We estimate the relationship between right-to-work laws to three economic variables: growth in total employment, real personal income and population. These relationships are analyzed in three distinct time periods: 1947 through 1970, 1971 through 1990 and 1991 through 2011. The purpose of this approach is to evaluate both the impact of right-to-work on these variables and how this impact varied across time periods. These estimates are displayed in the graphics below.[*]

**Graphic 4: Growth in State Total Employment, 1970-2011**

**Graphic 5: Growth in State Personal Income, 1947-2011**

**Graphic 6: State Population Growth Rate, 1947-2011**

Our analysis assumes that growth rates for employment, personal income and population are measures of overall economic well-being, and that right-to-work laws affect them through a labor-demand function. This labor-demand function yields conflicting theoretical possibilities as to the impact of unions, which has been the challenge to existing research in this area for some time.[15]

We also assume that the results above permit us to interpret that the right-to-work laws’ dummy variable is clean in the sense that it does not capture other policy variables which are not perfectly coincident. While the estimation process leads to this assumption in our interpretations, the relaxation of this assumption simply alters the interpretation from a strict right-to-work effect to that of a combined suite of policies of the type offered by Holmes.[16]

The first observation from the results displayed above is that right-to-work laws have a positive and statistically meaningful influence on growth during the length of the observed period (the first column of results in each graphic). This varies from 1971-2011 for total employment, and from 1947-2011 for real personal income and state population growth.

Interpreting the size of these coefficients requires some assumptions about the spatial relationship of right-to-work laws. If we assume that the enactment of right-to-work laws holds no spatial relationship, then the coefficient estimates above may be directly applied to the growth rates in the estimates. While this might appear attractive due to the very low level of statistical inference directly attributable to the adjacent right-to-work variable, this is a very restrictive assumption. Moreover, a casual glance at Graphic 1, as well as a cursory reading of the debate surrounding Michigan’s adoption of a right-to-work law, suggests that the presence of right-to-work laws in an adjacent state affects the adoption of these laws in other states. James LeSage and Matthew Dominguez offer a very clear approach to displaying the effects of a policy which is applied with some spatial dependence.[17] In the case here, we find an effect of right-to-work laws in both the own state and the effect of right-to-work in adjacent states (first-order contiguity), which LeSage and Dominguez refer to as indirect or spillover effect. These estimates appear in Graphic 7.

**Graphic 7: Spatial Dependence Estimates, 1947-2011**

*
*Not statistically significant
*

These impacts are relatively large averaged over the entire period, with growth rates boosted by up to 0.76 percentage points for employment (1970-2011), 0.75 percentage points for real personal income and 0.54 percentage points for population, noting that the early periods of right-to-work laws were the least affected, with no statistically significant impact on population or income growth.

We believe these results are not sensitive to alternative specifications which address the gravest concomitant variable problem concerns. For example, we include real per-capita taxes as a proxy for concomitant fiscal variables in both the estimations presented above and in an FGLS estimate without the endogeneity correction. In the first case, the fiscal variable is not statistically meaningful at any significant level, in any of the three measures of aggregate economic activity. The coefficient on both right-to-work and adjacent state right-to-work rose slightly (0.02 to 0.03 in all three estimates).

These estimates were not statistically different from each of the original estimates using a Wald test. No other results from the original estimate varied meaningfully. There are other concomitant variables; most especially labor-related regulations which we must concede remain a problem. However, we feel that the estimation strategies described above relieve many of the larger concomitant variables and omitted variable bias. So, we interpret the right-to-work variables as representing the laws and a small set of closely associated labor market regulations in states.

[*] The asterisks used in Graphic 4, Graphic 5 and Graphic 6 denote the level of significance for each statistic. One asterisk means the finding was significant to the 10 percent level, two the 5 percent level and one the 1 percent level.