Conditions of Contact: Reexamining the Relationship between Prison Visitation and Recidivism

Abstract

While in prison, incarcerated persons can be subjected to qualitatively different conditions of confinement ranging from minimum to maximum security settings. In this study, we use data on 17,542 incarcerated men to assess whether the relationship between visitation and recidivism varies across the types of settings (i.e. minimum, medium, close, and maximum custody units) in which individuals are housed and receive their visits. We find that the relationship between prison visitation and rearrest varies across conditions of confinement, and that this association is highly attenuated for individuals confined in close and maximum security units. We also find that these patterns are robust to alternate various measures of recidivism (reconviction and return to prison) and visitation (frequency and timing). These results raise questions about the nature and effectiveness of visitation in highly restrictive settings, and suggest that an ecological framework should be applied to future studies of prison visitation and recidivism.

Keywords:

• Prison visitation
• recidivism
• conditions of confinement

Disclosure statement

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

Data availability statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy restrictions.

Notes

1 Individuals serving time in prison for parole violations were not included in the sample.

2 All DOC policies described are specific to the time period in which the sample was incarcerated. Some policies have since changed.

3 Visitation did not capture meetings with attorneys or prison clergy, and the DOC did not allow conjugal visits.

4 Unlike in some state prison systems, individuals were not automatically reclassified or allowed to step down to a lower custody unit as they neared release from prison. Indeed, 80.6% of the sample who served most of their time in maximum security housing in the year prior to reentry were released directly from maximum custody units.

5 Less than 0.8% of individuals in the sample were incarcerated more than 550 miles from home.

6 Disciplinary histories were categorized as follows: violent infractions (e.g., homicide, assault on an officer or on another incarcerated person), property infractions (e.g., possession of a communication device, theft, possession of stolen property), drug infractions (e.g., possession of drug paraphernalia, drugs, or narcotics), threat/intimidation infractions (e.g., intimidating, harassment, stalking), security infractions (e.g., rioting, escape, promoting prison contraband), defiance infractions (e.g., disorderly conduct, resisting or disobeying a verbal or written order), and other non-violent infractions (e.g., conspiracy or attempt to commit an offense).

7 Prior to estimating multivariate models, various diagnostics were examined to rule out the presence of harmful collinearity. The highest variance inflation factor was 2.12 (average = 1.33) and bivariate correlations between independent variables did not exceed an absolute value of .495. The two strongest correlations were between violent discipline reports and other discipline reports. The key findings regarding the relationship between visitation and recidivism were not sensitive to the inclusion/exclusion of these variables.

8 As an alternative to clustered standard errors, we specified multilevel logistic regression models that nested individuals within housing units, within prison complexes. In the unconditional model, variance components indicated that the likelihood of rearrest varied across housing units (.223, p < .001) but not prison complexes (.022, n.s.). The intraclass correlation coefficient showed that 6.35% of the variation in rearrest existed between housing units, and 0.66% existed between prison complexes. Once all covariates were added to the model, the variance component at the housing unit level was reduced to non-significance (.0017, n.s.). Further, the regression coefficients in the fully specified model were nearly identical in terms of magnitude and statistical significance as those produced from logistic regression with clustered standard errors. Given the similarity in findings, and the non-significant variance components in the fully specified multilevel models, we proceed with presenting logistic regression models.

9 Models were also estimated with a measure that ranged up to 103 visits (the maximum number of visits received by anybody in the sample) and the results were the same. Less than 1% of the sample received more than 48 visits in the year prior to release.

Additional information

Funding

This material is based upon work supported by the National Science Foundation under Grant No. 153531. Any opinions, findings, and conclusions or recommendations expressed in this article are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Notes on contributors

Jillian J. Turanovic

Jillian J. Turanovic is an associate professor in the College of Criminology and Criminal Justice at Florida State University. Her research is aimed at examining various issues in criminological theory and correctional policy, with a special focus on victimization, violence, and health.

Melinda Tasca

Melinda Tasca is an associate professor in the Department of Criminal Justice at the University of Texas at El Paso. Her research focuses on corrections, the collateral consequences of incarceration, and disparities in the criminal justice system.