major_missing_because_of_controls <- 
  filter(df, !complete.cases(select(df, all_of(controls)))) %>% 
  filter(pop_tot>1 & ! & country != "NA")

dropped <- major_missing_because_of_controls$country %>% 
  as.character %>% 

major_missing_because_of_controls_pc <- 
  filter(df, !complete.cases(select(df, all_of(controls_deaths_capita)))) %>%
  filter(pop_tot>1 & ! & country != "NA")

dropped_pc <- major_missing_because_of_controls_pc$country %>% 
  as.character %>% 

##  Get measures and labels

families <- c("state_cap_vars", "pol_account_vars", "social_vars",
              "econ_vars",  "phys_vars", "epi_vars", "health_sys_vars")

for(v in families){  
  assign(v, filter(measures, family == v)$vars %>% as.character) 
  assign(paste0(v, "_labels"), filter(measures, family == v)$labels %>% as.character)

Existing social science research has implications for the kinds of features of societies that might help explain variation in a society’s ability to respond to adverse shocks. In some accounts, for example, more democratic societies will be more responsive; in other accounts, state capacity or social cohesion are more important. We gather together major measures suggested by these literatures and report on the evolving covariance between these measures and the cross national distribution of reported Covid-19 burdens.

As of the most recent data date, 22 November, 2021:

Notes / warnings:


Deaths/million (logged)

Predicted and observed mortality

Each point in the figure shows the actual and predicted (log) per capita mortality for a given country. To predict mortality, we use four covariates selected by a Lasso procedure that maximize predictive accuracy as well an additional control for the quality of health data reporting. The controls are share 65+, respiratory disease prevalence, healthcare quality index (GHSI), healthcare spending/capita, and health data quality.