↵Notes: *p< 0.1, **p<0.05, ***p<0.01. Two-way cluster-robust standard errors within census tracts and individuals are reported in parentheses. Each coefficient reported is from a separate specification with the outcome variable indicated by the column heading and the key regressor of interest described in the row heading. All other control variables are as described in the General Estimation Note below. We start with incarcerated individuals who are tied to the same census block neighborhood. The second through sixth rows estimate the influence of peers with increasingly similar characteristics (indicated by the variable description in each row). Full model results for the primary regressor measuring neighborhood peers of similar age, race, and gender are reported in Online Appendix Table A.2. General Estimation Note: All regressions include indicator variables for gender, race, age (in years) at the time of exit, type of offense associated with the incarceration spell, number of months incarcerated, number of prior arrests, and number of prior incarceration spells. Neighborhoods are defined by census block delineations unless otherwise noted. When splitting the neighborhood peers by attribute type, specifications also include a variable measuring the number of other types of neighbors incarcerated (for example, the effect of “Nbhd peers age within 1 year, same race, same gender” is from a specification that includes a variable measuring all of the other neighborhood peers incarcerated). We also include year-by-month of release fixed effects, as well as neighborhood fixed effects (census block). Our estimation sample is defined in Table 1.