Stock Price Predictability and the Business Cycle via Machine Learning

Paper: arXiv 2304.09937

Authors: Li Rong Wang, Hsuan Fu, Xiuyi Fan

Abstract

We study the impacts of business cycles on machine learning (ML) predictions. Using the S&P 500 index, we find that ML models perform worse during most recessions, and the inclusion of recession history or the risk-free rate does not necessarily improve their performance. Investigating recessions where models perform well, we find that they exhibit lower market volatility than other recessions. This implies that the improved performance is not due to the merit of ML methods but rather factors such as effective monetary policies that stabilized the market. We recommend that ML practitioners evaluate their models during both recessions and expansions.

Complexity vs Empirical Score

• Math Complexity: 3.5/10
• Empirical Rigor: 6.5/10
• Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: The paper provides a practical macro-level analysis of machine learning predictions across different market regimes (recessions vs. expansions). While mathematically light, it offers useful empirical insights into model degradation during high-volatility economic downturns and caution against over-reliance on ML during macroeconomic shocks.

Research Flowchart

  flowchart TD
    A["Research Goal: Assess ML Stock Price Predictability Across Business Cycles"] --> B["Inputs: S&P 500 Index Data, Recession History & Risk-Free Rates"]
    B --> C["Train ML Prediction Models & Evaluate in Expansions vs. Recessions"]
    C --> D{"Performance Analysis"}
    D -->|General Finding| E["ML Models Perform Worse in Most Recessions"]
    D -->|High-Performance Recessions| F["Analyze Volatility & Monetary Policy Effects"]
    E --> G["Key Outcome: Success Linked to Low Volatility/Policy, Not ML Superiority; Recommend Cycle-Specific Evaluation"]
    F --> G