Evaluating the effects of shelter-in-place policies during the COVID-19 pandemic

Abstract

We estimate the effects of shelter-in-place (SIP) orders during the first wave of the COVID-19 pandemic. We do not find detectable effects of these policies on disease spread or deaths. We find small but measurable effects on mobility that dissipate over time. And we find small, delayed effects on unemployment. We conduct additional analyses that separately assess the effects of expanding versus withdrawing SIP orders and test whether there are spillover effects in other states. Our results are consistent with prior studies showing that SIP orders have accounted for a relatively small share of the mobility trends and economic disruptions associated with the pandemic. We reanalyze two prior studies purporting to show that SIP orders caused large reductions in disease prevalence, and show that those results are not reliable. Our results do not imply that social distancing behavior by individuals, as distinct from SIP policy, is ineffective.

Keywords: COVID-19; disease spread; government policy; mobility; shelter-in-place policies.

Fig. 1.

Mobility by state over time. Day indicates days after February 24. Mobility is the proportional change relative to the beginning of the time series. The dashed lines indicate the start of a state SIP policy, and the dotted lines indicate the end. There are clear nationwide trends, and mobility was already declining in most states before a SIP policy.

Fig. 2.

Effects of SIP policies over time. The figure shows the estimated effects of having a SIP policy in place for 0 d to 14+ d in a row. Estimates that are statistically distinguishable from zero (P < 0.05) are red. We regressed the outcome of interest on state fixed effects, day fixed effects, lags of the dependent variable, our SIP policy variable, and 14 lags of the SIP policy variable. To estimate the effect of 2 d of SIP policies, for example, we add the coefficients associated with the SIP policy variable, the 1-d lag, and the 2-d lag, and we conduct an F test of the null hypothesis that this sum is equal to zero. Complete results for these regressions are shown in column 7 of SI Appendix, Tables S1–S4.