We study a model of earnings management and provide predictions about the time-series properties of earnings quality and reporting bias. We estimate the model to empirically separate two components of investor uncertainty: fundamental economic uncertainty, and information asymmetry between the manager and investors due to reporting noise. We find that (1) the null hypothesis of zero reporting bias is rejected; (2) the ratio of the variance of the noise introduced by the reporting process to the variance of earnings shocks is, on average, 45 percent; (3) the reporting noise plays a significantly less prominent role in valuation, due to the persistence of shocks to economic earnings; (4) the magnitude of investors' uncertainty created by reporting noise about firms' assets in place and about future earnings is similar; and (5) ignoring the possibility of reporting distortions would bias the estimates of variance and persistence of economic earnings.
