Updating Dual Momentum

Dual Momentum is a very robust and simple trading system that can be followed by any individual managing their own money. It's a great strategy for RRSPs, TFSAs, RESPs, and non-registered investment accounts.

Gary Antonacci publicized the Dual Momentum with some white papers on asset class momentum shortly after the Financial Crisis.

The papers were a big hit at a time when investors were still fresh experiencing massive portfolio drawdowns, so Gary wrote a book on the strategy which you can buy on Amazon. I recommend you read it if you are interested in this strategy.

About Dual Momentum

If you are looking for a simple strategy that requires very minimal effort to maintain and reduces drawdowns over market cycles, Dual Momentum may be the answer. It takes its name from the two evaluations you performance once every month: a relative momentum evaluation and an absolute momentum evaluation.

In Gary's original form, he does a 12-month lookback on two equity assets. The first equity asset class is the U.S. market via the S&P 500 Index. The second is the MSCI All Country World ex-USA Index which includes most of the world's investable equity markets outside of the United States.

The equity asset with the highest total return over the past 12-months is put up against the risk-free rate of return over that same time period, represented by Treasury bills.

If the equity asset return beats the risk-free (T-bill) return, you invest in the winning equity asset via a low-cost option like ETFs or index mutual funds. If the T-bill wins, you invest in a broad bond fund.

Since 1950, Gary estimates the Dual Momentum model would have returned 15.8 percent annually before fees and taxes. Drawdowns, measured monthly, would not have exceeded 17.8 percent. It's a truly amazing result that handily beats a buy-and-hold approach.

My Time Averaged Dual Momentum

Approximately one year ago, I created my TADM model based on Gary's research. It is designed for Canadian investors who want to trade with low-cost ETFs available on the local stock exchange.

In my model, I have taken the average of the 6-month return and 12-month return, and used that to evaluate the relative momentum of the two equity assets instead of the traditional 12-month return.

This averaging technique creates a more responsive system. Trades are often entered a month or more sooner than the standard Dual Momentum model. However, the cost is a few more whipsaw trades.

The assets I originally used are the MSCI USA Index, the MSCI EAFE Index, and short-term bonds.

Although the MSCI USA Index is not available in ETF, the S&P U.S. Total Market Index is, and it covers much of the same exposure. We have several options for the MSCI EAFE Index and the comparable FTSE Developed ex-North America Index.

My TADM model has shown great results in backtests as well. Although I don't have access to the monthly data on a broad range of measures like Gary does, since 1970 my TADM model would have returned about 16 percent per year.

The nice part about Dual Momentum is that it is extremely robust. No matter what timing period or which precise index is used, the results are always very similar. Annual returns in the 15 to 17 percent range, low maximum drawdowns under 25 percent, and relatively few trades to achieve these outstanding results.

Playing Dual Momentum Games

After doing a lot of testing of different Dual Momentum inspired models in the past few weeks, I've decided I'm going to change my shared signal and effectively kill the MSCI EAFE-based TADM model.

There's a few reasons for this, some of which will be explained in more detail in some upcoming posts. But it essentially boils down to Occam's razor. When two competing strategies are compared side-by-side with similar results, you should choose the strategy that is the simplest and most efficient.

TADM is the product of unnecessary complexity without the meaningfully different results. In a lot of other areas in my trading and when I communicate with readers, I frown on data mining to search for the "perfect" investment approach.

To be clear, data analysis is very useful when you are trying to test the robustness of a particular investment strategy. You want to invest in a strategy that shows similar results across similar parameters.

If you have a system that's based on a 12-month lookback period which returns 10 percent per year, but if you change to an 11-month lookback your results are wildly different—say 25 percent per year—you should be very suspicious of the likelihood of the system being valid going forward.

However, if your 12-month system returns 10 percent annually, the 11-month system returns 10.6 percent, the 10-month system return 9.5 percent, and so on, the system is showing consistency and is more likely to meet your expectations going forward.

If you have a robust system, like Dual Momentum, then it is best to stick with the simplest approach with the easiest execution. The simple 12-month backtest does exactly that.

Testing Dual Momentum Timing

In the following backtests, I ran systems using data that is closer to Gary's data. Namely, the MSCI USA Index (I can't get monthly return data on the S&P 500 or S&P U.S. Total Market Index), the MSCI World ex-USA Index (1970-1988), the MSCI ACWI ex-USA Index (1988-2017), 3-month Treasury bill rates, and 10-year Treasury bond returns.

In my backtests, it is clear that a 6-month lookback on its own is sub-optimal. The returns are still impressive at a little over 15 percent annually, but the number of trades increased quite a bit over the backtest period.

A 12-month lookback produces quite spectacular results. Again, although my index data and bond data is different, it approximates the results of Gary's model with a 16.95 percent annual return from 1970 through 2017.

The TADM model with the same data as the other tests, but uses an average of the 6-month and 12-month performance, produces great results as well with a 15.95 percent annual return.

Here's what the results of each model look like in a chart format.

Sources: TheRichMoose.com, MSCI Inc., FRED Federal Reserve St. Louis

If you look closely at the chart, you can pick out the winning periods for each timing mechanism. The 6-month lookback did well in the 1970s and 2000s, the 12-month lookback did well in the 1990s and 2010s, and the TADM lookback did well in the 1980s.

None of the models showed any horrible drawdowns, just like none showed any outlandishly better returns in any given period. Instead, it was a matter of one system shining a bit brighter than the rest at certain times.

When all taken together, using the broadest data I can get my hands on (the MSCI World ex-USA instead of the MSCI EAFE) and the safest measurement for cash (the 3-month Treasury bill), the 12-month lookback shone the brightest overall.

That is despite the signals being the exact same for the TADM model and the 12-month model almost all the time.

Changing My Dual Momentum Signal

Going forward, I'm going to be changing the method and terminology of my Dual Momentum signals.

First, I will only state: "U.S. Stocks", "International Stocks", or "Bonds". I will leave it to you to decide on how you exactly want to execute the signal on your end.

Here are some low-cost options to consider:

  • U.S. Stocks
    • Buy a Canadian-listed ETF that tracks the S&P 500, S&P U.S. Total Market, or other U.S. stock index
    • Buy a U.S.-listed ETF which tracks the S&P 500, S&P U.S. Total Market, Russell 1000, Russell 3000, or another U.S. stock index
  • International Stocks
    • Buy a Canadian-listed ETF that tracks the MSCI EAFE, FTSE Developed ex-U.S., or another similar developed market index
    • Pair the developed market ETF with a 25 percent exposure to a Canadian-listed emerging market ETF tracking the FTSE Emerging Markets, MSCI Emerging Markets index, or similar emerging market index
    • Buy a U.S.-listed ETF which tracks the All Country World ex-USA, FTSE All-World ex-U.S., or a similar global stock index
  • Bonds
    • Buy a Canadian-listed ETF that holds short-term or broad bonds with a mix of government and corporate debt
    • Buy Canadian government bonds through your brokerage
    • Buy tax-friendly Canadian bond ETFs
    • Buy a U.S.-listed ETF which holds short-term Treasury bonds, medium-term Treasury bonds, short-term corporate bonds, or a broad bond fund
    • Buy U.S. government bonds through your brokerage

Second, I will be basing my signals on a 12-month lookback period only using the MSCI USA Index, the MSCI ACWI ex-USA Index, and rolling 3-month Treasury bill returns over a one year period. All will be measured in U.S. dollars.

This changes from my former model where I used Canadian-listed ETFs for the performance signal and had the time averaged timing periods.

If you have been following my TADM model so far this year, don't freak out. The signals only changed slightly and my TADM model actually outperformed the 12-month model by 1.15 percent this year. Going forward it means fewer trades and very similar returns.

Comments & Questions

All comments are moderated before being posted for public viewing. Please don't send in multiple comments if yours doesn't appear right away. It can take up to 24 hours before comments are posted.

Comments containing links or "trolling" will not be posted. Comments with profane language or those which reveal personal information will be edited by moderator.

2 Replies to “Updating Dual Momentum”

  1. Thanks for your explanation!

    I’ve read Gary Antonacci’s Dual Momentum book after reading your blog, and I have some questions about how to apply this in practice.

    How does one actually check the 1 year total return for MSCI USA / ACWI ex-US or rolling 3-month T-bill?

    The process of selecting where to allocate funds seems overly complicated, but it’s very likely I’m missing some important distinction being made! It seems to me that you can simplify the process by first determining which of MSCI USA / MSCI ACWI ex-US / 3-month T-bill is giving the best 1 year total return. Then allocating funds in an ETF that corresponds to the winner, e.g.: US S&P500 / MSCI ACWI ex-US ETF / Canadian bond ETF.

    Is that accurate? Or am I missing an important detail about relative vs absolute momentum?

  2. Daren (Editor) says:

    I use MSCI data for each index to calculate 1 year total returns. You can easily find this on their website. For T-bill returns, I use FRED data for 3-month T-bills and simulate holding each T-bill until maturity and rolling them over to a new T-bill. For example, you would buy Oct/17 bills at 1.07 percent, roll that into a Jan/18 bill at 1.41 percent, roll that into a Apr/18 bill at 1.76 percent, and finally roll that into a Jul/18 bill at 1.96 percent. Your annualized return on rolling over 3-month T-bills would be about 1.56 percent.
    It’s true that you are effective investing in the highest performing asset of U.S. stocks, International stocks, and cash (holding bonds when signal is cash). The process is broken down into a relative momentum evaluation and absolute momentum evaluation so the investor understands the process and the mechanics behind the strategy.

Comments are closed