Discussion of Detecting Earnings Management: A New Approach

Despite the large number of papers purporting to identify earnings management, the current techniques used to identify it are sorely lacking. In general, they are based on the assumption that accruals unexplained by a linear projection on firm-level observables (i.e., discretionary accruals) represent either explicit earnings management or poor quality earnings.1 Namely, these techniques assume that the residual from a linear regression represents earnings management. They suffer from measurement error and correlated omitted variables, which lead to Type 1 errors (i.e., rejection of a true null hypothesis of no earnings management) and Type 2 errors (i.e., failure to reject a false null hypothesis of no earnings management). Furthermore, they are typically estimated in the cross section and therefore do not take into account the fact that earnings (and its components) are best described by dynamic processes. These techniques have not