Paper: arXiv 2510.16503

Authors: Domenica Mino, Cillian Williamson

Abstract

Artificial intelligence techniques have increasingly been applied to understand the complex relationship between public sentiment and financial market behaviour. This study explores the relationship between the sentiment of news related to the Russia-Ukraine war and the volatility of the stock market. A comprehensive dataset of news articles from major US platforms, published between January 1 and July 17, 2024, was analysed using a fine-tuned Bidirectional Encoder Representations from Transformers (BERT) model adapted for financial language. We extracted sentiment scores and applied a Generalised Autoregressive Conditional Heteroscedasticity (GARCH) model, enhanced with a Student-t distribution to capture the heavy-tailed nature of financial returns data. The results reveal a statistically significant negative relationship between negative news sentiment and market stability, suggesting that pessimistic war coverage is associated with increased volatility in the S&P 500 index. This research demonstrates how artificial intelligence and natural language processing can be integrated with econometric modelling to assess real-time market dynamics, offering valuable tools for financial risk analysis during geopolitical crises.

Complexity vs Empirical Score

  • Math Complexity: 7.0/10
  • Empirical Rigor: 6.0/10 (capped from 8.5: no out-of-sample test; see penalty flags)
  • Quadrant: Holy Grail — high math complexity, high empirical rigor (effective empirical rigor capped from 8.5)

Why this score: The paper employs advanced mathematical concepts like BERT embeddings, GARCH modeling with Student-t distributions, and statistical hypothesis testing, while demonstrating strong empirical rigor through a comprehensive dataset from a defined period (Jan-July 2024), specific S&P 500 index analysis, and statistically significant results.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Analyze Sentiment-Volatility Link<br>during Geopolitical Crises"] --> B
  subgraph B ["Data Collection & Preparation"]
      B1["News Articles<br>Jan-Jul 2024"]
      B2["S&P 500 Market Data"]
  end
  B --> C["Sentiment Analysis:<br>Fine-tuned Financial BERT Model"]
  C --> D["Volatility Modeling:<br>GARCH with Student-t Distribution"]
  D --> E{"Statistical Analysis"}
  E --> F["Key Finding:<br>Negative Sentiment<br>Significantly Increases<br>Market Volatility"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes integrating a fine‑tuned BERT sentiment model with a Student‑t GARCH framework to assess the impact of news sentiment on market volatility.

Source sentence
By integrating NLP techniques like the BERT transformer model with a Student‑t GARCH econometric framework, the research d…

Assumes 3

  • The error term ϵt is assumed to follow a normal distribution. model
    Source sentence
    ϵt is the error term assumed to follow a normal distribution
  • Residuals are modeled with a Student‑t distribution to capture heavy tails. statistical
    Source sentence
    we opted to model the residuals using the Student‑t distribution, which accounts for the heavier tails commonly seen in financial returns
  • Sentiment scores derived from news headlines represent investor sentiment affecting volatility. data
    Source sentence
    negative sentiment extracted from major U.S. news headlines showed a statistically significant effect on volatility

Data

  • Goperigon news headlines · Goperigon text-news public
    2024-01-01 to 2024-07-17 · major U.S. news sources (≈10,000 headlines)
    Source sentence
    from January 1 to July 17, 2024, collecting over 10,000 headlines from major newspapers
  • S&P 500 daily returns market-prices public
    2024-01-01 to 2024-07-17 · S&P 500 index
    Source sentence
    Yt represents the daily returns of the S&P 500 index at time t

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

  • Negative news sentiment is significantly associated with higher S&P 500 volatility (negative coefficient). statistical
    Source sentence
    negative sentiment extracted from major U.S. news headlines showed a statistically significant effect on volatility
  • The VIX coefficient is statistically significant and negative, confirming its role as a volatility predictor. statistical
    Source sentence
    the VIX coefficient was statistically significant and negative
  • Traditional macro indicators (bond yields, OFR, EPU) are not statistically significant in explaining volatility. statistical
    Source sentence
    traditional financial indicators such as bond yields, the Financial Stress Index (OFR), and the Economic Policy Uncertainty Index (EPU) were not found to be
  • The Student‑t GARCH model improves robustness and captures heavy‑tailed return distributions during crises. statistical
    Source sentence
    The empirical performance of the GARCH model with a Student‑t distribution further supports the view that financial returns during crises are prone to heavy tails and extreme deviations
  • BERT sentiment extraction outperforms traditional dictionary‑based methods for capturing nuanced geopolitical narratives. other
    Source sentence
    ... In contrast, as a transformer‑based model, BERT captures semantic nuance, word order, and bidirectional context

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 34 related papers). Coverage is partial, so this is neither replication nor refutation.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (41)

  1. Estimation and Prediction for Algorithmic Models of Investor Behavior (2022)
  2. Who writes the news? Corporate press releases during merger negotiations (2014) — link
  3. A study on usability and security features of the Android pattern lock screen (2014) — link
  4. FinBERT : Financial Sentiment Analysis with Pre-trained Language Models (2019) — link
  5. Investor sentiment in the stock market (2007) — link
  6. Volatility connectedness between geopolitical risk and financial markets: Insights from pandemic and military crisis periods (2024) — link
  7. Psychology-based models of asset prices and trading volume (2018) — link
  8. Twitter mood predicts the stock market (2011) — link
and 33 more
  1. Generalized autoregressive conditional heteroskedasticity (1986) — link
  2. A conditionally heteroskedastic time series model for speculative prices and rates of return (1987) — link
  3. The impact of uncertainty shocks (2009) — link
  4. New avenues in opinion mining and sentiment analysis (2017) — link
  5. BERT : Pre-training of deep bidirectional transformers for language understanding (2018) — link
  6. Terrorism activity, investor sentiment, and stock returns (2010) — link
  7. The causal impact of media in financial markets (2011) — link
  8. Modeling financial time series with GARCH models and heavy-tailed distributions (2017) — link
  9. GARCH Models: Structure, Statistical Inference, and Financial Applications (2019)
  10. Deep Learning (2016)
  11. Empirical asset pricing via machine learning (2020) — link
  12. Comprehensive review of text-mining applications in finance (2020) — link
  13. DeBERTa : Decoding-enhanced BERT with disentangled attention (2021) — link
  14. Sentiment Analysis and Opinion Mining (2012) — link
  15. RoBERTa : A robustly optimized BERT pretraining approach (2020) — link
  16. When is a liability not a liability? Textual analysis, dictionaries, and 10-Ks (2011) — link
  17. Opinion mining and sentiment analysis (2008) — link
  18. Investor sentiment and stock returns: Some international evidence (2009) — link
  19. Deep learning in neural networks: An overview (2015) — link
  20. Intraday forecasts of a volatility index: Functional time series methods with dynamic updating (2019) — link
  21. News sentiment and the investor fear gauge (2014) — link
  22. Tweets and trades: The information content of stock microblogs (2010) — link
  23. Giving content to investor sentiment: The role of media in the stock market (2007) — link
  24. Attention is all you need (2017) — link
  25. Econometric Analysis of Cross Section and Panel Data (2010)
  26. Geopolitical risk and stock prices (2024) — link
  27. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of United Kingdom Inflation (1982) — link
  28. A Practical Guide to Splines (2001)
  29. Modelling skewness dynamics in series of financial data using skewed distributions (2001) — link
  30. War, uncertainty and oil prices: Evidence from the Russia–Ukraine conflict (2022) — link
  31. VADER: A Parsimonious Rule-Based Model for Sentiment Analysis of Social Media Text (2014) — link
  32. War, economic sanctions and stock market volatility: The case of the Russia–Ukraine conflict (2023) — link
  33. Deep learning for financial sentiment analysis: A survey (2018) — link

0 of 41 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.