Dual Momentum in 2018

Last week I updated the annual performance and charts for the Dual Momentum model portfolio I track at TheRichMoose blog. You can see the updated charts by clicking on this link:

TheRichMoose Model Portfolios

In my method of calculating Dual Momentum—which varies slightly from Gary Antonacci’s more U.S.-centric method—we closed out 2018 with the worse annual performance since 1970.

The strategy, measured by index performance, lost 8.09 percent for the year. That’s worse than 2015’s performance of -7.08 percent and the prior historical lowest performing year of 1973 which returned -7.33 percent.

At first glance you might say this is proof of Dual Momentum failing as a strategy. In fact, there has been a popular research post by Newfound talking about the "fragility" of Dual Momentum which I'll address as well.

I have received several emails from readers concerned about Dual Momentum's performance in 2018. Particulary how Dual Momentum signaled "U.S. Stocks" for December, experiencing a large drawdown, only to signal "Bonds" for January and miss the recent rally.

While two historically bad years in a single global market up-cycle (2009-2018?) and a "bad signal" is not pleasant, before we write off a strategy we need to dig a little deeper into the true numbers.

We should also seek to understand fragility and the true options we have at our disposal as investors to reduce portfolio fragility.

Understanding the Return Drivers of Dual Momentum

While longer is always better, if we want to effectively compare two unique strategies over the short-term, we can measure market cycle performance two ways: from top-to-top, or from bottom-to-bottom.

As much as we love to ignore market down-cycles, measuring performance only from bottom-to-top is a foolish method. It's just as silly as judging a strategy by monthly or calendar year returns.

Dual Momentum is a strategy that is not designed to outperform in the market's upward moving half-cycle. Dual Momentum is characterized by whip-saw trades in the up-cycle, harming return relative to the global benchmark.

Dual Momentum derives its long-term performance advantage from two aspects. The first is the avoidance of long, deep downwards movements in equity markets. In these events, Dual Momentum signals the investor into bonds and avoids the larger part of those large portfolio drawdowns.

The second advantage is the dollar cycle, driven largely by the U.S. economic performance. When the dollar is strong and the U.S. economy does well relative to other economies (as we've seen this current up-cycle), the Dual Momentum investor is mainly in U.S. stocks.

When the world's other large economies are doing well and the U.S. dollar underperforms (we saw this in the 2003 to 2008 half-cycle), a Dual Momentum investor is largely in International stocks.

Investing with the dollar cycle allows the investor to gain a comparative advantage in the up-cycle, offsetting some of the losses caused by whip-saw trades into and out of bonds.

Current Top-to-Top Performance

Although it is still early to declare 2018 to be the end of the equity market up-cycle (after all, global stocks measured in U.S. dollars have dropped just 18.9 percent from peak-to-trough at this point), for the purposes of this analysis we need to use that number.

This means we'll consider the global market top to be in January 2018 in this cycle.

Global Benchmark Performance

November 2007 to January 2018

Total Cumulative Return:  +81.20 percent
Compound Annual Return:  +5.97 percent

Dual Momentum Performance

November 2007 to January 2018

Total Cumulative Return:  +110.27 percent
Compound Annual Return:  +7.52 percent

We can see here that on a top-to-top basis, Dual Momentum has handily outperformed the Global Benchmark. In fact, you would be more than 15 percent wealthier over the period choosing Dual Momentum.

Current Bottom-to-Bottom Performance

Again, just as it may be premature to call January 2018 as the top of this market cycle, it is even more premature to declare December 2018 to be the bottom of the market down-cycle. However, we’ll use December to illustrate the performance comparison.

Global Benchmark Performance

March 2009 to December 2018

Total Cumulative Return:  +147.63 percent
Compound Annual Return:  +9.83 percent

Dual Momentum Performance

March 2009 to December 2018

Total Cumulative Return:  +100.32 percent

Compound Annual Return:  +7.32 percent

In this bottom-to-bottom comparison, we see the Global Benchmark outperforming Dual Momentum. This should come as no surprise.

The Global Benchmark suffered much greater losses from top-to-bottom (2007-2009) and Dual Momentum did not even get signaled over to bonds in 2018.

Further, Dual Momentum’s bottoms do not correspond to the Global Benchmark. In the prior down-cycle, Dual Momentum hit bottom in June 2010. That's with a much lower total drawdown than the Global Benchmark at just -19.4 percent.

Using that measure, this is Dual Momentum’s bottom-to-bottom performance.

Dual Momentum Performance

July 2010 to December 2018

Total Cumulative Return:  +127.10 percent
Compound Annual Return:  +10.13 percent

Over the shorter timeframe of Dual Momentum’s bottom-to-bottom period, the compound annual growth rate was higher than the Global Benchmark.

Fragility and Dual Momentum

Upon hearing about Newfound's post, Gary Antonacci published a response post on his blog. Gary's post was on point and is a notable read for any Dual Momentum investor.

To understand the argument, I think it is first important to understand true fragility. Loosely using Nassim Taleb's work, fragility is a system that is damaged by shocks in an exponential fashion. Robustness is a system that is able to withstand great shocks with no significant damage or benefit. And anti-fragility is a system which benefits from shocks in an exponential fashion.

Fragility is not a difference in return of a few hundred basis points in a single calendar year, an equity market half-cycle, or even over several decades. Particularly not when the returns vary because of a slightly different lookback period applied to a single investing model.

Rather, fragility is a concept regarding the likelihood of exponential (or total) loss in adverse conditions. In some aspects, all financial markets and models are fragile. As investors, we take certain things for granted: political stability, currency stability, property ownership rights, and so on.

Addressing True Fragility with Robust Elements

Using a better definition of fragility, Dual Momentum as a broad model is indeed fragile to a certain degree. It is a long-only form of momentum investing with a restricted number of broad markets.

However, we should not ignore the elements introduced in Dual Momentum which ultimately make the system quite robust—though not anti-fragile.

First, Dual Momentum uses very broad holdings that reduce catastrophic shocks. Unlike say the inverse volatility trade that blew up in February 2018, the broad U.S. stock market, the broad International ex-USA stock markets, and the broad bond index are not likely to blow up overnight. These are highly diversified holdings backed by underlying hard assets and substantial revenues.

However, a long and steady unwind is not impossible. We saw this in the Japanese stock market from 1989 through 2003. That said, a Japanese investor using a local version of Dual Momentum (MSCI Japan, MSCI Kokusai, and Japanese CDs) would have done extremely well. Read this post.

In long equity market drawdowns, Dual Momentum puts the investor into bonds. Setting aside inflationary pressures for the moment, a broad basket of bonds is very robust. Particularly if the bond holder uses a mix of short duration and long duration, local and international bonds.

As stated earlier, Dual Momentum (nor any other strategy I can think of) will not protect investors from hyperinflation or seizure of private property. However, within the bounds of our current economic environment, I would categorize Dual Momentum as comfortably robust.

Lookback Period Variation and Fragility

In Newfound's post, Corey Hoffstein suggests Dual Momentum is fragile precisely because it uses a single lookback period (12-month) which does not always provide the best performance. He calls this specification risk in his post.

Hoffstein in particular alludes to the 10-month lookback as being superior in this latest market half-cycle while the 9-month lookback is an under-performer. This, he claims, shows evidence of fragility in the traditional 12-month Dual Momentum system.*1*

This idea is reinforced by 2018's performance where 12-month Dual Momentum was in U.S. stocks for December and recorded a loss for the year while 10-month Dual Momentum would have been in bonds, recording a positive result for the year.

Hoffstein suggests a better approach would be using multiple lookback periods (6 month, 7 month, 8 month, 9 month, etc.) for smoother results and to hedge against any bad performance caused by a single lookback period.

Hoffstein isn't necessarily wrong in his analysis from a simple mathematical perspective. But he may not be presenting a reasonable solution when looking at the bigger picture.

In his short backtest, he treats all lookback periods both as equally effective and meaningfully different—they are not. And he doesn't analyze if it is mathematically advantageous to track seven lookback periods compared with a more reasonable one, two or three periods over a longer timeline.

While I don't mean to be hypercritical, Hoffstein's opening statement in his conclusion is especially weak: "The odd thing about strategy diversification is that it guarantees we will be wrong.  Each and every year, we will, by definition, allocate at least part of our capital to the worst performing strategy.  The potential edge, however, is in being vaguely wrong rather than precisely wrong.  The former is annoying.  The latter can be catastrophic."

First, choosing slightly different lookback periods within the narrow bounds of this long-only strategy does not amount to strategy diversification. Second, picking the "wrong" lookback period does not equal a catastrophic result.*2*

I would challenge him to post to the full stats on implementing his model over a period longer than the current half-cycle. Specifically, go back to 1970 (when MSCI EAFE data is available) and analyze each lookback period individually, then combine the periods into a portfolio, and share the real costs of running such a complex system.*3*

I suspect Hoffstein will find the number of trades to increase significantly and discover using seven lookback periods offers little advantage to the single 12-month lookback. Even if he narrowed it down to a more reasonable two or three periods, the results are still not that advantageous by most measures. The trading and tax costs, however, increase dramatically.*4*

(I have extensively examined combining multiple lookback periods—specifically the 6-month and 12-month—but ultimately decided the benefits are slim to none over the long term.)

Further, I would challenge Hoffstein's suggestion that using different lookback periods amounts to strategy diversification. The historical returns across longer Dual Momentum lookback periods—which I'll suggest are lookback periods between 6 months and 12 months—are highly highly correlated.

If you want strategy diversification, don't choose 6-month Dual Momentum and 12-month Dual Momentum as two separate strategies. Instead, pair Dual Momentum with something completely different such as a long-short approach, currency trading, or managed futures. While not perfect, only non-correlated strategies provide a reasonable opportunity for real diversification.


Corey read my post and was kind enough to contact me and clarify a few points. I don't want to change my original content, but here are some of the explanations he pointed to by linking several other of his posts.

1* Regarding his concept of fragility, Corey states he was speaking specifically to fragility in the idea of inconsistency of returns within a specific model when slight variations are made. If I'm understanding Corey's point correctly, he believes Dual Momentum lacks robustness since the returns with a 9-month lookback period can be significantly different from the 10-month lookback when measured over shorter time periods. Also, he points out each month is independent from the prior month, so there should not be an inherent expectation that things will even out in the long run.

2* Regarding the idea of strategy diversification, again Corey states he was referring to diversifying within the Dual Momentum model to more closely align potential investor results with investor expectations. A system using multiple lookback periods is less likely to experience a meaningfully anomalous result due to a poor signal in one timing period.

3* Corey suggests most investors are not true long-term investors and results within a shorter time period like this half-cycle do matter. He states even one (or more) unlucky year could have a big impact on overall wealth.

4* Corey runs a research firm for institutional clients and financial advisors. These individuals may need to make transactions regularly to account for redemptions and so on as a normal part of their business. Also, he argues transaction costs are extremely low and taxes would only apply in regular investment accounts.

**Overall this was a great discussion and Corey brought up some fair points. I will be researching his suggested strategy for Dual Momentum some more over the coming weeks to better understand his position. If it turns out to be better, I may once again adjust the Dual Momentum system I share.

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.

One Reply to “Dual Momentum in 2018”

  1. “I will be researching his suggested strategy for Dual Momentum some more over the coming weeks to better understand his position. ”

    Please, share your conclusions Daren. Thanks.

Leave a Reply

four × = four