In recent years, the so-called Yale Model has been extremely popular with investors. The model is an attempt to mimic the investment strategy used by Ivy League endowment funds, which have an outstanding track record of beating the market indexes. David Swensen, the superstar manager of the Yale endowment fund, delivered returns of 10.1% annually from 2002 to 2011, a decade when stocks returned 3.9%. The Harvard endowment returned 9.4% over the same period and has grown 12.9% annually over the last 20 years.
The Ivy Portfolio, by Mebane Faber and Eric Richardson, describes how Yale and Harvard use an asset allocation model that is broadly similar the Couch Potato strategy. The key difference, however, is that the endowments include a number of asset classes that are not available to retail investors, including private equity, hedge funds, and direct ownership of timber resources and commercial real estate.
The first half of Faber and Richardson’s book is a fascinating look at how individual investors can mimic the Yale Model. The authors quote both Swensen and Jack Meyer (Harvard’s former endowment fund manager), both of whom recommend building diversified portfolios with low-cost index funds. Then they offer three possible “Ivy Portfolios” that anyone can assemble with ETFs: the simplest one allocates 20% each to US stocks, foreign stocks, government bonds, real estate and commodities. Two other chapters explain that private equity and hedge funds can be profitable for institutional investors, but are best ignored by the great unwashed.
It’s all in the timing
So far, so good—but the latter half of The Ivy Portfolio goes a giant step further. Rather than simply encouraging investors to diversify widely and rebalance, Faber explains his strategy for protecting oneself from dramatic drawdowns, or “winning by not losing.” He points out that four of the five asset classes in the Ivy Portfolio have experienced declines of about 50% or more since 1973. (Government bonds were the only exception.) “So, is there a way to avoid these long bear markets and losses?” he asks.
The suggestion Faber offers in the book is based on his paper A Quantitative Approach to Tactical Asset Allocation, first published in 2007 and updated in 2009. It’s still the most downloaded paper on the Social Sciences Research Network, which testifies to its extraordinary popularity.
Faber’s strategy is actually very simple. You just check the price of each asset class on the last day of each month, and if it is greater than its 10-month simple moving average (SMA), you buy (or continue to hold). If the price is less than the 10-month SMA, you sell and move to cash. In other words, it’s straightforward market timing.
If one had employed this timing strategy with U.S. stocks from 1900 through 2008, Faber reports, it would have delivered returns of 10.45%, compared with 9.21% for the S&P 500. More remarkable, it would have protected you from almost all the worst drawdowns: it even earned slightly positive returns in 1931 and 2008, when the market lost 44% and 37%, respectively. Faber’s analysis shows similar results when applied to international stocks, real estate, commodities and even 10-year government bonds. In all cases, the timing strategy produced higher returns with much lower volatility, while avoiding the largest drawdowns.
Backtested from 1973 through 2008, the the Ivy Portfolio with the market timing strategy trounced a buy-and-hold approach and delivered returns that would have made Swensen and Meyer smile:
| Buy and Hold | Market Timing | |
|---|---|---|
| Return | 9.77% | 11.27% |
| Volatility | 9.73% | 6.87% |
| Maximum Drawdown | –35.98% | –9.53% |
| Best year | 26.58% | 26.20% |
| Worst year | –30.09% | –0.59% |
These are very impressive results, especially when you consider that they came about with an average of just three to four round-trip trades per year.
The Ivy Portfolio is an excellent book that I would highly recommend to index investors (though I would discourage them from building a portfolio that is 40% real estate and commodities). But how about Faber’s timing strategy? Is it a recipe for lowering the volatility and improving the long-term returns on a Couch Potato portfolio? I’ll consider that question in my next post.
This smells a little of data mining. Is there some explanation of why the 10-month moving average is best? Do the results change much if we use 9- or 11-month moving averages?
@MJ: Faber does make reference to Jeremy Siegel’s data in Stocks for the Long Run, which considered a timing strategy based on the Dow Jones Industrial Average and a 200-day moving average. That analysis showed the timing strategy improved risk-adjusted returns, although it also involved about twice as many trades.
To be fair to Faber, he did do out-of-sample testing and the timing strategy works for all five asset classes. As I’ll explain in my next post, my problem with this strategy is not that the data are flawed, but that it is unlikely to work in the real world.
Standard and Poors did a study of market timing using moving averages of the SP500 from 1970 to July 2011.
The 200 exponential moving average, which is comparable to the 10 month in Fabers study, returned 7.3% (CAGR) with an average of 7 signals per year – this means a lot of trading and taxes which would probably reduce this return considerably. It beat the SP500 33% of the time on an annual basis. The market returned 6.6% CAGR
However the 50 X 200 simple moving average cross returned 7.4% with only 1 trade per year on average so there would be less drag on performance. It beat the SP500 34% of the time.
In the 2000’s the 50X200SMA returned 5.7% vs minus 0.6% for the index and beat the index 70% of the time annualized.
Using this signal with the TSX60 one would have exited the market in early Sept 2008 and missed about a 40% drawdown. There were 4 signals since then.
The most recent signal was only last Friday May 11 2012 with the 50 crossing below the 200 indicating time to reduce risk and raise cash or hedge some or all of the beta of the portfolio.
@Andrew: Thanks for joining the discussion: I was waiting to hear from you, as I know you have looked at Faber’s work extensively. I’m curious, though: do you use a timing strategy yourself? Do you know anyone who does?
Michael/CCP,
He addresses it in his paper. The results are insensitive to the length of the moving average over the range he tested. He used a variety of durations between 3 months and 12 months, IIRC. He also does extensive out of sample testing, specifically calling out the risk/concern of data mining and addressing it.
The paper is really worth reading. It’s pretty quick–you should be able to get through it in under 1 hour. The book provides a lot more explanation and of basic concepts.
I’m quite interested in hearing why you believe this wouldn’t work in the real world. It’s a strategy I have considered implementing for a while now and I have done a plethora of testing on it and all of it suggests a good return.
@Mike: Thanks for the comment. I will have more to say about this in my next post!
The success of the IVY might be including other uncorrelated asset classes such as commodities and real estate besides just the traditional equity/bond ones which are missed in a traditional 60/40 portfolio. But it’s interesting that Faber’s timing model works for many portfolios!
http://seekingalpha.com/article/27167-lazy-etf-portfolios-inspired-by-the-gurus
Hi Dan,
Yes I do use timing in a limited way. My registered accounts are CP, mostly bond ETFs and I rarely touch them. But in my open account I have used longer term moving averages three times in the last 5 years to add or subtract to equity within an asset allocation range – two periods of subtractions and one period of additions. I look at the 40 week and 10 week SMA crosses and also stochastics on a weekly chart and I look at the 10 month alone now. I have a “neutral” allocation to equity but with a low to high range. When there is a negative signal I go to the lower end of my range and with a positive signal I go to the higher end.
For instance early last summer I reduced my equity by about a quarter (all ETFs) because of the 10 month signal.
There was another signal last week (registered Monday as first day of week) a 10×40 SMA weekly cross – one of only 5 cross signals in the past 3 years (and one of only 12 signals in the past 10 years). Is the market telling us something?
I think using the very simple rule of the 10 month could be useful as a signal to reduce equity % and/or to hedge.
I wonder if anyone would like to comment on this – using the timing signal to allocate within a range. Should I be going to cash if the asset class is below the 10 month as Faber describes? Or skip this tactic completely and just rebalance?
Andrew, How do you get the data you mention above? I looked on TMX and you can’t select anything above 255 days for SMA in the “Upper Indicators” on the “Charts” tab. You mentioned that you use a 10 month (~304 days) and a 40 week (280 days); just curious where you get your data?
Thanks!
Different Andrew, but the answer to that question is that moving averages in days refers to trading days. 255 days is roughly a calendar year (5 x 52 – bank holidays). Months and weeks refer to calendar periods, with about 21 and 5 days in each period respectively.
10 month is roughly equivalent to 200 day.
@Andrew: It’s interesting that you have modified Faber’s strategy. This is one of my concerns about timing strategies: there is always the temptation to stray from the quantitative elements and introduce human judgment (and perhaps emotion). I worry that it would be very difficult to execute a strategy like this with the rigorous discipline that it would require.
Dan
thanks for this post and digging a bit deeper on the Faber approach to better protect CP assets. This question came up in discussion a while ago, I look forward to your next post.
I have implemented a “shadow” Faber approach on my CP portfolio – not actually implementing yet in practice, just watching for the signals as Andrew indicated. I have only been doing this for about 6 months, and as Andrew indicated, the first and only signal over that time just recently occurred to sell XIC to cash and then wait it out for the signal to rebuy it. At worst it would be earning 1% sitting as cash instead of losing more. The rest of the asset classes are fine and would remain untouched. The frequency of required action is very low, as pointed out above.
It seems relatively emotionless, taking a couple of minutes end of each month to check the 50 and 200 day MA, to date its been pretty boring. I don’t yet plan to do it, as I’m still not quite convinced and confident enough as relatively rookie investor, but the compelling attraction to me is enhanced protection based on a non emotional decision that still seems compatible with CP approach.
Que – the data are free on Yahoo.ca finance. Set up a 200 day MA watch list for your CP assets.
Looking forward to more on this discussion.
@Len: Thanks for joining the discussion. One thing I would argue with is the idea that, “At worst it would be earning 1% sitting as cash instead of losing more.” That’s actually the best that can happen. The worst is that you’re sitting in cash when the market moves sharply upward unexpectedly.
I wish I could share this article (as with your others) with my friends on Facebook, etc. You have a consistent, solid, educational message that us niche online investors need to share with others. Keep doing what you’re doing.
Is it decidedly less cost effective to buy puts whenever the 50 X 200, for example, suggests one should sell (I know someone already mentioned hedging)? It seems to me like the natural choice, given that the indicator doesn’t beat the index a majority of the time (limiting your upside potential) and it’s mostly meant to protect against big drawdowns anyway.
I prefer using the 5 month moving average in conjunction with the 12 month moving average. Jim Otar at http://www.retirementoptimizer.com has written a nice piece on this under “Hurricane Warning”. Technical Analysis sounds easy and it isn’t always accurate, and few have the discipline to consistently act on the signals, second guessing themselves along the way. IMO for most investors a disciplined strategy reflecting their goals and appetite for risk using low cost funds will likely yield better results with less confusion.
It’s also worth noting that Faber runs an ETF based on the philosophy outlined in his paper/book, ticker GTAA. He does not follow this strategy exactly, this is more of a simplified proof of concept. He used more granular asset classes, and claims to use a variety of moving averages (diversifying over various MAs). I have to say that the performance of the fund has been disappointing since inception, but it’s worth pointing out that this strategy works best in trending markets, and the period since the fund launched has been volatile and sideways.
Owen, I haven’t looked into it. It really depends on how you use puts. Often this strategy will have you exit before the market becomes panicked and option prices spike, so it might be viable.
Len, Andrew, or other, I am would like to see what you are seeing, which settings did you use to find:
Len, “I have only been doing this for about 6 months, and as Andrew indicated, the first and only signal over that time just recently occurred to sell XIC to cash and then wait it out for the signal to rebuy it.”
and,
Andrew, “There was another signal last week (registered Monday as first day of week) a 10×40 SMA weekly cross – one of only 5 cross signals in the past 3 years (and one of only 12 signals in the past 10 years).”
Go to:
http://www.google.ca/finance?q=xic
Below the chart is a link that says ‘technicals’. Select Simple Moving Average from the first drop down and punch in 10, then Simple Moving average from the second one and punch in 40. Then go back to the chart and expand the range to 5 or 10 years. When the two coloured lines cross each other, a signal is generated.
And this is fascinating. Rebalancing that greatly boosts returns and reduces volatility. Somewhat similar in that they are momentum strategies, holding what is working, selling what is not, and rebalancing often/monthly. These would be ‘The Trend is for Friend’ portfolio model(s).
http://seekingalpha.com/article/587491-adaptive-asset-allocation-a-true-revolution-in-portfolio-management
Here is a link that I use for my technical analysis http://stockcharts.com/h-sc/ui. It is free and you can use a variety of moving averages and oscillators. The thing with technical analysis is it takes emotion out of the equation. It will give you an indication, not always accurate as to what the trend is. In an up market you will always be a little late to the party and in a down market you will experience some of the decline. I like to think of it as a risk management practice with the potential for a slight enhancement of returns. I like looking at the longer term trends as opposed to the short term trends. Following the longer term trends you will trade far less frequently resulting in lower costs. It is always easy to look back at the chart and identify the trends, however it is sometimes difficult to read the signals in the moment. Hind sight is always very illuminating.
It is easy to implement this strategy but perhaps it means the behavioural bias is easier to overcome because it protects against large drawdowns – which can scare one into selling at bottoms. But you have to act on the data which is hard as well. For example at this very moment the indicators are all signalling negative and to reduce the amount of equity (Canada market):
The market is below its 10 month moving average and the 20 week moving average.
The 10 month moving average is downward sloping (but maybe flattening and may be bottoming). The 20 week is rolling over.
The 10 week is downward sloped.
The 10 week moving average just went below the 40 week days ago.
This implies the need to sell equity ETFs outright or to reduce their amount or to hedge beta. It may turn out to be a headfake but this is the cost of the insurance against losing capital.
My problem is that I studied economics. I see TED and EuroLIBOR spreads, and the rate of change of the correlation between US 10 year breakevens and the S&P500, and the latest release of the statistical series “NIPA corporate profits before inventory valuation adjustments” dancing in my head. And its hard but I try to ignore them from the point of view of acting on them rashly. I don’t ignore them however because they are telling me a story which like any story is more interesting when it gets to extremes.
So the only time I would not ignore data other than the price data is as the data move to extremes – and extremes are not that hard to identify. When a number of these indicators are flashing extreme redzone at the same time something is beginning to happen (or has happened). Also when these data are doing extreme things the simple indicators like the moving average trend lines are also showing something. Usually but not always. For example the 50x200SMA cross only beats the SP500 36% of the time. But it still beats it. The 65×200 exponential moving average is even better it beats the index 48% of the time. Both add marginally to performance over long periods of time. But both also had a worst drawdown in 40 years of only 15.2% and 10.7% respectively versus -38.5% for the SP500 on an annual basis. I can handle losing 10-15% over 38%.
So from this perspective maybe using a system like this is may be easier from a behavioural bias perspective because you will not see your hard earned and hard saved equity cut almost in half and at that very moment say “I can’t take it any more” and sell instead of calmly rebalancing which would take a lot of guts. As peoples portfolios get larger over time this becomes more of a consideration as relative performance has less psychological weight than absolute (people generally freak out less knowing they lost 7% of 100K which is 7K, than 7% of 1000K which is $70,000 – its still just 7% ) Also this method forces one to buy nearer to the bottom than the top when adding new money.
BTW some current issues: negatives: corporate profits have rolled over, multiples are starting to compress as bond yields drop to extreme lows, 10 year breakevens have turned down mid march (the longer term inflation indexed bond behaviour indicates what institutional investors are doing)insider selling is increasing dramatically, credit spreads are widening, Euro CDS at record highs, slowing economy in China(sales, money supply, loan levels, exports) volatility is rising, US dollar rising in recent weeks (flight to quality like the bond yield drop), the ratio of US long bonds to the canadian dollar index is rising, the ratio of junk bonds to the canadian dollar is rising, Bank runs in Greece, Spanish credits spreads widening, etc….)
positives: better employment numbers (USA, Canada, Australia), inflation expectations fall, Chinese inflation slowing, energy prices falling, confidence indicators rising.
Andrew, you said:
“The 10 week moving average just went below the 40 week days ago. This implies the need to sell equity ETFs outright”
Wouldn’t it make more sense to sell ETFs when they are above the moving averages? Buy low and sell high?
As pointed out by Andrew, all those buy and sell transactions would trigger more capital gain taxes then a simple couch potato portfolio. I would suspect this strategy to be be much less effective in a taxable account versus in a registered account.
Hi Andrew,
I don’t understand this statement:
[quote]For example the 50x200SMA cross only beats the SP500 36% of the time. But it still beats it. The 65×200 exponential moving average is even better it beats the index 48% of the time. Both add marginally to performance over long periods of time. [/quote]
If a strategy “beats the SP500 36% of the time”, doesn’t that imply that it underperforms 73% of the time? How can such a strategy “add marginally to performance over long periods” under these circumstances?
I suppose this is possible… but more likely I’m misinterpreting your meaning.
Thanks,
Stephen
Sorry about the math in my post… 100 – 36 = 64 (not 73) :/
Que
I use TD Waterhouse charts. You can also use Stochcharts.com which is free. Just play around with the settings to make the chart you need. The only problem is the free data only goes back 3 years.
Stephen D
Sorry I was not clear. The “annual frequency of beating the S&P500” in the study showed that the 65x200EMA was 48% with an average of 0.8 signals (crosses) per year and the 50x200SMA was 36% with an average of 1.0 crosses per year. The best year for the 65x200EMA was 32.1% vs 34% for the index and the worst year was minus 10.7% vs minus 38.5% for the index. Also overall this cross outperformed buying and holding the index over the period of time of the study by 0.6%. This doesn’t sound like much but it is the CAGR or compound annual growth rate. The compounding adds up : if you invested $1000 at the beginning of the study period the index returned just $13, 812 vs a return of $17, 293 for the 65x200EMA cross. This is $3481 or 25% more than the index.
Adam
You are right in the sense that if the ETF is above its upward sloping trend line it may be making a peak. But we do not know this so you wait until the price has dropped below the trend as a kind of confirmation that a peak has been made. A peak is just a change in direction. You miss the peak if you use the cross as a signal because it comes later so obviously you miss some of the gains. Using the crosses of the shorter MA vs the longer MA is a kind of confirmation of the trend change.
I like to use the weekly charts and monthly because there is less “noise” in them.
Hi Michael,
A beginner’s question – what does it mean when it says “the price of each asset class?”
what does “asset class” mean?
Thanks,
William
@William:
Fear not. I started as an absolute novice in early June to the idea that passive index investing was not inferior to active management (by fund managers, or anyone else) and from a pragmatic viewpoint was actually superior. I have learned a lot from following this website closely and researching or re-questioning the unanswered questions (or the ones not answered to my comfort-level) that came out of it. This is generally a civil and educated and knowledgeable crowd, interested in sharing and comparing their knowledge while exploring the Couch Potato concept, not necessarily all with exactly the same mind-set, but their biases and reliability will be self-revealed over time, and you adjust your internal knowledge base accordingly.
As this is a relatively easy question, I feel safe in answering it without screwing it up. As always, this website cannot take responsibility for the opinion of any independent poster such as myself. That’s how it works.
An ASSET, generally, is something of value that you own, after you acquire it. In this context it means all the tangible things in your investment portfolio, like the free money sitting around not gaining interest, the money you have put into interest bearing accounts, the bonds you have bought and the shares of companies, whether these latter two groups are bought as individual shares, or packaged into groups of mutual funds or index funds or ETF’s (Exchange Traded Funds). A CLASS is a grouping or category of any larger aggregate. In this context, ASSET CLASSES are a convenient way of dividing your portfolio into different well defined components that behave in distinct and characteristic ways (going up and down in value, hopefully in different directions at any given time, but generally up over the long haul.)
You will note that in my description of what ASSETS in a portfolio were, I went through a list — the things in my list were actually grouped basic ASSET CLASSES, that is, CASH (free money sitting around not generating interest), INCOME (Money generating interest, plus BONDS, including BOND FUNDS and ETF’s), and EQUITY (shares of stocks of individual companies, or index, mutual funds or ETF’s containing such shares). Some Classification Systems include money gathering low interest that can be withdrawn at any time without penalty as CASH rather than INCOME — I’m not entirely clear on that, but you get the picture.
Now these are the broadest categories of ASSET CLASSES. You can break them down into finer divisions of “SUB-CLASSES” such as EQUITY being divided into Equities (ETF’s and Stocks etc) from different geographic regions (EQUITY-CANADA, EQUITY USA, EQUITY REST OF WORLD etc.), or sliced in a different way, by the size of the underlying companies the individual shares or funds represent (EQUITY-LARGE CAPITALIZATION, EQUITY-MID-CAPITALIZATION, EQUITY-SMALL CAPITALIZATION). The discussion in the posts above has referred to these thinner slices of the ASSET pie as “more granular asset classes.”
The above familiarity was gleaned from merely following the website discussion. As time goes on, further esoteric ASSET CLASSES will arise in the discussion, I’m sure, limited only by the imagination of financial economists, and the pathetic desire on our part to try believe this is incrementally meaningful, and to try and squeeze out an extra buck from our portfolios (just kidding, but with some grounding cynicism).
Can anyone explain to me how and where (website) to setup an alert when an ETF price crosses a 10 month SMA for example?
Thanks, Que
Len or anyone else: How do you set up a 200 day MA watch list on Yahoo.ca finance?
Couch potato always manages to burst my bubble when I finds something new.