ABSTRACT A vast literature following Hayn [1995] and Burgstahler and Dichev [1997] attributed the so‐called “discontinuities” in earnings distributions around zero to earnings management. Despite recent evidence that these discontinuities are likely caused by other factors, researchers and teachers continue to point to the shapes of these distributions as evidence of earnings management. We provide three sets of further evidence that these discontinuities are likely caused by factors other than earnings management: (1) we provide, as an example, a detailed analysis of the severe effects of sample selection in a recent study; this study erroneously concludes that the shape of an earnings distribution is evidence of earnings management, (2) we provide a simple explanation for the shape of the earnings distribution that is most often cited as evidence of earnings management; the relation between earnings and prices differs with the magnitude and the sign of earnings, and (3) we provide further examples that support the main point of our paper; evidence beyond the mere shape of a distribution must be brought to bear before researchers can draw conclusions regarding the presence/absence of earnings management.

We provide evidence that the shapes (particularly around zero) of the frequency distributions of earnings metrics examined in the extant earnings management literature are affected by (1) deflation (using, for example, price or market capitalization), (2) sample selection criteria that lead to differential inclusion/exclusion of observations to the left of zero versus observations to the right of zero (implicit in studies focusing on firms followed by I/B/E/S and explicit in studies partitioning on a variable differing between loss observations and profit observations), (3) differences between the characteristics of observations to the left of zero and observations to the right of zero (such as market pricing and analyst optimism/pessimism), or (4) a combination of these factors. Since the shapes of the frequency distributions of earnings metrics at zero are likely due to one of the above effects, we conclude that the shapes cannot be used as ipso facto evidence of earnings management.