EcoSta 2024: Start Registration
View Submission - EcoSta2024
A0402
Title: Nonparametric estimation and forecasting of time-varying parameter models Authors:  Yu Bai - Monash University (Australia) [presenting]
Abstract: Issues of using a local estimator in a forecasting model affected by parameter instability are addressed. The choices of kernel weighting function are analyzed, and the bandwidth parameters are associated with the local estimator. The asymptotic optimality of the bandwidth selection procedure is proven, and an analytical criterion is provided for the choice of the kernel weighting function. The theoretical results are examined through an extensive Monte Carlo study and an empirical application on bond return predictability.