I recently wrote quite a long post on swap spreads (click here to see that post), covering some general intuition about swap spreads: what does a swap spread represent, why does it move, when does it move, which direction does it go, etc.
My blog stats show that a lot of people have read it, so it seems that it has been quite useful.
Browsing through quant.stackexchange I found my way to a very interesting article written by Paul Teetor which builds a trading model for 10-year swap spreads. Paul calls himself a quant developer and has written a book and plenty of articles on R, a programming language designed for statistical analysis.
The model uses a linear regression of the 10-year swap spread onto:
- 10-year treasuries
- 5-year treasuries
- 5-year swap rates
- S&P 500
- USD Bank stocks
The model is effectively using the 5-year swap spread as part of the signal generation for the 10-year spread. The use of LIBOR, S&P and Bank stocks are not surprising if you have read my post: they are indicators of activities which can move swap spreads.
The buy/sell signal is generated from the residuals combined with a momentum indicator on the residuals.