Testing multiple equation systems for common nonlinear components (Q1379913)

The paper considers a multivariate nonlinear system of regression equations which contains common nonlinear components. A generalized method of moments (GMM) test for common components in multiple time series is provided. The test statistic is derived in terms of canonical correlations between the series and a judicially chosen set of test regressors. The test performance for small samples is evaluated using Monte Carlo simulations. The authors show how to adapt and use the GMM test to investigate if nonlinear components in multivariate systems are common, and illustrate the usefulness of the proposed test with two applications for US and Canadian business cycles and for a single nonlinear factor model as part of an empirical real business cycle model.