The last few posts (here and here) have injected a dose of reality into the debate around future return expectations for 'Buy and Hold' investors. Markets have been expensive for over 80% of periods since 1994, and never reached the genuinely low valuations associated with secular bear markets, even at the depths of 2003 and 2008-9. By most traditional measures, stock markets around the world currently range in valuation from above average (UK) all the way up to nosebleed (China). Returns to a buy and hold strategy consisting of global stocks from these lofty valuations are unlikely to be robust, and may actually be negative over the next 10 years or more.
We believe that investors need not be held captive to the common buy and hold doctrine preached by mutual fund companies and banks. These institutions are motivated to keep investors locked into homogeneous mutual funds and managed account products. This benefits banks and fund companies because they are able to charge more for stock mutual funds (or managed accounts) than bond funds, despite abundant proof that most managers add virtually no value. Not surprisingly, the Buy and Hold approach requires the least amount of effort, training, or skill, and offers investors virtually no accountability. It does, however, advocate constant exposure to stock mutual funds under any and every market condition.
We assert that cap-weighted buy and hold is just one of a variety of broad investment strategies, albeit the one most commonly embraced by adherents to the dominant investment paradigm. Unfortunately for these adherents, the dominant investment theory does a very poor job of describing actual market behaviour, especially over time horizons that are meaningful for most investors. So what is an investor to do?
The Trend-Following Alternative
There are many alternatives to cap-weighted buy and hold, but the strategies that we feel hold the most promise for investors can be broadly described as 'trend-following'. Although trend-following is an investment strategy with strong empirical roots, it is helpful to think of it as a natural extension of actual human behavior. For many years, scientists in disciplines as seemingly unrelated as evolutionary anthropology and modern psychology have demonstrated that humans are prone to the same herding instincts as other animals. When faced with a choice in the absence of trusted information, humans will usually choose to follow the crowd rather than act against it. Many can relate to the experience of choosing a restaurant in a foreign city. Even when presented with several positive reviews of a restaurant from trusted official sources, people will often choose not to eat at that restaurant if it is empty, especially if a restaurant next door is full. They will usually follow the crowd into the busy restaurant, even at the expense of waiting for a table, rather than eat at the well reviewed but empty one.
A couple of years ago, psychologists at Columbia University performed an experiment to test the power of social herding. They set up an online music exchange, dubbed MusicLab, where over 14,000 participants registered to listen to, rate, and download songs by a variety of bands they had never heard of. Some of the participants saw only the names of the songs and the bands, while others, in what the experimenters called the 'social influence' group, were also shown which songs were most highly rated by others, and/or most frequently downloaded. The 'social influence' group was further divided into 8 separate 'worlds'. Participants in each world could see only the ratings and downloads of others in their world.
One of the researchers, Duncan Watts, explained the setup and their observations in a 2007 New York Times article:
"This setup let us test the possibility of prediction in two very direct ways. First, if people know what they like regardless of what they think other people like, the most successful songs should draw about the same amount of the total market share in both the independent and social-influence conditions — that is, hits shouldn’t be any bigger just because the people downloading them know what other people downloaded. And second, the very same songs — the “best” ones — should become hits in all social-influence worlds.
What we found, however, was exactly the opposite. In all the social-influence worlds, the most popular songs were much more popular (and the least popular songs were less popular) than in the independent condition. At the same time, however, the particular songs that became hits were different in different worlds... Introducing social influence into human decision making, in other words, didn’t just make the hits bigger; it also made them more unpredictable."
So how does this relate to investment trend-following? Trend following recognizes that one of the most powerful forces in human decision making is 'social influence', and actively takes advantage of this human condition. We see trends in fashion, eating, lifestyles, religion, education, child-rearing, baby names, car colours, medical surgeries, traffic, and almost everywhere else you might look in human society. These trends show up in the markets too, first as a recognition of real profits flowing to a company (Nortel), sector (Internet stocks), asset class (commodities), or geographic region (China, India, Brazil), and then as an extrapolation of this trend to infinity.
"It is not the strongest of a species that survives, nor the most intelligent, but the one most adaptable to change." - Charles Darwin
Trend-following systems are designed to find the most popular trends in the market place, and then ride those trends until they end, which they invariably do. These systems are not predictive; they will not tell you what toy, baby name, stock, or asset class will be most popular in a year or 3 years. Instead, they identify where the trends are currently, and adapt to changes. Some react to very short-term trends, like 2 or 3 month rotations in and out of stock market sectors, while others follow longer-term trends that occur over many months and years. Trends occur everywhere, from cotton to corn, stocks to silver, bonds to pork-bellies, and trend-followers trade them all.
Trend-following systems are designed to find the most popular trends in the market place, and then ride those trends until they end, which they invariably do. These systems are not predictive; they will not tell you what toy, baby name, stock, or asset class will be most popular in a year or 3 years. Instead, they identify where the trends are currently, and adapt to changes. Some react to very short-term trends, like 2 or 3 month rotations in and out of stock market sectors, while others follow longer-term trends that occur over many months and years. Trends occur everywhere, from cotton to corn, stocks to silver, bonds to pork-bellies, and trend-followers trade them all.
Although trend-following systems differ from other investment strategies in a variety of ways, perhaps the most important difference is that all systems have an exit condition. In other words, all systems have a set of conditions that indicate that a trade is not working, and a strategy for exiting the trade in order to minimize losses.
A Scientific Approach
Trend-following systems are predicated on the human condition, but rooted in empirical data. Systems are developed by applying the scientific method to market pricing data, in much the same way as pharmaceuticals are (or should be) developed and tested before being brought to market. A drug development team applies a deep knowledge of human biochemistry and physiology to identify a chemical that may influence one or more important disease pathways. It is then hypothesized that the chemical will influence the presentation of a disease in a certain way, and this is tested empirically. A successful test is one that meets the expectations of the researchers at a certain level of statistical confidence, and with manageable side effects. Once a chemical is shown to work against a disease, researchers work to optimize dosage levels, delivery methods, etc. before attempting to bring the drug to market.
Credible practitioners of trend-following apply a similar approach to create and test their investment systems. Based on an understanding of human behaviour, and a mountainous body of empirical knowledge about the mechanics of markets, systems developers formulate a hypothesis and run tests against real historical market data to test this hypothesis. To avoid a minefield of potential biases, professional system developers test very simple systems and, if they deliver statistically and economically significant results, proceed to optimize their systems by testing the sensitivity of the system to small changes in the parameters. A robust system shows fairly consistent results over a wide band of parametric dispersion, but certain parameters will be more likely to deliver optimal results. Once the system is tested and optimized, it is ready to be put to work with real capital.
Developing A Proprietary System
Our team has been working on a trend-following system that will work well for unsophisticated investors with a meaningful time horizon of 5 to 50 or more years. Besides statistical and economic significance in testing, we limited potential strategies to those that meet the following criteria:
1. Invests only in public stocks or ETFs traded on North American exchanges.
2. Is simple enough to explain to non-professional investors.
3. Acknowledges the behavioral biases of regular investors and makes it easy to stick to.
4. Possesses risk and return characteristics that validate the assumptions of actuarial-style financial planning tools.
Our efforts have led us to several promising systems, but we have focused on one strategy in particular which meets all of the above criteria, and passes tests of statistical and economic significance. Our system applies a combination of simple trend and moving average signals to a basket of global stock, bond, commodity, and real-estate ETFs. The moving average signals tell us whether each asset class is in a strong positive trend, or a weak or negative trend, while the momentum signals tell us where the trends are strongest. When markets show a weak or negative trend, our system sells out and goes to cash until a strong positive trend re-emerges.
Our systems show strong, consistent results across many time periods, and are resistant to time-period biases. We began by investigating results from several promising studies by Asness et al. (1997), Montier (2006), and Faber (2006, 2009) that show persistent outperformance by using momentum and moving average signals independently. The momentum strategies demonstrated extremely strong long-term performance, but are vulnerable to periods of large losses. The moving average strategies presented extremely consistent results with very infrequent, small losses, but returns were unspectacular. We decided to test a combination of the two systems to see if we could achieve strong, consistent returns with lower risk.
Montier (2007) builds on Asness' research to demonstrate that country stock market indices exhibit a very strong momentum effect. Country stock markets that have risen the most over the prior 12 months continue to do so in the following month, while countries that have risen the least in the prior 12 months continue to be weak. The following chart from Montier shows the power of a very simple strategy that invests in 16 of the top performing global stock markets over the past 12 months while going short the 16 worst performing stock markets, with holdings rebalanced monthly. This simple strategy delivered average annual returns of 16% going back to 1976.
Chart 1. Cumulative performance of global country 12-month momentum long-short portfolio.
Source: James Montier (2007)
Results are pro-forma and for illustrative purposes only
A Compelling Solution
Faber (2006, 2009) demonstrated the power of a simple moving average strategy to identify changes in trends among 5 different assets classes: US stocks, international stocks, REITS, commodities and bonds. The asset class allocations advocated by this researcher's strategy echo the allocations used by many of the top university endowments, such as those at Harvard and Yale. These endowments have a very small allocation to bonds, and a large allocation to real estate, global stocks, and alternative assets like commodities and timber. Due to their large size, however, these endowments are necessarily 'buy and hold' investors, as significant short-term changes to their allocations would cause noticeable market dislocations. Smaller investors can utilize Faber's moving average system, also called a Multi-Asset Tactical Asset Allocation system, to move to cash when market trends turn negative, mostly avoiding large losses.
The charts and tables below show how this strategy delivers very consistent returns with minimal losses. Investors would have received better returns than with stocks alone (11.3% vs 10.6%). More importantly however, investors would have experienced positive returns over 98% of all 12-month periods going back to 1973, and never lost more than 3.8% from their peak value at months' end. In contrast, stocks were positive only 76% of the time, and investors had to endure losses of 40% or more on their portfolio.
The charts and tables below show how this strategy delivers very consistent returns with minimal losses. Investors would have received better returns than with stocks alone (11.3% vs 10.6%). More importantly however, investors would have experienced positive returns over 98% of all 12-month periods going back to 1973, and never lost more than 3.8% from their peak value at months' end. In contrast, stocks were positive only 76% of the time, and investors had to endure losses of 40% or more on their portfolio.
Chart 2. Mutli-Asset Moving Average Strategy Demonstrates More Consistent Returns
Source: Faber (2009), Butler|Philbrick & Associates (2010)
Results are pro-forma and for illustrative purposes only
Chart 3. Cumulative Returns from S&P500 Buy and Hold versus Multi-Asset Moving Average Strategy
Source: Faber (2009), Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Given the strong performance of Asness' and Montier's simple momentum-based country allocation system, and the independent strong performance of Faber's simple multi-asset moving average system, we decided to investigate a strategy that combines these two systems. First, we tested a system that combines a momentum strategy with a moving average strategy using U.S. stocks. We tried this first because we had easy access to the necessary data and systems to run the test ourselves. We used a quantitative momentum model to select stocks, and combined it with a simple moving average model as a signal to move to cash. The quantitative model does quite well on its own, but the moving average overlay adds very significant value. Aside from delivering better absolute returns, the moving average overlay dramatically reduces periods of large losses (see blue line versus red and green lines in Chart 4 below).
Chart 4. Cumulative returns of S&P500 'Buy and Hold' versus momentum strategy versus momentum with moving average overlay
Source: CPMS (2009), Shiller (2009), Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Once we demonstrated that momentum works well in concert with a moving average signal in a strategy using individual stocks, we decided to test a similar strategy using allocations to global stock markets. This is a very simple and investable strategy because of the availability of ETFs that track the performance of over 35 global stock market indices. To implement this test we partnered with one of the most experienced systematic investing teams in Canada, Jason Russell and Nicholas Markos of Acorn Investments. Using MSCI country data going back to 1970, we tested a simple momentum model that invests in top-performing countries, and overlayed a simple moving average system to signal a change in the trend of global stocks. The following table and chart show the results of our non-optimized test.
Table 1. Performance characteristics of Global Tactical Allocation System (All Ex-Dividends)
Top row: Country Index Buy and Hold
Middle row: Country Index with Moving Average Overlay
Bottom row: Country Allocation System with Moving Average Overlay
Chart 5. Comparison of Cumulative Performance
Source: Butler|Philbrick & Associates, Acorn Investments
Results are pro-forma and for illustrative purposes only
Note that the Country Allocation System above, which combines the momentum model with the moving average signal, compounds at over 16% annually versus 6.5% for Buy and Hold. Further, the largest cumulative loss over the period is under 26%, less than half the maximum loss from a buy and hold strategy (56%). The longest period that investors are underwater on their investments is 40 months with the Allocation strategy, about half of the 78 month underwater period for Buy and Hold. Volatility is also less, at 14% annualized versus 15% for buy and hold. Any way you slice it, the Country Allocation System is superior to Buy and Hold.
The Country Allocation System in its current form was only applied to global stock markets. Our next step is to integrate our momentum/moving average system, which has demonstrated such excellent risk/return characteristics for stocks (Chart 4.) and allocations to countries (Chart 5.), into Faber's multi-asset model. We hypothesize that a momentum overlay will significantly enhance risk-adjusted returns versus the existing moving average system for each of the other 3 asset classes.
Maintain Your Lifestyle In Retirement
Buy and Hold is likely to underwhelm over the next decade or so, but investors have alternatives that may dramatically enhance their financial opportunities. Although a low-return environment for stocks is likely to also pull-down expected returns for the strategies we have described above, the historical risk-adjusted performance premium is likely to persist. For example, the current non-optimized Country Allocation System delivered a return premium over stocks of almost 10% per year from 1970 through 2009. Even assuming a 2% management fee, and some performance decay, it is reasonable to assume that this strategy might deliver 5% per year over a buy and hold strategy going forward.
The following charts demonstrate the massive impact that this return differential can have on a retirement plan. If we use a 4.6% expected real return for stocks as derived in this post, and assume that our strategy can deliver an extra 5% per year with slightly lower risk, then Chart 6. describes the potential difference in retirement income. The chart shows the income that can safely be withdrawn from a $2.5 million retirement portfolio under different return and risk assumptions.
Charts 7, 8, 9 and 10 use the 'Risk of Retirement Ruin' model described by Moshe Milevsky in his paper "A Sustainable Spending Rate without Simulation", and illustrate how the safe withdrawal rates in Chart 6. were modeled. The model uses lifespan assumptions for a healthy non-smoking male retiring on his 60th birthday, which is currently another 20 years.
Chart 6. Modeled safe annual income from a $2.5 million retirement portfolio for various strategies.
Source: Milevsky (2005), Acorn Investments, Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Chart 7. Milevsky Modeled Risk of Ruin for Global Stocks
Source: Milevsky (2005), Acorn Investments, Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Chart 8. Milevsky Modeled Risk of Ruin for Country Allocation Strategy
Source: Milevsky (2005), Acorn Investments, Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Chart 9. Milevsky Modeled Risk of Ruin for 50% Bonds / 50% Stocks
Source: Milevsky (2005), Acorn Investments, Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Chart 10. Milevsky Modeled Risk of Ruin for 50% Bonds / 50% Country Allocation
Source: Milevsky (2005), Acorn Investments, Butler|Philbrick & Associates
Results are pro-forma and for illustrative purposes only
Conclusion
In conclusion, the money management industry, which is dominated by large banks and mutual fund companies, have a vested interest in preaching a buy and hold doctrine. They earn greater profits from keeping investors in equity mutual funds which pay greater fees to these firms rather than suggesting that investors attempt to earn superior returns by using active timing strategies.
We have presented a few alternatives to the buy and hold paradigm, and offered compelling reasons to specifically consider trend-following systems. Trend following strategies operate on the principal that humans are influenced by powerful social factors that manifest in trends in life and markets. The dominant investment paradigm embraced by almost all contemporary investors emphatically denies that this human condition exists, preferring to think of human decision makers as computational engines that operate independently and have a perfect understanding of the odds. By acknowledging the human condition, trend-followers have an opportunity to deliver out-sized performance over time. Further, this superior performance can have a dramatic impact on the lifestyle expectations of retirees.
