Analysis of Diagnostic Tasks in Accounting Research Using Signal Detection Theory

Many accounting judgments are diagnostic tasks in which accountants, auditors, managers, or investors discriminate among possible states and decide which one exists. To measure the accuracy of such decisions, most accounting research employs percentage correct, a measure proven to be invalid and unreliable, primarily because it does not control for response bias. This paper describes Signal Detection Theory (SDT), a theoretical model of diagnostic tasks that has been supported empirically in many fields. SDT provides superior measures of accuracy and response bias. We discuss the benefits of employing SDT in accounting research and describe an SDT-based reanalysis of data related to two published accounting studies that results in revised conclusions and important additional insights.