Extended Data Table 1.

EC, Exposure and Friending Bias: Variation across Areas and Settings

Panel A shows how much of the group-level variation in EC, exposure and friending bias is between counties versus settings. The first column reports the adjusted R² from two separate regressions: one where mean economic connectedness (EC) in each group is regressed on county indicators (a group’s county is defined as the modal residential county of individuals assigned to that group), and another where EC is regressed on the six setting indicators. The unit of observation in these regressions is a single group such as a particular high school, neighbourhood, or recreational group. EC is defined as twice the average share of above-median-SES friends in the group among below-median-SES people within each group. Columns 2 and 3 replicate Column 1 with exposure and friending bias as the dependent variables. Exposure is defined as twice the share of above-median-SES individuals in the group. Friending bias is defined as the one minus the mean ratio of the share of above-median-SES friends to the share of above-median-SES peers in the group, averaging over below-median-SES individuals in the group (see Supplementary Information B.5). Panel B focuses on the variation in mean EC, exposure and friending bias at the ZIP-code level. The first row reports the standard deviation of EC, exposure and bias across ZIP codes. The second row shows the adjusted R²s from regressions of ZIP-code-level means of these three variables on county indicators. To construct ZIP-code-level means, we aggregate individual-level measures of EC, exposure and friending bias as described in the notes to Fig. 4 and Supplementary Information B.5. All statistics are weighted by the number of below-median-SES individuals in each group (Panel A) or ZIP code (Panel B).