Please rotate your device to landscape mode to view the charts.

Bond Risk Premiums with Machine Learning

Journal: Review of Financial Studies

Date: 20210201

Author: Bianchi, Daniele; Buchner, Matthias; Tamoni, Andrea

Abstract:
We show that machine learning methods, in particular, extreme trees and neural networks (NNs), provide strong statistical evidence in favor of bond return predictability. NN forecasts based on macroeconomic and yield information translate into economic gains that are larger than those obtained using yields alone. Interestingly, the nature of unspanned factors changes along the yield curve: stock- and labor-market-related variables are more relevant for short-term maturities, whereas output and income variables matter more for longer maturities. Finally, NN forecasts correlate with proxies for time-varying risk aversion and uncertainty, lending support to models featuring both channels.

Link: Google Scholar

Background and Context

Research Focus

The study examines how machine learning methods can improve bond return predictions compared to traditional statistical approaches.

Methodology

Researchers analyze U.S. Treasury bonds using neural networks, extreme trees, and other machine learning techniques to predict excess returns.

Data Used

The analysis uses Treasury yield curve data from 1971-2018 along with 128 macroeconomic variables to test predictive accuracy.

Superior Performance of Neural Networks vs Traditional Methods

  • Neural networks achieve significantly higher predictive accuracy (26.4% R²) compared to traditional PCA methods (-0.7% R²)
  • Even simple machine learning methods like extreme trees (24.6% R²) outperform traditional approaches
  • Results demonstrate the value of nonlinear modeling approaches for bond return prediction

Enhanced Performance with Macroeconomic Data

  • Adding macroeconomic variables improves predictive accuracy by approximately 10 percentage points
  • Demonstrates that bond yields alone do not capture all relevant information for predicting returns
  • Supports the value of incorporating broader economic data in bond return forecasting

Economic Value of Predictions

  • Neural network predictions translate into significant economic gains across all bond maturities
  • Highest gains achieved when considering all maturities together (4.829% CER)
  • Economic benefits increase with bond maturity

Performance Across Economic Cycles

  • Neural network predictions perform better during recessions
  • Model maintains positive performance during expansions
  • Demonstrates robustness across different economic conditions

Contribution and Implications

  • Machine learning methods significantly improve bond return predictions compared to traditional approaches
  • The combination of yield curve and macroeconomic data provides the best predictive performance
  • Results support investment strategies that consider both time-varying risk prices and macroeconomic uncertainty

Data Sources

  • Method Comparison Chart: Based on Table 1, comparing R² values for different prediction methods
  • Macro Comparison Chart: Based on Tables 1 and 2, showing improvement from adding macro variables
  • CER Chart: Based on Table 5, showing certainty equivalent returns
  • Variable Importance Chart: Based on Figure 5 variable importance analysis
  • Cycle Performance Chart: Based on Table 6, showing Sharpe ratios across economic conditions