Perhaps my biggest disappointment with Gary Antonacci’s book on “Dual Momentum” was that he only covered his basic Global Equity Momentum (GEM) system that chooses between US Equities, International Equities or US Bonds. The back test results of applying this strategy to liquid ETFs representing these asset groups was presented in Part 1 of this series of Posts.
However, in his 2013 paper entitled “Risk Premia Harvesting Through Dual Momentum” (available for download at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2042750) Antonacci extends this concept to other “Risk” groups.
His first “Risk” group (Equity/Sovereign Risk) is the same as that used in the GEM strategy, i.e. it compares the “Relative Momentum” of US Equities with International Equities and selects the asset with the highest “Relative Momentum”. If the “Absolute Momentum” of that asset is positive, it is included in the portfolio. The methodology changes slightly from the “strict” GEM strategy at this point in that, rather than investing in an Aggregate Bond asset, if the Equity asset has negative “Absolute Momentum” funds are allocated to T-Bills rather than to an Aggregate Bond asset. However, as we saw in Part 1, this change has very little impact on performance. In this post I will use VTI, VEA and SHY (as used in Part 1) to represent the selection options for this “Risk Group”.
Having taken Bonds out of the equation in the first “Risk Group”, Antonacci defines his second “Risk Group” as “Credit Risk” and defines this as a choice between High Yield Bonds and Credit Bonds. The methodology is the same as before – a comparison of “Relative Momentum” between the two Bond assets, followed by a check that “Absolute Momentum” is greater than returns on T-Bills. In the back tests that follow I will use HYG to represent High Yield Bonds and CIU to represent Credit Bonds. SHY remains as the check against T-Bills.
To expand portfolio diversity beyond just equities and bonds Antonacci next turns to Real Estate to define his third “Risk Group”. He breaks this group into equity and mortgage REITs and I will represent these assets through the use of VNQ (equity) and REM (mortgage). Selection methodology is the same as above, again with SHY as proxy for the T-Bill cut-off.
Finally, Antonacci defines his fourth “Risk Group” as his “Economic Stress” group – where investors might go in times of weakness in the economy. This group is represented by Gold (to reflect uncertainty concerns) and Long Term Treasuries (to reflect a flight to quality). In the following back-tests I shall use GLD and TLT to reflect this “Risk Group”. Again, the methodology is the same – a “Dual” asset comparison, followed by a check of positive “Absolute Momentum” against T-Bills (SHY).
Having defined his 4 “Risk Groups” Antonacci then selects 1 asset from each group to include in his portfolio. The selected assets are equally weighted. The maximum allocation to each asset will be 25% with the possible exception of T-Bills (SHY) that could be higher if other assets are determined to have negative “Absolute Momentum”.
When set up in the Ranking Spreadsheet the outputs look as follows:
Note that, for a single look back ranking period, the “Overall Rank” is equal to the single period (ROC2) rank, making the ranking process very simple.
Note also that the assets are split into the 4 “Risk Groups” and are identified as “Clusters”.
Allocations are shown in the above figure and also in the allocation table:
In the above example, all ETFs except VEU (International Equities) and Gold (GLD) have positive 12 month “Absolute Momentum” and the ETF with the highest Relative momentum in each “Risk Group” is chosen for inclusion in the portfolio.
Using Antonacci’s suggested 12 month look back period, our 4-group “Dual Momentum” portfolio performance is shown below (solid dark blue line) and compared with VTTVX (solid brown line) and the GEM (Equity/Bond) portfolio described in Part 1 (light blue line):
Immediately we notice the significant improvement in the performance of the (12 month look back period) 4-group, diversified portfolio, compared with the single group (Equity/Bond) portfolio.
CAGR = 9.41%; Volatility = 8.90%; Return/Risk (Sharpe) = 1.06; MDD = 8.69%
We note low volatility and draw-down with a healthy 1.06 Sharpe ratio.
As in Part 1 I also ran tests on this 8 asset list using a 6 month look back period (dashed dark blue line) and our “standard” 3/6 month momentum + volatility ranking system (dashed red line). Performance is shown below:
6 month look back:
CAGR = 8.61%; Volatility = 8.19%; Return/Risk (Sharpe) = 1.05; MDD = 9.35%
3/6 month + volatility (50/30/20):
CAGR = 9.73%; Volatility = 7.57%; Return/Risk (Sharpe) = 1.28; MDD = 7.13%
Although the statistics for the “standard” 3/6 system are a little better than the other 2 systems over this specific 8 year period I would classify all systems as being acceptably “robust”.
However, before leaving this analysis, let’s remind ourselves what we found using the 6 month (dashed light blue) and 3/6 month (solid red line) systems with the 3 asset GEM model:
Initially, the superior performance of the 3 asset GEM systems might seem a little confusing. However, we have to remember that the US Equity markets have outperformed all other market sectors over the past ~3 years – so it is not too surprising that an undiversified portfolio (equities or bonds) that is allowed to invest 100% in the best performing market segment should perform better than a diversified portfolio that can only invest 25% in that market sector. This is an excellent example of the need to look at market conditions within “floating” time windows over long time periods.
I would suggest that the diversified 4 (risk) group system is probably a far better (more “robust”) system to adopt for long term investment. It will likely continue to exhibit low volatility and drawdowns with “acceptable” (risk-adjusted) returns. However, I would question whether it is necessary to achieve this (low volatility and drawdown) performance by requiring all 4 groups be represented (and restricting maximum allocations to 25%) i.e. why not just take the 2 or 3 highest ranked assets (keeping the group/cluster filter) or at least use a weighting methodology that allows higher (than 25%) allocations.