The use of criminal background checks has been growing over the last 20 years, and there are substantial policy concerns that these background checks significantly impact the employment outcomes of individuals with criminal history records.  Prior work with audit studies has shown a reduction in callbacks for those with records, but no prior study has been able to estimate the actual impact of these checks on the overall employment outcomes of people denied employment because of a background check. 

This paper responds to this challenge by focusing on background checks under New York Consolidated Statute Article 28-E, which requires the State Department of Health (DOH) to conduct thorough criminal record background checks on all individuals offered employment in residential health care facilities (e.g. nursing homes, assisted living facilities), with home health care agencies, or with long-term home health care programs licensed by the state. The Criminal History Record Check Legal Unit at DOH has been conducting background checks on potential employees since Article 28-E took effect in 2006.

In this paper, we examine the impact of being denied employment as the result of a criminal background check on subsequent labor market outcomes. Our sample consists of all persons who applied and were provisionally hired to work for a health care agency subject to Article 28-E background checks during the 2008 and 2009 calendar years (approximately 139,000 individuals). This sample is 85% female and 42% black with a median age of 36. Roughly 7% of the sample had at least one criminal conviction prior to their application. The New York Division of Criminal Justice Services has provided three years of follow-up data on subsequent offending and we have unemployment insurance data for 12 quarters before and after DOH’s clearance decision.

To generate a causal estimate of the impact of this clearance decision on working in the post-period, working in the healthcare industry in the post-period, and total earnings, we use four estimation strategies.  First, we use standard regression models, controlling for observable demographic and contextual factors, prior employment, and prior criminal history, to establish a baseline.  We compare these results to a matching analysis to focus in on comparable groups that only differ on observables by the clearance decision.  To address the issue that we are potentially omitting important variables that may also explain employment outcomes, we then estimate a panel data model with a fixed effect estimator to control for all unobserved factors that are either stable or have a stable impact on behavior over the course of six years (12 quarters before and 12 quarters after the initial offer of provisional employment).  Finally, we exploit a DOH clearance rule that provides a directive to offer clearances to individuals whose records are more than 10 years old.  In this regression discontinuity model, we control for all unobserved differences, both stable and time varying.