Structural Break: Gregory-Hansen Cointegration Test in Eviews

[@omarahmed8957]

thank you brother for your interesting video. otherwise when I execute the 3rd model with regime shift, it displays this message " Near singular matrix error. Regressors may be perfectly collinear in "DO_ GHC.LS Y C (@TREND>4-2) G". ". please help me.

[@ChekwubeMadichie]

Thanks for your kind comment. If you have the Gregory-Hansen program code, the only place you need to edit is at the top. You can only remove X and add your independent variable(s), then remove Y and add your dependent variable. For the model 4 (regime shift), you may need a relatively large sample size. When you have more than 3 independent variables, you also need a relatively large sample size. This is because the regime shift assumes break in the entire coefficient vectors, meaning that you will need dummy variables for all and this will definitely extend the the of parameters.

[@omarahmed8957]

@@ChekwubeMadichie brother, what is the decision criterion for rejecting or accepting the null hypothesis?

[@ChekwubeMadichie]

@@omarahmed8957 see Gregory and Hansen (1996a & 1996b) for the critical values. We reject the null hypothesis if the modified ADF and Phillip statistics are less than the critical values. Good luck!

[@omarahmed8957]

@@ChekwubeMadichie thank you very much bro.

[@smsm314]

Hello my Professor, Thank you for this research.

[@ChekwubeMadichie]

Anytime brother!

[@mancobamndzebele6050]

Thank You for the informative video. I am also estimating using the Gregory-Hansen procedure. What is a setback on my side is that I do not have the Gregory-Hansen program and I am unable to download it. Could you please direct me to the link where I can download the program. I am using Eviews 9. Thank You.

[@arfaibtissem8830]

thank you for the video. my question is what are the conditions to use gregory hansen test , also do we use dummy variables while running the test ? and how to choose models ? thank you!

[@firûzeeeeeee]

Do you know how to do Maki cointegration test and Carrion-i-Silvestre Test? Or do you have the code of Maki and Carrion-i-Silvestre tests for eviews?

[@firûzeeeeeee]

Hello, If we don't have a program code for Gregory Hansen, how do we do it? Group statistics have the Johansen Cointegration test but there is no Gregory Hansen test. In this case, how can I get the Gregory Hansen test? Thank you already.

[@ChekwubeMadichie]

Thank you for your interest in my video. Kindly download the Gregory-Hansen program code from the link below forums.eviews.com/download/file.php?id=217&sid=55c202091b8249415268071b9f755250

[@firûzeeeeeee]

@@ChekwubeMadichie aa, thank you very much

[@ChekwubeMadichie]

@@firûzeeeeeee you're welcome

[@mancobamndzebele6050]

Thank you for the video sir. When I run the program code in Eviews 9 I get the following error message: Near singular matrix error. Regressors may be perfectly collinear in "DO_ GHC.LS Y C @TREND G (@TREND>20-2)". What should I do to fix this? Thanking you already.

[@ChekwubeMadichie]

You maybe working with the wrong gh_prg code.

[@dilpreetdhillon4725]

Hello sir, your video was very informative. kindly clear my doubt that in case of presence of structural break in our time series data, if our variables are I(1) , do we have to use Gregory-Hansen cointegration only or we can use Johansen Cointegration also by introducing dummy variable in the model?

[@ChekwubeMadichie]

You can use GH because it assumes all variables are I(1) but it's for a single equation setting just like the Engle-Granger residual-based. However, there is currently a program code for implementing johansen in the presence of break.

[@snowdeinpessey926]

Hi sir, thank you very much your video. Please is it possible to share the Gregory Hanson Program for the analysis .

[@ChekwubeMadichie]

Send me email at [e-mail]

[@harpreetkaur-bo4ko]

Thank you sir for this video. I am interested to find cointegration between 2 time series . since group of independent variable is not feasible then how can i run this model with 2 variable. Kindly suggest modification in need to do.

[@ChekwubeMadichie]

Dear harpreet, you can still run this model on 2 variables. All you have to do is click on "Object" in your workfile, then select "New Object". Click on "Group" and go to "Name for object" at the top-right and rename "Untitled" as "Independents" and click OK to open a new spreadsheet. Enter the only independent variable (with data) into the spreadsheet and click on "Edit +/-" icon to lock it, and then close the spreadsheet. A "group" object named "independents" will be added to your workfile. You can go on from there. I hope this helps.

[@harpreetkaur-bo4ko]

Thank you for prompt reply, sir. Problem related to creation of independents variable resolved as per your suggestion but when I run this model a error msg pop ups ( RES is not defined in "Delete res level ghtest break min_t_lag" on line 125)

[@manishtomar426]

how we can find the critical values for the same like we can do in stata ? you have also told us about critical value at the end of the video but it was not clear. With Critical values I meant to say Critical Values at 1%, 5% and 10%

[@ChekwubeMadichie]

Dear Manish, kindly refer to Gregory and Hansen (1996a & 1996b) for the critical values. I will advise you read the paper thoroughly to gain useful theoretical insights.

[@manishtomar426]

Respected Sir, I very glad that you replied. As you had asked my to go through Gregory and Hansen (1996a & 1996b), I have read the paper and the problem that is still rising that I am not been able to run the this test on EViews and even if I am, than I am not been able to find the values for Critical Values at 1%, 5% and 10% . Although I have understood the theory very well after reading the paper. I think that might be because I am new to this language. Hence you are requested to help in any way you can by helping me to find the Critical Values at 1%, 5% and 10% for my research work. Looking for a positive response from your side. Thank you

[@ChekwubeMadichie]

@@manishtomar426 Kindly refer to Table I titled "Approximate Asymptotic Critical Values" in page 109 of: Gregory, A. W. & Hansen, B. E. (1996). Residual-based cointegration tests for models in regime shift. Journal of Econometrics, 70: 99-126.

[@soumikdey645]

Hi @@manishtomar426 r You from India?

[@manishtomar426]

@@soumikdey645 yes sir

[@halilerdemerdem7571]

I can use GH for m=2,3,4 but i need to use m=1 which means one dependent variable-one independent variable. Please help me, thank you.

[@ChekwubeMadichie]

You can create group "independents" and have only one independent variable in it and still apply the same steps I used in the video.

[@zulfiqaralikhilji4736]

Can we apply gregory hansen test on more than 4 independent variables?

[@ChekwubeMadichie]

Yes, you can.

[@zulfiqaralikhilji4736]

@@ChekwubeMadichie well stata command of this test clearly states that it can't be applied on more than four independent variables.

[@ChekwubeMadichie]

This is Eviews and not STATA. Just follow the steps provided in the video. There is no limit to the number of independent variables to be added. Please read Gregory and Hansen (1996a & 1996b) for the theoretical model building. You will find that there is no limit to the number of X's variables.

[@tariqrahim1129]

Dear, Thank you for putting in a great effort and making such a difficult task so easy. While doing it I might a little bit of help, in regards to which I had already sent you an email. Looking forward to hearing from you.

[@ChekwubeMadichie]

I will attend to your email shortly. Thanks for your kind review.

[@soumikdey645]

Hello sir...if possible would you help in doing the analysis. As because I am using the Student version it says ...Program can't be done is student version. If possible can you analyse my data...of one dependent and 2 independent variables..I am stuck bcz of using the Student version of Eviews

[@ChekwubeMadichie]

Send a dm to my email: [e-mail]

[@soumikdey846]

@@ChekwubeMadichie Sir, I am send an Email

[@celebritiesclub534]

I m not able to copy firstly dependent and after independent from eviews when I m trying to copy variables resid is also copy along my variables plzzzz help me ????

[@rafeequeahmed7282]

thank you, brother, for such a nice contribution, But when I run the code, it gives the following error. Near singular matrix error. Regressors may be perfectly collinear in "DO_ GHC.LS Y C (@TREND>8-2) G". ". please help me.

[@ChekwubeMadichie]

Dear Raja, near singular matrix may mean that you're working with a relatively small sample size but have too many independent variables. What's the sample size? How many regressors? Which model are you estimating (level shift only?, level shift & trend? or regime shift?), which one? Note that estimating regime shift would require additional variables up to the number of regressors via the interaction of dummy with regressors. Thus, if the sample size is small and there are more regressors, there is likely going to be near singular matrix error and regressors may be perfectly collinear.

[@rafeequeahmed7282]

Dear Chekwube, the sample size is 53 and working with yearly data from 1965 to 2017. There are 5 independent regressors with 1 dependent regressor and I am trying to find the regime shift model. For the regime shift model as you are mentioning above that, I should use dummy variables. But I have a bit of confusion to select the dummy variables. Do I need to select dummy as zeros or ones for the regressors?

[@ChekwubeMadichie]

@@rafeequeahmed7282 The dummy variable is automatically added by GH after determining the breakpoint. You can also create a dummy by yourself, that is, if the break year is 1980, you create a dummy with 0 for 1965-1979 and 1 for 1980-2017. Just watch my video very well to be able to replicate what I did.

[@fatima2374]

I use EVIEWS 10 sir but I could not find the test. Can u help me where is the problem pls. Nd thanks so much for ur videos. All best to you from Algeria 🇩🇿.

[@ChekwubeMadichie]

To use any Eviews add-in, you have to connect your computer to the internet. Then go to Eviews add-in and click on the download add-in. It will navigate you to the website where you can find all the available Eviews add-ins. You can connect with me via email @ [e-mail]

[@owendeboer5525]

@@ChekwubeMadichie Dear Sir, I followed your instructions, but there appears to be no add-in for the Gregory Hansen cointegration test in EViews. I looked on the website of EViews as well and I could not find it. Would you perhaps have any other suggestions?

[@carlitoselloco5723]

Hello, Very much thank you for your great video. This eviews prg file allow for the inclusion of exogenous variables (dummies for crisis, for example)

[@ChekwubeMadichie]

Crisis is one of the causes of break in time series and the GH test detects break endogenously. That is to say that the analyst is non-informative about the breakpoint so that the informal data analysis wouldn't contaminate the choice of breakpoint. In order words, you don't need to add the dummy by yourself since you are assumed not to have knowledge of the breakpoint. If the crisis had caused a significant break in the data, the GH test would detect it and automatically add dummy to capture it effect.

[@carlitoselloco5723]

@@ChekwubeMadichie Very much thank you for your explanation. I was considering this test to check for breaks in the VECM and in case of breaks, to use the Johansen et al (2000) procedure to deal with breaks in the cointegration relationship. However, from your explanation, I wonder if by endogenously adding dummy variables the Gregory Hansen test always "properly " addresses the break in the VECM. Can you share your thoughts about this?

[@ChekwubeMadichie]

@@carlitoselloco5723 recall that the GH test is a residual-based test with the Engle-Granger test as a special subcase. Thus, the GH test is suitable in a single equation setting just like the Engle-Granger test. The VECM is a system of equations system and thus follows the Johansen system cointegration setting. However there is a program code for the Johansen cointegration test in the face of structural break. In addition, if the GH test was able to detect a break point, you can then use dummy to implement the VECM.

[@abdelatisaeid6240]

Thank you for your video. how can i get Gregory-Hansan program brother ?

[@ChekwubeMadichie]

Thank you for your kind comment. Kindly download the Gregory-Hansen program code from the link below forums.eviews.com/download/file.php?id=217&sid=55c202091b8249415268071b9f755250

[@yasemincoban9049]

hi can I give gh program please help

[@nirmalabhatt1221]

If you have code of Geogory and hansen plz send me link..

[@chiobichukwuebuka4082]

Zivot and Andrew