Heterogeneous sentiment contagion among firms during financial crisis

Abstract

We study the spread of sentiment among firms during the Silicon Valley Bank (SVB) crisis using a heterogeneous staggered difference-in-differences model, where sentiment is measured via dictionary-based analyzers. We merge our constructed sentiment datasets with stock and firm information from the Center For Research in Security Prices (CRSP) database, identifying the treatment timing for a firm as the date when it is first co-mentioned with SVB in a news article during the crisis period. Our baseline results reveal a significant negative heterogeneous treatment effect among treated cohorts. Notably, the earliest treated cohort exhibits a lag in information processing, whereas later cohorts do not. The results are robust to two alternative sentiment measures at an aggregated level, but less so at the cohort-specific level. Further heterogeneity analysis by partitioning samples into subgroups based on firms’ characteristics reveals that firms who have prior connection to SVB experienced the largest shift in their news sentiment. Our findings suggest the necessity for a timely intervention during the delayed sentiment reaction period and a more tailored communication strategies for firms with prior connections to shock originator (SVB) from policymakers and regulators.