以馬可夫轉換模型優化美林時鐘： 動態資產配置的實務應用與英國市場實證

Abstract

本研究探討了利用馬可夫轉換模型優化美林時鐘動態資產配置策略的潛力，並結合 ARIMA 模型進一步提升預測能力。傳統的美林時鐘策略以 PPI 與 CPI 年增率差值的正負變化作為股債配置的判斷依據，具有簡單易行且有效的市場擇時能力。然而，總體經濟變數的時間序列並不總是呈現出明顯的四階段循環，這可能導致傳統美林時鐘在某些情況下的適用性受限，特別是在經濟指標頻繁變動或市場波動加劇時。

為優化美林時鐘策略，本研究提出基於二狀態馬可夫轉換模型的動態資產配置方法。該模型通過平滑機率動態劃分牛市與熊市的機率分布，提供了更為量化的股債轉換時機判斷方式。此外，結合 ARIMA 模型對 PPI 與 CPI 年增率差值的未來走勢進行預測，能有效估算當前經濟狀態的持續期間，進一步提升動態配置策略的準確性。同時，馬可夫轉換模型與美林時鐘的結合，還能通過計算定態分布，提供簡便的資產權重計算框架，適用於長期動態資產配置。

基於 2012 年 10 月至 2024 年 8 月的英國市場實證結果顯示，傳統美林時鐘與基於馬可夫轉換模型的動態配置策略均顯著優於富時100指數（FTSE 100 Index）與60/40股債配置，展現出卓越的市場擇時能力。此外，基於馬可夫模型優化的美林時鐘配置策略在某些時間段（如 2020 年疫情期間和 2022 年市場波動時）明顯優於傳統的美林時鐘策略，突顯出馬可夫轉換模型靈巧地捕捉經濟信號變化的優勢。整體而言，基於美林時鐘的動態資產配置策略展現了顯著的市場擇時能力，而結合馬可夫轉換模型的策略進一步提升了該能力，為投資者提供了一種兼具直觀性與精確性的動態資產配置框架，幫助在不同經濟階段中實現更高效的投資決策。

This study explores the potential of using Markov Switching Models to optimize the dynamic asset allocation strategy of the Merrill Lynch Clock, further enhancing forecasting capabilities by integrating ARIMA models. Traditional Merrill Lynch Clock strategies rely on the positive and negative changes in the year-on-year growth rate difference between PPI and CPI as a basis for equity-bond allocation decisions. While simple and effective in market timing, the cyclical patterns of macroeconomic variables are not always evident, which may limit the applicability of traditional Merrill Lynch Clock strategies, especially during periods of frequent economic indicator fluctuations or heightened market volatility.

To optimize the Merrill Lynch Clock strategy, this study proposes a dynamic asset allocation approach based on a two-state Markov Switching Model. By dynamically segmenting the probability distribution of bull and bear markets through smoothed probabilities, the model provides a more quantitative method for determining equity-bond switching timings. Additionally, integrating ARIMA models to forecast the future trends of the year-on-year growth rate difference between PPI and CPI enables the effective estimation of the duration of the current economic state, further improving the accuracy of dynamic allocation strategies. Moreover, the combination of Markov Switching Models and the Merrill Lynch Clock can calculate steady-state distributions, providing a simplified framework for asset weight calculations suitable for long-term dynamic asset allocation.

Empirical evidence from the UK market between October 2012 and August 2024 demonstrates that both the traditional Merrill Lynch Clock and the dynamic allocation strategy based on the Markov Switching Model significantly outperform the FTSE 100 Index and a 60/40 equity-bond portfolio, showcasing superior market timing capabilities. Furthermore, the Merrill Lynch Clock strategy optimized with the Markov Model outperformed the traditional Merrill Lynch Clock in certain periods, such as during the COVID-19 pandemic in 2020 and the market volatility in 2022, highlighting the Markov Switching Model's advantage in adeptly capturing shifts in economic signals. Overall, the dynamic asset allocation strategy based on the Merrill Lynch Clock exhibits remarkable market timing capabilities. The strategy integrated with the Markov Switching Model further enhances this ability, offering investors an intuitive yet precise dynamic asset allocation framework to achieve more efficient investment decisions across different economic cycles.