Modern Portfolio Theory and Asset Pricing

Introduction and context

Asset management is fundamentally an exercise in allocating scarce capital among uncertain opportunities. An investor must decide not only which securities to buy, but also how much to hold, how exposures interact, and whether the expected compensation for risk is adequate. Modern Portfolio Theory (MPT), developed principally from Harry Markowitz’s 1952 work, transformed investment analysis by treating the portfolio—not the individual security—as the primary unit of risk management.

MPT provides the analytical foundation for diversification, efficient frontiers, portfolio optimisation and risk-adjusted performance measurement. Asset-pricing models extend this foundation by explaining how systematic risk may be related to expected return. The Capital Asset Pricing Model (CAPM), arbitrage pricing theory (APT) and multifactor models remain central to institutional portfolio construction, valuation, performance attribution and wealth advice.

These models are powerful simplifications rather than complete descriptions of financial markets. Their usefulness depends on the quality of inputs, the stability of relationships and the appropriateness of their assumptions. A professional investment banker or wealth manager must therefore use them critically, combining quantitative outputs with economic judgement, governance and client objectives.

Learning objectives

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

1. Explain diversification and distinguish total, systematic and unsystematic risk.
2. Calculate expected return, variance, standard deviation, covariance and correlation.
3. Construct and interpret efficient frontiers and investor utility functions.
4. Explain CAPM, beta, the security market line and the market portfolio.
5. Describe multifactor models and arbitrage pricing theory.
6. Evaluate the assumptions, empirical evidence and limitations of asset-pricing models.
7. Apply these ideas to portfolio design, advice, performance analysis and governance.

1. Return and risk: the building blocks

1.1 Holding-period and expected return

For an asset purchased at price P₀, sold at P₁, and paying income D₁, the simple holding-period return is:

Here, R is a decimal return, P₀ and P₁ are prices in currency units, and D₁ is dividend or other cash income. The formula assumes that the cash flows are measured over the same investment period. A continuously compounded return is ln(P₁/P₀), when no interim cash flow is present.

If possible future states s have returns R_s and probabilities p_s, expected return is:

Expected return is a probability-weighted mean, not a guaranteed outcome. In practice, probabilities may be estimated from historical frequencies, economic scenarios, statistical models or subjective judgement.

1.2 Variance and standard deviation

Variance measures dispersion around expected return:

Standard deviation is:

Both return and standard deviation are usually expressed in percentage terms. Variance is measured in squared-return units, so standard deviation is generally easier to interpret. These measures are most informative when returns are reasonably represented by their distribution and when the investor treats upside and downside deviations symmetrically.

For a sample of historical returns, an unbiased sample variance commonly uses n−1 rather than n in the denominator. Historical estimates are uncertain and may not represent future conditions.

1.3 Covariance and correlation

Covariance indicates whether two returns tend to move together:

Positive covariance means that above-average returns on one asset tend to accompany above-average returns on the other. Negative covariance indicates offsetting movements. Covariance depends on the units used, whereas correlation is standardised:

Correlation has no units. A correlation of +1 indicates perfect co-movement, 0 indicates no linear co-movement, and −1 indicates perfect inverse linear co-movement. Zero correlation does not prove independence and does not exclude nonlinear dependence.

2. Portfolio return and portfolio risk

For a portfolio containing n assets with weights w_i, where weights sum to one:

Consequently:

Weights may be negative when short selling is permitted. With no leverage and no short selling, weights are generally non-negative and sum to one.

For two assets A and B, portfolio variance is:

Using correlation:

The covariance term is the mathematical source of diversification. Portfolio risk is not normally the weighted average of individual risks. Even if both assets are individually volatile, imperfect correlation can reduce total portfolio volatility.

For n assets, matrix notation is more efficient:

Here, w is the column vector of portfolio weights, Σ is the covariance matrix, and w' is the transpose of the weight vector. The matrix must be symmetric and positive semidefinite for a valid variance calculation.

Worked example: two-asset diversification

Suppose an equity fund has expected return 10%, standard deviation 18%, and a bond fund has expected return 5%, standard deviation 7%. A portfolio invests 60% in equity and 40% in bonds, and their correlation is 0.20.

Expected return:

E(R_p) = 0.60(10%) + 0.40(5%) = 8.0%

Variance, using decimal units:

σ_p² = (0.60²)(0.18²) + (0.40²)(0.07²) + 2(0.60)(0.40)(0.20)(0.18)(0.07)

σ_p² = 0.011664 + 0.000784 + 0.0012096 = 0.0136576

Therefore, σ_p = √0.0136576 ≈ 11.69%. The portfolio’s risk is materially below the weighted average of individual standard deviations, which is 0.60(18%) + 0.40(7%) = 13.6%.

3. Diversification and the efficient frontier

Diversification reduces asset-specific, or idiosyncratic, risk when assets do not move perfectly together. It cannot eliminate broad economic or market risk. A portfolio of technology companies may diversify individual-company events but remain exposed to interest rates, recession, regulation and equity-market conditions.

For a chosen expected return, the minimum-variance portfolio is the portfolio with the lowest feasible variance. The set of minimum-variance portfolios across all target returns forms the minimum-variance frontier. The efficient frontier is the upper portion of that frontier: portfolios offering the highest expected return for a given risk, or the lowest risk for a given expected return.

In a mean–standard-deviation diagram, the efficient frontier is typically upward sloping and curved. The global minimum-variance portfolio is the frontier portfolio with the lowest standard deviation. Portfolios below the efficient frontier are inefficient because another feasible portfolio offers either more return for the same risk or less risk for the same return.

Optimisation requires estimates of expected returns, covariances, constraints, transaction costs and a rebalancing horizon. Small changes in expected-return estimates can produce extreme weights, especially when assets are highly correlated. Practical optimisation therefore often uses weight limits, turnover constraints, robust estimates, shrinkage methods or resampling.

4. Investor preferences and utility

MPT alone identifies risk–return opportunities; it does not determine which opportunity is best for a particular investor. Utility theory introduces preferences. A common quadratic utility approximation is:

U is investor utility, A is the coefficient of risk aversion, and σ_p² is portfolio variance. Higher A means greater aversion to risk. The formula assumes that mean and variance adequately summarise preferences, which is most defensible under normally distributed returns or quadratic preferences.

If a risk-free asset with return R_f is available and borrowing and lending occur at that rate, combinations of the risk-free asset and a risky portfolio lie on a capital allocation line:

Here, T denotes the selected tangency portfolio. The slope is the Sharpe ratio of the tangency portfolio. The tangency portfolio maximises the reward-to-risk ratio among risky portfolios. Investor risk tolerance determines the allocation between the risk-free asset and the tangency portfolio, not the identity of the tangency portfolio under the strong assumptions of the model.

5. CAPM and beta

The Capital Asset Pricing Model links an asset’s expected return to its non-diversifiable risk. Its central equation is:

R_f is the risk-free rate, R_M is the market portfolio return, and β_i measures the asset’s sensitivity to market risk. The market risk premium is E(R_M) − R_f.

Beta is defined as:

A beta of 1 implies market sensitivity equal to the market’s; a beta above 1 indicates greater sensitivity; a beta below 1 indicates lower sensitivity. A negative beta may provide hedging value, although it is uncommon for ordinary equities.

The security market line (SML) plots expected return against beta. Securities above the SML offer positive abnormal expected return relative to CAPM; securities below it offer negative abnormal expected return. This is distinct from the capital market line (CML), which plots expected return against total standard deviation for efficient portfolios.

CAPM assumptions

The traditional one-period CAPM assumes:

• investors are risk-averse mean–variance optimisers;
• investors have homogeneous expectations;
• investors can borrow and lend at a common risk-free rate;
• markets are frictionless, with no taxes or transaction costs;
• assets are infinitely divisible and tradable;
• investors share the same investment horizon; and
• the market portfolio includes all investable assets.

Under these assumptions, all investors hold the same risky market portfolio in different combinations with the risk-free asset. In practice, borrowing rates differ from lending rates, taxes and costs matter, investors face constraints, and the true market portfolio is unobservable.

6. Multifactor models and arbitrage pricing theory

Empirical research shows that a single market beta often does not explain all cross-sectional differences in returns. A general linear factor model can be written as:

Here, F_k are factor returns or risk premiums, β_ik are asset sensitivities, α_i is an intercept or abnormal component, and ε_i is asset-specific residual risk. A well-specified model generally assumes residuals have zero expected value and are not systematically explained by the included factors.

Common factors include market excess return, company size, value, profitability, investment, momentum, term risk, credit risk, inflation and liquidity. Fama–French models, for example, supplement market exposure with size and value factors; later versions add profitability and investment factors. Commercial risk systems may use sector, country, currency, duration, spread and style factors.

Arbitrage Pricing Theory argues that, with sufficiently diversified portfolios and no persistent arbitrage opportunities, expected returns can be represented as compensation for exposure to common factors:

λ_j is the risk premium associated with factor j. Unlike CAPM, APT does not require the market portfolio to be mean–variance efficient. However, it does not uniquely identify which factors should be used; factor selection remains an empirical and economic judgement.

7. Market efficiency and active management

The efficient-market hypothesis (EMH) states that security prices reflect available information to varying degrees. The weak form concerns past prices and trading data; the semi-strong form includes publicly available information; the strong form includes private information. If markets are semi-strong efficient, public information should be rapidly incorporated into prices, making consistent risk-adjusted outperformance difficult.

Efficiency is not equivalent to perfect pricing. Prices may deviate from fundamental value, but competition and arbitrage may limit the persistence of deviations. Behavioural finance identifies overconfidence, representativeness, loss aversion, anchoring and limits to arbitrage as potential sources of predictable mispricing.

For wealth managers, the practical implication is not that active management is impossible. Rather, an active strategy should have a defensible source of expected value, realistic implementation costs, capacity analysis, risk controls and evidence that apparent alpha is not merely compensation for an omitted factor or leverage.

8. Measuring performance and attribution

The Sharpe ratio is:

It measures excess return per unit of total volatility. The ex post version uses realised average excess return and realised standard deviation. It is sensitive to the benchmark, time horizon, non-normal returns and serial correlation.

Jensen’s alpha under CAPM is:

This realised form estimates performance relative to the return predicted by CAPM. A positive alpha may reflect skill, omitted risk, measurement error or luck. Multifactor attribution replaces the single beta term with several factor contributions.

Other measures include the Treynor ratio, which uses beta rather than total volatility, and information ratio:

R_B is benchmark return and the denominator is tracking-error volatility. These measures should be interpreted alongside drawdown, liquidity, tail risk, turnover, costs and mandate compliance.

9. Illustrative case: institutional balanced portfolio

Background

An illustrative pension fund has a long-term liability stream, moderate tolerance for interim losses and a mandate to invest across global equities, government bonds, investment-grade credit and real assets. Its existing policy allocates 60% to global equities and 40% to government bonds.

Problem

A risk review finds that the equity allocation is concentrated in a small number of large technology companies, while bonds have high duration exposure. The portfolio’s historical volatility appears moderate, but stress testing shows substantial losses under an inflation shock accompanied by rising interest rates.

Analysis

The investment team estimates forward-looking returns and a covariance matrix rather than relying exclusively on historical averages. It identifies equity beta, duration, credit spread, inflation sensitivity and currency exposure as relevant risk dimensions. The team also tests correlations during crises, recognising that correlations can rise precisely when diversification is most valuable.

Approach

The revised process uses a constrained optimisation. It limits single-sector exposure, imposes a maximum portfolio duration, reserves a liquidity bucket, and includes inflation-sensitive real assets. Expected returns are stress-tested rather than treated as precise forecasts. The committee compares the proposed portfolio with the old policy using expected shortfall, drawdown scenarios, turnover and funding-ratio outcomes, not merely standard deviation.

Outcome

The revised portfolio has a slightly lower central expected return but materially reduces concentration and interest-rate vulnerability. It remains exposed to market risk; diversification does not guarantee positive returns. Rebalancing bands are established, and governance requires periodic review of factor exposures and model assumptions.

Learning points

1. Optimisation is only as reliable as its inputs and constraints.
2. Correlations are dynamic and may increase in crises.
3. Liability characteristics should influence strategic asset allocation.
4. Factor and scenario analysis can reveal risks hidden by aggregate volatility.
5. Governance, liquidity and implementation costs are part of portfolio construction.

10. Limitations, governance and contemporary practice

Estimation risk and model risk

Expected returns are particularly difficult to estimate. Covariance matrices may be unstable when the number of assets is large relative to the history available. Historical data can contain regime changes, survivorship bias, stale prices and structural breaks. Model risk arises when users mistake an elegant formula for a reliable forecast.

Non-normality and downside risk

Returns may exhibit skewness, kurtosis, jumps and volatility clustering. Standard deviation treats upside and downside variability alike and may understate tail exposure. Conditional value at risk, maximum drawdown, scenario analysis and stress testing can complement variance-based analysis, although each has its own estimation limitations.

Liquidity, costs and taxes

Theoretical portfolios may require turnover, leverage or short positions that are impractical. Bid–ask spreads, market impact, financing costs, taxes and fund fees reduce investor returns. A portfolio should be evaluated on net, after-cost outcomes and under realistic trading capacity.

Ethical and regulatory considerations

Advisers must ensure that portfolio recommendations are suitable for the client’s objectives, risk capacity, time horizon, liquidity needs and legal constraints. Risk tolerance questionnaires should not substitute for a complete assessment of financial circumstances. Marketing should not present model outputs as guarantees. Explainable assumptions, documented approvals, independent validation and monitoring of breaches are essential governance practices.

Technology and AI

Modern systems can estimate covariance matrices, run Monte Carlo simulations, optimise portfolios and attribute returns in near real time. Machine-learning methods may identify nonlinear relationships, but they introduce overfitting, data leakage, instability and explainability concerns. AI-generated recommendations require human oversight, validation, version control, audit trails and controls against biased or unsuitable advice. Technology improves process discipline; it does not eliminate economic uncertainty.

Best-practice framework for professional use

A robust portfolio process should:

1. Define objectives, liabilities, constraints and risk capacity.
2. Select transparent asset and factor classifications.
3. Estimate returns, volatilities and covariances using appropriate horizons and robust methods.
4. Compare optimisation outputs with simple, diversified benchmarks.
5. Incorporate costs, taxes, liquidity and implementation capacity.
6. Use sensitivity analysis, scenarios and stress tests.
7. Monitor realised exposures, not merely intended weights.
8. Rebalance according to documented rules while considering tax and transaction costs.
9. Attribute performance to allocation, selection, factors, currency and fees.
10. Review assumptions and model performance regularly.

Chapter Summary

Modern Portfolio Theory shows that portfolio risk depends on the interaction of assets, not simply on individual volatility. Expected return is a weighted average, while portfolio variance incorporates variances, covariances and correlations. Diversification can reduce unsystematic risk, but systematic risk remains.

The efficient frontier identifies portfolios that are not dominated in risk–return space. Utility theory links the frontier to investor preferences, while the capital allocation line describes combinations of a risk-free asset and a risky portfolio. CAPM explains expected return through beta and the market risk premium; multifactor models extend this logic to several systematic sources of risk. APT offers a no-arbitrage framework without requiring the market portfolio to be efficient.

Market efficiency, performance ratios and factor attribution help evaluate investment strategies, but none provides certainty. Estimation error, changing correlations, non-normal returns, liquidity, costs, behavioural biases and governance constraints must be addressed. Professional application therefore combines quantitative models with scenario analysis, client suitability, transparent assumptions and disciplined oversight.

Key Terms

Expected return: Probability-weighted average of possible returns.

Variance: Probability-weighted squared deviation from expected return.

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

Covariance: Measure of joint movement between two returns.

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

Diversification: Combining assets whose returns are imperfectly related to reduce portfolio-specific risk.

Systematic risk: Risk associated with common market or macroeconomic factors.

Unsystematic risk: Asset-specific risk that may be diversified away.

Efficient frontier: Portfolios offering the highest return for each risk level or lowest risk for each return level.

Utility: Numerical representation of an investor’s preferences.

Risk-free rate: Return on an asset assumed to have negligible default risk over the relevant horizon.

Capital allocation line: Risk–return combinations of a risk-free asset and a risky portfolio.

Beta: Covariance of an asset with the market divided by market variance.

Market risk premium: Expected market return minus the risk-free rate.

Security market line: CAPM relationship between expected return and beta.

Multifactor model: Model explaining returns through exposure to several common factors.

Alpha: Return unexplained by a specified asset-pricing model.

Sharpe ratio: Excess return per unit of total volatility.

Efficient-market hypothesis: Proposition that prices incorporate information to a relevant degree.

Tracking error: Volatility of a portfolio’s active return relative to its benchmark.

Review Questions

Short Answer Questions

1. Why can a portfolio’s standard deviation be lower than the weighted average of its assets’ standard deviations?
2. Distinguish systematic risk from unsystematic risk.
3. What does a correlation coefficient of −1 imply in a two-asset portfolio?
4. Define beta and explain its role in CAPM.
5. What is the difference between the capital market line and the security market line?

Long Answer Questions

1. Explain the construction and interpretation of the efficient frontier, including the role of the global minimum-variance portfolio.
2. Present the assumptions of CAPM and critically assess their relevance to institutional investment.
3. Compare CAPM, APT and multifactor models in terms of factors, assumptions and practical application.
4. Explain how utility theory determines an investor’s preferred portfolio when a risk-free asset is available.
5. Discuss why historical covariance estimates may be unreliable inputs to portfolio optimisation.

Analytical / Application Questions

1. A portfolio invests 70% in an asset with expected return 12% and standard deviation 20%, and 30% in an asset with expected return 4% and standard deviation 6%. Their correlation is 0.10. Calculate expected return and portfolio standard deviation.
2. An equity has beta 1.2, the risk-free rate is 3%, and the expected market return is 9%. Use CAPM to calculate its expected return and interpret the result.
3. A fund has a high Sharpe ratio but a negative information ratio against its benchmark. Explain how both statements can be true and identify additional evidence required before judging performance.

Multiple Choice Questions with Answers

Question 1

Which term captures the joint movement of two asset returns before standardisation?

A. Variance

B. Covariance

C. Beta

D. Alpha

Correct Answer: B. Covariance

Question 2

A portfolio invests 60% in an asset returning 10% and 40% in an asset returning 5%. What is its expected return?

A. 6.0%

B. 7.0%

C. 8.0%

D. 9.0%

Correct Answer: C. 8.0%

Question 3

Which risk is most directly reduced by holding a broad range of imperfectly correlated securities?

A. Systematic risk

B. Inflation risk

C. Unsystematic risk

D. Market-wide risk

Correct Answer: C. Unsystematic risk

Question 4

In CAPM, beta is calculated as:

A. Asset variance divided by market variance

B. Asset covariance with the market divided by market variance

C. Market return divided by the risk-free rate

D. Excess return divided by standard deviation

Correct Answer: B. Asset covariance with the market divided by market variance

Question 5

The efficient frontier consists of portfolios that:

A. Have zero risk

B. Contain only risk-free assets

C. Offer the best feasible return–risk combinations

D. Always outperform the market

Correct Answer: C. Offer the best feasible return–risk combinations

Question 6

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

A. Greater than both individual risks only

B. Exactly equal to the market risk

C. Zero

D. Undefined

Correct Answer: C. Zero

Question 7

According to CAPM, an asset with beta 1.5 has:

A. No systematic risk

B. 50% greater market sensitivity than the market portfolio

C. A guaranteed return 1.5 times the market return

D. A correlation of 1.5 with the market

Correct Answer: B. 50% greater market sensitivity than the market portfolio

Question 8

Which measure uses total volatility in its denominator?

A. Sharpe ratio

B. Treynor ratio

C. Information ratio

D. Jensen’s alpha

Correct Answer: A. Sharpe ratio

Question 9

A multifactor model explains an asset’s return using:

A. Only its dividend yield

B. Several common risk exposures and a residual

C. Only its accounting profit

D. Only the risk-free rate

Correct Answer: B. Several common risk exposures and a residual

Question 10

Which statement best describes a major limitation of mean–variance optimisation?

A. It cannot calculate portfolio return

B. It assumes correlations and expected returns are known with certainty

C. It prohibits diversification

D. It applies only to risk-free assets

Correct Answer: B. It assumes correlations and expected returns are known with certainty