Post #972: The rain falls on the just and the unjust.

Posted on January 23, 2021

Source:  Calculated from NY Times Github COVID repository data.  Data reported through 1/23/2021.

U.S. new case counts are down to where they were before Thanksgiving.  And they are falling in all 50 states and DC.  That’s the only solid takeaway from this posting. The rest is just an explanation of why that’s so annoying.

The current situation

  • The U.S. rate of new COVID cases  per day is back to where it was before Thanksgiving.
  • The new cases rate is falling in all 50 states and DC.

Here are six regional graphs, showing the individual states.


And here’s the U.S. number, excluding California, just to be show that there’s nothing up my sleeve.  The current sharp decline isn’t an artifact of that one, large state.  And, in fact, the situation actually looks a little better if you exclude California throughout.


The sun rises on the evil and the good, the rain falls on the just and the unjust.

And it’s so annoying when that happens.

First, things seem a little too good to be true.  It’s as if someone rang a bell, and now it’s time for COVID new case rates to decline.   It’s not like new case rates are kinda-sorta falling in most places.  Not like they fell here, then they fell there, scattershot.  They are falling, everywhere.  And all of them started falling at virtually the same time.  With rare exception, the state peaks are within three days of 1/11/2021.

Is this just a temporary post-holiday un-surge?

I can offer one possible cause-and-effect explanation for this current sharp decline appearing simultaneously across all the states.  It’s so far outside of conventional wisdom that I hesitate to do it.  But here goes:  maybe this is the post-holiday un-surge.

Empirically, the “post-holiday surge” story has been wrong three times now.  Wrong for Canadian Thanksgiving (Post #916).  Wrong for Thanksgiving (Post #922).  And wrong for Christmas/New Year’s (Post #948).

What was the logic behind expecting a surge?  The story is that people travel and mingle with their families.  And that might spread disease.  And yet, e.g., increased air travel over the holidays is only minimal, relative to baseline (Post #939).  So the empirical basis for expecting a surge always has been thin, at best.  Which is a nice way of saying that people just kind of made up a story that seemed plausible.

But that story — travel and within-family socializing — is only half the story.  In addition, people don’t go to work, shops are often closed, and people do relatively little retail or other shopping. 

We’re all sitting home, watching TV, and eating leftovers.  Which, when you think of it, has the same effect as a lockdown.  Only voluntary.

And so, if telling a story is all the evidence you need, well, I can tell stories.  And the complete  story of the end-of-year holidays is that there’s a slight increase in air traffic, a large increase in family gatherings, and a large decrease in commercial activity of all sorts, including both going to work and visiting retail establishments.

When you tell the whole story, it’s not clear that there’s any case at all for a post-holiday surge.  If the reduction in cases from the reduced public activities more-than-offsets any with-family spread, you’re going to see a net reduction in new cases, starting 12 to 21 days after the holidays.  All at once.  Across all the states.

Kinda just like what we’re seeing now.   So I really would not dismiss the potential for the public health establishment not merely to have gotten it wrong about the post-holiday surge, but to have gotten it backwards.

And if that story is true, the current rapid declines will peter out 12 to 21 days after everybody got back to work and started shopping again. Or right about now.

And that makes the continuation of trend through the 23rd (and, we hope, beyond) interesting, in that it seemingly puts the myth of the post-holiday un-surge to rest.  If only we could do the same for the myth of the post-holiday surge.

The upshot being that the longer this trend goes on, the more we can be sure that it’s not some mere temporary post-holiday un-surge, but is instead a more permanent change in the direction of the pandemic.

Otherwise, there’s no rhyme or reason

Second, rates are falling for no apparent reason.  Got a mask mandate?  No mask mandate?  Under lockdown?  Bars open for business?  Arctic winter climate?  Tropical winter climate?  Got herd immunity yet?  Still have a ways to go to get to 70% immune?

It just doesn’t matter.  None of it.  The sun is shining on all of us.

And that’s unsatisfying.  In a mean-spirited way.


Source:  Harry Potter and Chamber of Secrets.

No reason means no guarantees

Third, there’s no way to get even a glimpse of the future of this trend.  And it sure would be nice to have that.  Because the flip side of “nobody knows why rates are going down” is “nobody knows when that’s going to stop”.

I looked at a handful of indicators and just gave up.  Near as I can tell, the only way to know if the trend is continuing is to wait for each additional day of new-case data.

I’ll just lay out a couple of the dud indicators, because these seem to get a lot of traction in the popular press.  But their “predictive ability” is purely folklore.

Test positivity.  Fact is, people get tested when they are sick.  And so, nationally at least, the fraction of tests that are positive just mirrors the new case rate.  It’s coincident with new cases, it doesn’t actually predict anything.

I could not get the raw data, so I have to rely on an existing graph.  But the peak and valleys on the right half of this graph coincide with the peaks and valleys of the US graph above.  The don’t lead (predict) the rates at all.

Source:  Johns Hopkins.

Another promising candidate for a leading indicator is hospital ER and OPD visits for COVID-like symptoms.  Presumably, that should rise at least a few days before the reported new cases rise, because it takes that long for test results to get reported.

But, in fact, the only daily data I could get turns out to be hilariously wrong, starting with the holidays.  The fraction of physician visits for COVID is almost perfectly negatively correlated with the COVID new-case rate.  (The likely explanation is that this is the percent of hospital OPD visits that are for COVID-like symptoms, and that people aren’t going to the hospital for elective (non-emergency) services over the holidays.  Leaving only the emergency cases, and so increasing the COVID portion of all cases.)

Source:  Calculation from data from Delphi COVIDcast,


All of a sudden, the sun is shining.  On the just and the unjust.  Maybe as an unexpected bonus from the holidays.  And so, temporarily.  And maybe not, and so with an expectation that it will continue.  The longer this persists, the less likely it is to be temporary (duh).

And if this isn’t an artifact of the holidays, then it may be just the natural progression of the pandemic.  No rhyme or reason to it.  All human coronaviruses are seasonal to some degree.  Maybe now is the time for COVID-19 to go out of season.

As I argued at the end of December, the odds have been shifting in favor of a decline.  But I did not expect to see anything like this current decline.  It’s just a little too neat, too correlated, and too universal.

But it is what it is.

On the bright side, for the time being, all the predictions of hundreds of thousands of new deaths in the coming weeks seem to have dried up in the face of this recent decline.  So at least you aren’t seeing that tossed in your face when you read the papers.

But there’s no data-based way to know if this will continue.  The only way to know if this trend continues to be a trend is to look at the next day of data, when that arrives.

A trend is a trend until it ceases to be a trend.  That’s cold comfort, but that seems to be the reality of it.