Debiasing Earnings Persistence Estimates

We offer a theoretical framework to help isolate persistence estimates of fundamental earnings innovations from the effects of accounting measurements. We show that a downward bias results when persistence of earnings innovations is estimated using reported earnings. We show that the greater the accrual estimation errors, the greater this downward bias, thus explaining the empirically observed positive association between accrual quality and estimated earnings persistence. However, when we debias reported earnings persistence as guided by our theoretical framework, we fail to detect any such association. We further show that market returns around earnings announcements are associated with the debiased measure of earnings persistence incremental to reported earnings persistence indicating that the market is able to undo the bias in a manner consistent with our theoretical model. Overall, our results help bring clarity to the literature by pointing out that accrual quality per se does not bear any relation to persistence of earnings innovations; rather, improving accrual quality merely improves the precision in estimating the persistence of these innovations