Mean Variance Optimization Limitations

Mean Variance Optimization (MVO) is currently one of the mainstream asset allocation approaches used in modern finance and is a core component of the modern portfolio theory. The theory distills securities to just return and risk (as measured by security price standard deviation) and assumes investors seek the highest return for the lowest level of risk (again as measured by the standard deviation of portfolio returns)

Today computer programs take into account the correlations between asset class returns to identify combinations of asset classes that offer highest possible returns for a given level of acceptable risk.

Key assumptions behind MVO:

1. Risk averse investors
2. No taxes or transaction costs
3. Investor returns are normally distributed
4. All returns, correlations and variances are known

As you can see these assumptions are extremely strong and discredit the approach significantly

Here are just a few criticisms

(CRISLS)

1. Concentrated positions - MVO produces highly concentrated asset allocation positions that may not be acceptable for investors

2. Investors care not only about returns and variance but other factors as well

3. The MVO models are highly sensitive to small changes in the inputs

4.  Sources of risk may be different so in essence the MVO asset allocation could be poorly diversified due to high exposure to a single risk source.

5. No Liabilities - MVO does not have a mechanism to account for liabilities

6. Single period model