Wednesday, March 11, 2009

Right to Work and Unemployment

The January state level unemployment figures came out today. The rates range from 3.7% in Wyoming to 11.6% in Michigan.

Over at Carpe Diem, my friend Mark Perry correctly points out that the 8 states with the lowest unemployment rates are all "right to work" states.

But that's not really the whole story. There are data for 50 states plus DC and there are 22 right to work states. We'd want, at a minimum to know, if right to work states have significantly lower unemployment rates than the others.

A couple minutes of slaving over EVIEWS produced the following simple regression

State Unemployment = 7.631 - 0.990*Right to work State

While this says that right to work states have about 1% point less unemployment on average, the t-statistic testing the null hypothesis that the coefficient on the right to work state variable is actually zero is computed to be -1.935. Thus we cannot reject that there is no difference between right to work and non right to work states at the traditional 5% significance level (though we could at the 10% significance level). If we adjust the standard errors for heteroskedasticity the t-statistic falls to -1.87 but the inference remains the same.

If we add an additional dummy variable for the 4 states hit hardest by the foreclosure crisis (California, Arizona, Nevada, and Florida (three of which are right to work states)) we get this:

State Unemployment = 7.47 - 1.183*Right to work State + 1.557*Foreclosure crisis State

In this case both the right to work coefficient and the foreclosure crisis coefficient are significantly different from zero at the 5% level.

So at the simple level of differences between means, there is no significant difference in current unemployment rates between right to work and non right to work states. Multiple regression may reveal a different picture. Of course in all these cases, without a theoretical model behind it, the results are only correlations and can't really be taken as anything causal.

4 comments:

Norman said...

This is the kind of research question I thought I'd be asking when I was applying to graduate schools. Interesting stuff.

MaxSpeak said...

If you put the #s on a map, the regional pattern seems pretty obvious. The SE is in the dumps, Red (RTW) and Blue together, also reddish states IN and MO.

Anonymous said...

"Of course in all these cases, without a theoretical model behind it, the results are only correlations and can't really be taken as anything causal".

Arn't the statistical results only marginal correlations (controlling for other covariates) even with a theoretical model?

Angus said...

If I had a solid theoretical model that said causation flowed one way only from right to work to unemployment and that no other relevant variables were excluded, then I could take the estimated relationship to be causal.