Affiliations: School of Business, Economics and Public Policy, University of New England, Armidale, Australia. E-mail: Lansun@une.edu.au
Abstract: Empirical studies have documented various approaches in detecting earnings management behaviour. Since the middle 1980s, the accrual approach has become the primary focus in the literature as the accrual is suggested as a desirable vehicle to achieve managerial manipulation and it is less likely to be detected [Journal of Accounting and Economics 18 (1994), 3–42; Journal of Accounting and Public Policy 19(4,5) (2000), 313–345]. However, the accruals approach and its associated varies models in generating a valid measurement of earnings management are often criticized. A long-standing issue lies with the model misspecification where variables that explain non-discretionary accruals have been omitted from the expectation models and so wind up on the residual term, which represents the management manipulation component of discretionary accruals. As a consequence, empirical evidences on earnings management often result in misleading inferences about earnings management behaviour. Given the potential for model misspecification in measuring earnings management, this research aims to examine the validity of alternative earnings management measurement and proposes a new aggregate measure of Principal Component (PC) to be used in detecting earnings management behaviour. The principal components that explain the highest variation among accounting variables are expected to capture earnings management behavior and reveals hidden dynamics of financial reporting without model misspecification constrains.
Keywords: Earnings management, principal component analysis, model misspecification