Bayesian Extensions of the Black-Litterman Model

To reduce the estimation error portfolio managers encounter when determining their portfolio allocation, this paper focuses on the application of the Black-Litterman model within a more extensive Bayesian framework. This framework allows to extend the existing Black- Litterman models to include prior specifications on the covariance matrix, hierarchical scaling parameter and non-normal data. Applying these models to a S&P 500 data set, the models using a Bayesian framework, taking into account additional parameter uncertainty, are in some challenging environments able to obtain a superior result compared to the benchmarks, making it a useful framework for portfolio managers. However, the models can not consistently outperform the less complex benchmark models.