News Dispersion, Asymmetric Verification and Conservatism

Asymmetric verification takes into account both the type of news (i.e., good versus bad news) and the level of uncertainty regarding that news. Good news should be more certain (i.e. more verifiable) before it is reflected in earnings than bad news. To test this principle, we use data from Thomson Reuters News Analytics to construct a measure of uncertainty based on news articles on public firms. These articles are measured at the firm level and are assigned positive, negative, and neutral scores that sum to a total of one. If the positive score is largest, then the article is coded as good news (bad news articles are coded similarly). We then create a dispersion measure for each article where higher numbers represent more polarized news. Overall, we find that good news is more highly associated with earnings when uncertainty is low consistent with asymmetric verification. In contrast, we find no such association between bad news and uncertainty (as we would expect)