The result of a spur of the moment.

Photographer: Scott Olson/Getty Images

How Econ Got Crime and Punishment Wrong

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion.
Read More.
a | A

A number of observers, me included, have commented on the big shift happening in the economics field. Theory is giving way to data. Economists who focus on statistics and empirics are soaring to the top of the field, while old-style mathematical philosophers are becoming less prominent. But does this really matter? Were all those theorems and proofs really doing any harm? 

Perhaps they were. In a magisterial blog post at Marginal Revolution, George Mason University economist Alex Tabarrok recounts a battle he had with Gary Becker. Becker, who received a Nobel Prize in 1992, was perhaps the most famous economic theorist of his generation. He took the basic tools of economic theory available at the time and applied them to social problems like workplace discrimination and marriage

Becker also took on the issue of crime. But as Tabarrok notes, Becker’s models of criminal behavior were suspiciously simplistic, even for the time. In Becker’s model, criminals decide whether to commit a crime after making a careful cost-benefit analysis. The cost of committing a crime is the probability of punishment multiplied by the severity of the punishment. If the penalty for robbery goes from 10 years in prison to 20 years, the cost of committing a robbery goes way up, even if the chance of being caught stays the same. 

That already sounds suspicious. Can any human being really conceive of what it’s like to serve a 10-year prison sentence? And can anyone really tell the difference between a 10-year sentence and a 20-year sentence? It seems unlikely that someone who has never been to prison will be able to form a concrete idea of what various long prison sentences will do to someone's mental state, health, personal relationships and job prospects. As Tabarrok points out, criminals are likely to act on the spur of the moment, or in the heat of passion. Becker’s perfectly rational, perfectly knowledgeable, perfectly forward-looking model of human behavior has no room for the heat of passion. 

This simplifying assumption leads Becker to conclude that severe punishment is more effective than certain punishment. If you know that every time you commit a crime you will be caught, but will receive a light sentence, there’s very little uncertainty involved. But if there’s a smaller chance of being caught, coupled with a very severe punishment, then there is lots of uncertainty. And since economists generally assume that people are risk averse, this leads Becker to conclude that rare but severe punishments are more cost-effective at deterring crime than a “broken windows” policing strategy. In other words, if we have only a few police, who occasionally catch criminals, we can still deter crime if the punishment for getting caught is fantastically huge. 

If potential criminals are short-sighted or impulsive, however, Becker’s theory breaks down, and implementing it would lead to disaster. 

Tabarrok tried to warn Becker about this at a dinner one time. He also warned that there were lots of important features Becker’s theory was simply ignoring, such as the impact on poverty and social resentment that severe but seemingly random sentences might create. Becker waved away Tabarrok’s protests, asserting that any such problem could be solved simply by making punishments more and more severe. 

This is what happens when economists take their theories seriously. Defenders of theory-led econ protest that models are simply tools to get from assumptions to conclusions. Theories don’t tell you what to believe, these defenders say -- they simply make sure that your beliefs are internally consistent. But unfortunately, human minds don’t work this way. Models do tell people which assumptions to make. Becker’s unwavering belief in his own model of crime shows that even the most intelligent economists can easily fall prey to the temptation to believe that models equal reality. 

Of course, as we now know after decades of mass incarceration, severe punishments are not very effective at reducing crime. As University of Michigan economist Michael Mueller-Smith recently found, locking people away often simply turns them into career criminals, probably because it shuts them out of good jobs and replaces their social networks with prison networks. Meanwhile, Mueller-Smith finds no deterrent effect of very long sentences. So the best evidence we have right now is against the Becker hypothesis. 

What was missing from Becker’s analysis? Sociologists who disapprove of Becker-style theorizing will often claim that by reducing the world to an oversimplified mathematical picture, economic theorists leave out lots of important features and nuances of reality that would be gleaned by a more qualitative study.

They may be right. But the main thing that was missing from Becker’s analysis was data. No matter how nuanced your mental model, whether it’s written in math or in English, your theory is only a conjecture until it is tested with real-world data. For all his brilliance, Becker’s theorizing was very hit-or-miss. Sometimes, as with Becker’s theory that economic competition reduces workplace discrimination, the data ended up broadly vindicating his ideas. But sometimes, as with his theory on crime, the things he left out turned out to be more important than the things he put into his models. 

Now, I’m not blaming Becker for mass incarceration (any more than I’m crediting him with the reduction in workplace discrimination). But it’s reassuring to see the econ world transitioning from an age of theory to an age of data. The risk of making big policy mistakes based on faulty models will now be reduced.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Noah Smith at nsmith150@bloomberg.net

To contact the editor responsible for this story:
James Greiff at jgreiff@bloomberg.net