Dynamic Asset Allocation based on Hidden Markov Model Regime Sequences

Nikolaj Abrahamsen & Dimitrij Nakovski

Student thesis: Master thesis

Abstract

This thesis attempts to answer whether Dynamic Asset Allocation (DAA) strategies that are formulated on uni- and multivariate Hidden Markov Models (HMM) can significantly outperform Strategic Asset Allocation (SAA) methods and other comparable benchmarks. Building on previous studies, an adaptive estimation approach with time-varying HMM parameters is applied. It is examined how well a 2- and 3-state Gaussian Hidden Markov Model replicates stylized facts of financial returns. The findings demonstrated that the state labelling accuracy is marginally improved when adding an additional input feature in combination with changes to the covariance matrix structure under the Hidden Markov Model. Moreover, with the application of state probability smoothing methods, the persistence of the predicted sequences is further strengthened. The market-neutral academic equity factors that are the most pro-cyclical as measured by the returns in the 2- state HMM are Momentum (MOM) and Size (SMB) factors, whereas the "Betting-againstBeta" (BAB) and "Robust-Minus-Weak" operating profitability factors are found to show the highest returns in the bearish HMM states. Based on the smoothed state sequences, mean-variance efficient portfolios are formed both in the unconditional "all-weather" SAA approach, which in the DAA strategies are optimized conditionally on the regimes and based on the in-sample (2001- 2015) asset performances. Whereby, the portfolios are formulated on MSCI USA equity factors, the S&P GSCI commodity index and the Bloomberg Barclays US Long Treasury Unhedged Bond Index. The findings suggests that for all DAA strategies, the performance in terms of the Sharpe ratios in the out-of-sample data is improved over the SAA portfolio. However, the significance tests of differences in returns establishes that such outperformance is found to be insignificant for all DAA strategies

EducationsMSc in Finance and Investments, (Graduate Programme) Final Thesis
LanguageEnglish
Publication date2021
Number of pages157