School Choice Reduces Crime

That’s according to this study from David J. Deming of Harvard. (Thanks to Don Boudreaux of Cafe Hayek and Andrew Coulson at Cato for the pointer).

I’m always skeptical of statistical studies, even those that confirm my biases.  Reading over the design of the study, my main concern is that there is a selection bias.

That means the kids who attended the school of their choice are likely to be the ones who care more about education and their future anyway.  So, it’s not necessarily the school choice that caused the lower crime, rather the school choice process just weeded out the kids who are less likely to commit crimes anyway.

After reading over the study design, I think that is possibly the case here.  My main beef here is that parents could request up to three schools, not including their neighborhood school.  I think what is important here is how many schools the parents requested.

But, even so, I’m not going to nitpick too much.  Even if there is selection bias in the study, I think it is safe to conclude that giving students and parents a choice, worst case, doesn’t cause more crime.

2 thoughts on “School Choice Reduces Crime

  1. Perhaps I’m missing something, but as I read the study it appears to be a random assignment experiment, which all but eliminates selection bias. All the studied students sought to attend a school of their choice and a random lottery decided which ones won that lottery and which ones lost it. So there is no reason to expect the treatment (winning) group to differ appreciably from the control (losing) group.
    There is a separate objection, which is the generalizability of results from lottery applicants to the entire population, but that, it seems to me, is a lesser concern than selection bias.

    • Hi Andrew, I appreciate the comment.

      You might be right. I couldn’t quite tell from the write-up. I had two key concerns. First, Deming mentioned that 95% of parents submitted at least one preferred school and they could submit up to three outside of their neighborhood school.

      That made me wonder if a parent took the time to submit three choices had a better chance of “winning” than a parent who submitted only one or two preferred choices and that may reduce the randomness.

      The second concern I had was with the preferences built into the lottery system. First preference to students who attended the school previously and their siblings and then to income status. That is not random.

      I can think of several different implications on the results for both of these concerns and I’m sure there are more that I haven’t thought about.

      But, I decided not to belabor these potential issues in my write-up. It’s quite possible the effects of my concerns are minimal or non-existent. However, I like to look at studies like this and assume the worst case that these concerns are problems, and ask if there’s something I can be reasonably sure the study is telling me?

      I’d say yes. I don’t think I trust that winning the school selection lottery CAUSED the students to commit less crime. But, I think it is safe to conclude that the lottery certainly did not cause them to commit more crime. Which still isn’t a bad result (though, it could be if losing the lottery caused the losers to commit more crime than students not exposed to a lottery).

      I also agree with your generalizability objection. But, I think my worst case conclusion still applies.


Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s