Estimating Markov Switching Models When Baum-Welch Converges Slowly

Summary

The document raises a practical estimation problem in using Markov switching models to forecast financial risk regimes. It reports that the Baum-Welch algorithm, commonly used for hidden Markov model estimation, can converge very slowly or fail to converge in some cases. It frames the problem in the context of regime forecasting similar to work by Kritzman, Page, and Turkington.

The author asks whether alternatives used in speech recognition and other fields have been applied to financial time series, naming Bayesian estimation and entropic-prior-based estimators as possibilities. The document offers no comparison, implementation guidance, or empirical findings, so it serves as a research question rather than a recommendation. It highlights convergence as one concern and asks whether other criteria should guide method choice, but leaves those tradeoffs unresolved.

Key ideas

• Markov switching models can be used to forecast changes in financial risk regimes.
• Baum-Welch estimation may converge slowly or fail to converge in some applications.
• Bayesian estimation and entropic-prior estimators are raised as possible alternatives.
• The document provides no financial comparison or evidence favoring one estimation method.

Tags

# Markov switching model estimation


# Markov switching model estimation












We are testing Markov switching models to forecast risk regimes, similar to the paper by Kritzman, Page and Turkington. We find that in some cases the Baum-Welch algorithm converges very slowly or not at all. Apparently this issue is known for hidden Markov models in speech recognition and other fields, where several alternatives to Baum-Welch have been proposed: for example, Bayesian estimation by M. Johnson or estimators based on an entropic prior by M. Brand.

Have these alternatives been applied to financial times series? Is there a reason to favor one particular approach, apart from slow convergence?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.