Asymmetric timeliness tests of accounting conservatism

Recent accounting research employs an asymmetric timeliness measure to test the hypothesis that reported accounting earnings are "conservative." This research design regresses earnings on stock returns to examine whether "bad" news is incorporated into earnings on a more timely basis than "good" news. We identify properties of the asymmetric timeliness estimation procedure that will result in biases in the test statistics except under very restrictive conditions that are rarely met in typical empirical settings. Using data series that are devoid of asymmetric timeliness in reported earnings, we show how these biases result in evidence consistent with conservatism. We conclude that the biased test statistics inherent in the asymmetric timeliness research design preclude using this method to measure conservatism; that these biases are irresolvable as they originate in the test's specification; and that studies employing asymmetric timeliness tests cannot be interpreted as providing evidence of conservatism.