Paper: arXiv 2601.04608

Authors: Jinjun Liu, Ming-Yen Cheng

Abstract

We study U.S. Treasury yield curve forecasting under distributional uncertainty and recast forecasting as an operations research and managerial decision problem. Rather than minimizing average forecast error, the forecaster selects a decision rule that minimizes worst case expected loss over an ambiguity set of forecast error distributions. To this end, we propose a distributionally robust ensemble forecasting framework that integrates parametric factor models with high dimensional nonparametric machine learning models through adaptive forecast combinations. The framework consists of three machine learning components. First, a rolling window Factor Augmented Dynamic Nelson Siegel model captures level, slope, and curvature dynamics using principal components extracted from economic indicators. Second, Random Forest models capture nonlinear interactions among macro financial drivers and lagged Treasury yields. Third, distributionally robust forecast combination schemes aggregate heterogeneous forecasts under moment uncertainty, penalizing downside tail risk via expected shortfall and stabilizing second moment estimation through ridge regularized covariance matrices. The severity of the worst case criterion is adjustable, allowing the forecaster to regulate the trade off between robustness and statistical efficiency. Using monthly data, we evaluate out of sample forecasts across maturities and horizons from one to twelve months ahead. Adaptive combinations deliver superior performance at short horizons, while Random Forest forecasts dominate at longer horizons. Extensions to global sovereign bond yields confirm the stability and generalizability of the proposed framework.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced mathematical concepts like distributionally robust optimization, moment-based ambiguity sets, and expected shortfall, indicating high math complexity. It also details a comprehensive empirical study with monthly data from 2006-2025, out-of-sample testing across multiple horizons, and extension to global sovereign bond yields, demonstrating substantial data and implementation rigor.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Forecast Treasury Yields<br>under Distributional Uncertainty"] --> B["Data Inputs:<br>Monthly U.S. Treasury Yields<br>Macroeconomic Indicators"]
  B --> C["Methodology:<br>1. Rolling Factor Augmented<br>Dynamic Nelson Siegel Model<br>2. Random Forest Model"]
  C --> D["Distributionally Robust<br>Forecast Combination:<br>Minimizes Worst-case Expected Loss<br>Adaptive Weights & Ridge Covariance"]
  D --> E["Computational Process:<br>Out-of-Sample Evaluation<br>1-12 Month Horizons"]
  E --> F{"Key Findings/Outcomes"}
  F --> G["Short Horizons: Adaptive<br>Combination Dominates"]
  F --> H["Long Horizons: Random Forest<br>Dominates"]
  F --> I["Generalizability: Robust<br>Performance on Global Bonds"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a distributionally robust ensemble forecasting framework that combines a rolling‑window Factor‑Augmented Dynamic Nelson‑Siegel model with high‑dimensional Random Forest forecasts via adaptive forecast combinations.

Source sentence
We propose a distributionally robust ensemble forecasting framework that integrates parametric factor models with high dimensional nonparametric machine learning models through adaptive forecast combinations.

Assumes 5

  • Only lagged predictor information is used to reflect data availability. data
    Source sentence
    Considering the release of many key economic data is lagging 1 month, only lagged predictor information is used to reflect data availability.
  • Forecast errors are evaluated under a moment‑based ambiguity set allowing uncertainty in mean and covariance. model
    Source sentence
    DRO seeks decisions that perform well against the worst-case distribution within a prescribed ambiguity set.
  • Ridge‑regularized covariance matrices stabilize second‑moment estimation in the combination step. statistical
    Source sentence
    instability in second‑moment estimation is mitigated through ridge‑regularized covariance matrices.
  • The severity of the worst‑case criterion is adjustable, allowing the forecaster to regulate robustness versus efficiency. model
    Source sentence
    The severity of the worst case criterion is adjustable, allowing the forecaster to regulate the trade off between robustness and statistical efficiency.
Show 1 more
  • Structural breaks are present in the yield series, as indicated by rejected CUSUM tests. statistical
    Source sentence
    The null is strongly rejected across all maturities (p < 0.001), indicating the presence of structural

Data

  • U.S. Treasury zero‑coupon equivalent yields · LSEG Reuters Workspace market-prices commercial
    January 2006 to August 2025 · 15 maturities from 3 months to 30 years
    Source sentence
    We use zero‑coupon equivalent yields on U.S. Treasury securities obtained from the LSEG Reuters Workspace. The sample spans January 2006 to August 2025 at a monthly frequency and covers 15 maturities ranging from 3 months to 30 years.
  • High‑dimensional panel of economic indicators macro unknown
    111 indicator series
    Source sentence
    Let Zt ∈Rp denote the vector of economic indicators observed at month t, where p = 111.
  • Global sovereign benchmark bond yields (10‑year) · LSEG Reuters Workspace market-prices commercial
    January 2010 onward · major economies including Canada, China, Germany, Japan, Malaysia, United Kingdom, United
    Source sentence
    We extend the Random Forest forecasting framework to a cross‑country setting by examining 10‑year benchmark government bond yields for a set of major economies, including Canada, China, Germany, Japan, Malaysia, the United Kingdom, and the United States.

Recognised sources: Refinitiv / LSEG (Datastream, Eikon, Tick History)

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • Adaptive forecast combinations outperform individual models at short horizons, while Random Forest forecasts dominate at medium and longer horizons. performance
    Source sentence
    Empirical results based on monthly data and forecast horizons from one to twelve months show that adaptive combinations outperform individual models at short horizons, whereas RF forecasts dominate at medium and longer horizons.
  • Random Forest models deliver stable RMSFE across horizons, with short‑end yields showing the highest errors (≈24‑25 bps) and long‑dated maturities the lowest (≈13‑14 bps). predictive-accuracy
    Source sentence
    The RF model delivers stable forecast accuracy across horizons, with RMSFE remaining broadly flat from one‑month to twelve‑month horizons for all maturities
  • The Factor‑Augmented Dynamic Nelson‑Siegel (FADNS) model improves short‑horizon performance relative to the baseline DNS model, reducing RMSFE by roughly 5‑15 bps at the one‑month horizon. predictive-accuracy
    Source sentence
    ... consistently reduce RMSFE relative to DNS across maturities. The gains are most pronounced at short and intermediate maturities, where RMSFE declines by roughly 5–15 bps compared to the baseline DNS model.
  • Both DNS and FADNS models exhibit rapidly increasing RMSFE for horizons beyond six months, performing worse than Random Forest models. negative
    Source sentence
    While FADNS continues to outperform DNS at the three‑month horizon, forecast errors increase rapidly beyond six months for all PCA specifications. At the nine‑ and twelve‑month horizons, RMSFE exceeds 150 bps at the short end of the curve and remains above 120 bps even at long maturities. The results are much worse tha

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 53 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (35)

  1. Parsimonious Modeling of Yield Curves (1987)
  2. Forecasting the Term Structure of Government Bond Yields (2006)
  3. A Dynamic N elson– S iegel Model with Forward-Looking Macroeconomic Factors for the Yield Curve in the US (2019) — link
  4. Random Forests (2001)
  5. The Limited Virtue of Complexity in a Noisy World (2025)
  6. Predicting the Yield Curve Using Forecast Combinations (2016) — link
  7. Improved Methods of Combining Forecasts (1984)
  8. Thick Modeling (2004)
and 27 more
  1. Stacked Regressions (1996)
  2. Combining Forecasting Procedures: Some Theoretical Results (2004)
  3. Robust Model Averaging Prediction of Longitudinal Response with Ultrahigh-Dimensional Covariates (2025)
  4. Distributionally Robust Optimization under Moment Uncertainty with Application to Data-Driven Problems (2010)
  5. Data-Driven Distributionally Robust Optimization Using the W asserstein Metric: Performance Guarantees and Tractable Reformulations (2018)
  6. Wasserstein Distributionally Robust Optimization: Theory and Applications in Machine Learning (2019)
  7. Distributionally Robust Local Non-Parametric Conditional Estimation (2020)
  8. Residuals-Based Distributionally Robust Optimization with Covariate Information (2024) — link
  9. Robustifying Conditional Portfolio Decisions via Optimal Transport (2025)
  10. Determining the Number of Factors in Approximate Factor Models (2002)
  11. Asymptotics of the principal components estimator of large factor models with weakly influential factors (2012) — link
  12. Factors That Fit the Time Series and Cross-Section of Stock Returns (2020)
  13. Regression Shrinkage and Selection via the Lasso (1996)
  14. Sharp Thresholds for High-Dimensional and Noisy Sparsity Recovery Using $\ell_1$-Constrained Quadratic Programming ( L asso) (2009)
  15. Honey, I Shrunk the Sample Covariance Matrix (2004) — link
  16. Optimization of Conditional Value-at-Risk (2000)
  17. Distribution of the Estimators for Autoregressive Time Series with a Unit Root (1979)
  18. Factor-Adjusted Regularized Model Selection (2020)
  19. Variable Selection via Nonconcave Penalized Likelihood and Its Oracle Properties (2001)
  20. Eigenvalue Ratio Test for the Number of Factors (2013)
  21. Robust Estimation of a Location Parameter (1964)
  22. The Stationary Bootstrap (1994)
  23. Comparing Predictive Accuracy (1995)
  24. Measuring Economic Policy Uncertainty (2016)
  25. The Combination of Forecasts (1969) — link
  26. Testing the Equality of Prediction Mean Squared Errors (1997)
  27. Term Premia and Interest Rate Forecasts in Affine Models (2002)

0 of 35 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.