Reentry and the Ties that Bind: An Examination of Social Ties, Employment, and Recidivism

Abstract

Scholars consistently find that reentering offenders who obtain steady work and maintain social ties to family are less likely to recidivate. Some theorize that familial ties may operate through employment to influence recidivism and that such ties may also serve a moderating role. The current study employs an integrated conceptual framework in order to test hypotheses about the link between familial ties, post‐release employment, and recidivism. The findings suggest that family ties have implications for both recidivism and job attainment. In fact, the results suggest that good quality social ties may be particularly important for men with histories of frequent unemployment. The implications of these findings are discussed with regard to theory and future research on prisoner reentry and recidivism.

Keywords:

• prisoner reentry
• recidivism
• social ties

Notes

1. Using an electronic sorting process, staff from the state department of corrections (DOC) placed all parolees released in 2000 into a single electronic file and randomly selected every fifth parolee from the total parole file until a sample of 570 offenders was compiled (401 males; 169 females). We focus here on the 401 males in the sample. Random sampling was used to reduce the costs to the department of corrections associated with data collection. The sample is reflective of the population of parolees from the state in 2000. Further details on the study state are available from the authors by request.

2. The efficacy of the LSI‐R has been well documented (Gendreau, Little, & Goggin, 1996; Petersilia, 2003); however, it is important to note that there is an ongoing debate surrounding the utility of this instrument for women (see Holtfreter, Reisig, & Morash, 2004; Smith et al., 2009). DOC officials from the state in which the data originates conduct annual assessments to determine the internal reliability of LSI‐R; this is done using a test–retest method. The assessment conducted during the time frame in which these datasets were collected showed high test–retest reliability from 2001 through 2004.

3. Due to confidentiality reasons and bureaucratic oversight we were not allowed to access answers to the open‐ended questions from the LSI‐R interview.

4. Parole officials reported to the authors that some reentering offenders learned of job opportunities while they were incarcerated, often through family connections. And therefore some were able to make a transition immediately into employment upon release (for similar see Nelson et al., 1999).

5. The measures of ties to relatives and parental ties were originally scored on an ordinal scale (see Appendix 1). We recoded both the relative and parental ties variables into binary measures because both of the variables (i.e., ties to relatives and ties to parents) were sharply skewed. For the parental ties construct, five offenders received a score of (3), and nine offenders received a score of (0). In terms of familial ties, six offenders received a score of (3) and eight a score of (0). To generate reliable estimates unaffected by rare and extreme scores, we recoded the two family ties variables into dichotomous measures, with scores of 0 and 1 recoded as (0) and scores of 2 and 3 recoded as (1).

6. The LSI‐R scoring guide states that corrections officers should “Look for one solid relationship (if the offender has regular contact with 1 out of 5 siblings and that 1 sibling is a good, pro‐social support for the offender, then you may score that 1 positive person).” Also, corrections officers are urged to “look at the density of the relationship (with who does the offender have the most contact).” The parental ties measures are to be scored using a similar logic of that used to score the parental measures (see Note 4).

7. We re‐estimated all analyses with the two family ties variables (i.e., ties to parents and ties to relatives) in their original four‐score ordinal metric (0) to (3) and the results were virtually the same as those reported here (available upon request). However, it is important to note that we found that the overall model fit was more robust with the dichotomized measures relative to the model that employed the ordinal versions of the family ties measures. In fact, we found that the models showed a moderate level of improvement in structural fit with the implementation of the dichotomized measures. Based on this evidence, we have greater confidence in the reliability and validity of the estimates generated with the dichotomized measures versus those from the ordinal metric.

8. With regard to our measure of race, the category of non‐White includes 15 individuals who identified either as Hispanic, Asian, or Native American. Due to the small cell counts, it was not feasible to conduct analyses with separate measures created for individuals in these three groups. However, we conducted alternate analyses where the 15 non‐White members were recoded in the data file and included as White; the results under the new coding scheme were virtually identical to those reported here.

9. Information on the specific diagnoses from the psychological assessment was not made available to the researchers due to concerns surrounding subject confidentiality.

10. The LSI‐R also contains an item related to whether parolees maintain close ties with family members who are involved in crime. To examine its potential role in predicting recidivism, we added this measure to the models in Table 2 in a supplementary set of analyses. Slightly less than 10% of the sample indicated having such ties. Results from multivariate analysis showed that the measure had a non‐significant effect on the timing of recidivism (β = .148, p = .101).

11. We conducted supplementary analysis to probe the nature of the relationship between familial ties and recidivism, particularly the effect of parental ties. First, we re‐estimated Models 1 and 2 of Table 1, but analyzed separate models for each of the familial ties measures. We carried out this procedure primarily to determine if the effect of parental ties was diminished by the presence of ties to relatives and vice versa. The correlation between parental ties and ties to relatives was statistically significant (r = .45, p < .05), and potentially large enough to partially mask the effects of one another. Results of this exercise indicated that absent the measure of ties to relatives, parental ties did not have a significant effect on recidivism with or without employment in the model.

12. In theory, much like parents and other relatives in one’s social network, a spouse or intimate partner also perhaps represents a tie that facilitates job attainment (Lin, 2001). However, this idea is not fully articulated in theory, thus we largely focus on relations with the family of origin in this regard. Recall that Table 3 shows that offenders who are involved in an intimate relationship are more likely to be employed, which is consistent with theoretical expectations. We created product terms between intimate ties and frequent unemployment and between intimate ties and educational attainment in order to examine whether intimate partner ties moderate the effects of deficits in personal capital on employment. The two interaction terms failed to reach statistical significance, suggesting that intimate partners do not play a significant role in moderating the effect of educational attainment or work history on the acquisition of a job.

13. We would like to thank an anonymous reviewer for suggesting this potential explanation.

14. We examined the possibility that the findings in Table 2 with regard to family ties and employment may not be generalizeable across time points during the post‐release period. Common to reentry research of this type, we use static measures of post‐release conditions in our analysis, while recidivism is a time‐varying outcome. However, the nature of the post‐release context likely changes over time, which may limit the predictive efficacy of our fixed measures. Indeed, the effects noted in Table 2 may be time‐dependent. To gain some insight into this possibility, we re‐estimated the models in Table 2 but restricted the outcome to capture recidivism at 6‐, 12‐, and 18‐month time periods. In other words, we estimated Models 1 and 2 in Table 2 with the outcome measure censored at three different periods. Results from the supplementary analyses were very similar to those reported here, with the exception of the drug abuse history variable—it was positive and significant in the 6‐ and 12‐month models but not in the other set of models restricted at 18 months. These results are not shown owing to space concerns; however, they are available upon request from the authors.

15. In fact, Ericson (1977, p. 24) asserts that “employment is the indication of respectability” it signifies to the wider community that offenders have made an alignment with legitimate society.

16. We would like to thank an anonymous reviewer for prompting this idea.