Attrition bias and inferences regarding earnings properties
51 Pages Posted: 21 Sep 2017 Last revised: 8 Jul 2021
Date Written: August 18, 2020
On average, across the years 1980 to 2019, almost 8.4 percent of firms on the Compustat Annual data set, which had earnings and return observations in year t-1, did not have earnings observations in year t. Because these disappearances were not random, there is attrition bias in estimates of fundamental properties of earnings including: earnings persistence, mean reversion, accuracy of earnings forecasts, asymmetric timeliness, and market pricing of earnings persistence. We show that conclusions about these properties change when we impute the disappeared earnings observations using four parsimonious methods and when we use inverse probability weighting in least squares regressions.
Keywords: earnings, forecasts
JEL Classification: G14
Suggested Citation: Suggested Citation