Finite-Sample Inference on NGARCH Models With Flexible Distributions

Abstract

This paper aims to investigate the use of the Exponential Power distribution (EPD), a parametric flexible distribution, in the context of auto-regressive models (NGARCH- EPD). The EPD represents an instance of the flexible distributions implemented by Zhu and Galbraith in the context of NGARCH models. The EPD allows for capturing departures from the usual assumption of normality in terms of heavier and lighter tails than those of the normal distribution. A simulation study is presented in order to investigate the finite sample properties of the resulting NGARCH- EPD model both in the frequentist and the Bayesian frameworks. The inference in NGARCH-EPD model in the latter approach has not been done before.

Zhu and Galbraith present a practical application of flexible distributions, while this dissertation covers theoretical and empirical aspects of the same distributions. Therefore, this paper is complementary to the work of Zhu and Galbraith. The results presented in this work can be of interest for industry practitioners because they provide some guidelines on the use of these distributions in practice and their benefits to volatility modelling. The R codes required to implement these models as well as the simulation studies presented in this work are original contributions of the author.