With an introduction by Markus Brunnermeier, Director of the Princeton Bendheim Center for Finance
On Thursday, October 22, Bengt Holmstrom joined Markus’ Academy for a lecture on the seasonality of Covid-19.
Bengt Holmstrom is a Professor of Economics at MIT, and a 2016 recipient of the Nobel Prize in Economics.
Historically, viruses show a pattern of seasonality with peaks at the turn of the year and troughs in the summer. This holds for the past 10 years of across various corona viruses and the Spanish flu of 1918 as well, and holds across countries. Examining a 14 day sum of new COVID-19 cases across 9 countries, this pattern holds for the past eight months, with the exception of the US.
This seasonality of COVID-19 may be due to UV variation. For instance, in France, increased UV in the summer was correlated with a decrease in COVID cases and a low UV was correlated with higher cases in March and in the recent weeks. The same is true for a string of other European countries. In contrast, for countries in the southern hemisphere, such as Australia and South Africa, there was a peak in our summer, when UV was the lowest.
Mobility and policy stringency are also correlated with COVID-19 cases. With mobility data from Apple and Google maps and looking at Western European countries, there was a large decrease in mobility during the March lockdowns (to varying degrees between the countries). However, mobility has returned to normal levels since the summer with varying degrees of infection rates across the countries, so mobility cannot explain all the infection rate changes. Policy stringency is reactive to the virus, so there is a lag in the correlation data between policy changes and infection rates, leading to endogeneity issues.
Regressions show that the estimated long-term elasticity of UV as an explanatory variable of new COVID-19 cases is quite stable across specifications controlling for mobility and stringency independently.For example, in France, a 1% permanent increase in UV radiation was estimated to reduce new COVID-19 cases by 3.43%. The long-term elasticity was stable also across countries like Germany and Finland. This shows that the UV effect is quite strong, especially compared to the 70-day lag present in mobility changes explaining infection rates.
A full cycle is needed to further examine the effect of UV radiation. In addition, while UV radiation is exogenous, it may impact behavioral patterns (i.e. people will go outside more when it is warm). Further research can look at the impact of altitude — which drives variation in UV and can be used as a potential instrument — as there is already evidence of less cases and shorter virus lengths with higher altitudes. Within-country variation in latitude may prove to be useful in a similar vein.
If UV radiation is found to affect infection rates of Covid-19 as suggested by these preliminary results, peak infections will occur around the end of the year in the northern hemisphere. In the fall, declining UV accelerates the spread of the virus and in the spring it decelerates it. Summers are likely to be relatively virus free if the virus keeps returning, a pattern familiar from earlier corona viruses.
A seasonal pattern allows proactive policy measures and helps people cope.