Disparities in Reaching COVID-19 Vaccination Benchmarks: Projected Vaccination Rates by Race/Ethnicity as of July 4
Stanford University and KFF (Kaiser Family Foundation) researchers used current state-reported vaccination data by race/ethnicity to project vaccine coverage, by state and nationally, among people ages 12 and older for four racial/ethnic groups (White, Black, Hispanic, and Asian). Specifically, we used data on distribution of vaccines administered by race/ethnicity extracted from state reporting dashboards by KFF, total numbers of people who have received at least one dose from the Centers for Disease Control and Prevention, and total population data from the 2019 American Community Survey to estimate the share of people ages 12 years and older receiving one or more COVID-19 vaccination doses, by state and race/ethnicity, through June 7. We then estimated coverage rates through September 1 based on the average daily vaccination rate implied by the change in coverage between May 24 and June 7 for each racial/ethnic group, by state.
Data on vaccination coverage by race/ethnicity vary by state in terms of reporting groups and completeness. We applied the following data processing steps to produce comparable estimates. We assumed vaccinations reported as “unknown” race/ethnicity were distributed proportional to shares of vaccinations with known race/ethnicity in each state. Examining vaccinations reported as “other” race/ethnicity, we found that in most states, the shares attributed to “other” greatly exceeded population shares (implying coverage >100%). We therefore adjusted shares by assuming “other” were vaccinated proportional to eligible population, and proportionally redistributed remaining vaccinations among specified racial/ethnic groups. We adjusted shares to avoid double-counting in states that report shares by race separate from shares by ethnicity. For racial/ethnic groups not reported by specific states, we assumed these groups were vaccinated proportional to population size and scaled down shares of vaccines to reported groups accordingly. We capped coverage among any racial/ethnic group at 100% of the eligible population, and in cases where implied coverage exceeded 100%, we proportionally redistributed the excess across other groups.
A handful of states required exceptions to the standard approach. The share of vaccinations by race/ethnicity from Nebraska was unavailable on June 7. As a result, projections for Nebraska were based on race/ethnicity-specific vaccination rates spanning May 10 to May 24. The share of vaccinations by race/ethnicity from Idaho and Tennessee were unavailable from May 24. As a result, projections for these states were based on the three-week period spanning May 17 to June 7. CDC reported coverage in New Hampshire decreased slightly between May 24 and June 7, likely due to reconciling reporting issues. As a result, projections for New Hampshire were also based on the three-week period spanning May 17 to June 7. The share of vaccinations by race/ethnicity for Pennsylvania reported in the state dashboard do not include vaccinations for Philadelphia County. Since Philadelphia County includes a substantial fraction of the Black, Hispanic, and Asian population living in Pennsylvania, we separately extracted and included data from the Philadelphia County dashboard.
Limitations of this analysis include reliance on several assumptions to address incomplete and heterogeneous reporting of vaccination data by race/ethnicity across states. Previous reporting on racial/ethnic disparities in vaccination through the CDC and other sources has not adjusted for these data discrepancies, resulting in reported coverage levels that likely underestimate actual population coverage. Although we have adopted a standard set of definitions and rules for reconciling unknown or discrepant data elements to enable transparent and comparable estimation of coverage over time and place, results must be interpreted as approximations in the context of missing and sometimes noisy data. Future work should continue to update these estimates and further assess uncertainty due to model assumptions.
Replication code and data are available at: https://github.com/PPML/covid_vaccination_coverage_disparities