Fig. 3.

OLS and double-lasso regression estimates of the relationship between multigroup Relative Diversity Index and COVID-19 death and infection rates. OLS models include no controls. Fixed effects models include state fixed effects. Double-lasso models include the 18 controls selected by the lasso procedure (shown in SI Appendix, Table S3) and state fixed effects. All regressions include population weights. Standard errors are clustered by state. Bars around estimated coefficients reflect 95% CIs. The multigroup Relative Diversity Index and covariates have been standardized to have mean 0 and SD 1. Outcomes are in log rates. The sample includes 2,174 counties. COVID-19 outcomes are measured as of September 30, 2020.