When in doubt: the value of uncertainty for release success among incarcerated women

Abstract

Reentry can be complicated and daunting, yet people leaving prison tend to be overly optimistic about their capability to remain crime-free. This optimism could be good—where a hopeful mindset could be indicative of a prosocial future narrative. Or this optimism could be bad—where a naïve mindset could be indicative of a lack of preparation for the challenges ahead. Our goal is to explore the concept of perceived uncertainty for reentry success with a focus on how it may be useful in better preparing people to rejoin society. Using data from 200 women incarcerated in Arizona, we explore (1) the demographic characteristics that are associated with perceived uncertainty for reentry success and (2) the extent to which perceived uncertainty is associated with the identification of specific obstacles upon their release. Our findings suggest that most women are optimistic about their capability to stay out of prison, but that age, education, and custody level are associated with perceived uncertainty in this capability. Women with uncertainty are more likely to identify employment as a barrier that could bring them back to prison. We suggest that a focus on perceptions of uncertainty is critical for research and programming on prison reentry.

Keywords:

• reentry
• correctional programming
• perceptions of success
• rehabilitation
• female offenders

Notes

1 The wide range indicates that respondents were not required to report difficulty ratings for all 13 goals, as other work out of the study highlights that 75% or more of respondents were optimistic across release goals (see Visher et al., 2003, for example)

2 Women tend to recidivate at a lower rate than men, but recidivism rates among women are still substantial, and women have unique pathways to recidivism (Huebner et al., 2010).

3 The study protocols were approved by ASU’s Institutional Review Board (STUDY00008660).

4 The Arizona Department of Corrections, Rehabilitation, and Reentry provides information on the end of fiscal year combined population (Arizona Department of Corrections, Rehabilitation, and Reentry, 2019a). The average age of females in ADCRR in 2019 was 37.2, compared to 37.4 in our sample. The average length of stay in ADCRR for females was 61.2 months, compared to 60.0 months in our sample. Fifty-nine percent of women in ADCRR were serving their first term in an Arizona prison, compared to 65% in our sample. This difference could largely be due to asking about any prison term, rather than Arizona-specific. The level of education is higher in the current study compared to national estimates, with 25% of women in the current study having less than a high school diploma compared to 39.1% nationally (Harlow, 2003).

5 We also measured uncertainty as a categorical variable containing three groups, with cut-off points for group inclusion determined by median and standard deviation. The first group, ‘certainty,’ represents respondents who said they were 100% prepared to stay out of prison after release. This is the median response of the original variable and captures 72% of the sample (n = 144). The second group, ‘moderate uncertainty,’ represents women who were at the mean or within one standard deviation of the mean. This group represents respondents who said they were between 73% and 99% prepared to stay out of prison after release (15.5% of sample, n = 31). The third group, ‘high uncertainty,’ represents women who are within two or more standard deviations below the mean for preparedness for release. This group represents respondents who said they were between 0% and 73% prepared (12.5% of sample, n = 25). We estimated separate models with the three-category variable, which did not significantly alter the results presented here. Results are available upon request.

6 Race and ethnicity were originally recorded across seven dichotomous variables: White, Hispanic, Black, Asian, American Indian/Alaska Native, Native Hawaiian/Pacific Islander, and Other. Respondents could identify more than one race or ethnicity. Fifty-three percent of the sample identified as White, 32% identified as Hispanic, 10% identified as Black, 9% identified as American Indian or Alaskan Native, 7% identified as Other, 1% identified as Native Hawaiian or Pacific Islander, and 0% identified as Asian. Analyses were run using the white/non-white variable and using all racial/ethnic groups. Analyses produced similar results and white was not significantly more likely to be uncertain compared to any of the other racial/ethnic groups. For simplicity in reporting and interpretation, we use the dummy white/non-white variable.

7 The variable for months until release and medium security have a correlation of 0.41. Due to collinearity, supplemental analyses were run using each of the variables separately. These analyses produced two changes. When months to release was dropped from the multivariate analysis for research question one, times to prison was significant (1.323, p < 0.044). Further, we observed an improvement in model fit for the multivariate analyses for research question two on the release obstacle of employment. Medium security was dropped in the multivariate employment release obstacle model to improve model fit.

8 Pseudonyms are used in this section. Additional demographic information is not provided to protect anonymity.