A0921
Title: Integrated modified OLS estimation and inference in I(2) cointegrating regressions
Authors: Sebastian Veldhuis - University of Klagenfurt (Austria) [presenting]
Martin Wagner - University of Klagenfurt, Bank of Slovenia and Institute for Advanced Studies, Vienna (Austria)
Abstract: Integrated modified OLS (IM-OLS), as developed in a past study, is extended for the I(1) cointegrating regressions, to cointegrating regressions with unknown mixtures of I(1) and I(2) processes as regressors. The IM-OLS estimator is extended to this setting and it is shown that its limiting distribution is mixed normal, leading to standard asymptotic inference. In addition to standard asymptotic theory, fixed-b inference is also developed and fixed-b critical values are provided for a large set of specifications, kernels and bandwidths. The setting is closely related to the so-called residual-based fully modified OLS (RBFM-OLS) estimator developed in a past study. The performance of OLS, RBFM-OLS and IM-OLS are compared by means of a simulation study. Finally, the estimator is used to test the nominal-to-real transformation for long-run money demand data.