- Improve diversification through Cluster Weighting Momentum Analysis.
- How to avoid the next Bear Market.
- Using SHY as a "circuit breaker.".
One sample portfolio presented in Mebane Faber and Eric Richardson's book, The Ivy Portfolio: How to Invest Like the Top Endowments and Avoid Bear Markets, allocates equal percentages to the 20 ETFs listed on page 76. With this allocation one ends up investing the same percentage (20%) in four commodities as are assigned to U.S. equities. Is it possible to improve on the equal percentage plan and what might motivate one to make any changes?
To investigate how a portfolio would have performed during the Great Recession, I "invested" in the Ivy 20 ETFs that were available on 6/30/2007 and held them through 7/29/2014. This starting date includes the run-up before the market crashed and captures the bull market from March 2009 through the present date.
Due to data issues, I substituted the ETF WOOD for TREE.L. If an ETF did not exist on 6/30/2007, I purchased 5% on the day the ETF became available.
From 6/30/2007 through 7/29/2014 the Internal Rate of Return (IRR) of the Ivy 20 was 3.8% while the benchmarks, VFINX and VTSMX, were 6.1% and 6.6% respectively. While dividends are included for both benchmarks, they are not included for the Ivy 20 portfolio, so that accounts for part of the difference. As mentioned above, not all of the Ivy 20 were available on the starting date. Two were purchased just in time to catch the bear market of 2008.
Those ETFs that carried the brunt of the positive growth load were: VTI, VO, VB, ESD, DBP, BND and TIP. What I did not do, and this no doubt would have increased the portfolio IRR, is to sell the ETF when its price dropped below its 200-Day Simple Moving Average. This timing model is discussed in great detail in The Ivy Portfolio book.
The Cluster Weighting Momentum model (explained below) had the Ivy 20 entirely in cash or SHY on 9/30/2008.
In addition to diversifying over 20 ETFs, and applying a timing model, is there any other way to enhance performance? We know that certain ETFs add value during certain periods while others are a drag on the portfolio. The following model, known as Cluster Weighting Momentum (CWM) is designed to diversify the portfolio into low correlated groups and sell off underperforming ETFs.
CWM analysis ranks the ETFs, finds how they are correlated, breaks the ETFs into 10 clusters, and recommends percentages to be invested in the top ETF of each cluster so long as it is outperforming SHY, our cutoff ETF.
Ivy 20 Portfolio: Here we have the portfolio recommended in The Ivy Portfolio book. As mentioned above, WOOD is substituted for TREE.L due to data issues. The 100 shares assigned to each ETF is only for illustration purposes and certainly does not represent 5% invested in each ETF.
Ivy 20 Rankings: Using current data, the following table shows the ETF rankings. Three ETFs, DBE, DBP, and IWC are sold out of the portfolio as all three are lagging the performance of SHY, our "circuit breaker" ETF. We do not hold ETFs that occupy this position. An exception might be made for IWC as it has a positive absolute acceleration percentage.
Instead of buying the top ranked ETFs, the next step in the analysis is to locate the best ETFs that also carry low correlations.
Ivy 20 CWM Recommendations: The table below breaks the 20 ETFs into 10 clusters when the correlation screen is set to 0.65. For example, VO, VTI, VEU, GWX, WOOD, VB, and IWC all have a correlation of 0.65 or higher. VTI is the top performer from Cluster #1 and the recommended percentage is 8.7%. Cluster #8, for example, is represented by TIP.
No ETFs are recommended from Clusters 5, 6, 9, or 10. In several cases the ETF representing a particular cluster is under-performing SHY and in other cases it is tied to the cluster filter setting. In this example the cluster filter screen is set to extract the best performing ETFs from the top six clusters.
Due to the short life of many ETFs in the Ivy 20, it is not possible to test how well the Cluster Weighting Momentum (CWM) model works going back to 6/30/2007. When back-tested with ETFs that go back over ten years, the results are positive or the performance tops the market benchmark. The real test is - will CWM work going forward? Such an experiment is in progress with real portfolios that are reviewed every 33 days.