Is There an Advantage in Splitting a Portfolio into Tranches? – Part 4: Combining Momentum with Strategic Asset Allocation (SAA)

In previous posts in this series I have used equal (dollar) weighting to determine fund allocations to selected qualifying assets.

When Lowell first started this blog (and before I corrupted it with momentum ? ) it was focused directly on passive (Buy and Hold) investing (with maybe occasional rebalancing) of portfolios containing selected assets with strategic allocation of funds consistent with theories of minimum variance through diversification of asset types.

It was when I asked the question “Why would anyone hold an asset, purely for the sake of diversification, when it was obvious that the asset was going down in price?” that Lowell invited me to become a contributing author on his blog. Our objective was to try to provide ITA members with an education on the strengths and weaknesses of alternate strategies and to combine the best features of available investment strategies into the management of different portfolios. This effort would be supported by applications to portfolios with “real” money invested (to keeps us honest ?)

In this post I thought it would be appropriate to take a look at momentum applied to a portfolio of assets to which allocations were strategically defined. Since, throughout the earlier tests in this series of posts, I have used the 10 asset “Rutherford” list to demonstrate the application of various asset selection strategies and tranching techniques, it is appropriate that I continue to use the same asset list in this this post.

I will therefore start by assigning strategic allocations to the Rutherford Assets:

Note that this allocation plan is consistent with many “classic” SAA portfolios (see e.g. https://portfoliocharts.com/portfolios/) – It comprises 60% Equities (split between US and International markets), 20% bonds (US and International), 10% Real Estate (US and International) and 10% in other inflation (gold) and commodity products.

While the allocation to bonds may seem a little low, I have considered the fact that, under the “rules”, funds not qualifying for inclusion in the portfolio might be allocated to SHY (short term treasuries) – hence increasing the allocation to the “bond” category. [In the following tests, allocations to assets not qualifying for inclusion in the portfolio are assigned to Cash rather than SHY. This is purely laziness on my part since this would require more work to build the logic into my spreadsheets – but, since interest rates are so low, this is not expected to materially change the results of the tests].

As for the “rules”, they remain very similar to the tests reported in Parts 1-3 of this series except that for the “Rank” strategy ALL assets ranked higher than SHY are included in the portfolio (rather than only the top five) – the big difference being that the asset allocations are fixed by the above SAA Plan rather than being equally weighted and dependent on the number of qualifying assets.

As in previous posts, the portfolio is split into 4 tranches:

…. And we see the normal variances (although not looking quite as severe).

Combining the 4 tranches into the “Rank” portfolio we see the following performance:

$1,704 return (8.5%) in 3+ years.

In the second strategy we again take the assets in the top 3 groups – but allocated in accordance with the SAA Plan rather than the equal weighting used in Part 2 of the study series:

… combined:

$1,075 (5.4%) in 3 + years.

Finally, we will again take assets from the top 4 groups with positive Heikin-Ashi signals – but allocated in accordance with the SAA Plan rather being equally weighted:

…. And combined:

$605 (3%) in 3+ years.

Comparing the 3 strategies when applied to the SAA Rutherford Portfolio:

Performance in tabular format is shown in the following table:

This can be compared with the table summarizing the results of the “equal weighting” tests (from Part 3 of this series):

Observations and Conclusions:

1. It is still difficult to beat an “equally weighted” portfolio in terms of returns – however, volatility in the SAA portfolios are lower and return/risk is generally higher (not quite as good for the Group_F (HA filter) tests).
2. The differences in results for Group_NF and Group_F tests are due entirely to the weighting strategy since the chosen assets at all points in time are the same for the “equal weight” and “SAA weight” tests.
3. Since the “SAA Rank” portfolio allows all assets ranked higher than SHY to be included in the portfolio, this portfolio will invariably hold more than (or at least equal) the number of assets held in the equally weighted portfolio – this generally leads to lower returns and volatility.
4. Based on the above results, although the SAA Rank strategy does not generate the highest returns (2 other options show higher returns) the low volatility and superior return/risk performance make it look attractive.

Caveats:

Although I have deliberately tried not to “mine” the data I am very sensitive to the fact that, over the past 3 years, the US equity markets have outperformed all other asset classes represented in this portfolio. Therefore, assigning a 30% weighting to VTI might inadvertently introduce an unintended “bias” to the results. I don’t know what else I can say about this – it is what it is – I did consider a 40/15/5 split between the equity assets (since I wanted something like a 60/40 portfolio – 40 bonds with benefits) but chose 30/20/10 because I thought it was more conservative. Recall that Swensen suggests a 30/15/5 split for his 6 asset portfolio – so I may still be a little heavy on equities.

3 years is a very short period for a test – no argument there – but I think the last 3 years represent a very significant and difficult period for investors – and a good period over which to check for robustness.

I encourage all members to share their reactions to this data – I would expect a wide diversity of opinion – and none of it wrong – just appropriate to each investor.

In the future I plan to run similar tests on a different set of assets – probably 40 assets – to see if there might/might not be an advantage in adding more specific sector or regional ETFs to the “quiver” of available assets. I haven’t yet started this project – so don’t hold your breath waiting for the results ?

Update: 15 May, 2017

In response to Jimmy Smith’s question in the Comments section below I am adding the plot of performance of the Rutherford SAA portfolio as a buy and hold strategy:

Here we see that the return of the buy and hold portfolio is about two times that of the momentum strategy. Obviously, over this specific time period, buy and hold would have been the better strategy (at least in terms of eventual returns). The big advantage of a momentum strategy comes in times of deep distress and strong pullbacks/bear markets. However, for investors looking for some solace it should be noted that maximum drawdown of the momentum strategy is only ~5% ($1,000) compared to ~15% ($3,000) for the buy and hold strategy – i.e. less volatility/risk. [The difference in drawdown between that shown in the above figure and Jimmy’s link figure is due to the fact that the above figure shows drawdown on a day over day basis rather than a month over month measurement].

Note that the Rutherford SAA buy and hold performance runs pretty close to that of the 70/30 (Equity/Bond) VTTVX Fund that I tend to use as a reference and difficult benchmark to beat.

David