Contra Caplan on COVID Consequentialist Calculations

Bryan Caplan writes this week about life years lost due to COVID and in particular what he sees as an overreaction to the pandemic:

Well, we’ve now endured 8 months of COVID life.  If that’s worth only 5/6ths as much as normal time, the average American has now lost 4/3rds of a month.  Multiplying that by the total American population of 330M, the total loss comes to about 37 million years of life.  That’s about 15 times the reported estimate of the direct cost of COVID.

Tyler Cowen has an important critique. Given very low government restrictions in the U.S. South, Sweden, and Brazil, we can see that even as a baseline, lots of anti-pandemic actions are actually taken by citizens voluntarily rather than by force of the state. If true, then there is little we can do to mitigate these costs even if Caplan is correct. And moreover, if we could change people’s behavior to pretend that everything was normal, then why wouldn’t we also change people’s behavior to perfectly isolate when necessary and wipe out the pandemic within a couple weeks?

Cowen argues that the best way to improve the situation if private citizens’ actions are in fact greatly damaging is to push for a vaccine as fast as possible. Robin Hanson has objections here, but putting all that aside, I wanted to look at Bryan Caplan’s numbers.

If Caplan is correct, I should change my behavior; I’ve been avoiding movie theaters and indoor restaurants and perhaps I shouldn’t. Caplan suggests 37 million QALYs lost due to “COVID” time being worse than normal time. However, this is based on a Twitter poll and Caplan points out that when asked personally about how to compared “COVID” time to normal time, his median follower says it’s almost even. But let’s take the 37 million figure at face value because it’s really the best we’ve got.

Caplan points out that what we are not comparing the 37 million QALYs to the total loss in QALYs we’ve had with COVID in this timeline, but rather a counterfactual one where we didn’t overreact.

You have to ask yourself: If normal life had continued unabated since March, how many additional life-years would have been lost?  I can believe that the number would have been double what we observed, even though no country on Earth has done so poorly.  With effort, I can imagine that the number would have been triple what we observed.  There’s a tiny chance it could have been five times worse.  But fifteen times?  No way.

Actually, it’s pretty easy to imagine! Every country on Earth has had a strong reaction to COVID and that’s why it’s hard to imagine a doubling or tripling of QALYs. If people simply went about their business, or very nearly so, then there really is little to stop COVID from spreading exponentially. In all likelihood, we would have seen overwhelmed hospitals. Instead of the per infection death rate being 0.55%, it seems quite plausible to me it could be up around 1%, especially given we were worse at treating the disease towards the beginning of the pandemic.

Moreover, if lots of people got this disease, they’d have to deal with the aftereffects which can be remarkably unpleasant. Certainly there’s a guaranteed loss of quality of life for half a month to a month as you battle the virus. Many people then have extended loss of taste and smell, extended fatigue, difficulty breathing, and perhaps even cognitive effects.

If we taken Caplan’s citation that people who die from COVID on average lose 10 QALYs (I would guess higher if more people got it, but we’ll go with this number) and say people who survive COVID on average lose half a QALY from lingering issues, the formula for this calculation of QALY loss is:

[Total QALYs lost] = [US population] * [% who get COVID] * ([% who die] * [10 QALYs] + [% who don't die] * [.5 QALYs])

If we say two thirds of the country gets COVID under a “no change in behavior” scenario (again, this seems very conservative, I would think it’d be higher) and 1% of infected die, we get:

[Total QALYs lost] = 330,000,000 * .67 * (.01 * [10 QALYs] + .99 * [.5 QALYs])
[Total QALYs lost] = 2,211,000 * [10 QALYs] + 218,889,000 * [.5 QALYs]
[Total QALYs lost] = 147,262,500

Notably, this is much higher than 37 million. Interestingly, the vast majority of the cost actually comes from the people who get the disease and survive. Perhaps I was too pessimistic on how difficult their lives are, after all, we are still uncertain. Let’s change it to a quarter of a QALY lost. Some people might have trouble breathing for years, but most people won’t, so perhaps this is a better estimate.

[Total QALYs lost] = 330,000,000 * .67 * (.01 * [10 QALYs] + .99 * [.25 QALYs])
[Total QALYs lost] = 2,211,000 * [10 QALYs] + 218,889,000 * [.25 QALYs]
[Total QALYs lost] = 76,832,250

Again, bigger than 37 million. And it’s not close. In fact, even if you say it’s only one month QALY lost when contracting COVID (which seems way too low to me considering that’s pretty much the baseline of getting and recovering from the disease) that still gets you over 40 million QALYs in our model. Add back in that the two thirds assumption is definitely an underestimate, that 10 years of QALY loss per death might also be an underestimate, and 37 million is again completely unattainable.

In other words, I don’t think the consequentialist calculation recommends going to movie theaters or indoor dining.

StatCounter - Free Web Tracker and Counter