Title: A fixed-b cointegration test for cointegrating polynomial regressions
Authors: Sebastian Veldhuis - University of Klagenfurt (Austria) [presenting]
Martin Wagner - University of Klagenfurt (Austria)
Abstract: A fixed-b inference is developed for a type of cointegration test for full design cointegrating polynomial regressions, which include, e.g., specifications used in the environmental Kuznets curve literature or Translog-type production and cost functions. Full design refers to a situation that allows reformulating the test statistic as a functional of standard Wiener processes. A detailed simulation study shows that fixed-b inference reduces size distortions significantly at the expense of only modest losses in size-corrected power, a typical feature of fixed-b inference in both stationary as well as unit root and cointegration settings. We illustrate the performance of the test with an application to the environmental Kuznets curve. A supplementary appendix provides critical values for a large number of specifications, a fine grid of bandwidths and several widely-used kernels.