A new paper by Steven Kreft and Russell Sobel, "Public Policy, Entrepreneurship, and Economic Freedom" (Cato Journal, Fall 2005), examines the causal relationship between venture capital (VC) and entrepreneurship. Do increases in venture funding lead to increases in entrepreneurial activity, or does VC flow to areas in which entrepreneurship is already taking place? Kreft and Sobel apply Granger causality testing to state-level panel data from 1992 to 2001 and conclude that VC follows, rather than leads, entrepreneurial activity.
I'm largely in agreement with the second part of their paper, showing that economic freedom (measured by a version of the Gwartney et al. index) is the primary determinant of state-level entrepreneurship. I have reservations about the main part, however. First, knowing the direction of Granger causality doesn't necessarily tell us much about (what Roger Garrison calls) "Webster-causality." Is there a reasonable explanation for the finding? One interpretation suggested by Kreft and Sobel is that VC is more mobile than labor, so the binding constraint on entrepreneurial is the availability of human, rather than financial, capital. But VC is heavily concentrated geographically (in the US, primarily on the East and West coasts), as is innovation more generally, and it's not clear that VC can flow freely to those areas that need it. Location matters, for obvious Hayekian and other reasons. Second, state-level measures of entrepreneurship and VC (and, for that matter, economic freedom) miss the within-state variation in entrepreneurship and VC, both of which are heavily concentrated in urban, rather than rural areas.
Third, the results contradict those in Samuel Kortum and Josh Lerner's influential paper on the causal relationship between VC and innovation ("Assessing the Contribution of Venture Capital to Innovation," Rand Journal of Economics, Winter 2000). Kortum and Lerner use a 1979 law allowing pension plans to invest in venture funds as an instrument for VC and find a significant effect of VC on patent rates. I find their instrumental-variables approach more convincing than Kreft and Sobel's Granger causality test.
