What are the practical limitations of mean-variance optimization, and how do risk managers address them?
Markowitz's mean-variance framework seems elegant in theory, but I keep reading that it produces 'garbage in, garbage out' portfolios. What specific problems arise when you actually run an optimizer, and what techniques are used to produce more robust allocations?
Mean-variance optimization (MVO) finds the portfolio weights that either maximize expected return for a given risk level or minimize risk for a given return — tracing out the efficient frontier. While mathematically elegant, MVO has well-documented practical problems that every FRM candidate should understand.
The Core Problem: Input Sensitivity
MVO is extremely sensitive to small changes in the input estimates (expected returns, volatilities, and correlations). A 50-basis-point change in one asset's expected return can cause the optimizer to swing wildly between 0% and 60% allocations to that asset.
Specific Limitations
| Problem | Description |
|---|---|
| Estimation error amplification | Optimizer maximizes exposure to the most overestimated returns |
| Concentration | Unconstrained MVO often produces extreme, undiversified portfolios |
| Instability | Small input changes produce large allocation shifts — impractical for rebalancing |
| Ignores higher moments | Only uses mean and variance; ignores skewness and kurtosis |
| Single-period | No consideration of multi-period rebalancing, liabilities, or liquidity |
| Symmetric risk | Treats upside and downside volatility equally |
Worked Example
Glenfield Endowment runs MVO across 8 asset classes using 10 years of monthly data:
Unconstrained optimal portfolio: 72% US Large Cap, 31% Emerging Market Bonds, -3% TIPS (short)
This is clearly impractical — dominated by two asset classes with negative allocations.
Practical Remedies
Black-Litterman Example
Instead of using raw historical returns, Glenfield starts with the equilibrium implied returns (reverse-engineered from market-cap weights) and overlays their investment committee's views:
- View 1: EM equities will outperform DM equities by 2% (moderate confidence)
- View 2: Commodities will return 5% (low confidence)
The resulting allocation is much more stable: 28% US LC, 18% Intl Dev, 12% EM, 15% US Bonds, 10% TIPS, 8% Real Estate, 5% Commodities, 4% EM Bonds.
Resampled Frontier
Run the optimizer 500 times, each time drawing inputs from a distribution around the estimated means and covariances. Average the resulting weights. This produces portfolios that are statistically indistinguishable from the "true" optimum but far more diversified and stable.
Exam Tip: FRM questions often present an optimizer output and ask you to identify which limitation is causing the problem (concentration = estimation error; instability = input sensitivity; poor out-of-sample performance = overfitting to historical data).
For more on portfolio construction, explore our FRM Part I course.
Master Part I with our FRM Course
64 lessons · 120+ hours· Expert instruction
Related Questions
How exactly do futures margin calls work, and what happens if I can't meet one?
How do you calculate the settlement amount on a Forward Rate Agreement (FRA)?
When should I use Monte Carlo simulation instead of parametric VaR, and how does it actually work?
Parametric VaR vs. Historical Simulation VaR — when does each method fail?
What are the core components of an Enterprise Risk Management (ERM) framework, and how does it differ from siloed risk management?
Join the Discussion
Ask questions and get expert answers.