Defining accurate: a look into the Nelson-Siegel model for yield curve forecasting

Abstract

This thesis examines the Nelson-Siegel model, with a focus on the Mexican and United States yield curves over a 13-year period. The study investigates the model’s accuracy and adaptability in the face of exogenous shocks that produce sudden changes to the curve. Utilizing the dynamic Nelson-Siegel model and a cross-validation process, we optimize the λ parameter to enhance forecasting precision. The results demonstrate that while the Nelson-Siegel model is generally accurate, it faces challenges during periods of yield curve stress. Our findings indicate that short-term maturities are more accurately forecasted for the United States, whereas long-term maturities yield better forecasts for Mexico. This research contributes to the literature by identifying the model’s limitations and proposing an improved forecasting methodology that adjusts to sudden yield curve changes. The implications of these findings are significant for central bankers and private investors, offering insights into better monetary policy and investment decisions.