Dynamic OLS + Restricted Trend

[Gruenspan]

In Eviews 7 you can estimate a cointegrating regression with dynamic OLS (DOLS). It is also possible to include a (restricted) trend in case that the deterministic trends do not cancel.
My problem is that Eviews does not report the estimated trend coefficient. Does anybody know how to find it? Thanks in advance for any help on this.

[EViews Glenn]

I'm not certain that I understand the question. But let me take a crack at an answer anyway :)

My understanding is that the presumption in the specification that you propose is that there is no trend coefficient in the cointegrating equation, thus there is no trend coefficient to estimate (and thus no way of retrieving a trend coefficient). One simply augments the proposed specification with the appropriate leads and lags and off you go. In some sense, the presence of the extra trend term in the regressors portion of the triangular system is irrelevant since the presumption is that the cointegrating equation is correctly specified.

Note that in the FMOLS and CCR specifications, the presence of the extra trend term does matter since it is used to obtain the system residuals. In the DOLS, it's irrelevant since the estimation is performed in one step using the augmented cointegrating equation specification.

I hope that this answers your question.

[Gruenspan]

Not exactly, if I understood your answer correctly.

I know that the base specification does not include a trend. However, I estimated also a C-VAR model with Johansen MLE. One of the vectors I am most interested in contains a restricted linear trend in my final model. To check the robustness of these results I want to match the econometric model I estimate as closely as possible, which means I also include a restricted trend when estimating with D-OLS. If I do so the estimation output shows me all estimated coefficients apart from the one for the trend. However, I do also need that estimate.

Thanks for your help!

[EViews Glenn]

The Johansen system estimator is just that, a system estimator which estimates based on a likelihood using all of the equations in the system. As a consequence, information about the restricted trend comes naturally from the estimation of the complete system.

The DOLS estimator (and its "friends" the CCR and FMOLS) are single equation estimators which use information about the rest of the system in a restricted fashion. The CCR and FMOLS estimators use the restricted trend specification in a first step, then estimate the cointegrating equation after adjustments to obtain an orthogonality condition. The DOLS simply uses the orthogonality condition for an augmented cointegrating regression.

This is a drawback of the single equation methods. I can't really think of an easy way to obtain the equivalent coefficient. Perhaps someone else might be able to shed some additional light on this.