Entropy-Balanced Discretionary Accruals
69 Pages Posted: 29 Jan 2015 Last revised: 5 Mar 2018
Date Written: Feb 27, 2018
To estimate discretionary accruals, we employ a recently developed multivariate matching approach (entropy balancing) to adjust for underlying accrual determinants at the sample-level instead of relying on a first-stage linear determinants model. Entropy balancing identifies weights for each control sample observation to equalize the mean, variance, and skewness of underlying determinants across treatment and control samples. Addressing covariate imbalance using either propensity score matching (PSM) or entropy balancing significantly improves discretionary accrual model specification in samples of extreme financial performance. However, we find that entropy balancing enhances test power over PSM in simulations with seeded accrual management. To empirically validate these improvements, we examine the accruals of firms issuing new equity as a setting in which differences in growth and financial constraint across treated and control samples are particularly egregious. In contrast to existing discretionary accrual measures, we find that combining entropy-balanced discretionary accruals with tests for accrual reversals results in an inference of insignificant (significant upward) accrual earnings management in the year of an initial public (seasoned equity) offering, consistent with the stringent reporting requirements and investor scrutiny of initial public offerings limiting earnings management.
Keywords: discretionary accruals; entropy balancing; propensity score matching; covariate balance; initial public offerings; seasoned equity offerings
JEL Classification: M04
Suggested Citation: Suggested Citation