Fake Science and Bad Public Policy: Lessons from Venture Capital

Abstract

I review statistical methods used to estimate crowding out of private venture capital by government venture capital in recent work. I explain the importance of using data that pre-dates the creation of government venture capital in order to examine whether or not government has crowded out private venture capital. I show that statistical inference in recent work on topic that does not follow this straightforward statistical rule can give rise to remarkably incorrect conclusions; including, for example, a bizarre and clearly false inference that a market with more than 89% investment by government funds exhibits no evidence of displacement of private funds. Further, I point out an alarming number of inaccuracies and misstatements and an unusual refusal to explain problems with data and analyses amongst authors and editors alike.