Textual Analysis of Short-seller Research Reports

Abstract

Using survey cash-flow expectations, we find that investors underreact to negative cash flow news included in short-seller reports. On average, target firms earn abnormal returns of -4.9% on the publication day, and subsequent price revisions equal -18% over the next 12 months. Using large language models, we introduce a novel text-based fraud measure and find that reports more related to fraud predict larger negative long-term abnormal returns. Other allegations, such as claims of overvaluation, do not have predictive power. Furthermore, short-seller research reports predict significant reductions in future real investment and stock issuances, with those heavily emphasizing fraud predicting larger declines. A model featuring expectation stickiness combined with limits-to-arbitrage is consistent with the long-term price impacts of short-seller reports.

Keywords: Short-seller Research, Real Investment, Asset Pricing, Real Impact of Financial Markets, Analysts Expectations