Markowitz mean–variance model
Finds efficient portfolios from expected returns, variances and covariances.
How it works
Mean–variance optimisation combines expected returns, variances and covariances to find efficient portfolios under stated constraints. The portfolio mean is the weighted sum of asset returns. Variance depends on weights and joint movements, so allocating to a volatile asset can still help if its returns offset other positions.
Worked example
Take two assets with volatility 20% each and equal portfolio weights. At zero correlation their portfolio volatility is about 14.14%. At correlation 0.5 it is about 17.32%. At perfect positive correlation it is 20%. Those outcomes use identical individual volatilities and demonstrate why the covariance estimate matters.
Validation and limitations
Stress expected returns and correlations, use a coherent covariance matrix and check sensitivity to position limits. Estimated means often produce unstable optimal weights. Benchmark against simple allocations and assess realised costs, turnover and concentration out of sample.