極大化平滑度與精確度之利率期限結構估計

Estimating the Term Structure of Interest Rates with Maximum Smoothness and Accuracy

本研究以台灣公債市場為研究對象，並利用Nelson－Siegel模型與Nelson－Siegel－Svensson模型作為研究基礎，比較利率期限結構配適能力之優劣。此外在增進模型平滑度的前題下，利用GCV模型、VRP模型與Anderson－Sleath模型對Nelson－Siegel模型與Nelson－Siegel－Svensson模型加以 結合。實證結果發現，Ne1son－Siegel模型在結合GCV模型、VRP模型與 Anderson－Sleath模型之創新後，在平滑度上皆較原Nelson－Siegel模型更為優 良。而Nelson－Siegel－Svensson模型則以結合GCV模型後，能提供較佳之平滑度。當同時考量精確度與平滑度之配適能力後，本文認為在Nelson－Siegel模型中，以結合Anderson－Sleath模型能提供較佳之估計結果；Nelson－Siegel－Svensson模型則以結合GCV之模型，可以提供較佳的估計結果。

The purpose of this paper is to compare the fitting performances of the estimation of the term structure of Taiwan Government Bonds market based on the Nelson-Siegel and the Nelson-Siegel-Svesson model. Three fitting-smoothness improving models (the GCV, VRP and Anderson-Sleath model) are used to increase their fitting performances in accuracy and smoothness. The empirical results indicate that the Nelson-Siegel corrected by the GCV, VRP and Anderson-Sleath model produces better smoothness. The Nelson-Siegel-Svesson corrected by the GCV model also has a better fitting-smoothness result. If we take both the fitting accuracy and smoothness into consideration, the Nelson-Siegel corrected by Anderson-Sleath model and the Nelson-Siegel-Svesson corrected by the GCV model are better choices.