I haven’t written anything for this blog for quite a while, due to a backup of work for clients, but I have cleared a couple of days for administrative tasks and figured I’d write a post I’ve been wanting to do for a bit. I’d like to try to answer a question that has existed in feeble minds since the mid-19th century: Does increased government control of alcohol distribution in these United States keep people safer? Well, that’s a big question, so I can’t say I can fully answer it with certainty, with one blog post and a couple hours of Excel work. But I did want to see if there was a correlation between state control of alcohol distribution and reduced drunk driving.
Now, I’m a bit biased on this subject. And by a bit biased, I mean that I genuinely believe we should eliminate not just state control of alcohol distribution and not just the three-tier system, but state-level alcohol regulatory agencies, too. So, before analyzing the data, I decided to define the methods I will use first to eliminate the possibility of subconscious data manipulation – generally a good idea in any case.
I looked at data from each state and the District of Columbia. To measure drunk driving I settled upon drunk driving fatalities per million miles driven in 2013, which for reasons of data availability and decreased potential for confounders, seemed to be the best proxy. The data was pulled from the National Highway and Transportation Safety Administration website.
I then looked at the laws of each state and found that 18 states are considered control states. I grouped them into six tiers of state control, as is delineated below, in order of least to most state control of alcohol distribution:
Some Substantial State Control: Maryland
Some Level of State Wholesale Monopoly: Iowa, Michigan, West Virginia
True State Wholesale Monopoly: Maine, Wyoming
Some Level of State Retail Monopoly: Alabama, Idaho, New Hampshire, North Carolina, Ohio, Oregon, Virginia
True State Retail Monopoly: Mississippi, Montana, Vermont
Pretty Much Stalinist: Pennsylvania, Utah
Then it was time to take a look at the data. The chart below shows the average number of drunk driving deaths (DDD) per million vehicle miles travelled (MVMT) for a few groupings of states:
All States 0.34 DDD/MVMT
Non-Control States 0.34 DDD/MVMT
All Control States 0.35 DDD/MVMT
Some Control (Maryland) 0.25 DDD/MVMT
Some Wholesale Monopoly 0.36 DDD/MVMT
True Wholesale Monopoly 0.29 DDD/MVMT
Some Retail Monopoly 0.33 DDD/MVMT
True Retail Monopoly 0.52 DDD/MVMT
Stalinist Control States 0.26 DDD/MVMT
Three things stick out to me here. First, the average DDD/MVMT is nearly identical between control and non-control states. In fact, it is nominally higher for control states. Second, states that fall into the true retail monopoly category have a tragically high frequency of drunk driving fatalities. Third, the Stalinist control states are below the US average. That is most certainly due to the high number of teetotalers in Utah. The other state in that category, Pennsylvania, has 0.37 DDD/VMT, slightly above the national average.
I think this more or less puts the debate to rest, with a few caveats I’ll go into in a bit. First, though, I did use regression analysis to take a deeper look at the data. I found absolutely no correlation between control states as a whole and DDD/MVMT.
I did find a correlation, however, between states with a true retail monopoly and DDD/MVMT. When measured in a vacuum – again, wait for the caveats – there is a correlation of 0.36 between status as a true, state retail monopoly and DDD/MVMT, a p-value of 0.01 and a coefficient of correlation of 0.18. In imprecise, layman’s terms that means that there is a high likelihood that state control of retail alcohol sales translates into, on average, 0.18 more drunk driving fatalities per million miles driven. I found no correlations for any other sub-category of control.
OK, so that is not the end of the story. To really study this, we would want to control for other factors that may influence drunk driving, which could be confounders. For instance, cultural influences keep drunk driving fatalities lowest in Utah. Washington, D.C. is second lowest, despite a high level of alcohol consumption. This is likely because it is easy to get around without driving when drunk; a high number of miles logged by sightseers; and a concentration of politicians and professionals with plenty to lose from a DUI and enough money for a driver. Educational, economic and infrastructural factors are likely strong influences on drunk driving across the nation. Without controlling for these with dummy variables, we cannot be sure that state control of retail sales really does have a true correlation to, much less cause, DDD/MVMT. We can, however, be pretty darn sure that state control of alcohol distribution decreases drunk driving accidents by somewhere between zero and none.
Listened to While Writing This Blog Post:
My own mix of The Brothers Comatose