Econometricians have been claiming proudly since World War II that significance testing is the empirical side of economics. In fact today most young economists think that the word “empirical” simply means “collect enough data to do a significance test”. Tjalling Koopmans’s influential book of 1957, Three Essays on the State of Economic Science, solidified the claim. A century of evidence after Student’s t-test points strongly to the opposite conclusion. Against conventional econometrics we argue that statistical significance is neither necessary nor sufficient for proving commercial, human, or scientific importance. A recent comment by Thomas Mayer, though in parts insightful, does nothing to alter conclusions about the logic and evidence which we and others have assembled against significance testing. Let’s bury it, and get on to empirical work that actually changes minds.
This article is a response to Ziliak and McCloskey’s Criticisms of Significance Tests: An Assessment by Thomas Mayer (EJW, September 2012).
Response to this article by Thomas Mayer: Reply to Deirdre McCloskey and Stephen Ziliak on Statistical Significance (EJW, January 2013).