Interest rates are going up. Interest rates are going down. The 30 year bull market in bonds is coming to an end. The FED cannot afford to raise rates significantly without killing the economy. Bonds will have negative real returns going forward. We’re in the biggest bond bubble of all time. Bonds are still the best diversifier of equity risk. Have you heard any of these lately? How about over the last few years? Probably incessantly.

There seems to be more hand wringing and fretting over the future of bonds than even equities. And that is saying something. We could jump into a bunch of fundamental and historical analysis to see which scenarios are more likely. We could look at international markets like Japan for analogues to the US situation, etc…But I really have zero interest in being an active bond picker again, or for that matter, a stock picker. As I detailed in this post, most of the time it’s not worth it and most of the time you can’t even get our of your own way. Can we then apply some of the same quant investing concepts that we use for equities to bonds and not have to fret over the future of bonds? The short answer is yes.

The most powerful factor, some people call it an anomaly, in investing is momentum. Eugene Fama, Nobel prize winner and one of the fathers of Efficient Market Theory, has called momentum the biggest challenge to the efficient market hypothesis.  And it applies just not to equities but to pretty much any asset class. I’ll just point you to the research by Antonacci and Moskowitz but basically momentum works across all asset classes, across different time frames, including bonds. In Antonacci’s book Dual Momentum, he specifically mentions applying a momentum approach exclusively to bonds as part of his GBM portfolio (page 129 if you want to follow along). Let’s see how that works.

The Antonacci bond trend following portfolio couldn’t be any easier. It chooses from the universe of US long treasuries, US high yield, Global government bonds, and US treasury bills. Then once a month it ranks these bond classes by 12 month return and choose the top one to invest in. Re-do once a month. It definitely improves the performance of the Antonacci GBM portfolio as described in the book. I took a look at just the bond portion of GBM to see how it did since 2007. Since the history of many ETFs is limited I was not able to go back very far. I used the following ETFs to represent the 4 basic bond classes; TLT, JNK, IGOV, and SHY. The results compared to Vanguard’s Total Bond Market ETF (BND) are shown below.

GBM bond vs BND 2007 to 2014

Not bad. 60% higher returns over the total bond market. And you never need to worry about rising rates, falling rates, etc.. The model gets you into the best performing bond market segment. It’s a great model but I think it has some difficulties for the individual investor. First, though over time 12 month returns has been the best predictor of future returns markets seem to move faster these days. In addition, investors don’t seem to have the patience to wait for 12 month returns to signal an investment change. And finally, I think most investors would not be able to stick with switching their entire bond portfolio between 1 ETF. So, I took a look at a few permutations of the basic model. In the first permutation I simply used 6 month returns vs 12 month returns in the original model. In the second permutation, I added some more segments of the fixed income market to the portfolio and chose the top 1. I added intermediate US gov’t bonds, US inflation protected bonds, US corporate bonds, and US municipal bonds. In the final permutation I expanded the portfolio to the top 3 ETFs by 6 month returns instead of the top 1. I think more investors could stick with such a portfolio. The results of the various permutations are shown below.

Various GTAA Bond Models Dec 2014

Simply going to 6 month returns (GBM Bond 6mo) increases returns and sharpe/sortino ratios significantly. Adding more segments of the bond market (GTAA Bond 1) doesn’t do much when only choosing the top 1 ETF. Finally, choosing the top 3 ETF (GTAA Bond 3) from the expanded universe has the best risk adjusted return ratios, Sharpe and Sortino, and the lowest drawdowns and risk (standard deviation). This is probably the portfolio most easily implemented by the majority of investors.  Below I show the year by year results of the GTAA Bond 3 portfolio vs the original GBM Bond and the total bond index (BND).

Summary of GTAA vs GBM Bond Models Dec 2014

There you have it. An automatic quantitative bond model for any interest rate environment. No need to worry. No need to fret over the FED’s next decision. Apply one of these bond models with your favorite quant equity strategy and you’re probably going to do OK.


11 Comments

fjpenney (@fjpenney) · December 7, 2014 at 2:30 am

The challenge we all have with bond ETF strategies is that we have a very limited data series and all bond ETF’s have operated in a bull market given that interest rates have been falling since 1981. I agree with your approach other than I would not buy a bond ETF if it had a negative return over the lookback period. You mention Antonacci’s book but you didn’t state if you are using an absolute momentum rule.

    libertatemamo · December 7, 2014 at 6:27 am

    That is very true for the ETFs. And my biggest concern when I started this effort. But we do have data for several types bonds going back to pre-70s. For example, I’ve looked at 3 mo T Bills, 10 year Bonds, corp bonds, and muni bonds using the monthly FRED data series from the Federal Reserve and they support a trend following model in a rising rate environment as well. This is quite a painstaking analysis but so far has supported the approach. In a rapidly rising rate environment the model keeps you in short bonds most of the time. Also, Antonacci’s GBM model at least goes back to 1974.

    I used the method he uses in his model which is dual momentum, both absolute and relative momentum. Yes, I agree with not buying the ETF if it has a negative return over the look back period. Since the selection includes very short term US govt bonds I think this would be a rare event unlike the case with the equity model (GEM) which needs to overcome the risk free rate as a hurdle.

    Paul

Rich D. · December 7, 2014 at 7:02 pm

Great Blog, Please keep it up. Les Masonson has a great site, Buy-Don’t Hold and has an interesting book on market timing. Take care, Rich

    libertatemamo · December 7, 2014 at 7:24 pm

    Thanks Rich. I have the book and visit the site.

    Paul

Steve Nelson · December 7, 2014 at 8:48 pm

We are all fortunate to have your advice available to us as we make our own personal financial decisions!

    libertatemamo · December 8, 2014 at 7:54 am

    Thanks Steve. Happy to help where I can.

Ben O · December 29, 2014 at 3:28 pm

Hi Paul, nice post, did yous see that Gary Antinacci recently posted something exactly along those lines?

Quick question: in your GTAA versions, did you use the 6 or 12 months momentum?

Regarding the question of what happens in a raising rate environment, Faber stated that using Tbonds instead of Tbills in the IVY portfolio raised returns since the 70s and importantly did not decrease them during the 70s when interests where raising.
Cheers
Ben

    paul.novell@gmail.com · December 30, 2014 at 4:52 am

    Hi Ben, yes I saw Antonacci’s post. In my GTAA bond momentum portfolio I used 6 month returns.

    As for Faber’s analysis of Tbill vs bonds, the only drawback of his analysis is that his data goes back to only 1973. That means it only captures the last part of the great rise in rates (1973 to 1981), albeit the worst part of the increase.

    Paul

Carl · May 22, 2015 at 2:17 pm

Nice work. When you added the other bond types (intermediate US Gov, US inflation protected bonds, US corporate bonds, US municipal bonds), what ETFs did you use? Finally, did you consider Emerging Market bonds?

Thanks
Carl

    paul.novell@gmail.com · May 23, 2015 at 7:46 am

    Thanks Carl. Doesn’t really matter what ETFs you use. Results are similar. I think I used TIP, LQD, and MUB. Yes, I have done backtests using EMB and it works well.

    Paul

Andrew · November 10, 2015 at 5:34 pm

Have you tested using international bonds?

Comments are closed.