Self-Normalization Approach to Hypothesis Testing on Cointegrating Vectors

This study investigates the properties of a self-normalization inferential method in cointegrating regressions. The test statistic is constructed using recursive integrated modified ordinary least squares estimators. The proposed test does not involve a long-run variance estimator, and the test statistic has a null limit distribution free of nuisance parameters. A simulation experiment shows that the proposed test has more accurate size in finite samples than other methods