Systematic Macro, Factor Investing, and MachineLearning in G10 Fixed Income : Systematiska makrostrategier, faktorinvesteringar och maskininlärning på G10-räntemarknaden

Sammanfattning

When inflation returned in 2021 and ended four decades of falling yields, it sparked a curiosity among market makers to better understand systematic hedge fund behavior in sovereign bond markets, and it is this curiosity that the present thesis seeks to address. Conducted in collaboration with SEB’s Rates desk within the FICC division, the study aims to give market makers a better understanding of how systematic hedge funds (CTAs) may trade in fixed income markets. Using macroeconomic data synchronized with daily bond futures prices from September 2019 to April 2025, the thesis implements and back-tests ten rule-based strategies across the classic factor categories: Momentum, Value, and Carry. The strategies are evaluated both as outright directional positions and as market-neutral cross-sectional portfolios. A multi-strategy composite portfolio is constructed using equal weighting, Global Minimum Variance, and Enhanced Portfolio Optimization with covariance shrinkage. Finally, an XGBoost classifier is trained on pre-2019 data and tested out-of-sample.The results show that macroeconomic momentum and value strategies performed best in risk-adjusted terms compared to the passive equal-weighted G10 benchmark, especially during the 2021-2023 inflationary shock. The best-performing strategies were Unemployment Momentum (Sharpe 0.70), Inflation Breakout (Sharpe 0.57), and Yield Curve Slope Reversal (Sharpe 0.54), while Carry strategies struggled in the inverted yield curve environment. Multi-strategy portfolios delivered the strongest overall performance, achieving the highest risk-adjusted returns with substantially lower drawdowns than any standalone strategy. Multi-strategy portfolios improved Sharpe ratios and reduced drawdowns, with moderate covariance shrinkage (λ = 0.25–0.50) providing the best balance between diversification and turnover. No individual strategy reaches statistical significance at the 5% level under Newey-West adjusted tests. The results indicate that these strategies do not work equally well at all times. They performed best during the 2021–2023 period when inflation surged and central banks raised interest rates aggressively, but tended to add less value during calmer periods. This thesis also gives a clear and practical way to evaluate and replicate systematic macro strategies. For SEB’s Rates desk, the results offer a starting point for understanding what drives systematic CTA client activity.