intercept/trend in DOLS

[unknown]

Hi,

After installing the lattest eviews 7.2 update (released on february, 28), it seems that the DOLS procedure doesn't allow for constant or trend in the space of deterministic regressors. Indeed, the estimation works only when I select "none" in the trend specification, while in any other cases I got an error message of "invalid specification". Thanks for your support.

[EViews Glenn]

We've become uncomfortable with the added cointegrating equation trend assumptions that the restricted specification would entail so we've disabled it. The induced trend terms in the cointegrating equation would have to be accounted for which we currently don't handle in a general fashion. If you want to allow for trends in the regressors, you'll have to include the trend in the main specification.

But before you despair, I will note that these qualifications are for "added trend" specs only (those that are in the regressors but not in the cointegrating equation). That is, if you already have a trend in the cointegrating equation then that trend will eliminate the equivalent trend in the regressor specification. It's only if you *don't* have a trend in the cointegrating equation and have one in the regressors that there's an issue. So if you have a trend in the cointegrating equation, you're covered since the presence of the trend terms in the specification orthogonalizes the regressors.

I hope this answers your question. Let us know if you need further clarification.

[unknown]

Dear Glenn,

Many thanks for your reply. I understand the new framework you have implemented regarding DOLS procedures. By "main specification", I guess you mean the first blank line in the top of the DOLS windows "equation specification". I have already tried to add a constant in the cointegrating space, by putting a "c" there, but if the estimation was done by EViews, I could not reach the coefficient associated with the constant term since it did not appear in the estimation output. I suppose that the coefficient it's just removed when making the estimation since results are exactly the same whereas I add c or not in the equation specification.

In addition, if the "additional trend" menu is indeed locked as you said, it is not the case for the "trend specification" menu in the first part of the windows. But here, I always get an error message unless I select "none". When I try to put a constant in the following line "deterministic regressors", EViews suggests me to use the trend specification menu.

Thanks.

[EViews Glenn]

Think it's a bug. Let me check. [edit] Yes it is. The check for specification (which we put in for safety) was looking at the total regressor specification not the excess specification. This relates to the discussion above. I've put the fix in and it'll appear in an update shortly.

Sorry for the inconvenience.

[unknown]

Thanks for the patch. Unfortunatly, it seems that it does not fix the error that disabled the first trend specification in the DOLS estimation since it produce the same error message. Hence, there is still no way to add a constant or trend in the set of cointregrating regressor. Thank you for you help.

[EViews Glenn]

I'll check on it. [edit] Seems to be a problem with the patch. Please download again. Sorry for the trouble.

[lex]

Hi there, I am encountering the exactly same issue, and suspecting that the bug has not been fixed (the constant has been removed automatically). Would you please advise? FYI My version is 7.2. Thank you in advance.

[EViews Gareth]

What is the build date of your copy? (Help->About EViews)

[lex]

Feb 24 2012. Would it matter? thx

[EViews Gareth]

Um, you suspect that "the bug may not be fixed", yet, as stated in this thread, the bug was fixed in March, yet you are using a version from February.

[lex]

Understood, thx