Compares three definitions of earnings management used by accounting researchers and three methods of estimating it: aggregate accruals, specific accruals and discontinuities in earnings distribution. Discusses evidence relating to the reasons for income‐increasing earnings management, income‐decreasing earnings management and specific contexts, e.g. financial institutions with regulatory constraints. Concludes that, although the evidence is limited, managers are more likely to manipulate income up rather than down; and identifies some opportunities for further research.