Context-Based Interpretation of Financial Information

Does the context surrounding the numbers in financial statements help investors to interpret these numbers? Answering this question empirically presents a major challenge. We leverage recent advances in natural language processing to study the role of context in the interpretation of numeric information. When the contextual information surrounding earnings disclosure is taken into account, we show that earnings become more informative about a firm's future prospects. We also show that the contextuality of earnings, i.e., the importance of the earnings context for their interpretation, has been increasing over time. Finally, we use earnings context to identify rich firm-year heterogeneity in earnings persistence, which is of interest as a firm-year measure of earnings quality.