—-
Efficient market theorists have long been puzzled by momentum, and exclaim that it should not be possible to make money from buying past winners and selling past losers in well-functioning markets. Practitioners have been ignoring these efficient market theorists and collecting money for decades. There are probably more academic pieces written on momentum than almost any other subject in finance. (Besides the efficient market hypothesis of course.) I count nearly 900 when searching momentum on the SSRN.
Which is why it makes me shake my head when people talk about how technical analysis doesn’t work. It is such an uninformed opinion it is embarrassing at this point. At least three of the Alpha Hall of fame members use technical analysis. I am not defining TA as the subjective form of “charting”, but rather the simple analysis of price. Some call it tape reading (
Cohen), some call it
quant analysis (
Simons), and some are fine with the TA label (
Jones).
(Disclaimer: Before I start getting loads of hate mail on why technical analysis doesn’t work, please realize I do not bow at the alter of any discipline, but am simple guided by what works for me and most importantly
what makes money – whether it be value factors like price/book, valued-added fundamental analysis, sentiment analysis, mean reversion techniques, or systematic
arb strategies
I don’t care.)
Fama and French found momentum to be the most predictive of their four factors, and countless other speculators have used momentum as parts of their models. Just about every
CTA out there uses momentum under the label
trendfollowing.
One of the most comprehensive studies was performed by
Dimson, Marsh, and
Staunton of “
Triumph of the Optimists” fame. They found that winners (top 20% past returns) beat losers (bottom 20%) by 10.8% per year in the UK equity market from 1956-2007. Even using the top 100 UK stocks by market cap still produced a 7%
outperformance. Taking a look at these top 100 stocks since 1900, they found a 10.3% per year
outperformance.
Momentum strategies have been in existence for the majority of the 20
th Century (and probably longer).
Alfred Cowles and Herbert Jones found evidence of momentum as early as the 1930s [1937]. H.M. Gartley [1945] mentions methods of relative strength stock selection in his Financial Analyst’s Journal article “Relative Velocity Statistics: Their Application to Portfolio Analysis.” Robert Levy [1968] identified his own system in “The Relative Strength Concept of Common Stock Price Forecasting”. Other literature penned by investors who suggest using momentum in stock selection include O’Shaughnessy’s [1998] book “What Works on Wall Street”, Martin Zweig’s “Winning on Wall Street”, William O’Neil’s [1988] “How to Make Money in Stocks”, and Nicolas Darvas’s [1960] “How I Made $2,000,000 in the Stock Market.”
One of the best reasons as to why momentum (and by default, indexing) works is the distribution of stock returns. I have a few charts in my book that I think are fantastic, but you will have to wait until that comes out in January.
CROSS-MARKET MOMENTUM (or, relative strength)
I published a paper about a year ago that focused on a very simple
trendfollowing system to reduce risk. The paper could have easily been called “A
Quant Approach to Risk Management” as it is really the same thing. However, many investors are not interested in reducing risk, but rather maximizing returns. One could simply leverage the existing system, but with retail rates for margin and cash that is not ideal and will compromise results.
Below we examine a simple cross-market momentum system that stays fully invested at all times. This system compares assets to each other (is real estate going up more than bonds?) rather than my paper which compares assets to themselves (is the S&P going up or down?).
This can also be called a rotation system as you are rotating into what is performing best over a given time period. Many people have researched such systems over fifty years ago and they have continued to work decades after publication.
The system uses the same five asset classes as before – US Stocks, Foreign Stocks, US Bonds,
REITs, and Commodities.
Each month, the 3, 6, and 12 month total returns are recorded for each asset class (and then averaged for the combo). The actual time frame selected does not matter much as the 3, 6, and 12 month time frames all produce similar results. I prefer using all three (combo) because it picks the asset classes that are outperforming in numerous time frames.
The investor then simply invests in the top X asset classes for the following month. For example, at the end of 2007 the order of returns from best to worst was Commodities, Foreign Stocks, Bonds, US Stocks, and Real Estate. The portfolio for the next month (January) in 2008 would be in that same order.
Below we show the results of taking the top one, two, and three asset classes, updated monthly, based on the rolling 3,6, and 12-month total returns. (Top 1 means you just take the top asset class each month. Top 2 means you select the top two asset classes each month and put 50% of the portfolio in each, Top 3 is the top three assets with 33% in each, etc).
Click on table to enlarge.
While simply taking the top performing asset class may seem like a good idea because it experiences high returns, in reality it is not. Investing 100% of your portfolio in only one asset class leaves the investor exposed to market shocks, and consequently the turnover, volatility, and drawdowns are higher for a single asset class. A better idea would be to invest in the top 2 or 3 asset classes each month which equates to the top 40-60% of asset classes. (So, traders could run this with 10 asset classes and select the top 50%, or 5 asset classes.) A generic 10 asset class allocation is below (I offer two ETFs in each asset class in case people are utilizing tax harvesting):
US Stocks – VTI, SPY
US Stocks -VB, IWM
Foreign Stocks –VEU, EFA
Foreign Stocks –VWO, EEM
Bonds – IPE, TIP
Bonds – BND, AGG
REITs – VNQ, IYR
REITs – RWX, IFGL
Commodities – DBC, DJP
Commodities – GSG, RJI
For similar risk as buy and hold, taking the top 3 positions for the “combo” outperforms by over 4% per year, with a similar Sharpe Ratio as the timing model. We expect 0.80 to be a consistent target for a momentum approach to tactical asset allocation regardless of the exact strategy employed. This strategy outperforms the buy and hold portfolio about 70% of all years, and 10 of the past 15 years. Taxes are obviously a drag, so it makes the most sense to run this in a tax-defered account. For those who feel that commissions will be restrictively high, the system could be updated quarterly.
I believe that as humans are involved in the financial markets, the markets will continue to be driven by the emotions of greed and fear. This aspect of the market is a simple example of an alpha generator that is “timeless and universal”. Beta asset classes go up about 70% of the time. My research has shown that returns, on average across the five asset classes back to 1973, are about 40% lower and volatility is 20% higher when asset classes are below the 10 month moving average. This “volatility clustering” is one of the simple reasons the timing model works – when markets are declining people become more fearful and use a different part of their brain when markets are going up.
There are many, many variants and offshoots one can take the model. (For example, invest in the top 40% of asset classes and sell one when it drops out of the top 50% to reduce turnover.) For the most part, the take away is that for similar risk, a momentum model generates some excess annual returns. This is not the investing Holy Grail, but I consider this a method for a simple, timeless alpha that is rooted in human psychology.
