Is There a Confidence Interval for That? A Critical Examination of Null Outcome Reporting in Accounting Research

Abstract

This study evaluates how accounting researchers analyze and report null outcomes based on an examination of recent accounting research publications. As null outcomes reflect failures to provide compelling evidence against a proposed null hypothesis, in and of themselves they have little inferential value. Nevertheless, we find that articles routinely interpret such outcomes in a highly conclusive manner, using terms such as: did not, is no difference, find no effect, equals, no association, etc. We also find little evidence that articles subject the available evidence to further statistical examinations that might provide relevant insights about the studied phenomenon. The analysis also illustrates the use of confidence intervals as a useful approach for providing insights about the location, inconsequentiality, and associated estimation precision of studied relations in null outcome settings.