Paper: arXiv 2604.13458

Authors: Songrun He

Abstract

In this paper, I present the first comprehensive, around-the-clock analysis of systematic jump risk by combining high-frequency market data with contemporaneous news narratives identified as the underlying causes of market jumps. These narratives are retrieved and classified using a state-of-the-art open-source reasoning LLM. Decomposing market risk into interpretable jump categories reveals significant heterogeneity in risk premia, with macroeconomic news commanding the largest and most persistent premium. Leveraging this insight, I construct an annually rebalanced real-time Fama-MacBeth factor-mimicking portfolio that isolates the most strongly priced jump risk, achieving a high out-of-sample Sharpe ratio and delivering significant alphas relative to standard factor models. The results highlight the value of around-the-clock analysis and LLM-based narrative understanding for identifying and managing priced risks in real time.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper expertly combines high-frequency financial econometrics (continuous-time Fama-MacBeth regressions) with state-of-the-art open-source reasoning LLMs (Qwen) to analyze 24-hour systematic jump risk. It delivers robust empirical results with significant out-of-sample portfolio performance, bridging cutting-edge AI techniques with modern asset pricing.

Research Flowchart

  flowchart TD
    A["Research Question: Identify & Price Interpretable<br>Around-the-Clock Systematic Jump Risk"] --> B["Data Inputs: High-Frequency Market Data<br>& Contemporaneous News Narratives"]
    B --> C["LLM Processing: Reason & Classify Jumps<br>into Narrative Risk Categories"]
    C --> D["Risk Premium Analysis: Decompose Risk & Estimate<br>Heterogeneous Premia via Fama-MacBeth"]
    D --> E["Portfolio Construction: Build Factor-Mimicking<br>Portfolio Isolating Macroeconomic Risk"]
    E --> F["Key Outcome: Superior Out-of-Sample Sharpe Ratio<br>& Significant Alpha vs. Standard Models"]

Cited References (75)

  1. Continuous-Time Fama-MacBeth Regressions (2025)
  2. Comparing cross-section and time-series factor models (2020)
  3. Slope factors outperform: evidence from a large comparative study (2023)
  4. High-frequency financial econometrics (2014)
  5. News and asset pricing: A high-frequency anatomy of the SDF (2025)
  6. Jump regressions (2017)
  7. The conditional CAPM does not explain asset-pricing anomalies (2006)
  8. On estimating the expected return on the market: An exploratory investigation (1980)
and 67 more
  1. On the estimation of beta-pricing models (1992)
  2. When moving-average models meet high-frequency data: uniform inference on volatility (2021)
  3. Common risk factors in the returns on stocks and bonds (1993)
  4. Liquidity measurement problems in fast, competitive markets: Expensive and cheap solutions (2014)
  5. High-frequency factor models and regressions (2020)
  6. DeepSeek-R1 : Incentivizing reasoning capability in llms via reinforcement learning (2025) — link
  7. Chain-of-thought prompting elicits reasoning in large language models (2022)
  8. A five-factor asset pricing model (2015)
  9. Choosing factors (2018)
  10. Deep learning in asset pricing (2024)
  11. The natural language of finance (2025)
  12. What triggers stock market jumps? (2021)
  13. Business news and business cycles (2024)
  14. Narrative asset pricing: Interpretable systematic risk factors from news text (2023)
  15. Does Overnight News Explain Overnight Returns? — → 2507.04481
  16. Roughing up beta: Continuous versus discontinuous betas and the cross section of expected stock returns (2016)
  17. Granular betas and risk premium functions (2025)
  18. Expected returns and large language models (2022)
  19. Does finance benefit society? A language embedding approach (2025)
  20. Chronologically Consistent Large Language Models — → 2502.21206
  21. The value of information from sell-side analysts — → 2411.13813
  22. Can chatgpt forecast stock price movements? return predictability and large language models — → 2304.07619
  23. ChatGPT and DeepSeek: Can they predict the stock market and macroeconomy? (2025) — link
  24. Unusual financial communication: ChatGPT, earnings calls, and financial markets (2024)
  25. Economic representations (2024)
  26. The ghost in the machine: Generating beliefs with large language models (2023) — link
  27. News implied volatility and disaster concerns (2017)
  28. Fundamentals of perpetual futures — → 2212.06888
  29. The fundamentals of commodity futures returns (2013)
  30. Common risk factors in currency markets (2011)
  31. The cross-section of intraday and overnight returns (2021)
  32. The granular origins of aggregate fluctuations (2011)
  33. Information, trading, and volatility: Evidence from firm-specific news (2019)
  34. Stock return variances: The arrival of information and the reaction of traders (1986)
  35. Lookahead bias in pretrained language models (2024)
  36. Optimal versus naive diversification: How inefficient is the 1/N portfolio strategy? (2009)
  37. Using stocks or portfolios in tests of factor models (2020)
  38. Trading costs (2018)
  39. Machine learning and the implementable efficient frontier (2024)
  40. A taxonomy of anomalies and their trading costs (2016)
  41. Factor models, machine learning, and asset pricing (2022)
  42. Data-snooping biases in tests of financial asset pricing models (1990)
  43. Why We Think (2025) — link
  44. Thinking, fast and slow (2011)
  45. Qwen3 technical report (2025) — link
  46. Identification in macroeconomics (2018)
  47. High-frequency identification of monetary non-neutrality: the information effect (2018)
  48. Does monetary policy matter? A new test in the spirit of Friedman and Schwartz (1989)
  49. Federal Reserve information and the behavior of interest rates (2000)
  50. Asset pricing: A tale of night and day (2020)
  51. Risk preferences and the macroeconomic announcement premium (2018)
  52. The pre-FOMC announcement drift (2015)
  53. So many jumps, so few news (2024)
  54. News as sources of jumps in stock returns: Evidence from 21 million news articles for 9000 companies (2022)
  55. Warp speed price moves: Jumps after earnings announcements (2025)
  56. Probable inference, the law of succession, and statistical inference (1927)
  57. Do Sell-side Analyst Reports Have Investment Value? (2025) — link
  58. Chronologically Consistent Generative AI (2025) — link
  59. An intertemporal capital asset pricing model (1973)
  60. Tests of statistical hypotheses concerning several parameters when the number of observations is large (1943)
  61. The overnight drift (2023)
  62. Does finance benefit society? A language embedding approach (2025)
  63. Expected Returns and Large Language Models (2023) — link
  64. Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models (2023) — link
  65. Instruction Tuning Chronologically Consistent Language Models (2025) — link
  66. The Natural Language of Finance (2025)
  67. Lookahead Bias in Pretrained Language Models (2024) — link

5 of 75 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.