Financial News Impact: Rumors Drive Prices, News Confirms Volatility

Alireza Kargarzadeh, Nariman Khaledian, Navid Parvini, Sid Ghatak, Arman Khaledian· August 17, 2026 View original

Key takeaways

  • Most price movement related to news happens before or at publication, driven by rumors.
  • Markets underreact to quantitative news, causing prolonged drift.
  • Markets overreact to qualitative news, often leading to price reversals.
  • News increases volatility before publication, which then declines.

Who benefits

Financial ServicesInvestment ManagementAlgorithmic TradingFinTech

Summary

A study analyzing 4.57 million financial news articles found that price movements associated with news largely occur before or at publication, with rumors capturing the entire move. Markets underreact to quantitative news, leading to drift, while overreacting to qualitative stories, which often reverse.

A comprehensive study examined 4.57 million financial news articles related to approximately 3,000 US stocks to understand how quickly markets absorb public information. The research utilized a large language model to categorize articles by event tags and attributes, clustering them into stories to differentiate initial reports from follow-up coverage. The findings confirm that the significant price movements linked to news largely concentrate before and at the time of publication. For events flagged as rumors, the entire price shift occurs on the rumor day, with subsequent official news contributing no further movement. This supports the adage "buy the rumor, sell the news." Furthermore, the study revealed differential market reactions based on news type. Markets tend to underreact to quantifiable fundamental news (e.g., earnings, dividends), leading to a prolonged price drift in the news direction. Conversely, they overreact to qualitative, story-driven news (e.g., product launches, leadership changes), which often sees its initial gains reversed. News also increases volatility before publication, which then subsides as uncertainty is resolved.

Why it matters

Financial professionals and quantitative traders can gain a deeper understanding of market efficiency and information absorption, informing their trading strategies and news-conditioned forecasting models.

How to implement this in your domain

  1. 1Develop news-conditioned trading strategies that prioritize acting on early signals or "rumors" rather than waiting for official announcements.
  2. 2Differentiate between quantitative fundamental news and qualitative story-driven news in your analysis, adjusting expectations for price drift or reversal accordingly.
  3. 3Incorporate pre-publication volatility increases into risk management models for stocks frequently covered by news.
  4. 4Utilize the provided event tag drift table as a prior for building or refining news-conditioned forecasting models.
  5. 5Design systems to quickly process and categorize financial news using LLMs to capitalize on the rapid market reaction window.

Original post by Alireza Kargarzadeh, Nariman Khaledian, Navid Parvini, Sid Ghatak, Arman Khaledian

"arXiv:2608.14014v1 Announce Type: new Abstract: Two old market sayings hold that news is already priced in by the time it is published, and that the rumor is bought while the news is sold. Both place the price move associated with a piece of news before and at publication rather…"

View on X

Originally posted by Alireza Kargarzadeh, Nariman Khaledian, Navid Parvini, Sid Ghatak, Arman Khaledian on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Investing

AI Engineering & DevToolsAI ResearchAI Investing

MINT Enables Zero-Shot Prediction for Transaction Data

Researchers introduce MINT (Multimodal Instruction Network for Transactions), a framework connecting a transaction sequence encoder to an LLM for zero-shot prediction on financial transaction data. MINT achieves state-of-the-art performance, reducing input tokens, latency, and memory compared to text serialization methods.

Parameswaran Kamalaruban, Viktor Drobnyi, Maeve Madigan, Julia Rozanova, David Sutton, Stuart BurrellAug 17, 2026
AI ResearchAI Investing

Time-Series Models Face "Forecast Collapse" in Finance

This research identifies "forecast collapse" in time-series foundation models (TSFMs) when predicting hourly equity returns, where predictions become flat and show poor stock ranking. The phenomenon is linked to low target predictability and per-series objectives, leading to a calibration-ranking tradeoff, which a new objective called CalibRank aims to balance.

Shu Wan, Miles Ma, Hank Zhu, Guangqi Liu, Stephen Wang, Qingsong Wen, Huan LiuAug 17, 2026
AI News & ToolsAI Engineering & DevToolsAI Investing

Language Models Evolve: Cost Drops, Specialized Agents Rise

Over eight years, AI language models progressed from BERT to advanced agents, with coding capabilities improving sixfold annually since late 2024. Costs have dramatically decreased, with new budget models matching flagship performance at a fraction of the price, leading to a rise in specialized models excelling in specific tasks.

Pranav Kumar KaliaperumalAug 17, 2026