Abstract Accrual models have been central to the earnings management literature to measure the extent of earnings smoothing, yet consensus on the relationship between earnings and discretionary accruals remains limited. This systematic review investigates the statistical relationship between discretionary accruals and earnings to identify which accrual types are used to smooth earnings. In total, the study analyzes 85 accrual regression models reported in 13 accrual-modeling studies that report both an earnings proxy and discretionary accruals in the same regression model. The results reveal a mixed relationship between discretionary accruals and earnings due to the diverse nature of discretionary accrual types, indicating that all types of accruals are used to manage earnings. The results indicate that the modified Jones model of Dechow et al. (1995) and the performance matched model of Kothari et al. (2005) are most frequently used, whereas more recent work highlights the advantages of cash-flow-based models (e.g., McNichols 2002) in addressing accrual reversals and revenue-related discretionary accruals. The review concludes that methodological differences across all residual-based accrual models are small and that all models remain vulnerable to similar conceptual limitations. Future research should prioritize models that analyze accrual components directly rather than relying on residuals as proxies.
Heikki Häkkinen (Wed,) studied this question.