Market data:

Momentum Trading in Currency and Commodity ETFs

The following article is a draft for an upcoming article on Active Trader Magazine

In the realm of actively managed strategies, the four stand-outs are: value, carry, volatility and momentum.  All four tend to show positive performance across most asset classes, but have very different time horizons and risk profiles.

Momentum always seems to fascinate aggressive traders as it fits with their usually shorter trading time frame and frequency of execution.  I wrote extensively on momentum in the past and ran different studies utilizing futures or Exchange Traded Funds (for a sector rotation study implemented with ETFs please see Activetrader , March 2011, Volume 12, No.3) and validated the expected positive rate of return often occurring with de-correlating characteristics versus traditional assets performance.

The practice of buying assets that are showing persistent price increases or selling assets that are in a negative price pattern gained popularity with trend following strategies implemented successfully in the 1970’s by commodity traders and Commodity Trading Advisors.  The ebullience of equity markets in the following decades during the Great Bull market that started in 1981 and ended in 2000, also proved that momentum could be highly profitable in stocks as well as commodities and foreign exchange.

The difficulty of the markets in the last few years created conditions where momentum showed successful pockets but also required more sensitive parameters.  For example, momentum in FX seems to have retained an edge albeit not as significant as in the past.

But why does momentum work?  Explanations vary from the banal to the exotic but ultimately, momentum’s success is another crack in the misguided theory of efficient markets.  Antti Ilmanen in his book “Expected Returns” (Wiley, 2011) quotes different ideas such as investor under-reaction to fundamental news which leads to a persistence in price series or in the case of commodities he refers to  theories such as hedging pressure and the theory of storage.  In my studies, I find most theoretical frameworks related to commodity roll yields to be flawed but I do believe behavioral components and markets structure are high suspects in determining the success of momentum strategies.

Over twenty years ago, George Soros published a book titled “The Alchemy of Finance,” where it detailed his vision of highly behavioral markets always in a state of disequilibrium.  He called his investing framework “the theory of reflexivity,” in essence a manifesto of momentum investing.  The interesting element of his analysis was not limited to observing price patterns that should not have existed should markets be efficient, but in the determination of the mutual relationship between investors’ involvement and fundamentals.  Soros observed that investors did not have the luxury of natural scientists that could study their subjects in a vacuum and draw conclusions; in fact the very act of investment analysis and subsequent action or inaction had an effect on the fundamentals of the assets taken in consideration.  This subtle feedback loop creates reflexive price waves that constantly move away from equilibrium until the deviation becomes unsustainable and causes a fundamental reversal (i.e. in recent history we can see the case of the real estate bubble as a perfect example).

Ilmanen also quotes additional studies which illustrated successful momentum strategies in 58 liquid investments (Moskowitz-Ooi-Pedersen, 2010).  Based on these findings, the most successful momentum patterns seem to center on the following categories:

  • Stock selection; in this case I believe momentum is underscored by the increase in high frequency trading.
  • Industries and countries.
  • FX; I believe this reflects a mixture of politically driven moves and a high component of technically driven trading.
  • Alternatives; Ilmanen quotes price persistence in assets besides commodities such as real estate, hedge funds and private equity but points out – and I am in agreement – that stale prices and returns smoothing may be contributing.

In line with my thinking of taking studies and ideas and turning them into easily replicable and implementable strategies, I decided to run an update on the 2011 ETF study and concentrate on currencies and commodities.  The idea was not to replicate with statistical correctness the mentioned studies but to look at implementing a cost efficient, liquid ETF strategy centered on currencies and commodities.  Choosing currencies as a main element of trading reflected a strong belief that the current macro-economic decisions that persistently alter markets will continue to find amplified responses in the most macro tool of all i.e. currencies.  On the other hand, the selection of commodities reflected more the need to provide investors with possible solutions to help them manage a portion of their allocation which should really not be left passive; in fact in my view commodities must be actively traded.

Study Methodology

In the currency portion of the study, I selected ETFs which provide exposure to significant economies and that show enough trading liquidity.  The following currency ETFs were chosen:

  • Euro (FXE)
  • Yen (FXY)
  • Australian Dollar (FXA)
  • British Pound (FXB)
  • Canadian Dollar (FXC)
  • Swiss Franc (FXF)
  • Mexican Peso  (FXM – delisted as of March 2012)
  • Swedish Krona (FXS)
  • US Dollar as benchmark (UUP)

In the commodity section of the study, I chose an eclectic mix of single commodities and diversified baskets since not enough single commodity ETFs provided the necessary liquidity.  This inconsistent sample set may have skewed the results and this should be noted; another problem with commodity ETFs is a frequently significant tracking error due to their constructions which may include derivatives and/or credit risk by the issuer of the note (sometimes commodity baskets are issued as Exchange Traded Notes and not as Funds).  The following commodity ETFs/ETNs were chosen:

  • Gold (GLD)
  • Oil (USO)
  • Silver (SLV)
  • Natural Gas (UNG)
  • Coal (KOL)
  • Lithium (LIT)
  • Agricultural Basket (DBA)
  • GSCI Commodity Index (GSG, oil heavy)
  • Rogers Index (RJI)

The investment period covered March 2008 to July 2012 and two different sets of analysis were run; the first back-test calculated a 12 month ranking period and a 1 month holding period while the second study had a 3 month ranking period and a 1 month holding period.  The ranking period served the purpose of calculating the performance of all the currencies/commodities for the purpose of ranking them from strongest to weakest; then the three best currencies/commodities were selected in equal weight and held for one month. At the end of the holding period, the process was run again.  The idea of utilizing different ranking and holding periods (a difference from the 2011 study) was directed at mitigating the impact of the inevitable major turning points when trends end abruptly and revert.

In the study I also incorporated an opposite strategy as a comparison tool; I also ran numbers on going long the three worst currencies or commodities after each ranking period (a contrarian strategy).

I benchmarked the currency portfolio against UUP, a long US Dollar ETF, and the commodity portfolio against GSC, the Goldman Sachs Commodity Index ETF (GSG was also part of the sample).

Tables of Results

   

CURRENCIES STUDY

 
   

Sample: 8 ETFs

 
   

Portfolio: 3 ETFs

 
   

Weight per ETF: 33.33%

 
   

Period: March 2008- July 2012

 
       
     

Momentum

Value Contrarian

UUP

 

Annualized Avg. Returns

-3.12%

6.08%

0.36%

 

Period Average Returns

-0.26%

0.49%

0.03%

12   - 1 period

Average Gain

1.93%

2.76%

2.61%

 

Average Loss

-2.92%

-2.24%

-2.27%

 

Period Maximum Return

6.39%

8.72%

8.43%

 

Period Maximum Loss

-9.23%

-11.72%

-7.01%

 

Winning Periods

29

29

25

 

Losing Periods

24

24

28

 

Annualized Std Deviation

3.15%

3.38%

3.20%

         
 

Annualized Avg. Returns

20.30%

-1.05%

0.36%

 

Period Average Returns

1.55%

-0.09%

0.03%

 

Average Gain

2.47%

2.35%

2.61%

 

Average Loss

-1.26%

-2.62%

-2.27%

3   - 1 period

Period Maximum Return

8.73%

7.59%

8.43%

 

Period Maximum Loss

-5.27%

-11.66%

-7.01%

 

Winning Periods

40

27

25

 

Losing Periods

13

26

28

 

Annualized Std Deviation

2.57%

3.42%

3.20%

         
           
   

COMMODITIES STUDY

 
   

Sample: 10 ETFs

 
   

Portfolio: 3 ETFs

 
   

Weight per ETF: 33.33%

 
   

Period: March 2008- August 2012

 
       
     

Momentum

Value Contrarian

GSG

 

Annualized Avg. Returns

1.37%

-20.56%

-8.66%

 

Period Average Returns

0.11%

-1.90%

-0.75%

12   - 1 period

Average Gain

5.25%

4.39%

4.72%

 

Average Loss

-7.71%

-6.72%

-7.89%

 

Period Maximum Return

15.74%

19.95%

21.50%

 

Period Maximum Loss

-26.01%

-16.10%

-29.67%

 

Winning Periods

32

23

30

 

Losing Periods

21

30

23

 

Annualized Std Deviation

8.00%

7.39%

8.36%

 

Sharpe Ratio

     
 

Annualized Avg. Returns

1.35%

-18.24%

-8.66%

 

Period Average Returns

0.11%

-1.66%

-0.75%

 

Average Gain

5.56%

5.42%

4.72%

 

Average Loss

-7.56%

-5.62%

-7.89%

3   - 1 period

Period Maximum Return

20.20%

19.58%

21.50%

 

Period Maximum Loss

-22.20%

-23.52%

-29.67%

 

Winning Periods

31

19

30

 

Losing Periods

22

34

23

 

Annualized Std Deviation

8.42%

7.75%

8.36%

 

Sharpe Ratio

     
           

Table calculations provided by Rosario Rivadeneyra

Conclusions

The best result was produced by the three month ranking – one month holding study in currencies; this portfolio vastly outperformed its benchmark even with a lower annualized standard deviation; in absolute terms, the annualized average return – 20.3% - was significant enough to make it appealing as a stand-alone strategy.  This result is in line with my previous studies and much more positive.  The longer twelve month ranking period did not produce positive results for the momentum approach but produced outperformance for the contrarian strategy; this is also in line with other studies I have done in the past which reveal that value based approaches require longer time frames to arbitrage price anomalies.  One idea would be to combine the two strategies in one portfolio as their combination would probably lower the overall risk of the portfolio.

As far as the commodity study, I did not produce results as appealing as in the currency space yet the momentum approach did better than just a plain long index exposure.  While the GSCI Commodity Index is significantly down since the starting point of the study, a rotation strategy has actually produced positive annualized rates of return.  A striking element was the largely negative performance of a value contrarian strategy in either ranking periods.

The foundations of this approach are solid; I have found validation across different asset classes, different time periods and by implementing slight methodology variations; it would be interesting to pursue more testing by including for example the following elements:

  • a more comprehensive sample of commodity ETFs as they gain popularity and liquidity
  • test with futures (this would help construct a much wider sample for commodities)
  • test spot currencies (this would help increase the sample of currencies)
  • calculate transaction and slippage costs
  • test alternative ranking periods.

Sources

Asness, Moskowitz, Pedersen, Value and Momentum  Everywhere, 2009

Barberis, Schkeifer, Style Investing, Harvard Institute of Economic Research, 2000

Antti Ilmanen, Expected Returns, Wiley Finance, 2011

Moskowitz, Ooi, Pedersen, Time Series Momentum, 2010

Geert Rouwenhorst, International Momentum Strategies, Yale School of Management, 1997

George Soros, The Alchemy of Finance, Wiley, 1994