Homeownership and Unemployment: Not So Fast

Print
Friday, 10 May 2013 05:12

Folks have been asking me about a new study showing a strong link between homeownership and unemployment. The study finds a long-term of elasticity of the unemployment rate with respect to homeownership close to 1. This means that if the homeownership rate in a state doubles then we should expect its unemployment rate to double. For the country as a whole, since the homeownership rate has risen by roughly 20 percent from its 1950 level we should expect the unemployment rate to be roughly 20 percent higher, after controlling for other factors.

These are striking results, but even as a critic of the cult of homeownership, I am not buying. The paper does include many tests for robustness, so there is no simple story of cherry-picking the data. But there are important questions of reverse causation. Suppose that states have weak economies so that many people leave for states with more job opportunities.

In this story the state losing people is likely to have a higher homeownership rate (homeowners are less likely to move) and the state getting people is likely to have a lower homeownership rate since the new arrivals are less likely to be homeowners. The study tries to control for this issue by having lags of up to 5 years, but it is certainly possible that trends in economic growth and stagnation are longer than this. It might have been useful to try lags of 10 years.

It is also striking that the states with the largest increase in homeownership are all in the south. If there was a rise in unemployment in these states was this a regional effect or due to homeownership? Including a regional variable might be helpful to see how it affects the results. In the same vein, immigrants are likely to be associated with both a lower unemployment rate (immigrants go to areas with jobs) and a lower rate of homeownership (recent immigrants don't own homes). I may have missed it, but it doesn't look like immigrant status is one of the control variables in the regressions.

One finding that may have a simple explanation is their finding that in the years 2000-2010 there was a strong tie between commute times and homeownership. I'll be a bit of a cynic here. Areas like Los Vegas and Phoenix were booming in the bubble years, these states saw substantial increases in homeownership. This was probably associated with an increase in commute times. The bust and drop in homeownership was especially pronounced in these areas. My guess is that they also saw a drop in average commute times. I don't know if this is really the story, but at first glance that would be my guess.

Anyhow, I can believe that homeownership has some negative impact on employment. There is the story of reducing mobility. This can be exaggerated (people do rent out homes and couples separate for work), but surely it is not zero. Also, policies that favor homeownership, like the mortgage interest deduction, undoubtedly pull capital away from productive investment. However, the relationships found in this paper seem too large to be plausible.

So chalk me up as a skeptic on this one, but it is an interesting paper that deserves serious consideration.   

 

Note:

It is interesting to see that the union density variable in these regressions is always negative and sometimes significant. This would suggest that higher union density is associated with lower unemployment rates. Much as I might like to say this is the case, my guess is that something else is at work.

States like Michigan and Ohio saw the percentage of workers in unions fall at the same time they lost hundreds of thousands of jobs in the auto and related industries. This could give the sort of correlation found in these regressions. This is the same sort of reverse causation that I suspect we see with homeownership and unemployment rates.

Further Note:

Danny Blanchflower, a co-author of the paper, notified me that he ran a regression that included a variable for the southern states to pick up any regional effect. He said that this actually made the results stronger, so clearly their findings are not driven by some peculiarity of the south that led to both higher rates of homeownership and higher rates of unemployment in the region.