Sunday, October 04, 2020

COVID-19: Statistical Sauce for the Goose is Statistical Sauce for the Gander

Quoth Jay W. Richards at National Review:

If lockdowns really altered the course of this pandemic, then coronavirus case counts should have clearly dropped whenever and wherever lockdowns took place. The effect should have been obvious, though with a time lag. It takes time for new coronavirus infections to be officially counted, so we would expect the numbers to plummet as soon as the waiting time was over. ... To judge from the evidence, the answer is clear: Mandated lockdowns had little effect on the spread of the coronavirus.

I've been quick in recent months to point out a fairly obvious non sequitur among proponents of various "pandemic control" measures. This is the same fallacy, coming from the other side.

It does not necessarily follow from the fact that cases/hospitalizations/deaths go up or down, or remain the same, after a particular thing that that particular thing caused the increase/decrease/non-change.

Not long ago, I was arguing about mask mandates with a fellow who pointed to a general interest news article (not a peer-reviewed "scientific study," many but not all of which deserve those scare quotes anyway) asserting a correlation between areas with or without mask mandates and reductions or increases in cases/hospitalizations/deaths.

By way of pointing out that non sequitur, I noted that areas which had or did not have mask mandates tended to share other characteristics as well. Areas with mask mandates probably had more stringent, and more stringently enforced, "social distancing" regimes as well. Areas with mask mandates were more likely to be running aggressive "contact tracing"/quarantine programs. And so on and so forth. To attribute outcome X to measure Y is simply to assume too much.

Vis a vis Richards's claim, a couple of thoughts:

Areas with stricter "lockdown" rules tend to be areas that got hit earlier and harder in the first place, and that have higher population densities than areas without strict "lockdown" rules. COVID-19 was likely well into "community spread" in New York City before Blue Eye, Missouri saw its first case (if it's even seen a case yet). NYC has a fairly densely packed population of 8.4 million and is a major travel/trade nexus and port of entry to the United States. Blue Eye is a rural community with a population of 162 that probably doesn't see as many international travelers in a year as the JFK international terminal sees in a day.

Different diseases tend to prefer different climates. Coronaviruses apparently don't like warmth and humidity that much, so you're probably going to see different levels (and routes) of spread in Des Moines, Iowa than in Deland, Florida. Areas that got hit earlier might see reductions as some semblance of herd immunity emerges, while communities that got hit later might still be trending upward because they haven't hit that point yet. And there are also different strains of the SARS-COV-2 virus hitting different communities that might be more or less easily passed on to others.

I am NOT arguing FOR "lockdowns." In fact, I am very much against them, for reasons extending well beyond whether I think they're effective in reducing the spread of COVID-19. But a society of 320 million people living in various climates, at various densities, etc., is not a laboratory environment for which all plausible variables can be controlled, or even known. That's a fact that all sides of the debate over how to fight the thing need to remember.

No comments: