Mean-Variance Model · Portfolio Theory
The Core of Harry Markowitz's Modern Portfolio Theory (MPT) — Finding the Optimal Balance Between Risk and Return
Core Formulas
Portfolio return is the weighted average of each asset's return, where the weight $$w_i$$ is the proportion of asset $$i$$ in the portfolio.
Portfolio risk depends not only on each asset's own risk, but even more critically on the correlation coefficient $$\rho_{ij}$$ between assets.
When $$\rho_{12} < 1$$, portfolio risk is lower than the weighted average of the two assets' risks—this is the diversification effect.
Parameter Description
$$w_i$$: Investment weight of asset $$i$$; the sum of all weights is 1 (i.e., 100%). $$E(R_i)$$: Expected annualized return of asset $$i$$. $$\sigma_i$$: Annualized standard deviation of asset $$i$$, measuring the asset's volatility risk. $$\rho_{ij}$$: Correlation coefficient between assets $$i$$ and $$j$$, ranging from $$[-1, 1]$$. A negative value means the two assets move in opposite directions, giving better diversification.
Interactive Simulation · Two-Asset Portfolio
Standard Introduction
The mean-variance model was proposed by Nobel laureate Harry Markowitz in 1952 and is the cornerstone of Modern Portfolio Theory (MPT). The model uses expected return (mean) to measure investment return and variance/standard deviation to measure investment risk, and through mathematical optimization maximizes return for a given level of risk or minimizes risk for a given return target, thereby constructing the efficient frontier—the set of all optimal portfolios.
Plain-Language Introduction
Imagine you have two friends: one runs a business (high return but unstable), and the other works a salaried job (stable but ordinary return). You lend money to both at the same time instead of lending it all to one—this way, your overall return is smoother and risk is lower. The mean-variance model uses mathematical methods to calculate how much to invest in each so as to achieve the best balance between "earning the most" and "losing the least." Its core insight is that holding assets together is safer than holding them separately, which is the mathematical expression of "don't put all your eggs in one basket."
What Problem Does It Solve?
- Quantifying the risk-return trade-off: For the first time, investment risk is precisely measured using statistical methods (variance).
- Mathematical foundation of diversification: It proves that as long as the correlation coefficient between assets is less than 1, diversification can reduce overall risk.
- Optimal asset allocation decisions: It provides a scientific method for allocating funds among multiple assets rather than relying on intuition.
- Introducing the efficient frontier concept: It helps investors identify which portfolios are "efficient"—that is, no other portfolio can provide higher return at the same risk.
Why Is It Important?
The mean-variance model completely transformed the investment management industry. Before this, investment decisions mainly relied on subjective judgment. Markowitz's theory provided a quantifiable and replicable scientific framework for asset management and directly gave rise to modern financial tools such as index funds, quantitative investing, and Robo-Advisors. Its influence extends from academia to Wall Street, and it remains the core theoretical foundation of global portfolio management today, as well as required content for professional qualifications such as the CFA.
Application Scenarios
- Pension and insurance fund management: Pursuing stable returns under strict risk constraints.
- Robo-Advisors: Platforms such as Betterment and Ant Fortune automatically build portfolios based on users' risk preferences.
- Major asset class allocation: Allocating funds among asset classes such as stocks, bonds, commodities, and cash.
- Fund of Funds (FoF) optimization: Building a "fund of funds" and optimizing allocation among funds with different strategies.
- Strategic Asset Allocation (SAA): Determining the policy benchmark weights for long-term investment.
Model Deficiencies and Limitations
- Relies on historical data: It uses past returns and risk to predict the future, but the market environment may undergo structural changes.
- Assumes normal distribution: Actual financial data exhibit "fat tails," and extreme events (such as the 2008 financial crisis) far exceed model predictions.
- Parameter sensitive: It is highly sensitive to small changes in expected returns, variance, and correlation coefficients, and easily produces unstable extreme weights.
- Single-period static model: It does not consider multi-period dynamic adjustment and rebalancing costs.
- Ignores skewness and liquidity: It focuses only on mean and variance and does not consider the asymmetry of return distributions or the difficulty of liquidating assets.
- Homogeneity assumption: It assumes all investors have the same expectations and investment horizon, which does not match reality.