The Black-Litterman model, developed by Fischer Black and Robert Litterman at Goldman Sachs in 1990-92, is a way to build expected returns for portfolio optimization by starting from the returns implied by market weights and adjusting them for the investor's views, in proportion to confidence in each view.
The idea
- Start from equilibrium. Assume the market portfolio is optimal and work out the expected returns that would make it so (reverse optimization).
- Add views. For example, Stock A will beat Stock B by 2%, with a stated level of confidence.
- Blend. The model combines the two, weighting each by its uncertainty, giving a set of expected returns.
- Optimise using these returns. See mean-variance optimization.
An illustrative example
| Equilibrium | View | Blended (low confidence) | Blended (high confidence) | |
|---|---|---|---|---|
| Expected return of A | 7% | 9% | 7.4% | 8.6% |
The more confident the view, the closer the blended return is to it. Other assets' expected returns also shift if they are correlated with A. The numbers are made up.
Why use it
- It avoids the extreme, concentrated weights that unconstrained optimization can create.
- With no views, it gives back the market weights, a sensible neutral starting point.
- Views can be partial and relative, instead of a forecast for every asset.
Limits
- The results depend on the confidence levels and a scaling parameter, which are judgement calls.
- It assumes the market portfolio is a good anchor.
- Views from a quantitative signal need translating into returns and confidence, and poor views give poor portfolios. See information coefficient.
Estimation of risk still matters. See covariance matrix.
Frequently asked questions
What is the Black-Litterman model?
It is a method that blends market-implied expected returns with an investor's views to give stable inputs for portfolio optimization.
Why was Black-Litterman developed?
To fix the unstable, concentrated portfolios that mean-variance optimization gives when expected return inputs are noisy.
What are views in Black-Litterman?
Statements such as a stock's expected return, or that one asset will beat another by a given amount, each with a confidence level.
What happens if there are no views?
The model returns the market-implied returns, so the optimal portfolio is the market portfolio.