On Monday afternoon, President Trump indicated that while the mascot of the Republican Party is an elephant, his personal mascot is an ostrich.
It should be immediately obvious both that this is true and that it is a baffling thing to say. Sure, if you stop actually testing people for the virus, you’ll see fewer confirmed cases. But like that apocryphal ostrich, covering your eyes doesn’t make the bad things go away. A surefire way to decrease the number of reported murders in a city is to stop arresting people for murder. But that doesn’t solve the problem.
In fact, it would almost certainly make the problem worse — as would reducing the number of tests we conduct aimed at tracing the spread of the coronavirus. But we’ll come back to that.
Let’s first adjudicate the point Trump was trying to make more broadly, a point that Vice President Pence later reportedly recommended that state governors embrace as they fought a rhetorical battle against continuing efforts to contain the virus. The administration’s theory is that the recent increase in the number of cases in a number of states is largely, if not solely, a function of an increase in testing. Trump’s been beating this drum for a while, insisting both that the nation’s increased testing capacity is a sign of his administration’s success and that it is the reason that the pandemic still seems so problematic.
By now, it is true that the United States is conducting more tests than other countries, though it wasn’t always. Back in early March, when the country was conducting relatively few tests and the virus was spreading undetected, Trump’s frustration about the steadily climbing number of cases couldn’t be blamed on test counts — so Trump tried to tamp the numbers down in more obvious ways.
Any current analysis of how the virus is spreading in the United States is hampered by the haze of uncertainty that surrounds the data. There are a variety of reasons for that murkiness, including uneven reporting by states, delays in testing relative to infection and dependence on the starting point for any comparison. So let’s look at new cases relative to May 25, Memorial Day, using state data compiled by the New York Times.
The biggest increases have been in Hawaii and Alaska, a reminder that relative change is generally more dramatic in places with lower populations because adding 100 cases to 100 existing cases yields a bigger rate of change than adding 1,000 cases to 100,000 existing cases.
It’s hard to tell from this graph how many states have seen increases (25, as it turns out) and how that relates to testing. In fact, it’s hard in general to compare new cases with increases in tests because of additional layers of murkiness. Who’s being tested? What’s being tested? Those sorts of things.
So let’s do a direct comparison of the change in conducted tests — from the Covid Tracking Project — and the change in reported cases. (To be consistent with the data in other charts, we’re using the change in the seven-day average of new cases and tests conducted per day on May 25 and June 14. That seven-day average allows us to smooth out some of the jerkiness otherwise present in the state data.)
There are two things in particular here that we want to pay attention to. The first is the shaded area, representing states in which testing has dropped (to the left of the vertical line) and the number of cases has increased (above the horizontal line). Since May 25, seven states fall into that range. The other interesting factor is that diagonal line at right. States above that line have seen more rapid increases in new cases than daily tests. Fifteen states fall into that group.
This suggests that, in many places — especially those seven states at upper left — the link between new cases is independent of additional testing. But, again, it’s hard to know how to treat those other 15 states where new cases are growing faster. Are they just testing more at-risk populations? It’s hard to say.
There are other metrics to consider, though. For example, the rate of positive tests. In Washington and Mississippi, for example, a higher percentage of tests are coming back positive than on May 25. In Maine, the rate of positives has fallen.
This rate is obviously independent of how many tests are being conducted, though it may still be subject to differences in who’s being tested. (If a state suddenly constricts its testing to patients exhibiting symptoms in hospital emergency rooms, for example, its positive rate will probably spike.) Nonetheless, it’s worth noting that in 21 states, the rate of positive tests has increased since May 25.
(If you don’t like using May 25 as the baseline: Fair enough! On Monday, we made a tool allowing you to use whatever range of dates you wish.)
There’s another useful metric that has driven a lot of the concern about new cases: increases in new hospitalizations per day. This data is a bit spottier because not every state reports it. But here we see, for example, that Hawaii’s increase in new cases isn’t only manifested in increased testing. It’s also putting more people into hospitals.
Of the 33 states for which there is hospitalization data compiled by the Covid Tracking Project, 13 have seen increases in the number of new hospitalizations per day since May 25. That includes five of the seven states where testing has dropped as new cases have increased.
These data yield a spotty measure of how states are doing. If we simply indicate how states are faring on the four metrics (cases, tests, rate of positive tests and hospitalizations), we get something like this, where darker blue means doing better on the metric and darker orange doing worse. New York is doing well! Georgia, not so much.
Again, though, the point of testing isn’t simply to track new cases. It’s to better allow states and the country overall to contain outbreaks. We test a lot because that has helped states to know where the virus is spreading and to quarantine those who might have been exposed, preventing new infections.
When the administration began touting how many tests were being done, comparing the numbers to those in South Korea, which was seen as the gold standard for testing, we pointed out that the value of South Korea’s testing regimen was that it geared up quickly — and did a much better job of preventing outbreaks of the virus.
This chart, comparing the number of tests conducted in each country with the number of cases, shows how South Korea’s quick increase in tests led to a quick reduction in cases.
Yes, more testing will necessarily yield more cases, but then it will ideally lead to fewer overall cases as the virus is contained. Trump for some reason didn’t learn his lesson in early March when he tried to present the number of U.S. cases as being smaller than it actually was. He again wants people to think that the number of cases is somehow inflated, that measuring the spread of the virus is itself making things worse.
It’s not new that Trump would be frustrated by objective analysis that makes his policies or his efforts look flawed. In this case, though, Trump’s argument at least makes some intuitive sense: Sure, more tests means more cases.
But, again, it’s still more cases. More deaths. And we still can’t be confident that we’re capturing every infection because we’re still not sure that we’re testing everyone who needs to be tested. As in early March, more testing will mean a quicker containment of the pandemic. As in early March, Trump would apparently rather suggest that everyone’s worrying unnecessarily and hope for the best.
We can at least hope that things turn out better moving forward than they did in the two months after early March.