EViews.com

I am very new to VEC models. But when I run a test on one of my cointegration vectors I seem to get weird degrees of freedom. Shouldn't the degrees of freedom for when 'Hypothesised No. of CE(s)' =2 be 5 instead of 17? [see picture] I keep getting similar results when I test other restrictions. Any help is appreciated; I feel like I am making a very dumb error/ oversight. Thanks.

Would it be possible for you to post your workfile, along with the steps you took to get that (odd looking) result?

I just use the 'Quick-> Estimate VAR' function. Choose VEC. Use all my variables (yt pt it et yt_ pt_ it_) for the endogenous variables. Keep it at '1 2' for lags. For 'Cointegrating' I use rank 2 and option 3. I don't enter any restrictions in this window and then click 'Ok'. I then do 'View->Cointegration Test' I choose 'Osterwald-Lenum' and for the restriction I enter 'B(1,1)= 0, B(1,2)=0, B(1,3)=1, B(1,4)=0, B(1,5)= 0, B(1,6)=0, B(1,7)=0' I attached the workfile also. Thanks for your help!

The odd looking result of dof=17 is coming out of code where we need to numerically determine the rank of the restrictions and there's some very unpleasant scaling in the results from your estimation in the coint rank=2 case which are throwing things off.

I'd probably say that the whole restricted estimation is actually kind of suspect in that case.

I'm not sure if we could do better, but I think this is probably happening because you are trying to do some slightly strange things:

- Most of your series appear to be I(0) not I(1) based on unit root tests. It doesn't make sense to put I(0) series into a cointegrating vector since cointegration is looking for stationary combinations of non-stationary variables. If the variables are already stationary, any combination of them is already stationary. This is probably why you're getting such a large number of cointegrating vectors showing up in cointegration rank tests.

- The restrictions you specified are kind of odd too because they are constraining a cointegrating vector to contain a single variable, which is a sort of rounabout way of saying that you already know that the single variable is stationary. If this is the case, that variable should really be dropped from the cointegrating vector to begin with.

So I agree that we're behaving a little oddly, but you probably need to rethink some of what you're trying to do here.

Okay, thanks... I will go back and take a closer look at my methodology I guess. I really appreciate your help and explanation of the issue.