Modern Portfolio Theory

Portfolio risk depends on how assets move together

Modern portfolio theory, or MPT, treats the portfolio as a whole rather than judging each security in isolation. Expected return is a weighted average of component expected returns, while portfolio risk depends on each asset’s variance and how returns move together. Two volatile assets can form a less volatile combination if their returns are not perfectly correlated.

Correlation ranges from −1 to +1. A correlation of +1 means returns move perfectly together in the measured data; combining them does not produce diversification benefits from imperfect co-movement. A correlation near zero means little linear relationship in the sample. A negative correlation means returns tended to move in opposite directions. Correlations are estimates and can change, especially in stressed markets.

MPT does not say that diversification eliminates all loss. A portfolio can remain exposed to market-wide declines, interest-rate shifts, inflation, credit events, or unexpected correlations. It explains how portfolio construction can reduce some risk while maintaining expected return, given assumptions about the assets and the investor’s objectives.

Expected return is a weighted average

For a portfolio with assets i, expected return equals the sum of each asset’s weight multiplied by its expected return. If 60% is invested in an asset expected to return 8% and 40% in an asset expected to return 4%, portfolio expected return is 0.60 × 8% + 0.40 × 4% = 6.4%.

This calculation is linear in weights, but actual return is uncertain. The 8% and 4% values are estimates, not guarantees. If the assets’ actual returns differ, the realized portfolio return will differ. Rebalancing, fees, taxes, and timing of contributions or withdrawals can also change an investor’s realized outcome.

The weights should sum to 100% for a fully invested unlevered portfolio. If borrowing or cash holdings are permitted, the interpretation changes. A problem may use percentages or decimals; convert consistently. A frequent error is adding asset returns without weighting them or using the wrong portfolio proportions.

Two-asset portfolio variance

For two risky assets, portfolio variance depends on the weights, each asset’s variance, and their covariance. In standard notation: σp² = w1²σ1² + w2²σ2² + 2w1w2Cov(R1,R2). Since covariance equals correlation times the two standard deviations, the final term can also be written 2w1w2ρ12σ1σ2.

Assume two assets each have 10% standard deviation, equal 50% weights, and correlation of 0. The portfolio variance is 0.25 × 0.01 + 0.25 × 0.01 + 0 = 0.005, so standard deviation is √0.005, or about 7.07%. The portfolio is less volatile than either asset in this simplified example because their returns are not perfectly correlated.

If the same assets have correlation +1, variance is 0.01 and portfolio standard deviation is 10%; the diversification benefit disappears. If correlation is −1 and the standard deviations and weights match, the simplified portfolio variance can be zero. Perfect negative correlation is rare in practice, and estimates do not guarantee future offsets.

Worked example with unequal risk

Suppose Asset A has an expected return of 7% and standard deviation of 12%; Asset B has an expected return of 4% and standard deviation of 6%. A portfolio invests 60% in A and 40% in B. Expected return is 0.60 × 7% + 0.40 × 4% = 5.8%.

If correlation is 0.25, covariance is 0.25 × 0.12 × 0.06 = 0.0018. Portfolio variance is (0.60² × 0.12²) + (0.40² × 0.06²) + 2 × 0.60 × 0.40 × 0.0018. The first terms are 0.005184 and 0.000576; the covariance term is 0.000864. Total variance is 0.006624, so standard deviation is about 8.14%.

A weighted average of individual standard deviations would be 0.60 × 12% + 0.40 × 6% = 9.6%, which is not the portfolio standard deviation. Risk does not average linearly because covariance matters. This distinction is central to MPT and a common exam calculation trap.

The efficient frontier

The investment opportunity set contains the portfolios that can be formed from the available assets under specified constraints. The minimum-variance frontier shows the lowest-risk portfolios for each expected return. The efficient frontier is the upper portion that offers the highest expected return for a given level of risk, or the lowest risk for a target expected return.

A portfolio below the efficient frontier is inefficient in the mean-variance sense if another available portfolio has higher expected return for the same risk or lower risk for the same expected return. A portfolio on the frontier is not automatically suitable for every client. The investor’s risk preferences, liabilities, liquidity needs, and constraints help determine which efficient portfolio is appropriate.

The global minimum-variance portfolio is the portfolio with the lowest variance among the available risky-asset combinations. It is one point on the frontier, not necessarily the portfolio with the highest return or the best choice for every investor. A highly risk-averse investor may prefer a low-risk point; another may select a higher-return point, subject to capacity.

Add a risk-free asset and capital allocation line

In the basic MPT framework, a risk-free asset can be combined with a risky portfolio. The capital allocation line shows combinations of the risk-free asset and a risky portfolio. Its slope is the portfolio’s excess expected return over the risk-free rate divided by its standard deviation, often described as the Sharpe ratio for that line.

Under the CAPM assumptions, all investors hold a combination of the risk-free asset and the same market portfolio, with different weights depending on risk preference. The line tangent to the risky-asset efficient frontier is the capital market line. This result depends on strong assumptions, including common expectations and the ability to borrow or lend at a common risk-free rate.

In real markets, investors may face different borrowing rates, taxes, transaction costs, restrictions, and forecasts. The market portfolio is not directly observable. The capital market line is therefore a theoretical benchmark, not a promise that every investor can access the same tangent portfolio or risk-free borrowing terms.

Risk preferences and the optimal portfolio

MPT can identify efficient portfolios, but it does not by itself identify which one a client should own. The client’s preferences over risk and return matter. A risk-averse investor may choose a lower-volatility portfolio even when another efficient portfolio offers higher expected return. The choice reflects the client’s utility, goals, and capacity for loss.

An indifference curve represents combinations of expected return and risk that provide the same utility to an investor. The investor’s optimal portfolio is often described as the point where the highest attainable indifference curve touches the efficient frontier or capital allocation line. The model formalizes the tradeoff; it does not measure a client’s preferences without reliable information.

A client may have multiple goals and constraints that are not captured by one utility curve. A near-term purchase, tax limits, legal restrictions, and liquidity needs can affect the feasible set. Portfolio optimization should be interpreted alongside the client profile rather than treated as an automatic recommendation engine.

Assumptions and estimation risk

MPT calculations require expected returns, standard deviations, and correlations or covariances. These inputs are forecasts or historical estimates, not known facts. Small changes in the inputs can produce large changes in optimized weights. If the expected return estimate is wrong, the model may recommend a concentrated portfolio based on a false precision.

Historical covariance can change when inflation, rates, business cycles, or market structure shift. Correlations may rise during crises, reducing diversification just when it is most needed. A model that uses long historical averages may miss structural changes; a short sample may be dominated by unusual events. Analysts should test multiple scenarios and understand the source of each input.

Basic mean-variance analysis usually treats variance as the measure of risk and expected return as the reward. Investors may care more about downside loss, shortfall probability, liquidity, or permanent impairment than symmetric volatility. Alternative risk measures and goal-based approaches can supplement MPT. The model remains useful, but it does not capture every investor concern.

Worked example: comparing two portfolios

Portfolio X has expected return 6% and standard deviation 8%. Portfolio Y has expected return 7% and standard deviation 12%. If an investor’s target is 6%, Portfolio X achieves that expected return with lower modeled risk. If the investor seeks 7%, Y offers the higher expected return but also greater volatility. Neither comparison indicates which portfolio is suitable without the client’s goal and risk capacity.

Suppose a third portfolio has expected return 5.5% and standard deviation 10%. If Portfolio X has both a higher expected return and lower risk, the third portfolio is dominated and lies below the efficient frontier in this simplified comparison. If no other portfolio has both advantages, it may still be efficient for some risk-return target.

Expected return and standard deviation do not describe the full distribution. Two portfolios with the same mean and variance can have different tail risk, liquidity, tax treatment, and concentration. A client who cannot tolerate a severe shortfall may need additional analysis beyond the mean-variance chart.

MPT and diversification across asset classes

MPT encourages investors to consider how assets interact, not merely how each performed on its own. Combining stocks, bonds, cash, real estate, commodities, or other exposures can change expected return and risk depending on covariance. The asset categories are not automatically diversifying; the actual holdings and relationships matter.

A fund may own many securities yet be concentrated in one sector or risk factor. A portfolio can hold several funds whose top holdings overlap. Look through to underlying exposures and evaluate correlations under more than one period. If relationships shift, the portfolio’s risk can differ from the model estimate.

Diversification is a tool for reducing some unsystematic risk, not a guarantee against loss. A diversified portfolio can decline when markets fall broadly. Investors also face inflation, interest-rate, credit, currency, and liquidity risks. MPT helps describe tradeoffs but should not be marketed as a method that eliminates risk.

Practical uses and limitations for an adviser

An adviser can use MPT concepts to compare asset mixes, explain diversification, evaluate covariance, and discuss risk-return tradeoffs. The calculations can help show why a lower-correlated asset may improve a portfolio even if its standalone volatility is high. They can also show when a security adds little diversification because it moves closely with existing holdings.

The adviser should explain that model outputs depend on inputs and constraints. A portfolio optimizer may use historical expected returns or theoretical forecasts that are uncertain. The client should understand whether short selling, leverage, taxes, transaction costs, and liquidity limits are included. A precise percentage allocation can create an illusion of certainty if the inputs are weak.

Optimization should be constrained by the client’s objectives, restrictions, time horizon, and risk capacity. A mathematically efficient portfolio may be inappropriate if it creates unacceptable drawdown, illiquidity, or concentration. The purpose of the model is to inform judgment, not replace it.

Common MPT errors

One common error is calculating portfolio standard deviation as a weighted average of individual standard deviations. Another is ignoring covariance or using correlation directly without multiplying by both standard deviations. A third is confusing variance with standard deviation; variance is in squared units, while standard deviation returns to percentage units.

A second error is thinking the global minimum-variance portfolio is automatically optimal for every investor. A third is assuming a portfolio on the frontier is guaranteed to earn its expected return. A fourth is assuming diversification prevents losses in a broad market decline.

Finally, optimization can be sensitive to assumptions. If an output recommends a very large allocation to one asset based on a small expected-return difference, test the assumptions and impose client-specific constraints. A model’s mathematical answer is only as credible as its data and objectives.

A repeatable solution method

For expected return, multiply each asset’s expected return by its portfolio weight and add the results. For two-asset variance, calculate each weighted variance term and the covariance term using correlation times both standard deviations. Sum the terms, then take the square root to obtain standard deviation.

For conceptual questions, identify whether the question asks about an efficient frontier, minimum variance, correlation, or investor preference. The frontier describes available risk-return combinations; the investor profile selects among them. Correlation affects diversification; expected returns alone do not determine portfolio risk.

State assumptions and interpret the result in context. If an input is hypothetical, say so. If the client has constraints or a short horizon, note that the mathematically efficient portfolio may not be appropriate. This method keeps the calculation accurate while preserving the model’s real-world limits.