Does a Rising Tide Lift All Boats? Labor Market Changes and Their Effects on the Recidivism of Released Prisoners

Abstract

The dramatic growth in incarceration nationally has increased attention to the factors that influence recidivism among ex-prisoners. Accordingly, scholars have called for research that identifies factors, such as employment opportunities, that may influence reentry experiences. Few studies, however, have examined how changes in labor market conditions affect ex-prisoner offending. Drawing on prior scholarship, this study examines the effect of such changes on the recidivism of ex-prisoners and, in particular, how the recidivism among blacks and whites may be differentially affected by changes in labor market conditions in the areas to which they return. The analyses indicate that, among black male ex-prisoners, labor market declines increase violent recidivism. They also indicate that, among white male ex-prisoners, the effects are more tenuous, influence only property recidivism, and are moderated by prior labor market conditions and criminal history. Implications of the study are discussed.

Keywords:

• unemployment
• recidivism
• prisoner reentry

Acknowledgments

The authors thank the Florida Department of Corrections for providing data for this study. Views expressed in this paper are those of the authors and do not necessarily reflect those of the Department of Corrections. We also thank Eric Stewart and Brian Stults for insightful suggestions during the development of this paper and the Editor and anonymous reviewers for their helpful comments and recommendations.

Notes

1. Sentencing points are assigned according to the Florida Criminal Punishment Code and are based on the primary (most serious) offense. Following Burton et al. (2004), we use the 1999 and 2000 sentencing guidelines offense points assigned to 52 different offenses.

2. Parker (2008, p. ix) has argued that resource deprivation and labor markets tap “different aspects of the local urban economy.” We found no statistically significant correlation between the race-specific measures of unemployment rates and resource deprivation (.022 for black males and .184 for white males) or between change in these rates and deprivation (.007 for black males and .050 for white males). In addition, all VIFs were below 4, and the largest VIF was 2.192.

3. To address concerns about spatial dependence, we used the nearest-neighbor criterion (Baller, Anselin, Messner, Deane, & Hawkins, 2001) and neighbor weight matrices for 5, 6, and 10 nearest neighbors (all weights equal “1”; larger counties were assigned larger weights), with proximity defined as the distance between county centroids. We computed global Moran’s I statistics using offense-specific reconviction rates for black male and white male ex-prisoners, respectively. Using S-plus’s spatial module, we undertook 1,000 permutations for each Moran’s I statistic and found no evidence of statistically significant spatial autocorrelation for white ex-prisoner recidivism outcomes. For black ex-prisoners, there was evidence of spatial autocorrelation for property recidivism and drug recidivism, respectively, and so we included the spatial lag specific to each offense. The lag was created by taking the average of the reconviction rates for the five nearest counties neighboring each county; few counties in Florida have more than five counties that adjoin them.

4. The negative effect of the 1990 unemployment rate on black male ex-prisoner drug recidivism in Table 2 is surprising. We do not discuss it here because the focus centers around changes in employment contexts, but speculate that, as suggested by Mears et al. (2008), black ex-prisoners may return to areas in which illegal drug activity is more entrenched and defended. In such a context, individuals may be more likely to engage in drug offending but less likely to be caught.

5. To assess whether the effects of changes in race-specific unemployment rates on violent recidivism were different for black vs. white male ex-prisoners, we performed a z test (Brame, Paternoster, Mazerolle, & Piquero, 1998). This test indicated that the effect of increases in black unemployment rates among black ex-prisoners (b = .05) was not significantly greater than the effect of increases in white unemployment rates among white ex-prisoners (b = −.00) (z = .98, p > .05). Per Bushway, Sweeten, and Wilson (2006), we compared 95% confidence intervals associated with these coefficients. The true coefficient of changes in black unemployment rates ranges from .03 to .07; the true coefficient of changes in white unemployment rates ranges between −.1 and .1. The comparison highlights that increases in black unemployment rates increase black violent recidivism, whereas increases in white employment may either increase or decrease white violent recidivism.

6. To conserve space, we report here the main and interactive effect estimates, less the controls, from the two-way interaction models, which used HGLM regression models similar to the ones in Table 2. For black property recidivism, the estimates were as follows: criminal history (b = .31, se = .03, p < .01), change in black male unemployment rate (b = .00, se = .01, p > .05), and criminal history x change in black unemployment rate (b = .02, se = .01, p < .05). For black drug recidivism, the estimates were: criminal history (b = .22, se = .03, p < .01), change in black male unemployment rate (b = −.02, se = .02, p > .05), and criminal history × change in black unemployment rate (b = −.02, se = .01, p < .01). For both models, the cross-level interaction between criminal history and the change in black male unemployment rate was included. Analyses involving three-way interactions, where the unemployment rate in 1990 was included, revealed no statistically significant three-way interactions. (All interaction model results are available upon request).

7. As per note 6, we report here only the main and interactive effect estimates to conserve space. For white property recidivism, the estimates of the main effect and interaction terms were as follows: criminal history (b = .43, se = .03, p < .01), change in white male unemployment rate (b = .06, se = .04, p > .05), and criminal history × change in white male unemployment rate (b = .05, se = .02, p < .05). No evidence of a three-way interaction, one that included the unemployment rate in 1990, surfaced. (All interaction model results are available upon request).

Additional information

Notes on contributors

Daniel P. Mears

Daniel P. Mears, PhD, is the Mark C. Stafford Professor of Criminology at Florida State University's College of Criminology and Criminal Justice. He conducts basic and applied research on a range of crime and justice topics, including juvenile justice, supermax prisons, homicide, and prisoner reentry. His work has appeared in the Journal of Research in Crime and Delinquency and Law and Society Review, among others, and in a recent book, American Criminal Justice Policy (Cambridge University Press 2010).

Xia Wang

Xia Wang, PhD, is an Assistant Professor at Arizona State University's School of Criminology and Criminal Justice. She is undertaking studies of reentry, sentencing, corporate crime, and the use of multilevel and spatial analyses to extend and test crime theories. Her work has appeared in Criminology, the Journal of Research in Crime and Delinquency, and the Journal of Quantitative Criminology.

William D. Bales

William D. Bales, Ph.D., is a Professor at Florida State University's College of Criminology and Criminal Justice. He focuses on a range of crime and policy topics, including factors that contribute to recidivism, the effectiveness of electronic monitoring, and tests of labeling theory. He has published in Criminology, the Journal of Research in Crime and Delinquency, Criminology and Public Policy, and other crime and policy journals.