A0728
Title: Cointegrating multivariate polynomial regressions: Fully modified OLS estimation and inference
Authors: Oliver Stypka - TU Dortmund (Germany) [presenting]
Martin Wagner - University of Klagenfurt, Bank of Slovenia and Institute for Advanced Studies, Vienna (Austria)
Abstract: A fully modified OLS (FM-OLS) estimator is developed for cointegrating multivariate polynomial regressions, i.e., regressions that include as explanatory variables deterministic variables, integrated processes and products of non-negative integer powers of the integrated processes. The stationary errors are allowed to be serially correlated and the regressors are allowed to be endogenous. The FM-OLS estimator is extended from cointegrating polynomial regressions to cointegrating multivariate polynomial regressions, with the difference being the inclusion of cross-products of powers of the integrated processes, which overcomes the additive separability restriction typically used in nonlinear cointegration analysis for the polynomial case. The FM-OLS estimator has a zero-mean Gaussian mixture limiting distribution that allows for standard asymptotic inference. In addition to hypothesis testing on the parameters also Wald and LM specification tests are derived, as well as a KPSS-type test for cointegration. The theoretical analysis is complemented by a simulation study. Since the developed estimator immediately leads to a RESET-type specification test, we also compare in the simulation section the performance of our FM-OLS RESET test with the more restrictive RESET-type tests as well as the integrated modified OLS.
