Incorporating Non-Gaussian Stochastic Disturbances in Panel Time Series 1 Models to Account for Structural Breaks

Abstract

This paper addresses scenarios in real-valued series, such as exchange rate and gross 10 production rate, where non-normal disturbances are occur due to structural breaks resulting 11 from events like policy changes, market fluctuations, financial burdens, stock price effects, etc. 12 To justify this, spherically symmetric families of distributions have been explored. We propose 13 a new covariate panel autoregressive model that accounts for multiple structural breaks when 14 the error distribution belongs to the family of spherically symmetric distributions. Bayesian 15 estimation and testing methodology are introduced to estimate model parameters and detect 16 significant deviations from normality using a multivariate t-distribution for analysis. The use of 17 symmetric loss functions for estimating parameter values is compared to classical maximum 18 likelihood estimation, and the Bayes factor is derived to provide strong evidence for the error 19 distribution. A simulation study and an analysis of practical economic series are provided to 20 illustrate the performance of the proposed model.