Data dropouts
Well, this had to happen sooner or later. More states are reporting less data, less frequently, regarding COVID-19. You can read the full writeup at this link.
This less-than-daily report is not a completely new phenomenon. It has always been true that some states — typically low-population rural states — did not report data on Sundays. This added a harmless little “flat spot” in their seven-day-moving average curves, but did not meaningfully distort the overall picture.
But now Florida (of course, Florida) has decided to report information just once a week. With some lag, as well. This is the only state where the new-cases data are now, in effect, missing for an entire week at a time, driving the seven-day moving average down to zero on a real-time basis. That began June 4. And that puts a serious kink in any sort of data analysis.
Across the states, the data reporting situation is summarized in the table below. Out of 51 states (including DC), about half continue to report every day. Those tend to be the larger states, and they contain about 69% of the population. The rest of the states skip reporting one or more days per week. For those states, the daily new case count will show as zero for those days, followed by some larger number as multiple days’ worth of cases are reported on the next reporting day.
Source for this and other graphs of new case counts: Calculated from The New York Times. (2021). Coronavirus (Covid-19) Data in the United States. Retrieved 6/11/2021, from https://github.com/nytimes/covid-19-data.” The NY Times U.S. tracking page may be found at https://www.nytimes.com/interactive/2020/us/coronavirus-us-cases.html.
For most of these states, the non-reporting remains a minor issue. One or two days per week, their time-series will show a flat spot. This will be followed by a jump as the time series corrects back to the true underlying level of new cases. A trend line that ends on (say) a Thursday (such as today’s analysis) will show all those states at their true levels.
But Florida is a special case. I can either drop it, or gap-fill it by using the prior week’s new-case rate as the gap-fill, until new data appear again. When all is said and done, it’s probably smarter just to drop Florida, for now, at least.
These changes, along with ongoing reporting issues in (e.g.) Texas and elsewhere, make it hard to say much about ongoing trends.
A leveling-off off of the new cases curve
With the data reporting caveats in mind, as far as I can tell, this past week, the downward trend of the U.S. COVID-19 fourth wave appears to have stopped. In my last look at trends, I overstated the downward trend due to missing data from Florida. But when I now drop Florida, and add the last few days’ worth of data, the new-cases curve appears to have flattened.
That may slightly overstate the flattening of the curve, due to some data reporting issues in Texas. (Texas added then removed a considerable number of old cases to their data last week, but without enough information to allow me to correct the time-series.)
If I divide the states by (my estimate of) the fraction of the population already immune to COVID-19, it seems fairly clear that this is not merely the consequence of bad data reporting in a few states. There’s still a slight downward trend in the states with the highest levels of immunity (black line), but a complete leveling-off in states with the lowest levels of immunity.
Why?
Why has the downward trend in case cases now flattened out?
There has been a lot of talk about the new Delta variant of COVID-19 (formerly the Indian variant, still identifiable as B.1.617.2, now simply termed “Delta”). But that’s still far too infrequent in the U.S. to have any material impact. As of now, it accounts for about 6% of new cases.
Source: CDC COVID-19 data tracker.
There are plenty of reasons to worry about this for the near future.
-
- It’s about 40% more transmissible (infectious) than the British variant that is now the dominant strain in the U.S.
- It causes much worse symptoms, particularly among younger people.
- The current vaccines are not as effective against it as they are against other strains.
You can read all that good news by clicking the link to this WebMD article. But, again, it’s not prevalent enough yet in the U.S. to be causing a flattening of the curve.
Seasonality is always a good guess. COVID-19 appears to have a strong seasonality to it. As we made the transition from winter to summer, disease transmission fell — just as it would for influenza, say. But now were in summer, and should expect no more seasonality-related declines.
Memorial Day weekend is not really a plausible suspect for this leveling-off of new cases. The main reason is that, despite the hype, there never was any post-holiday “surge” in the U.S., for any prior holidays (e.g., Post #922, Post #1029). Given that, it’s hard to believe that, uniquely, there would be one for Memorial Day, when most social activities are in safe outdoor locations anyway. The only thing that makes this even remotely plausible is the timing. Infections acquired at the start of Memorial Day weekend would start showing up in the data at the end of this current week. (June 9 is 12 days from the Saturday of Memorial Day weekend. That 12-day figure is my best estimate for the typical lag between infection and reporting in the U.S.)
If I had to pick a cause, I’d say seasonality. Part of the decline in new cases, after the peak of the U.S. fourth wave, was due to vaccination. And part was just ongoing seasonal trend. New vaccinations have slowed to a crawl. And we’re now fully into summer, and should expect no more help from seasonality this year. Absent those two drivers, it seems reasonable that the decline in cases would cease.
That said, it’s only been a week now. It’s too soon to make any firm conclusions. And the picture is muddled by data reporting issues from the Memorial Day holiday, and by changes in data reporting by the states. Give it another week and it should be clear whether or not the U.S. fourth wave has ended.
