Mean-Variance Optimization
Mean-variance optimization is the mathematical framework introduced by Harry Markowitz that identifies the portfolio with the highest expected return for a given level of risk (variance), or equivalently, the lowest risk for a target expected return.
Mean-variance optimization (MVO) forms the mathematical core of Modern Portfolio Theory (MPT). Markowitz demonstrated in 1952 that an investor who cares only about the mean (expected return) and variance (risk) of portfolio returns can construct the set of efficient portfolios — the efficient frontier — by solving a quadratic optimisation problem. Every point on the efficient frontier dominates all portfolios to its right (higher risk for the same return) and below (lower return for the same risk).
The inputs to MVO are the expected return vector, the covariance matrix of asset returns, and optionally a set of constraints (e.g., no short-selling, maximum weight per stock). In practice, these inputs must be estimated from historical data or forward-looking models, and the final portfolio weights are extremely sensitive to small estimation errors — particularly in expected returns. This input sensitivity is the central practical criticism of MVO.
For a portfolio of N Indian stocks, the covariance matrix has N(N+1)/2 unique entries. Estimating these accurately requires substantial historical data. For Nifty 500 constituents, this means estimating over 125,000 covariance pairs, a statistically noisy exercise. Practitioners mitigate this using shrinkage estimators (Ledoit-Wolf), factor models (Fama-French three-factor or Barra models), or by constraining the universe to a smaller set.
In the Indian mutual fund industry, index construction committees use MVO-adjacent approaches in smart beta strategy design. The Nifty Alpha Low Volatility 30 index, for example, blends momentum and low-volatility criteria in a way that implicitly approximates a risk-adjusted optimisation. SEBI regulations for quantitative funds require transparent disclosure of the optimisation methodology used.
Key assumptions of MVO — investor rationality, normal return distributions, no transaction costs — are violated in practice. Fat-tailed return distributions (kurtosis > 3) are common in Indian individual stocks, particularly during earnings shocks or regulatory events. Extensions such as mean-CVaR (Conditional Value-at-Risk) optimisation and robust optimisation methods address some of these limitations, and professional portfolio managers increasingly adopted these extensions in India's growing quantitative investing ecosystem.