Sunday, February 14, 2010

Social Host Laws Revisited

We at Twenty-One Debunked have repeatedly stated that there was no hard evidence that "social host" laws (laws that impose civil and/or criminal liability on those who merely allow (not give) those under 21 to drink on property they control, especially if injuries or fatalities subsequently occur) save any lives or reduce underage drinking.  In fact, a 2008 study by Fell et al. (a true believer in the 21 drinking age no less) found no effect of such laws, at least not for criminal ones.  That is not surprising since even district attorneys find these laws difficult to enforce, including the notorious law in Massachusetts. 

But a new study by Dills (2009) appears to have found a lifesaving effect for social host laws among 18-20 year olds, at least according to the author.  And it supposedly remained even after several other variables (drinking age, 0.08 BAC limit, seat belt law, zero tolerance, beer tax, etc.) and fixed effects were controlled for.

However, this claim does not appear to stand up to closer scrutiny.  After reading the paper ourselves, we find the following issues with the study:

  • The fatalities were divided into three categories:  drinking, drunk driver, and sober.  The fact that data from the 1977-2005 were used would likely introduce biases relating to BAC testing rates.  Testing rates were much lower in the 1970s and early 1980s, and determination was often subjective.
  • Restricting the data to 1982-2005 (the only years for which that FARS has alcohol-related data, and likely less biased) reduced the size and significance of the effects of both social host laws and the drinking age.  The former was only significant at the 10% level, while the latter was not even statistically significant at all. 
  • None of the models showed a "dose-response" relationship when the effects of various drinking ages (18, 19, 20, and 21) were tested.  In fact, some even had the "wrong" sign.
  • Many of the covariates such as BAC limit, beer tax, zero tolerance, and seat belt laws were statistically insignificant, suggesting something wrong with the models.
  • Dram-shop laws were not controlled for, and since many social host states have these as well, this may be a potent confounding factor.  Some past studies have found effects of dram-shop laws, while others have not.
  • Other variables that were not controlled for include sobriety checkpoints, roving patrols, 0.10 BAC laws, harsher DUI penalties, administrative license revocation, police per capita, and several others.
  • There was no distinction between statutes and case law, which suggests a potential endogeneity problem.
  • There was no over-21 comparison group.
  • In general, states that adopted social host laws already had declining fatalities before adoption.
  • Using survey data among 18-20 year olds, effects of social host laws were not significant (even at the 10% level) for drinking, "binge" drinking, and drunk driving in the past 30 days when other variables and state trends were controlled for.  For the frequency of drunk driving per respondent, it was only significant at the 10% level despite a very large sample size of over 52,000 people.
  • Effects on those under 18 were not tested in any sense.
For those who don't know, statistical significance refers to the likelihood that a result did not occur by chance.  However, it only controls random errors, not systematic ones (such as bias or confounders).  Traditionally, a 5% level is chosen, meaning that if the p-value (the probability of getting another result at least as extreme) is below 0.05, it is considered statistically significant.  Results with a p-value above 0.05, such as many of the results in the Dills study, are traditionally rejected.

In other words, the evidence from the study in support of social host laws is rather weak, and is likely to be a spurious correlation.  In fact, the reported 9% decrease in drunk driving fatalities is both too small to be conclusive evidence of causality (especially when statistical significance is examined), and too large to be plausible given that the majority of people are likely unaware of the existence of these laws, especially civil ones (which are usually only sporadically enforced).  Most likely, social host laws are acting as a proxy for something else, such as tougher DUI laws and/or enforcement.

To our knowledge, there has been only one other study of social host laws to date. A 2000 study by Stout et al. did manage to find a significant negative correlation between all-ages civil social host laws and both self-reported "binge" drinking as well as self-reported drinking and driving among adults over 21 in national survey data from 1984-1995.  While numerous variables were controlled for, the following were not:  state fixed effects, state trends, blood alcohol limit, sobriety checkpoints, overall DUI enforcement, and drinking age.  Again, it could have been a proxy for something else.  And traffic fatalities or any other consequences of drinking were not examined in the study, so questions relating to those are left unanswered by that study.  Generalizability is also limited, and it says nothing about the persistence of the reported effects or whether they are applicable to anyone under 21, especially in today's world.  Those who use the study to justify social host laws as worthwhile should bear in mind that the same study found that mandatory fines and especially mandatory jail for a first DUI offense to be more effective in reducing self-reported drinking and driving than social host laws.

What about the "collateral damage" that occurs from social host laws?  For one, a host (however broadly defined) can be fined, sued, or even jailed--in some cases for several years.  And civil liberties often need to be violated to enforce such draconian prohibitions.  But what about the "children," you know, the ones these laws were ostensibly written to "protect"?  In the five years since San Diego passed their own local social host ordinance in 2003 (California itself has none), there have been more police responses to parties, and more alcohol-related teen hospital admissions.  In other words, there was likely just as much drinking if not more so, but the bigger teen drinking parties of the past seem to have broken up into many more smaller (and more dangerous) ones with presumably more booze to go around.  And only the bold and reckless (and/or ignorant) are hosting them now that it is a crime.  Looks like the Law of Eristic Escalation in action yet again.

We at Twenty-One Debunked do not support any type of social host law, civil or criminal, as they contravene the very idea of personal responsibility and likely do more harm than good by forcing alcohol deeper underground, making it more dangerous than it has to be.  At the very least, we do not think that such laws should apply to drinking by those over the age of majority (18), regardless of the current legal drinking age.  Of course, we want that to be 18 as well.  And we have repeatedly noted that social host laws are just another pathetic attempt to prop up the greatest alcohol policy failure since Prohibition.  We would be better off going after those who actually do drive drunk and endanger the public, regardless of age.

No comments:

Post a Comment