![Effect of stay-at-home orders. Note: Figure plots estimated treatment effects ωk\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\omega _{k}$$\end{document} of stay-at-home orders on different outcomes, using the event-study specification at the CSA-day level outlined in Eq. (1) for the mobility and economic outcomes and Eq. (14) for the health outcomes. Panel A shows the effect on mobility, using the log of total POI visits in the SafeGraph data. Panel B shows the effect on consumer spending, using the log of total spending in Facteus’ debit card sample. Panel C shows the effect on employment, using the log number of individuals with positive work hours from the Homebase sample. Panel D examines log new cases as the outcome and sets γ=0\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma =0$$\end{document} which implies log(Iit)=log(Cit)\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\log (I_{it})=\log (C_{it})$$\end{document}. Panel E is the same as Panel D, except that it sets γ=1/6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma =1/6$$\end{document}. All regressions include date fixed effects δt\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\delta _{t}$$\end{document}. Panels A-C include CSA fixed effects μi\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mu _{i}$$\end{document}; Panels D–E include the event window indicator ξit\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\xi _{it}$$\end{document}. CSAs are weighted by population in the regression. Standard errors are clustered at the CSA level irrespective of order timing](https://www.researchgate.net/publication/387175728/figure/fig1/AS:11431281298804689@1734576930431/Effect-of-stay-at-home-orders-Note-Figure-plots-estimated-treatment-effects_Q320.jpg)
![Temporal variation explained by policy. Note: Figure plots the share of the total change in each outcome that is attributable to a given policy PolicyΔ/TotalΔ\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text {Policy}\Delta /\text {Total}\Delta $$\end{document} following Sect. 5.1. Panel A reports estimates using stay-at-home orders and business closure orders for POI visits from SafeGraph, total wages from Homebase, and employment from Homebase. For each policy treatment, we restrict attention to the CSAs treated by the given policy. Panel B reports estimates of the policy-induced change in the contact rate βt\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\beta _{t}$$\end{document} from stay-at-home orders, varying the assumed basic reproduction number R0\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\mathscr {R}}_{0}$$\end{document}. In both panels, the bars depict 95% confidence intervals. See Table 5 for additional details on Panels A and B. Panel C plots observed and counterfactual cases following the methodology outlined in Sect. 5.3 for our sample of cumulative statistical areas (CSAs) using γ=1/6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\gamma =1/6$$\end{document}. Panel C reports the observed number of cases along with the estimated number of cases if a uniform stay-at-home order was implemented on March 17 with proportional effect on social distancing behaviors, a uniform stay-at-home order was implemented on March 17 that caused social distancing behaviors to fall to a fixed level of 35% of March 1 levels, and the estimated number of cases if all areas followed the same social distancing behavior as the San Francisco Bay Area (defined to be the counties in the San Francisco CSA that were first to implement a stay-at-home order)](https://www.researchgate.net/publication/387175728/figure/fig2/AS:11431281298794375@1734576930594/Temporal-variation-explained-by-policy-Note-Figure-plots-the-share-of-the-total-change_Q320.jpg)
![Determinants of geographic variation in log contact rates. Note: Figure plots the coefficients of regressing the CSA log contact rate fixed effect estimates θ^i\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat{\theta }_{i}$$\end{document} from Eq. (13) on CSA-level determinants. The θ^i\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\hat{\theta }_{i}$$\end{document} and all covariates have been standardized to have a mean 0 and a standard deviation of 1. Panel A plots the standardized coefficients and 95% confidence intervals from univariate regressions. Panel B repeats Panel A but the regressions also control for the log of population. Population weights are not used. Robust standard errors are used to compute the confidence intervals](https://www.researchgate.net/publication/387175728/figure/fig3/AS:11431281298757478@1734576930757/Determinants-of-geographic-variation-in-log-contact-rates-Note-Figure-plots-the_Q320.jpg)
December 2024
·
1 Read
·
5 Citations
Review of Economic Design
We provide new evidence on the drivers of the early US COVID-19 pandemic and develop a methodology that future researchers can use to similarly analyze the outbreaks of new diseases. We combine an epidemiological model of disease transmission with quasi-random variation arising from the timing of stay-at-home-orders to estimate the causal roles of policy interventions and voluntary social distancing. We then relate the residual variation in disease transmission rates to observable features of cities. We estimate significant impacts of policy and social distancing responses, but we show that the magnitude of policy effects was modest, and most social distancing was driven by voluntary responses. Moreover, we show that neither policy nor rates of voluntary social distancing explained a meaningful share of geographic variation. The most important predictors of which cities were hardest hit by the pandemic were exogenous characteristics such as population and density.
