Estimating the Permanent Component of Accounting Earnings Using the Unobservable Components Model: Implications for Price-Earnings Research

Abstract

This study proposes the unobservable components (UC) model as a conceptually and statistically superior approach to modeling the stochastic properties of accounting earnings. Based on the UC model, accounting earnings are decomposed into permanent and transitory components and the earnings-return relation using these components is reexamined. Once the transitory component is eliminated from accounting earnings, the average firm-specific earnings response coefficient (ERC) estimates from regressing annual returns on earnings changes increase more than 300 percent—from 2.192 to 7.770. Using a 10-year returns model, the study documents that the stock prices in the long run predominantly reflect the permanent component of earnings. Results also suggest that the success of aggregate earnings in explaining long-window returns is due to the aggregate earnings proxying for the permanent component of earnings. Finally, the study provides evidence suggesting that no simple income statement classification scheme can effectively isolate the transitory component of earnings.