Skip to 0 minutes and 16 seconds Up until now, we saw a number of empirical studies that try to recover a causal relationship in the absence of a random experiment. For example, if I’m interested in knowing the effect of more severe punishment on crime, perhaps the most straight-forward way to find this out would be to run the following experiment. I will impose more severe punishment on randomly chosen individuals but not on others, and compare their subsequent criminal behavior. But running an experiment like this can easily become infeasible and that’s why we try hard to look for identifying variation to recover a causal relationship from non-experimental data.
Skip to 0 minutes and 58 seconds Last week we saw how the age cutoff of 18 for legal minority status can provide a variation that allows us to identify the deterrent effect of more severe punishment. Individuals who are just above age 18 and individuals who are just below age 18 should be highly comparable in most aspects, except that those below 18 are subject to more lenient juvenile court system. And because of that, we can plausibly attribute the difference in their offending rates to the difference in the criminal sanctions that they face. But if we could run a random experiment that is well designed and well implemented, finding out a causal effect of some factor X on crime can be a lot simpler and easier.
Skip to 1 minute and 49 seconds But running a good random experiment is usually going to be very difficult, and instead we often rely on variations that are caused by natural events, existing and newly introduced laws, or some unexpected changes to identify the causal relationship of interest. But what if we can actually implement a random experiment to find out more about the causal relationship of interest? The advantage of using a random experiment in finding a causal relationship has been well-known for many years by now, and policy makers and researchers worked very hard to advocate, design, and actually implement a number of random experiments in recent years.
Skip to 2 minutes and 34 seconds In this final week of the course, we will examine the examples of these policy experiments, and what they tell us about how to fight crime in efficient and innovative ways. We will see how crime may be lowered by imposing very quick and certain punishments, providing housing support to low-income households, and offering cognitive behavior therapy for high-risk students.
Welcome to Week 6
Do you know of any randomized policy experiment that took place in your city and/or country?
© Songman Kang, Hanyang University