Cointegration Testing in Single Error‐Correction Equations in the Presence of Linear Time Trends
指出Banerjee等人提出的误差修正检验在序列存在线性趋势时，若使用未去趋势的回归，其临界值不适用，并提供了正确的百分位数。
Banerjee, Dolado and Mestre (J. Time Ser. Anal. 19 (1998) 267−283)introduce an error‐correction test for the null hypothesis of no cointegration. The present paper supplements their work. They provide critical values forregressions with and without detrending. Here it is shown that the latter arenot appropriate if the series display linear trends. This does not mean thatdetrending is required. Correct percentiles are suggested for the case thatseries follow linear time trends but tests are based on regressions withoutdetrending. They are readily available from the literature.