Extended Data Fig. 3: Effect of smoke PM2.5 on different mobility measures, and heterogeneity by income.
a Percent of mobile phones estimated to be completely at home on a given day at the US county level, 2019–2020. Black lines are regression point estimates from spline fits conditional on fixed effects, with shaded areas showing bootstrapped 95% confidence intervals. Number of observations in each regression is shown in upper left corner of each panel. Histograms at the bottom show the log distribution of smoke PM2.5 exposure in each sample. b Same but for % fully away from home on that day. c-d Effect of smoke PM2.5 on mobility as a function of income. Lines show the marginal effect of a heavy smoke exposure (50ug smoke PM2.5 on that day) on percent of individuals completely at home on that day (c) or completely away from home (d), as a function of median household income in that county. Colors represent models run with either date fixed effects (blue) or state-by-date fixed effects (orange). Dark lines show regression point estimates, shaded area the bootstrapped 95% CI.
