ABSTRACT
The classic interest rate term structure modeling approach constructs a single yield curve for interpolation and forecast purposes. This chapter shows the implementation of the single-factor Cox–Ingersoll–Ross (CIR) process in a multi-curve framework to extract the market participants’ aggregate expectation of the future short rate at various available tenors from a cross-sectional set of zero coupon bond prices. It seeks approximate solution to the systems of non-linear equations via an algorithm to capture the information contained in the prices of the triplets of zero coupon bonds and represent them as the corresponding inferred CIR model parameters. The chapter uses the various sets of inferred CIR model parameters inferred from zero coupons of different tenors to construct predictors to forecast short rates at different tenors. It compares its forecast performance against a predictor constructed based on the classic single-curve framework. The numerical strategies used in this context follow the strand of model calibration literature for equity options and interest rate model.
