Discretionary accounting choices: an information perspective

Frantz, Pascal (1994) Discretionary accounting choices: an information perspective. Doctoral thesis, University of London: London Business School.

Abstract

This thesis seeks to explain discretionary accounting choices in a world in which managers have discretion and do exercise discretion when reporting their firms' financial statements. It provides a review of the main perspectives adopted by the accounting literature on the discretionary reporting choices made by managers. It reviews both the papers explaining discretionary disclosures and the papers explaining discretionary accounting choices. Most of the accounting literature adopts an "opportunistic behaviour" perspective on accounting choices by assuming that managers use their discretion, at best, to attempt to fool outside claimholders on behalf of the existing shareholders, and, at worse, to increase their own welfare at the expense of the existing shareholders. However, a number of scholars have alluded to the potential for discretionary accounting choices to act as signals for the quality of the firm. This thesis provides the accounting literature with a number of signalling models, alluding to an "information perspective" and illustrating how accounting choices may enable the manager to reveal his private information concerning future earnings or cash-flows in a world characterised by an information asymmetry between the manager and the financial market. In each of these models, it is the adoption of conservative accounting policies, that is, the adoption of either income-reducing or balance sheet-weakening policies, which signals strength. This thesis finally provides a unifying treatment of the common features of the models and an exposition of the empirical implications of these models versus the main alternative categories of discretionary accounting choice models. It offers an interpretation of the implications of the existing empirical work for the relative merit of the various models.