Ex-Offender Variables used by Doleac with additional 6 variables added with DePaul replication
After baseline data were collected, participants were assessed every 6 months for the next 24 months. There were 270 participants, with 90 randomly assigned to each of the three conditions (more details about the characteristics of the sample and the methods are in Jason et al., 2015). We then tried to replicate Doleac et al.’s (2020) difference-in-differences approach, as we used their method in the current study, and we selected a baseline variable that more validly measures criminal justice involvement. In the Doleac et al. (2020) article, the replicators had found that over the past 30 days at baseline, those in the recovery home had 2.54 fewer days incarcerated than controls (means of 1.2 versus 3.7 days incarcerated for the recovery home versus the controls). As mentioned in the introduction, we found conceptual and methodological problems with this baseline measure. In contrast, in the current study, we employed a more valid variable collected at the start of the study, which involved the length of time of the most recent incarceration, and on this measure, recovery home residents evidenced 76.80 more days incarcerated than controls (Table 1 in Jason et al., 2015 indicated means of 16.59 versus 14.03 months for most recent incarceration for recovery home versus controls-- months were calculated into days for the current study). As indicated in the introduction, the baseline variable was changed in this current study to length of time of the most recent incarceration, to reflect what we considered to be a more valid measure of baseline criminal justice involvement. In addition, we investigated two different outcome variables for the waves following baseline. We used the outcome variable employed by Doleac et al. (2018), days incarcerated over the past 30 days, for each of the 4 follow-up assessments that occurred every 6 months. But we also used another outcome measure, a variable involving total months incarcerated over the assessment periods (we recoded months into their respective days for this variable as well as the baseline months of most recent incarceration). This second outcome variable used the full period rather than the final 30-days between each 6-month wave. In doing so, we compiled a more robust method for assessing incarceration activity in between each wave by asking participants how long they have been detained since we last interviewed them. In a study that can be obtained from Leonard Jason, our group assessed how changing the baseline measure of criminal justice involvement alters the difference-in-differences findings reported by Doleac et al. (2020). As their findings focused on the treatment-on-the-treated (TOT) effects (those who participated in the recovery homes for at least 30 days), we also focus on these results that are in a paper that can be obtained from Leonard Jason.