摘要：
This study investigates the importance of accounting for the sign of earnings as well as disaggregating earnings in empirical value relevance research. The paper presents evidence that value relevance as measured by the explanatory power of regression analysis more than doubles if both the sign and the disaggregation effect are incorporated into the analysis. Thus, traditional value relevance regressions may seriously understate the value relevance of accounting information. However, value relevance is not equally underestimated across sub-samples. Hence, the conclusions of prior studies that have compared value relevance between sub-samples from different time-periods, industries, countries, etc. may be biased.
