Toward a Positive Theory of Information Evaluation: Relevant Tests of Competing Models in a Principal-Agency Setting.

Abstract ABSTRACT: This paper reports the results of an experiment to test four models of the principal's information evaluation behavior in a private, pre-decision, principal-agency setting, The four models tested were an expected utility model, a prospect theory model, a linear model and a multiplicative model. In response to recent criticism that tests of positive theories in accounting lack sufficient power to discriminate among competing theories, the tests of the models were designed to disconfirm the models' predictions without assuming initial conditions. The model which best resisted the attempts to falsify it was the multiplicative model. The results demonstrate the generalizability of theories of information evaluation behavior from an information evaluator-decision maker setting to a principal-agency setting and show that the descriptive validity of principal-agency theory can be enhanced by incorporating models of individual behavior other than those based on expected utility theory. The study also provides a rigorous test of prospect theory and offers guidance for the future development of this theory.