GVAR: A Case of Spurious Cross-Sectional Cointegration

Global Vector Autoregressive models came to be used quite widely in empirical studies using macroeconomic non-stationary panel data for the global economy. In this paper, it is shown that when the loading matrix of the cointegrating vectors is not block-diagonal and the cross-sectional spillovers of disequilibrium exist, the use of the GVAR model leads to spurious cross-sectional long-run relationships. Moreover, the results of Monte Carlo simulation show that the GVAR model is outperformed by other valid econometric approaches in terms of the maximum likelihood estimator of long-run coefficients, when the cointegrating vectors matrix is block-diagonal.