Cointegration question

[lbkova]

Hello All,

Could you please tell what is the difference in using Engle-Granger and Johansen method for testing cointegration? I want to estimate if there are cointegration relations among let's say 3 stock indices (A, B, C). Now what if I find cointegrations between A-B and B-C but not between A-C using Engle-Granger method but I find only 1 cointegrating equation among the 3 variables using Johansen? Which result shall I accept as correct? Also given the limitations of EG method can I actually use Johansen to test cointegration between only 2 variables?

I hope I was understandable...

Best regards,

Laszlo

[nayan]

Johansen cointegration is recommended because it allows us to test number of (more than one) cointegrating relations among a set of more than two variables (Hamilton, 1994 p. 630).

Hope that u got the answer.