Methods

If licensing and licensing burdens are necessary to protect health and safety, this would mean that businesses like barbershops and nail salons with unlicensed or less onerously licensed staff present a greater threat to health and safety than similar businesses with licensed or more onerously licensed staff. Using inspection outcomes as a measure of health and safety risks, I test whether licensing and licensing burdens are related to inspection outcomes.

To do this, I employed a design that approximates a randomized controlled trial. In a true randomized controlled trial, I would randomly assign nail salons and barbershops to states with different licensing conditions and compare their inspection outcomes. This would allow me to attribute any differences in inspection outcomes between businesses in the states to their licensing conditions and not to other potential differences between the states. Obviously, such a research design was not feasible. However, there are various ways to approximate a randomized controlled trial, and one of those ways is to use a geographic regression discontinuity design, a research design that takes advantage of variation around geographic boundaries. ¹  Subject to qualifications, the basic idea is that, on average, businesses and business environments equidistant from the border between two states are similar, with that similarity increasing as the distance to the border decreases. Intuitively, this makes sense. For example, one would expect businesses and business environments in Kansas City, Kansas, to be more comparable to businesses and business environments a few miles away in Kansas City, Missouri, than to those hundreds of miles away in St. Louis.

The reason the design approximates a randomized controlled trial is that, within a certain, relatively short, distance of the border—known as a “bandwidth”—it is as if the businesses were randomly assigned to one side of the border or the other. Assuming this is true, it is possible to estimate the expected difference in inspection outcomes between a business with unlicensed or less onerously licensed workers on one side of the border and a business with licensed or more onerously licensed workers on the other. In essence, each inspection outcome receives a weight that is a function of the businesses’ distance to the border. The estimated expected difference is a weighted average effect of licensing requirements on inspection outcomes, with more weight given to businesses assumed to be more similar (i.e., those closer to the border) and less weight given to businesses assumed to be less similar (i.e., those farther from the border). In this way, this design allows us to attribute differences in health and safety outcomes, as measured by inspections, to differences in licensing regimes.

This study uses nail salon inspection data from Connecticut and New York and barbershop inspection data from Alabama and Mississippi to answer this question: All else constant, do businesses with licensed staff or more onerously licensed staff commit fewer violations and have a greater probability of passing inspections compared to businesses with unlicensed staff or less onerously licensed staff?