What are the main criticisms of MVO and how do you address them for CFA Level III?

Mean-Variance Optimization (MVO) is the foundational framework for asset allocation, but it has well-documented problems. Here are the key criticisms and the solutions the CFA curriculum emphasizes: **Criticism 1: Extreme Sensitivity to Inputs** Small changes in expected returns produce dramatically different optimal portfolios. The optimizer 'maximizes errors' — it overweights assets with overestimated returns and underweights those with underestimated returns. *Fixes:* - **Reverse optimization (Black-Litterman):** Start with market-implied equilibrium returns and adjust with your views. This anchors the inputs to a sensible starting point. - **Shrinkage estimation:** Blend historical estimates with a structured estimate (like the global minimum-variance portfolio's implied returns). This pulls extreme estimates toward a central value. - **Resampled MVO (Michaud):** Run MVO hundreds of times with slightly perturbed inputs (Monte Carlo), then average the resulting allocations. Produces more stable, diversified portfolios. **Criticism 2: Concentrated, Unintuitive Allocations** MVO often allocates 0% to many asset classes and extreme amounts to a few — portfolios that no practitioner would implement. *Fixes:* - **Constraints:** Add minimum and maximum weight bounds (e.g., no asset class below 5% or above 40%). Simple and practical. - **Black-Litterman:** Produces naturally more diversified portfolios because it starts from a diversified equilibrium. **Criticism 3: Single-Period Framework** MVO is a one-period model that doesn't account for changing opportunity sets, income flows, or multiple time horizons. *Fixes:* - **Monte Carlo simulation:** Simulate thousands of multi-period paths to test how portfolios perform over time with realistic features (cash flows, rebalancing, taxes). - **Dynamic programming / multi-period models:** More complex but theoretically correct for lifetime planning. **Criticism 4: Assumes Normal Returns** MVO uses only mean and variance, ignoring skewness and fat tails. *Fixes:* - **Mean-CVaR optimization:** Replace variance with Conditional Value at Risk to account for tail risk. - **Non-normal optimization approaches** that incorporate higher moments. **Criticism 5: Illiquid Assets Don't Fit** Private equity, real estate, and infrastructure have stale pricing that understates their volatility and correlation, making them appear artificially attractive in MVO. *Fix:* - **Unsmoothing adjustments:** Correct for appraisal-based return smoothing before inputting into the optimizer. **Exam Priority:** Black-Litterman, resampled MVO, shrinkage, and adding constraints are the most heavily tested fixes. Know when to recommend each. Test your understanding with our CFA Level III Asset Allocation practice questions.