Is Driving Drunk Rational?

I doubt this statement will shock you or light up the blogosphere, but drunk driving is bad. Our own Levitt has looked at the costs, and found that those who have had even one drink are seven times more likely to cause a fatal crash, while for those over the legal BAC limit the risk is multiplied by 13. This equates to a cost to society of more than 40 cents per drunk mile driven (2013 dollars), implying that a fine of $10,500 would be appropriate if drunk drivers were to bear the full cost of their actions.

The good news is that we have made tremendous progress. According to the National Highway Traffic Safety Administration, road fatalities due to drunk driving have dropped from 21,113 in 1982 to 9,878 in 2011. The decrease is even more remarkable given that total miles driven almost doubled during that period. So the drunk driving fatality rate per billion miles traveled has dropped from 13.4 to 3.4 in the last 30 years.

Some of this is due to general improvements in driving safety, such airbags and increased seatbelt use. But this is only a part of the equation. A suite of policies specific to alcohol has also been implemented, with considerable success. These have been recently analyzed by Susan A. Ferguson and Koyin Chang, Chin-Chih Wu, and Yung-Hsiang Ying, among others. Successful policies have included toughening laws and their enforcement, such as reducing permissible blood alcohol content (BAC) levels, especially for underage drinkers. Sobriety checkpoints are a very effective enforcement mechanism, particularly if properly publicized. Other policies that have been found to be effective are higher alcohol taxes (very), and to a lesser extent laws banning open alcohol containers in cars and higher legal drinking ages.

What about tougher penalties? It seems pretty obvious that these must work, but the evidence is surprisingly mixed. “Zero tolerance” for drunk drivers below 21 seems to be effective, and license suspension has shown some results, although the story here is somewhat equivocal. (For example, many of those who have lost their license continue to drive, illegally, albeit less frequently and more carefully. One strategy for dealing this might be impounding the offender’s vehicle). Breath alcohol ignition locks are also very effective—but only when in use: The evidence shows offenders tend to return to their old habits when they are removed. The brief incarceration that often follows a DWI stop has not been shown to be much of a deterrent.

This goes to the larger debate about the link between severity of punishment and criminal recidivism, which is not as cut and dried as you might think. In fact, some studies actually show harsher prison conditions increase recidivism upon release. For example, it is possible that prison is a “crime school” and that more time behind bars hanging around with other criminals may make one more likely to learn new tricks. Other work shows severe punishments do work, but it is not crystal clear why. Because punishment persuades people crime doesn’t pay? Or because incarceration keeps criminals off the street?

The theory that harsher punishment causes people to alter their calculus about the profitability of crime is in line with the assumptions of classical economics, which hold that man is a perfectly rational computer, with complete information, who calculates the potential value of future actions and chooses the path with the highest expected returns. In the case of crime, economist Gary Becker pioneered the theory that a potential criminal should multiply his chances of getting caught by the potential costs of the penalty (presumably with time spent behind bars converted to a monetary value) and weigh this against the benefits he gets from lawbreaking. However, we all know that in reality we’re rarely smart enough to do this.

The flaws in the rational man conceit are why behavioral economics is such a hot topic in the field; economists have moved beyond thinking about the way that people should behave to the way they do behave. How do tough penalties affect us psychologically?

Benjamin Hansen, an economist at the University of Oregon whose work I’ve written about in the past, recently presented a conference paper that looks at this question in the case of drunk driving. Specifically, he examines whether people commit fewer crimes because of fear of what may happen to them if they do (“deterrence”) versus what they’ve learned emotionally from punishments that have already happened to them (called “specific deterrence”).

Hansen’s study takes advantage of a large database of DUI offenders in Washington State, which has laws similar to the rest of the US. The key to the study is that there are two definitions of DUI: the regular old garden variety offence if BAC is over .08, and “aggravated” DUI with a level over .15. Aggravated DUI brings a harsher punishment for the offense in question, but it has no special effect on the level of punishment for the next offense (which rises in severity the same amount no matter the level of the prior conviction).

First, Hansen finds that having been convicted and punished for a DUI reduces the chances of getting caught with a DUI in the future. Since the punishment scale for DUI ramps up with prior convictions, this appears to fit with the standard theory that harsher punishment is a deterrent for reoffending.

However, one aspect of Hansen’s findings call the model of the rational criminal into question. Since there is no extra future penalty for those convicted of an aggravated DUI, if drunk drivers rationally calculated the future cost of punishment they would all have the same odds of reoffending regardless of the level of their past conviction. However, Hansen finds that those who got aggravated DUIs were actually less likely to reoffend, all else equal. This suggests that the experience of having received harsh punishment teaches you a lesson. This is called a “learning effect.” Another implication is that drunk drivers may not fully understand the system of punishment, and base their estimation of the cost of the next conviction on past experience and not accurate knowledge of the law.

These findings are interesting in terms of both psychology and deterring crime. They suggest that increased sanctions for drunk driving—such as lower permissible BAC levels, longer license suspensions, ignition lock systems, and harsher fines—are indeed likely to have additional deterrent effects. In fact, Hansen computes that raising penalties by 10 percent would reduce drunk driving by about 4 to 7 percent, and this is just counting the effects on repeat offenders and not deterrence for those who have never been caught. However, there is one catch; the positive impacts of such a learning effect would take a while to play out, as the deterrent will only mount as people get caught and punished over time.

As for psychology, it probably doesn’t come as a surprise that criminals suffer from “bounded rationality,” and with possible rare exception of a Bernie Madoff or a Unabomber probably lack the mathematical acumen to calculate the present and expected future value of their crimes. I’d be quite interested to know if, given that the extra penalty from an aggravated DUI is a “sunk cost” which according to economic theory any rational person should disregard,  chronically drunken economists are more likely to reoffend than the rest of us.

Steve Cebalt

Is drunk driving rational? Are drunks rational? It's a great and complicated question. The "Sober" version of most of us would never drive drunk; many see it as abhorrent behavior and would NEVER want to hurt anyone, let alone suffer all the lesser consequences. But hypothetically let's say I have a beer. I think, "I can drive; just one beer." I have another. With each subsequent drink my ABILITY to ratonalize is the same, but the QUALITY of the relationalizations deteriorates exponentially. Later I might say "I've driven after 12 beers before and did just fine...." My cognitive abilities are impaired and I am incapable of making a sensible risk/reward calculation. I don't know any answers to the problem, except that remedies must address two people: the sober version of the person, and, more importantly, the drunk version of that same person -- a person who can rationalize, but not effecively.



Of course they are rational before they drink. And other models of time inconsistent prefences would suggest that a person who is prone to drunk driving that really cares about others, will engage in commitment devices to prevent drunk driving (such as finding a designated driver, or not even going to a bar, or only bringing cash to buy 2-3 drinks only).

The question posed in the Hansen paper is really whether or not they are rational in their behavior after being caught, and therefore are they deterred in a rational manner.

Friend of Bill W.

I have been sober in AA for 35 years so I have had the opportunity to watch many people get sober, or sink like a stone. One pertinent observation is that people who ultimately get sober do so from being able to visualize some reasonable, sober, livable future for themselves.

Recently my brother-in-law's mother got caught for her second DUI in a tough drunk-driving-law state. She hit another car. She will get 120 days in jail, loss of license for 5 years, loss of employment (rules of her employer), endless court fines and costs, and since she was the sole support of her family, loss of home, and probably (a reasonable guess because her husband is way beyond merely pissed-off...) loss of marriage.

Could it get worse? Well yes, but this punishment is enough to effectively destroy any reasonable possibility of future life quality for her.

Does this make the public safer generally? I doubt it. Society will pay some heavy costs for her now unlivable future.

I am not so forgiving about these issues as some may think, but I have to reflect that many are now seeing addictions as more of a medical problem than a legal problem.



I don't like the way you conflate alcoholism and drunk driving. Sure, alcoholism is arguably a disease. Refusing to call a cab or take the bus is not, and I have no sympathy.


Given the probability of a drunk driver being caught is about 1 in 20,000, or .0002, and the cost of a dui is about $10,000, the expected cost of driving drunk is about $2 per night out, or less than the cost of that craft beer I just had. This is not deterrence. While the average pub crawler might not be savvy enough to calculate the expected cost, the anecdotal cost is so low - I don't know anyone immediately who has been caught driving drunk, for example - drivers will continue to drive drunk.


It may be that drunk drivers don't calibrate their level of intoxication finely enough to hit the lower range -- and know that so they rationally use the higher penalty if deterrence figures in at all.

Jim Bang

Some recent work from a friend of mine:
Zero Tolerance Laws Don't Work:
Optimal Drunk Driving Penalty Structure:
Some Political Economy Considerations:

Joel Upchurch

If you have seen the development of self-driving cars, then you would be aware that one advantages of self-driving cars is the elimination drunk driving fatalties and injuries. It doesn't make much sense to waste resources on a problem that is going away.


Once driverless cars have come down in price enough to be required for drunk drivers and common otherwise, we will wonder how we got along without them.

Honest Abe

The fight against drunk driving has become a crazed mob of irrational hate. It seems that no punishment is harsh enough for evil drunk drivers. “Impound their car, crush it, then used the crushed ball to destroy their house and burn it to the ground! Revoke their license for life!!!”

It might partly be because the big media hubs are in the major cities, where public transportation and cabs are readily accessible, but for everyone else….I hate to be the one to break this…everyone else drives drunk. All.The.Time. It is a fact. You can’t get a cab in my city without a 30 minute wait. Then it costs a minimum of $20 to go anywhere, and I’m only going a few miles. My entire bar tab was only $17 (happy hour), I’m not about to double that just to go four miles in a cab.

I drove drunk this week. I’ll drive drunk again. I can guess that I’ll probably drive drunk 4-8 times in a year for the next 30-40 years. I am not sorry about this and I’m not worried about it either. Literally, every single person in the bar will be driving home drunk. Not just some, everyone. You just know to not be an idiot about it. Don’t drive home wasted.

The police don’t want to stop drunk driving. They don’t. They want to bust the wasted drivers. If the police cared about busting drunk drivers, then they’d park squad cars in the parking lot of every bar in the city and just bust people as they pulled out of the parking lot. For that matter, they could bust every single car in line for Taco Bell after 10pm.

Drunk drivers very, very rarely kill people. I know because there are literally thousands of drunk drivers on the road in the metro every single night. Wasted drivers kill people, and even they only do so every once in a while.

Obviously, no one wants people to die. There should be punishments for DUI, I totally agree with that. But this attitude that no punishment is harsh enough is both wrong and life destroying….and a little hypocritical given how many alcoholic lawyers, judges, reporters, and politicians there are.



Sorry, but "everyone else" doesn't drive drunk. I don't, for instance, because (for say the last 20-30 years or so) I've chosen never to drink enough to become drunk. I also have many, many better things to do with my evenings than sit in a bar. Even my semi-alcoholic neighbors don't drive drunk: they'll often call me to pick up something (usually more beer) from the store because they've already had a few.

To the original post, though, I wonder just how much of the decrease in drunk driving and/or drunk driving accidents is due to laws & penalties, and how much to the fact that, like smoking, it has increasingly become socially unacceptable?


I don't think the study addresses one of the imponderables, and that is that repeat offenders often have alcoholism or have developed a dependence on the drug. Like any other addict, an alcoholic who is active NEEDS his alcohol. Since it is legal, and thus relatively inexpensive, many who become addicted are not bowery bum, sleeping under a bridge types, at least initially. They have jobs, they have families, and they have cars. An alcoholic in a blackout may not even be aware of what he's doing when he gets behind the wheel after drinking. The only accepted way for an alcoholic not to get more DWI's is not to avoid driving drunk, but to avoid drinking in the first place. But once that first, legal, drink is consumed, all bets are off. The alcoholic may very well wake up in jail with no knowledge that he's driven drunk again and possibly killed or hurt some innocent. And even worse, will have to bear the guilt that comes with the knowledge that if he hadn't picked up that first drink...



People who are time inconsistent in their preferences can still engage in commitment devices to make their future selves behave a manner that their current self would like. This could include not drinking, or not bringing keys to a bar.


When Candy Lighnter first started MADD back in the early 80's her focus was on preventing "drunk driving" and shone a light on the root causes of intoxicated driving and looked for remedies to educate a society on it's disastrous consequences. But against Lightner's wishes MADD quickly morphed into a vindictive vendetta against "drunk drivers' and their propaganda today spouts out all sorts of nasty rhetoric calling people "killers" and "murderers" for breathing over .08 on a breathalyzer machine. Deterrence may have an affect on a specific individual by destroying his livelihood and stigmatizing him with the 'criminal' label (absent any victim) but "drunk driving" is still an essential part of our society, just go down to your local sportsbar, restaurant, stadium etc. and see how many cars are parked in the parking lot waiting for the keys to be put in the ignition.


People are seldom rational.


I get a little frustrated when the effect of drinking on driving is drawn from the likelihood of people who choose to drink and drive. You're inherently sampling a biased group. Sure, it increases the risk (which is important that society address) but any group that has like 80% more arrogant drivers is also going to cause more accidents.

Ray Daugherty

Very interesting. However, DUI offenders are not just punished, there are also efforts to inform and change their future behavior. In some cases this is a required DUI class (of varying quality in different places) and in other cases mandated treatment. So assuming that fewer DUIs are due just to the punishment fails to take into account that maybe the interventions they are required to receive also help. In our own work we provide programs used in such interventions with tens of thousands of offenders in multiple states. These have been subjected to evaluations but I would love to see economists take a look. At the very least, it seems important to factor this into the equation when considering the impact of punishment; to what extent does the event itself, the punishment, or the type of intervention contribute to change.


If we really cared about reducing drunk driving, there would not be minimum parking requirements for bars.