Testing for conditional conservatism - or not

Following Basu’s (1997) influential paper many widely cited studies have used linear piecemeal regressions of net income modelled by stock returns, split between negative and positive returns, to identify conditional conservatism. However, I show that the convex relationship between stock returns and net income will lead to results that are consistent with conditional conservatism whether it exists or not and regularities in the data related to low price stocks also introduce apparent conditional conservatism where it doesn’t exist. Taken together these results suggest that investigations of conditional conservatism using linear piecemeal regressions should be treated with great caution.