Panel Error Correction Testing with Global Stochastic Trends

This paper considers a cointegrated panel data model with common factors. Starting from the triangular representation of the model as used by Bai et al. (2008) a Granger type representation theorem is derived. The conditional error correction representation is obtained, which is used as a basis for developing two new tests for the null hypothesis of noerror correction. The asymptotic distributions of the tests are shown to be free of nuisanceparameters, depending only on the number of non-stationary variables. However, the tests are not cross-sectionally independent, which makes pooling difficult. Nevertheless, the averages of the tests converge in distribution. This makes pooling possible in spite of the cross-sectional dependence. We investigate the nite sample performance of the proposed tests in a Monte Carlo experiment and compare them to the tests proposed by Westerlund (2007). We also present two empirical applications of the new tests.