Robustness to Alternative Definitions of Neighborhood Peers Incarcerated
| Peers Incar. as a Ratio | Released First Half of Month | Released Second Half of Month | Peers Incar. at 3 Months | Peers Arrest within 1 Year | |
|---|---|---|---|---|---|
| (1) | (2) | (3) | (4) | (5) | |
| Nbhd peers incarcerated: age within 1 year, same race and gender | –0.046* (0.024) | –0.051 (0.034) | –0.039 (0.039) | –0.035*** (0.010) | –0.033** (0.013) |
| Mean of dependent var. | 0.220 | 0.233 | 0.203 | 0.220 | 0.220 |
| Mean of key independent var. | 0.045 | 0.044 | 0.045 | 0.058 | 0.065 |
| Observations | 17,361 | 9,612 | 7,749 | 17,361 | 17,361 |
↵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. This table presents specifications that assess robustness to potential measurement issues of our key regressor of interest: neighborhood peers incarcerated within one year of age, and of the same race and gender. The first column measures the key independent variable as the ratio of potential criminal peers incarcerated dividing the number incarcerated at the time of release by the total number of individuals previously incarcerated from the same neighborhood with the same demographic characteristics; the second and third columns estimate effects separately by the timing of release (first half of month in Column 2 and second half of month in Column 3) since our measure of peers incarcerated at the time of release only counts those incarcerated for the entire month of release. Column 4 estimates the impact of the number of peers incarcerated for the entire three-month recidivism window on the three-month recidivism outcome instead of our preferred measure capturing those incarcerated during the month of release; finally, Column 5 reports estimates from a specification that redefines peers to include only those who are arrested within one year of the individual in our estimation sample.