by Jon Sanders
Research Editor and Senior Fellow, Regulatory Studies, John Locke Foundation
On June 24, Gov. Roy Cooper announced that on June 26 he would order everyone in North Carolina to “wear face coverings when in public places where physical distancing is not possible.” He was also not lifting any restrictions, ordering the state not to open any further beyond not-yet-Phase-2.
In so doing, Cooper told everyone that “North Carolina is relying on the data and the science to lift restrictions responsibly, and right now our increasing numbers show we need to hit the pause button while we work to stabilize our trends.”
Cooper specifically made stabilizing our trends the goal for lifting restrictions. This was in keeping with past announcements.
Mandy Cohen, NC Health and Human Services secretary, told everyone that the “best way” North Carolinians can be “helping our neighbors” is “by taking the simple action of wearing a face covering that covers your nose and mouth. If we each do our part, we can get back to the people and places we love.”
Cohen, who prefers to speak to adults as if they’re fidgeting in line to see Big Bird, had thereby ratified the administration’s position that wearing face masks would lead to stabilizing trends so that Cooper would relax his restrictions on people and places.
A week before, Cohen had presented to the NC House Health Committee the administration’s “science and data” on face masks. They were emerging studies with questionable application to NC, but they promised immediate and dramatic turnarounds in the numbers of COVID-19 cases. On them Cooper and Cohen determined to force rather than merely recommend face masks.
A nagging question remained: if “the science and data” backed forced face-mask wearing, then why didn’t Cooper lift any restrictions?
Tellingly, that’s a question that went largely unasked for several weeks following the order. That could be because the media talking point — which even made it to Newsweek and NPR on July 30 — was that NC was in the midst of a spike in coronavirus cases.
The very next day, July 31, Cohen was telling NPR that the masks had worked, and that they had started working 2–3 weeks after Cooper’s June 26 order. This was in fact the administration’s updated talking point, which they had unveiled at a July 28 press briefing.
It was then Cohen announced that the state’s trends had shown “early signs of stabilizing.” Cohen specifically credited the face mask order: “Specifically, we see a direct correlation to the start of the statewide mask requirement. At the end of June, two to three weeks after implementing this requirement, we started to see the beginning of these more stable trends.”
Politifact, perhaps spurred on by a tweet from Brent Woodcox, special counsel to the General Assembly, deigned to “fact-check” Cohen’s declaration of a “direct correlation to the start of the statewide mask requirement.”
To be clear, this was a fact check of declared direct correlation. Politifact called it “Mostly True”:
Data shows that the number of new reported COVID-19 cases is beginning to stabilize. Experts told PolitiFact the mask law and case count could be related, but that it’s hard to quantify the law’s impact.
Several quibbles: Data is a plural noun, the mask order is not a law, and incidentally, could be related and hard to quantify aren’t the usual companions of “direct correlation.” But like Cohen’s emerging research, Politifact chose to center on COVID-19 cases.
Also now taking stabilization as a given, Politifact consulted academic experts. They were cautious but obliging. A Harvard University research professor of public health, Barry Bloom, pointed out that it is “difficult to disaggregate a single intervention” from the others, but said it was possible the mask order could have led people to adopt masks and other precautions, which he called “safe to say they may have contributed to more stable trends.”
An epidemiologist from the University of Nebraska, Ali Khan, could not agree to credit the mask order directly for “the leveling-off of cases,” not yet, but said it could be a “fair” claim, based on “other data that [purport to show] masks work.”
A UNC researcher, Ryan McNamara, acknowledged “a lot of variables at play” and explained that it would take months to review all the data. This portion is worth quoting in full:
There’s a popular saying in research: “correlation does not mean causation.” In this context, it means that the mask law may not have caused the dip in new COVID-19 cases, even though there are signs a cause-and-effect relationship could exist.
“That said, Dr. Cohen and her team are in possession of much more data than I am. I think her statement was carefully worded and has scientific merit.”
The idea that Cohen is telling the truth because she could be telling the truth and could have access to data the rest of us can’t see may strike some as compelling, but “safe to say” it “may be” “possible” that others “could” find this line of reasoning rather flimsy.
More of interest, we see there is question over what “caused the dip in new COVID-19 cases” amid “signs a cause-and-effect relationship could exist.”
Here is a graph showing the trend from May 1 to August 11. I’ve marked when the mask order took effect:
As advertised, there is a continued increase for weeks past the order. The slope of increase seems, if anything, to become steeper for a few weeks, suggesting an increasing amount, not the beginnings of a tapering off.
Then there is the taper, followed by a clear decline. If you’re looking for evidence of the mask order working, and you’re told to expect not to see any reduction for several weeks, then this fits. You can’t call it a direct correlation, but it fits.
There’s a big problem with that, however. Cohen’s research didn’t say the effect of a mask order would take weeks. They promised the effects would be immediate and large.
In fact, the research Cohen used held that the effects of a mask order were so immediate, it was better to begin measuring them on the day a mask order was announced, not on the day it actually took effect. (See a discussion of the research here.)
So a continued, greater upward slope for weeks after the mask order was implemented (or announced) is not what we were promised in “the science and data” provided by the Cooper administration. Cohen had, with presumed “scientific merit,” shifted the goalposts for the order several weeks after the fact.
But there still is that tapering off and decline. Let’s look at it again:
I posted the wrong graph, both times. That’s the graph for testing, not cases. Seems I left off the title the first time.
Here’s the graph on cases, also marked with when the mask order took effect. Remember, Cooper, Cohen, Cohen’s research, Politifact, and Politifact’s experts all focused on COVID-19 cases.
You can clearly see how I got those two graphs confused. They’re very similar. In fact, I don’t have to change anything in the paragraph above where I described a continued increase for weeks or the slope of increase getting steeper, or an eventual tapering and falling off.
These two graphs track remarkably close. There’s a reason for that. And if you’re looking for correlation, start there.
Here’s a reminder:
It’s important to know that what media report as “cases” are not the same as infections. An infection is any time someone has the virus, whether that person feels sick or not. A case is when an infection has been confirmed by testing.
Back on July 31, I explained that the “spike” in cases that had earned NC bad press in Newsweek was a spike in testing bringing about “a proportionate spike in known cases.”
Why proportionate? Because the proportion of tests returning positive has been relatively stable for months. For that reason, a change in the amount of testing brings about a proportionate change in the amount of cases. More testing, “spike in cases”; less testing, “masks are working.”
So no, in no way is there proof of a “direct correlation” between the mask order and falling cases several weeks later. If you wanted to be really academic about testing for a direct correlation with such significant impact as Cohen claimed, you’d have to (off the top of my head and among other things): find significant evidence that the progress of the virus is impacted by policy interventions rather than taking an independent, natural course here; quantify and take out the effects of the other policy interventions; isolate and quantify the effect of the mask order; quantify and take out the effects of the people who would have voluntarily donned a mask (no claiming credit for “forcing” people to do what they were already doing); and quantify and tabulate the various levels of effectiveness of the masks selected by involuntary mask-wearers.
Cooper and Cohen made stabilizing the trends the goal before we could enjoy lifting of any restrictions. Then they wanted to take credit for the trends stabilizing, but not lift any restrictions.
Cooper said it was “encouraging to see our numbers are stabilizing.” Then he changed what that meant for North Carolinians: “In order to start a downward trend, we have to double down on actions that slow the spread of the virus.” He later added, “Stable is good, but decreasing is better.”
What? We achieved the goal, so we get more punishment?
Cohen put it this way: “Seeing glimmers of potential progress does not mean we can let up. It means it’s time to double down while we’re stabilizing.”
Clearly, Cooper and Cohen moved the goalposts again. They told us stabilizing our trends would mean lifting restrictions. Then they said our trends had stabilized, but they redefined it to mean “it’s the time to double down” on restrictions.
It’s a bad sign of things to come.