Occupational Licensing of Social Services and Nursing Home Quality: A Regression Discontinuity Approach

Occupational licensing has grown dramatically in recent years, with over 25% of the U.S. workforce having attained a license as of 2008, up from 5% in 1950. The associated debate as to whether licensing improves quality or is simply rent-seeking behavior has correspondingly grown in intensity. We contribute to the growing literature by estimating the impact of occupational licensing on quality, focusing on the staffing of social services departments in skilled nursing facilities (SNFs). Unlicensed paraprofessionals and qualified social workers provide the same services in social services departments, but a staffing provision of the Omnibus Budget Reconciliation Act of 1987 (OBRA-87) requires all SNFs with 121 or more beds to staff at least one full-time equivalent (FTE) qualified social worker.

Qualified social workers (QSWs) must be licensed to practice in the state the facility is located in, or if the state does not have its own licensing requirements, the federal government requires QSWs to meet educational and experience requirements similar to state licensing requirements. Using a regression discontinuity design, we find that this provision binds at that margin for a significant share of SNFs. QSW staffing increases by 8-10%. However, the overall increase in social services staffing is negligible because firms tend to meet this requirement in the lowest cost way – substituting from unlicensed paraprofessional social services staff to QSWs, effectively increasing the licensure level of the marginal social services employee. This policy creates a natural experiment whereby SNFs just above the threshold are induced to use a significantly higher share of QSWs than slightly smaller facilities.

Our study is different from existing studies on occupational licensing and quality in two important ways. First, existing studies exploit differences in licensing requirements across states. Estimated effects could be biased by sorting of individuals into areas that are more or less regulated. In contrast, the policy requirement we focus on leads to variation in licensure within a geographic region at a given point in time, based on facility size. Second, our study focuses on firms substituting between two possible inputs on the margin: a licensed employee or a non-licensed employee performing similar duties. In other words, we estimate the quality impact of requiring the marginal employee to be licensed.

We find that the OBRA-87 social services staffing provision led to about a six percentage point increase in the share of social services hours that are staffed by QSWs. Importantly, we find no evidence of changes in any other type of staffing levels (e.g. nurses) or facility characteristics (e.g. for-profit vs. non-profit) at the threshold, suggesting that the only discontinuous change is in the licensure composition of social services departments. Using the same regression discontinuity design with outcomes on quality, we find no evidence of a corresponding change in overall SNF quality or provision of social services, as measured by the deficiency score received by a facility during their annual recertification inspection. There is tepid evidence that increased licensure improves deficiency scores in the narrow category of admission, transfer, and discharge rights.