Mean-Variance Optimization

AdvancedPortfolio Management3 min read

Quick Definition

A mathematical framework for constructing portfolios that maximize expected return for a given level of risk using asset correlations.

What Is Mean-Variance Optimization?

What Is Mean-Variance Optimization?

Mean-Variance Optimization (MVO) is the mathematical foundation of Modern Portfolio Theory, developed by Harry Markowitz in 1952. It constructs portfolios by finding the optimal combination of assets that maximizes expected return for a given risk level (or minimizes risk for a target return), using expected returns, volatilities, and correlations.

The MVO Framework

Inputs Required:

InputDescriptionExample
Expected Returns (μ)Forecasted return per assetStocks: 10%, Bonds: 4%
Standard Deviation (σ)Volatility per assetStocks: 16%, Bonds: 5%
Correlation Matrix (ρ)Co-movement between assetsStocks-Bonds: -0.2

How It Works

  1. Define the universe of investable assets
  2. Estimate expected returns, volatilities, and correlations for each asset
  3. Generate thousands of possible portfolios with different weight combinations
  4. Plot the efficient frontier: The curve of optimal portfolios
  5. Select the portfolio that matches your risk tolerance

The Efficient Frontier

The efficient frontier represents all portfolios where:

  • No portfolio offers higher return at the same risk level
  • No portfolio offers lower risk at the same return level
  • Any portfolio below the frontier is suboptimal

Example Calculation

Two assets: Stocks (μ=10%, σ=16%) and Bonds (μ=4%, σ=5%), correlation = -0.2

Stock WeightBond WeightExpected ReturnPortfolio Risk
100%0%10.0%16.0%
70%30%8.2%10.9%
40%60%6.4%6.1%
0%100%4.0%5.0%

The 40/60 portfolio achieves 6.4% return with only 6.1% risk—better risk-adjusted return than 100% bonds.

Limitations

  • Garbage in, garbage out: Results highly sensitive to return estimates
  • Unstable weights: Small changes in inputs produce dramatically different allocations
  • Backward-looking: Historical data may not predict future relationships
  • Ignores fat tails: Assumes normal distribution of returns

Why It Matters

MVO provides the theoretical basis for portfolio construction used by robo-advisors, pension funds, and institutional investors. Understanding its principles helps investors appreciate why diversification works and how correlation drives portfolio risk.

Formula

Formula

Minimize: w'Σw subject to w'μ = target return, Σwi = 1

Mean-Variance Optimization Example

  • 1Finding that a 60/40 stock-bond portfolio has better risk-adjusted returns than 100% stocks
  • 2A robo-advisor using MVO to determine optimal allocation across 7 asset classes