Everything is great as long as the cities burn

July 10, 2020

When the George Floyd demonstrations began, there was a lot of justified concern that they would act as super-spreader events and set back the progress we have made against covid. (One couldn’t help but notice that the same people who been highly critical of public gatherings just days before were fine with gatherings in what they deemed to be a good cause.) This paper argues that, against all expectation, the demonstrations actually slowed the spread of covid.

The paper has met with some skepticism, but the methodology seems reasonable to me. I think it’s very plausible. However, its results are widely misunderstood, and to some degree it is the authors’ fault.

The paper explicitly acknowledges that they did not even attempt to look at whether covid spread at the demonstrations. Instead, the paper looks at cell-phone data to determine whether, on balance, the public overall congregated more during them. (ASIDE: The paper also looks a little at growth in covid cases, and that part is less convincing.) They found that the general public was so scared of the violence at these riots that they stayed home, and that effect outweighed the effect of the demonstrations themselves.

This makes sense. As large as the riots were, the vast majority of people did not participate, so even a modest negative effect among the majority could outweigh the riots themselves.

This is being spun as a defense of the demonstrations. Partly, that is the fault of the authors who asserted in their conclusion that “public speech and public health did not trade off against each other in this case.” This is a bizarre take.

Yes, public speech and public health may have coexisted, but the price of their coexistence was burning cities, many deaths, immeasurable property damage, and countless ruined lives. This is not a positive result.

The paper compounds the problem by failing to look at the key question: what happened when the riots settled down but the protests continued? At that point we would expect the negative effect to fade and the positive effect to dominate. Alas, the paper does not look at that question, so we don’t have the cell-phone data, but we can look at public data on covid spread.

According to the notes I took at the time, there was relative peace in the cities starting June 2. We expect a week’s incubation time between exposure and symptoms, so we would expect to see a surge in covid cases starting June 9. And, if we look at the data, that is exactly what we see.

To summarize, the burning of our cities masked the anticipated effect of the demonstrations on covid, but the anticipated rise happened as soon as the demonstrations became peaceful.

(Via Daily Wire.)

Divergent charts

July 10, 2020

I think this chart is really interesting:

chart of covid cases versus deaths

(Credit for the chart.)

You can see that at the start of the epidemic, deaths were lagging behind cases by 6 days. But now it’s been a month-and-a-half since the two curves diverged. Cases have been soaring for about a month, while deaths have continued their steady decline.

The rejoinder is often “wait two weeks and deaths will rise.” But the chart shows that we’ve already waited more than four weeks, and, if the early pattern prevailed, we should have seen a surge in deaths in only about one.

Some say that the rise in cases is merely more testing. I don’t think that explains it (although it might be part of it) since the percentage of tests that were positive started rising at about the same time. My theory is that the people getting sick are younger (and thus less vulnerable) than the ones who were getting sick before.

In any case, I think the chart shows convincingly that although the rise in cases is surely troubling, we can temper our alarm.

UPDATE: The day after I posted this, the death rate started to rise, which obviously is bad. Nevertheless, the point still remains that deaths are not mirroring cases.

Heterogeneous measures

June 26, 2020

The CDC observes that the coronavirus outbreaks in Arizona, California, Florida, and Texas skew towards younger people, so the consequences will probably be less bad than the earlier outbreaks in the northeast:

Ongoing outbreaks of COVID-19 in Arizona, California, Florida and Texas are “significant,” but the younger average age of confirmed cases in these states might mean the “consequences” will be less severe, U.S. Centers for Disease Control and Prevention officials said Thursday.

In light of that, I wanted to revisit this paper out of CMU and Pitt. It shows that heterogeneous measures lead to significantly fewer deaths than homogeneous measures. In fact, heterogeneity matters much more than the strictness of the measures: even a more-moderate heterogeneous measure is significantly better than a stricter homogeneous measure. In plain English, what that means is we should be protecting the vulnerable tightly, but be looser with the less vulnerable. (And keep in mind this is just about deaths, without any consideration of ruined livelihoods or economic damage.)

This CDC observation suggests to me that these heterogeneous measures are finally happening. That’s a good thing.

POSTSCRIPT: Heterogeneous measures are better assuming they are not upside-down — as they were in New York and some other states — which locked down the general public tight but pushed coronavirus patients into nursing homes. That was insanity.

(Via Instapundit.)