A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
The efficient frontier is the set of portfolios that maximize return for a given variance (or minimize variance for a given return). Multi-asset efficient frontiers require solving constrained optimization problems using covariance matrices and expected returns.
Use historical returns to estimate covariance matrix. Solve for optimal weights across multiple assets subject to constraints (e.g., no short-selling). Plot the resulting efficient frontier and compare to naive portfolios.
From two-asset portfolio optimization, you know that combining two assets reduces portfolio risk below the weighted average of their individual risks whenever correlation is less than 1. The diversification benefit grows as correlation falls. With N assets, the same logic applies — but now there are N(N-1)/2 pairwise correlations to exploit. The efficient frontier is the systematic answer to: given all available assets and all their pairwise relationships, which portfolios make efficient use of diversification, and which waste it?
The central object in multi-asset optimization is the covariance matrix Σ, an N×N matrix where element σ_{ij} captures how assets i and j move together. Portfolio variance is σ²_p = w'Σw, where w is the vector of portfolio weights. This quadratic form means that when you add an asset to a portfolio, its contribution to portfolio risk depends primarily on how it covaries with everything already in the portfolio, not on its standalone variance. An asset with high individual volatility but low correlation with your existing portfolio can actually reduce total risk — something the two-asset intuition already hinted at but becomes even more powerful with many assets.
To build the efficient frontier, you solve a parametric optimization: minimize w'Σw subject to w'μ = μ̄ (achieve a target expected return) and w'1 = 1 (weights sum to 1). By varying μ̄ over all feasible values, you trace out the full set of minimum-variance portfolios. In mean-variance space, this set forms a parabola (or a hyperbola in standard deviation-return space). The minimum variance portfolio is the leftmost point on this frontier — the lowest-risk portfolio achievable. The efficient frontier is the upper portion of the curve, above the minimum variance portfolio: for any lower-return portfolio on the frontier, there exists an efficient portfolio with the same risk and higher return, so rational investors won't hold the lower portion.
Comparing efficient frontier portfolios to naive portfolios illustrates the cost of ignoring correlations. An equally-weighted portfolio, or a simple 60/40 stocks-bonds allocation, lies inside the frontier — it is inefficient in the sense that you could rearrange the same assets to get either higher expected return for the same risk, or lower risk for the same expected return. The gap between where naive portfolios sit and the frontier quantifies what diversification is worth. For a moderately sized portfolio of, say, 20 assets with diverse correlations, the efficiency gains from optimization over equal-weighting can be substantial — reduced volatility without sacrificing expected return.
In practice, the frontier is only as good as its inputs. Expected returns are the most problematic input: historical average returns are noisy estimates of forward-looking expected returns, and small errors in expected returns translate into large swings in optimal weights (the portfolio optimizer is "error-maximizing" with respect to expected return estimates). Covariance matrices estimated from historical data can also be ill-conditioned when N is large relative to the number of observations. Practitioners address this through shrinkage estimators (pulling individual estimates toward a structured prior), factor models (constraining the covariance matrix to lie in a lower-dimensional space), or constraints on short selling and concentration. These modifications trade theoretical efficiency for robustness in estimation — a real-world concession that the elegant theory of the frontier requires messy adjustments to be usable.