Bazzi Blasques Koopman Lucas 2017
“Time-Varying Transition Probabilities for Markov Regime Switching Models” (Journal of Time Series Analysis 38:458-478; earlier Tinbergen Institute / SSRN working paper, 2014) proposes a Markov switching model in which the transition probabilities are not held constant but evolve over time through an observation-driven (generalised autoregressive score) updating law, with the innovation generated by the score of the predictive likelihood. It appears in this vault as the econometric formalisation of the remedy for the Non-Stationary Transition Matrix failure mode: it treats a constant transition matrix as a misspecification rather than a default, building on the earlier covariate-driven time-varying-transition-probability work of Diebold, Lee & Weinbach (1994) and Filardo (1994). It is a methodological / time-series-econometrics contribution, not a trading study, so it carries no profitability claim.
Connections
- Non-Stationary Transition Matrix — relates, provides the score-driven time-varying transition probability remedy, source: https://onlinelibrary.wiley.com/doi/10.1111/jtsa.12211
- Markov Regime-Switching Model — proposes_model, extends the model to time-varying transition probabilities, source: https://onlinelibrary.wiley.com/doi/10.1111/jtsa.12211