Women, Criminal Records, and Certificates of Relief: An Experimental Study

Abstract

Few studies have examined the impact of criminal history on hiring outcomes for women and even fewer studies have explored the effectiveness of mechanisms aimed to improve hiring outcomes for women with criminal history. The overall objectives of the current study were to examine (1) whether criminal history negatively influenced hiring outcomes for women and if so, (2) whether Ohio’s certificate of relief could improve hiring outcomes for women. These objectives were achieved with the use of an experimental correspondence audit that addressed limitations of prior research and utilized innovative methodological approaches. The results showed that callback (i.e., invitation to continue in the hiring process) point estimates for those with no record were higher than those with a record and no certificate. However, the differences between these two conditions lacked statistical significance. Further, while callback probabilities for certificate holders were statistically indistinguishable from those with no record, the callback probabilities for certificate holders were also statistically indistinguishable from those with a record and no certificate. Finally, there were no statistically significant racial differences. These results were supported in several robustness checks. Despite the lack of statistical significance, several findings have substantive significance and these findings are discussed in detail.

Keywords:

• Certificates of Relief
• Collateral Consequences of Conviction
• Employment
• Rights Restoration

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 See Ortiz (2014) for a detailed discussion surrounding system-involved women and employment.

2 The current study was part of a larger effort that resulted in several papers. These papers inevitably share various components given the overlap in topic and methods.

3 The indeed.com website would sporadically present new out-of-order postings after page changes or refreshing (even when the postings were sorted by date). While efforts were made to identify these postings, a complete census cannot be guaranteed. Nonetheless, the descriptive statistics show that the overall sample of postings is an accurate representation of lower-level employment in Cleveland, Ohio (see Bureau of Labor Statistics, 2019c).

4 Resume-only submission was recommended by the work readiness program because individuals could apply to more positions during their job searches and because the listings offering resume-only submission were often positions that those with criminal history would be more likely to secure (given the prevalence of lower-level postings).

5 A small number of postings (approximately five) included language that explicitly discouraged those with criminal records from applying (e.g., “no felonies”). The decision to exclude these postings was based upon the small number of postings that included such statements and the bias which would be introduced to the study because of the criminal record condition without a CQE.

6 Sample size calculations were largely based upon our primary research question (differences between criminal record conditions). This would equal 200 observations for each condition. However, the study was also designed to have adequate power to detect statistically significant effects for our secondary research question (differences between criminal record conditions across racially distinct names). This would equal 100 observations for each condition. Leasure and Andersen (2016) used very comparable numbers and was able to detect statistically significant effects in their similar study that focused on male applicants. Sample size calculations did not include the robustness check models.

7 A corrupted file resulted in the loss of information regarding control variables for eleven observations. However, the authors were able to extract ten job types from the title of the listing. Further, because of the low numbers of lost observations, their classification as missing completely at random, and because treatments were randomly assigned, the authors felt this issue could be safely ignored.

8 Robustness check results are presented in the Technical Appendix. The individual name, interaction, and multiple call models were all estimated with and without controls. However, adding controls to those models largely did not alter the statistical or substantive significance of any results. Therefore, the results of those models with controls added are not presented in the Technical Appendix. Nonetheless, the individual name, interaction, and multiple call models with controls added are available upon request.

9 See the Technical Appendix for regressions presenting odds ratios as well as information on regression diagnostics.

Additional information

Funding

This work was supported by a York College of Pennsylvania Research Grant

Notes on contributors

Peter Leasure

Peter Leasure, J.D., Ph.D. ([email protected]) is an Assistant Professor at York College of Pennsylvania in the Department of Criminology and Criminal Justice. His research focuses on experimental design, corporate compliance, data quality control, and collateral consequences of conviction. His research has appeared in various scholarly journals such as Crime and Delinquency, the Journal of Experimental Criminology, the Journal of Financial Crime, and the Journal of Money Laundering Control.

Gary Zhang

Gary Zhang, Ph.D. ([email protected]) received his Ph.D. in Criminology and Criminal Justice from the University of South Carolina in 2018. His recent publications have appeared in a variety of peer-reviewed journals, including American Journal of Criminal Justice, Crime & Delinquency, Deviant Behavior, and Policing: An International Journal.