Skip site navigation

Last week, for the first time since the pandemic began, the number of tests fell week over week. This week, it happened again. This week, in fact, every major metric we track fell, including a slight drop in weekly deaths, the first time deaths have dropped since the end of June.

The falls in cases, hospitalizations, and deaths should be good news, and in reality, they likely do reflect an improving reality on the ground. But tests are also dropping, nearly as much as cases are. In an ideal world where the Sunbelt surge is genuinely easing, we’d see increased testing combined with falling case counts. That happened in state after state in the Northeast as the spring turned to summer. 

In the South, that’s not what’s happening. Tests have precipitously declined over the last two weeks. While complicated situations in Florida and Texas have led the way in that drop, just about every state in the region shows a decline in testing in the first half of the month. 

The different regions differ widely in population size, and the South is by far the most populous. When we look at the per million testing rates, you can see that the Northeast is now testing at higher rates than the South, despite vastly fewer cases. In fact, every region is. 

Add it up to the national statistics and these dynamics produce a chart where tests and cases are uncomfortably correlated in recent weeks.

Data disruption

The data that we compile from states can look complete and polished in the charts and data tables that we produce. But as we’ve noted time and time again, our data is a patchwork of data put out by 56 jurisdictions and multiple entities within each. The Federal government influences the data states report through guidelines and control of reporting systems, but the states themselves have not received guidance on what they should publish themselves. That is to say, we are acutely aware of the data production processes that must all work to bring reliable numbers to our users. 

And some weeks, these data systems break down, marring individual state numbers and skewing the national statistics. Just this week, we’ve seen major data problems from California, Florida, North Carolina, Texas, and Washington

Some of these issues are easy to understand. Florida’s Department of Health is dealing with both storm-related backlogs and an individual lab that had not reported 4,000 cases in the Miami area. Other times, the technical issues are difficult, but the system at fault is clear; in California and North Carolina, specific technical systems broke down, leading California to add missed tests and North Carolina to pull out 220,000 misfiled results

And then there is whatever is happening in Texas.

It is not at all clear why the state would suddenly do a fraction of the tests it completed a few weeks ago. And the most likely answer is that the state is not doing fewer tests, but that something strange has happened within the reporting system. Some Texas data sleuths have pointed to a category of test that had not been assigned to a particular county. As these tests have gotten a county designation, the number of tests the state reports having completed overall is falling. Why? That’s what nobody can figure out. Is it counties finding duplicate tests? Is it antigen tests being shaken out of the state data? Or something more mysterious? The state has launched an investigation, and we look forward to the publication of its results. Our researchers are also looking into the mystery, as are journalists throughout Texas. 

In the meantime, we continue to compile the data and to annotate the problems we find as part of our ongoing mission to contextualize all these deceptively tidy-looking data points in a way that brings them closer to the reality on the ground.

To keep up to date on our work, follow us on Twitter and join our low-frequency email list.


Alexis C. Madrigal is a staff writer at The Atlantic, a co-founder of the COVID Tracking Project, and the author of Powering the Dream: The History and Promise of Green Technology.


Peter Walker is Head of Marketing & Growth at PublicRelay and Data Viz Co-Lead at The COVID Tracking Project.


More Weekly Updates

Vaccines Begin to Arrive as Cases and Deaths Keep Rising: This Week in Long-Term Care COVID-19 Data, Dec 16

Cases are up and known deaths in long-term-care facilities are the highest they’ve been since late May.

Our Final Week: This Week in COVID-19 Data, Mar 4

Tests are up, while cases, hospitalizations, and deaths continue their declines. We are at a crucial moment in the pandemic, with vaccinations ramping up but multiple variants of SARS-CoV-2 gaining footholds across the US. In our final weekly report, we urge continued vigilance in reducing the spread of the virus, and direct readers on how to follow the course of the pandemic without us.

Good News Despite Data Wobbles: This Week in COVID-19 Data, Feb 25

Cases, hospitalizations, and deaths are still declining, though holiday reporting and winter storms have probably caused fluctuations in several metrics. We reiterate that deaths reported each day don’t represent people who died that day—and they may even include deaths that occurred several months ago. And now is the time to switch over to federal data sources, because The COVID Tracking Project has only a little over a week of data compilation left.