Abstract The article informs that behavioral research in accounting simultaneously must consist of theory-building and methodology refining. The problems of identifying significant variables and generalizing from experimental findings are inextricably linked. These joint problems are generally acknowledged, widely deplored and little studied. This article reports the results of a lab experiment testing the effects and interactions of three factors--subject type, consistency of information, and earnings per share trend--imbedded in the context of a financial analysis task. Multiple dependent variables were employed so as to obtain data about both the observable outputs and the non-observable process used by subjects in arriving at their predictions. The findings of the experiment have several interpretations. On a substantive level, they confirm some of the conventional wisdom about the use of accounting information, occasionally display counter-intuitive results, and provide insights into the relative impact of different elements of the traditional annual report.
Thomas R. Hofstedt (Sun,) studied this question.