(EViews10):Estimate Johansen Cointegration Test
Вставка
- Опубліковано 11 бер 2018
- This video shows you how to perform the Johansen cointegration test using EViews10. After performing stationarity test, there are three (3) likely outcomes: the series may turn out to be I(0), I(1) or a combination of both. So what do you do next? This hands-on tutorial shows you what to do in EViews10 when series are I(1), that is, first difference-stationary series.
Here is the link to the ex21-1.wf1 dataset (EViews file) used for this tutorial (endeavour to have a Google account for easy accessibility): drive.google.com/drive/u/1/fo...
Follow up with soft-notes and updates from CrunchEconometrix:
Website: cruncheconometrix.com.ng
Blog: cruncheconometrix.blogspot.co...
Forum: cruncheconometrix.com.ng/blog/...
Facebook: / cruncheconometrix
UA-cam Custom URL: / cruncheconometrix
Stata Videos Playlist: • (Stata13):Estimate and...
EViews Videos Playlist: • (EViews10):Interpret V...
UA-cam recently changed the way my content will be monetised. My channel now needs 1,000 subscribers. So it would be amazing if you show your support by both watching my videos and subscribing to my channel if you haven’t done so already. Monetising my videos allows me to invest back into the channel with some new equipment so this small gesture from you will be extremely huge for me. Many thanks for your support….CrunchEconometrix loves to teach, help me stay online.
Thank you dear. i am Econometrician from Sudan. I do support you because what you have posted is 100% refined and solid with almost zero error. All interested Students, specially from Africa are encouraged to follow your tutorials, for sure, i will let my students joining your group of the followers around the globe. I have many friends from all Africa, i will be proud to be one of them.
Thanks Mohd for the encouragement and show of support. Please encourage your students and friends to subscribe to my UA-cam Channel. My videos are tailored for beginners and intermediate-level users. Thanks!!!
From Zambia here, thank you for your videos, really helpful 👏
Hi Chixy, thanks for the encouraging feedback. Deeply appreciated! Much love from Nigeria 🇳🇬!
Watching your video for the first time today got me really excited and interested in learning more econometrics. I've decided to watch every lesson you've ever thought online. Dr Adeleye you are a blessing, please keep doing this great job. THANK YOU
Thanks Elijah for the encouraging feedback and kind words about my UA-cam videos. Deeply appreciated!🙏❤️ Please may I know from where (location) you are reaching me?
Your content is mind blowingly good!!!
Glad you think so, Afra...thanks!
Thank you Dr, for this nice tutorial
U're welcome, Doc! Positive feedback is deeply appreciated!
Fantastic explanations. Thank you,
Thanks for the encouraging feedback, Dennis...deeply appreciated! May I know from where (location) you are reaching me?
Nice video. Well explained in simple terms. Thank you. Keep posting. Good luck and God bless you
My pleasure, Dhamod! Thanks for the encouraging feedback, deeply appreciated!
Thank you so much for your videos, you're a wonderful teacher ! 😊
Hi J, I'm humbled by your positive feedback and commendation. Deeply appreciated! Please may I know from where you are reaching me?
@@CrunchEconometrix From France. You provide very valuable information so it's only fair you receive praise. 👏
i think your video is very well constructed@!
f some variables are at level and some others are first difference so which can i combine at level or first difference pl help me
Madam, Its as if you are god sent! These tutorials are exactly my homework. Only explained in a manner that much more coherent than my lectures. i have learned more in your videos than I did in my uni course! God bless you!
Hahahaha, c'mon Gaucho don't make me blush more than necessary. I'm just trying to teach with simplicity and clarity for my viewers to understand basics. Thanks for watching my videos and kindly help me share too. Comments humbly taken...gracias! 💕 😊
@@CrunchEconometrixI have shared it with my classmates and I am sure word will get out! :)
I love ur tutorial.
Thanks for your encouraging feedback, deeply appreciated 🥰🙏
You are a Godsend
Thank you madam 💕🙏🏾
From Haïti 🇭🇹
Thanks, ChildofGod for the encouraging feedback. Deeply appreciated. Much love from Nigeria 💗
Brilliant content. Thanks a lot!!!!!
Glad you liked it! 🥰
Thank you very much for these simple and highly informative vedios on econometrics 😊😊
U're welcome! 🥰🙏
My first time learning such econometrics stuff but I'm getting the understanding so quick and easy. Thanks darling for the videos
Wow! Thanks for the kind words and positive feedback on my video, Bella. Deeply appreciated! 💕 😊 May I know from where (location) you are reaching me?
@@CrunchEconometrix Currently in China
@@bellabae3661 Awesome!!! I will appreciate if you subscribe to my Channel and also share my videos and link with your friends and academic community in China. They'll learn some useful skills and tips too...😀
Dr Adeleye, i am grateful to you for your videos. You explain in such a beautiful manner that even a layman can understand the technical aspects of the topic. Thank you so much for making such beautiful videos. These were a great help in taking a step forward in my Ph.D. I request you to please make videos on SVAR, TVP-SVAR, Trimmed mean too. These techniques are often used and barely any help is available on these topics. Thankyou so much for helping people like me.
Thanks for the encouraging feedback, Gopal. Deeply appreciated! Thanks for the suggestions. I included SVAR among the techniques to understand before creating simple video tutorials. But I've never heard of TVP-SVAR. What technique is that?
Dr. Adeleye, TVP-SVAR is time varying parameter-structural var model. The model is generally used to check the time varying properties of the coefficients e.g. say to Check if determinants of inflation are varying with passage of time. I hope I am able to explain it.
Thanks Gopal, for the heads up. I will read more about this technique for familiarity. Regards.
Wonderful explanation 🙏
Thanks, Muyiwa for the encouraging feedback. Deeply appreciated 🙏🥰
Thank you for the videos
You are welcome, Paul😊
I love your channel. I am from Nigeria.
Awesome! Thank you, Veronica...I'm encouraged!
I wasn't able to access the data set using the google drive, I love your videos they have helped me a lot with my studies I'm currently doing my Post Graduate Diploma in Applied Economics
Hi Lelethu, I deactivated the link to my Google drive. Kindly know that due to abuse and unethical conduct some datasets and all Stata dofiles used in my videos are no longer free but available on my website upon payment. Here's the link cruncheconometrix.com/view/datashop.php
Very helpful thank you sir
Thanks, Black Hero for the encouraging feedback. Deeply appreciated!
Love u Dr
Same here, Alief😊...thanks!
thank you very much
Glad it helped! 🥰
Again, I found this useful
Glad it was helpful, Isaac!
great content! i'm a student from Malaysia! keep up the good work madam!
Thank you, Fakhput! Will do!
Thank you !
You are welcome, Nina! 💞 😍
Hello, first thank you so much for your video! I have a question about cointegration test. I conducted multivariate cointegration test and found 1 cointegrating vector among five prices. In that case, can I say there is a long-run relationship between prices? How can I interpret the result if I get 1, 2, 3, or 4 cointegrating vectors? Thank you so much in advance.
thank you maaa💯👏👏👏👏👏👏
You are so welcome, Susan! 🙏🥰
Thanks maa'm
U're welcome, Ashomi. Please tell others too by sharing my videos! 💕 😊
Greetings.
Love your work, it is really helpful for my analysis. I have just one question.
If the data is not in logarithmic form, is the EViews tehnique the same, or do we need to change anything? I am researching the cointegration between nominal exchange rate and domestic and foreign prices, for the validity of PPP.
Best regards.
Hi Keyarslo, log transformation is at the discretion of the researcher. Not compulsory but advisable. I will advise you do more readings on its merits.
Hi! Thank you for helpful video tutorials. At around 02:45, you said that cointegration should be performed in either level or log-transformed series. Agreed. But should one use the original or the seasonally adjusted data series? Thanks a lot!
Hi Julie, the original series should be appropriate or you can confirm from other online resources. Thanks
Hi, I'm currently using Eview 13 and the selection criteria for deterministic trend assumption is different than what you have in this video. May I know which one I should select? (as in your video you choose no.3 which no longer available in Eviews 13)
I use EViews10.
Hey crunch! I was wondering in terms of no Co integration, would I estimate a var model with log values or first difference of raw data? Also, if no cointegration, would i be able to explore the system using granger causality tests? And could i further explore these results with impulse response function models in light of no cointegration?
Kind regards :)
Whether you use logged values is at your discretion. If there's no cointegration, estimate only the basic VAR model.
Thank you! I find the recording useful and will certainly utilize the knowledge obtained from it in my thesis. My thesis is about the short-run and the long-run effects of foreign direct investment (FDI) on economic growth in a set of countries using panel data (10 countries, 20 years of annual observations each). I was wondering how I can distinguish short and long-run relationships but it turns out that the VAR model would be an answer for the short-run and the VEC model for the long run. Even better - you've already published videos on those too - awesome!
I am slightly lost in all of those assumptions that need to be satisfied to conduct those tests. For instance, in order to use OLS, that are supposed to be used in my final linear regression model, I need to fulfill the OLS assumptions. I would certainly begin with conducting the ADF test to detect unit roots, to achieve stationarity. I am thinking about estimating the VIF model to detect multicollinearity too, but I have also read that differencing the data can help me get rid of the multicollinearity issue, by achieving stationarity. Beyond what I wrote above,
1) could you please clarify if while estimating the final linear regression model, is it a good or bad sign when the variables are cointegrated - or perhaps it has nothing to do with 'good' or 'bad' and merely implies the long or short-run relationship?
Thanks for the feedback. Appreciated! Your query is TOO long.
@@CrunchEconometrix Excuse me, madame. I have contacted you through the form on your official website. I wanted to request personal consultations with an expert so I can sort out a step-by-step process of my research. I haven't received any response yet, though sent a message at least four days ago. Thank you.
My sincere apologies. Pls send your email to cruncheconometrix@gmail.com
@@CrunchEconometrix Thank you. Upon your reply, I would like to ask you a question. I am conducting research using heterogenous long panel data where T>N. In case all of my independent variables and dependent variable are either I(0) or I(1) (after performing the ADF tests), I think I could utilize the ARDL model to test the short-run and long-run effect of FDI on economic growth.
Is there any other way of estimation that would allow the use of I(0) and I(1) variables together? - As I don't think I would be entitled to use Johansen's cointegration test, engle granger approach, var, Toda-Yamamoto, the VECM, or the ECM (because the integration orders are different - both I(0) and I(1)). Is it correct? Would you suggest any other model than ARDL? Thank you in advance.
Kindly watch my videos on panel ARDL. You will find them helpful.
Hello Dr Adeleye. I'm from The Bahamas and doing econometric work at The Central Bank there and 90% of what I have learned is from your videos. I appreciate you taking a practical approach to explaining these procedures and theories. In regards to the Johansen Test, how do I go about choosing the optimal lag for this test ? I watched your video on how to select the optimal lag for a VAR, but is it the same process ? My process said that 1 lag is optimal (Im using annual data) however, this is problematic because my VEC must be 1 less than my optimal lag which gives me 0. So in a case like this, is it practical to use 2 lags for my VAR ? If so, should I also use the same amount for my Cointergration test ?
Hi Lorenzo, thanks for reaching and the positive feedback on my videos. Deeply appreciated! This is really not a problem. Relax. Simply estimate the VECM with 1 lag because it cannot be a static model (that is, with 0 lag). Put a footnote in your paper stating this.Thanks.
Thank you so much for this extremely informative information. I have a question, please: if some of my time series variables are I(0) and some are I(1) is VAR no longer an appropriate model of choice? I know I can use ARDL in this case (thanks to watching your videos) but I wanted to confirm that VAR is inappropriate for the mixed-order time series variables that I am working with.
You know the answer already. VAR modelling is inappropriate.
How can I before testing cointegration(using Johansen Cointegration or Engle Granger) we assume that the group of time serries are cointegrated? Is there a test I can do on a group of timeserries data ?
Hi JJM, the Ho of no cointegration is tested against the alternative that a cointegration relationship exists.
Hi,
Thanks for the great videos and explaining these concepts in a simple way. I have a question regarding what needs to be done after the Johansen test. Do you have a video or can you elaborate how to construct the stationary time series after we have determined that there is 1 or more stationary time series from the test results? I am getting confused at this stage. And also, if we are using the ln transformation on the input time series, how do we apply the weights to the original time series?
Any clarifications would be very helpful.
Thanks
Vin
Hi Vineet, please watch my VECM videos. Thanks.
Hello,
I have a variable whose distribution does not follow the normal law, I tried to correct this problem by the logarithmic transformation on the software Eviews and the p-value of jarque-bera remained lower than 0.05 what to do?
Hi Fatiha, modify the lag structure and re-estimate.
Thank you for the video. I would like to ask about the role of the outliers. What if there are multiple outliers in the data set, and the variables are integrated in I(0). Should I remove the outliers via the winterization method in Eviews, or it is ok to conduct the Johansen cointegration test on the data set where I have multiple outlier values? thank you beforehand,
Hi Ibrahim, thanks for the encouraging feedback. You can use the winterization method if you are familiar with it. But I know that outliers are taken care of once you take the log of the variables.
Thank you for the video. One question, what if series are not stationary even after log-differencing?
Hi Ksenija, test with trend and if it is still not stationary, you may need to change the variable... Or explore a different technique that accommodate I(2) series.
Hi professor, I have a question regarding the Johansen cointegration test. Say I am running 4 variables with spot prices as my dependent variable, and futures 1,2 and 3 as my independent for the Johansen test. And the output leads to all is insignificant and there is remark given by eviews that at most 4 equations are co-integrated. What des this financial interpretation mean?
Hi Gilbert, it means that you have 4 cointegrating relationships. But follow my guide and use just I when estimating the VECM.
I don't get it, so in this case we are accepting that there's cointegration? What we are looking for is a Pvalue lower than 0.05 at 'none' but bigger at 'at most 1 and 2'?
You may want to watch the clip again to fully understand the intuition and interpretation of results.
Hello!, If my series are integrated of different orders how shall I proceed? There is 1 cointegration within my data. Can i still do VAR or VECM?
Catalin, kindly watch my video on ARDL.
Hi, firstly thanks a lot for these videos which are really helpful!! Can you please tell me what will happen if the series are integrated of different order (that is, having a combination of I(0) and I(1) series. Thanks!
Hi Ajit, thanks for the positive feedback. Deeply appreciated! Read Pesaran and Shin (1997); Pesaran, Shin, and Smith (2001) for in depth details to your query. Thanks.
In model VECM ...If i found the C1 negative but non significant at the level 5%. we say there is relation a long run between variables but without significant and there is not adjustment to equilibrium iin the short run to a long run
or say there is no relation a long run between variables but without significant ?
Hi Bettayeb, once the coefficient is NOT statistically significant it implies that there is NO relationship between the independent and dependent variable.
Hi. I am following all your videos and i finished doing the stationarity test and my series were stationary at first difference. I just finished conducting the johansen test as well and I'm confused with my results. all my trace values are smaller than the 0.05 critical values (including the none hypothesis) and all my probability values are greater than 5%. what does it mean? is that a bad thing?
Hi Zama, I interpreted the results of the JCT. Please watch the clip again and modify my explanations to suit your outcome.
Hello ma'am, why do you choose assumption no 3 arbitarily when doing Johansen Cointegration test? Is there any test to make sure which assumption that should be used (about deterministic trend)? some literatures said that we can use Johansen Summarize Test (6) and choose based of AIC or SC, Is that right? Regards from Indonesia 🇮🇩
Hi Christ, I chose model 3 (unrestricted constant and no trend) for simplicity. Nothing stops the use of other models if you understand the underlying assumptions. My love to Indonesia 🇮🇩 please share my UA-cam Channel link with your friends, students and academic community. Thanks! 💕 😊
For my thesis, my all variables arr stationery at 1st difference and one varioable is stationary at 2nd difference. Now which is best for my data?
I have responded to your previous question on this.
madam appropriate lag for my data is 8. When I use 8 lags data shows no cointegration. but when I use 2 lags data shows cointegration. I should accept the result as shown by 8 lags, right?
Sunita, you are the researcher. Take your decision if you either want cointegration in the model or otherwise. Thanks.
Is it possible for 2 variables to have no short-run relationship but have a related long-run relationship?
Yes, very possible.
On which variables we can run the Johnson cointegration means first difference of log series or raw data series?
Hi Mohammed, if you watched this clip you will agree that I answered your query early on the video. Kindly watch again and please pay attention to the screen and explanation. Thanks
Thank you Ma. How do we estimate the short-run if there is no cointegration
Hi Iwasam, if there's no cointegration you are to report only the results from the underlying VAR model.
Hello. thank you for your video. Here, you say do not use the differenced data, if unit root testing finds unit root, that is I(1). Why is that? Is it because EVIEWS will difference the data for us? Thank you.
it is advisable to test the variables in their raw forms for cointegration.
Hello ma'am ,, first of all thank you for guiding us.
one query i am having, what if i am getting situation 3(mix of stationary data at level and at first difference) after performing stationarity test
can't i perform co-integration test?
Hi Priyanka, kindly watch my video on BOUNDS COINTEGRATION. Thanks
what should I do if all the "Hypothesized No.of CE(s)" are significant? What have I done wrong?
Laksmita, you have done nothing wrong. Your result is as it is. Follow my explanation to decide on the final outcome.
Dear Professor, if I have dummy variables in my model, should I include them in the group for the Johansen cointegration test? Thank you very much for your answer.
Hi Eric, I'm not sure if VAR accommodates dummy variables. You may need to check other online resources.
@@CrunchEconometrix Thanks, Professor. Will check other resources. Your videos are very helpful. Thank you for sharing your knowledge.
Hi
I'm getting the results showing co-integration of variables without log transformation but after taking log there is no co-integration. Which one should I choose?
Hi Javed, I'm quite sure you know the answer to your query. You estimated 2 different models, hence you have the prerogative to choose either result.
I realle love the way you explain. Here İn the notes you said that appropriate estimation technique would be ARDL and ECM in no.4. But at the end you said VAR and ECM. Which öne is correct? Pls response to my query. Keep up the good job👏
Hi Asfia, thanks for the encouraging feedback. Appreciated! Since this clip is on JCT, then VAR-VECM is appropriate. Thanks.
Thank you for the video, in the case that Trace test indicates 1 cointegrating equation at 5% but Max test indicates no contegration at the 5% level. Is it fine for me to continue with the cointegration or i should use another regression technique even though my variables are stationary in first difference.
Hi Takunda, this is easy. You decide which of the tests to take. If you go with Trace then perform the VECM and if it's the Max Eigen then you perform only the baseline VAR.
@@CrunchEconometrix thank you very much. Assuming i have yearly data on eviews then i change that data to quartely using eviews. Will it be fine if i use that data as quartely not as yearly
Takunda, data conversion is permitted. You can read more on the pros and cons online. Thanks
you said cointegration is applied on level form variables does that mean original data?
Yes.
Hello, your work is very respectful, How can I follow the right order of the videos for VAR and VECM, for better understanding. thank you
Hi Jennifer, thanks for the positive feedback. I always mention and list the PREREQUISITE videos to watch.
coodmorning madame, thank you so much for the video, its really hopeful, im algerian student in tunisia, im using this modelisation for stres testing by impulse fonction (if you have an idea about stress testing, do you think is it possible to make shocs by impulse function?),
but the real problem that i have is: that in my data, i have four variables, three are stationary at level I(0), et the last one stationnary in first difference so at I(1), which model should i use? VAR or VECM?? and after doing the test of johansen, there is no cointegration relations? could you help me please
Hi Walid, ARDL technique is the most appropriate when you have mixed levels of integration. I have very good videos on the technique. Kindly watch them. Thanks.
@@CrunchEconometrix Hello maam Thank you for your great video it really helps a lot. i have a same problem as him and as you have suggested to use ARDL model so can i use ARDL model in panel data too ? need your help
Thank you so much for the video!
What diagnostic tests should we do after doing the Johanssen test?
To the best of my knowledge, none. But you can watch my VAR and VECM videos for post-estimation tests.
Thank you!!@@CrunchEconometrix
Hello Professor. Thank you so much for your videos! Can a log transformation be done on all variables except the ones specified in percentages or in rates? And can this log transformed variables be used for your analysis?
Hi Kehinde, you can log transform variables that are in "% of..." but not rate variables like inflation rate, interest rate, exchange rate etc.
@@CrunchEconometrix Thank you for the swift response. Does Squalli and Wilson(2011) measure of trade openness expressed as a percentage fall into a rate variable? Can you perform an analysis with a mixture of variables that are log transformed and variables in their raw form with the exception of rate variables?
A variable that is a percentage of another is NOT a rate variable. Yes, you can have a mix of level and log variables in the same model.
@@CrunchEconometrix Thank you!!! This sums it up.
Thanks alot so far your help, u tutorial has really been helpful.
After carrying out the johassen cointegration test and u discover that this is no cointegration, what test will u perform.
Hi George, I must have mentioned that you perform just the basic VAR and not the VECM.
Hello
Thanks so much for the videos. I am learning a lot from you. What do we do if like one serie (or more) is I(0) and the others I(1) in the case of Johansen Cointegration test?
Hi John, in that case VAR/Johansen is not appropriate. Use ARDL/Bounds test. Please may I know from where (location) you are reaching me? Thanks.
@@CrunchEconometrix Thank you. I am reaching you from Haiti.
@@johnbobymesadieu8922 Awesome! Kindly share my videos with your colleagues. Thanks!
@@CrunchEconometrix Sure! Keep doing this amazing work.
Nice work.my question is that mostly researcher shows the Johansen result through its coefficients not through trace stat and eignvalue .how we can pick such results from this test
Hi Muhd, for the positive feedback. Deeply appreciated! This video wraps up the nitty-gritty about the JCT. Nothing more to add or deduct. Thanks.
Can you please mention the name of this software used for conducting the test?
Hi Bilal, EViews. Always indicated in the 1st line of the video description.
If cointegration exists either hy johenson or bounds test... But then none or only 2 out of 6 variables come significant.. Is that a problem.. Does cointegration mean long run coefficients must be significant.??
I explained the concept of cointegration. You may need to watch the clip again, thanks.
Thanks very much, Professor for the detailed and painstaking explanation on all topics ma.... .. How can I find your tutorial on the co-integration topic on the application on series integrated at different order ma....
Hi Simon, thanks for the encouraging feedback. Deeply appreciated. Search within my Channel for video on BOUNDS COINTEGRATION TEST.
where can I find the part 1?
Hi Selin, Part 1 is the video on "Augmented Dickey-Fuller" stationarity test.
Hello Ma'am, If trace statistics indicate the presence of cointegration among the variables but not the max-eigenvalues. What is to be done in the case?
Hi Saakshi, you are at liberty to choose whichever since you are disposed to using either of them. I explained that in my cointegration videos.
@@CrunchEconometrix Thank you Ma'am, and another question if not bothering you much. Can we use Engle-Granger (EG) co-integration test? So far my knowledge this test has been criticized a lot. Why I am thinking about the test as I have two variable model that is I am interested to check the long run relationship between two variables only. So is it okay if I only use Johansen and not EG or should I use EG rather than the former? This is part of my research and any help is highly appreciated.
@@saakshijha9689 I often don't take positions on which to use because both tests are still relevant in the literature just that for whatever reasons JCT is now more popular than EGCT. Again, I'll say use any and you can't go wrong.
@@CrunchEconometrix Thank you
@@saakshijha9689 U're welcome, Saakshi...kindly share my videos with your friends and students too, thanks! 💕 😊
mam i confused a little bit in your vedio ... first i took log and than i make stationary my data.My all variables are
stationary at first difference.now can i used johensan co-integration test.
Yes MTK. You are on track. Nothing to be confused about.
Whats the difference between Johansen and Bounds test for cointegration ?
Both are tests for determining long-run association. JCT for VAR/VECM and BT for ARDL models.
Thank you for your knowledge sharing Mam. Please help. what do I do when my dependent variable is stationary at second difference but all exogenous variables are stationary at I(1). I was going to perform a cointegration test then VECM if all variables were stationary at first difference, but with this unit root result, I don't know what to do.
For VAR/VECM estimations, all series must be I(1).
@@CrunchEconometrix Yes ma. What do I do in this I(1) , l(2) case
Use the Toda-Yamamoto procedure. Read about it online because I've never had cause to use it.
"The variables you are using must be on their level form and not on their first difference..." But the video is how to perform cointegration at order 1. Am confused. Somebody help...
Hi Beedee, nothing confusing here. I suggest you read some resources on testing for cointegration to support the estimation guide shown in this video.
Must be of the same order of difference ..take first order of difference assuming they all happen to be non stationary.
Dear Dr. I have 4 variables (lngdp, lnfdi, lncpi and lnir). At levels, lngdp is non-stationary in intercept but stationary in trend & intercept whereas lnir is non-stationary in both. However, all the variables are stationary at first difference. So Dr., is it safe to say that all the variables are integrated of order i(1) and then apply the Johansen test for cointegration?
Yes, all are I(1) series only make sure to make a note to the table.
@@CrunchEconometrix ok. Thanks Dr. for your prompt response.
Can the Johnsen co-integration test be applied for the data which is stationary at levels?
Hi Pavan, no. JCT is applied to NONSTATIONARY variables.
Okay ma'am thank you 🙂 then how we can test the co-integration for the data which are stationary at levels.
No test required.
Which model should i apply as my variables are stationary at 1st difference and one variable is stationary at 2nd difference.
Consider the Toda-Yamamoto procedure which allows I(2) variables.
@@CrunchEconometrix thank you Sir. Do you have a video on that?
Not yet. But I have listed it as a to-do video.
how can i find the pdf of the presentations
Hi Mucee, no pdf version. You can copy the notes from the video description and save as a pdf file. May I know from where (location) you are reaching me from?
I checked VEC model there is no causality running from tax revenue, income inequality and consumption to GDP in long run..
No causality Tax revenue to GDP in short run also . no problem of serial co relation, heterosecidiesty ,Granger and linearity .Do you suggests me I'll accept these results. If I accept then suggest me any solid rational .
Unfortunately, I don't get involved in personal studies. I offer general advice. Your model seems okay having passed all diagnostics.
if some variables are at level and some others are first difference so which can i combine at level or first difference pl help me
Hi Jemis, if that's the case, perform the Bounds cointegration test. Kindly watch my video on the procedure...may I know from where (location) you are reaching me?
@@CrunchEconometrix somalia am student and i study statistics and planning and i get so many things for your videos and now i studied how to test co-integration test so send me your email as to get from you many knowledge
Good to know, Jemis. Please inform the academic community and your friends in Somalia about my UA-cam Channel. No need for my email address just follow my tutorials and practice along. Thanks.
Hi madam, I wanna ask about I have run Johansen cointegration and the result of the research is interest rate has a positive impact on share price, normally this should be negative relation, where is my mistake? Thank u
Surf the literature to find supporting evidence for your result.
@@CrunchEconometrix ok thank u madam
Hello Ma'am, thank you so much for your video tutorials. I have a query - My model consists of 5 variables - One dependent and rest of the variables are independent. ADF unit root test confirmed that all variables are integrated of order one. So then I applied the Johansen Cointegration test. The results of Cointegration test show that there are at most FIVE cointegration equations. The problem is that I have read papers that state that the no. of cointegration equations should be ONE less than the total no. of variables used in the model. What should I do in this case???
Hi Saima, as explained in the clip use one cointegration equation for easy interpretation. This is simply because the normalization is on your outcome variable which is the 1st variable listed in the system. But if you are on top of your interpretation then you can use more than one CE. Kind regards.
One more question please - Is it ok to use different lag length for Cointegration test and VECM test for a study? For instance, can I choose lag length 2 for Johansen Cointegration test and lag length 3 for VECM test since I get better results in doing so. Is this possible? Thank you in advance!
As explained in my VECM videos, estimation is with p-1 lags except the optimal lag length is 1 then VECM is also estimated with 1 lag.
@@CrunchEconometrix Thank you so much for your prompt response and suggestion! But ma'am, I m sorry I still don't get it. What i mean is - I have 5 variables in total ( including the dependent variable) and then both Trace test and Max Eigen text show that there are 5 cointegrating equations in the Johansen cointegration test results. So, in such a case can I further test the long run relation with VECM or I must use VAR because no. of variables are equal to no. of cointegrating equations?
Kindly follow the steps shown in the VECM videos. They are well explained. Thanks.
Can you please explain how we can create restrictions on the cointegrating vectors?
Hi Ann, you may need to check other online resources for that.
@@CrunchEconometrix Can you recommend any channel?
Ann, you can easily search the Internet for resources.
Dear Bosede, I have two variables , REER dependent and Public Debt independent, I know that it is EG that I ve to use. My question is can I use Johansen in this case because some authors and articles say that it is possible to use. Can you please clarify this issue.
Hi Azer, your choice of variables has nothing to do with the estimation technique. No relationship at all.
@@CrunchEconometrix I see, thank you.
@@azerdilanchiev3781 U're welcome.
www.omicsonline.org/open-access/impact-of-debt-threshold-level-on-real-effective-exchange-rate-a-scenario-of-developing-countries-2162-6359-1000555-107438.html
What about this ???
erepository.uonbi.ac.ke/handle/11295/95259
How do you decide on the optimal lag intervals?
Hi Antoniya, kindly watch my videos on Optimal Lag Selection and those on VAR and VECM on how to go about it. Thanks.
Thank you for your quick response!
What happens if there is a contradiction between the results of max statistics trace statistics. that means one test suggesting cointegration and the other is suggesting no cointegration???????
Hi Brijesh, I mentioned that you have the prerogative to choose either of the results.
@@CrunchEconometrix 🙏🙏 thankyou for responding mam. My trace statistics indicating for a cointegration but max stat. Suggestion no cointegration . In this case what should I do. Once again thanks for your videos, it helped me a lot.🙏🙏
Hello Ms, I have one question, what if When I do my Johansen Cointegration test, i have the only * I have is in None, I dont understand why it's cointegrated when:
None * 0.482373 51.36303 40.07757 0.0018
At most 1 0.290746 26.79620 33.87687 0.2744
At most 2 0.200102 17.41512 27.58434 0.5445
At most 3 0.170329 14.56465 21.13162 0.3206
At most 4 0.112543 9.312804 14.26460 0.2610
At most 5 0.000651 0.050830 3.841466 0.8216
So in this case I have to do a VECM with 1 cointeg equation? What If I have the * in None and At most 1?
Thank you very much
Hi Maskim, I gave clear interpretations of the Johansen Cointegration test results. You may want to watch it again and apply to your results. Thanks.
@@CrunchEconometrix Thank you, I saw it again. I have other question, How important is the rank of cointegration? Because sometimes the Johansen test is telling me that my rank is 4 but for example If I have 6 variables in my model I "lost" the long run information of the first 4 of them, is it so bad to do rank 1 of 2 even if the johansen test tells me that i have rank 4? or how can i recover the long run information of this 4 variables? because I have 1.000 in the first and then 0.0000 in the 3 next, thank you very much Ms
Maskim, it is important you read up on RANK from any econometrics textbook for better understanding. Video tutorials are insufficient to equip your understanding.
It is now clear that it is left to the discretion of the researcher to decide whether to use the trace stats or max eigen value results in case they are in conflict. I hope it therefore means that it is not mandatory to obtain uniform results for both trace and max eigen stats.
Yes.
I wish you could demonstrate to us such an example because we have been made to believe that both trace stats and max eingen value must always agree. It would be informative to see how a VAR vis a vis a VECM of the same data set will look like.
Denis, this is the application of common sense. Researchers ain't robots but humans with the ability to think, reason, and take decisions based on outcomes. These results are AUTOMATICALLY generated. And they do DIFFER at times. So, what do you do? You DECIDE if you want to perform VAR (no cointegration) or VECM (presence of cointegration) and I have videos on both. Researchers have the ability to own their studies and defend it. There's no big deal here if you know what to do.
This response is a great confidence booster. You are a Godsend to us researchers. Be blessed.
How do you know which assumption ypu you should use?
Hi Kamran, what assumptions pls?
@@CrunchEconometrix hi, the specification for the Johansen test
Hi Kamran, I don't use any assumptions.
Hi this is Shaik from India.
I have dought on this,
whether cointegration will be applicable on order level or first difference ?
Hi Shaik, I have edited my initial response from "1st diff" to the raw or log-transformed form of variable as explained in my JCT video. Tx
@@CrunchEconometrix thank you for your quick response, but in your video you mentioned that NOTE: Co-integration test should be performed on the level form of the variable and not on THEIR FIRST DIFFERENCE ? How come its level form
Ma for clarity you asked us to use data on the first level form wen performing co integration or the log of raw data. I want to ask for the first level form I believe it's the raw data uploaded from excel u meant ma
Yes Emmanuel, use the raw untransformed data.
Hlo Ma'am,
In video you took lag interval 1 2. I followed lag length criteria and optimal lag is 4. Should i used lag interval 1 3 (p-1)??. Should i abstract one lag in johansen test as i noticed in video related to VECM.
Hi Ajay, if the optimal lag is 4,then for JCT it is 1 4; and 1 3 for VECM.
Thanx a lot Ma'am... plz specify the reason.
Why lag is 4 for JCT and 3 for VECM??
@@ajaysidhu4091 Hi there, my videos are well explained. I will advise that you watch all videos related to VAR and VECM, then you will get the understanding.
@@ajaysidhu4091Kindly, watch my videos on VAR and VECM for better understanding.
Hi could I book a hour of consultation with you for my project?
Greetings
Kindly send me an email cruncheconometrix@gmail.com for us to conclude on your request. Thanks
Good video but still not clear enough,***If TRACE and Max Eigen is Higher than 5% critical level reject Ho, but in this case we cannot reject Ho from your stated hypothesis. I need to be clear your choice of word.
Hi Oyebade, you may need to watch the clip again. Thanks.
Madam, I feel that in this video it should be Reject the Null Hypothesis if the value of the Trace and Max statistics < 5% critical values otherwise, fail to reject the Null Hypothesis. This is because for example in Trace, at 'At Most 1' the trace statistics is 11.35331 and the 0.05 critical value is 15.49471. In your explanation you are rejecting the Null hypothesis (No cointegration equation) and accepting that there is atmost 1 cointegration equation. Please clarify.
Hi Manoj, u've stated exactly what I said. In the JCT, H0 = no cointegration. So once Trace>CV, that hypothesis is rejected at the 5% level. Kindly watch the clip again and follow my explanation. Thanks.
Madam, when Trace > CV it means that there is no cointegration and hence the Null Hypothesis is accepted. Only when Trace < CV, there is cointegration. And hence the Null Hypothesis (No Cointegration) is rejected. Please clarify whether I am right.
Manoj, I've given you my explanations, but I'll advice that you read up and decide on your own. No need for arguments. Hope this settles it, thanks!
Thank you so much for this video but we only accepte H1 when when the trace is greater than critical value n conclude that there is a relationship btw the variables...you swiped your conclusions that when we accepte H1 we fail to conclude the existence of a relationship yet it's the vice versa of that! you can re-check the video
Prior, are you sure about this?