Fig. 2: Larger, less segregated, and more diverse cities have lower implicit bias levels.
From: Implicit racial biases are lower in more populous more diverse and less segregated US cities

a Scaling relationship, diversity adjustment, and segregation adjustment for IAT data from 2020 in 149 cities with > 500 IAT responses per city. The shaded region is the 95% confidence interval for the scaling relationship. For visualization purposes, the segregation shown in this figure is estimated using only the mean deviation segregation measure. Results are similar with cutoffs of > 250 and > 1000 IAT responses per city and for other measures of segregation (Supplementary Tables 19–26). b Variance explained (R2) by segregation (measured via residential racial segregation), diversity, and scaling relationship. Data for n = 20 models are shown for 2016–2020. Medians are shown by a horizontal line and have values of 0.094, 0.097, 0.147, and 0.346, respectively. Variance explained by segregation is from all four models with different segregation measures. Noise ceiling estimates are obtained by computing correlations of bias levels between split halves of IAT participants within cities.