Multi-regime Markov-switching models with time-varying transition probabilities: An application to U.S. Treasury yields
Abstract
This paper studies Markov-switching (MS) models with time-varying transition probabilities (TVTP) under alternative specifications of the transition probability matrix. We extend the two-regime common-variance setting of the Generalized Autoregressive Score (GAS) model of Bazzi et al. (2017) to the general K-regime case with regime-specific means and variances, and develop an open-source R package, multiregimeTVTP, for simulation and estimation of such models. A Monte Carlo study shows that the regime means, variances, and transition probabilities are reliably recovered, whereas the TVTP driving coefficients are harder to identify. The GAS score coefficient appears to be statistically non-identifiable, owing to a ridge in the likelihood surface linking it to the regime variance. The filtered regime probabilities are accurately recovered under correct specification, whereas the one-step-ahead conditional mean is insensitive to misspecification of the transition dynamics. An empirical application to U.S. Treasury zero-coupon yield changes (1961-2024) at four maturities shows that a specification driven by the lagged yield level provides the best fit, attaining the lowest AIC at all four maturities and the lowest BIC at the short end of the yield curve, and that the estimated regimes align with documented episodes of U.S. monetary history.
Related
- Nonlinear parameter-varying embeddings for nonlinear state estimation with application to a two-link robot manipulator
- Stochastic linear-quadratic control problem for regime-switching jump-diffusion system and its application to finance
- Trial design considerations in the face of overdispersion and time-varying event rates: a meta-regression approach with an application in asthma
- Large Deviations in Switching Diffusion: Dynamical Condensation and Diffusivity BBP Transition
- Unidentified prediction-risk rankings under spatial sampling with fixed inclusion probabilities
BibTeX
@misc{modee2026multiregime,
title = {{Multi-regime Markov-switching models with time-varying transition probabilities: An application to U.S. Treasury yields}},
author = {Samuel Modée and Yushu Li and Sjur Westgaard and Stein Andreas Bethuelsen},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.14976},
eprint = {2605.14976},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.14976},
}
RIS
TY - UNPB
TI - Multi-regime Markov-switching models with time-varying transition probabilities: An application to U.S. Treasury yields
AU - Samuel Modée
AU - Yushu Li
AU - Sjur Westgaard
AU - Stein Andreas Bethuelsen
PY - 2026
DA - 2026/09/13
PB - arXiv
DO - 10.48550/arxiv.2605.14976
UR - https://arxiv.org/abs/2605.14976
AB - This paper studies Markov-switching (MS) models with time-varying transition probabilities (TVTP) under alternative specifications of the transition probability matrix. We extend the two-regime common-variance setting of the Generalized Autoregressive Score (GAS) model of Bazzi et al. (2017) to the general K-regime case with regime-specific means and variances, and develop an open-source R package, multiregimeTVTP, for simulation and estimation of such models. A Monte Carlo study shows that the regime means, variances, and transition probabilities are reliably recovered, whereas the TVTP driving coefficients are harder to identify. The GAS score coefficient appears to be statistically non-identifiable, owing to a ridge in the likelihood surface linking it to the regime variance. The filtered regime probabilities are accurately recovered under correct specification, whereas the one-step-ahead conditional mean is insensitive to misspecification of the transition dynamics. An empirical application to U.S. Treasury zero-coupon yield changes (1961-2024) at four maturities shows that a specification driven by the lagged yield level provides the best fit, attaining the lowest AIC at all four maturities and the lowest BIC at the short end of the yield curve, and that the estimated regimes align with documented episodes of U.S. monetary history.
ER -