This estimate has been widely circulated, in everything from newspapers to Congressional testimony, but Heinzerling and Ackerman find it is not credible. They find it reflects a misunderstanding of the definition of the relevant data, fails an elementary question on the normal distribution, pads the analysis with several years of near-identical data, and fails to recognise the difference between correlation and causation, among other problems.

The authors also note that while it would be bad enough if this were a private study, undertaken with private funds, this a study requested, funded, reviewed, and edited by a government agency, the SBA’s Office of Advocacy. They argue that the Office of Advocacy’s sponsorship and official embrace of the study, including its defense in testimony before Congress even after it had been severely criticized, ‘embroils this public agency in an unwholesome blend of ineptitude and bias’. The agency should acknowledge the study’s many failings, they argue, and publicly disavow it.

Download the article (PDF, 278kb)