The Black-Litterman model combines investor’s personal views with historical data and gives optimal portfolio weights. In (Andrei & Hsu, 2020), they reviewed the original Black-Litterman model and modified it in order to fit it into a Bayesian framework, when a certain number of assets is considered. They used the idea by (Leonard & Hsu, 1992) for a multivariate normal prior on the logarithm of the covariance matrix. When implemented and applied to a large number of assets such as all the S&P500 companies, they ran into memory allocation and running time issues. In this paper, we reduce the dimensions by considering Bayesian factor models, which solve the asset allocation problems for a large number of assets. In addition, we will conduct sensitivity analysis for the confidence levels that the investors have to input.
On Posterior Consistency of Bayesian Factor Models in High Dimensions
