Thursday, April 17, 2008

Marginal Deterrence and Child Rape

Slate's always excellent Dahlia Lithwick has a good summary of the recent Supreme Court cases regarding capital punishment. One of the cases is regarding whether or not discomfort or pain during lethal injection (or any execution method) makes it cruel and unusual; the court announced its decision (the court supported lethal injection 7 to 2). The other case is regarding whether or not rapists can be executed for the rape of children; the court heard arguments on this yesterday. It is the latter of these two that interests me.

Let's go ahead and get this over with first, just to be sure that no one accuses me of being a fan of rapists or child rapists: rape is a horrible evil, and the rape of a child is even more horrible than the rape of an adult. Clearly rapists should be severely punished.

The question is, what should that punishment be? Should it be death? Rape seems such a horrible crime that death does not seem unreasonable. Yet there is an important reason to believe that capital punishment for rapists would be a bad idea, not for the sake of the rapists, but for the sake of their victims.

The problem is what economists call marginal deterrence. If the punishment for rape is execution, and if that is the same as the punishment for murder, then the marginal cost of committing murder after committing rape is zero. That is, punishing rapists with death gives them an incentive to kill witnesses, particularly their victim. The result of execution of rapists may be slightly fewer rapes, but possibly a significantly higher number of the remaining rapes may end up with dead victims. I am personally uncomfortable with the idea of reducing rapes at the cost of additional murders, particularly if these victims are children.

When deciding the punishments for crimes, the absolute punishment is not the only thing that matters. The relative punishment for different crimes also matters. Unfortunately there was no discussion (from what I have read, anyway) of this at all in the Supreme Court. My low opinion of the court remains unchanged.

I anticipate three possible responses to this argument (there are surely more):
  • Marginal deterrence could still be preserved if rape + murder results in a higher probability of execution than rape alone, resulting in higher expected punishment.
The problem with this argument is that I can't really imagine many circumstances in which one could claim mitigating factors in raping a child, resulting in a lesser sentence. Furthermore, any factors that might matter--insanity, for example--might also reduce the probability of murder conviction. It is hard for me to imagine someone wriggling out of the death penalty for the rape of a child, but then being convicted for the murder of that child. If this really did happen, it would suggest that the death penalty for rape doesn't really have teeth.

  • Punishment isn't about deterrence at all.
I don't think that anyone is really this indifferent to the effects of poor punishment design, are they? Wouldn't someone interested in justice care if the result of this change was more dead children?

  • Changes in punishment do not affect the behavior of criminals, i.e., deterrence doesn't work.
The usual argument behind this claim is that criminals are irrational and do not respond to incentives. I think there is evidence to counter this claim, although I do not have time to summarize the entire literature here. More importantly, I think that most people really know that this argument is false, as a result of simple everyday experience. Why do parents discipline their children? Because children tend to respond. Yes, different children respond to different degrees, but a response is usually expected, and it usually occurs. If punishment does not affect behavior, why do we have laws and regulations that impose monetary punishments? Imprisonment at least makes sense on the grounds that the criminal is removed from the population, and unable to commit additional crimes. Fines and fees and similar punishments only make sense if we believe that people will respond by altering their behavior. Why should criminals--particularly the cold, calculating sort--be any different?

Monday, April 14, 2008

Consumer Reports Reveals that Sun Rises in Morning, Sugar Tastes Good, Fire Burns

Consumer Reports' automotive blog has an entry on a questionable government study. The government study concluded that the cost of making roofs in vehicles stronger was not justified by the number of lives saved. It estimated that if the strength of a vehicle's roof were raised so that it could support 3 times the vehicle's weight (as opposed to the current 1.5 times), it would save up to 135 lives per year. This would cost at least $1.2 billion, or almost $9 million per life saved. Increasing the strength to 2.5 times the vehicles weight would cost at least $88 million and save at most 44 lives per year, around $2 million per life saved. There is a new Insurance Institute for Highway Safety study which finds that raising all vehicles to the roof strength of the Xterra (which is around the "3 times" mark, depending on the weight of the vehicle as configured) would save more like 212 lives. They don't seem to dispute the cost estimates, so that puts us at around $5.6 million per life saved.

The frustrating thing about the Consumer Reports entry is that the title concludes "more lives could be saved". This seems to suggest that this is the end of the story--more lives can be saved, so more lives should be saved. But this does not follow. More lives can always be saved. We could reduce speed limits everywhere to 30 mph or less. We could require that all cars be surrounded with a layer of polystyrene foam. We could require that everyone undergo two years of intense driving training before being allowed to drive. All of these would save lives, and even if we did all of them, more lives could still be saved. The observation that "more lives can be saved" is trivial and unimportant. The question is not whether more lives can be saved, but rather, whether or not it is worth the cost. All the measures I just suggested likely have costs far in excess of benefits. Implementing them would be unwise. What about the suggested regulation mandating roof strength? Would it be wise?

It seems to me that it is a tough call. The value of of a life that is typically used for cost-benefit analysis is usually around $3 million to $6 million dollars, although I have seen numbers as high as $10 million dollars (these figures are based on observing the amount that people are willing to pay to avoid risks; it is not without controversy, resulting in the wide range I just reported). Now you can see how difficult this decision becomes. Some of these cost estimates are inside this interval, while others are outside or on the fringe. It looks like the 2.5 standard might be justified on cost-benefit grounds, using the NHTSA's numbers (which suggests to me that is even more likely to be justified if the IIHS's results are correct), while the 3.0 standard would probably be too costly according to the NHTSA's numbers, and questionable using the IIHS's numbers. This is much more interesting and useful, I think, than "more lives can be saved".

So perhaps the 2.5 standard should be implemented, and the 3.0 standard should not. Yet I think we can do better. Some people may be willing to pay for a safer vehicle, while others may not. The best policy would let people buy the car that best satisfies their risk preferences. I suggest an alternative to imposing a single standard: Require each manufacturer to report the strength of their roof. That is, suppose we say that a car meets the 2.5 standard if it can support 2.5 times its weight without suffering more than, say, 2 inches of crush (presumably tested just as cars are tested for rear-, side-, and front-impact safety). Cars that can do better can report a higher number. Get the producers competing over safety, and let consumers decide how much they want to spend on safety. Now everyone--including those consumers who are unwilling or unable to pay the extra for a slightly safer car--is better off.