MARKET REGIME DETECTION USING HIDDEN MARKOV MODEL AND TACTICAL ASSET ALLOCATION STRATEGIES IN ASIAN EQUITY MARKETS

Abstract

Asian equity markets are highly vulnerable to sudden regime shifts, yet most quantitative strategies assume return stationarity. This paper evaluates whether a three-state Gaussian Hidden Markov Model (HMM) can reliably identify economically distinct market regimes across five major Asian indices (BSE SENSEX, Nikkei 225, Hang Seng, STI, and SET), and whether regime-aware trading rules can generate superior risk-adjusted returns. Parameters are estimated using the Baum-Welch EM algorithm with twenty random restarts, and the most probable state sequences are decoded via the Viterbi Algorithm. To prevent look-ahead bias, parameters are locked at a January 2022 cutoff for a four-year out-of-sample testing window. We compare three distinct strategies: a Hybrid rule (following momentum in trending states and fading the market during mean-reversion states), a pure Contrarian rule, and a passive Buy-and-Hold benchmark. The HMM demonstrates strong out-of-sample stability, with over 74% of forecasted days maintaining posterior probabilities above 0.80 in four of the five markets. The Hybrid strategy consistently outperforms the Buy-and-Hold benchmark on a Sharpe basis in three markets, showing the most significant gains in the Nikkei 225 (Sharpe Ratio +0.71 vs. +0.47) and the STI (SR +0.84 vs. +0.53), while successfully limiting drawdowns in Thailand’s structurally declining market. Conversely, the strategy fails in Hong Kong due to a prolonged Bear-state misclassification triggered by post-2022 structural breakdowns, which erased the model's predictive edge and allowed the pure Contrarian rule to deliver the superior performance.

Keywords: Hidden Markov Model, Market Regime Detection, Tactical Asset Allocation, Asian Equity Markets, Hybrid Strategy