The Hype Index: an NLP-driven Measure of Market News Attention

Zheng Cao, Wanchaloem Wunkaew, Hélyette Geman

TL;DR

This paper introduces the Hype Index, a novel NLP-driven metric that quantifies investor attention by measuring news volume rather than just sentiment tone. By adjusting news coverage against market capitalization, the researchers isolate "disproportionate" attention, providing a powerful new signal for market volatility and behavioral anomalies in large-cap equities.

What Problem Does This Paper Address?

In financial NLP, most researchers focus on sentiment analysis—determining whether news is "good" or "bad." However, the authors argue that the intensity of attention (the "hype") is a distinct and equally important signal.

A major challenge in measuring hype is "size bias": a massive company like Apple naturally generates more news than a smaller one. Without adjusting for economic footprint, it's impossible to tell if a stock is actually being "hyped" or if the news volume is simply commensurate with its size. The paper seeks to bridge this gap by creating a framework that identifies when media exposure deviates from a benchmark of proportionality.

Methods

The researchers developed two primary metrics:

  1. News Count-Based Hype Index: A "raw" measure that calculates a stock’s or sector’s share of total news articles within the S&P 100 universe on a given day.
  2. Capitalization Adjusted Hype Index (CapHype): The core innovation. This is the ratio of a stock’s news weight to its market capitalization weight.
  3. A value of 1.0 (Hype Neutrality) means news coverage is perfectly proportional to the company's size.
  4. A value > 1.0 indicates the asset is receiving "excess" attention relative to its economic importance.

The team utilized Natural Language Processing (NLP) via the LSEG (formerly Refinitiv) entity recognition system to map unstructured news headlines to specific tickers, ensuring high accuracy in news-to-firm mapping.

Datasets

  • Universe: Constituents of the S&P 100 index (as of March 2024).
  • News Data: Headlines and metadata from LSEG Data & Analytics, covering 326 business trading days from December 21, 2023, to April 10, 2025.
  • Market Data: Daily market capitalization, shares outstanding, and adjusted closing prices from Refinitiv Workspace.

Key Insights and Ideas

  • The "Big Tech" Paradox: In raw news volume, Information Technology is the most "hyped" sector. However, once adjusted for market cap, it actually falls into the "Less Prominent" category. This suggests that while we hear about Nvidia and Microsoft constantly, the volume is actually lower than expected given their multi-trillion-dollar valuations.
  • Real Estate is the Real Hype: Conversely, sectors like Real Estate and Utilities are "Relative Hyped." They have small market caps but receive a disproportionately high amount of news coverage relative to their size.
  • Non-Linear Scaling: The relationship between market cap and news volume isn't linear; it follows a power-law distribution ($R^2 = 0.631$). This suggests that as companies grow, the way the media scales its coverage of them changes fundamentally.
  • Hype as a "Fear Gauge": The CapHype Index moves in tandem with the VIX (Volatility Index). Spikes in adjusted hype often coincide with market "meltdowns" (like the August 5, 2024 crash), suggesting that "excess attention" is a precursor to or a real-time indicator of market instability.

Key Results

  • Sector Dominance: Financials and IT dominate raw news attention, each holding roughly a 19% share of all S&P 100 news.
  • Adjusted Rankings: In the CapHype Index, Real Estate leads with an average score of 3.12, followed by Industrials (2.24). IT sits at the bottom with 0.51, indicating it is "under-covered" relative to its massive economic weight.
  • Statistical Properties: Changes in the Hype Index are not normally distributed. Normality tests (Shapiro-Wilk, Jarque-Bera) yielded p-values near zero, indicating "fat tails"—meaning extreme hype events happen much more often than a standard bell curve would predict.
  • Correlation: There is a very high correlation (up to 0.98 in Communication Services) between the raw Hype Index and the Adjusted Index, suggesting that when market caps are stable, news volume alone is a reliable proxy for adjusted attention.

Limitations and Future Work

  • Proprietary Mapping: The study relies on LSEG’s proprietary entity recognition. While accurate, it acts as a "black box" for the underlying mapping logic.
  • Sentiment Integration: While this paper focuses on volume, the authors note that combining Hype (intensity) with Sentiment (tone) is the next logical step for predictive modeling.
  • Stochastic Modeling: The authors suggest that future research could treat the Hype Index as a stochastic process (like interest rates or volatility) to enhance option pricing models and risk management tools.
  • Causality: The paper establishes correlation with market moves but notes that more work is needed to formally assess the "predictive utility" and lead-lag relationships between hype spikes and price reversals.

Summary updated

Abstract

This paper introduces the Hype Index as a novel metric to quantify media attention toward large-cap equities, leveraging advances in Natural Language Processing (NLP) for extracting predictive signals from financial news. Using the S&P 100 as the focus universe, we first construct a News Count-Based Hype Index, which measures relative media exposure by computing the share of news articles referencing each stock or sector. We then extend it to the Capitalization Adjusted Hype Index, adjusts for economic size by taking the ratio of a stock's or sector's media weight to its market capitalization weight within its industry or sector. We compute both versions of the Hype Index at the stock and sector levels, and evaluate them through multiple lenses: (1) their classification into different hype groups, (2) their associations with returns, volatility, and VIX index at various lags, (3) their signaling power for short-term market movements, and (4) their empirical properties including correlations, samplings, and trends. Our findings suggest that the Hype Index family provides a valuable set of tools for stock volatility analysis, market signaling, and NLP extensions in Finance.
BibTeX
@misc{cao2025hype,
  author = {Zheng Cao and Wanchaloem Wunkaew and Hélyette Geman},
  title = {{The Hype Index: an NLP-driven Measure of Market News Attention}},
  year = {2025},
  eprint = {2506.06329},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2506.06329},
}
Figure