On the Power of Cointegration Tests: Dimension Invariance vs. Common Factors

This paper considers the trade-off, for cointegration tests, between dimension and power: that is, we compare the power performance of test-statistics which are dimension-invariant but impose common-factor restrictions with tests which are not dimension free but do not impose those restrictions. As a by-product of the analysis, we consider cases where the t-ratio form of the tests have better power properties than the coefficient form, in spite of the latter diverging at rate O (T) and the former at O (T ), under the alternative hypothesis of cointegration.