Determinants of error attribution in accounting estimates

Accounting researchers (e.g., Lundholm 1999) and standard-setters (e.g., AICPA 2002 and SEC 2002) have suggested that ex post reporting and examination of the accuracy of prior-period accounting estimates may increase investor welfare by increasing the perceived reliability of current-period accounting estimates. In this dissertation, I experimentally examine two main research questions related to decision makers' interpretations of prior-period accounting estimate accuracy disclosures: (1) Are decision makers' interpretations of such ex post reports on prior-period accounting estimate accuracy affected by the properties of the time series observed, and (2) Are decision makers' interpretations of such ex post reports susceptible to biases resulting from directionally motivated reasoning? I hypothesize that directionally motivated reasoning moderates decision makers' misconceptions of the properties of bias and noise in the observed sequences of accounting estimate errors, such that potential stockholders are more likely than other decision makers to attribute misestimations to bias, while current stockholders are more likely than other decision makers to attribute misestimations to noise, and that these differences will decrease as reasonableness constraints increase. Also, results suggest that when reasonableness constraints are low, potential stockholders are more likely than current stockholders to attribute misestimations to bias, while current stockholders are more likely than potential stockholders to attribute misestimations to noise, consistent with directionally motivated reasoning. Results from a second experiment show that increasing the number of observations to the maximum likely to be seen in a financial reporting context does not mitigate the effects observed in the first experiment.