Modern Portfolio Theory explained: how Markowitz's framework guides portfolio construction
Modern portfolio theory (MPT) is a framework for constructing portfolios that maximise expected return for a given level of risk. Introduced by Harry Markowitz in his 1952 paper, Portfolio Selection, it replaced the intuitive but unstructured approach of selecting individual assets with a mathematically rigorous method of combining them—one that treats the portfolio as a whole rather than as a collection of separate bets.
What modern portfolio theory is
Before Markowitz, portfolio construction was largely a matter of judgement. Investors sought strong companies or assets and accumulated them, with diversification understood only loosely—not putting all eggs in one basket. Markowitz’s contribution was to formalise the relationship between individual asset risk and the risk of a combined portfolio, and to show that the two are not the same thing.
The central insight is that what matters is not the risk of each asset in isolation, but how assets move relative to one another. An asset that is volatile on its own may reduce overall portfolio risk if it tends to move independently of—or in the opposite direction to—the other assets held. This relationship is captured by correlation.
MPT assumes investors are rational and risk-averse: given two portfolios with the same expected return, a rational investor will prefer the one with lower risk. It follows that there is an optimal way to combine assets—one that extracts the maximum return per unit of risk accepted. Every other combination is, by definition, inefficient.
How it works
MPT quantifies each asset’s expected return and standard deviation (volatility), and the pairwise correlations between all assets in the portfolio. From these three inputs, it calculates the expected return and risk of any given combination. Plotting all possible combinations generates what Markowitz called the efficient set—an arc of portfolios that deliver the highest possible expected return for each level of risk accepted.
The practical step from this geometry is portfolio optimisation. Mean-variance optimisation—the mathematical process derived directly from MPT—identifies which portfolio on the efficient set is best suited to a given investor’s risk tolerance. The key inputs are: expected returns, expected volatility, and the correlation matrix between assets. Correlation is the variable investors most often underestimate; two assets that appear well-diversified in isolation may move together precisely when diversification is most needed.
A portfolio on the efficient set is described as mean-variance efficient. A portfolio below it accepts unnecessary risk for the return it delivers. A portfolio above it does not exist—no combination of the available assets can achieve it.
What the evidence shows
Markowitz’s framework has been tested extensively since its publication. Its core prediction—that diversification across low-correlated assets improves the risk-return profile—has been consistently supported by the evidence. Portfolios constructed using MPT have historically delivered better risk-adjusted returns than concentrated portfolios over long periods, particularly across multi-asset-class portfolios that include equities, fixed income, and commodities.
The Sharpe ratio, which measures return per unit of risk, is in many respects the quantitative expression of what MPT seeks to maximise. Portfolios closer to the efficient frontier tend to produce higher Sharpe ratios over time, though this relationship depends heavily on the accuracy of the inputs and is not guaranteed in any single period.
Limitations and trade-offs
MPT’s practical limitations are significant and well-documented. The framework depends entirely on the quality of its inputs: expected returns, volatility estimates, and correlation coefficients. All three are estimated from historical data, and all three change over time.
The most serious limitation is estimation error. Small changes in expected return assumptions can produce dramatically different “optimal” portfolios—a problem sometimes described as error maximisation. If input estimates are inaccurate, the optimised portfolio may perform no better than a simpler equal-weight allocation. This sensitivity is the primary motivation for alternative optimisation methods, including Hierarchical Risk Parity, developed specifically to address the instability of the covariance matrix that undermines MPT in practice (López de Prado, 2016, Building Diversified Portfolios that Outperform Out-of-Sample).
A second limitation is the assumption of a stable correlation structure. In practice, correlations tend to rise during market stress—precisely when diversification is most needed. A portfolio that appears well-diversified under normal conditions may offer less protection during a crisis than the model suggests. Investors should treat MPT as a framework for structured thinking about portfolio construction, not as a guarantee of optimal outcomes.
Modern portfolio theory in pfolio
pfolio implements mean-variance optimisation as its default portfolio construction method, placing portfolios on the efficient frontier given the selected assets and historical data. Two alternatives are also available—Hierarchical Risk Parity and equal weight—for investors who prefer methods less sensitive to estimation error in the covariance matrix. All three are described in the how we build portfolios help article. You can explore the results across model portfolios at pfolio.io/portfolios.
Related articles
Disclaimer
Get started now
