Phillips-Ouliaris test for cointegration

[oksanakim]

Hi guys

I am looking for assistance with the Phiilips-Ouilaris test for cointegration of two series. I selected the series as a group, then specified the P-O test option, and here is the output:

Date: 12/09/12 Time: 10:56
Series: RETHOME RETGDR
Sample (adjusted): 1 7258
Included observations: 7258 after adjustments
Null hypothesis: Series are not cointegrated
Cointegrating equation deterministics: C
Automatic lags specification based on Schwarz criterion (maxlag=35)


Dependent tau-statistic Prob.* z-statistic Prob.*
RETHOME -56.21539 0.0001 -11205.10 0.0000
RETGDR -34.69869 0.0000 -13145.23 0.0000

*MacKinnon (1996) p-values.


My question is, how can I interpret the output? Specifically, why does EViews return outputs for two variables, while I am effectively testing for cointegration - moving of two variables as one - and hence expect one output for the combination of these two series? In this case, I reject the null for both rethome and retgdr variables and I assume, I am simply rejecting the null that the two series are not cointegrated, correct? What if the test rejected the null for rethome but did not reject the null for the retgdr? How would I interpret the overall output in this case?

Thanks a lot!

Oksana

[EViews Glenn]

You get a different test result depending on which is considered to be the dependent variable in the long-run regression. Much as in you will get different results in an Engle-Granger framework depending on which you treat as the dependent variable. This is one of the downsides of the single equation methodology.

[oksanakim]

Thank you for helping with the test. So, in case the first line of the output rejects the null and the second line of the output does not reject the null, how shall I interpret the results? Shall I say, series are not cointegrated? Or shall I say, one series lead the relationship and the other series adjusts to it, but not vice versa?

Thanks again!

Oksana

[EViews Glenn]

That's a problem, as you might imagine. There's no right or wrong answer to this one. All it suggests is that the evidence for and against the null is ambiguous.