It is not surprising that moving out of lockdown resulted in more COVID-19 cases. In addition, the increasing number of tests undoubtedly increases the number of known infected.
But could there be something else going on?
Could increased use of air conditioning, particularly in the southern tier of states, be a significant driver of increasing number of COVID-19 cases?
This blog will attempt to help answer this question.
So where is the virus really spreading? A good way to see the problem locations is to view the percentage of positive tests. A worsening epidemic is signaled by a higher percentage of positives, assuming there is widespread testing. Positive percentage is far better than number of positive tests, which, of course, varies by the amount of testing.
Here is a plot of the positive percentage on July 7th. The big problem states were Arizona, followed Mississippi, Florida, South Carolina, Texas, Georgia, Alabama, Nevada, and finally Idaho.
Below is a different type of plot that shows the same thing, but provides the actual numerical values. The bottom line: the situation is far worse for states along the southern tier of the U.S. Arizona is the worst, with Mississippi and Florida right behind. These are states with very different demographics.
But what do these states have in common? Some media outlets are pushing the fact that most of these states are dominated by the Republican party and have been quicker to open up. But they have something else in common: these states have had high temperatures with a lot of air conditioning use. (And no, there is no reason to think that heat turns people into Republicans).
If we look at the high temperatures in June (shown below, NOAA division dataset), southern Arizona (including Tucson) is the nation’s hot spot–and yes, it is the hot spot for COVID-19 as well. Mississippi, South Caroline, Florida, Texas are all very hot. And according to U.S. Census data nearly all homes and most restaurants in these states have AC.
And an independent graphic, showing the high temperatures averaged over the 30 days ending July 7th (Climate Prediction Center), has a similar pattern. Arizona has the highest temperatures.
Now is this hypothesis consistent with observations? We can begin by looking at the total tests and percent of positive tests in Arizona (see below). Tests went up substantially in May and June, but so did the percentage of positive tests, which has progressively risen since mid-May (the largest increase was in mid-June)
So what happened in Tucson, located in southern Arizona during June? Temperatures exceeded 100F on many days and over half of the month was above normal (green shows the normal range). Some days were way above normal. June is the worst month in southern Arizona–very, very hot without the relief of the southwest monsoon in July. Air conditioning was a necessity and this miserable period is exactly when the virus surged.
Florida had a similar story. Positive percentages surged in middle and late June.
And this is exactly when temperatures surged to way above normal in southern Florida (see below). And Florida has terrible humidity as well. Folks were forced to flock to air conditioned spaces.
You want something more rigorous? No problem.
If I was writing a paper on this topic, I would present a scatter diagram plotting the temperatures against positive percentages of COVID-10. And I have done exactly that below. Specifically, I found the June average maximum temperature for every state in the continental U.S. and its corresponding positive percentage for COVID-19 (Y-axis percentage, X-axis is average high temperature). Each state is shown by a blue dot. I only plotted states with max temperatures in June of 75F or more, which excluded a handful of states that are very cool and have very few air conditioners (e.g., WA, OR, and Montana).
I also plotted a best-fit line (red). There DOES appear to be a relationship between COVID-19 infection rates and temperature. The correlation coefficient is .481, which suggests this relationship explains about 23% of the variability. The point in the upper right corner?–Arizona.
Now certainly there are a number of factors that help explain the variability of COVID-19 infection rates around the U.S. But I do think the above results are very, very suggestive that very warm temperatures result in increasing infection rate. Not because the virus likes warm temperatures (it does not, as shown by a number of studies), but because warm temperatures push people indoors into air conditioned spaces in which spread is greatly enhanced. Restaurants and bars are probably key here.