Bryan Caplan  

Crime and Sheepskins

PRINT
Progress on OTC Contraceptives... Did the Bush tax rebates actua...
Criminals are poorly educated.  About 68% of state inmates dropped out of high school.  Many researchers study whether this effect is causal.  As usual, though, I'm more interested in whether the causal effect stems from signaling.  

Education could reduce crime by enhancing students' job skills.  But it could just as easily reduce crime by certifying students' job skills.  If you only want to stop your kid from pursuing a life of crime, the mechanism is a red herring.  If you want to stop kids in general from pursuing lives of crime, however, the mechanism is all-important.  In the signaling model, to paraphrase a great meme, "When everyone has a diploma, no one does." 

Empirically distinguishing human capital from signaling is notoriously tricky, but sheepskin effects -  discrete benefits from crossing academic finish lines - are a strong symptom of signaling.  Sheepskin effects for income are enormous.  Are there comparable sheepskin effects for crime?  The literature is sadly thin.  But these figures from Lochner and Moretti's influential  2004 AER piece reveal big sheepskin effects of high school graduation on incarceration, with graduation year providing roughly 2/3 of the benefit. 
crime2.jpg

Anyone know of other sources on sheepskin effects and crime?  Google Scholar, for all its wonders, hasn't been too helpful so far.




COMMENTS (9 to date)
Sam writes:

I suspect they should have done their regression in log-probability space so that "regression adjusted probabilities" don't end up negative (!) for college-educated blacks. I wonder if the uptick for blacks with post-graduate degrees is statistically significant? Or are the bulk of people with 18 years of schooling college-educated after having been held back in school twice? (In which case one could imagine that the incarcerations occur before high school graduation.) For that matter, it seems like any of these studies should be careful to restrict to studying incarceration events that take place after X years of schooling, rather than incarceration events occurring for people who eventually get X years of schooling. Or perhaps the difference between these two is what is really interesting.

Peter Gerdes writes:

I don't see sheepskin effects as very good evidence of the relative components of education that are provided by signalling and value.

I mean any reasonable approach has to recognize both that there is some signaling value to graduation. Moreover, part of the signal is evidence of responsibility and ability to wait for greater rewards.

Now the closer to a degree one drops out the stronger the signal it sends that the individual lacks the ability to refrain from immediate gratification even in the face of large rewards. Thus, we can't evaluate the relative value of the degree signal compared to the value of education because there is a (potentially unboundedly large) negative signal generated by dropping out (assuming there is some non-zero signal value of a degree) closer to the degree.

Charlie writes:

"I suspect they should have done their regression in log-probability space so that "regression adjusted probabilities" don't end up negative (!)"

There is a long discussion in Angrist and Pischke about the benefits of using the linear probability model, rather than making arbitrary functional form assumptions to restrict probabilities to being between 0 and 1.

Kendall writes:

It isn't clear to me it is possible to divide up how much of the education premium is signalling and how much is human capital. One leg of a 3 legged stool holds 1/3 of the weight but it is 100% necessary for the stool to function. The odds of the top 50 high school football recruits playing in the NFL is well over 50% but if they didn't play for three years that percentage would drop significantly. Just because you can predict who will make it before the training doesn't mean the training isn't adding value. ACT/SAT scores are good indicators of who will get good jobs after college. That does not mean if the top 20% of the ACT/SAT students skipped school and sailed around the world for 4 years they would do as well.

John Becker writes:

That uptick at year 16 of education is super interesting. Why does actually graduating from college make you more likely to become a criminal than not graduating?

Bob Schadler writes:

Surely some have suggested that both may have another cause. For starters, look at Ed Banfield's The Unheavenly City. A predisposition
or inability to visualize the future might well be one such, as he argues. For someone who can't comprehend that what I do now may affect my future, education is more like punishment than "investing in my human capital." And getting caught and punished is less salient than for those who can visualize spending a long time in prison. Whether that ability to visualize the future is cultural or innate or some third possibility and the degree to which it can be enhanced by deliberate effort becomes the focus.

Floccina writes:

When you write about signalling in this case do you mean that the degree signals an individual who is unlikely to commit a crime, or do mean the degree signal allows one get a better job and so lowers his chances of turning to crime, or both. I assume you mean both but I would guess that on the former is significant.

Greg Koralewski writes:

This ruins a bit the romantic myth of very clever and sophisticated criminals, like say - Hannibal Lecter ;)

Aleksandar writes:

"If you only want to stop your kid from pursuing a life of crime, the mechanism is a red herring. If you want to stop kids in general from pursuing lives of crime, however, the mechanism is all-important."

Isn't this the opposite of what you wanted to say? Assuming that by 'mechanism' you mean education, your point should be that wanting to stop kids in general from crime is a red herring not your own kid?

Comments for this entry have been closed
Return to top