Does COVID-19 have seasonality effects?
An underpowered and too-preliminary analysis.
Starting point: this paper which suggested COVID-19 spreading most rapidly in a temperature range 5-11 C and Relative Humidity 47-79
papers.ssrn.com/sol3/papers.cf…
An underpowered and too-preliminary analysis.
Starting point: this paper which suggested COVID-19 spreading most rapidly in a temperature range 5-11 C and Relative Humidity 47-79
papers.ssrn.com/sol3/papers.cf…
We revisit the JHU CSSE data previously used here; merged against NOAA temperature records here ncdc.noaa.gov/sotc/national/…
Generate three temperature bands (no humidity info for now):
- Colder than 41 F (5 C)
- "Critical Temp" of 41 - 52 F (5-11 C)
- Warmer Than > 52F
Use actual Feb temp; last year's Mar temp
- Colder than 41 F (5 C)
- "Critical Temp" of 41 - 52 F (5-11 C)
- Warmer Than > 52F
Use actual Feb temp; last year's Mar temp
First we look in levels, regressing against total daily state cases and clustering at state level. If anything, states in the "critical temp" zone have fewer cases.
Then I add a control for previous daily cases. Unsurprisingly that matters a ton. The monthly temperature effects are really noise.
Last I try interacting previous daily caseload with temperature. There is VERY weak evidence that the colder, warmer states have faster transmission rates. Not much evidence this is true in that "critical region."
I can't emphasize enough the caveats here:
- Relying on reported COVID caseload, and state differences reflect testing capacity as much as "true" cases
- Using monthly average temperatures
- Not incorporating humidity
- CX error correlation not adjusted by State clusters
- Relying on reported COVID caseload, and state differences reflect testing capacity as much as "true" cases
- Using monthly average temperatures
- Not incorporating humidity
- CX error correlation not adjusted by State clusters
Still, given the prominence of the "seasonality" paper I thought it was worth at least trying to check out the hypothesis
Mostly, I find that COVID cases at state level are increasing more or less parallel trend across climate zones
Mostly, I find that COVID cases at state level are increasing more or less parallel trend across climate zones
Of course, this is only using the variation across American states.
The most favorable data points for a seasonality perspective, IMO, have been limited exposure so far in places like Thailand or Malaysia - with many Chinese contacts and decent testing capacity
The most favorable data points for a seasonality perspective, IMO, have been limited exposure so far in places like Thailand or Malaysia - with many Chinese contacts and decent testing capacity
Could be that such tropical areas have very different transmission properties outside of US range of variation.
Could also be that testing has been too minimal there, or variety of other factors. Or a matter of time.
Could also be that testing has been too minimal there, or variety of other factors. Or a matter of time.
So I wouldn't conclude from this much more than: there is limited evidence of weather patterns, across US, in driving cross US transmission rates; using too short of a sample and limited data; in this particular analysis
I am obviously no expert here, so welcome any feedback
I am obviously no expert here, so welcome any feedback
