January 25, 2017

Country ETF Rotation

My recent research has been focused on finding strategies that are not highly correlated with the S&P500 index. One of my most popular posts is ETF Sector Rotation. The idea for this post is to apply those concepts to a list of country ETFs. Would this produce decent returns that were not highly correlated to the S&P500 index? I would like to see the correlation under .50. What about adding a filter to not enter an ETF when it is highly correlated with the S&P 500?

ETF Universe

We will be using this list of country ETFs.

  • iShares MSCI Israel Capped (EIS)
  • WisdomTree India Earnings (EPI)
  • iShares MSCI Australia (EWA)
  • iShares MSCI Canada (EWC)
  • iShares MSCI Sweden Capped (EWD)
  • iShares MSCI Germany (EWG)
  • iShares MSCI Hong Kong (EWH)
  • iShares MSCI Italy Capped (EWI)
  • iShares MSCI Belgium Capped (EWK)
  • iShares MSCI Switzerland Capped (EWL)
  • iShares MSCI Malaysia (EWM)
  • iShares MSCI Netherlands (EWN)
  • iShares MSCI Austria Capped (EWO)
  • iShares MSCI Spain Capped (EWP)
  • iShares MSCI France (EWQ)
  • iShares MSCI Taiwan Capped (EWT)
  • iShares MSCI United Kingdom (EWU)
  • iShares MSCI South Korea Capped (EWY)
  • iShares MSCI Brazil Capped (EWZ)
  • iShares China Large-Cap (FXI)
  • VanEck Vectors Russia (RSX)
  • iShares MSCI Thailand Capped (THD)
  • iShares MSCI Turkey (TUR)
  • iShares MSCI South Africa (EZA)
  • iShares MSCI Mexico Capped (EWW)
  • iShares MSCI Singapore Capped (EWS)
  • iShares MSCI Japan (EWJ)
  • iShares MSCI Chile Capped (ECH)

Testing will be from 1/1/2007 to 12/31/2016.

Simple Rotation Test

The first test is a simple momentum method.

Rotation Rules

  • At the end of the month,
    • Consider ETFs only with the 21 day moving average of Close * Volume over $5 million
    • Rank the ETFs from high to low of their (6,12) month returns
  • Buy the top (1,2) ETFs on the next open

The six and twelve month values were chosen because those are most often referenced in the research I have seen. The spreadsheet also contains 3 and 9 month rotation periods.

Results

Not a great start. I could not believe how bad the results were. The correlations are all above .50. On to the next idea.

Rotation with trend filter and backup ETF

Dual Momentum has the concept of when an ETF does not pass some filter, instead of investing in that ETF you invest in some alternative ETF. This ETF could be SHY (iShares 1-3 Year Treasury Bond) or IEF (7-10 Year Treasury Bond).

New Rotation Rules

  • At the end of the month, rank the ETFs from high to low of their (6,12) month returns
  • Buy the top (1,2) ETFs on the next open
  • But if the top ranked ETF is below the their (6,12) month moving average, then instead of buying the ETF buy the alternative ETF(SHY,IEF)

In this example we are using 6 month return and IEF as the alternative ETF. If EWA and EWC are the two top ranked ETFs by 6 month return. Before buying them we check and see if they are trading above their 6 month moving average. Say that EWA is not, then instead of buying EWA we buy the alternative ETF, IEF.

Results

The results have improved. Higher returns and lower drawdowns but still not very good. The correlation has dropped under .50. I would like to see the CAR over 10% and MDD under 30%.

Rotation with dual ranking, trend filter and backup ETF

Another concept with ETF rotation strategies is using more than one ranking. We rank the ETFs by two time periods, add the rankings and then rank again. Taking the top ranked.

New Rotation Rules

  • At the end of the month,
    • Rank1 = rank the ETFs from high to low of their (3,6,9,12) month returns
    • Rank2 = rank the ETFs from high to low of their (3,6,9,12) month returns
    • Rank3 = Rank1 + Rank2
    • Rank rank3. In case of ties, use Rank1 as the tie break.
  • Buy the top (2) ETFs on the next open
  • But if the ETF is below the their (6,12) monthly moving average, then instead of buying the ETF buy the alternative ETF(IEF)

Results

A slight improvement but still not very good. The correlations are under .50.

Correlation Filter

For the last test, we will try adding a correlation filter. An ETF will not be considered if the 21 day correlation between it and the S&P500 index is over .75.

Correlation Results

Well this is surprising. The results got worse and the correlation with the SPX went up. That was a fail. I looked at a maximum correlation of .50 and the results are even worse which you can see in the spreadsheet.

Spreadsheet

File the form below to get the spreadsheet with lots of additional information. This includes yearly breakdown, (3,6,9,12) results, and much more.

Final Thoughts

This test did not come close to what I was hoping for. I am surprised on how poor the results were. Got any ideas?

Edit 2/22/2017: See the results of reader’s suggestions at Country ETF Rotation – Reader’s Suggestions

Good quant trading,

Fill in for free spreadsheet:

spreadsheeticon

Click Here to Leave a Comment Below

Tony Roylance - January 25, 2017 Reply

Cesar, I’ve been studying similar strategies, with similar results. However, how about adding the country CAPE values into Aux1, and then looking for positive momentum with the best CAPE value? I’d be happy to provide you my dataset of country CAPE values if you’d like.
Kindest Regards,
Tony Roylance

    Cesar Alvarez - January 25, 2017 Reply

    That is an interesting idea. Is there site that I can get the CAPE values for free? I appreciate you offering but if I find something I like, I want to know that I can get access to the data.

Gerry Wollert - January 25, 2017 Reply

You might want to add two additional periods: 1 month and 3 month returns to your analysis. In other words, rank the ETFs on the sum of the 1, 3, 6, and 12 month returns. I would surprised if the results didn’t show significant improvement.

    Cesar Alvarez - January 25, 2017 Reply

    I have not tried that. So simply add the four numbers and rank from highest to lowest? If I get a couple more ideas, I can do this for the next blog post. Thanks.

Tony Roylance - January 25, 2017 Reply

For current values, Star Capital in Germany (http://www.starcapital.de/research/stockmarketvaluation) keeps their site up to date, I believe quarterly. Unfortunately, for historical values, Global CAPE values can be difficult to obtain.
Tony Roylance

Aaron Smith - January 25, 2017 Reply

The main problem is your testing period. Almost all international countries have sucked as far as returns are concerned for the past 10 years. Compare your rotation system to the returns of the general market. The Vanguard All-World Ex-US ETF has returned a whopping 1.3% CAGR w/ a 61% drawdown since 2007. I’d say your system is doing pretty well all things considered.

To increase performance, run the system from 2003-2007, I’ll bet returns will be outstanding.

You will also need to increase the number of positions. 2 positions when you have a basket of 30 possible ETF’s is not enough. Try testing between 5-10 positions each month and see what that looks like.

    Cesar Alvarez - January 30, 2017 Reply

    I don’t know if enough of the ETFs go back to 2004 but if there are enough it would be an interesting test to see how this does from 2003-2007. In general I have found the more ETFs you add the worse your results. I can try doing 5 or 10 and see what the results are.

Curt Carlson - January 25, 2017 Reply

Cesar,

Could also try adding, either as a rank value, or as a filter, a short term oversold indicator. One idea for the indicator are % below 5, 10, or 20-day high. Another idea is 100-RSI(3). So you avoid the risk of buying something overbought, get a bit a reversion to mean benefit, but add a slight bias into ETFs that are just beginning a correction.

If you tried this as a filter, and ended up with an SHY position, then mid month you could check again for a highly ranked, oversold ETF to buy. But only allow a rotation from SHY to an ETF mid month. Keep the full rotation (ETF sells) at the end of the month.

Complicated, but eliminates the bias toward the most overbought ETFs that your current approach has.

    Cesar Alvarez - January 25, 2017 Reply

    Another good idea. It is starting to look like I have enough ideas to test out and for the next blog post.

Kapil - January 26, 2017 Reply

Seems like the system is getting whipsawed to death. Maybe momentum isn’t the best idea for country funds. Have you considered buying the weakest funds instead?

Peter - January 26, 2017 Reply

A few suggestions Caesar.
1) Create a separate emerging markets universe. I’ve found that EM doesn’t seem to mix well with developed countries in ranking schemes.
2) Use a fast 3 mo MA filter for the EM universe. Seems to fit better with extreme volatility of EM universe. (6mo MA filter works better for developed mkts)
3) Add volatility as a 3rd ranking factor

Here are the ranking parameters I use, which seem to work well. They’re my default ones in ETFReplay.
Return A 6 mo, 35% Weight
Return B 3 mo, 30% Weight
Volatility (daily) 3 mo, 35% Weight

I think it will be hard to get any equity model’s SPY correlation much under 50%. You need bonds in the rotation to do that. Good luck! I really appreciate your informative posts.

    Cesar Alvarez - January 30, 2017 Reply

    I decided to leave the volatility filter out in this test to see if just using returns would produce anything interesting. My goal is to find something less correlated to what I am already trading in the US markets. I figured with so many countries to pick from, that it might be possibile.

Sergey Vedernikov - January 27, 2017 Reply

Hi Caesar,

Yep, simple momentum doesn’t work for stock indices. That would be too simple, wouldn’t it?

Try to add reversion logic to your momentum – say, rank short-term returns, if it is too high – ETF gets low score, if it is too low – high score. Then add up your best momentum and reversion scores – results will improve. The only problem – reversion will increase turnover of your strategy.

    Cesar Alvarez - January 30, 2017 Reply

    One typical idea is to ignore the most recent month return or rank it inversely. Simple ideas can still work. I am not looking for home run type returns.

Mike Moody - January 27, 2017 Reply

These results are not as bad as you imagine. Look, for example, at a benchmark like CWI. That is the all-cap world index minus the US. Total return over the last decade is around 16%. Not annually–total. Returns of 5-6% per year are not so bad under those circumstances. Momentum will work just fine with country funds, but during the 2007-2016 there hasn’t been much trend. Sometimes it also helps to hold slightly more securities. You might try looking at a top 3-4 strategy also, with the bond option during lousy markets.

    Cesar Alvarez - January 30, 2017 Reply

    I think you are right about setting my goals to high for this period. It do have a bond option, IEF during bad times.

Carl - February 12, 2017 Reply

Rank 1 and 2 seem to be the exact same thing. What is different?

Thanks
Carl

    Cesar Alvarez - February 13, 2017 Reply

    For example, you can do rank 1 by 3 month return and rank 2 by 6 month return. Or rank1 by 6 month return and rank 2 by 12 month returns. Yes sometimes they can be the same. Does this make sense?

Pete - May 15, 2017 Reply

Hi Cesar-

If you want to look at something with limited correlation to the S&P then why not look at currencies? I appreciate that it’s not what you normally do, and as it’s an unregulated market there could be a higher risk of getting stung by a Corzine like character. But it’s a diversification that may have potential.

Thanks

Pete

    Cesar Alvarez - May 16, 2017 Reply

    What I have done is look at futures ETFs with no real luck. I have not done much with currencies in years.

Vitaly Belitskiy - May 31, 2017 Reply

Hi Cesar,

I’m far from an expert in trading, but after back-testing quite a few types of rotational strategies the only thing that seems to hold true is that market just doesn’t reward traders for having exposure 100% of the time. The biggest punishment for these strategies comes in crisis periods like 2008 or 2011 and with ETFs modest returns it takes a long time to recover from a drawdown. So one would logically assume that it is better just to refrain from trading altogether in times of violent environment. But the tricky part is in determining when the market switches its regime from bull/sideways/bear, because there is just not enough data for statistical significance of such an indicator. It would be interesting to hear your thoughts (maybe post) on this topic.

Thanks.

    Cesar Alvarez - May 31, 2017 Reply

    You are correct. What these type of strategies general out perform is over an entire bull bear market cycle. They tend to underperform in the bull and overperform in the bear.

Leave a Reply: