On Estimating Conditional Conservatism

The concept of conditional conservatism has provided new insight into financial reporting and has stimulated considerable research since Basu (1997) developed it. While the concept encapsulated in the adage “anticipate no profits but anticipate all losses” is reasonably clear, estimating it is the subject of some discussion, notably by Dietrich et al. (2007), Givoly et al. (2007), and Ball, Kothari and Nikolaev (2011). Recently, Patatoukas and Thomas (2011) report important evidence of possible bias in firm-level cross-sectional estimates of conditional conservatism (asymmetric earnings timeliness) which they attribute to scale effects. They advise researchers to avoid using conditional conservatism estimates or making inferences from prior research using them, a view we regard as excessively alarmist. Our theoretical and empirical analyses suggest the explanation is a correlated omitted variables problem that can be addressed in a straightforward fashion, for example by fixed-effects regression. We show that cross-sectional correlation between the expected components of earnings and returns confounds the relation between the news components, and biases estimates of how earnings incorporates the news in returns (e.g., timeliness). We also show that the correlation between the expected components of earnings and returns depends on the sign of returns, biasing estimates of asymmetric timeliness. When firm-specific effects in earnings are taken into account, estimates of asymmetric timeliness do not exhibit the bias, are statistically and economically significant (though smaller in magnitude and perhaps more consistent with priors), and behave as a predictable function of market-to-book, size and leverage. It would be surprising if this was not the case. Conditional conservatism accords with the long-standing accounting principle of anticipating losses but not gains, with specific asymmetric accounting rules such as the lower-of-cost-or-market method for inventories and the rules for impairment of long term assets, and with loss recognition practices that occurred prior to the promulgation of formal rules