Post #1026. Has the six-week downward trend bottomed out?

Posted on February 24, 2021

I’m not sure I can blame Texas for all of the recent change in the COVID-19 new case trend.  Today it’s starting to look like the steeply downward trend of the past six weeks is bottoming out.

Let me do something a little out-of-order here.  Take a look at the picture below.  See the little upturn in the case count that just happened?

That’s Canada.  This is the US, below.

Given that the peaks occurred on almost exactly the same day, having the troughs occur on almost the exact same day is startling.  More on that at the end of this post.  But, for sure, it’s not like the Texas grid failure caused an uptick in Canadian cases.

Next, here are the charts I regularly produce.

Texas:  OK, first, sure, the return to normalcy in Texas has some effect on the US numbers.  The button-hook at the end of the gray line (South Central states) is due to Texas, as discussed in prior posts.  And Texas has a ways to go yet to return to prior trend.  So Texas’ data reporting issues are affecting and will continue to affect the U.S. trends for some time yet.

Much broader than that, though.  And yet, if I return to my stock-chart-style gainers-versus-losers chart, the majority of states had an uptick in rates, as of the last available day of data.  Only 22 states saw a decline in the new case count yesterday.

And, in fact, if I simply toss Texas out of the U.S.A. entirely …

And so, ex-Texas, we’ve still got a pronounced leveling-off of the downward trend, at the end of the ex-Texas line.

And now, a brief musical interlude:

Finally, if I take just the February state data, and graph that, by eye, there appears to be a bottom forming in the new case rates.  It’s clearest when I use a log scale, so that’s what I present below.

Your eyes are often pretty good at integrating data.  I see a fairly distinct shallow bottom occurring in the red circled area.  That bundle of lines went from a clear, linear downward trend to, well, something else, right at that point.

Now for the hard part:  What, if anything, is causing this?

To be clear, with a couple of day’s worth of upturn, there’s no definitive answer yet.  And there may never be.  But it’s worth laying out the leading candidates.

Just a weather artifact?  Probably not.  Plausibly, what appears to be a coordinated change might just be the last little ghost of an impact of last week’s cold snap.  For Texas, most of what you are seeing is a pure data reporting issue.  They just literally didn’t tabulate the test data from a lot of places.  But there might also be a behavioral issue, in that people might have postponed getting tested due to the intense cold.

And so you might see some “catch up” testing results coming out of the system now.  Not a genuine increase in infections, but a jump in infections being caught due to a jump in testing.

A problem with that explanation is that, for the U.S. as a whole, there has been no jump in testing following the cold snap.  You can see the day-by-day totals at this website sponsored by The Atlantic.

A second problem with that “post-cold-snap-catch-up-testing” explanation is that most states don’t show any obvious dip in testing during or just after the cold snap.  Just look at a few of the states in the graphs below.  There are few or no obvious cases of a steep fall followed by a rebound.  Instead, in most cases, where there is a change, the line simply stops falling and levels out.

I note that it’s still too soon to see any impact of that cold weather event on the actual underlying rate of new infections.  That ought to take (by my estimate) 16 days to filter into the reported numbers.

Super Bowl? Maybe.  The timing is pretty good for this to be the post-Super-Bowl “explosion” of cases.  I date the turn of that shallow bottom to 2/21/2021, which is exactly 16 days after the Super Bowl.  And 16 days is exactly the lag that I would expect, and have talked about at length in prior postings.

(I realize that in prior posts I mocked the whole idea of a post-Super-Bowl “explosion” of cases, but the data are what they are.  Right now, that’s at least as plausible as any other explanation due to the timing of the upturn.)

But what about Canada?  Turns out, a high a proportion of the Canadian population watches the Super Bowl.  This source shows a peak of 11M viewers, or about 29% of the Canadian population, versus 150M in the U.S., or 45% of the U.S. population.   So you can’t rule out a post-Super-Bowl explosion based on the simultaneous bottoms in the U.S. and Canadian data.

That said, the patterns by state don’t really support this well.  First, it’s far from uniform.  For example, there’s no change in either New York or California.  Don’t they have Super Bowl parties there?  Beyond that, the upticks are not disproportionate in Florida, Kansas or Missouri, where one would expect there to have been a disproportion fraction of the population involved in Super Bowl parties.

End of seasonality?  Possibly.  An alternative explanation is that the natural seasonality of COVID-19 is running its course.  Which is not intrinsically different from saying “stuff happens”.

This has some support in terms of the places where we are and aren’t seeing this leveling off or upturn.   I first spotted a leveling-off and upturn among the states that led the way up on the U.S. Third Wave.  And now I’m still not seeing much of an upturn at all on the U.S. coasts, which were the areas that lagged on the third wave.  E.g., in the chart above, almost all the U.S. South Atlantic states continue on a downward trend in daily new cases.

The U.K. variant?  No, it’s too soon for this to be due to the U.K variant.  That remains on track to become the most prevalent strain in mid-March.  The most recent data from Helix Corporation, based on moving averages posted through 2/21/2021, shows that the B.1.1.7 (U.K.) variant accounted for (my estimate) 13.5% of all U.S. coronavirus cases.  (That’s from combining the two SGTF tables).  Given sample sizes and such, that’s consistent with (my) predicted 15%.  It’s not yet prevalent enough to be driving new case rates upward.