Tuesday, May 17, 2011

"if incentives don't matter in individual cases, they can't matter in the aggregate"

So says Mr. Gillette, in critiquing Andrew Leonard's critique of Tyler Cowen's blog post about that French dude who allegedly raped a hotel maid, a blog post which I lampooned a few minutes ago.

Mr. Gillette is ....


What Mr. Gillette is overlooking is the concept of probability.  Incentives, properly understood, are probabilistic.  The idea is that a guy like DSK is less likely to do something batshit crazy than somebody with much less to lose.  "Incentives" are not a law of nature that forecloses certain choices.  They are, rather, nudges that have effects on a population-wide basis, over time.  In other words, they may not make a difference in an individual case, but they matter in the aggregate.  

To understand this, imagine two slightly psychopathic people.  One is accomplished and rich, with a lot to lose; the other is a failure, and poor, with nothing to lose.  Since both are slightly psychopathic, both of them occasionally have an urge to do something really naughty.  In fact, both of them usually resist this urge.  After all, even the worst criminals have only a few dozen criminal moments during their trillion-moment-long lifetimes.

(The "true economist" would say that this low level of crime-committing even among the slight psychopaths is a matter of incentives.  Even very naughty people understand that they're likely to get punished for being naughty.   So that pragmatist on their right shoulder is usually able to out-argue the devil on the left.  But the devil gets his due....)

Anyhow, it's reasonable to assume that, at any given moment, the poor slight-psycho is somewhat more likely to commit a crime than the the rich slight-pyscho.  It's just a matter of probability - the poor guy's got a 2% chance of giving into his naughty side at any given moment, and the rich guy has a 1% chance.  This is because these guys -- though equally naughty -- are still rational actors with some ability to control themselves.  The more reason they have to control themselves, the less likely they are to be naughty.

This is how "incentives" can "not matter" in individual cases but still still matter in the aggregate.  It's only in 1 out of 100 moments that the difference in incentives has an effect.  Still, over the course of a billion temptations, incentives will take their toll.


  1. Wrong about probabilities (at least according to you) and also wrong for not including Cowen's block quote. That will teach me to discuss economics. At least I wasn't wrong about, apparently, about the traitors on the Supreme Court.


Comments on posts older than 30 days are moderated because almost all of those comments are spam.