Source: Calculated from case counts in the NY Times Github COVID data repository.
Above, the colored lines are plots of new COVID-19 cases/ 100,000 /day in six broadly-defined regions of the US. The thick black line is the U.S. average. The broad “dips” in that black line are artifacts of the holidays. The first is Thanksgiving, the second is the year-end (Christmas, New Year’s) holidays.
A lot of things affect reported COVID-19 case counts around the holidays. In some cases, public health departments simply shut down or are short-staffed, resulting in extremely low counts on the holiday itself, and high “catch-up” counts in the following days. Presumably, those should be a wash, more-or-less, within a few days of the holiday.
But as I noted earlier (Post #929), you actually see a temporary-but-real dip in diagnosed COVID-19 cases. Above, you see two of them now — one for Thanksgiving, one for Christmas/New Year’s. My best guess is, people “on the margin” with low-severity cases don’t bother to get diagnosed during the holidays. But that’s just a guess.
With all the ups and downs of the data, it’s not obvious that such a thing is happening. When this first occurred, at Thanksgiving, I didn’t notice it. I only noticed that when I tried to “smooth out” some of the Christmas anomaly, and couldn’t. And then found that the same thing had happened at Thanksgiving. And then realize that the dip in the case count is a consequence of the holiday.
I have crudely filled in those dips, above, with dotted lines that continue the immediate pre-holiday trend. Because I think that’s the true story, if you’re trying to figure out what the trend is, right now. You can’t take either that sharp downslope into the dip, or sharp upslope out of the dip, as an indication of the likely trend. All you can do is wait until you’re passed the dip, to see where you’re going.
Which finally gets me to my point. I’m getting ready to look for a post-Christmas surge in COVID-19 cases. In theory, that should start to show up between 12 days to three weeks after the start of travel and socializing for the holidays.
Which leads to the obvious question: when does that period start? With Thanksgiving, it’s such a brief holiday that this wasn’t really much of an issue. Given how loosey-goosey this whole “surge” concept is, I wouldn’t do any real damage by simply starting the clock on Thanksgiving itself.
But for Christmas, when does that 12-day to three-week clock start? The clock that will mark the point at which those “surge” cases should begin appearing in the data.
It seems like air travel for the holidays is the signature example of bad behavior during the pandemic. We’re always being told about the high rates of air travel use over the holidays, and how that presages the forthcoming holiday surge. (Which, if you follow this blog, you realize did not occur after Thanksgiving).
I finally decided to look up the numbers. And got a surprise. The vast wave of air travel that hits the U.S., just prior to the holidays? It’s more like a ripple.
Here are the data from the U.S. Transportation Safety Administration (TSA) website. This is from all “federalized” airports (which I think means all or nearly all that provide commercial passenger service). I’m unclear on whether this counts screening services provided via contractors, but I believe it does.
This is the fourth quarter of the year, and the blue line is 2020. The thin wavy line is the actual daily data, and the thicker line is a seven-day moving average. Around the Xmas holidays, the peak travel days are the Sundays — the Sunday before Christmas, the Sunday after Christmas, and the Sunday after New Year’s day. (And so I answer my question: The “holiday travel season” starts the Sunday before Christmas.)
But an interesting secondary point is that there is only slightly more air travel over the holidays than at other times. Maybe the peak, crunch times are vastly worse. But the average isn’t so different. The 2020 baseline rate of travel (from October, say) is about 900,000 persons screened per day. The Christmas/New Year’s travel is about 1,000,000 persons screened per day.
In short, for the 2020 travel season, we had maybe an additional 100,000 persons per day flying, for maybe 10 days. Or about 1M excess air trips. In the context of a population of 330M people, it’s hard to think that you could possibly identify the impact of additional infections from those additional flights, against the background rate.
(Don’t believe me? Let’s do the quick and dirty calculation: In the U.S., we’re now seeing about 65 new cases/ 100,000/ day. Let me assume that each case is infectious for a generous average of 5 days prior to onset of symptoms. (And that people who are obviously ill have the good sense not to fly.) Let me take an estimated 0.32% chance of each infected flyer infecting any of 15 fellow passengers (based on this research on Chinese train travel) When I sum that up, I get (100,000 persons/day x 10 days) (total additional passengers) x (65 x 5/ 100,000) (rate of boarding persons being infected) x (0.0032 x 15) likelihood of a boarding infected person passing on that infection = 156 new infections. The direct effect of those additional 1M flights should be to add 156 infections to the US total.)
I’m not saying that it’s smart to fly over the holidays. I’m just saying that there is no huge increase in air travel over the holidays. Yeah, about 1M people per day were flying over the holidays. But about 0.9M people per day were flying before the holidays. And that, given how small that increment is, the direct effect of that air travel, on US infection totals, is not going to be noticeable.
I tried to find the relevant car travel data, but they haven’t been published for the last week of 2020 yet. This is where you’d get the report on U.S. Interstate highway travel.
This is no way implies that there won’t be a post-Xmas surge. But if a surge occurs, it’s pretty clearly crazy to blame it on air travel. Air travel occurs all the time, and rises only slightly during the holiday season.