benefit analysis of the 21 drinking age should take these deaths into account as well.

We did our own research as well, using the CDC WONDER database to look up any deaths that mentioned "alcohol" in the cause of death for 15-24 year olds.  They group ages by 15-19 and 20-24 instead of 18-20 and 21-24, but the data are still useful for our purposes.  Shown below are the alcohol-related death rates per million people in each of the two age groups.

In other words, raising the drinking age to 21 appears to have been a shell game at best, and those who support it have quite a bit to answer for.  It's a lot like the movie White Noise 2, hence our term for this unfortunate phenomenon.

White Noise Syndrome

Has anyone ever seen the 2007 horror film White Noise 2? (Spoiler alert) A man has a near-death experience that has left him with the supernatural ability to predict exactly who will die and when.  He acts on his premonitions, and saves several lives, only to find out that exactly three days later, the people he saves start killing others upon being possessed by an evil entity.  In other words, the net effect is an increase in deaths.  He then realizes he wasn't supposed to save those people, regrets his choices, and actually considers killing those he saved in order to rectify this horrible, unforseen tragedy.  We watch such films with revulsion and assume they are mere fiction.

But what if there was a government policy, at the expense of tax dollars and civil liberties, that at best delayed deaths of young people by a few years and potentially even increased the number of premature deaths over the lifecycle?  What if there were entire orgainizations who wholeheartedly endorsed such a policy as "saving lives" or "for the children" while ignoring or minimizing its dark side?  And what if anyone who questions such a policy is subjected to a heckler's veto and even occasional censorship to chill debate? 

Well, that describes the 21 drinking age perfectly.  Fans of this blog already know about a study done by Dee and Evans (2001) which showed that raising the drinking age merely shifted deaths into the future by a few years, and perhaps even increased them.  Asch and Levy (1987 and 1990) and Mike Males (1986) were some of the first people to notice this redistribution of mortality.  And remember, the longer a drunk driver lives, the more innocent people he or she can take to his grave with him or her.  Of course, not every study agrees with Dee and Evans' conclusion, but there is some new evidence that supports this view.

Kitt Carpenter and Carlos Dobkin (2009) have a new study out that shows a discrete and significant jump in mortality at exactly age 21.  The effect is true only for external causes of death, including motor vehicle accidents, suicides, deaths labled as "alcohol related," and those labeled as "other external," but not homicides or drug-related deaths.  The effect also occurs for self reported alcohol consumption as well.  Ruling out alternative explanations, they conclude it is due to the effect of the drinking age.   But unfortunately, they also make the specious claim that such an effect is not merely a delay in deaths but a true lifesaving effect of the policy.

We at Twenty-One Debunked who have read the paper fail to see a true lifesaving effect over the lifecycle.  First of all, only deaths between one's 19th birithday and 23rd birthday are included, and the data are rather grainy, making longer range projections very difficult for what would happen in the absence of the observed drinking age effect.  Yes, the effect persists to an extent, but one can clearly see it gradually decline over time.  It would have been better if they expanded the data to include ages 18 through 24 (are 18 year olds somehow irrelevant to the debate?).  And there could be other age-related factors that give an illusion of persistence, such as a "toning down" of drinking in the few months just before turning legal as well as the fact that 22-23 year olds are more likely to have cars and live away from their parents than 19-20 year olds.  Indeed, our own crude back-of-the-envelope calculations after reading the paper (and its graphs) in which we project while excluding ages 20.5-21.5 suggest exactly that--it is most likely just a temporary effect overall.

Of course, death rates are merely the tip of a very large iceberg.  Carpenter and Dobkin (2008) also conducted another similar study, this time concerning various types of crime, with similar results overall.  Arrest rates were used as the proxy measure of crime.  They found a discrete and significant jump in the arrest rates of several offenses, such as assault, drunkenness, disorderly conduct, and especially DUI, at exactly age 21.  However, there was no noticeable effect for other crimes.  Interestingly, even rape, which has a reputation for being alcohol-related, appeared to be unaffected.  (We suspect this is due to the fact that drinking is less likely to be done "underground" after 21, and thus in environments less conducive to rape, which may outweigh the increase in drinking.  Or perhaps the supposed causal link between alcohol and rape has been overstated.)  The authors draw the same conclusions that they did in the other study, which is unfortunate for precisely the same reasons.

While our own back-of-the-envelope projection estimates suggest that the increaes in assault and disorderly conduct arrests are merely temporary and seem to wear off by age 22, the effects on DUI and drunkenness arrests do still seem to persist to at least age 23.  However, the fact that 21-23 year olds can drink in bars may make drunk drivers more likely to get caught, and also the increase in both DUI and drunkenness may be an artifact of the fact that people over 21 can no longer be charged with underage drinking, as evident in the simultaneous sharp decrease in "liquor law" (i.e. underage drinking) arrests upon turning 21.  Thus, some behaviors that would lead to underage drinking arrests before 21 would likely lead to DUI and/or drunkenness arrests instead after turning 21.

In other words, these studies show that banning young people from drinking until age 21 (when they are more likely to have cars, and family controls are much weaker) may not be the best way to introduce them to alcohol.  In fact, it appears on balance to be one of the worst ways, and is akin to setting a time bomb.  There is zero evidence that people magically become mature enough to handle alcohol upon turning 21.  Indeed, the aforementioned studies suggest quite the opposite, at least in the short term.

We all know what the road to hell is paved with.  Let's defuse this ticking time bomb and lower the drinking age to 18, legalizing alcohol for all legal adults in America.  What better time than now?