Paper: arXiv 2412.04063

Authors: Kevin Benson, Ing-Haw Cheng, John Hull, Charles Martineau, Yoshio Nozawa, Vasily Strela, Yuntao Wu, Jun Yuan

Abstract

We study the drivers of the Gilchrist and Zakrajšek (2012) excess bond premium (EBP) through the lens of the news. The monthly attention the news pays to 180 topics (Bybee et al., 2024) captures up to 80% of the variation in the EBP, and this component of variation forecasts macroeconomic movements. Greater news attention to financial intermediaries and crises tends to drive up the EBP and portend macroeconomic downturns, while greater news attention to politics and science tends to drive down the EBP. Attention-based estimates of EBP largely drive out the forecast power of direct sentiment measures for macroeconomic fluctuations and predict the business cycle going back to the early 1900’s. Overall, we attribute predictive variation about the EBP for macroeconomic movements to variation in news attention to financial intermediaries, crises, and politics.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper relies on straightforward statistical linear combinations, Lasso, and SHAP analyses rather than heavy theoretical math, placing it low on math complexity. However, its empirical design is highly rigorous, deploying extensive textual data spanning over a century to analyze macroeconomic forecasting, out-of-sample backtests, and variance decompositions.

Research Flowchart

  flowchart TD
    A[Research Goal: Understand Drivers & Predictive Power of Excess Bond Premium] --> B[Data Inputs: Monthly EBP & News Attention on 180 Topics 1900s-Present]
    B --> C[Methodology: Textual Analysis & Topic Modeling]
    C --> D[Computational Process: Regress EBP on Topic Attention & Forecast Macro Variables]
    D --> E[Outcome 1: News Attention Explains up to 80% of EBP Variation]
    D --> F[Outcome 2: Intermediary/Crisis News Drives EBP Up; Politics/Science Drives EBP Down]
    E & F --> G[Key Finding: Attention-based EBP Robustly Forecasts Macro Downturns & Outperforms Sentiment]

Cited References (56)

  1. Belief Overreaction and Stock Market Puzzles (2024) — link
  2. Diagnostic Expectations and Credit Cycles (2018) — link
  3. Intermediary asset pricing: New evidence from many asset classes (2017) — link
  4. Liquidity and leverage (2010) — link
  5. Credit-Market Sentiment and the Business Cycle (2017) — link
  6. Banking Crises Without Panics (2020) — link
  7. Historical Banking Crises: A New Database and a Reassessment of their Incidence and Severity (2022)
  8. Macroeconomic Attention and Announcement Risk Premia (2022) — link
and 48 more
  1. Macroeconomic Shocks and Their Propagation (2016) — link
  2. Financial Crises: A Survey (2022) — link
  3. When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10-Ks (2011) — link
  4. Business News and Business Cycles (2024) — link
  5. Credit spreads and real activity (2008)
  6. How Credit Cycles across a Financial Crisis (2020)
  7. Mutual fund flows and fluctuations in credit and business cycles (2021) — link
  8. The Term Structure of the Excess Bond Premium: Measures and Implications (2021)
  9. The Bond Market’s q* (2009) — link
  10. Macroeconomic Conditions and the Puzzles of Credit Spreads and Capital Structure (2010)
  11. The levered equity risk premium and credit spreads: a unified framework (2010)
  12. Uncertainty, Financial Frictions, and Investment Dynamics (2014) — link
  13. Credit market shocks and economic fluctuations: Evidence from corporate bond and stock markets (2009) — link
  14. Issuer Quality and Corporate Bond Returns (2013) — link
  15. Bond Spreads and Economic Activity in Eight European Economies (2016) — link
  16. Market-Based Monetary Policy Uncertainty (2021) — link
  17. Narrative Monetary Policy Uncertainty (2023)
  18. Forecasting Using Text-Based Uncertainty Measures (2023)
  19. Measuring Economic Policy Uncertainty (2016) — link
  20. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding (2019) — link
  21. (Almost) 200 Years of News-Based Economic Sentiment (2022)
  22. Monetary policy uncertainty (2020) — link
  23. What explains the stock market’s reaction to Federal Reserve policy? (2005)
  24. Monetary policy surprises and interest rates: Evidence from the Fed funds futures market (2001)
  25. Measuring the effects of federal reserve forward guidance and asset purchases on financial markets (2021)
  26. Macroeconomic Attention and Announcement Risk Premia (2022) — link
  27. Multifrequency news and stock returns (2007) — link
  28. Long-Run Risk and the Persistence of Consumption Shocks (2013) — link
  29. News implied volatility and disaster concerns (2017)
  30. Realised Volatility Forecasting: Machine Learning via Financial Word Embedding (2021) — link
  31. The colour of finance words (2023) — link
  32. Predicting Returns With Text Data (2019) — link
  33. FinBERT: Financial Sentiment Analysis with Pre-trained Language Models (2019) — link
  34. Text Selection (2019) — link
  35. Expected Returns and Large Language Models (2022) — link
  36. The Structure of Economic News (2020) — link
  37. Business News and Business Cycles (2021) — link
  38. Notes on the n-Person Game – II: The Value of an n-Person Game (1951) — link
  39. A value for n-person games (1988)
  40. Analysis of regression in game theory approach (2001) — link
  41. Explaining Prediction Models and Individual Predictions with Feature Contributions (2014) — link
  42. A Unified Approach to Interpreting Model Predictions (2017) — link
  43. Text Selection (2021) — link
  44. The colour of finance words (2023) — link
  45. Multinomial Inverse Regression for Text Analysis (2010) — link
  46. Predicting Returns with Text Data (2019) — link
  47. Business news and business cycles (2021)
  48. Expected returns and large language models (2022)

0 of 56 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.