Marijuana: An Update

By Scott Alexander

[Originally to be titled “Marijuana: I Was Wrong”, but looking back I was suitably careful about everything, and my reward is not having to say that.]

Five years ago, I reviewed the potential costs and benefits of marijuana legalization and concluded that there wasn’t enough evidence for a firm conclusion. I found that using some made-up math, the effects looked slightly positive, but this was very sensitive to small changes in how made-up the math was.

The only really interesting conclusion was that most of the objective costs or benefits of legalization came from road traffic accidents. Either stoned driving would increase such accidents, killing thousands. Or people using marijuana instead of alcohol would decrease those accidents, saving thousands. I concluded:

We should probably stop [emphasizing direct] health effects of marijuana and imprisonment for marijuana-related offenses, and concentrate all of our research and political energy on how marijuana affects driving.

Using the best evidence available at the time, I predicted that marijuana legalization would probably decrease road traffic accidents. Now several states have legalized marijuana, data are in, and we have some preliminary evidence on how marijuana affects driving. And I was wrong.

A study by the Highway Loss Data Institute in June of last year finds that states that legalized marijuana saw insurance claims for auto accidents increase about 3% over the general national trend for the time. An updated study by the same group finds 6% according to insurance claims, and 5.2% according to police reports.

These are usually contrasted with a 2017 study that finds legalization states did not have significantly increased rates of car accident fatality. What is going on?

This study finds a (non-significant) increase of 2.7%. This very nicely matches the non-fatal collision study covering the same period, which finds a 3% increase in total collisions. But because there’s a lower sample size of fatal collisions compared to total collision, the fatality result fails to reach significance. Probably the reason these two results are lower than the 5.2% – 6% result is because the 5.2% – 6% result is newer, and marijuana sales have been increasing every year after legalization.

If this interpretation is true, we should expect that a mature legal marijuana industry causes about a 5% increase in car crashes and fatalities. Score one point for “obvious things” in its fight with “clever attempts to draw counterintuitive conclusions because of substitution effects”.

In the current set of nine states with legalization, the 5% increase would amount to an extra 300 deaths per year. If the country as a whole legalized, that would make about 1800 extra deaths per year. Using my totally made-up math model from the previous post, this is enough to shift the net effect of marijuana legalization from positive to slightly negative. This is especially true if the alternative to legalization is decriminalization, which has many of the benefits of legalization but fewer costs.

But again, given how weak the math here is and how dependent it is on a lot of assumptions, this probably shouldn’t taken too seriously. Tomorrow we could find out that I interpreted the fatality study wrong and marijuana really does cause uniquely non-fatal car accidents. Or that we should be ignoring all of this and paying attention to the effects on chronic pain. Or that marijuana causes cancer. Wait, no, that one was last week. Screw it.

People pointed out on the original thread that all this quantification of the objective harms and benefits of marijuana left out something important: a lot of people like it. Fair. This is hard to think about, but here are some things that help guide my intuitions:

1. Marijuana still is definitely not as bad as alcohol or smoking, which aren’t banned

2. Marijuana still is probably worse than SSRIs, which are banned without a prescription (though it’s hard to go to jail for having them; consider them “decriminalized”). Don’t tell me this a fake comparison; they’re both psychoactive drugs that purport to make you calmer and happier.

3. About two thirds of drunk driving deaths are the drunk driver themselves, and stoned driving is probably the same way. We might choose to focus only on the one-third of fatalities that happen to bystanders if we believe people should be allowed to make bad choices that only hurt themselves.

4. Everyone expects the marijuana market to keep expanding in the states where it already exists, so these numbers may increase.

5. Marijuana taxes, spent intelligently, could easily save more lives than these accidents cost.

6. Marijuana taxes won’t be spent intelligently

7. If marijuana really does increase cancer risk by a few percent, that could easily outweigh everything else and make it a giant public health disaster.

8. But bacon also increases cancer risk by a few percent, is already a giant public health disaster, and we don’t worry about it that much.

9. If the above calculations are true, preventing national legalization of marijuana would save half as many lives as successfully implementing Australia-style gun control in the US.

There wasn’t meant to be a conclusion to all of these: they help guide my intuition, but in so many different directions that I still don’t have a real position.