On the Seasonality of the Virus

What are the odds that the coronavirus will recede on its own during the spring because of warmer temperatures or a higher ultraviolet (UV) index? This has been a question from the beginning.

There has been some research in support of the idea that the warmer season would force the virus to retreat. And there has been other research that concluded that the virus would retreat but not disappear, that it would survive in the southern hemisphere and that it could then stage a comeback in the northern hemisphere in the fall when cooler temperatures return.

Single Variables

Looking at the United States state by state, we find little correlation between the number of deaths per capita and the UV index. For example, Wisconsin with a UV index of 4 in March has so far suffered 25 deaths per million inhabitants, but Rhode Island also with a UV index of 4 saw as many as 60 deaths per million. At one extreme, New York, New Jersey, Michigan, Connecticut and Massachusetts, all with a March UV index of 4, had over 100 deaths per million. At the other extreme, South Dakota also with a March UV index of 4 had only 7 deaths per million. (All deaths figures are as of 12th April 2020 per Worldometer).

Clearly the UV index on its own is not a determinant of the spread, whether we measure that spread by the number of deaths or by the number of confirmed cases.

Another important single variable is the international exposure of each state. This approach is based on the premise that a higher international exposure would lead to a wider spread of the virus. In order to run this exercise, we quantified international exposure as the number of international air passengers that arrived in each state in 2018, via its gateway airports. Here again we found no clear correlation by using this variable on its own.

For example, there were 3.7 million arrivals in Michigan through Detroit Airport and Michigan has so far had as many as 149 deaths per million. But California, Florida and Texas had 42.8 million, 36 million and 20.4 million international arrivals, each through its own multiple gateway airports. And none of these states has had more than 22 deaths per million so far.

So here again, international exposure alone does not provide a good indication. It is true that the virus is occurring in winter and early spring and the figures for annual international travel differ from month to month. It is likely that these sun belt states see many more international arrivals in the summer than in winter. Nonetheless the difference between 3.7 million in Michigan and 20+ million in CA, FL and TX means that it is not necessary to make the adjustment to monthly figures. In any given month, there is no doubt that Los Angeles plus San Francisco plus San Diego receive far more international passengers than Detroit does.

Multiple variables

The next task then was to combine the two variables and to see if states with 1) a low UV index and 2) a high international exposure have seen a larger number of cases and of deaths than states with only one or neither of these conditions.

The result here was more encouraging.

The table shows all states with a March UV index of 5 or less. Click to enlarge.

Screen Shot 2020-04-13 at 9.22.03 AM

If we divide the states in four quartiles by number of deaths and separate out all the states with a March UV index of 5 or lower, the outcome is very clear.

In the top quartile (the most deaths per capita, in yellow in the tables) are the tri-state area encompassing New York, New Jersey and Connecticut (fed by JFK, Newark and La Guardia in this order), the Massachusetts-Rhode Island corridor (Boston airport), the Illinois-Northern Indiana region (Chicago O’Hare and Midway), the Washington DC region (Dulles, Reagan National and Baltimore) and the states of Michigan, Colorado, and Pennsylvania (Detroit, Denver and Philadelphia).

At the top of the second quartile (in tan in the tables) are Vermont impacted by visitors from Boston and New York, Maryland and Delaware impacted by the tri-state area on one side and the DC area on the other, and Wisconsin marginally affected by Chicago.


So far so good. But there are three notable outliers. There are two states, Georgia and Louisiana, that are performing worse than expected because they have a high death count despite the fact that they do not meet both of the conditions above. And there is one state, Minnesota, that is doing better than expected because it has a low death toll despite the fact that it does meet both conditions.

The table shows all states with a March UV index of 6 or more. Click to enlarge.

Screen Shot 2020-04-13 at 9.22.03 AM 2

Louisiana meets neither condition. Although it has a March UV index of 8 and New Orleans is not a significant gateway for international arrivals, Louisiana has one of the highest death tolls in the nation at 180 per million.

Georgia meets one condition by having a high number of international arrivals through Atlanta airport. But it has a higher March UV index at 6 than the most impacted states of the Northeast. Georgia is at the top of the second quartile with 43 deaths per million.

Finally Minnesota has a low UV index of 3 and a relatively high number of international arrivals and should be squarely in the first quartile. Instead it is at the bottom of the third quartile with 13 deaths per million. In theory, Minnesota should be doing as badly as Michigan or Colorado.

Testing disparities

So what explains these outliers? One answer we can offer has to do with the amount of testing. On average across the United States (excluding New York), the percentage of people testing positive is 15.3%. For every 100 people tested, 15 are found to have the virus.

However, in Louisiana and Georgia the percentages of positives are 20.7% and 23.5%, higher than the national average. This suggests that these states are not testing enough and that they are behind the curve in identifying and isolating people who may spread the virus. In the case of Louisiana (a much bigger outlier than Georgia), it is also said that Mardi-Gras parades may have played a role in accelerating the spread.

In Minnesota by contrast, the percentage of positives is 4%. At this level, testing seems adequate to identify and isolate the people who are infected. For comparison, the percentages of positives in Michigan and Colorado, two states that are most similar to Minnesota in our analysis, are much higher at 31.6% and 19.9%.

Generally, none of the states in the first quartile is testing enough. The percentages of positives in New Jersey, New York and Connecticut are very high at 48.4%, 40.9% and 28.9%. In Illinois and Massachusetts, the percentages of positives are 21.4% and 21% and so on.


Our view is that there is sufficient evidence to support the hypothesis that there is at least some seasonality. The UV index will rise in all states in the spring and summer. On our two conditions, the New York, Massachusetts, Illinois, Michigan and Washington DC regions will look in June-August like Texas in March or like California in April. Colorado will look like Florida in March or Texas in April.

This all may not be the last word on seasonality but we view it as supportive of the notion that the change of season will cause the virus to subside. Whether it returns in the fall will clearly depend on testing. Minnesota offers an important lesson here.

Related: A Few Certainties About Covid-19

In addition, there are commentary and charts on the coronavirus in the Wednesday Briefs. Access all Wednesday Briefs here.


Leave a Reply

Your email address will not be published.