One of the most disturbing aspects of the holiday season is the sharp rise in automobile fatalities, in part due to drunk driving from overdrinking at parties and other celebrations. The United States has almost 40,000 deaths per year from automobile accidents, and about 40 percent are due to drunk driving. This exceeds the annual number of deaths from major diseases like prostate and breast cancers, and drunk driving deaths are also concentrated among young persons, while deaths from most diseases come at much later ages.
That the number of deaths per mile driven is not fixed but varies greatly with different conditions is seen from the wide differences among countries in the number of deaths per passenger vehicle. The United States is at the high end, and in 2004 had about 1.8 road traffic deaths per 10,000 vehicles, compared to less than 1 per 10,000 vehicles in Sweden and Norway, 1.1 in Germany, and 1.0 in the United Kingdom. A few countries in the more economically advanced nations have higher death rates than the United States, including S. Korea at 3.6, and Hungary at 3.9. A number of factors help explain these international differences, including the quality of the roads, speed limits, minimum driving ages, age distribution of the driving population, density of traffic, amount driven per vehicle (presumably higher in America), and other factors. I concentrate on accidents due to drinking.
The American approach to drunk drivers has been more laissez faire than other nations, with relatively light punishment for drunk driving, often even when it caused serious accidents. In the last couple of decades, however, under pressure from groups like Mothers Against Drunk Driving (MADD), states and the federal government have begun to crack down on drunk driving, and driving by youth. Largely in response to such pressure, the minimum drinking age in all states has been raised from what was typically age 18 to age 21. States also established more standardized criteria for what constitutes driving while drunk, and have lowered the minimum level of blood-alcohol concentration that is taken to indicate driving while drunk from 0.10 per cent to 0.08 per cent. More checkpoints and patrols have been put on the road to give sobriety tests to suspected drunk drivers.
As a result, alcohol-related fatalities fell dramatically from 1982 to 1994. But the trend has stalled since that year, and alcohol-related driving fatalities have bottomed out at about 17,000 per year. When adult drivers cause their own accidents or that of their passengers because they were drunk, one can reasonably assume they and their passengers are capable before they get drunk of determining what risks to take, such as whether to drive, speed, how much alcohol to absorb, and other factors that increase their risk of an accident. But drunk drivers often kill or injure persons in other cars, or pedestrians, and in this way they impose what are called negative externalities on these innocent victims. This is the main case for public policies to reduce drunk driving and other externality-causing driving behavior.
One approach is to tax gasoline and perhaps cars to cut down the amount of driving. Another way is to tax alcohol to cut down the amount of alcohol consumed. But these are very blunt policy instruments because they punish also people who drive without driving while drunk, or people who drink even in large quantities without driving or otherwise endangering others. By punishing only people who are discovered to be driving while drunk, or who get into accidents because they were drunk, or who are much more likely than others to drive while drunk, such as teenagers, one avoids much punishment of drinkers or drivers who do not risk the lives and property of others, and concentrates punishments on those who either are likely to, or actually did, impose external costs on others. At the same time, this would discourage, perhaps greatly, the tendency to drive after heavy drinking.
My colleague, Kevin M. Murphy, prepared estimates for the United States for the year 2000 of the total cost imposed by drunk drivers who get into accidents on drivers and passengers in other cars, and on innocent pedestrians. These costs include estimates of the statistical value of the lives lost (see my post on September 3, 2006 for a discussion of this concept), the value of medical expenses for those injured, and the value of the property lost. The total of all these external costs from drunk driving in the year 2000 is about $15 billion, with the great majority of this total coming from the $5 million value placed on each of the more than 2,000 innocent persons who lost their lives in that year because of drunk driving.
About 1.4 million persons in that year were arrested for drunk driving. Given the calculations in the previous paragraph, this means the cost imposed on others from driving while drunk amounts to about $10,000 per person arrested for drunk driving. This is a large amount, and provides a first order guidance to the punishment that should be imposed in some form on the drunk drivers arrested: large fines, suspended licenses, and jail terms in some cases. Such large punishments would match the damages done from drunk driving, and at the same time would encourage many persons to avoid driving while legally drunk. To show this with a little algebra, suppose that g per cent of all drunk drivers (=N) are arrested, and let that be the same for everyone- presumably it is higher for those who actually cause accidents. Let the punishment to those arrested be d per person, which is Murph''s $10,000 figure, where d=D/gN, and D is total damages from all accidents caused by drunk driving. Then the total punishment to those arrested =gNd=D, which is the right number in order to have drunk drivers pay for all the damages to others that they cause.
On incentives to drunk drivers, the expected damages per drunk driver =gd=gD/gN=D/N= damages per drunk driver=p D/pN, which is what it should be, where p is the assumed common probability that a drunk driver gets into an accident. The assumption of a probability common to all drunk drivers of causing an accident is clearly not perfectly accurate, but far better than the assumption that all persons who drink are equally likely to harm others. Morover, the size of the punishment might not be the same for all those arrested, but could depend on alcohol-blood levels, the recklessness of the driving behavior, and the severity of the accident if they had been in an accident, past offenses, and other factors that try to relate punishment to the magnitude of the "crime" committed and the degree of individual responsibility.
To my knowledge, no state in the United States imposes anything approaching such serious punishments on persons arrested for drunk driving. In fact, they are generally much too lenient, and do not treat this as the serious crime that it is. Countries like Sweden are much harsher in their punishment of persons arrested for drunk driving. Driving with a blood alcohol level exceeding only 0.2 parts per thousand may lead to prison terms of up to six months, and the driver's license is usually suspended. Driving with a blood alcohol level of over 1.5 may lead to one year of prison. My experience there and in Norway is that these deterrents work very well, and induce people who are planning on drinking any significant quantities to be careful not to drive afterwards. Several econometric articles by H.L. Votey indicate that punishments by fine, revoking of driving licenses, and imprisonment are important in explaining the lower tendency in Sweden to drive while drunk and the lower rates of accidents due to drunk driving, although some of the results are disputed by H. Lawrence Ross.
The American approach to drunk driving is surprisingly soft, often including the treatment of drunk drivers who seriously injure or kill innocent persons because of reckless driving. Given the sharp increase during the past 20 years in the severity of punishments for felonies, and the over 2 million persons in jails or prisons, this reluctance to try to cut down the disgraceful number of highway deaths due to driving while drunk is anomalous and disturbing.