Stationary bootstrapping for cointegrating regressions

Abstract

The validity of stationary bootstrapping is investigated for cointegrating regressions in large samples as well as in finite samples. The bootstrap ordinary least squares estimator (OLSE) is shown to be valid in large samples having the same limiting distribution as the OLSE under a similar normalization. Large sample validity of a bootstrap test regarding cointegration parameters is also established. Finite sample size and power properties of the bootstrap test are investigated via a Monte Carlo experiment. © 2012.