Thanksgiving
Reducing portfolio risk is at the forefront of investor thinking after a flat market in 2015. Momentum principles, as laid out in the Kipling 2.5.1 spreadsheet, do not always work as well as preferred, due in part to the 33-day portfolio review waiting period. We see the rankings change when we use 12 Portfolio Offsets. It has been suggested one might split a portfolio into smaller units so the portfolio reviews comes up at different times, much as the 14 portfolio rotations I track on this blog. While I am not recommending one split a portfolio into 14 sections, one suggestion is to divide a portfolio into units of $100,000 to $250,000. Portfolios of this size provide more flexibility than very small portfolios. In addition to looking at different parts of the portfolio at different times, different securities can easily be used for diversification purposes.
Assume a portfolio is $500,000 in size. Breaking it into halves or thirds makes sense. In families this may come naturally as a couple may have two taxable and two tax deferred accounts. If this is the case, the portfolio divisions, while not equal, are present by default.
Even if a portfolio is divided in half and one uses 12 Portfolio Offsets, over a month we have good coverage as to what securities were ranked high during nearly every trading day of the month. Give serious thought to dividing a portfolio into different parts.
Discover more from ITA Wealth Management
Subscribe to get the latest posts sent to your email.
Lowell, I agree this is a good strategy. One question is what assets to use for the momentum calculation? It seems one would want to keep diversity in each portfolio, but diversity over what universe of ETFs? It would not seem optimum to use the same ETFs in every portfolio. If not, what are some alternative portfolios to use?
Thanks.
Ernie
Ernie,
I’ve not given your question a lot of thought, but here are a few ideas. Let’s assume one is going to try to cover most, if not all the critical asset classes. I assume them to be: U.S. Equities, Developed International Equities, Emerging Markets, U.S. REITs, International REITs, Commodities, U.S. Bonds, International Bonds, and Precious Metals.
If one is going to broaden out the number of ETFs kiss commission free ETFs off the list as one will be branching, most likely, into securities that are not commission free. For example, one might focus on sectors in one of the portfolios. That move would branch into non-commission free ETFs. Another portfolio might have a focus of dividend oriented ETFs which would include VIG, VYM, REM, DVY, etc. It makes sense to confine high yield ETFs in tax-deferred accounts.
In one portfolio you might use PCY as the international bond surrogate while BWX might be included in another portfolio. One portfolio might be managed following the strict rules of Dual Momentum as advocated by Gary Antonacci where only three ETFs are used and the look-back is 252 trading days. One of the three or four portfolios might be passively managed as I do with the Schrodinger.
There are both options of what securities to use and how one wants to manage the different portfolios.
What ideas were you thinking of?
Lowell
Lowell, you have some really good suggestions. So far all I have come up with is a set of sectors, a mainly U.S. set of somewhat diversified funds, and a broad international collection. But, there is a lot of duplication in those, if not the same identical fund, a similar fund. I like your idea of dividend-oriented funds.
I also like your management ideas. Right now I manage them all with Kipling with a dash of input looking at a couple of oscillators and EMAs (dangerous, I know). If others reading this thread have ideas, I’d like to hear them. I have several portfolios but have too little diversification in both content and management style among them right now.
Ernie
Ernie,
There is some duplication I want in all my personal portfolios. For example, VTI, VEA, and VWO are ETFs I want available in all portfolios.
One other thought I to pass on. In every portfolio I want several ETFs that can be thought of as bear market “brakes.” ETFs such as BND, BIV, SHY, AGG, PCY, and BWX fall into this category. To a lesser extent JNK and HYG would be part of one or more portfolios. In other words, find ETFs that have a low correlation with VTI.
A portfolio should have 30% to 50% of the securities with a relatively low correlation with equity ETFs such as VTI, VEA, and VWO. DBC and GLD are two that fit that requirement.
Lowell
Ernie,
This what I have been “playing” with. I have 3 IRAs in my name, one in my wife’s name, a cash account and a TIAA account. I break those into 4 groups. Unfortunately the TIAA account is ~1/3 of my investable assets and they have their own proprietary family of funds, so I group the others up into 3 additional “portfolios”. I tend to use the usual ETF mix and a few stocks. For the TIAA account I use their equivalent of our standard ETFs. I than space the rebalance of these 4 portfolios one week apart and review each on the 33 day schedule. Too soon to tell if this improves things but so far it looks promising. I’m always surprised by how much the top 4 rankings change in the market we are now experiencing, so at least psychologically I feel better: ^)
Bob W.
Bob, that’s exactly what I am trying to do. Right now I find myself often investing in the same ETF in the different portfolios because that ETF floats to the top ranking in each portfolio. That’s why I want to diversify the management strategy more and also the holdings. I, too, don’t have a track record yet.
Ernie
Not that I have a nugget large enough to split, but I thought you’d just split it by strategy, and ensure some level of diversification in each:
1. Fundamental (value/ growth etc)
2. Technical (momentum I guess)
3. Income/dividend etc
4. Options hedge
5. Alternatives (long/short, arbitrage, risk parity, business ventures, real estate or something)
6. ????
Heh, maybe you could rank each strategy by performance and apply an investment weighting factor according.
I think doing the same thing with a sight variation 14 times opens you to a risk of common cause failure.
In theory you would need a tool (morningstar?) that allows you to see underlying stock in your ETFs across all your portfolios, cumulatively.
That way, if across your 14 portfolios, you see that a 50% (Or something) exposure to Apple, or Exxon, or whatever, is far more than you want right now and can make an informed decision about that.
Jimmy,
Regardless as to how the portfolios are divided and managed, I would definitely track the performance and benchmark each. It would take a lot of thought to come up with appropriate benchmarks for the management models you suggest.
Lowell
Jimmy, I agree with Lowell. This would be hard to manage.
In my mind, there are two distinct issues: a) portfolio asset content and b) management strategy. I think it would best to diversify both of these. So, my initial question concerned different types of portfolios where there was a diversification within each portfolio and among the portfolios (somewhat difficult to accomplish). Second, it would be nice to manage each with a different strategy where the strategies were diversified. That is, it would be optimum to have some strategies that provided protection or decent performance in sideways and bear markets, and other strategies that performed best in bull markets. At the very least, the management strategies should not be the same class (e.g., they should not all be momentum or trend following). That would be my goal. Not sure how to accomplish that.
Ernie,
I think Lowell’s Mosaic approach tries to partially achieve that within one portfolio. A portion is based on fundamental allocation based on portfolio theory and a portion on momentum. One could use this to cover two separated portfolios.
Bob W.
Bob W.
In some ways the Mosaic Model attempts to find the best of both the passive and active worlds. With the passive portion of the portfolio I am always exposed to the seven critical asset classes. There is abundant research arguing how difficult it is to beat the broad market.
Our own research shows that the Tranche Momentum Model (TMM) outperforms the broad market so that is where we invest another portfolio of the total portfolio. I’ve only been using the TMM approach for about a year so it is early to know whether or not it will work in the future.
Here is how one might attack the Mosaic Model. Invest passively in VTI, VEA, VWO, VNQ, RWX, TLT, and PCY to cover the seven critical asset classes. Commodities and Precious Metals are not included. Then run ETF and stock screens as I’ve done in the latest blog post. Here is the link.
https://itawealth.com/2016/05/24/top-ranking-etfs-stocks/
Decide what percentage goes to the passive side and what percentage should be actively managed using the tranche momentum model.
Unfortunately this model is not easy to back-test, but not impossible either.
Lowell
As I read this thread, I think it is important to be clear on what one is solving for when contemplating making a change to the prescribed plan. I’ve come to some of these conclusions in my own journey, and have had specific reasons for doing so – which filters out certain approaches.
I like the Tranche model and the Baker’s Dozen assets. Much has been posted about both on this site – and I think it all has merit. US tax laws influence the desire for a 30-day holding period, shorter periods are greatly penalized.
As I have set out to implement Tranche / Baker’s Dozen (TB12), I too have been concerned about diversification, protection, and holding period. It’s one thing when your cash is small and your years of income production are long – it’s quite another when those items are reversed.
I set out to simulate trading TB12 to learn how it works under the hood. After my first round of simulation from 2006 to the present, I concluded I would allow up to 3 assets and would impose a max allocation of 33% to any single asset – putting the balance to SHY. My simulated returns were acceptable. However, I noticed a sizeable number of instances where a single position lost at least 1% of portfolio value in a single 33-day period (let’s call these “meteors”). As I tallied up the results, the value of the meteors was more than 80% of the gains – in other words, if the multi-year simulation yielded a return of $200K, then the tally of the meteors was in excess of minus $160K. So, I concluded (1) the TB12 was doing well IN SPITE of the meteors, and (2) there was far more value in trying to manage exposure to the meteors than in additional strategies or assets.
In the rest of this note, I will share what I have come up with and will be implementing soon. And, I fully realize that my concept of a simple tweak is someone’s else’s idea of excruciating complication. Finally, DISCLAIMER – nothing that follows is statistically rigorous enough for an academic paper. I have tested to the limits of my data, abilities, and emotional needs. In other words, for me it’s “Good enough.”
1. The Tranche metrics or Recency / EMA / Golden Cross / Abs Acceleration
Collectively, I refer to these as REGA. I have felt there is some inherent predictability in these metrics. I have arrived at a set of tests that are used to validate the Tranche recommendations. I now look at 4 assets and select the top 3 that also pass the REGA filter. Net – my REGA tests are in effect attempting to filter false positives from the Tranche lists. (I am not unwilling to share the rules, but unfortunately the math is very complicated / tedious. I will gladly share with Lowell / David if they care to consider expanding the tranche evaluations.)
Relative to the first round of backtesting, the results of the REGA filters were to (1) increase total return, (2) reduce the number of ETF-months by 20%, (3) reduce the number of meteors by 20%. So, I concluded REGA was having the intended effect of reducing meteors without reducing total return.
2. Looking at the charts
However, there were still a large number of meteors. So, I decided to look at the charts. Being a day trader, I am always looking at charts, so this is not a new or scary activity for me. For this exercise, my basic grid is a Daily and Weekly chart, plotting Heiken-Ashi candles (I also have a few other studies, but they are not important for the main point here.) Looking at this grid, I noticed a sizeable percentage of the meteors fell into a couple of scenarios:
(a) The review date occurred when the Daily or Weekly candle was red. I am not going to enter a position when the candle is red and is likely to stay red for a few more candles. NOTE: This seems to occur regardless of the Tranche ranking – rank 1 or 2 are just as likely to encounter this condition.
(b) The position has been on for more than 33 days, and the weekly chart turns over in the middle of a cycle. What were good gains, become less good due to something that can be observed.
So, I have concluded I need to look at the charts each week. I will gladly enter a position in any week – and with the intention of holding for at least 33 days. But, I will pull a position that goes red on the weekly chart due to the likelihood of that condition running for 3-4 candles.
(I understand the risk of fiddling too much when influenced by charts. I am simply suggesting that learning how to avoid meteors may be easier than additional strategies and assets.)
So, at this point I will be sticking with the TB12 strategy looking to deploy 3 assets, and reviewing each week for meteor exposure.
3. Adaptive risk-based allocation percentages.
I have been concerned about assuming all capital is available for allocation, regardless of prevailing global conditions. I feel much better about Tranche ranks 1 and 2 when SHY is ranked 6, than I do when SHY is ranked 3. (I have no data to back up my discomfort – but the discomfort is there.)
So, I was interested to recently come across this newly published paper that builds on Faber and Antonacci by adding an adaptive feature to aid in crash protection (what the authors call Protective Asset Allocation)
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2759734
The details are in the paper, but the gist is a model for determining what percent of the portfolio is eligible for risky assets. The PAA model breaks the suite of ETFs into: Risk On assets, and Risk Off assets. At each monthly review, the Risk Off % is determined, and the Risk On assets are allocated 1/N to the balance of 1 – Risk Off %.
The modeling results indicate a reduction in ETF-months, a big reduction in max expected draw down, and a modest reduction in expected returns. In my mind, the additional risk management is worth the reduced return.
The PAA authors reduce the Risk On portion by 20% for each Risk On asset that has a negative 12-month momentum measure. So, 5 negative assets suggests no Risk On allocation. I think this is WAY TOO restrictive given the rigorous modeling / testing of TB12. Given that I already have a limit of 33% for any single Risk On asset, I’m thinking of applying the PAA rules to reducing the Risk On pie by only 5% per non-qualified ETF.
I expect this will have a reduction in return, but also a reduction in drawdown.
4. Swing Trade the ultras.
This is my decision related to adding a new strategy. I have a lot of observation and a small (but growing) experience with using equity index Ultra ETFs with holding periods of a few to several days. I am allocating 10% of my retirement assets to be available to this strategy. I have watched others over a number of years make well over 100% per year with relatively infrequent exposure to the market. My intent is to use the weekly chart review to be aware of opportunities – my annual goal is around is 50% on risk capital which adds 5% to the portfolio which is enough to compensate for any drag on returns from the PAA.
=============================================
NET – My approach
90% of funds available to Tranche / Bakers Dozen (TB12)
10% of funds available to swing trading ultra ETF on SPX, NDX, DOW, RUT
TB12 funded with the top 3 assets from Ranks 1-4. These assets also pass a chart review before entry.
TB12 allocations will be allocated equally into the Risk On portion as determined by PAA, up to a max allocation of 33%.
I will review charts weekly to avoid entries that seem to be in a decline pattern (if it’s short, I’ll get in next week) and exit positions that begin to fail on the weekly Heiken-Ashi candles.
Along the way, I hope to do trades on ultra ETFs to goose the returns (much more to my temperament than options – which I did extensively for a number of years.)
=============================================
I hope this furthers the discussion.
Thank you.
Mike Saylor
Seattle
Mike,
I recently read the Protective Asset Allocation paper. What did you think of the back-testing methods?
Lowell
Lowell –
Thank you for your question.
As far the specific back-testing methods, I will say the authors paint a credible story. They seem to go to great lengths for significant in-sample and out-of-sample data. I have no basis to comment on their fabrication of data back to 1970, nor their statistical methods. I will say simply it all sounds good.
As to what they are back-testing, the crash protection aspect got my attention. As I mention in my earlier post, I think their conclusion of reducing Risk On assets by 20% for each non-qual asset is way too restrictive for the Rutherford portfolio.
I do like dividing the assets into Risk on vs Risk off, I do like reducing the Risk On allocation based on a less bullish global context, and I do like the idea of IEF (or equivalent) as a possible alternative to SHY.
My intention is to allocate funds as follows….
The RISK OFF portion will be the greater of:
=> the 10% I will always keep in cash to be available for ultra ETFs
=> 5% for every Risk On asset that has negative 12-mon momentum
Invest in the winning Risk Off asset for any portion above 10%. If there are 4 non-qual Risk On assets, the RISK OFF portion will be 20%. 10% will always be in CASH, so in this example 10% would be allocated to the RISK OFF asset.
Divide the RISK ON portion equally between a max of 3 qualifying ETFs, with a max allocation of 33% per individual ETF.
If the RISK ON assets cannot fully absorb the allocation, the difference will be considered RISK OFF.
In my mind, this captures the intent of PAA while substantially preserving Tranche / Bakers Dozen – and lets me have a single set of rules for managing sizeable retirement accounts.
/MCS
Mike, thanks for sharing this.
I actually backtested the PAA method a couple of months ago and shared it privately with HedgeHunter, Lowell, and a couple of other people. It did not back test very well, as it gave up a ton of return in exchange for the protection. It did reduce drawdown and volatility, however.
I did the Monte Carlo back test from 2006 to this spring using the Rutherford set, which I believe is the the same as or almost the same as the Baker’s Dozen. (I’d have to check.) I came to the conclusion that a better approach would likely be to use position sizing based on an MAE study something like what Howard Bandy describes in his book. HedgeHunter and I started down that path but we never completed the study, as Bandy did his study for single assets. The situation for a portfolio gets a bit more complicated.
If you want a copy of the study I can send it to you.
In the meantime, you provide much more food for thought that will take time to digest.
Ernie
Back to the question of splitting the portfolio. So, suppose one splits a single large portfolio into, say, a dividend-oriented set of ETFs, a broad basket of diversified ETFs passively managed, and either a set of sector fund ETFs or perhaps a U.S. centric portfolio.
Then, the question comes as to how to diversify management. Possibilities are dual momentum (Antonacci), Kipling, SAA, … and infinitely others. Some of these might be more appropriate to certain portfolios than others. For example, I am not sure I would feel comfortable trying to use a passive method on a dividend-oriented portfolio.
Are there any thoughts on this? Mike, you provided another management strategy. Do you have ideas on this question?
Ernie
Ernie –
Thank you for the continuing dialog.
Personally, my centering point is being able to clearly state what I am solving for with strategy selection. (I think it’s a point everyone should consider and decide.)
I’m looking for market-beating returns at reduced risk that fits my ability to achieve a high level of competency and execute over a long period of time.
In all of my reading / study / experience over many years, I’ve reached a couple of conclusions:
=> multiple strategies aimed at the same timeframe have the potential to create excessive activity and dilute the ability to be expert in any one. So, a single primary strategy is better if something worthy can be identified.
=> the Tranche Model with the Baker’s dozen (“TB12”) is the best fit I’ve seen as a worthy foundational strategy for long-term funds.
Clearly TB12 is not “perfect” by any means. But given my goal of competency and longevity, my efforts have been directed to adjusting / improving TB12 to my requirements, rather than augment with multiple alternative strategies. (My intended adjustments to TB12 are in my earlier post.) This leaves me with one primary strategy for the long-term money, with rules that should take less time each review cycle as competency is gained.
BTW, a final “tweak” I am considering is replacing (where possible) the low-cost ETFs with the more common alternatives that provide liquid options. While I do not expect to actively trade options with the same diligence as Hedgehunter, I do think having viable options as part of the arsenal is worth considering.
So, my adjusted version of TB12 is my strategy for ~90% of my long-term funds. My other 2 main strategies are for shorter timeframes.
/MCS
Thanks, Mike. I agree with your goals, but I think the idea of multiple portfolios is to further diversify both the investment horizon and also the management strategy, so that one doesn’t have all of the eggs in one portfolio or invested using one strategy. Admittedly, this leads to a lot more work. The question is, “Is it worth the extra work to reduce the risk in the portfolio(s) by using multiple portfolios and strategies, vs using a single portfolio where one hones a single strategy.” I don’t know that there is any hard data to support an answer to this. My sense is the more one can diversify things, the lower the risk, but at some point one might as well invest in a diversified buy-and-hold portfolio or a single ETF such as VTTVX. This raises interesting questions. Is it worth the effort?
Ernie —
The question you pose is key. And, I believe, the answer is influenced by one’s circumstances and how one is wired. If someone is truly going to manage these portfolios mechanically, then conceptually a mix of 2-3 (hopefully somewhat uncorrelated) strategies might provide an improved aggregate outcome.
In my case, I don’t have time to recover another “correction”, but I do have time to review charts more frequently. So, I have arrived at the conclusion TB12 does a good job of getting me in and a weekly review of Heiken Ashi candles can be a very useful final check on an entry as well as an effective mid-cycle exit strategy.
I think managing exposure to losses from mid-cycle breakdowns is a huge boost to the risk-return metrics, and it’s quite possible with the TB12 assets. And… it’s easy to simply expand the number of funded assets in order to be more diversified.