期刊名称：International Journal of Economics and Finance
印刷版ISSN：1916-971X
电子版ISSN：1916-9728
出版年度：2019
卷号：11
期号：9
页码：1-11
DOI：10.5539/ijef.v11n9p1
出版社：Canadian Center of Science and Education
摘要：The level of the yield curve is strongly associated with a very important macroeconomic variable for developing economies: the inflation. Therefore, it becomes relevant for economic studies the development of a time series model that can accurately predict this variable. This article proposes the estimation and prediction of the yield curve level using the GAS (Generalized Autoregressive Score) class of time-varying coefficient models. The formulation of these models facilitates a general framework for time series modelling presenting a series of advantages, including the possibility of specifying any conditional distribution deemed appropriate for the yield curve level. In addition, the complete structure of the predictive distribution is transported to the mechanism that updates the time-varying parameters, via score function. When analyzing the evaluation criteria, the measures of adherence, and both Wilcoxon and Diebold & Mariano tests, it was verified that the adjustment of the GAS model (2,2) with gamma distribution to the series containing the Brazilian Yield Curve level of January 2006 and February 2017 presented a satisfactory result.
