Weighted Bayesian Model Averaging for Portfolio Decisions in Matrix Variate Dynamic Linear Models

In this paper, we assess Bayesian model averaging (BMA) techniques for dynamic linear models (DLMs) with variance matrix discounting. In previous research, the discount factors for the variance matrices and the auto-regressive lag have typically been pre-determined and held constant over time. Using posterior model probabilities, we average DLMs employing different discount rates and lag parameters, allowing flexibility in the weights on these parameter values. We also include an additional discount factor for previous model likelihoods to emphasize the most relevant data and make the averaging process more inclusive. We apply this model averaging and extension to the daily prices of nine exchange rates, two commodities, and two stock indices, evaluating fiveday forecasts and portfolio returns. Afterward, we apply the same model to the series of realized returns of the averaged models to create a “fund of funds” portfolio. We find that there is shifting of model probabilities over time across both discount factors and lags, and the inclusion of likelihood discounting leads to higher portfolio returns.