EViews.com

Dear all,

I estimated a (Consistent)M-TAR model for two prices (and 5 exogenous variables, one shift dummy and four impulses) and results indicate the existence af asymmetries and cointegration between the two series. Now, I'd like to estimate the relative AECM in order to investigate short run dynamics and test the significance of the zplus and zminus (with a Wald test), but I don't know how to generate positive and negative ECT. Anyway, I tried with the following code:

GENR DRES=D(RESID) (First difference of residuals from the cointegration regression)
GENR MTAR=DRES(-1)>=-0.001863 (Here, the value is the estimated threshold value calculated by Eviews (Chan's method))
GENR ZPLUS=MTAR*RESID(-1) (Positive ECT)
GENR ZMINUS=(1-MTAR)*RESID(-1) (Negative ECT)

After that, I estimated my AECM by Least Square, in which the dependent was the first difference of my series and the independent the other one + zmins and zplus:

dps c dps(-1 to -3) dpp(-1 to -3) zplus zminus

And run a Wald test in which I tested the H: Zplus=Zminus.

I repeated the same process for the other series, in order to check for its exogeneity.

Is the process right?

The next question would be how to display IRFs for positive and negative shocks, in order to show the asymmetric adjustment.
I attached the workfile for if it could be of any help. "mtarcons" is the model to which I refer and ecmps the AECM for the dependent DPS-

Thanks in advance for any helpful comment,

Bests

Federico

Any suggestion?

I don't understand the question. The output generated by TARCOINT add-in already gives an "F-equal" statistic, which corresponds to test result of the hypothesis you mention.

In order to investigate short run dynamics of the model (therefore the IRFs) I need to build up the AECM, and the two ECT+ and ECT- have to be different, otherwise there is no asymmetric adjustment, right?

If it could be of any help, the main reference is the work of Abdulai (2002) "Using threshold cointegration to estimate asymmetric price transmission in the Swiss pork market," Applied Economics, Taylor & Francis Journals, vol. 34(6), pages 679-687.
In his work, after assure the existence of cointegration and asymmetries through a TAR and M-TAR estimation, he built the relative AECM, checked the significance of the two ECT terms and compute the IRFs.

I hope this may help,

Thank you again,

Bests

FA

Testing of the asymmetry is and should be done prior to estimating the error correction function. Once you have found evidence in favor of asymmetric adjustment, then you go on to estimate AECM equation. In this step, coefficients of zplus and zminus are interpreted according to their expected behavior towards the long run equilibrium level. As for the IRF, I am afraid you'll have to compute it on your own as there are no readily available routines for that purpose.