A New York Times story last week about the rise of Covid in Florida said "even a large state that emphasized vaccinations in combating the coronavirus can be crushed by the Delta variant when no other measures are put in place." As the figure shows, there's a strong relationship between the vaccination rate and the Covid hospitalization rates, but also a good deal of variation even after taking vaccination into account.* For example, North Dakota and Georgia both have low vaccination rates, but Georgia has one of the highest hospitalization rates while North Dakota is about average. Is this because North Dakota has more "other measures" in place than Georgia? There are often news stories about restrictions, or the absence of restrictions, in different states, but I haven't seen any general measure, so I made one myself. It was based on a summary of restrictions by state constructed by the AARP. I coded them as follows:
-1 state limitations on local mask mandates; 0 "fully reopened," no mention of any mask mandates; 1 indoor mask mandates for certain people (e. g.. unvaccinated people or general mandates in some cities); 2 indoor mask mandates for all people; 3 restrictions for outdoors.
If you regress the logarithm of hospitalization rates on vaccination rates and the restrictions variable, the estimate for restrictions is positive--more restrictions go with more hospitalization--but the t-ratio is only about 1.6.
The problem with this approach is that states might tighten restrictions in response to a rise in cases, or loosen them in response to a decline. So it might be better to just look at a contrast between the "-1" states and everyone else. That is, limiting local mask mandates is a political move, not something a state is likely to do just because things are getting better. The states with limits on local mask mandates (Arkansas, Florida, Iowa, Montana, Tennessee) average about 30% higher hospitalization rates, controlling for vaccines. But that's all due to Florida--if you just look at the others, there's essentially no difference (and even with Florida, the t-ratio of the estimate is only about 1.3).
Another possibility is to use the party of the governor and Biden's share of the vote as predictors, on the grounds that Democratic states are likely to have stronger requirements for masking an social distancing than Republican states. If you do that, both the Biden share and having a Democratic governor go with a higher rate of hospitalizations given the vaccination rate. But the estimates aren't statistically significant (individually or jointly).
So overall, there's no evidence that "other measures" have an effect. Of course, I'm not saying that they don't--it seems certain that some measures will make some difference--but the way that the story frames the situation seems odd. A more accurate way to put it is that Florida shows that even a state with a moderate rate of vaccination can be crushed by the delta variant. Although Florida "emphasized vaccines" in the sense that it didn't implement other measures, it's only about average in vaccination rates (and has tried to prevent employers from mandating vaccines, although I think that a court has issued an injunction against that). I don't think there are any examples of states that have high vaccination rates and a history of few other restrictions.
*The vaccination rate is shots per hundred people.
I like this commonsense discussion and analysis. I'd prefer just plotting the state abbreviations without the dots. The dots seem distracting to me.
ReplyDeleteI had the same thought as I was making that plot, but couldn't figure out how to substitute abbreviations for the dots. I use Minitab, which is easy but sometimes limited in what it allows.
ReplyDelete