Impression Management: AI-based Evidence from Earnings Guidance
A study of over 18,000 earnings guidance press releases from 3,600 US firms reveals companies use upbeat headlines to influence investors, even when the news is negative. Using FinBERT, an AI language model, researchers found headlines are more positive than the body text, especially with bad news. This impression management affects stock prices and trading activity. The positive tone is strongest at the start of releases, a pattern unique to company disclosures, not media or AI-generated headlines.
What it examines
This study uses AI (FinBERT) to analyze the tone and sentiment in over 18,000 earnings guidance press releases from 2001 to 2022. It investigates whether companies use positive language in prominent parts of disclosures to influence investor perceptions, focusing on differences between titles and the main text.
What it concludes
The research finds firms use positive sentiment in key parts of disclosures to manage impressions, affecting stock prices and trading. These findings help investors and regulators spot impression management. Future work could improve AI tools for financial analysis and explore how investors respond to such language strategies.
Evidence objects
Companies often use upbeat headlines in earnings guidance press releases to shape investor perceptions, even when the underlying news is negative, revealing a strategic approach to impression management in corporate communications.
key_findings bullet 1 · key_findings · validation V0
Analyzing over 18,000 press releases from 3,600 firms using FinBERT AI, researchers found headlines are consistently more positive than the body text, especially with bad newsa pattern unique to company disclosures.
key_findings bullet 2 · key_findings · validation V0
Headline sentiment significantly influences stock prices and trading activity beyond the full text, showing investors react to these cues. The studys large dataset and AI methods highlight subtle tactics, but focus mainly on U.S. firms.
key_findings bullet 3 · key_findings · validation V0
This paper uniquely examines sentiment differences between titles and body text in earnings guidance using FinBERT, linking impression management to investor reactions. While not groundbreaking in methodology or predictive modeling, its moderately novel focus offers valuable insights for those interested in AI, sentiment analysis, and market microstructure within financial disclosures.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
These features of the title have implications for both firms and investors. First, they make the title a potentially attractive tool for firms to manage
Source row: 1074 · abstract type: snippet