The efficient frontier: how to construct portfolios with optimal risk-adjusted returns

The efficient frontier is the set of portfolios that deliver the highest possible expected return for every given level of risk—the mathematical boundary beyond which no further improvement is achievable with the available assets. Every portfolio on it is optimal; every portfolio below it wastes risk.

What the efficient frontier is

The efficient frontier was introduced by Harry Markowitz in his 1952 paper Portfolio Selection, as a direct output of Modern Portfolio Theory. Before Markowitz, the idea that combining assets could reduce overall portfolio risk without reducing expected return was not formalised. The efficient frontier made that insight geometric: plot expected return on one axis, volatility on the other, and the efficient frontier appears as a curved arc at the top-left boundary of all possible portfolios.

The frontier is derived from the full universe of possible weight combinations across the selected assets. Each point on the frontier corresponds to a specific set of weights—a portfolio. Portfolios below the frontier are suboptimal: for the same risk, a higher-return portfolio exists on the frontier. Portfolios above the frontier are impossible given the available assets and their correlations.

The shape of the frontier is determined primarily by the pairwise correlations between assets. When assets are imperfectly correlated, the frontier bows outward to the left, reflecting the diversification benefit. The lower the average correlation, the more pronounced the bow—and the greater the risk reduction that diversification can achieve.

How it works

Constructing the efficient frontier requires three inputs for each asset: expected return, expected volatility (standard deviation of returns), and the correlation between every pair of assets in the portfolio. From these, it is possible to calculate the expected return and volatility of any weighted combination.

The mathematical process—mean-variance optimisation—sweeps through target return levels and, for each, identifies the asset weights that minimise portfolio volatility. Plotting these minimum-variance portfolios at each return level traces the frontier. The portfolio at the leftmost point of the frontier—with the lowest achievable volatility—is called the minimum-variance portfolio. The portfolio that maximises the Sharpe ratio (return per unit of risk) is the tangency portfolio, found at the point where the capital market line just touches the frontier.

In practice, investors choose a point on the frontier that matches their risk tolerance. A conservative investor selects a lower-return, lower-risk position; an investor with a higher risk appetite moves up the frontier. All choices on the frontier are efficient; the choice between them is a preference, not a quality judgement.

What the evidence shows

Markowitz's original paper and subsequent empirical work have consistently confirmed that well-diversified multi-asset portfolios outperform concentrated ones on a risk-adjusted basis over long horizons. Blume and Friend (1975) demonstrated that the benefits of diversification are largely exhausted well before portfolios become fully diversified by stock count alone—a finding that motivates asset-class diversification rather than simple stock picking.

More recent evidence from Ilmanen and Kizer (2012, The Death of Diversification Has Been Greatly Exaggerated) showed that diversification across asset classes—equities, fixed income, commodities, and alternatives—delivers more durable risk reduction than diversification within a single asset class. During the 2000–2002 and 2008–2009 downturns, multi-asset portfolios suffered significantly lower maximum drawdowns than equity-only portfolios, even after accounting for the correlation increases that occur during stress periods.

Limitations and trade-offs

The efficient frontier is only as reliable as its inputs. Expected returns, volatilities, and correlations are all estimated from historical data—and all are unstable over time. Small errors in expected return estimates can shift the frontier substantially and produce portfolios that appear optimal on paper but perform poorly out of sample. This problem—known as estimation error—means that portfolios appearing to lie on the efficient frontier based on historical data may not do so in real time. Correlations, in particular, tend to converge towards one during market crises, compressing the diversification benefit precisely when it is most needed.

A second practical constraint is the sensitivity of optimal weights to minor input changes. The optimiser may recommend extreme positions—very high weights in one asset, near-zero weights in others—that reflect noise in the data rather than genuine opportunity. Position constraints and robust estimation techniques are commonly applied to address this, but they move the portfolio away from the theoretically optimal point on the frontier.

Efficient frontier in pfolio

pfolio constructs the efficient frontier using historical return data for each asset selected and places your portfolio at the point that maximises the Sharpe ratio, subject to position constraints. You can adjust the asset mix using the asset selection and allocation controls and observe how the frontier shifts. For an overview of the three optimisation methods available, see the how we build portfolios help article.

Related articles

• Modern Portfolio Theory explained: how Markowitz's framework guides portfolio construction
• Mean-variance optimisation: the algorithm behind optimal portfolio construction
• Portfolio diversification: why spreading risk across asset classes beats spreading across stocks
• Sharpe ratio explained: measuring risk-adjusted portfolio returns