For a long time, those on the left who oppose the “tough on crime” policies of the last few decades have argued that the experience of incarceration itself makes those convicted of crime more disposed to future criminality. In prison, one learns from peers how to be a better criminal, makes criminal contacts and also acquires a pemanent record that severely inhibits the possibility of future employment. The conservative argument is that the unpleasant experience of prison serves as a useful deterrent and discourages released prisoners from committing more crime. Both of these frameworks would predict that the effects of incarceration would be amplified by harsher, more restrictive prison conditions. Under the first theory, higher-security confinement would introduce prisoners to more expert criminals, reinforce more anti-social behavior and create a larger stain on one’s resume, whereas under the second, the more burdensome the experience of prison itself, the larger the deterence effect

Remarkably, there’s very little empirical evidence to suggest which of these two theories are correct. Steven Levitt, along with two coauthors, did find in a 2003 paper that there is a detectable deterrence effect, but there’s been no empirical study of the effect of harsher prison conditions on recidivism rates.

Until now. Recently, economists Jesse Shapiro and Keith Chen posted a working paper titled Does prison harden inmates? A discontinuity-based approach . In it, the co-authors use an ingenious bit of statistical sleight of hand to lend empirical support evidence to those of us in the first camp: harsher prisons do make people more likely to commit crimes once they’re released.

Here’s how the methodology works. They took a data set of approximately 1,000 federal prisoners from the 1980s, whose rearrest rates were tracked for three years. In the federal prison system, each new prisoner is assigned a score of 0-7 for a number of risk factors (prior record, the severity of the crime, etc…) and those points are totaled to compute a score of 0-36. Using that score, the prisoners are sorted into different security categories. For example, prisoners with scores of 0-6 get put into minimum security while those in 7-9 get put into low security, all the way up to high security for those with the highest scores.

Now, the tricky thing about figuring out whether prison conditions affect recidivism rates, is that you can’t just cite higher recidivism by those in maximum security, because their increased criminality might be just because they’re more hardened criminals, which is why they’re in maximum security in the first placea. But Shapiro and Chen exploit the discontinuity between those prisoners with scores around the cut-offs, to show that the prison conditions themselves are likely contributing to more criminal activity after release. If each point on the scale means a criminal is marginally more likely to commit another crime, there’s no reason there should be a bigger difference between those with a score of 5 and 6, and those with a score of 6 and 7. But it turns out there are differences, big ones. As you step up from prisoner with a score of 6, who gets placed in minimum security and a prisoner with a score of 7, who gets put in low security, you get a big increase in the recidivism rate. The same effect, though less pronounced happens between those in low security and those in high security.

In their conclusion, Shapiro and Chen write:

By exploiting discontinuities in the assignment of inmates to different security levels, we attempt to isolate the causal impact of prison conditions on recidivism. Our findings suggest that harsher prison conditions cause higher rates of post-release criminal behavior, behavior which is also measurably more violent.

The criminal justice system is both the most dysfunctional aspect of American democracy and the most insulated from reform, thanks to a continuing legacy of the spike in crime in 1970s, and the political benefit of “get tough on crime measures” that exploit racial fears without ever giving them explicit mention.

At some point this has to change. It would be naive to think that facts alone are going to be the undoing of America’s prison-industrial complex. But they certainly don’t hurt.