We empirically test the effects of right-to-work legislation on employment trends by estimating a model to describe county-level industry employment as a percentage of total private employment for 18 industries for the year 2018.
One advantage of examining industry employment as a percentage of total private employment is that the variable is automatically adjusted for ups and downs in the business cycle. That is, both the numerator and the denominator are impacted by general economic trends.
Control variables used in the model include various county-level population demographics obtained from the U.S. Census Bureau, such as total population, percentage of population in poverty, percentage of population aged 25 and older with at least a bachelor’s degree, percentage of population who are nonwhite, percentage of population who are female, and percentage of population aged 20 to 64, the typical working age.
While we are interested in the effect of right-to-work legislation, we must also control for each state’s general economic and policy environment. A state that has opted for right-to-work legislation might tend to enact other types of policies that may impact employment and economic performance. To control for such policies, we include the 2017 “economic freedom score” from the Fraser Institute’s Economic Freedom of North America report. This index ranks states based on how market-friendly their policies are — the higher the ranking, the more market-friendly. We use this in the model to help avoid inappropriately assuming certain economic outcomes are the result of right-to-work laws.[*]
The direct effect of right-to-work legislation on a county is measured through the inclusion of two binary variables. There are roughly two waves of states enacting right-to-work laws — one in the 1940s and 1950s and another in the 2010s. Early adopters of right-to-work legislation may have different employment patterns than those states that more recently adopted right-to-work, and the effects of the enacting the law may diminish over time. We attempt to take these potential differences into account by dividing the analysis and results into two groups: pre-2000 right-to-work adopters and post-2000 adopters.[†]
Six states enacted right-to-work legislation after 2000: Oklahoma in 2001, Indiana and Michigan in 2012, Wisconsin in 2015, West Virginia in 2016 and Kentucky in 2017. It is possible that the full effects of the most recent adopters will not have been observed in the 2018 data. In both instances, however, we expect that counties in right-to-work states will enjoy higher employment share than they would have without a right-to-work law, particularly in industries traditionally represented by higher union membership.
To measure the indirect effect on counties in nearby non-right-to-work states that border a right-to-work state, we include variables indicating whether a given county borders another in a right-to-work state. To the extent that job opportunities move toward right-to-work states, one might expect such counties to observe lower employment share in union-dense industries, especially if employers tend to move from nearby states to ones with new right-to-work laws. However, it is also possible that right-to-work policies create general growth in a region, which could spill over into nearby non-right-to-work counties, potentially leading to increased employment in counties bordering right-to-work states.
The estimation model employed for the analysis is a spatial error model. SEMs control for the influence of spatially correlated omitted variables. While researchers attempt to include the most theoretically important explanatory variables, omitted variables are commonplace in all empirical studies, in part, because of data limitations, but also because throwing in all possible variables (the proverbial kitchen sink) can lead to other empirical issues, such as multicollinearity. Particularly in the case of county-level data, many variables, including omitted ones, will be correlated systematically across space. The spatial correlation of omitted variables can lead the estimated coefficients from a standard regression to be inconsistent — that is, the expected value of the estimated coefficient does not approach the true (yet unknown) value as the sample size increases. The SEM tries to correct for this problem and reduces the potential negative influence from omitted variables. Further details on the SEM can be found in the technical appendix.[‡]
We first discuss estimation results employing data for the continental U.S. states. Sample sizes vary across the 18 industries examined due to county-level employment data availability in the U.S. Census Bureau’s County Business Patterns. Full results can be found in the technical appendix for all industries; however, we limit the discussion here to the six industries with the highest levels of union membership among employees. These include utilities, construction, manufacturing, transportation and warehousing, information, and education services. These industries had a higher than average rate of unionization in 2018, ranging from 9% in manufacturing to 20.1% in utilities, according to the Bureau of Labor Statistics.
[*] From the EFNA, we remove the union density (area 3Aiii) variable and recalculate the economic freedom score to reduce the likelihood that this variable measures similar policy influences as our right-to-work variables.
[†] The exact year chosen to divide the states into two groups is unlikely to have a large impact on the results. We chose the year 2000 in part because Oklahoma was the first state to change its policy on right-to-work in 16 years when it did in 2001, and within the next several years, more states followed suit, suggesting that a general shift in the political possibility of right-to-work laws kicked off around that time. Further, 2000 is a census year, which means more data is available for this year than others.
[‡] For robustness, we also estimate a spatial autocorrelation model, or SAR, which addresses the spatial correlation in a different fashion. Additional details on this specification and the discussion of the results from the SAR are discussed in the technical appendix.