Abstract: To estimate abnormal accruals, prior research employs a wide variety of models and estimation procedures. We evaluate the performance of three representative models – modified Jones model (MJ), MJ with operating cash flows (MJOCF), and MJ with return on assets (MJROA) – and two estimation procedures – industry-specific and firm-specific regressions. In evaluating the performance of various models, we use mispricing tests (i.e., relating current accruals to future returns) as our main test. We find that the best performing model is the firm-specific MJOCF model followed by the industry-specific MJROA model. However, when we use the recursive firm-specific procedure (i.e., based on current and previous years’ information only), we do not find the firm-specific MJOCF model outperforms the industry-specific MJROA model. We recommend that researchers use the industry-specific MJROA model to estimate abnormal accruals when investigating earnings management. For evaluating earnings quality that can include management estimation error, the firm-specific MJOCF remains clearly the best model from the perspective of abnormal accrual mispricing.