6.1 Mean-variance analysis

Mean-variance analysis is a cornerstone of modern portfolio theory. It quantifies the trade-off between risk and return, helping investors construct optimal portfolios that maximize expected returns for a given level of risk.

This approach revolutionized investment management by emphasizing portfolio-level analysis over individual asset selection. Key concepts include expected return, variance as a measure of risk, the efficient frontier, and the Sharpe ratio for evaluating risk-adjusted performance.

Definition of mean-variance analysis

• Fundamental approach in modern portfolio theory quantifies trade-off between risk and return
• Developed by Harry Markowitz in 1952 revolutionized investment management and asset allocation
• Forms basis for optimal portfolio construction considering expected returns and risk tolerance

Key concepts and terminology

• Mean represents expected return of an investment or portfolio
• Variance measures dispersion of returns around the mean quantifies risk
• Efficient frontier plots optimal portfolios maximizing return for given level of risk
• Risk-free rate serves as benchmark for evaluating risk-adjusted returns
• Sharpe ratio measures excess return per unit of risk taken

Historical context and development

• Originated from Harry Markowitz's 1952 paper "Portfolio Selection" in Journal of Finance
• Challenged traditional focus on individual asset selection emphasized portfolio-level analysis
• Won Nobel Prize in Economics in 1990 for contributions to financial economics
• Evolved with advancements in computing power enabled more complex optimization techniques
• Influenced development of Capital Asset Pricing Model (CAPM) and other asset pricing theories

Portfolio theory fundamentals

• Focuses on constructing optimal portfolios balancing risk and return
• Emphasizes diversification to reduce overall portfolio risk
• Introduces concept of systematic (market) risk and unsystematic (specific) risk

Expected return calculation

• Weighted average of individual asset returns in portfolio
• Formula $E(R_p) = \sum_{i=1}^n w_i E(R_i)$
  • $E(R_p)$ expected portfolio return
  • $w_i$ weight of asset i
  • $E(R_i)$ expected return of asset i
• Considers historical data, analyst forecasts, and economic projections
• Adjusts for different time horizons and market conditions

Risk measurement using variance

• Variance quantifies dispersion of returns around mean
• Portfolio variance formula $\sigma_p^2 = \sum_{i=1}^n w_i^2 \sigma_i^2 + \sum_{i=1}^n \sum_{j \neq i} w_i w_j \sigma_i \sigma_j \rho_{ij}$
  • $\sigma_p^2$ portfolio variance
  • $\sigma_i$ standard deviation of asset i
  • $\rho_{ij}$ correlation coefficient between assets i and j
• Incorporates individual asset variances and correlations between assets
• Square root of variance yields standard deviation commonly used risk measure

Covariance and correlation

• Covariance measures how two variables move together
• Formula $Cov(R_i, R_j) = E[(R_i - E(R_i))(R_j - E(R_j))]$
• Correlation standardized measure of covariance ranges from -1 to 1
• Negative correlation reduces portfolio risk through diversification
• Positive correlation indicates assets tend to move in same direction

Efficient frontier

• Graphical representation of optimal portfolios in risk-return space
• Represents portfolios with highest expected return for given level of risk
• Crucial tool for portfolio selection and asset allocation decisions

Construction of efficient frontier

• Plot all possible portfolios in risk-return space
• Identify portfolios with highest return for each level of risk
• Connect these points to form efficient frontier curve
• Utilize quadratic programming or numerical optimization techniques
• Consider constraints (short-selling restrictions, sector limits)

Properties of efficient portfolios

• Lie on upper portion of efficient frontier curve
• Offer best trade-off between risk and return
• Cannot improve return without increasing risk or vice versa
• Typically well-diversified across multiple assets or asset classes
• Vary in composition based on investor risk preferences

Capital allocation line

• Represents combinations of risk-free asset and risky portfolio
• Tangent to efficient frontier at optimal risky portfolio
• Slope of CAL known as Sharpe ratio measures risk-adjusted return
• Formula $CAL: E(R_p) = R_f + \frac{E(R_m) - R_f}{\sigma_m} \sigma_p$
  • $R_f$ risk-free rate
  • $E(R_m)$ expected return of market portfolio
  • $\sigma_m$ standard deviation of market portfolio
• Investors choose point on CAL based on risk tolerance

Utility theory in portfolio selection

• Incorporates investor preferences into portfolio decision-making process
• Balances desire for higher returns with aversion to risk
• Provides framework for selecting optimal portfolio based on individual utility function

Risk aversion concepts

• Describes investor's attitude towards risk
• Measured by coefficient of risk aversion in utility function
• Higher risk aversion leads to preference for lower-risk portfolios
• Influences shape of indifference curves and optimal portfolio selection
• Can vary across investors and change over time

Indifference curves

• Represent combinations of risk and return yielding same utility for investor
• Convex shape reflects diminishing marginal utility of wealth
• Steeper curves indicate higher risk aversion
• Tangent point with efficient frontier determines optimal portfolio
• Shift based on changes in investor preferences or market conditions

Optimal portfolio selection

• Occurs at tangency point between highest indifference curve and efficient frontier
• Maximizes investor's utility given risk-return trade-off
• Considers both expected return and risk tolerance
• May involve combination of risk-free asset and optimal risky portfolio
• Requires periodic reassessment as market conditions and preferences change

Markowitz model

• Foundational framework for modern portfolio theory
• Focuses on mean-variance optimization for portfolio selection
• Balances expected returns with portfolio risk to maximize utility

Model assumptions

• Investors are rational and risk-averse
• Markets are efficient and information is freely available
• Returns follow normal distribution
• No transaction costs or taxes
• Investors can lend and borrow at risk-free rate
• All assets are perfectly divisible

Mathematical formulation

• Objective function maximize expected utility $U = E(R_p) - \frac{1}{2} A \sigma_p^2$
  • A coefficient of risk aversion
• Constraints
  • Sum of weights equals 1 $\sum_{i=1}^n w_i = 1$
  • Non-negativity constraints (if short-selling not allowed) $w_i \geq 0$
• Solved using quadratic programming techniques
• Results in optimal portfolio weights for given risk tolerance

Limitations and criticisms

• Assumes normal distribution of returns may not hold in reality
• Sensitive to input parameters estimation errors can lead to suboptimal portfolios
• Does not account for higher moments of return distribution (skewness, kurtosis)
• Ignores transaction costs and taxes can overstate benefits of frequent rebalancing
• May lead to concentrated portfolios in absence of constraints

Single-index model

• Simplifies portfolio analysis by relating asset returns to single market factor
• Reduces number of parameters to estimate compared to full covariance matrix
• Provides framework for understanding systematic and unsystematic risk

Market model vs CAPM

• Market model descriptive relates asset returns to market returns
• CAPM prescriptive provides framework for asset pricing
• Market model formula $R_i = \alpha_i + \beta_i R_m + \epsilon_i$
  • $\alpha_i$ asset-specific return
  • $\beta_i$ sensitivity to market returns
  • $R_m$ market return
  • $\epsilon_i$ idiosyncratic risk
• CAPM formula $E(R_i) = R_f + \beta_i (E(R_m) - R_f)$
• Market model used for empirical analysis CAPM for theoretical asset pricing

Beta estimation

• Measures sensitivity of asset returns to market returns
• Estimated using regression analysis of historical returns
• Formula $\beta_i = \frac{Cov(R_i, R_m)}{Var(R_m)}$
• Beta > 1 indicates higher volatility than market
• Beta < 1 indicates lower volatility than market
• Can be adjusted for leverage or other factors

Simplification of covariance matrix

• Reduces number of parameters from $\frac{n(n+1)}{2}$ to $2n + 1$
• Covariance between assets $Cov(R_i, R_j) = \beta_i \beta_j Var(R_m)$
• Assumes all covariance due to common market factor
• Ignores residual correlations between assets
• Computationally efficient for large portfolios

Multi-factor models

• Extend single-index model to include multiple explanatory factors
• Capture additional sources of systematic risk beyond market factor
• Provide more nuanced view of asset returns and risk exposures

Arbitrage pricing theory

• Developed by Stephen Ross as alternative to CAPM
• Assumes returns generated by multiple macroeconomic factors
• Does not specify factors a priori allows for flexible model specification
• Formula $E(R_i) = R_f + \beta_{i1} \lambda_1 + \beta_{i2} \lambda_2 + ... + \beta_{ik} \lambda_k$
  • $\lambda_k$ risk premium for factor k
  • $\beta_{ik}$ sensitivity of asset i to factor k
• Relies on no-arbitrage condition for pricing assets

Fama-French three-factor model

• Extends CAPM to include size and value factors
• Developed by Eugene Fama and Kenneth French
• Factors market excess return, size premium (SMB), value premium (HML)
• Formula $E(R_i) - R_f = \beta_i (E(R_m) - R_f) + s_i E(SMB) + h_i E(HML)$
  • $s_i$ sensitivity to size factor
  • $h_i$ sensitivity to value factor
• Explains significant portion of cross-sectional variation in returns

Extensions and variations

• Carhart four-factor model adds momentum factor
• Fama-French five-factor model includes profitability and investment factors
• Industry-specific models incorporate sector-related factors
• Macroeconomic factor models use economic indicators (GDP growth, inflation)
• Statistical factor models use principal component analysis to identify factors

Practical applications

• Implement mean-variance analysis in real-world portfolio management
• Balance theoretical concepts with practical constraints and considerations
• Adapt techniques to changing market conditions and investor needs

Portfolio optimization techniques

• Quadratic programming solves mean-variance optimization problem
• Monte Carlo simulation generates scenarios for robust optimization
• Genetic algorithms search for near-optimal solutions in complex landscapes
• Black-Litterman model incorporates investor views with market equilibrium
• Risk parity allocates based on risk contribution rather than capital allocation

Rebalancing strategies

• Periodic rebalancing adjusts portfolio weights at fixed intervals
• Threshold rebalancing triggers when asset weights deviate beyond set limits
• Optimal rebalancing considers transaction costs and expected utility gain
• Dynamic rebalancing adjusts allocation based on changing market conditions
• Tax-aware rebalancing minimizes tax impact of portfolio adjustments

Performance evaluation metrics

• Sharpe ratio measures excess return per unit of total risk
• Treynor ratio assesses excess return per unit of systematic risk
• Jensen's alpha evaluates risk-adjusted performance relative to CAPM
• Information ratio gauges active return relative to tracking error
• Sortino ratio focuses on downside risk penalizes only negative deviations

Advanced topics

• Explore cutting-edge techniques in portfolio management
• Address limitations of traditional mean-variance analysis
• Incorporate advanced statistical and computational methods

Black-Litterman model

• Combines market equilibrium with investor views
• Addresses estimation error issues in mean-variance optimization
• Uses Bayesian approach to blend prior (market) and posterior (views) distributions
• Allows for varying degrees of confidence in investor views
• Results in more stable and intuitive portfolio allocations

Robust optimization

• Accounts for uncertainty in input parameters
• Minimizes worst-case scenarios rather than optimizing expected outcome
• Techniques include
  • Minimax optimization
  • Uncertainty sets
  • Distributionally robust optimization
• Produces portfolios less sensitive to estimation errors
• May lead to more conservative allocations

Machine learning in portfolio management

• Utilizes artificial intelligence techniques for asset allocation
• Neural networks for return prediction and risk assessment
• Clustering algorithms for asset classification and style analysis
• Reinforcement learning for dynamic portfolio optimization
• Natural language processing for sentiment analysis and news impact
• Ensemble methods for combining multiple models and strategies

Limitations and challenges

• Recognize potential pitfalls in applying mean-variance analysis
• Address practical issues in implementing portfolio optimization
• Consider alternative approaches to overcome limitations

Estimation error

• Input parameters (expected returns, variances, covariances) subject to uncertainty
• Small changes in inputs can lead to significant changes in optimal portfolio
• Methods to address
  • Shrinkage estimators
  • Resampling techniques
  • Bayesian approaches
• Use of longer historical periods or forward-looking estimates
• Incorporation of estimation error into optimization process

Transaction costs

• Can significantly impact realized returns especially for high-turnover strategies
• Types include commissions, bid-ask spreads, market impact
• Methods to address
  • Incorporating transaction costs into optimization objective
  • Implementing trading limits or turnover constraints
  • Using multi-period optimization models
• Trade-off between optimal allocation and cost of rebalancing
• Consideration of tax implications for taxable investors

Non-normal return distributions

• Asset returns often exhibit fat tails and skewness
• Violation of normality assumption in mean-variance analysis
• Alternative risk measures
  • Value at Risk (VaR)
  • Conditional Value at Risk (CVaR)
  • Lower partial moments
• Use of copulas to model complex dependence structures
• Consideration of higher moments (skewness, kurtosis) in optimization