Yesterday I shared a graph showing that states with the biggest spending cuts are suffering from the steepest employment losses. Deliberately oversimplified version of that post: States with lots of oil and barley are doing great. States with too much home building are not.
Today, author and graph-maker, Adam Hersh from the Center for American Progress, was kind enough to redraw the graph excluding some of the worst states (Florida, Arizona, and Nevada) and best states (Texas and the Dakotas). What he found was the story didn't change much. Spending cuts still had a statistically significant effect on job losses.
"State spending choices are not deterministic followers of economic conditions," Hersh concluded* in an email to me. States may be constrained in the short term by balanced budget laws to make spending match revenues, but they still get to choose whether to raise taxes or cut spending on public services and investments, and this choice has economic consequences.
It's clear that job cuts come from state cuts. But where do state cuts come from? They come from economic conditions. Another way to look at this question is to ask: What's the matter with the Sun Belt?
Look at the states with the worst job losses. Eight of
them are in the top ten for highest foreclosure rate: Oregon, Georgia,
Florida, Michigan, Idaho, Utah, California, Arizona and Nevada. Two
more, New Mexico and North Carolina, are in the Sun Belt. That doesn't mean solar exposure is bad for employment. It means lots of
families and developers were drawn to the southern U.S. in last
decade's rush for cheap credit and housing and today those are the same
states feeling the economic crisis the worst. Recent budget cuts in these states are an ingredient in a stew of economic trouble.
*His explanation via email: "I also tested for reverse causality, whether the economic downturn 'caused' the shrinking state spending. I say 'caused' because it is not really possible to determine causality strictly with econometrics (to determine causality requires good experimental design which is too often not possible in social science). However, I did some more regressions where (Y) state spending cuts in 2009 and 2010 is the dependent variable and (X) GDP growth and change in private employment in 2007-2008 are specified as independent variables. Since X precedes Y in time, it follows that X can be seen to "cause" Y if there is a statistically significant relationship between the two. And there is not--or at best an exceedingly weak relationship that does not meet generally accepted levels statistical significance (like a 35-50% probability, compared to a standard 95% probability; the probability of the relationship in the graphs is better than 99%)"