Working with Dr. Harry Markowitz on the Loring Ward Investment Committee has taught me to never look at any type of change to a portfolio until I’ve done extensive mean variance optimization (MVO) work. However, there are about 100 different decisions you must make when running an MVO process and it starts with your input assumptions.

There are two schools of thought on what to use for the expected characteristics of an investment: you can use an Asset Class proxy, such as the Russell 2000 index for small cap exposure; or you can use a Risk Factor proxy, such as the difference between the Russell 2000 – Russell 1000, the small cap premium.
The Asset Class provides you with the simple return of that particular index. This is easy to understand and compare but has the potential downside of having a high correlation to the overall market (the Russell 2000 returned 38.8% while the Russell 1000 returned 33.1% in 2013).
The Risk Factor isolates only the relative return of small cap compared to large cap. While it’s useful to know exactly how small cap companies did comparatively, it can leave you with a relatively lackluster number in a booming market year, just 5.7% for small cap in 2013.
Which is the better measure to use in an MVO process?
This question was asked and tested in a recent paper titled Factor-Based Asset Allocation vs. Asset-Class-Based Asset Allocation by Thomas Idzorek and Maciej Kowara, in the May/June 2013 issue of Financial Analysts Journal.
In an Asset Class scheme, all individual securities should be assigned to an asset class and the various asset classes should be mutually exclusive. With Factors, individual securities may have exposure to multiple risk factors, such as a corporate bond which would encompass both duration risk and credit risk. Because of the shorting needed to gain exposure strictly to a particular premium (for example small cap premium would be gained by buying small cap securities and shorting large cap securities), this type of portfolio would not be able to be universally held (not everyone in the market can short the same segment — someone has to be long).
The authors first compared an MVO of portfolios created using Asset Class to those using Factors for the period 1979-2011, which resulted in the Asset Class frontier dominating the Factors frontier. However, when run for different periods of time (for example the 10 years ending in 2011), the Factor inputs provide more efficient portfolios at some points of the efficient frontier.
Their conclusion is that neither model is better than the other. In some cases the Factors outperform; however, they do so with the constraint of a long-only exposure removed. The apparent superiority of the risk factors is a simple result of the fact that the risk factors are, in a guise, a set of asset classes with the long-only constraint removed.
This topic isn’t one most advisors would worry about, but it is another reason why it makes sense to partner with a team who is able to spend the time and resources to make well-designed decisions in the portfolio creation process.
Source: Morningstar Direct
Factor-Based Asset Allocation vs. Asset-Class-Based Asset Allocation, Thomas Idzorek and Maciej Kowara, Financial Analysts Journal Volume 69 number 3.

Past performance does not guarantee future. Indexes are unmanaged baskets of securities that are not available for direct investment by investors. Index performance does not reflect the expenses associated with the management of an actual portfolio. Stock investing involves risk including loss of principal. Securities of small companies are often less liquid than those of large companies. As a result, small company stocks may fluctuate relatively more in price. International and emerging market investing involves special risks such as currency fluctuation and political instability, and may not be suitable for all investors. Bonds (fixed income) are subject to market and interest rate risk if sold prior to maturity. Bond values will decline as interest rate rises, issuer’s creditworthiness declines, and are subject to availability and changes in price.
IRN R 14-137