Cointegration in Panel Data with Breaks and Cross-Section Dependence

Abstract

The power of standard panel cointegration statistics may be affected by misspecification errors if proper account is not taken of the presence of structural breaks in the data. We propose modifications to allow for one structural break when testing the null hypothesis of no cointegration that retain good properties in terms of empirical size and power. Response surfaces to approximate the finite sample moments that are required to implement the statistics are provided. Since panel cointegration statistics rely on the assumption of cross-section independence, a generalisation of the tests to the common factor framework is carried out in order to allow for dependence among the units of the panel.