Beyond words: the challenge of measuring financial text across borders

Dictionary-based methods are widely used to assess the information content of corporate disclosures, yet they are primarily developed within the U.S. context. This study applies a popular U.S.-based dictionary approach to analyze the disclosure sentiment of 200,960 earnings call transcripts across 85 countries. While disclosure sentiment is positively associated with short-window stock returns globally, the relationship is approximately 30% weaker for non-U.S. firms. Drawing on cultural accounting theory, we find that cultural differences play a central role in explaining this disparity. In terms of practical implications, we find that sentiment-based trading strategies are effective for U.S. firms and for non-U.S. firms operating in culturally similar environments, but not for firms from culturally distant countries. To address potential endogeneity, we use linguistic features as instrumental variables reflecting historical cultural values. Additionally, we apply a machine learning model trained on U.S. data, which continues to uncover cultural limitations when used internationally. Our findings underscore the importance of institutional and cultural context in international textual analysis and highlight the limitations of applying U.S.-centric empirical methods globally. The development of culturally sensitive and contextualized empirical approaches has the potential to advance the measurement and understanding of financial communication in international settings.