Cointegrated Continuous-time Linear State Space and MCARMA Models

TL;DR

This paper characterizes cointegrated continuous-time linear state-space models, showing their equivalence to cointegrated MCARMA processes and deriving conditions for cointegration, including an error correction form for discrete observations.

Contribution

It introduces a comprehensive framework for cointegrated continuous-time models, linking state-space and MCARMA models, and extends cointegration conditions to this setting.

Findings

Cointegrated continuous-time models can be represented as a sum of a Lévy process and a stationary component.

Cointegrated MCARMA processes are equivalent to cointegrated continuous-time linear state-space models.

Derived necessary and sufficient conditions for cointegration based on autoregressive polynomial.

Abstract

In this paper we define and characterize cointegrated continuous-time linear state-space models. A main result is that a cointegrated continuous-time linear state-space model can be represented as a sum of a L\'evy process and a stationary linear state-space model. Moreover, we prove that the class of cointegrated multivariate L\'evy-driven autoregressive moving-average (MCARMA) processes, the continuous-time analogues of the classical vector ARMA processes, is equivalent to the class of cointegrated continuous-time linear state space models. Necessary and sufficient conditions for MCARMA processes to be cointegrated are given as well extending the results of Comte for MCAR processes. The conditions depend on the autoregressive polynomial. Finally, we investigate cointegrated continuous-time linear state-space models observed on a discrete time-grid and derive an error correction form for this model. The error correction form is based on an infinite linear filter in contrast to the finite linear filter for VAR models.