The Black-Litterman model: incorporating investor views into mean-variance optimisation
Standard mean-variance optimisation is notoriously unstable: small changes in expected return inputs produce wildly different optimal portfolios. The root cause is that MVO amplifies estimation errors in expected returns. Black and Litterman (1990) proposed a solution: start not with subjective return estimates, but with the returns implied by the market portfolio under the CAPM. This provides a stable, diversified starting point. The investor then expresses views as deviations from that equilibrium, and the model blends the equilibrium and the views in proportion to the investor's confidence in each. The result is a set of blended expected returns that are less extreme than raw subjective views and more informative than pure equilibrium returns.
What the Black-Litterman model does
The model has two inputs: the market-implied equilibrium returns (derived by reverse-engineering the CAPM from market capitalisation weights) and the investor's views (expressed as absolute or relative return expectations for specific assets or asset classes). The investor also specifies a confidence level for each view, which determines how much the view shifts the equilibrium. The model outputs a set of posterior expected returns—a Bayesian blend of equilibrium and views—which can then be used as inputs to standard MVO. The resulting optimal portfolio is considerably more stable than one built from raw return estimates and tends to look more like the market portfolio when views are weak.
The framework
Posterior returns = [(τΣ)⁻¹ + P'Ω⁻¹P]⁻¹ × [(τΣ)⁻¹π + P'Ω⁻¹q]
Where π is the vector of equilibrium returns (implied by market weights), P is the matrix of views (each row expressing one view), q is the vector of expected returns under each view, Ω is the diagonal matrix of view uncertainty, Σ is the covariance matrix, and τ is a scalar controlling the weight given to the prior equilibrium. The formula is Bayesian: the posterior is a precision-weighted average of the prior (equilibrium) and the likelihood (views).
How to interpret Black-Litterman outputs
A view that deviates only modestly from equilibrium, held with low confidence, produces a posterior close to the equilibrium return—the model stays near the market portfolio. A view held with high confidence produces a posterior that tilts toward that view, generating a portfolio with meaningful active exposure. The model's key practical advantage is that weak or uncertain views produce small tilts, so the resulting portfolio is never extreme unless the investor has strong, precise views. This prevents the over-concentration problem that plagues unconstrained MVO. Investors using the model must decide how to quantify their confidence in each view, which introduces subjectivity at a different level than raw return estimation.
Limitations
The Black-Litterman model does not eliminate the need for judgement—it relocates it. Instead of estimating raw expected returns, the investor must estimate confidence levels and express views in a structured format. The equilibrium prior depends on the CAPM, which assumes that the market portfolio is mean-variance efficient; in practice, market capitalisation weights embed momentum and concentration biases that the model inherits. The model also requires a covariance matrix estimate, which has its own instability problems addressed by covariance estimation techniques. Implementing Black-Litterman correctly requires meaningful quantitative infrastructure.
Black-Litterman in pfolio
The Black-Litterman model is not currently implemented in pfolio. The platform's portfolio optimiser uses standard mean-variance optimisation as the default, with Hierarchical Risk Parity and equal weight as alternatives. Investors who want to apply Black-Litterman-style views would need to compute the implied expected returns externally and use those as inputs to a custom portfolio. The construction methodology pfolio actually uses is documented at how we build portfolios.
Related articles
Disclaimer
Get started now
