The purpose of this paper is to show that an affine model which incorporates the condition of no arbitrage enables improvements in forecasting the term structure of interest rates in Mexico. The three factors of the yield curve (level, slope and curvature) used in the model are estimated by the method of principal components. The forecasting model is specified as a linear relationship between each of the interest rates and these factors, for maturities of 1 to 60 months. Affine model predictions are compared with four benchmark models: a forward rate, an AR(1), a VAR(1), and a random walk model. The main finding is that the affine model has a performance comparable to benchmark models for horizons of 12 and 18 months, except for the random walk model. However, improving its forecasting performance for the 24-month horizon, and especially for 60-month maturities.
Affine Model Forecasts Yield Curve Principal Components Condition of no Arbitrage