Fig 1.
Posterior Random Effects Estimates: Risk Ratio Black, Armed-to-Unarmed.
(a) County-by-county posterior estimates of the risk ratio of being {black, armed, and shot by police} to being {black, unarmed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {black, armed, and shot by police} to being {black, unarmed, and shot by police}.
Fig 2.
Posterior Random Effects Estimates: Risk Ratio Hispanic, Armed-to-Unarmed.
(a) County-by-county posterior estimates of the risk ratio of being {hispanic, armed, and shot by police} to being {hispanic, unarmed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {hispanic, armed, and shot by police} to being {hispanic, unarmed, and shot by police}.
Fig 3.
Posterior Random Effects Estimates: Risk Ratio White, Armed-to-Unarmed.
(a) County-by-county posterior estimates of the risk ratio of being {white, armed, and shot by police} to being {white, unarmed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {white, armed, and shot by police} to being {white, unarmed, and shot by police}.
Fig 4.
Posterior Random Effects Estimates: Risk Ratio Black-and-Armed to White-and-Armed.
(a) County-by-county posterior estimates of the risk ratio of being {black, armed, and shot by police} to being {white, armed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {black, armed, and shot by police} to being {white, armed, and shot by police}.
Fig 5.
Posterior Random Effects Estimates: Risk Ratio Hispanic-and-Armed to White-and-Armed.
(a) County-by-county posterior estimates of the risk ratio of being {hispanic, armed, and shot by police} to being {white, armed, and shot by police}. Grey bars are county-specific estimates. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {hispanic, armed, and shot by police} to being {white, armed, and shot by police}.
Fig 6.
Posterior Random Effects Estimates: Risk Ratio Black-and-Unarmed to White-and-Unarmed.
(a) County-by-county posterior estimates of the risk ratio of being {black, unarmed, and shot by police} to being {white, unarmed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {black, unarmed, and shot by police} to being {white, unarmed, and shot by police}.
Fig 7.
Posterior Random Effects Estimates: Risk Ratio Hispanic-and-Unarmed to White-and-Unarmed.
(a) County-by-county posterior estimates of the risk ratio of being {hispanic, unarmed, and shot by police} to being {white, unarmed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {hispanic, unarmed, and shot by police} to being {white, unarmed, and shot by police}.
Fig 8.
Posterior Random Effects Estimates: Risk Ratio Black-and-Unarmed to White-and-Armed.
(a) County-by-county posterior estimates of the risk ratio of being {black, unarmed, and shot by police} to being {white, armed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {black, unarmed, and shot by police} to being {white, armed, and shot by police}.
Fig 9.
Posterior Random Effects Estimates: Risk Ratio Hispanic-and-Unarmed to White-and-Armed.
(a) County-by-county posterior estimates of the risk ratio of being {hispanic, unarmed, and shot by police} to being {white, armed, and shot by police}. Grey bars are county-specific 95% PCI estimates. The blue bar is the nation-wide pooled 95% PCI estimate. The points on the error bars are posterior medians. Data are plotted on the log scale, but are labeled on the natural scale. (b) Map of county-specific posterior median estimates of the risk ratio of being {hispanic, armed, and shot by police} to being {white, unarmed, and shot by police}.
Fig 10.
Data on Race-Specific Assault-Related Arrest Rates.
In these figures, only counties with greater than one arrest are plotted. (a) County-specific Department of Justice data on assault-related arrests (White), per 10,000 residents (2012). (b) County-specific Department of Justice data on assault-related arrests (Black), per 10,000 residents (2012).
Fig 11.
Data on Race-Specific Weapons-Related Arrest Rates.
In these figures, only counties with greater than one arrest are plotted. (a) County-specific Department of Justice data on weapons-related arrests (White), per 10,000 residents (2012). (b) County-specific Department of Justice data on weapons-related arrests (Black), per 10,000 residents (2012).
Fig 12.
AIDSVu Data on Income and Inequality.
(a) County-specific data on median income, in $1,000s. (b) County-specific data on inequality (Gini).
Fig 13.
(a) County-specific data on total population size, in 10,000s of residents. (b) County-specific data on the black-to-white population ratio.
Fig 14.
Data on Proxy Measure of Local Norms Concerning Racism.
(a) Designated Market Area-specific Google Search Racism Scores, 2004–2007.
Table 1.
Predictors of an increased county-level risk of being {black, unarmed, and shot by police} relative to being {white, unarmed, and shot by police}.
Values are: posterior mean (posterior standard deviation) of the regression coefficients. The symbol log referes to the natural logarithm. Pop refers to absolute population size. Pct. B. refers to the percentage of the county population that is black. Md. In. refers to median income. Gini refers to the Gini index of inequality. GRP refers to the Google search racism proxy. W. Ast and B. Ast refer to the white- and black-specific arrest rates for assualt, respectively. W. Wps and B. Wps refer to the white- and black-specific arrest rates for weapons violations, respectively. Posterior probabilty that a postive regression coeffcient is less than zero (or a negative one greater than zero) is coded as: * indicates a probability between 0.10 and 0.05, ** indicates a probability between 0.05 and 0.01, and *** indicates a probability of 0.01 or less.
Table 2.
Predictors of an increased county-level risk of being {black, unarmed, and shot by police} relative to being {white, armed, and shot by police}.
Values are: posterior mean (posterior standard deviation) of the regression coefficients. The symbol log referes to the natural logarithm. Pop refers to absolute population size. Pct. B. refers to the percentage of the county population that is black. Md. In. refers to median income. Gini refers to the Gini index of inequality. GRP refers to the Google search racism proxy. W. Ast and B. Ast refer to the white- and black-specific arrest rates for assualt, respectively. W. Wps and B. Wps refer to the white- and black-specific arrest rates for weapons violations, respectively. Posterior probabilty that a postive regression coeffcient is less than zero (or a negative one greater than zero) is coded as: * indicates a probability between 0.10 and 0.05, ** indicates a probability between 0.05 and 0.01, and *** indicates a probability of 0.01 or less.