By Michael Pearson
One of the sharpest controversies of the COVID-19 pandemic has centered on when it is safe to reopen state economies following outbreak-induced shutdowns. Scott Ganz, an assistant professor in the School of Public Policy, has proposed a solution to help answer that question in a working paper for the American Enterprise Institute, where he is on leave from the Georgia Institute of Technology as a research fellow.
In the weeks prior to most U.S. states beginning to reopen their economics, public health officials frequently cited the need for 14 days of declining COVID-19 cases as one benchmark for deciding when to ease restrictions introduced to slow the spread of the disease. But determining whether that threshold had been met proved difficult in many states.
“One reason the ‘sustained reduction’ criterion has so far proven difficult to interpret is the simple fact that the daily data on COVID-19 can be noisy,” Ganz wrote in his paper. “An isolated outbreak at a nursing home, a change in weather, or simple sampling variability can result in spikes or valleys in documented cases.”
Such “noise” in the data can result in a premature decision to open or a decision to stay shut down for too long, Ganz said.
Instead of simply waiting for 14 consecutive days of declining case counts, Ganz proposes that officials use a series of statistical tests to evaluate models based on theories about trends in coronavirus infections in their states over the previous 14 days. Models that significantly diverge from the data are eliminated through a series of tests until only models that largely mirror the data remain. Reopening would be supported by models that accurately “predicted” past trends and display what Ganz calls a “monotonic decrease,” or a consistent downward stepping pattern. These models form what is called a “model confidence set” (MCS).
“Policymakers basing their reopening decisions on the 14-day sustained reduction criterion can use this method to determine which hypothetical models of recent trends are supported by the data. For example, it might signify that public health indicators are following an inverse U-shape (IUS) curve (indicating that the peak in cases occurred in the prior two weeks), or that the indicators are constant (which would be consistent with a prolonged plateau).
“If, instead, the monotonic decreasing model is included in the MCS, and all other models that are inconsistent with a monotonic decrease have been culled, then policymakers should feel confident that the “sustained reduction” criterion is satisfied.”
While most states have already begun the reopening process, Ganz’ framework could be put to use in the future. Many public health experts expect another wave of COVID-19 cases as economies reopen, or during cooler fall and winter weather when respiratory infections tend to be more common.
Ganz notes that the statistical test he proposes should not be the sole benchmark used by policymakers. Recommendations by public health officials and the White House Coronavirus Task Force for robust testing, contact tracing, and effective treatment options are also important, he notes.
Ganz also warns that states that have not seen significant spikes in COVID-19 cases, have recently substantially increased testing capacity, or which have “noisier measurement of public health indicators” may find the results his methodology inconclusive or misleading.
Ganz’ paper, “Does Your State Have 14-days of Declining COVID-19 Cases? Finding straightforward answers to a surprisingly tricky question,” was published May 11, 2020, on the American Enterprise Institute website.
The School of Public Policy is a unit of the Ivan Allen College of Liberal Arts.
For more coverage of Georgia Tech’s response to the coronavirus pandemic, please visit our Responding to COVID-19 page.