The impact of right-to-work on county-level employment share is estimated in each of 18 different industries, as defined using 2-digit, NAICS codes. Sample sizes vary across the industries examined due to county-level employment data availability in the County Business Patterns. We limit the discussion here to the six industries for which union membership is high, including utilities, construction, manufacturing, transportation and warehousing, information, and education services. These industries have unionization rates that exceed the private sector average, ranging from 9% in manufacturing to 20.1% in utilities, according to the Bureau of Labor Statistics.
These results from SEM estimation can be found in Table 1. The results with respect to the impacts of right-to-work are color coded. Dark green is for results that have the expected sign and are statistically significant. Light green is for the expected sign but not statistically significant. The dark red is for results with a different sign and that are statistically significant. The light red is for results with a different sign than was expected without statistical significance. A mix of positive and negative results across the industries must be observed given that we are measuring shares of total private employment. That is, if employment share is rising in one industry, it must be falling in at least one other industry.
Table 1: Regression Results, Six Union-Dense Industries
In general, we find evidence that counties in states with right-to-work legislation observe higher employment share in certain industries as a percentage of total private employment — what we call the direct effect. Much of this increase could be the result of the relocation of employers from nearby non-right-to-work counties — what we call the indirect effect. The direct effects of right-to-work legislation can be observed from the estimated coefficients on the Right-to-Work pre-2000 and Right-to-Work post-2000 variables, displayed in Table 1. While not always statistically significant, they are positive in sign, with the exception of education services and transportation and warehousing. The indirect effects are observed from the estimated coefficients on the two interaction terms: W*Right-to-Work pre-2000 and W*Right-to-Work post-2000. The signs of the indirect effects are more mixed in these higher union density industries but tend to be negative more often than positive. This indicates that having neighboring counties in right-to-work states reduces the industry employment share in the bordering non-right-to-work county.
With some exceptions, the direct and indirect effects largely cancel out for interior counties. The more interesting cases concern the border counties between states which differ in their policies regarding right-to-work. There is a clear indication that, in most but not all high union density industries, border counties in states without right-to-work laws lose industry employment share while border counties in states with such laws gain employment share. It is possible that border counties in non-right-to-work states are losing jobs in some industries to neighboring counties in right-to-work states. However, this relocation of employment opportunities to right-to-work counties is not the only source of increased jobs for these counties.
To illustrate the effects of right-to-work policy in counties along each side of a border between two states with differing state policies on right-to-work, we make the following assumptions regarding our representative border counties: 1) the typical county borders five other counties; 2) right-to-work counties along the state boundaries in question typically border three other right-to-work counties; and 3) border counties along the state boundaries in question and in the non-right-to-work state border two right-to-work counties.
It is necessary to make such assumptions in order to interpret the effect of the interaction terms. Recall that the interaction of the weight matrix and the right-to-work variables is interpreted as the percentage of bordering counties in right-to-work states (pre- or post-2000 enactment); as such, a percentage must be imposed for interpretation purposes. These assumptions are intended purely for illustrative purposes and results can easily be computed with alternative assumptions.[*]
The results regarding the manufacturing industry (two-digit NAICS code 31), which accounts for just over 16% of private employment, are distinct from the other industries highlighted here. We find that right-to-work passage leads to increased manufacturing employment as a percentage of total employment in all counties of interest: interior right-to-work, border right-to-work, and border non-right-to-work, a finding that is particularly large for post-2000 adoption states, such as Michigan. Specifically, manufacturing employment share for interior right-to-work counties is estimated to be between 15.5% (pre-2000 adoption) and 31.5% (post-2000 adoption) higher. Border counties in right-to-work states experience between 12.1% and 20.7% higher manufacturing employment.[†]
The most interesting result regarding the effects of right-to-work on manufacturing employment might be from border counties in states without right-to-work laws. These counties also observe increased manufacturing employment, by between (pre-2000) 3.4% and 10.8% (post-2000). This finding might be the result of increased demand for inputs by manufacturers in the right-to-work state. With some of those suppliers being located across state lines in the non-right-to-work state, the employment benefits to manufacturing can spill over across state borders.
The results from an analysis of the construction industry (two-digit NAICS code 23), nearly 6% of private employment, differ from those of manufacturing as the results are more mixed. Construction industry employment share in interior counties is estimated to be 15% higher in states that established right-to-work before 2000. However, in those states which have more recently adopted right-to-work, construction employment share in interior counties is lower by 7.1%. Right-to-work border counties observe between 1.6% (post-2000 adoption) and 14.2% (pre-2000 adoption) higher employment share for the construction industry. Border counties in non-right-to-work states are estimated to experience between 0.8% higher (pre-2000 adoption) and 8.7% lower (post-2000 adoption) construction industry employment. Also of note: The pre-2000 effect of right-to-work laws appears especially strong for the construction industry.
Consider next the effect on the transportation and warehousing industry (two-digit NAICS code 48), which accounts for just over 4% of private employment. While the effect of right-to-work is not found to be statistically significant due to large standard errors, we will discuss the point estimates. Interior counties in right-to-work states observe industry employment that is between 4% (post-2000) and 11.7% (pre-2000) higher than would be otherwise as the result of the policy. Right-to-work border counties in states which adopted the policy prior to 2000 are estimated to also experience an 11.7% increase in employment share, yet the equivalent counties in post-2000 adoption states are estimated to observe a 3.5% reduction in transportation and warehousing employment. Interestingly, bordering counties from non-right-to-work states were unaffected if bordering a pre-2000 right-to-work state, but experienced a 7.5% increase in employment share in the industry if bordering a post-2000 right-to-work state. Like manufacturing, this might be the result of the coordination of firm activities and suppliers across state lines.
Given the smaller industry share of total private employment for the utility industry (1.4%), the information industry (1.6%) and the educational services industry (2.3%), we only quickly highlight the results for these three industries. In both the utilities industry (two-digit NAICS code 22) and the information industry (two-digit NAICS code 51), right-to-work is estimated to increase industry employment share for border counties in right-to-work states and decrease industry employment share in both interior right-to-work counties and border counties in non-right-to-work states. Right-to-work in the educational services industry (two-digit NAICS code 61) is estimated to reduce employment share in all cases except for border counties in non-right-to-work states, when right-to-work in the bordering state was passed post-2000.
Table 2 presents the SEM estimation results for the remaining twelve less union-dense industries. For conciseness, we only highlight the qualitative effects across a selection of the industries, particularly those with greater statistical significance. Before dividing the counties into border and interior categories to analyze the results, some interesting findings pop out of the spatial-based analysis. For instance, right-to-work laws are estimated to increase the accommodations and food services industry employment share for counties within a right-to-work state. Nearby counties, particularly those in non-right-to-work states, experience employment losses in this industry.
While exerting statistically insignificant direct effect on the mining industry employment share, right-to-work is estimated to benefit mining employment in neighboring counties, again, particularly for the nearby non-right-to-work counties. The finance and insurance industry is similarly affected; however, the direct effect of right-to-work laws passed prior to year 2000 is to reduce the employment share slightly. Right-to-work is estimated to reduce the industry employment share for health care and social services in right-to-work counties but has no statistically significant effect on neighboring counties. Results concerning the remaining industries are more sporadic and largely statistically insignificant and can be viewed in Table 2.
Table 2: Regression Results, Non-Union-Dense Industries
In addition to the estimates based on all counties (for which data is available) in the continental U.S., we also estimate our model for the manufacturing, construction, and transportation and warehousing industries with restricted samples to include Michigan and its border states and, separately, Indiana and its border states. We now turn to the discussion of those results. As was conducted for the full sample, we estimate the model using both the SAR and the SEM. The SAR takes the form as follows with the spatially lagged dependent variable included as an independent variable:
Y = ρWY + Xβ + ε
where Y is once again defined as industry employment share, ρ is the spatial autoregressive parameter to be estimated, W is the previously discussed spatial weight matrix, X is the vector of independent variables, β is the vector of coefficients on the independent variables, and ε is the well-behaved error term. The SAR estimator appears to be a better fit for these smaller samples; as such, it is the direct, indirect, and total effects from the SAR estimation that are discussed below. With that said, the qualitative results are generally consistent across the two spatial estimators.
[*] Given our assumption of the typical county, the estimated effect of right-to-work on RTW border counties is computed as the product of the coefficient on Right-to-Work pre-2000 and 0.6*W*Right-to-Work pre-2000. Likewise, the estimated effect of right-to-work on external border counties is computed as the product of the coefficient on Right-to-Work pre-2000 and 0.4*W*Right-to-Work pre-2000. The same calculation was done for the coefficients on Right-to-Work post-2000 terms. Appropriate t-statistics were then computed in order to carry out the test of statistical significance.
[†] The 15.5% increase in manufacturing employment share for interior counties is found by the following calculation: 1.118 + (1.386*1) for a 2.5 percentage point increase in manufacturing employment as a share of total private employment. We then divide this figure by the industry market share of 16.146 for a percentage change of 15.51%. In this calculation, the coefficient on the interaction term, 1.386, is multiplied by 1 because 100% of all contiguous neighbors for interior right-to-work counties are in a right-to-work state. The 12.1% figure for representative border counties in right-to-work states instead multiplies that coefficient by 0.6 because we assume 60% of its bordering counties are in right-to-work states: [1.118+(1.386*0.6)]/16.146 = 12.08%.