Post #909: Thanksgiving data artifact. I think. Now confirmed.

Posted on December 4, 2020

Edit:  One more day of data confirms that this was just a data reporting artifact.   True Thanksgiving-generated cases, if any, should begin showing up in the data next week.

Original post follows

Based on data reported through 12/3/2020, many (but not all) states are showing large upticks in the seven-day moving average of new cases per day.  As shown above, circled in red.

After looking at the details, I think that’s probably an artifact of Thanksgiving data reporting.  That is, I don’t think this is the start of a rapid increase in cases due to Thanksgiving-related travel.  Initially, I thought that all of that would have “washed through” the data reporting systems by Tuesday at the latest.  But here’s why I’ve changed my mind:

  1. The true increase in cases due to infections spread at Thanksgiving should not start to show up in great numbers for another few days yet.  My best guess was that those cases would begin appearing in the data around 12/7/2020.
  2. The upticks are highly uneven.  Some states (e.g., Virginia, New York, New Jersey) show nothing.  Others show huge apparent increases.
  3. The states with the largest current upticks are also (in the main) the states that showed the largest concurrent down-ticks around Thanksgiving day itself.  In other words, this seems to be occurring mostly in states with reporting systems that seem sensitive to the disturbance caused by Thanksgiving.
  4. Where there are upticks across the states, they are perfectly coordinated as to timing.  All the bottoms of the tick marks line up.

Source:  Washington Post, downloaded 12/4/2020.

For the US as a whole, this means that we didn’t actually see a true drop in new cases last week.  Likely, the entire “dip” at the end of the current US chart is entirely an artifact of temporary distruption of testing and reporting due to the holiday.  That’s circled in red above.

And if this is just a matter of swapping around the timing of when cases were reported — so both the decline and the following uptick are just a matter of shuffling of paperwork across dates — then the right thing to do is eyeball that peak, and average out the dip and uptick.  When I do that, it looks like things average out to flat-to-maybe-a-little-down, over that entire week or so.  And my guess is, that’s closer to the true situation than the little roller-coaster that’s pictured above.

Let me put it this way.  If those upticks were real, and represented the start of the post-Thanksgiving surge, then a lot of states are in serious trouble.  But I honestly don’t think that’s the case.  I think that’s the last short-term impact that Thanksgiving had on testing and reporting behavior.  And we’ll see the true impact of the holiday on infection rates starting next week.