Modern Portfolio Theory and Diversification

Introduction

Investment decisions are rarely made by selecting securities in isolation. A security with an attractive expected return may also expose an investor to substantial uncertainty, while a security with modest individual risk may improve the risk–return profile of an entire portfolio. Modern Portfolio Theory (MPT), developed principally from Harry Markowitz’s work, provides a formal framework for analysing this trade-off.

The central insight of MPT is that portfolio risk depends not only on the risk of individual assets but also on how their returns move together. Consequently, diversification can reduce risk without necessarily reducing expected return. The theory provides the foundations for efficient frontiers, minimum-variance portfolios, utility-based investment choice, asset allocation, and much of contemporary portfolio management.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Calculate expected return, variance, standard deviation and covariance.
2. Explain how correlation affects portfolio risk.
3. Construct and interpret the efficient frontier and minimum-variance portfolio.
4. Apply investor utility and risk-aversion concepts to portfolio choice.
5. Distinguish systematic from unsystematic risk.
6. Evaluate the practical benefits and limitations of diversification.
7. Apply MPT to a professional portfolio-management decision.

1. Foundations of Modern Portfolio Theory

MPT treats an investment as a probability distribution of future returns. An investor evaluates a portfolio primarily through two dimensions:

• Expected return: the probability-weighted average of possible returns.
• Risk: commonly represented by variance or standard deviation of returns.

MPT is a single-period, mean–variance framework. It is most directly applicable when investors have relatively similar expectations about returns, risk and correlations, and when return distributions can reasonably be summarised by their first two moments.

1.1 Expected return

If an asset can produce returns R₁, R₂, …, Rₙ with probabilities p₁, p₂, …, pₙ, its expected return is:

Here, pᵢ is the probability of outcome i, Rᵢ is the return in that outcome, and ∑pᵢ = 1. Expected return is expressed as a percentage or decimal return over the relevant investment period.

For a portfolio containing assets with weights wᵢ:

The weights generally sum to one for a fully invested portfolio:

Weights may be negative where short selling is permitted.

1.2 Variance and standard deviation

Variance measures the average squared deviation from expected return:

Standard deviation is the positive square root of variance:

Both variance and standard deviation measure dispersion, but standard deviation is easier to interpret because it has the same units as return. If returns are measured in percentages, standard deviation is also measured in percentage points.

For historical observations, sample variance is commonly estimated as:

where T is the number of observations and R̄ is the sample mean. The denominator T − 1 provides an unbiased estimator under standard random-sampling assumptions.

1.3 Covariance and correlation

Covariance indicates whether two asset returns tend to move together:

Positive covariance indicates a tendency to move in the same direction; negative covariance indicates a tendency to move in opposite directions. Covariance is expressed in squared return units and is therefore difficult to compare across asset pairs.

Correlation standardises covariance:

Correlation lies between −1 and +1. A value of +1 indicates perfect positive co-movement, 0 indicates no linear association, and −1 indicates perfect negative linear co-movement. Zero correlation does not imply independence or absence of all forms of dependence.

2. Portfolio Risk

For a two-asset portfolio, portfolio variance is:

Using correlation, the same equation becomes:

The final term is the covariance contribution. It explains why portfolio risk is not simply the weighted average of individual risks. For n assets, the general matrix representation is:

where w is the vector of portfolio weights, Σ is the variance–covariance matrix, and w′ is the transpose of w. This formula assumes that the covariance matrix is measured over a consistent time horizon and that weights and returns are defined consistently.

Worked Example 1: Two-asset diversification

Assume:

Suppose correlation is 0.20. Expected portfolio return is:

Covariance equals:

Portfolio variance is:

Therefore:

The portfolio’s standard deviation is 10.14%, which is below Asset A’s 15% and demonstrates the effect of imperfect positive correlation.

3. The Benefits and Limits of Diversification

Diversification combines assets whose returns do not move perfectly together. It can reduce portfolio variance because adverse performance in one holding may be offset by stable or favourable performance elsewhere.

3.1 Types of diversification

• Naïve diversification: spreading investment across several securities without explicitly estimating covariances.
• Efficient diversification: selecting combinations based on expected returns, variances and covariances.
• 跨-asset diversification: combining equities, bonds, cash, property, commodities or alternative investments.
• Geographical diversification: investing across countries and regions.
• Sector diversification: avoiding excessive exposure to one industry.
• Factor diversification: reducing concentration in factors such as value, growth, size, momentum or interest-rate sensitivity.

Unsystematic, or diversifiable, risk declines as the number of imperfectly correlated securities increases. Systematic risk—market-wide risk arising from interest rates, inflation, recessions, geopolitical events or broad liquidity conditions—cannot be eliminated simply by holding more securities in the same market.

If securities have independent and similarly distributed risks, the contribution of firm-specific risk tends to decline as the portfolio expands. However, the reduction is not linear: the first few additional holdings generally provide the largest risk reduction, while later additions produce smaller marginal benefits.

Diversification does not guarantee a profit. During systemic crises, correlations may rise sharply, reducing the protection expected from historical relationships. Diversification also introduces transaction costs, monitoring costs, tax consequences, implementation complexity and the possibility of over-diversification.

4. Feasible Set and Efficient Frontier

The feasible set is the collection of all portfolios that can be formed under specified constraints, such as available assets, budget, short-sale restrictions and investment limits.

For each feasible expected return, some portfolio has the lowest possible variance. These portfolios form the minimum-variance boundary. The upper portion of that boundary is the efficient frontier: portfolios offering the highest expected return for a given level of risk, or the lowest risk for a given expected return.

A portfolio is inefficient if another feasible portfolio offers:

• higher expected return at the same risk; or
• lower risk at the same expected return.

The global minimum-variance portfolio (GMVP) is the feasible portfolio with the lowest variance, regardless of expected return. It is not necessarily the portfolio selected by every investor.

Figure: The upper boundary of the feasible set is the efficient frontier; the global minimum-variance portfolio is the leftmost portfolio.

4.1 Portfolio optimisation

A common optimisation problem is to minimise variance subject to a target return and full investment:

subject to:

Here, μ is the vector of expected asset returns, R* is the target portfolio return, 1 is a vector of ones, and w is the vector of portfolio weights. Additional constraints may include wᵢ ≥ 0, maximum asset weights, turnover limits and regulatory requirements.

Without short-sale restrictions, the two-asset minimum-variance weight for Asset 1 is:

The weight in Asset 2 is w₂ = 1 − w₁. This formula applies to a fully invested two-asset portfolio and assumes the denominator is non-zero. If the calculated weight is negative, the unconstrained solution involves short selling; a long-only investor must solve the constrained problem instead.

5. Investor Utility and Optimal Portfolio Choice

MPT identifies portfolios that are efficient, but it does not determine which efficient portfolio is best for a particular investor. That decision depends on preferences, especially risk aversion.

A commonly used quadratic utility function is:

where U is mean–variance utility and A is the investor’s coefficient of absolute risk aversion. A larger A means that the investor places a greater penalty on variance. The factor ½ is conventional and affects the scale of A, not the ranking of portfolios.

For two portfolios, an investor prefers Portfolio X to Portfolio Y when:

This framework assumes that higher expected return is preferred and risk is undesirable. It is particularly defensible when returns are approximately normally distributed or when the investor’s preferences can reasonably be represented by mean and variance.

Risk aversion and indifference curves

An investor with high risk aversion requires a larger expected-return compensation for additional variance. In a risk–return graph, indifference curves for a risk-averse investor slope upward: higher risk must be accompanied by higher expected return to maintain the same utility. The optimal risky portfolio is where the highest attainable indifference curve touches the efficient frontier.

Risk aversion is not the same as risk capacity. Risk capacity is the financial ability to bear losses, while risk tolerance is the psychological willingness to do so. Professional advice should assess both, along with investment horizon, liquidity requirements, tax position and legal constraints.

6. The Risk-Free Asset and the Capital Allocation Line

A risk-free asset has a known return Rf and zero variance, with covariance zero against risky returns under the idealised model. Combining the risk-free asset with a risky portfolio produces a capital allocation line:

where y is the proportion invested in the risky portfolio B. If 0 ≤ y ≤ 1, the remainder is invested in the risk-free asset. If y > 1, the investor borrows at the risk-free rate and leverages the risky portfolio; this requires borrowing to be available at that rate, an unrealistic assumption in many settings.

The slope of the capital allocation line is the Sharpe ratio:

The tangency portfolio is the risky portfolio with the highest Sharpe ratio when a risk-free asset is available. Under the strong assumptions of the separation theorem, all investors hold the same tangency portfolio but combine it differently with the risk-free asset according to their risk preferences. In practice, taxes, borrowing constraints, heterogeneous expectations and institutional restrictions weaken this result.

7. Worked Portfolio Choice Example

Suppose an investor is comparing two efficient portfolios:

Let A = 4. Using decimal returns:

The investor prefers Portfolio Y because its utility is higher, despite its lower expected return. The result illustrates that return rankings alone cannot determine optimal portfolio choice.

8. Applied Case: Pension Fund Asset Allocation

Background

This illustrative case concerns a pension fund with a long-term liability horizon. Its investment committee is considering a portfolio of domestic equities, investment-grade bonds and global equities.

Problem

The existing portfolio has a high domestic-equity concentration. The committee wants to improve the expected risk–return profile without relying on unrealistic forecasts or excessive trading.

Analysis

The investment team estimates expected returns, volatilities and correlations using long-term historical data, forward-looking capital-market assumptions and stress scenarios. It finds that global equities have substantial equity-market risk but lower correlation with domestic bonds and some lower correlation with domestic equities. Bonds provide lower expected return but stabilise the portfolio under certain economic conditions.

The team constructs a covariance matrix and evaluates portfolios subject to constraints: no short selling, a maximum 60% equity exposure, a minimum 20% bond allocation and a turnover ceiling. The unconstrained global minimum-variance portfolio is rejected because it has an implausibly high bond allocation and does not meet the fund’s return requirement. A constrained efficient portfolio is selected instead.

Approach and solution

The committee uses a strategic allocation as the policy benchmark and permits limited tactical deviations. It conducts sensitivity analysis because small changes in expected returns or correlations can produce large changes in unconstrained optimisation weights. It also tests the portfolio under inflation shocks, equity-market declines, credit deterioration and currency movements.

Outcome

The revised portfolio has a slightly lower forecast return but materially lower forecast volatility and reduced concentration. The committee accepts the allocation because it better matches the fund’s liability horizon and risk capacity. The decision is documented with assumptions, constraints, monitoring triggers and rebalancing rules.

Learning points

1. The mathematically optimal unconstrained portfolio may be institutionally unsuitable.
2. Correlation estimates are central to diversification benefits.
3. Constraints can improve robustness and governance.
4. Scenario analysis should supplement mean–variance optimisation.
5. Strategic asset allocation should reflect liabilities, not only asset statistics.

9. Limitations and Critical Evaluation

MPT is powerful but not a complete theory of investment behaviour.

Estimation error

Expected returns, variances and correlations are estimated, not known. Expected returns are particularly difficult to forecast. Optimisation may amplify small estimation errors, producing extreme weights. Resampling, shrinkage estimators, Bayesian methods, factor models and robust optimisation can reduce—but not eliminate—this problem.

Non-normal returns

Standard deviation does not distinguish beneficial upside volatility from harmful downside volatility. Return distributions may exhibit skewness, kurtosis, jumps and fat tails. Investors may therefore supplement variance with semideviation, value at risk, conditional value at risk, maximum drawdown and stress testing.

Changing correlations

Historical correlation is not a permanent structural relationship. Correlations can rise during crises, precisely when diversification is most valuable. Regime-switching models and scenario analysis may provide a more realistic assessment.

Liquidity and market impact

MPT often assumes that assets can be bought and sold at observable prices without affecting those prices. Illiquid securities may have stale valuations, wide bid–ask spreads and significant liquidation costs.

Taxes, costs and constraints

Taxes, fees, transaction costs, leverage limits, short-sale restrictions and regulatory capital requirements can materially change the optimal portfolio. A theoretical efficient frontier may therefore be unattainable after implementation costs.

Behavioural and governance issues

Investors may display loss aversion, overconfidence, home bias, herding and inconsistent risk preferences. Institutional portfolios are also shaped by governance capacity, reporting requirements and stakeholder expectations. A portfolio that is statistically efficient but impossible for a client to maintain during a downturn is not practically optimal.

10. Professional Practice, Technology and Governance

Portfolio analysts commonly use spreadsheet models, statistical software, optimisation solvers and portfolio-management systems to estimate covariance matrices, generate frontiers and monitor exposures. AI and machine-learning tools may assist with forecasting, classification, scenario generation and anomaly detection, but they do not remove model risk. Their use should include:

• transparent data provenance;
• out-of-sample validation;
• prevention of look-ahead and survivorship bias;
• explainable assumptions;
• human approval of investment decisions;
• controls against data leakage and model drift.

Governance should require documented investment beliefs, approved benchmarks, independent risk oversight, periodic model review, stress testing and clear accountability. Fiduciary decisions should be based on the client’s objectives and constraints rather than on an optimisation output alone.

Best-Practice Guidelines

1. Use economically justified, forward-looking assumptions rather than blindly extrapolating history.
2. Check that return frequencies, compounding conventions and covariance horizons are consistent.
3. Examine the covariance matrix for errors, instability and near-singularity.
4. Compare unconstrained and constrained solutions.
5. Report marginal and total risk contributions, not only portfolio volatility.
6. Include transaction costs, taxes and liquidity considerations.
7. Test portfolios under adverse but plausible scenarios.
8. Rebalance according to a documented policy rather than emotion.
9. Distinguish statistical diversification from genuine economic diversification.
10. Communicate uncertainty and model limitations clearly to decision-makers.

Chapter Summary

Modern Portfolio Theory evaluates investments in the context of a portfolio. Expected portfolio return is the weighted average of asset expected returns, whereas portfolio variance depends on individual variances and all pairwise covariances. Correlation is therefore central to diversification: imperfectly correlated assets can produce a portfolio with lower risk than its individual components.

The feasible set contains all attainable portfolios. The efficient frontier contains those offering the highest expected return for each risk level. The global minimum-variance portfolio has the lowest feasible variance, but it is not automatically optimal for every investor. Utility functions incorporate risk aversion and identify the preferred portfolio on the efficient frontier. When a risk-free asset is available, the capital allocation line and tangency portfolio provide an additional framework for combining risk-free and risky investments.

MPT supports asset allocation, pension-fund management, risk budgeting and performance evaluation. Nevertheless, estimation error, non-normal returns, unstable correlations, liquidity, transaction costs, taxes and behavioural factors limit its practical reliability. Effective portfolio management therefore combines optimisation with constraints, scenario analysis, governance and professional judgement.

Key Terms

Expected return: Probability-weighted average return of an investment.

Variance: Average squared deviation of returns from their expected value.

Standard deviation: Square root of variance and a common measure of total risk.

Covariance: Measure of the joint movement of two asset returns.

Correlation: Standardised covariance ranging from −1 to +1.

Portfolio: Combination of assets held by an investor.

Diversification: Combining assets to reduce concentration and potentially lower risk.

Systematic risk: Market-wide risk that cannot be eliminated through ordinary diversification.

Unsystematic risk: Firm- or security-specific risk that can be diversified away.

Feasible set: All portfolios attainable under specified assets and constraints.

Efficient frontier: Portfolios offering the best expected return for each risk level.

Global minimum-variance portfolio: Feasible portfolio with the lowest variance.

Mean–variance analysis: Portfolio analysis based on expected return and variance.

Risk aversion: Preference for lower risk when expected returns are equal.

Utility: Numerical representation of an investor’s preference over risk and return.

Risk-free asset: Asset assumed to have a known return and zero variance.

Capital allocation line: Line showing combinations of a risk-free asset and a risky portfolio.

Sharpe ratio: Excess expected return per unit of portfolio standard deviation.

Covariance matrix: Matrix containing asset variances and pairwise covariances.

Model risk: Risk that an analytical model is incorrect, misused or based on unreliable assumptions.

Review Questions

Short Answer Questions

1. Define expected portfolio return and explain why it is a weighted average of asset expected returns.
2. Distinguish covariance from correlation.
3. What is the global minimum-variance portfolio?
4. Explain the difference between systematic and unsystematic risk.
5. What does a high coefficient of risk aversion imply about an investor’s portfolio choice?

Long Answer Questions

1. Explain the construction and interpretation of the efficient frontier.
2. Derive and discuss the two-asset portfolio variance equation.
3. Evaluate the theoretical and practical benefits of diversification.
4. Explain how a risk-free asset changes the portfolio-selection problem.
5. Critically assess the main assumptions and limitations of Modern Portfolio Theory.

Analytical / Application Questions

1. Two assets have expected returns of 12% and 7%, standard deviations of 20% and 10%, and correlation of −0.25. Calculate the expected return and standard deviation of a portfolio invested 50% in each asset.
2. An investor with A = 5 is comparing a portfolio returning 14% with standard deviation 20% and another returning 10% with standard deviation 12%. Calculate mean–variance utility for each and identify the preferred portfolio.
3. A pension fund’s optimiser recommends a 90% allocation to one illiquid asset. Explain why the recommendation may be unstable or unsuitable and propose at least four governance or modelling responses.

Multiple Choice Questions with Answers

Question 1

Which measure directly captures how two asset returns move together?

A. Expected return

B. Covariance

C. Variance

D. Arithmetic mean

Correct Answer: B. Covariance

Question 2

A portfolio contains 60% in an asset returning 10% and 40% in an asset returning 6%. What is its expected return?

A. 7.2%

B. 8.0%

C. 8.4%

D. 9.6%

Correct Answer: C. 8.4%

Question 3

If two assets have correlation of −1, under suitable weights their portfolio risk can be:

A. Greater than both individual risks only

B. Exactly equal to the weighted average risk

C. Zero

D. Undefined

Correct Answer: C. Zero

Question 4

Which portfolio lies on the efficient frontier?

A. A portfolio dominated by another portfolio with higher return and equal risk

B. A portfolio offering the highest return for its level of risk

C. Every portfolio containing a risk-free asset

D. Only the global minimum-variance portfolio

Correct Answer: B. A portfolio offering the highest return for its level of risk

Question 5

For a fully invested two-asset portfolio, the portfolio variance includes:

A. Only the weighted average of individual standard deviations

B. Only the sum of individual variances

C. Individual variance terms and a covariance term

D. Expected returns but no risk terms

Correct Answer: C. Individual variance terms and a covariance term

Question 6

Using U = E(Rₚ) − ½Aσₚ², what is the utility when E(Rₚ) = 0.10, A = 4 and σₚ = 0.10?

A. 0.08

B. 0.09

C. 0.095

D. 0.20

Correct Answer: B. 0.09

Question 7

Which risk is generally not eliminated by holding many securities in the same market?

A. Firm-specific risk

B. Unsystematic risk

C. Systematic risk

D. Operational risk of one company

Correct Answer: C. Systematic risk

Question 8

The Sharpe ratio measures:

A. Total return divided by variance

B. Excess expected return divided by standard deviation

C. Expected return divided by covariance

D. Risk-free return divided by beta

Correct Answer: B. Excess expected return divided by standard deviation

Question 9

A major practical danger of unconstrained mean–variance optimisation is that it may:

A. Always produce equal asset weights

B. Ignore expected returns entirely

C. Generate extreme weights because of estimation error

D. Eliminate all systematic risk

Correct Answer: C. Generate extreme weights because of estimation error

Question 10

A high coefficient of absolute risk aversion generally means that the investor:

A. Places a larger penalty on portfolio variance

B. Must prefer the highest-return portfolio

C. Is indifferent to volatility

D. Cannot hold a risk-free asset

Correct Answer: A. Places a larger penalty on portfolio variance