Understanding How the Effects of Conditional Conservatism Measurement Bias Vary with the Research Context

While the asymmetric timeliness (AT) measure of Basu (1997) underpins a large body of empirical research on conditional conservatism (CC), prior studies have demonstrated that it is biased and could lead to Type 1 error. To assess how this bias could affect inferences from prior literature, we re-examine previous CC studies that apply the AT measure, and we compare the outcomes with those based on the asymmetric conditional variance (ACV) measure of Dutta & Patatoukas (2017) confirmed to be less influenced by similar bias. We draw two primary conclusions. First, both measures yield similar inferences in interrupted time-series settings that examine the impact of exogenous accounting policy changes on CC. Second, the inferences drawn from the AT measure are not supported by the ACV measure in seminal studies that model the determinants of CC in cross-sectional settings. In addition, we find that the conflicting results between the AT and ACV measures in the cross-sectional settings disappear when the AT measure is allowed to vary across firms. Our findings have implications for both past and future empirical studies of CC