Should we burn the statistical significance tests?

Abstract

Inference plays a central role in management research as researchers are frequently led to draw conclusions or make generalizations from their observations or results. In many cases, they are able to do this rigorously through inferential statistics, which is the process of inference whereby the statistician tests the generalization of information collected in a sample to the entire population the sample is from. Statistical tests are thus at the heart of inferential statistics and, consequently, the process of inference. However, since they were first developed, statistical significance tests have been the object of sharp and repeated criticism regarding both their nature and their role (Nickerson, 2000). Such criticism has been longstanding in virtually all disciplines, with the notable exception of management that is just beginning to address the issue (Mbengue, 2007, Schwab & Starbuck, 2009). The main purpose of this paper is to provide researchers in management with clear information about the controversy surrounding statistical significance tests, to detail the content and issues and, most importantly, to offer recommendations for improving the testing of hypotheses and beyond, in other words, the process of statistical inference in management research.