EViews.com

This thread is about the TARCOINT add-in, written by Eren Ocakverdi (Trubador), that performs the Enders and Siklos 2001 Cointegration and Threshold Adjustment procedure.

Professor Siklos, one of the authors of the article "Cointegration and Threshold Adjustment", noted in an email exchange that using an exogenous threshold series may have uncertain statistical implications and therefore threshold series should be derived from the underlying equilibrium relationship. He also suggested that, it would be better to add any exogenous type variables in the ECM portion of the model for the completeness of estimation.

I revised the add-in in light of Professor Siklos' recommendations:
1) Removed the option to supply exogenous threshold series.
2) Added the option to enter regressors into the equilibrium equation.

trubador wrote:Professor Siklos, one of the authors of the article "Cointegration and Threshold Adjustment", noted in an email exchange that using an exogenous threshold series may have uncertain statistical implications and therefore threshold series should be derived from the underlying equilibrium relationship. He also suggested that, it would be better to add any exogenous type variables in the ECM portion of the model for the completeness of estimation.

I revised the add-in in light of Professor Siklos' recommendations:
1) Removed the option to supply exogenous threshold series.
2) Added the option to enter regressors into the equilibrium equation.

Trubador, this add-in is really helpful, thanks so much.
is it possible to update the add-in by replacing the current one by the corrected one? I mean the one that include 2) option to enter regressors into the quilibrium equation.
Thanks.

You already have access to this revised version. Just download and install it from the link given in the very first post of this thread...

A minor update for the add-in is available (and is already uploaded):

1) You can now adjust delta values in generating random data for endogenous variables. These are simply the coefficients of lagged changes used during simulation, where the lag length is exogenously or endogenously determined. For details, please refer to Equations 11' and 12' in the original study.

Hi,

I try to use TARCOINT add-in and I would like to ask you one thing:

You need to specify 2 endogenous variables but I would like to make sure that I estimate yt= a + xt + et and not xt = a + yt + et...

Is the first endogenous variable the dependent variable?

Thanks a lot!

Steph wrote:Is the first endogenous variable the dependent variable?

Yes, that is correct.

Thanks a lot for your prompt response.

I am sorry to bother you but I have another question. I would like to add a trend in the long run relationship but I am not sure we can add exogenous variable in the model. Is it possible?

Thanks in advance for your help!

Steph

Yes, it is possible. Just create a seperate series for the trend variable (e.g. series tr = @trend) and then enter the name of this variable into the relevant dialog box called "Exogenous variables". Please see the documentation (Add-ins/Manage Add-ins/Docs) for more details.

I don't know what I did wrong when I installed the package but I don't have a box called "exogenous variables" but I have 2 boxes called 'endogenous variables". When I enter the trend in the 2nd box called "endogenous variables", I get an output table where no exogenous variable have been included.

Do you know if anybody else has ever had this problem?

Thanks for your help.

It is the second box. I guess there was a typo, when you first downloaded the add-in. You can download the latest update using same link, which is given in the very first post of this thread. Other than that, as it is mentioned in the second post, any exogenous variable will be influential in the ECM model only.

Thanks!

if the threshold is more than one ,how to estimate it ?

Hello,

I've the same problem than Steph, when i enter an exogenous variable in the 2nd box called "endogenous variables", the output window doesn't contain the exogenous variable.

I've downloaded the version of the first post like you recommend.

Is there a problem with the program?

Thanks.

PS: I use eviews 7

As I mentioned in the second post and reiterated in my response; "any exogenous variable will be influential in the ECM model only". The technique operates on residuals and therefore the output shows only the relevant variables used in those equations. For more details, please refer to the original study...