LAO on Varshney Studies

Yesterday, the LAO released a review of the Varshney and Associates AB32 and cost of regulations studies that I have blogged about here and here. The LAO does a nice job of laying out the many flaws, so I won't repeat them here.

However, there is one sentence buried in a paragraph on page 3 of the LAO review that was interesting new information to me (emphasis added).

The regression analysis has a number of problems. The biggest is that the relative size of states was not taken into account by V&T in explaining interstate differences in GSP. One way to have done this would be to have focused on explaining GSP adjusted for a measure of the size of each state, such as by using per capita GSP. The authors indicated to us that they in fact tried this methodology, but got poor results and thus did not use them. Our own regressions indicate that when per capita GSP is used, the regression outcomes dramatically change and become questionable. For instance, the regression itself explains only a bit over 10 percent of interstate differences in per capita GSP, and the regression coefficient measuring the effect of the regulatory index on the economy changes its sign from negative to positive. The latter finding, which also has been noted by other economists who have reviewed the V&T study, is inconsistent with the hypothesis that a poorer ranking on the regulatory environment index hurts GSP.

If you understand regression analysis and research methods, that should certainly cause a raised eyebrow. It gives new meaning to this passage from today's Sacramento Bee article.

Varshney defended both studies in an interview Thursday, saying "we really gave
it our best shot and gave it our best efforts."