A Theoretical and Empirical Analysis of Discretionary Accrual Estimation Models in Earnings Management Research

Abstract

The main goal of this research is to verify the theoretical and empirical validity of operational models to measure discretionary accruals, which are used in the detection of accounting earnings management. Initially, a theoretical-critical analysis is presented about the specification of the operational models present in literature for discretionary accrual estimation. Based on observations about the construction of analytic-empirical models and empirical evidence from earlier studies, a new model was proposed to analyze accruals behavior and estimate discretionary accruals. This empirical-analytical study also uses bibliographic and descriptive research. The sample in this study is constituted by the group of publicly-traded companies in the North-American and Latin-American capital markets, covering the period between 1996 and 2005. The models’ specification and predictive power is analyzed through different statistical procedures. All operational discretionary accrual measurement models used to detect accounting earnings management are estimated through the pooling of independent cross sections approach for all economic environments. The model with the best predictive power is chosen by the adjusted R2 analysis, Akaike and Schwarz criteria and Voung’s Test. The results of this research suggest that the operational discretionary accrual estimation models present in current literature do not present adequate theoretical foundations and that some of these models are weakly specified, have low predictive power and are significantly affected by the economic environment. Moreover, the results prove that the operational model proposed in this study to estimate discretionary accruals for earnings management has greater power to explain accrual behavior in all economic environments under analysis.