Multivariate Markov-switching score-driven models: An application to the global crude oil market Articles uri icon

publication date

  • June 2022

issue

  • 3

volume

  • 26

International Standard Serial Number (ISSN)

  • 1081-1826

Electronic International Standard Serial Number (EISSN)

  • 1558-3708

abstract

  • A new class of multivariate nonlinear quasi-vector autoregressive (QVAR) models is introduced. It is a Markov switching score-driven model with stochastic seasonality for the multivariate t-distribution (MS-Seasonal-t-QVAR). As an extension, we allow for the possibility of having common-trends and nonlinear co-integration. Score-driven nonlinear updates of local level and seasonality are used, which are robust to outliers within each regime. We show that VAR integrated moving average (VARIMA) type filters are special cases of QVAR filters. Using exclusion, sign, and elasticity identification restrictions in MS-Seasonal-t-QVAR with common-trends, we provide short-run and long-run impulse response functions for the global crude oil market.

subjects

  • Economics

keywords

  • global crude oil market; markov regime-switching models; nonlinear co-integration; score-driven models; structural changes