Unraveling Managerial Tangents in Firm Disclosure: Concealing Issues or Being Exposed?

Abstract: Earnings calls influence stock prices and are traditionally analyzed using sentiment and linguistic traces. Our research introduces a "Topic-Switching Index," a novel metric quantified through the transformer model FinBERT, to measure managerial evasion during Q$&$A sessions in earnings calls. We find a negative correlation between this index and subsequent stock prices, indicating that investors penalize managerial evasiveness. This study is the first to quantify such evasive tactics, adding a new dimension to how earnings calls are understood and suggesting that topic shifting is an overlooked but significant factor. We also show the predictability of the index under three different classifier models and it stands out in all circumstances.