A1058
Title: When Black-Litterman meets decision-fusion for asset allocation
Authors: Xinyu Huang - University of Bath (United Kingdom)
David Newton - University of Bath (United Kingdom)
Emmanouil Platanakis - University of Bath - School of Management, UK (United Kingdom) [presenting]
Xiaoxia Ye - University of Exeter Business School (United Kingdom)
Abstract: Efficient decision-making often requires synthesizing insights from diverse knowledge domains. A novel method to model and consolidate inputs from multiple experts within the Black-Litterman model (BLM) is introduced. Each expert's beliefs and subjective judgements on stock trends are formulated as belief distributions in the decision fusion framework. Both misspecification and estimation errors are accounted for in the equilibrium allocations and covariance matrix. For equilibrium allocations in BLM, the mean-variance asset allocation rule is optimally combined with either the minimum-variance or the 1/N rule. Wishart stochastic volatility for modeling covariances is employed, initializing with a sparse conditional covariance matrix via the graphical lasso. After transaction costs, the method consistently boosts the Sharpe ratio compared to the 1/N rule and surpasses eight other established portfolio strategies for managing estimation risk. Robustness experiments show that the approach withstands various parameter choices.
