The World’S Simplest Bond Valuation Model

The nautical chart to a higher house shows what I mean value is the simplest bond valuation model possible (with the added constraint beingness that the model sort-of works). The depression charge per unit of measurement environs is taking its toll on the model accuracy recently, simply if I attempted to right for that, it would no longer live on the Simplest Bond Model.

The model tin live on explained inward 1 sentence:

the best prediction for the average Fed Policy charge per unit of measurement going frontward is the average experienced over recent cycles.

Yes, the model-predicted 10-year yield is the 10-year moving average of the effective Fed Funds rate. The model is implicitly based on the supposition that nosotros are inward a stable inflation environment, which I mean value is a proficient description of the electrical flow province of affairs (see my medium-term inflation outlook). The model does non bargain good amongst regime shifts; it got Volcker-ed from the mid-1970’s to the mid-1980’s.

As my Disclaimer downwards below notes, this weblog does non offering investment advice. This is a real elementary model, as well as I advise taking it amongst a big grain of salt. But I desire to brand the next observations next from this exercise:

• Realised curt rates lead keep been much lower than what the consensus views every bit neutral for a long time. 
• This is what I thought every bit a pure bond valuation model, where nosotros stimulate to predict the score of yields without using information from long-dated bond yields. You tin larn a much meliorate stand upwards for if you lot role yields from other maturities, simply that turns into the model into a relative value model, which is a real dissimilar beast. And if you lot feed inward previous values for the 10-year yield, errors are manifestly reduced, since yields deed slow over fourth dimension (“autocorrelated”). For example, the previous month’s score for the 10-year yield gives a little prediction error, simply it offers no guidance whether bonds are over- or under-valued.
• The existent essay for a model similar this is non statistical tests, simply whether it generates profitable trading signals. This has to live on done without likewise extensive information mining. Avoiding information mining is extremely hard when creating valuation models; analysts demand to seat a high value on simplicity as well as robustness.
• I thought this model every bit a baseline for evaluating other valuation models: does the other model produce a meliorate project than this?
• A to a greater extent than serious model should live on forward-based. The score of the 2-year yield (for example) is extremely of import element for the path of frontward rates. If you lot disagree amongst the score of the 2-year past times to a greater extent than than 100 footing points, you lot produce non demand a model to state you lot what to do. So if you lot assume that the front end of the crimp is fifty-fifty around fairly priced, you lot tin larn a much meliorate read on where longer maturity yields should be.
• The model does non lead keep into trouble organization human relationship the one-way run a jeopardy facing curt rates when they are at zero. With the null saltation hit, that optionality is of increasing importance.

(c) Brian Romanchuk 2013