14. Portfolio Theory

Name: 14. Portfolio Theory
Uploaded: 2015-01-06T18:18:57.000Z
Duration: 2 h 48 min 55 s

Portfolio Theory Overview

Introduction to Portfolio Theory

The content is provided under a Creative Commons license, supporting MIT OpenCourseWare's educational resources.

Today's focus is on portfolio theory, a crucial topic in finance, starting with Markowitz Mean-Variance Optimization.

Historical Context and Key Concepts

The discussion begins with the historical theory of portfolio optimization, emphasizing performance characteristics based on mean return and volatility.

The analysis extends to include risk-free assets, which significantly alters the investment landscape compared to only risky assets.

Utility Theory and Decision Making

Introduces von Neumann-Morgenstern utility theory for rational decision-making under uncertainty, essential for decision analysis and portfolio selection.

Constraints in Portfolio Optimization

Discusses practical constraints such as available capital, short-selling limitations, and asset capacity that affect investment decisions.

Estimation Challenges in Portfolio Analysis

Estimating Returns and Volatilities

Most theories assume certainty regarding returns' means, variances, and correlations; however, real-world estimation introduces complexities.

Alternative Risk Measures

Concludes with alternative risk measures that extend beyond simple mean-variance analysis.

Mean-Variance Analysis Fundamentals

Single Period Investment Strategy

Focuses on single-period investments in risky assets indexed by i (1 through m), using multivariate vectors for returns.

Portfolio Representation

A portfolio is represented as a weighting vector of investments across m assets; total weights equal one unit of capital invested.

Two Asset Case Study

Simplified Portfolio Analysis

Examines optimal portfolios involving two assets: one with a 15% annualized return (25% volatility), another with 20% expected return (30% volatility).

Expected Return Calculation

The expected return formula combines both asset returns based on their respective weights in the portfolio.

Markowitz's Feasible Portfolio Set

Defining Feasible Portfolios

Markowitz's mean-variance analysis defines feasible portfolios as pairs of volatility-return combinations from all possible portfolios.

Optimal vs. Sub-optimal Portfolios

Portfolio Optimization and Asset Correlation

Introduction to Asset Allocation

The discussion begins with plotting sigma_omega (volatility) against alpha_omega (returns) for two assets, denoted as sigma_1, alpha_1 for the first asset and sigma_2, alpha_2 for the second.

Points on the graph represent different weights in a portfolio: w = 0 corresponds to only asset 1, while w = 1 corresponds to only asset 2.

Simulation of Asset Returns

A simulation shows mean returns of 15% and 25% with volatilities ranging from 20% to 30%, indicating no apparent correlation between the two assets.

Cumulative return graphs illustrate that asset 2 has higher returns than asset 1, which is represented by their respective positions on the graph.

Feasible Set of Portfolios

The feasible set curve demonstrates how combining assets can reduce volatility without sacrificing returns; increasing allocation towards asset 2 improves portfolio performance.

The minimum variance portfolio is discussed mathematically, showing how to minimize volatility through weight adjustments based on each asset's volatility.

Optimal Portfolio Selection

The formula derived indicates that optimal weights are inversely proportional to each asset's squared volatility.

Observations reveal that portfolios with certain weights (like w1) are sub-optimal since they can be improved in terms of return and reduced in volatility.

Impact of Correlation on Portfolio Volatility

It’s emphasized that when assets are uncorrelated, diversifying across them can enhance portfolio performance.

Introducing negative correlations (e.g., -0.4 or -0.8), allows further reduction in portfolio volatility as shown by shifts in the scatter plot of returns.

Hedging Strategies and Zero Variability

A perfect negative correlation (-1) could lead to zero variability in a portfolio, allowing for effective hedging strategies using derivatives.

Understanding Portfolio Diversification and Optimization

Impact of Correlation on Diversification

Increasing correlation from 0 to 0.4 still provides diversification benefits, but the ability to lower variance diminishes. Higher correlations (e.g., 0.8) exacerbate this effect, reducing diversification effectiveness.

Simulation of Returns and Sample Statistics

Simulated returns are based on Gaussian distributions with specified means, volatilities, and correlations. The sample statistics reveal discrepancies between sample estimates and theoretical parameters, particularly in sample means which can vary significantly. For instance, the second asset's sample mean is notably higher than expected.

Variability in estimating covariances and correlations is generally less than that for sample means; thus, caution is advised when interpreting these estimates due to their inherent uncertainty.

Evaluating Portfolios: Mean vs Volatility

The evaluation of portfolios focuses on maximizing expected returns while minimizing volatility (risk). Markowitz formulated this as a quadratic programming problem aimed at minimizing squared portfolio volatility under a mean constraint while ensuring full investment.

Lagrangian Method in Portfolio Optimization

The optimization problem involves defining a Lagrangian that incorporates constraints related to mean return and total weights of assets in the portfolio. This leads to first-order conditions that help derive optimal weights for asset allocation.

The second-order derivative of the Lagrangian confirms minimization since it results in a positive definite covariance matrix, indicating successful optimization within multi-dimensional space akin to parabolic shapes in simpler cases.

Solving for Weights and Variance

Optimal weights can be expressed through lambda variables derived from first-order conditions; substituting these into equations allows solving for variances associated with target returns effectively. This relationship forms a parabolic structure when visualized graphically across multiple dimensions or assets.

Alternative Perspectives on Portfolio Problems

Understanding Portfolio Theory and the Efficient Frontier

The Concept of Asset Combinations

The discussion begins with two assets, sigma and alpha, forming a parabola representing their relationship. Adding more assets creates additional parabolas for each two-asset portfolio.

These combinations of two-asset portfolios lead to a feasible set of investments, which ultimately results in a convex set of all feasible assets.

Introducing the Risk-Free Asset

The lecture shifts focus to incorporating a risk-free asset into the investment strategy. This asset is defined as having zero variance and potentially some expected return.

A graph illustrates how combining this risk-free asset with other risky assets allows for new investment opportunities along a linear line between the risk-free rate (r_0) and the return on asset 2 (alpha_2).

Benefits of Including a Risk-Free Asset

Investing in the risk-free asset can improve returns while reducing variance compared to points on the efficient frontier, such as the minimum variance portfolio.

The mathematical approach involves minimizing volatility under constraints that align with target returns, leading to optimal portfolio allocations.

Mathematical Framework for Optimal Portfolios

The Lagrangian method is employed to derive optimal weights for risky assets based on target returns. This process reveals how lambda_1 influences allocation decisions across these assets.

As lambda_1 increases, so does allocation towards risky assets, maintaining proportional investments while adjusting overall weightings to achieve desired returns.

Market Portfolio Dynamics

A fully invested portfolio in risky assets is termed the market portfolio. Closed-form expressions are derived for its expected return and variance.

The optimal portfolio structure maximizes mean return across all portfolios by aligning with tangent lines that intersect at the market portfolio point.

Key Insights from Tobin's Paper

Optimal Portfolio Construction and Capital Market Theory

Understanding Optimal Portfolios

The optimal portfolio is defined by the amount of capital allocated, with all optimal portfolios investing in the same risky assets as the market portfolio but differing only in total weight.

For a portfolio P to achieve an expected return of alpha_0 , it will invest in both risk-free assets and market returns, with weights determined by specific expressions.

Capital Market Line (CML)

The volatility of portfolio P is calculated as the square of its weight in the market multiplied by market volatility, leading to the definition of the Capital Market Line (CML).

The CML's slope is derived from the difference between market return alpha_m and risk-free rate r_0 , divided by market volatility sigma_m .

Risk and Return Dynamics

The relationship between risk and return indicates that higher risks should yield higher returns; this principle underpins investment decisions regarding how much risk to take.

By borrowing at a risk-free rate, investors can leverage their investments in the market portfolio, achieving points on the efficient frontier beyond traditional limits.

Historical Context and Foundations

A list of foundational papers on classical finance theories is provided, emphasizing their accessibility online for further reading.

Notable contributors to these theories include Nobel Prize winners like Markowitz and Sharpe, highlighting their significant impact on modern finance.

Utility Theory in Portfolio Optimization

Von Neumann-Morgenstern Utility Theory

This theory emphasizes decision-making under uncertainty through utility functions that maximize expected wealth outcomes.

Rational decision-making involves preferring higher returns while managing variability based on individual utility function definitions.

Wealth Utility Function Characteristics

The utility function reflects greater satisfaction with increased wealth but may taper off as wealth grows due to diminishing marginal benefits.

Taylor Series and Utility Functions in Economics

Taylor Series Approximation of Utility Function

The utility function can be expanded using a Taylor series around a base wealth W^* , resulting in an expression that includes the first and second derivatives of the utility function.

The expected utility derived from this approximation is proportional to expected wealth minus half the product of risk aversion ( lambda ) and variance, indicating a trade-off between expected return and volatility.

Types of Utility Functions

Economists commonly utilize various utility functions such as linear, quadratic, exponential, power, and logarithmic functions for modeling investment choices.

Quadratic utility leads to mean-variance analysis being appropriate under von Neumann-Morgenstern expected utility theory, focusing on expected wealth and its variance.

Limitations of Quadratic Utility

Mean-variance analysis may not be optimal if other types of utility functions are used that account for skewness or kurtosis in returns.

If asset returns are assumed to be Gaussian distributed, mean and variance alone can characterize portfolio distributions effectively.

Portfolio Constraints in Optimization

Practical Portfolio Constraints

Real-world portfolio optimization involves constraints such as long-only positions (weights must be positive), holding limits on specific assets, and turnover constraints when adjusting portfolios over time.

Benchmark Exposure Constraints

Investors may want their portfolios to closely resemble market indices like the S&P 500 while aiming for better performance; thus, they impose limits on how much their allocations deviate from benchmark weights.

Tracking Error Constraints

Tracking error measures the variability between a portfolio's returns and those of an underlying benchmark. Controlling this error helps maintain alignment with desired benchmarks while managing risk exposure.

Risk Factor Constraints

Portfolio Optimization and Constraints

Understanding Weight Constraints in Portfolio Optimization

The discussion begins with the concept of constraining weights to achieve zero exposure across various market factors, highlighting the importance of managing risk.

Minimum transaction sizes are introduced, noting that trades typically occur in 100-share units, although this is evolving. Integer constraints become significant when trading high-value assets like Google stock compared to lower-value stocks like Ford.

All constraints can be expressed as linear and quadratic constraints on portfolio weights, allowing for a structured approach to portfolio optimization by incorporating additional Lagrange multipliers.

Example: US Sector Exchange-Traded Funds (ETFs)

An example involving US sector ETFs from 2009 to the present is presented. These funds represent different industrial sectors within the US market.

A graph illustrates cumulative returns across nine ETFs over time, indicating varied performance among sectors such as materials, healthcare, consumer staples, technology, and utilities.

Risk vs. Return Analysis

The analysis focuses on annualized risk versus return for these ETFs, revealing a volatility range between 0% and 30% and an annualized return potential between 0% and 25%.

A mean-variance optimization is applied with a constraint limiting investment to no more than 30% of capital in any single ETF. This leads to insights about optimal allocations based on varying target returns.

Impact of Capital Constraints on Allocations

As target returns increase from minimum values upward, allocation patterns shift; consumer staples receive high weight initially but face limits due to the imposed capital constraint.

When reaching the maximum allowable investment per asset (30%), reallocations occur towards other securities like consumer discretionary as investments in constrained assets cannot increase further.

Efficient Frontier and Portfolio Performance

A stacked allocation graph shows how capital distribution changes under constraints; higher expected returns necessitate higher risks while reallocating away from safer assets like consumer staples.

The efficient frontier graph demonstrates that optimized portfolios yield better risk-return profiles than individual ETFs. For instance, an optimal portfolio targeting a 10% return showcases improved outcomes despite capital constraints.

Discussion on Risk-Free Asset Constraints

Portfolio Optimization and Risk Management

Impact of Allocation Constraints on Portfolio Performance

The discussion begins with the effects of tightening maximum allocation constraints from 30% to 15%, leading to earlier capital constraints that necessitate diversifying into other exchange-traded funds (ETFs).

A graph illustrates how allocations change under a 15% constraint, revealing that the efficient frontier is lower, indicating reduced potential for higher returns due to limited investment in high-risk securities.

The example emphasizes insights into portfolio optimization, noting that historical performance data may not reflect realistic future settings when defining portfolios.

Market-Neutral Strategies and Sector Models

Transitioning from ETFs, the focus shifts to hedge fund strategies utilizing sector pricing models for long/short investments across various sectors, aiming for market neutrality.

A graph presents multi-factor pricing models for trading market-neutral programs over five years, showing modest total returns ranging from 20% to 60% across different models.

The discussion highlights diversification benefits from these models due to their lower correlations, with optimal allocations favoring certain sectors like utilities and industrials.

Benefits of Portfolio Optimization Techniques

Achieving a target volatility of 10% through combining different market-neutral trading models demonstrates significant advantages in portfolio optimization due to low correlation among assets.

Emphasizes the importance of using estimated returns, volatilities, and correlations in portfolio construction while acknowledging the impact of estimation errors on results.

Advanced Risk Measures Beyond Volatility

Various methodologies are discussed for estimating variance-covariance matrices more accurately than traditional sample estimates by employing dynamic factor models or exponential moving averages.

Alternative risk measures such as mean absolute deviation and semi-variance are introduced as potentially more appropriate depending on distributional assumptions compared to standard volatility measures.

Value at Risk (VaR) and Conditional Value at Risk (CVaR)

Value at risk (VaR), defined as the threshold level of loss at a specified probability (e.g., 5%), is presented as a simple yet effective risk measure widely used in portfolio management.

Conditional value at risk (CVaR), which assesses expected losses exceeding VaR thresholds, is highlighted as an important extension gaining traction in regulatory frameworks for banks.

Understanding Value at Risk and Coherent Risk Measures

The Role of Assets in Risk Analysis

The effectiveness of value at risk (VaR) depends on the type of assets being analyzed, such as stocks or bonds.

For simple investments like cash instruments, VaR is considered a reasonable measure for assessing risk.

When dealing with derivatives that have non-linear payoffs, the analysis becomes more complex and requires case-by-case handling.

There is an ongoing discussion about coherent risk measures that can provide deeper insights into risk analysis.