File(s) under permanent embargo
A threshold mixed count time series model: Estimation and application
journal contributionposted on 2023-05-20, 10:12 authored by Dungey, M, Martin, VL, Tang, C, Tremayne, A
A new class of integer time series models is proposed to capture the dynamic transmission of count processes over time. The approach extends existing integer mixed autoregressive-moving average models (INARMA) by allowing for shifts in the dynamics of the count process through regime changes, referred to as a threshold integer autoregressive-moving average model (TINARMA). An efficient method of moments estimator is proposed, with standard errors based on subsampling, as maximum likelihood methods are infeasible for TINARMA processes. Applying the framework to global banking crises over 200 years of data, the empirical results show strong evidence of autoregressive and moving average dynamics which vary across systemic and nonsystemic regimes over time. Coherent forecast distributions are also produced with special attention given to the Great Depression and the more recent Global Financial Crisis.
Publication titleStudies in Nonlinear Dynamics and Econometrics
PublisherBerkeley Electronic Press
Place of publication805 Camelia St, Second Floor, Berkeley, USA, Ca, 94710
Rights statementCopyright 2019 Walter de Gruyter GmbH, Berlin/Boston