Monday, July 20, 2009

Why buy a state you already own? 

Reading John Lott today reminds me yet again of the research of Gavin Wright on the New Deal. (JSTOR copy here.) Written in 1974, Wright found that the distribution of grants and of WPA jobs were driven in no small part by political considerations.
The distribution of WPA employment was "a storm center from the beginning" (Howard 1943, p. 596), and charges of discrimination were common, most especially from governors who believed they were being shortchanged (Patterson 1969, pp. 54, 57, 77, 81, 198-200). Apprehension about the motives of WPA was promoted also by the fact that "the WPA was most reluctant to give out any information about the way in which state employment quotas were determined" (MacMahon et al. 1941, p. 223).
Wright found that if a state was more "in play" -- by which was meant the degree to which states switched support between Republican and Democratic candidates in the prior half-century -- it was more likely to receive grants and WPA jobs.

Lott writes:

Politico reported on June 5 that the �Stimulus tour� � visits by Mr. Obama and other administration officials �across the country to tout the massive spending program or hand out stimulus cash to grateful local officials� � overwhelmingly took place in states that voted for Obama: �52 of the 66 events were in states that backed Obama.� The other 14 events were in states that Obama lost only narrowly. A new study released by USA Today also finds that counties that voted for Obama received about twice as much stimulus money per capita as those that voted for McCain.

In our results, Obama's share of the vote accounted for only a small percentage of the variation in how the stimulus money is being allocated. A one percentage point increase in Obama's vote share means an additional $13.52 in per capita spending, but even then the relationship rests on the large amount of money given to D.C.

I would argue John needs a fuller analysis here of voting patterns. It makes no sense to be lavish on Washington DC in this view because the District is reliably Democratic. I would look for states that have had votes switch more often, swing states like Ohio or Missouri or Florida. If Obama's spending is politically motivated, you should be able to use that variability of voting with the measures Lott includes like unemployment, foreclosure and bankruptcy and find a result.

Looking at his regression, I see he's included per capita income and vote share for Obama -- I'd like to see a correlation of those two, because I suspect richer states voted for Obama. I would also think there's some multicollinearity in the various measures of economic hardship included. I'll send John a link, and suggest he try some additional tests to flesh out this result a la Wright or John Wallis (REStat 1987, JSTOR link.)

UPDATE (7/21): John emailed correlation coefficients for per capita gdp growth and Obama's vote share: .60 with DC included, .46 without. The DC influence is VERY strong.

Labels: , , ,