Mass Reentry, Neighborhood Context and Recidivism: Examining How the Distribution of Parolees Within and Across Neighborhoods Impacts Recidivism

Abstract

Recent scholarship focuses on the role neighborhood context plays in reoffending. These studies lack an examination of how the size of the parolee population at the neighborhood-level impacts individual recidivism. We examine how the size and clustering of parolee populations within and across neighborhoods impacts individual-level recidivism. Using data from parolees returning to three Ohio cities from 2000 to 2009, we examine how concentrations of parolees in neighborhoods and in the surrounding neighborhoods impact the likelihood of reoffending. We also examine whether parolee clustering conditions the relationship between neighborhood-level characteristics and recidivism. Results show concentrated reentry increases recidivism, while parolees in stable neighborhoods are less likely to recidivate. Also, the positive effect of parolee concentration is tempered when parolees return to stable neighborhoods. These findings suggest that augmenting resources available in neighborhoods saturated by parolees, as well as bolstering residential stability in these same neighborhoods might reduce reoffending.

Keywords:

• parolees
• recidivism
• mass incarceration
• neighborhoods
• reentry

Acknowledgements

The authors wish to thank the anonymous reviewers for their helpful comments, as well as John Hipp, Lyndsay Boggess and Sara Wakefield for comments on earlier drafts.

Notes

1 On 1 July 1996, the Ohio legislature enacted Senate Bill 2 (SB2), Ohio’s truth in sentencing law (La Vigne, Thompson, et al., 2003). SB2 also affected Ohio’s parole policies; following the implementation of SB2, parole was eliminated for offenses committed after 1 July 1996 and was replaced with a period of mandatory post-release control (PRC), which is often required for serious offenders. Judges impose PRC status at the time of sentencing, and might also impose the conditions of supervision, although this is generally determined by the Adult Parole Authority upon release. For ease of reference, all releases under post release supervision (parolee and PRC) will be referred to collectively as parolees. The mixing of parolees and PRC subjects did not result in significant differences in the models tested here.

2 Addresses were updated for each parolee annually between 2000 and 2003, while addresses were available every six-months beginning in 2004. In order to take advantage of the more detailed address data in later years, data reported on an annual basis were coded in six month intervals. Thus, parolees can have multiple addresses attached to them during their time under supervision capturing their movements while under supervision. Even though addresses are updated frequently, the address updates may not capture all the moves a parolee makes while supervised. While it would be preferable to have address data for each move a parolee makes, residing at a residence for a relatively short period of time reduces the likelihood that a parolee will have much effect on recidivism and overall parolee clustering. A weakness in all studies examining the movements of parolees is that it is extremely difficult to track where parolees are residing. However, there is some evidence that parolees are fairly stable with regard to their living situation; a recent study in Cleveland, Ohio found that approximately 72 percent of a sample of parolees lived at the same location 1–2 years later (La Vigne & Parthasarathy, 2005).

3 Some parolees were designated as homeless in ODRC records. These individuals were scrubbed from the data, since no address information was available. Nonetheless, it is possible that an address was entered for a transient parolee, though this is unlikely to affect a significant proportion of individuals in the data.

4 A limitation of this measure is that it only captures offenses for which a parolee was detected. This is a standard complication with recidivism data (Kubrin & Stewart, 2006); as such, we have conservative estimates of reoffending.

5 Unfortunately, data restrictions do not facilitate knowing the location of offenders released from prison but who are not under correctional supervision. The concern here is that there is a non-parolee effect on individual recidivism, however, our current model likely accounts for this possibility. Because research suggests that ex-offenders generally cluster in only a few neighborhoods (Kirk, 2009; La Vigne, Thompson, et al., 2003; Sampson & Loeffler, 2010), non-parolees are likely similar to parolees in that they are returning to the same neighborhoods; as such, our current measures of parolee concentration would serve as a proxy for both parolees and non-parolees. Therefore, we capture the non-parolee effect in the current analysis.

6 Given that this measure only captures individuals on parole during the study period, this is likely a conservative estimate of these effects. However, this measure likely captures a population most at risk for recidivating.

7 A value of 1 was added prior to the log transformation.

8 Kubrin and Stewart (2006) examined the effects of concentrated disadvantage and the ICE measure. We do not include a measure of ICE here since it captures concentrated advantage and is highly correlated with concentrated disadvantage. Kubrin and Stewart (2006) note that they were unable to estimate a model with both the ICE and concentrated disadvantage measures as there was a high degree of multicollinearity between the two.

9 We use a slightly different combination of variables for this measure than has been used in other studies. Hipp et al. (2010) used a combination of (1) average length of residence; (2) percent of households that moved into their units in the last five years; and (3) the percentage of units that are currently vacant. Our construct includes the percent of homeowners instead of the percent of vacant units. As we have argued, residential stability is likely an important factor in promoting informal social control, which may reduce the likelihood of reoffending among parolees. The residential stability of homeowners represents an important source of neighborhood stability and cohesion (Boggess & Hipp, 2010) that may be more consequential than the presence of vacant properties or renters and is therefore included in the model.

10 Ancillary models examined subsequent spells yielding largely similar results. Given the reduced sample size of chronic recidivists, the magnitude of these effects was smaller, but the pattern of results were consistent.

11 The three cities examined in this study are the largest cities in their respective counties, and therefore likely draw a large number of ex-offenders. Using county release statistics from ODRC, we cross-checked our data with official reports to verify that the estimates used in this analysis were relatively accurate. Our estimates are consistent with official patterns of release and provide a reliable indicator of reentry.

12 The risk of failure for parolees supervised at intense levels is β = 1.103 compared to parolees supervised at low levels. Since the coefficients displayed in Table 2 reflect the change in the log of the hazard, they can be transformed by exponentiating the coefficient to calculate the actual rate. Using the coefficient in Model 1, Table 2 for parolees supervised at intense levels, we see that e^(β) = e^(1.105) = 3.01.

13 In survival models, variables that are log-transformed are not exponentiated. Instead, a 1 percent increase in e^(x) results in a 1 percent increase in e^(y) (Cleves, Gould, & Gutierrez, 2004).

14 Given that this finding contradicts prior research, we tested additional models in which concentrated disadvantage was the only predictor of recidivism, still finding null effects. These differences could suggest location dependent differences in the role neighborhoods play in recidivism across various contexts, which implies the need to examine multiple cities with different structural characteristics to determine how robust the findings are across settings. Another possibility is that our annual measure of parolee concentration may capture temporal variations in disadvantage that our interpolated measure of concentrated disadvantage does not.

15 Unfortunately, since we cluster our data within census block groups, we are only able to obtain a pseudo r-squared value. Given the uncertainty of this measure and the difficulty in interpreting its meaning, we do not feel that this is a good indicator of the explanatory value of our models. Instead, we perform a Wald test to determine whether the overall fit of the model improved with the inclusion of our neighborhood-level variables.

16 Aggregations are based on the non-logged version of the variable, which has a mean of 3.14 and a standard deviation of 6.56.

17 In ancillary models, we find that the spatial concentration of parolees only has an impact on recidivism on the extreme end of the distribution (>95th percentile) of parolee concentration. Given the limited number of data points, we do not include this finding in the analysis here. Nonetheless, future research should explore the possibility that neighborhoods experiencing extreme levels of parolee concentration not only have a unique relationship to recidivism, but likely impact the surrounding neighborhoods.

Additional information

Funding

This work was Prepared from the Department of Housing and Urban Development, Office of University Partnerships [grant number H-216035G]. Points of views or opinions in this document are those of the author and do not necessarily represent the official position or policies of the Department of Housing and Urban Development. This research was also sponsored in part by the Newkirk Center for Science and Society, University of California, Irvine.

Notes on contributors

Alyssa W. Chamberlain

Alyssa W. Chamberlain is an assistant professor in the School of Criminology and Criminal Justice at Arizona State University. Her research interests focus on the nexus between neighborhood dynamics and crime, more specifically, the spatial and temporal relationship between neighborhood structural characteristics, social inequality and crime and how those factors shape neighborhood change over time. She also examines issues related to prisoner reentry and corrections.

Danielle Wallace

Danielle Wallace is currently an assistant professor at Arizona State University in the School of Criminology and Criminal Justice. She received her PhD in sociology from the University of Chicago in 2009. Her research interests include neighborhoods and crime, theories of disorder, and parolee reentry and recidivism. Her current work examines the relationship between health and re-entry outcomes.