"It's not the number of asset classes that determines diversification; it's the correlation among those asset classes that determines diversification." - Richard Bernstein
How does one build a well-diversified portfolio using low correlated securities so as to reduce risk? This article explains the basics steps of how an investor can construct a portfolio though what is known as the "Cluster Weighting Momentum" model or pairing equity oriented securities with low correlated securities. While this process is more complicated than simply developing an asset allocation plan, The Feynman Study, covering both bear and bull markets, shows promising results for adding alpha while reducing risk. Here is a brief outline of the steps necessary to construct a "cluster" type portfolio.
- Determine what securities will be used in the portfolio. The number can range from 15 to 30 if using ETFs as ETFs provide diversity right out of the gate. In the following example I use over 20 ETFs.
- SHY is the cutoff ETF. In other words, don't invest in a security that is under performing the 1-3 year treasury bond.
- Some method for ranking securities to determine if they are outperforming SHY is required. An example of such a spreadsheet is demonstrated in this YouTube video.
- A second software program creates what we call a Cluster Hierarchy. An example is shown in the second screenshot below.
Here is a current screenshot of the ETF ranking. The ranking is based on three- and six-month price performance as well as a volatility factor. The two right-hand columns show a heat map of the acceleration of momentum, both absolute and relative. This type of ranking model was discussed in two earlier Seeking Alpha articles.
(click to enlarge)
In addition to the above ranking spreadsheet, a commercial add-in to Excel is necessary in order to develop what is known as a Cluster Hierarchy. I am only a user of this software and have no financial connections with the Hoadley software.
Clustering ETFs: From a correlation matrix the following clustering hierarchy is created. One needs to manually determine which securities fall into cluster 1, 2, 3, etc. up to 10 clusters. Once this decision is made, the ranking spreadsheet automatically calculates what percentage is to be invested in ETFs with the highest rank, but only if the ETF is outperforming the cutoff ETF, SHY.
If you look at the bottom of the cluster, GLD and DBC each fall into individual clusters (9 and 10) and they both have a very low correlation with all the equity ETFs found at the top of the cluster diagram. However, both are under performing SHY and are therefore eliminated from the portfolio at this time.
Rankings: The "Cluster Weighting Momentum" model narrows the number of securities to six. Large-Cap Growth (NYSEARCA:VUG), International Select Dividend (NYSEARCA:IDV), and Emerging Markets (NYSEARCA:VWO) are the equity trio that are paired with three low correlated ETFs. Those three are: 1-3 yr. Treasury (NYSEARCA:SHY), High Yield Bonds (NYSEARCA:JNK), and International Treasury Bonds (NYSEARCA:BWX).
This "Cluster Weighting Momentum" model is undergoing additional testing with several portfolios at ITA Wealth Management. Testing results from late June of 2007 through late June of 2013 is available on the ITA blog.
Additional disclosure: This model is the creation of "HedgeHunter" an author on the ITA Wealth Management blog. I am using of modification of this model with several portfolios.