Sequential monitoring of structural changes for cointegrated vector autoregressive models

Abstract

This study introduces a sequential monitoring method for detecting structural changes in cointegrated vector autoregressive (VAR) time series models. We present a novel approach that utilizes the location-scale cumulative sum (LSCUSUM) technique, which incorporates two key CUSUM processes: one for detecting changes in the vector autoregressive parameters, and another for identifying shifts in the covariance of the error terms. The method dynamically updates with each new observation and follows a closed-ended procedure, accounting for the unstable nature of the data. It either detects structural breaks or concludes at predefined endpoints. Through simulations and real-world case studies, we demonstrate the effectiveness of this approach in identifying change points in cointegrated time series data.