Cointegration question

[yoryiecn]

Hi everybody!
I am a newbie in the time series analysis.. Well.. i appreciate your help in this topic.

When i make a Johansen's cointegration test and the trace test indicates one (or more) cointegration relationships AND the Eigen. Max indicates no cointegration in ALL MODELS what is supposed that is the conclusion? these time series are cointegrated or not?

thanks !

Regards from Dominican Republic.

[sakamuk]

Hi

I have faced this problem and had to make a decision to opt for one. This was my problem. There is no clear-cut preference for either test. There are academics on both sides of the fence. Enders (1995) opts for the maximum eigenvalue approach check p393 something about this test having a “sharper alternative hypothesis”. Kennedy (2008) also says it is superior but Lüutkepohl, Saikkonen, and Trenkler (2001) argues otherwise and opts for trace statistic approach. You may want to look at these sources to help in making your decision.

Enders, W. (1995). Applied Econometric Time Series (First Edition ed.). New York: John Wiley & Sons.
Kennedy, P. (2008). A Guide to Econometrics (6 ed.). 6th edition, Cambridge, MA: Wiley-Blackwell.
Lüutkepohl, H., Saikkonen, P., & Trenkler, C. (2001). Maximum eigenvalue versus trace tests for the cointegrating rank of a VAR process. Econometrics Journal, 4(2), 287-310.

Best of luck making the decision just give a convincing argument as to why you choose one over another.