There’s a striking passage in Freakonomics that echoes an argument I’ve occasionally made myself: Namely, that the death penalty as it is now practiced couldn’t have much effect because it is so unlikely to actually be imposed.

[G]iven the rarity with which executions are carried out in this country and the long delays in doing so, no reasonable criminal should be deterred by the threat of execution. Even though capital punishment quadrupled within a decade, there were still only 478 executions in the entire United States during the 1990s. Any parent who has ever said to a recalcitrant child, “Okay, I’m going to count to ten and this time I’m really going to punish you,” knows the difference between deterrent and empty threat. New York State, for instance, has not as of this writing executed a single criminal since reinstituting its death penalty in 1995. Even among prisoners on death row, the annual execution rate is only 2 percent—compared with the 7 percent annual chance of dying faced by a member of the Black Gangster Disciple Nation crack gang. If life on death row is safer than life on the streets, it’s hard to believe that the fear of execution is a driving force in a criminal’s calculus.

For “reasonable criminals,” this is a slam dunk argument. But how reasonable are criminals? Freakonomics is also famous for publicizing the finding that the typical drug dealer is earning a McDonald’s wage. They’ve got a sort of rational choice explanation, but still, you’ve got to wonder.

More importantly, there’s a specific reason to doubt the Levitt/Dubner argument: Availability bias. As Nobel prize-winner Daniel Kahneman argued with his legendary co-author Amos Tversky, people tend to over-estimate the probability of VIVID, MEMORABLE events. And what’s more vivid and memorable than a big execution? I suspect that executions are to punishment as airplane crashes are to travel: Statistically rare events that people count all out of proportion to their objective probability.

The upshot is that the huge media circus that surrounds each execution is probably enhancing its deterrent effect. (For a more fleshed-out story, see Kuran and Sunstein’s brilliant model of availability cascades). And since you need death penalty opponents to keep the circus alive, there’s a good chance they’re actually increasing the deterrent effect of the policy they oppose. Funny how things work out.