Model contest and portfolio performance: Black-Litterman versus factor models

The aim of this article is to discuss and compare four models: single index model, Fama French three-factor model, Carhart four-factor model and the Black-Litterman model in the context of twenty largest capitalization stocks listed on NASDAQ and NYSE. It was found that among the three factor models, the Carhart four-factor model produced the best fitness of data with the highest adjusted R². When testing the portfolio performance built on the four models across the testing periods, the Black-Litterman model was superior to all other three models. The portfolio constructed based on the Black-Litterman framework generated relatively higher return with relatively lower risk than other models with a Sharpe ratio of 0.4437. Several recommendations can be made to improve the performance of portfolios. Firstly, Black-Litterman Model is favored when constructing portfolio rather than traditional optimization approach. Secondly, view return and view covariance matrix should be generated based on thorough equity research rather than merely quantitative method such as data mining. Finally, parameters involved in the Black-Litterman framework should be generated with care; in particular, τ and c can be combined into one parameter.