One of the purposes of “right to work” legislation, currently being debated in Indiana, New Hampshire, and other states, is to reduce the percentage of the workforce covered by collective bargaining agreements. Leaving aside the ethics of collective bargaining as practiced in the U.S. today, what are the political and economic consequences? Since unions donate almost exclusively to Democratic candidates and lobby heavily for more government regulation and spending, it would be unsurprising if more unionized states ended up with bigger state governments.
To examine the evidence, I ran statistical models of unionization and taxation over the 2000-2008 period. The dependent variable in the first analysis is state and local tax collections as a percentage of state personal income, excluding mineral severance and gas taxes (since large and resource-abundant states will otherwise look like states with large tax burdens), in Fiscal Year 2007-8, the latest year for which data are available. The main independent variable is the percentage of the state workforce covered by collective bargaining contracts in 2004. I measure the variable a few years earlier since changes in unionization rates should take some time to show up in changes in election results and public policies. For control variables, I use legislative professionalism from 2003 (a variable in political science that captures the extent to which state legislators can use their positions as a career, 2003 is the latest year available), the average Partisan Voting Index (PVI) in the 1996-2004 presidential elections (Democratic+Green+Nader vote shares in the state minus those shares in the nation; I “penalized” states with a Democratic candidate from that state 4 points if a challenger, 2 points if an incumbent, and “rewarded” states with a Republican candidate from that state 4 points if a challenger, 2 points if an incumbent – about the size of the home-state effects that political scientists have previously estimated), the square of the PVI, mineral severance tax collections as a percentage of personal income (more mineral wealth means you can afford to have lower taxes in other areas), and federal grants to each state and its local governments in FY 2007-8 (federal grants may either substitute for or incentivize state taxation).
Here are the results:
What they show is that even when you control for overall state ideology (Democratic states have higher taxes, and really Democratic states have much higher taxes), union density increases tax rates. Increasing union density from 10% of the workforce, as in Nebraska, to 25%, as in Hawaii and New York, increases the tax burden by about one and a quarter percentage points of state personal income. For a sense of scale, the mean tax burden was 10.0% of personal income in 2008, and the standard deviation was 1.23, so this is essentially a standard deviation increase.
By passing right-to-work, Indiana could expect its unionization rate to drop anywhere between four and nine percentage points, taking in the range of values observed in other right-to-work states. This change would decrease Indiana’s tax burden in the long run by between a third and three-quarters of a percentage point of personal income. The predicted effects in New Hampshire would be slightly less, since New Hampshire is slightly less unionized than Indiana.
I have run some robustness checks on these results. I tried instrumenting for the potentially endogenous controls legislative professionalism and federal grants with log state population, population per state house district, population per state senate district, and per capita personal income in 2000. The results continue to show a statistically significant, positive effect of union density on tax burden, and the coefficient remains essentially the same size. I also tried dropping grants and replacing it with per capita personal income and log state population – again with no significant changes to the other estimates.
As an aside, note also that federal grants appear to increase state and local taxation. If states think they can get away with just spending other people’s money by taking money from the feds, they’re mistaken.
Finally, I ran an analysis in which change in tax revenue from 2000 to 2008 was the dependent variable. The independent variables were unionization rate in 2000, tax revenue/personal income in 2000, PVI, PVI-squared, legislative professionalism in 1996, change in legislative professionalism from 1996 to 2003, change in severance revenue from 2000 to 2008, and change in federal grants from 2000 to 2008. Here are the results:
|Tax revenue, 2000||-0.26||0.11||0.025|
|Leg. professionalism, 1996||1.11||0.89||0.22|
|Change in leg. prof., 96-03||1.16||2.48||0.643|
|Change in severance taxes||-0.01||0.04||0.781|
|Change in federal grants||0.19||0.10||0.055|
These results are rather poor: nothing is highly statistically significant, although the model as a whole is marginally statistically significant. Nevertheless, the coefficient on union density is positive and “close” to statistical significance. Over this short a time period, it seems to be difficult to predict which states increase and decrease taxes. If the result on unionization rate is real, it means that each additional percentage point of union density in 2000 corresponded with a 0.036% (of personal income) tax increase over this eight-year period. Thus, New York would have had 0.5% of personal income less tax increase over these eight years if its union situation had been like Nebraska’s. Again, given the short time period, this effect is not small – there’s just a great deal of uncertainty about it.
In conclusion, the evidence suggests that policies discouraging collective bargaining help a state reduce its tax burden in the long run, but the short-run effects are less certain.
Update: Clarified PVI construction.