In my rant on the Zillow Book I brought up some issues related to how the authors did (and more appropriately did not) estimate the effect on real estate prices of a nearby Starbucks.

In hindsight that might have been a pretty dick move on my part. I wouldn’t say I was necessarily punching down (since I don’t wield any power in any domain…I’m essentially owned by a 2-yr old on a daily basis, I make very little money, and my professional work is almost exclusively published in a string of 3rd tier journals) but I can see how it could come off that way, since I criticized a group of pop-analysts for screwing up something I spend all-day everyday doing.

At any rate, the little nod I gave to the synthetic control methodology in that post, coupled with some NPR stuff I heard about legal weed and public health, convinced me to write a little primer on public policy analysis. I split that up into a couple posts covering difference-in-differences estimators, natural experiments, and the synthetic control method:

I like to think of this as my not-too-gentle-but-not-too-rough (it’s not the cadillac of policy effects estimation…but hopefully it’s not the Geo Metro either.) intro to public policy analysis.

Now that I’m satisfied that I’ve provided the appropriate background for talking about measuring policy impacts, a natural question arises: does anybody who wasn’t previously familiar with these methods (basically all the non-economists I know) give a shit about them? I tried to think of a set of policy questions that non-economists might be interested in:

  • minimum wage impacts? nah, I mean they get a little press but its mostly policy wonks and econ-nerds that like to talk about them
  • cigarette taxes? again, mostly just goofy economists arguing about elasticities

Then I heard an NPR interview with a doctor claiming legalizing weed was awesome because it leads to fewer overdoses from misuse of opioid pain medication. Since everybody loves talking about weed, I realized that was my in.

A little stroll though Google Scholar turned up some interesting social science research (some of it good interesting and some of “how did this get published” interesting).

1. We’ll start here:

Wen, Hockenberry, and Cummings. 2015. The effect of medical marijuana laws on adolescent and adult use of marijuana, alcohol, and other substances. Journal of Health Economics..an earlier version appears to be available on The NBER website.

These guys used a difference-in-differences approach to determine whether medical marijuana laws (MMLs) caused an increase in i) probability of adolescents trying marijuana for the first time, ii) probability of regular marijuana use among adolescents (age 12-20) and adults, and iii) whether MMLs increase the probability of binge drinking among those 21 and over.

2. For an example of slightly more dubious look at possible implications of marijuana policy have a look here. My intention here was just to highlight some of the weed-related social science research out there, not necessarily break down the legitimacy of each…but this piece has all sorts of problems: poorly defined metrics, arbitrarily defined explanatory variables, a regression table that I’m almost certain is a misprint since it indicates that all categories of drinker (ex-drinkers, heavy drinker, alcoholic, etc. are all less likely to engage in supervisory neglect than a life-time abstainer from alcohol)…Also, I’m pained that the scientific community actually has a peer-reviewed journal called, “The Journal of Abuse and Neglect.”

3. Related to #1 above this joint uses a difference-in-differences framework to evaluate the question, “do MML laws increase the prevalence of adolescent marijuana use.” They actually find that, when confounding factors are accounted for, MMLs actually lead to a decrease in the past month marijuana usage among adolescents.

4. Finally, I really enjoyed skimming this one. I haven’t read it close enough to comment on it’s legitimacy but I got a kick out of the research question. They used state-level death certificates to determine whether MMLs result in fewer state-wide deaths from heroine overdose (heroine and opioid prescription pain-killers).

It’s Friday, I made an early exit to post-up a bar with wi-fi, so I’ve kind of lost track of what my point was here. I think, if I even had a point to start with, it was something like this:

  • Policy analysis via D-i-D, natural experiments, synthetic controls is really cool.
  • By ‘really cool’ I don’t just mean really cool to economists I mean really cool generally.
  • I can sense you don’t believe me so here are some examples of empirical policy analysis methods applied to Weed-centric research questions.
Advertisement