Want to keep learning?

This content is taken from the Hanyang University's online course, Economics of Crime. Join the course to learn more.

Skip to 0 minutes and 0 seconds In this video, we will talk about a research paper that investigates how criminal behaviors of young adults are affected by more severe sentencing. This paper is done by Professor David Lee from Princeton University and Professor Justin McCrary from UC Berkeley. In particular, we will pay close attention to the empirical method used in the paper, called regression discontinuity, and how it helps them to recover the causal effect of more severe sentencing on crime. The main research question is the following. We want to know whether having to serve for a longer period of time in prison has any impact on the person’s criminal behavior. As we saw earlier, there are at least two important mechanisms by which longer incarceration can reduce crime.

Skip to 0 minutes and 51 seconds The first is what’s called an incapacitation effect: putting criminals in the prison prevents them from committing new crimes, at least when they are in the prison.

Skip to 1 minute and 1 second The second is what’s called a deterrent effect: if other potential offenders know that they will be punished with a longer prison term, this will deter them from committing the same crime. For obvious reasons, we want to have a better understanding of the magnitudes of these two effects. But to keep things simple, let’s focus on the deterrent effect for now. Theoretically, this effect is directly related to the economic model of crime, which predicts that potential offenders will become less likely to commit a crime when the cost of committing a crime goes up.

Skip to 1 minute and 37 seconds So, if empirical research based on actual data shows that the deterrent effect is indeed large and significant, then we may be more confident that the economic model of crime is relevant and can be used to explain the actual behavior of criminals. Let’s say we want to measure the magnitude of deterrent effects based on actual data. What should we do? One possibility is to compare the rates of crimes committed by individuals who face different punishments when they commit a same crime. It may sound strange why two people who commit a same crime will receive different punishments, but we actually see this all the time in real life.

Skip to 2 minutes and 22 seconds Even when two people commit a same crime, if one person is a first-time offender and the other had a long criminal career, the latter usually receives more severe punishment. But comparing the rates of crimes committed by these two individuals is unlikely to give us the deterrent effect. The obvious problem is that they may be systematically different in terms of their innate criminal risks. It is very well-known that individuals who commit multiple crimes in the past are much more likely to commit new crimes than first-time offenders.

Skip to 3 minutes and 1 second So even if we find that their recidivism rates are very different, we cannot tell how much of the difference is driven by their difference in innate criminal risk and how much is driven by the deterrent effect. A random experiment may help me avoid such a problem. One possible idea is to run the following experiment. I will recruit many individuals for this experiment, and randomly divide them into two groups. I will tell the first group, if they are convicted of a crime, their punishment will be a lot more severe than usual. For example, their prison terms would be 3 years longer than the usual sentencing guideline.

Skip to 3 minutes and 46 seconds I will tell the second group that they will be subject to the usual sentencing guideline. I will also cooperate with the criminal justice system, so that these two groups will indeed receive different punishment when they commit a crime. Few years later, I will compare the offending rates between the two groups. Since the group assignment was random at first, there should have been very little difference in innate criminal risk between the two groups. So if I find a big difference in crime rates between the two groups, I will take this as evidence that having to face three more years of incarceration does have large deterrent effects on their criminal behavior. Of course, there is no way I can run this experiment.

Skip to 4 minutes and 33 seconds It is very hard to imagine that the government will agree to impose different punishment on criminals who commit the same crime just for this experiment, which will cause large legal and ethical problems. This experiment I just described does not sound unrealistic, right? But our criminal justice system is doing something very similar to this hypothetical experiment on a daily basis. In the United States, and many other countries, juvenile criminals go through a different court system than adult criminals. For example, in many states, criminals under age 18 will go through a juvenile court system, which usually gives out more lenient punishment than the adult courts.

Skip to 5 minutes and 21 seconds What’s interesting is that this age 18 cutoff creates a large variation in the severity of punishment between two groups of offenders who are not that different from each other. For example, it’s reasonable to believe that, on average, there is very little difference in criminal risks and needs between individuals who are just above age 18 by few weeks and individuals were just below 18 by few weeks. But when they commit a same crime, one group will receive a lot more severe punishment than the other. And this sharp discontinuity in punishment severity creates a variation that’s very close to the one I was hoping to have in my thought experiment.

Skip to 6 minutes and 5 seconds After this video, we will talk about how this variation allows us to run what’s called a “regression discontinuity” analysis used in the paper.

Regression discontinuity

Legal and institutional requirements often provide an ideal setting to run the regression discontinuity analysis, where two seemingly comparable individuals are subject to very different treatments because of a pre-determined cutoff.

For example, the minimum legal drinking age of 21 in the U.S. creates a large variation in the pattern of alcohol consumption between individuals just below 21 and just above 21. Some university grants have a minimum GPA requirement of 3.0 so that, for example, students with 3.05 GPA are eligible for the grant but students with 2.95 GPA are not.

Now, if you observe a large difference in some outcome measure between these two groups, we can credibly interpret this difference as the causal effect of the treatment. (More frequent emergency room visits for individuals just above age 21 than those just under age 21? A higher chance of on-time college graduation for those with 3.05 GPA than those with 2.95 GPA?)

Can you think of other examples?

Share this video:

This video is from the free online course:

Economics of Crime

Hanyang University

Get a taste of this course

Find out what this course is like by previewing some of the course steps before you join: