This paper studies the reliability of financial reporting when the credibility of the manager, represented by his misreporting propensity, is unknown. We show that credibility concerns affect the time-series of reported earnings, book values, and stock prices in ways that seem consistent with empirical evidence. When investors are uncertain about the credibility of the reporting process, earnings response coefficients, as well as market-to-book values (MTB), are random and time-varying; relatively low MTB reflect poor credibility of financial reporting; stock prices are s-shaped in earnings surprises and relatively insensitive to bad news. Finally, when the manager is more likely to have reporting discretion, discretionary accruals tend to be larger and more volatile. We estimate the model using U.S. earnings announcement data during 2002–2012 and find that the probability of misreporting is 7 percent. A counterfactual analysis reveals that ignoring the possibility of misreporting leads to overestimation of the mean (3.5 percent), volatility (13 percent), and persistence of earnings (17 percent).
JEL Classifications: D82; D83; D84.