The surge in COVID-19 cases since June has led to much speculation that states began reopening too soon. But perhaps some states imposed stay-at-home (SAH) and business closure orders too soon, long before the virus truly reached their states. Premature shut downs maybe contributing to the ongoing spike in cases and have rendered a second shut down politically infeasible.
A premature shutdown could affect subsequent citizen behavior in at least two ways. The first is through learning about the virus. The shutdown of life across the U.S. (and the world) has been unprecedented, presumably portending a devastating pandemic. A state shut down followed by relatively few cases and deaths (perhaps in part due to the policies) may make people decide that the threat from SARS-CoV-2 was exaggerated, leading to additional risk-taking during reopening. Of course, numerous news accounts and statistics documented the extent of the pandemic, including 60,000 deaths by the end of April. The hypothesized learning effect would require that local events carry more weight with people than distant events. I will not argue this point further, just observe what the effect would require.
A premature shutdown could also reduce confidence in state leaders. The lock down policies imposed staggering economic costs and deprived residents of freedom. Elective surgery bans will have fatal consequences and in some places produced empty hospitals and furloughed nurses. Residents might decide that state officials lack the competence to manage COVID-19 and to ignore future edicts. Or businesses may ignore occupancy limits and make up for lost revenue from the premature shutdown and prior to a potential second shutdown.
If premature shutdowns are fueling current outbreaks, then states with a lower incidence of the virus at the time of shutdown should be experiencing worse outbreaks today. To identify premature shutdowns I will use an inferred infection rate on the day a state’s SAH order went into effect. Testing in March and early April was woefully limited, so official case counts then were largely meaningless. Deaths are probably measured far more accurately, although medical experts disagree over attributing deaths to COVID-19. Frequently described as a respiratory ailment, COVID-19 can cause heart and kidney damage, blood clots, and neurological problems. Plus, some deaths in March or April were almost surely not identified due to limited testing. I have nothing to add to these discussions, although it would be nice to have a uniform count of deaths across states and countries. Regardless of how these considerations turn out, deaths are still likely counted much more accurately than cases.
Deaths occur with a significant lag following infection, on average about 21 days. To use deaths as a proxy for infections when a SAH was issued, I take the state’s death toll from the COVID Tracker website 21 days after the SAH went into effect (the dates were taken from WalletHub). The infection fatality rate (IFR) then allows an inference on the number of infections based on the formula (1/IFR)*Deaths 21 days later. In June the CDC estimated the IFR at 0.27%. The CDC increased this estimate to 0.65% in July 10 planning guidance; but a review of seroprevalence studies by John Ioannidis reported a median IFR of 0.25%, so I will use the June IFR in calculating an inferred infection rate (IR), which I measure as a percentage of state population. I only use the IR to rank order the states, so applying the July CDC IFR would not change the states identified as issuing the most premature SAH order.
Deaths 21 days after the SAH ranged from 7 in Montana to 9,385 in New York, and the inferred IR’s ranging from 18.6% in New York and 9.8% in Michigan to 0.24% in Hawaii and 0.21% in West Virginia; the latter rates are one case for every 400 or 500 residents. For the eight states which did not issue SAH, I use deaths on April 25, 21 days after the last SAH was issued, to infer an IR. The states without SAH’s had inferred IR’s from 1.7% in Kentucky and 1.4% in Iowa to 0.47% in Wyoming and 0.44% in South Dakota. I will use 1% inferred IR’s as a threshold for potentially premature shutdowns. Twenty states have IR’s less than 1%, including six of the states that never issued a SAH order. The average IR in the SAH states was higher, 2.9% vs. 0.9%, and all states with inferred IR’s in excess of 2% issued SAH’s. This illustrates the endogeneity of policy problem which plagues policy evaluation.
I compare the fourteen states with potentially premature SAH orders with the six states that had inferred IR’s under 1%. The accompanying figures compare the averages across the groups of states for the increase cases over the past five weeks (6/17 to 7/22), the increase in deaths over the past three weeks (7/1 to 7/22), the state fatality rate to date, and the case fatality rate. Infections take about seven days to translate into official cases and 21 days to result in deaths as noted, so these windows reflect the progression of the outbreak in the states since June 10. Changes in cases and deaths are percentages of the totals on the initial date; the fatality rate is per 100,000 state residents; the case fatality rate is deaths on 7/22 as a percentage of reported cases on 7/8 (applying the 14 day difference in testing vs. fatalities).
In each case, the states that issued SAH’s with a very low inferred incidence of COVID-19 are suffering worse outbreaks now. The percentage change in cases in the SAH states is twice that of the no SAH states and the increase in deaths is 19% greater. Percentage changes can be biased downward when a state starts with substantially more cases or deaths, but these states all had similarly low incidences at the end of April. The measures on the outbreak to date are also better in the states that never shutdown. The fatality rate is 40% higher in the states that shut down, and the case fatality rate, which reflects any possible overwhelming of healthcare system capacity, is 76% higher in the states which shut down.
To be sure, all the states that issued SAH orders with low incidence of COVID-19 are not experiencing extreme increases in cases now. Arizona, California and Texas all qualify as premature shutdown states, but Minnesota is as well, where cases increased only 53% over the past five weeks. The evidence is reasonably strong though that the states that closed down most prematurely are doing worse now, relative to states which resisted the pressure to shut down at a similar incidence of COVID-19. Perhaps we need to change the narrative. Arizona and Texas didn’t reopen too early; they shut down far too early.
Daniel Sutter is Affiliated Senior Scholar at the Mercatus Center and Professor of Economics at the Manuel H. Johnson Center for Political Economy at Troy University.
