The CDC page that launched a thousand bad takes
TLDR: death reporting lags
There’s nothing wrong with the CDC’s webpage “Provisional Death Counts for COVID-19.” It does what it’s supposed to do: tally total mortality based on death certificates forwarded from local health authorities.
It’s just: social media is often dumb.
And forgetful, as Matthew Yglesias observed:
In this article, I explain what the CDC Provisional Deaths page shows and outline three distinct waves of bad takes based on accidental and deliberate misunderstandings of the page.
How CDC’s Provisional COVID-19 Deaths work
Most confusion about the page originates from the fact that it doesn’t offer the same up-to-the-minute (-ish) counts of COVID-19 deaths as state health department reports which get aggregated by COVID Tracking Project, Worldometers, HHS, and the CDC itself elsewhere on its site. These sources currently show about 340,000 deaths compared to only 303,823 as of December 31 on the Provisional Deaths page.
Why the gap? Well, since the beginning of the pandemic, states try to report COVID deaths quickly to reflect the current status of the pandemic. These “COVID dashboard” numbers usually report deaths that occurred days or sometimes even weeks ago, so they aren’t quite real-time. But dashboard reporting occurs much more quickly than it takes for coroners finalize death certificates which are then forwarded and manually coded by the CDC’s National Center for Health Statistics.
Careless and dishonest people have used the lag to launch three waves of bad takes.
First wave: CDC slashed COVID-19 deaths!
CDC created the Provisional Deaths page on April 3 and it was almost immediately misunderstood and misused by folks minimizing the pandemic.
While I don’t blame the CDC, the page confuses people who encounter it and don’t read the lengthy notes surrounding the data tables. For example, the very first time it was publicly linked on Twitter, it was a misunderstanding by a commentator essentially accusing California of exaggerating their COVID deaths. Yet the the Provisional Deaths page explained (as it does now) that the provisional count from death certificates varies from other sources and that “data currently are lagged by an average of 1–2 weeks.”
Mortality death reporting lags:
On April 19, Alex Berenson posted screenshots showing that overall mortality reported at the time was only 90% of normal, with comment “Judge it for yourself,” followed by tweets claiming that “because of the aggressive way in which we code #COVID deaths … changes in all-cause mortality will be the ONLY reliable way to judge the death toll.” Berenson did not understand at the time (or did not care to explain to his followers) that death reporting lags. The week ending April 18 was actually near the peak weekly death toll ever recorded in the United States, 76,675 deaths, eclipsed only by the previous week and a whopping 42% above normal.
The first wave of bad takes began in earnest the evening of May 1-2, 2020, when putative comedian Tim Young suggested that the death count had been “revised” to something much lower than the 64,000 or so deaths reflected by dashboards at the time:
Yes, he read it wrong.
Not content with his disingenuous question, and after dozens of people pointed out his mistake, Young falsely claimed the next day: “They've now segmented the deaths to show that only 37,308 are directly from COVID-19.”
The gates opened, flooding social media with thousands of bad takes about how the country had “been played.” Fact checkers quickly explained that this was nonsense, as the Provisional Death page itself explained. Even Berenson, who had weeks earlier misleadingly cited the page warned his followers: “Making sloppy mistakes makes it easy for Team Apocalypse to discredit serious and accurate arguments.”
Some who innocently spread the error, like Joe Rogan, conscientiously passed the correction on to their followers. Others, like Tim Young, demonstrated their complete indifference to truth:
But once hatched, viral nonsense is hard to dispatch. People continued to cite the Provisional Deaths page for months as if it contained “real” or more accurate death figures.
It would probably still be happening today except the difference between dashboard and Provisional Deaths count is “only” about 37,000 lives—and 303,000 deaths is still a big number that doesn’t make COVID-19 seem anything like the flu.
So the bad takes evolved.
Second wave: deaths are going to zero!
The second wave took off on July 14, when an attorney at a firm that filed dozens of lawsuits against California and its municipalities over COVID-19 restrictions tweeted this data from the Provisional Deaths page:
Other users jumped in to retweet, remix, or simply plagiarize the tweet.
The second wave of bad takes capitalized on the dominant COVID-minimizing narrative at the time: that cases were no big deal because deaths (until the beginning of July) were dropping. But of course, deaths follow weeks after cases are identified, and deaths inevitably started rising again in July, which was reported. This second wave of bad takes was a desperate attempt to deny the second wave of deaths.
This bad take is still with us. It reemerged in October for the same reason it popped up in July. The third wave was becoming clear, and minimizers wanted to exaggerate the low number of deaths in the face of spiking cases. Donald Trump Jr. posted a graph based on the Provisional Death date on October 27. This attracted fact checks when Trump Jr. absurdly told Laura Ingraham deaths were “almost nothing.”
I expect we will see more takes like this in the future, even though it’s possibly the dumbest of the three misinterpretations. Even a casual glance at the Provision Deaths table shows that recent weeks are significantly under-reported.
Obviously, deaths did not drop by 70% last week. That would be big news—like a science fiction story. But I’ll safely wager we’ll see more bad takes like this.
Third wave: annual deaths are normal!
The last, and currently most common Provisional Deaths bad take asserts that total mortality in 2020 is unremarkable compared to recent years. False. Even with incomplete reporting, we already know 2020 was an all-time record for deaths in the United States (surpassing 2.855 million in 2019):
There are two problems with using the Provisional Deaths page for this comparison. First, as discussed above, deaths lag. Second, [until Jan. 6 2021] the Provisional Deaths page started on the week ending February 1—it omits almost four weeks of deaths entirely. That’s because the page was never supposed to show all deaths in 2020, but instead count deaths coded for COVID-19.
TLDR: The Provisional Deaths total is incomplete on both ends.
Unlike the first two waves, this idea percolated for months before becoming popular. I first remember seeing it in August, and a non-competitive Congressional candidate was an early spreader:
But the meme didn’t really take off until late November on Instagram and Facebook, which is ironic because that’s about the same time that (incomplete) CDC-tallied deaths for 2020 passed every previous year on record.
Fact checkers wrote it up in November, but to no avail. The idea gained even more traction after an undergraduate student newspaper at John Hopkins University wrote up a webinar with the erroneous conclusion that COVID-19 caused no excess deaths compared to prior years.
There are excess deaths! Even more than the number of recorded COVID-19 deaths, which the CDC tracks on another page.
More examples of the meme were posted later in November, with the most successful version was posted on December 13 by Blaze TV personality Jon Miller.
Not content to rely on the then-current version of Provisional Deaths (which then totaled 2.72 million deaths without including January), Miller instead linked to an archived version of the page from November 17. Miller’s viral nonsense was widely spread and imitated on the right (oh look, it’s Tim Young again!).
More fact checks occurred, but this bad take continues to spread, un-updated as copypasta, even though the year is over and the numbers are obviously wrong.
I suspect the third wave will die out soon, but I’m certain that folks will find new ways to misuse the Provisional Deaths page.
Update Jan. 12: On January 6, CDC changed the provisional deaths page to include deaths back to the week ending January 4, 2020, so it should be much, much harder for the third wave of nonsense to spread. At this time, the total is already clearly over 3.2 million, which is almost 400,000 higher than any previous year, and will climb higher still.
Viral misrepresentations punctuated by a “source” spread much more quickly than comprehension about how the source has been misused. The recent excitement over the meta-analysis that “finds no evidence of asymptomatic or presymptomatic spread” (false!) illustrates this principle.
I hope that 2021 brings fewer bad takes than 2020, but seems doubtful.