Does Stable Employment after Prison Reduce Recidivism Irrespective of Prior Employment and Offending?

Abstract

Sampson and Laub’s (1990, 1993) life course theory posits that stable employment can rehabilitate a criminal trajectory, irrespective of criminal history. Others contend that post-prison employment could be attributable to pre-prison patterns. These hypotheses have not been fully evaluated. Drawing on a random sample of 1,607 restored citizens in Ohio, this paper analyzes matched administrative and unemployment insurance (UI) data across three-year pre- and post-prison periods and provides new evidence that employment stability reduces recidivism irrespective of pre-prison employment stability or extensive criminal history. Individuals we label “employment gainers” lacked stable employment before prison, comprise 41% of the stably employed after prison, with recidivism levels indistinguishable from those with stable employment in both periods. Reductions in recidivism associated with employment stability are consistent across levels of criminal history. Our results thus carve a clearer vision of the possibilities for redemption among those with marginal employment histories and extensive criminal histories.

Keywords:

• Recidivism
• employment
• reentry

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 An examination of competing measures of employment stability is beyond the scope of this paper. We acknowledge that job stability is an important measure. Unfortunately, data limitations preclude us from measuring it directly.

2 There have been policy changes regarding correctional practices surrounding work since the time of data collection, and the economy continues to shift towards precarious forms of employment. However, it is important to recognize that we are not evaluating specific policies or the changing economy in this paper. Rather, the key question is whether there is a reason to believe that the relationship between work and crime has changed today relative to the first decade of the 21st century. Our data capture the period of strong economic growth during the runup to the Great Recession. We believe that the long list of barriers to employment faced by returning citizens during our study period, and the kinds of jobs they obtained, were largely similar to today.

3 According to NIJ’s website, “recidivism is measured by criminal acts that resulted in rearrest, reconviction or return to prison with or without a new sentence during a three-year period following the person's release (https://nij.ojp.gov/topics/corrections/recidivism).” However, we argue that the best approach to measure recidivism is to focus on events or outcomes that reflect serious, felony offenses. We therefore prefer to employ a more restrictive measure of recidivism, because many arrests are for non-serious offenses, and some of those do not result in a new conviction. Using the same logic, we also exclude technical returns to prison, because they could result from potentially minor violations, and because they are more sensitive to policy changes. By restricting our focus to return to prison for a new sentence, we are focusing on events reflecting commission of a new felony crime, and obviously an arrest and conviction.

4 As noted above, our data captures legal, on-the-books employment. Our analysis therefore does not capture off-the-books employment, which tends to be temporary and unstable (Nightingale and Wandner 2011). As well, we considered including quarterly earnings but opted against it. Hourly wages, a measure of employment quality, cannot be determined from UI earnings data because the number of hours worked per quarter is not collected by ODJFS (i.e., it is not part of their formula for determining unemployment insurance benefits).

5 We considered other possible approaches, such as within-individual longitudinal models and change score models. However, neither approach clearly captures employment stability. We also experimented with imposing thresholds– a minimum number of quarters in each three-year period, above which we would consider an individual to be stably employed. It was difficult to arrive at a convincing rationale for any particular threshold. Higher values for the threshold were arguably more meaningful in terms of capturing stability, but identified very few individuals as being stably employed. On the other hand, lower values for the threshold identified more individuals as stably employed, but were potentially less meaningful for measuring stability. Given the absence of theoretical guidance with respect to a particular definition of stable employment, we elected to use GBTM to identify clusters of individuals following similar trajectories over time (Jones & Nagin, 2013).

6 The criminal history score was validated on independently drawn samples of formerly incarcerated individuals who were released to parole supervision. Validity was assessed by cross-classifying recidivism rates with level of risk, defined as low, medium, or high. Support was indicated by recidivism rates that increased sharply from the low to high-risk categories. Also, recidivism was cross classified with each indicator that comprises the six-item index. The results of that analysis confirm that recidivism increases by level of risk. The criminal history score has since been validated affirmatively on probationers using the same strategy as described above.

7 We combined all non-whites into the same group because when we further break that group out into black and other, which are the two categories that are combined to form nonwhite, there are very small numbers in the “other” group (less than 10). This decision does not impact any of the other estimates in the paper.

8 It is generally suggested that the best fitting model has the largest (or least negative) BIC score (Nagin, 2005). However, Nagin (2005, p. 74–75) emphasizes the importance of considering BIC scores in conjunction with substantive and theoretical considerations.

9 Importantly, when we derive the employment gainers and stably employed groups from the three- and four-group models, the same empirical pattern emerges (analysis not shown).

10 We note that this finding is consistent with results reported by Kolbeck et. al., (2022). Using the same data set but a very different unit of analysis, and an analytic strategy that includes multiple failures over a much longer period, they find the effect of employment after prison is contingent on having a recent work history, which they defined as being employed in the street interval prior to that prison spell. Most employment gainers studied here (181 of 215) did have a recent work history, although they did not have stable employment before prison.

Additional information

Funding

This work was supported by the National Science Foundation.