A0602
Title: Words matter: Forecasting economic downside risks with corporate textual data
Authors: Cansu Isler - Brandeis University (United States) [presenting]
Abstract: Forecasting downside risks to economic growth is increasingly critical for policymakers and financial institutions. A novel daily sentiment indicator is constructed using textual analysis of corporate disclosures from SEC 10-K and 10-Q filings. Firm-level sentiment is measured as the year-over-year change in the tone of forward-looking statements, based on counts of positive and negative words from the Loughran-McDonald dictionary. These firm-level scores are aggregated into a weekly sentiment index, weighted by market capitalization to reflect broader economic signals. The index is integrated into a mixed data sampling (MIDAS) quantile regression framework to forecast lower quantiles of US GDP growth. Results show that this sentiment-based indicator significantly improves the prediction of economic downturns, outperforming traditional metrics such as the national financial conditions index (NFCI). The approach offers a timely and interpretable measure of market sentiment drawn directly from firms' own assessments of risk and outlook. These findings underscore the potential of corporate textual data as a forward-looking tool for macroeconomic surveillance and policy design.
