Extended Data Table 2 Primary survey outcome estimates
From: Knowledge about others reduces one’s own sense of anonymity
- Estimates of treatment effects on residents’ perceptions of how likely it was that officers would find out if they did something illegal and how well officers knew them in general. Estimates presented as the percent change in resident survey response attributed to the intervention. Our primary specification presents results when covariates are selected using the double LASSO procedure. “Alternative Covariate Set 1” includes a large set of development-level covariates (crimes per resident one, two, and three years prior to the intervention, total population, % Black, % Hispanic, % working families, and randomization pair indicators). “Alternative Covariate Set 2” includes a small set of development-level covariates (crimes per resident in the year prior to the intervention, total population, % Black, and % Hispanic). “Keep Matched Control Devs” leaves the control developments paired with the four un-surveyed developments in the dataset. “Drop Lowest Control Devs” instead drops the four control developments with the lowest average response to the survey question, among all control developments in the study. P-values are from two-tailed tests.
