Forums  > General  > bond carry and "global bond factor"  
Page 1 of 1
Display using:  


Total Posts: 45
Joined: May 2006
Posted: 2017-07-07 15:15
I have a cross-sectional carry strategy that ranks the slope between the three month and ten year zero coupon yields (of US, Japan, UK, Germany, Australia and Canada) and buys the ten year bond futures of the top ranked countries and sells those of the bottom ranked countries (equally weighted save for an adjustment to equalize duration).

While I think this strategy is roughly bond market-neutral (the backtest shows nearly zero correlation with the average returns of the bond futures), I am wondering how (in a less rough way) to make this strategy neutral to the "global ten year bond factor" (with global defined to consist of the six 10Y bond futures markets)?

I am thinking the modification would involve a PCA of the returns of the six bond futures but would appreciate the board's thoughts.


Total Posts: 1013
Joined: May 2004
Posted: 2017-07-10 05:25
1. Do a PCA on your 6 bond futures (some caution is needed because you are working with asynchronous data). PC1 should be approximately each bond weighted by the inverse of its volatility. This is the "bond factor" you are looking for.

2. Compute your Long/Short EW portfolio's exposure to PC1, and neutralize it. Voila, your portfolio is now factor neutral. Note that it is not cash neutral anymore, but should still be pretty close.

3. Wonder if you're not splitting hairs and if it's all worth it Wink

"Earth: some bacteria and basic life forms, no sign of intelligent life" (Message from a type III civilization probe sent to the solar system circa 2016)


Total Posts: 45
Joined: May 2006
Posted: 2017-07-10 15:08
1. Is the "global bond factor" just the futures returns time the associated eigenvector?

2. Given that at any time the strategy will only be long two and short two out of the total of six bond futures, what would be a reasonable way to neutralize the portfolio's exposures to PC1?

3. Yes. I am already duration-weighting the notionals. it seems trying vol-weighting may be as far as I want to go.
Previous Thread :: Next Thread 
Page 1 of 1