The Black-Litterman model is a mathematical framework for portfolio allocation developed in 1990 at Goldman Sachs by Fischer Black and Robert Litterman.71 It addresses the challenges institutional investors face when applying modern portfolio theory in practice. Black and Litterman refined the mean-variance (MV) model by integrating it with the capital asset pricing model (CAPM), resulting in the Black-Litterman (BL) model. This model incorporates strategic asset allocation that embeds investors’ global views without assuming that expected returns are always at equilibrium, as posited by CAPM.
The Black-Litterman model allows portfolio managers to express views about portfolios rather than requiring a complete vector of expected returns on all assets. It acknowledges that market imbalances can drive returns towards equilibrium, enabling investors to potentially achieve higher returns by integrating their views with market information.71 In the simplest context—when there is no benchmark or constraints—the optimal portfolio is intuitive. It consists of deviations from market capitalization weights in the directions of portfolios about which views are expressed.
The model starts with an asset allocation based on the equilibrium assumption (assets will perform in the future as they have in the past) and then modifies that allocation by incorporating the investor’s opinions regarding future asset performance. The mechanism is a weighted average. The model takes the CAPM equilibrium returns — the returns implied by current market-capitalization weights — as a Bayesian prior, and treats each investor view as new evidence tagged with its own confidence level. The output is a blended expected-return vector sitting between the two: a high-confidence view pulls the blend strongly toward itself, a tentative one barely moves it.72 Because the inputs are deliberate views with explicit confidence weights rather than a brittle full vector of point estimates, the result is far less sensitive to estimation error, and the resulting weights are more intuitive and better behaved than those from raw mean-variance optimization. That stability is the practical failure BL was built to fix, and it makes the model well suited to investors willing to act on well-evidenced views.73