Beloved guest/subscriber, you have discovered my amazing UA-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Hi, thank you for being interested in a question, but I will be clear and precise. 1) I would like to know what is the logic in this estimate of GMM, because I have the impression that it is a method that uses trial and error. unless I made a mistake. 2) How to play with the variables so that they are not significant in AR (1), AR (2) and Hansen, but also for the interpretation of the variables while passing by the possibility of the student must be equal to P
@@martinbitsie7270 (1) GMM is an IV regression. It is advisable that you read those references I indicated at the end of each video to fully understand the concept of GMM; (2) watch my videos to see how different simulations can lead to different values of AR(1) and AR(2). Thanks.
Bosede Ngozi Adeleye thank for your video. I am beginner for this. May I ask for your advice? How we can know which variables are exo or endo geneous. How may lag should I use? My topic is about predict stock return (y) base on financial ratio( roe , dy, roa,....)
Thank you Professor, it is instructive and helpful. Thanks a lot. But I have a question that my result of AR(1) always is non-significant. Does this question will affect my results highly. And why does my result of AR1 is non-significant? and how to solve this question? Hope to hear from you soon, thanks a lot again
very nice explanation the more I saw the videos the more i understand the Difference and system GMM... I have only one question how come the endogenous variable and the internal instrumental variable are the same
Great tutorial Professor. Please using first and second difference Gmm I am getting great results but my F and Prob >F does not provide any value, it’s just dot( a point). It’s there a problem with my results?
Yes, Ray something is not quite right with your results as you should have the F-statistic and corresponding pvalue. You may want to estimate the model by adjusting lags and the instruments in IV( ).
Dear Professor, Thank you once again for your amazing efforts Could you please clarify on external instruments used in your GMM model. All the external instruments (IV) used in your model is already included in the main equation, I can see only 'lntrade' as an external instrument. Please clarify and guide me on this regard. Thank you.
I just want to know, is there any other command or way to perform the Two-step GMM dynamic panel data model in STATA because while I am giving the xtabond2 command STATA showing unrecognize command
Hi Joyti, it is unrecognized because you did not install the syntax. Type "help xtabond2" in the Command Window and follow the Stata prompts for installation. Thanks
Thanks Yeboua for watching my video. As to your question, I have no paper to back up that assertion but it's the academic norm. No reviewer or Supervisor who knows what he or she is doing should ask for unit root test when a series has less than 10 years observation. But keep searching, maybe you'll find a paper.
Thanks for your lecture! I have one question. Would weakly balanced panel make any differences in the result? A set of strongly balanced panel is necessary?
Instead of creating the year dummies, can i not just use the command i.year? Also, how do I choose between a one step or two step? should i just run one and two step models, and see which is better?
@@CrunchEconometrix thank you so much for this! your content is very highly appreciated. I shared your videos with my peers who are also doing their thesis in the pandemic and trying to learn :)
Dr. Ngozi, great the GMM approach explaining I would like to ask you about the result of hansen test p-value(0.99) in my diff estimation, how to emphasize it in my interpretation.
Any GMM result having a Hansen statistic of 0.99 should be disregarded (Roodman, 2009). Read the paper and watch my GMM videos from the beginning because I emphasized it and should what can be done in the event such occurs.
@@CrunchEconometrix thank you for replying me. I asked it since in your almost all video you included all explanatory variable in iv. Is there any criteria to help decide which explanatory and other variables should I include in instrumental variable? last question is I am a little bit worry about the fact that you used the explanatory variable in IV as a level term instead of firs difference?
@@rmdrayfelix7157 RM, you have to run different simulations to get the model that gives you the best results. Studies differ so what works in mine may not in yours. If putting all regressors as instruments fits well in your model, good for you...and if using a few of them works well, that's fine too. GMM in itself is a complex estimator. Keep it simple and run several simulations to pick the best.
Hi Nirmal, the reason I don't have the EViews version is because it is not straightforward. Very complicated...may I know from where (location) you are reaching me?
Hello Professor, I have a doubt about my own analysis. I am evaluating and comparing firm value through two different equations, one with a variable for firm Risk and one without. Can I have a different set of instruments for each equation?
Hello, thank you very much, the whole series is very hepful and clear! I was wondering if there are some guideline rules on when one should use one step or the two step GMM. Thanks a lot for your help!
Hi Alessandro, none. You can use either. Most times I estimate both and take that which gives the best results. May I know from where (location) you are reaching me?
@@CrunchEconometrix I will for sure :)! I have another question that is not 100% clear from reading Roodman 2009 and other papers: does system GMM account for unobserved heterogeneity (fixed effects) or not? In other words, if I am sure I need to control for unit heterogeneity, would difference GMM be more robust than system GMM? thank you in advance for your help
Hi - thanks for your video. I was wondering what do you do when your regressors outnumber your instruments? This is the case for me when I introduce y* (my year dummies are 40)..
Prof, A question just came to my mind while (re)listenning to this beautiful lecture of yours, What if AR(1) is not significant? but all the other yield good results that is AR(2) is not significant, p value of Hansen and Sargan too. Should the results be discarded due to poor instrumentation? must AR(1) be significant?
Yes, I didn't bother because the syntax for the Two-Step Difference GMM are minimal modifications to the One-Step Difference GMM which can be gleaned from the screen. Thanks for watching. May I know from where (location) you are reaching me?
Hello Dr. Adeyele, once again, I followed your previous teachings on xtabond2 and I find them really helpful. However, I am working on my dissertation whereby I have endogenous variables in my model, I noticed you did not make any example of such, I'd like to learn about how to specify these variable in my System dynamic panel model. My N=13 and my T=9. Also, I have a question about the no of instruments, no of groups or heterogeneity and no of observations, I mean must they exceed or remain bellow the no of observation or groups?
@@CrunchEconometrix Hi Adeleye, thanks to reply, i am learning how to estimate dynamic panel model in stata. and i wanted to download your dofile to try it in my data. but your website didnt open
Hi Dr, hope you are doing well. I have a question on t-statistic GMM. Assuming I want to compare 2 values of t-statistic on A and B, how should I interpret it, if A t-statistic is 2.94 and B is 1.94. Thanks again for your kind help.
Prof. Thanks for this great insight into Two step GMM. I am running a GMM estimator using a panel of 81 countries over 23 years. And in line with the work my study is modeled after, i used three years average data. the result of System GMM indicate that both Sargan and Hansen test are highly significant. Pls what do i do to solve this problem of over identification instrument even after including the collapse option.
Hi Ugo, thanks for the positive feedback on my video, deeply appreciated. From your query, 3-year averages reduces T to about 7years? That is good. GMM procedure can be quite naughty and you may have to try different simulations: reduce regressors, reduce instruments, estimate with and w/o year dummies. Try these and see the outcomes. Also, try use the 1-step system and the difference GMMs then compare results.
thanks very much for your response I sincerely appreciate. My understanding is that System GMM produces a better result than Difference GMM. What do you have to say about that?
@@ezeugonna5017 Not quite, only in terms of number of instruments used. Both difference and system GMM yield efficient results. One of the reasons I made this GMM series is to draw attention to the relevance of difference GMM. Do a Google to see several contemporary papers on difference GMM approach.
thanks so much Prof. Please I encountered another problem with my STATA while estimating the model. It started given this error message "Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm. J(): 3900 unable to allocate real [163296,2016] _Explode(): - function returned error _ParseInsts(): - function returned error xtabond2_mata(): - function returned error : - function returned error r(3900);" please do you have solution to this problem
Hello, first I will like to thank you for your work. I am working on the effect of remittances in the four pillars of food security, I have a data panel of N=34 and T=6, I was thinking in applying GMM, since then I performed a good model for one-step system but when I applied other types of GMM estimations, the regressors (except the lag dependent variable) become not significant, why do you think this is happening? The model has 111 observations (I don't know if that is enough) and due to the missing values at the end only use 25 groups, but there are 22 instruments, the Hansen is chi^2=0.112 and AR(2) z=0.438
Hi Francisco, thanks for reaching out. If one-step GMM give you the results you want then go for it. If you watch my videos you will see the measures involved in changing the status of coefficients.
@@CrunchEconometrix Thank you so much for your answer! I have another question, what do you suggest if the coefficient of the lag dependent variable in the GMM model is higher than in pooled OLS but lower than in LSDV. I also preformed ivhettest and showed a presence of homoskedasticity (as I understand GMM is recommended when you have a presence of heteroskedasticity).
thank you for your efforts can I ask why the following error appears when I apply Sys GMM : Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm. / / / / / / invalid.r(198); PLEASE NEED HELP THANKS IN ADVANCE
Hi! Thank you for your video! I find it very helpful why didn't I find it sooner?!! Anyway, I tried to open your website to get the do file but I cannot open your website. Is there any other way for me to ask for your do file? I really need help with my thesis right now. Thank you so much!
Hi Novia, thanks for the encouraging feedback on my videos. Deeply appreciated! 🙏 ❤️ My website is accessible cruncheconometrix.com.ng/shop/. Kindly try again.
@@CrunchEconometrix Hi, thank you for being interested in a question, but I will be clear and precise. 1) I would like to know what is the logic in this estimate of GMM, because I have the impression that it is a method that uses trial and error. unless I made a mistake. 2) How to play with the variables so that they are not significant in AR (1), AR (2) and Hansen, but also for the interpretation of the variables while passing by the possibility of the student must be equal to P
@@CrunchEconometrix hello Professor, thank you and congratulate you all for the work you do for science. I finally understood your method. But I have another concern, if I lead a study at the Aranese scale ie 54 countries, and later I would like to make an estimate on one of the regional economic communities so how am I to take when my base is made up of 54 countries? which command to execute or is there another method? Thank you in advance for your response and good luck to you.
Hello, Thank you for your videos, they are really helpful. I face a problem when I implement the GMM model. My problem is that my main independent variables is omitted!! Do you have any idea what would be the problem?
Dear Prof. I run the model for diff GMM but I get missing values for F test and its' probability. While I drop the option noleveleq and use system GMM I get results for F test. What would be the reason for that?
Hi Ruri, it's in the video title. Stata13. Have you tried installing with "ssc install xtabond2" or "findit xtabond2" or "help xtabond2". May I know from where (location) you are reaching me?
Hi, Ms, can you give an orientation please? I have a doubt: why when you use iv(variables), the xtabond and xtabond2 use the difference of this variables as instruments instead of levels?. I read arellano and bond (1991) and i don't find any reference, it is gonna makes me crazy.
Hi thank you for your explanation! I have a question. I used xtabond2 y l.y l2.y x1 x2 x3 gmm(x1 x2, collapse) noleveleq twostep. But I got omitted value in x3 Could you give me solution for this?
@@CrunchEconometrix so I change code to xtabond2 y l.y l2.y x1 x2 x3 gmm(x1 x2,eq(diff) collapse) twostep and now I got omitted cons Could you help me how to fix it? Btw from korea😁
Dear Professor, my dependent variable is ROA, i want to use two dummy variables (time periods ) as predictors in the main equation and as instruments other variables, the syntax is like : xtabond2 roa l.roa S2 P2 , gmm(roa ,eq(diff) collapse ) iv (inflation l.gdp pen corr tax l.size prem ) twostep small robust. N=24, T=15.Am I doing well ? or Am I completely in the wrong direction?
Dear Professor, I am trying to follow this method but i keep getting my instruments bigger than the number of groups. and also since my variables are only stationary in the first difference, so shall i be using them in the first difference as instruments?
Hi John, this evidences that one or more of your variables has so many missing observations. Check your data to know which and change them with closer proxies having more data points. Hope this helps. May I know from where (location) you are reaching me?
Beloved guest/subscriber, you have discovered my amazing UA-cam Channel tailored specifically for you and other beginners and intermediate users. Please do not keep me to yourself (lol). Kindly share my videos and links with your students, colleagues and academic community so that they too can SUBSCRIBE and learn with ease….and for the global community to be aware that applied econometrics can be simplified. My teaching approach is very practical. I adopt a do-as-I-do style. Many thanks to those who have supported me by telling others. Once again, CrunchEconometrix loves to teach, support my Channel with your subscription, likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Hi, thank you for being interested in a question, but I will be clear and precise.
1) I would like to know what is the logic in this estimate of GMM, because I have the impression that it is a method that uses trial and error. unless I made a mistake.
2) How to play with the variables so that they are not significant in AR (1), AR (2) and Hansen, but also for the interpretation of the variables while passing by the possibility of the student must be equal to P
@@martinbitsie7270 (1) GMM is an IV regression. It is advisable that you read those references I indicated at the end of each video to fully understand the concept of GMM; (2) watch my videos to see how different simulations can lead to different values of AR(1) and AR(2). Thanks.
Bosede Ngozi Adeleye thank for your video. I am beginner for this. May I ask for your advice? How we can know which variables are exo or endo geneous. How may lag should I use? My topic is about predict stock return (y) base on financial ratio( roe , dy, roa,....)
Ngozi ADELEYE
-Ngozi ADELEYE
Covenant University Ota Ogun State, Nigeria. I have learnt a lot from your videos. Very educative. Great stuff
Thanks for the positive feedback, Peter!
What great insight in GMM approach. Thank you so much Dr. Ngozi
Thanks Zeph, for the positive feedback and remarks on my video. Deeply appreciated! May I know from where (location) you are reaching me?
thank you for your video. please how to recuperate residuals after xtabond2
Hi Nadia, thanks for the encouraging feedback. Deeply appreciated! I'm not sure if possible but you can try the "predict" command.
Thank you Professor, it is instructive and helpful. Thanks a lot.
Thanks, Omer for the encouraging feedback... deeply appreciated!
Thank you Professor, it is instructive and helpful. Thanks a lot. But I have a question that my result of AR(1) always is non-significant. Does this question will affect my results highly. And why does my result of AR1 is non-significant? and how to solve this question? Hope to hear from you soon, thanks a lot again
AR(1) not significant implies there's no 1st order serial correlation which is good. Interpret your results.
@@CrunchEconometrix So does this mean i can not do GMM model?
You can.
very nice explanation the more I saw the videos the more i understand the Difference and system GMM... I have only one question how come the endogenous variable and the internal instrumental variable are the same
Yonatan, that will be at your discretion. It's your study. Your research.
Great tutorial Professor. Please using first and second difference Gmm I am getting great results but my F and Prob >F does not provide any value, it’s just dot( a point). It’s there a problem with my results?
Yes, Ray something is not quite right with your results as you should have the F-statistic and corresponding pvalue. You may want to estimate the model by adjusting lags and the instruments in IV( ).
@@CrunchEconometrix Okay Professor. Thank you
Thank you Madam but Ma'am how to proceed if we want to display results by country?
Kindly watch my videos on SUB-SAMPLE ANALYSIS and incorporate the idea into the GMM model specification. Thanks.
Dear Professor, Thank you once again for your amazing efforts
Could you please clarify on external instruments used in your GMM model. All the external instruments (IV) used in your model is already included in the main equation, I can see only 'lntrade' as an external instrument. Please clarify and guide me on this regard. Thank you.
Yes, Swathi. You have made the clarification. Exactly what I did. Some treat IVs differently but I chose the most used approach.
Professor
I dont have any constant term in two step system gmm estimation result.
Now what can i do?
Is it ok if i report without constant value?
Report the results you have.
I really enjoyed this video
Thanks for the feedback, Einstein...deeply appreciated, kindly tell others about my Channel! 💕 😊
would it be better if you name your variables differently so we can follow the logic; for example V1 V2 V3
Yahya, you mean you don't understand the variables used?
Thank you for your teaching. It's helpful. I have question, in performing preliminary, besides multicol test, is gmm method require normality test?
Thanks for the positive feedback. Appreciated! Pls watch the foundation GMM video to know what's about the technique. Tx
Thank you
I just want to know, is there any other command or way to perform the Two-step GMM dynamic panel data model in STATA because while I am giving the xtabond2 command STATA showing unrecognize command
Hi Joyti, it is unrecognized because you did not install the syntax. Type "help xtabond2" in the Command Window and follow the Stata prompts for installation. Thanks
Thank you for your video. It is usually said that when T
Thanks Yeboua for watching my video. As to your question, I have no paper to back up that assertion but it's the academic norm. No reviewer or Supervisor who knows what he or she is doing should ask for unit root test when a series has less than 10 years observation. But keep searching, maybe you'll find a paper.
Thanks for your lecture! I have one question. Would weakly balanced panel make any differences in the result? A set of strongly balanced panel is necessary?
Hi Donghoon, not much difference. I use unbalanced panel most of the time.
AOA
Madam which is P- Value of AR (2) and Hansen test in difference GMM?
Umar, if you check the result output you will see the pvalue. Thanks.
Is it OK to use external instrument in GMM?!
Hi Ayman, yes. They will be in the iv( ).
Is it necessary that both sargan test and hansen test to be more than 0.05 value?
What value of both is good for both?
Hi Mohd, I explained this in the GMM introductory video. Kindly watch it, thanks.
Good morning mam , how we can learn these commands.
Familiarity with the techniques speeds up understanding the commands.
why did dummies i and 2 omitted? what if IV is omitted?
Hi iftikhar, they were omitted by Stata due to multicollinearity. I have not encountered where an IV is omitted.
Thank you Madame
U're welcome, Debleme...kindly tell others about my Channel. They need to know. Gracias! 💕 😊
is xtdpdsys and xtabond2 different? can you refer the difference please
Yes, both are different dynamic panel data techniques. You may need to read more about their differences via Stata HELP menu or PDF documentation.
@@CrunchEconometrix thank you
Instead of creating the year dummies, can i not just use the command i.year? Also, how do I choose between a one step or two step? should i just run one and two step models, and see which is better?
Kiran, YES to both questions.
@@CrunchEconometrix thank you so much for this! your content is very highly appreciated. I shared your videos with my peers who are also doing their thesis in the pandemic and trying to learn :)
hi , correct me if i am wrong , does y* mean that we have to add every year dummy in the model like y1,y2,y3,y4,y5 ........ in our syntax
Junaid, with y* Stata automatically includes all the year dummies (less one) into the model.
Dr. Ngozi, great the GMM approach explaining I would like to ask you about the result of hansen test p-value(0.99) in my diff estimation, how to emphasize it in my interpretation.
Any GMM result having a Hansen statistic of 0.99 should be disregarded (Roodman, 2009). Read the paper and watch my GMM videos from the beginning because I emphasized it and should what can be done in the event such occurs.
@@CrunchEconometrix Thank you for your answer, N=50, T= 21 I think the obstacles here the T is it correct?
Firstly thanks for sharing your knowledge. Please, is it a must to include all explanatory variables in instrumental variable?
No, not at all.
@@CrunchEconometrix thank you for replying me. I asked it since in your almost all video you included all explanatory variable in iv. Is there any criteria to help decide which explanatory and other variables should I include in instrumental variable? last question is I am a little bit worry about the fact that you used the explanatory variable in IV as a level term instead of firs difference?
@@rmdrayfelix7157 RM, you have to run different simulations to get the model that gives you the best results. Studies differ so what works in mine may not in yours. If putting all regressors as instruments fits well in your model, good for you...and if using a few of them works well, that's fine too. GMM in itself is a complex estimator. Keep it simple and run several simulations to pick the best.
Ma'am, kindly share the GMM estimates for dynamic panel model using Eviews. I will be highly obliged.
Hi Nirmal, the reason I don't have the EViews version is because it is not straightforward. Very complicated...may I know from where (location) you are reaching me?
I am research scholar at IIT Kharagpur India.
@@Vizva-vloger-india Awesome Nirmal! 💕 Kindly spread the word about my videos to your students and academic community in IIT. Thanks! 💕 😊
Hello Professor, I have a doubt about my own analysis. I am evaluating and comparing firm value through two different equations, one with a variable for firm Risk and one without. Can I have a different set of instruments for each equation?
Hi Vincente, you can use different set of instruments for different models.
@@CrunchEconometrix thank you professor
Hello, thank you very much, the whole series is very hepful and clear! I was wondering if there are some guideline rules on when one should use one step or the two step GMM. Thanks a lot for your help!
Hi Alessandro, none. You can use either. Most times I estimate both and take that which gives the best results. May I know from where (location) you are reaching me?
Bosede Ngozi Adeleye thanks for your answer! :) i was in Lecce italy
@@alessandroguasti5779 Awesome Aless! I will appreciate if you can share my videos with your colleagues in Lecce...thanks!
@@CrunchEconometrix I will for sure :)! I have another question that is not 100% clear from reading Roodman 2009 and other papers: does system GMM account for unobserved heterogeneity (fixed effects) or not? In other words, if I am sure I need to control for unit heterogeneity, would difference GMM be more robust than system GMM?
thank you in advance for your help
Pls ma how were u able to generate ur year dummies??
tab year, gen(yr)
@@CrunchEconometrix thanks ma.
I got this "Warning: Uncorrected two-step standard errors are unreliable." for the result. how to correct it?
You may need to play around with lags and instruments set.
Hi - thanks for your video.
I was wondering what do you do when your regressors outnumber your instruments? This is the case for me when I introduce y* (my year dummies are 40)..
What is your N and T? I need to understand your data structure.
Prof, A question just came to my mind while (re)listenning to this beautiful lecture of yours, What if AR(1) is not significant? but all the other yield good results that is AR(2) is not significant, p value of Hansen and Sargan too. Should the results be discarded due to poor instrumentation? must AR(1) be significant?
Hi Asante, pay attention to AR(2) and Hansen statistic as I did throughout this GMM series.
@@CrunchEconometrix thanks prof, always great
thanks madame, please how to interprete THE method 3SLS of simultanous equaTions
Sorry Nasri, I don't have videos on 3SLS. You can do some online search on it. Thanks.
Hi Dr. Ngozi
The do file for "Two step Difference GMM" is probably missing in the Do files list, Would you pls add that to the folder.
Yes, I didn't bother because the syntax for the Two-Step Difference GMM are minimal modifications to the One-Step Difference GMM which can be gleaned from the screen. Thanks for watching. May I know from where (location) you are reaching me?
Hello Dr. Adeyele, once again, I followed your previous teachings on xtabond2 and I find them really helpful. However, I am working on my dissertation whereby I have endogenous variables in my model, I noticed you did not make any example of such, I'd like to learn about how to specify these variable in my System dynamic panel model. My N=13 and my T=9. Also, I have a question about the no of instruments, no of groups or heterogeneity and no of observations, I mean must they exceed or remain bellow the no of observation or groups?
Ur query is an evidence that you didn't watch the prerequisite videos as adviced. Do so and jot some notes while doing so...and you have your answers.
@@CrunchEconometrix Thank you very much, off I go.
problem of no observation, N=24 and T=3 years (cross-section)
Mahjoubi, check your data again to see if a country has 0 observations on a variable included in the specification. If there is, remove the country.
Hi, i couldnt open your website to download the dofiles. can you tell me how can i access please?
Hi Shiraz, my website is active. Can I know the challenge you are experiencing?
@@CrunchEconometrix Hi Adeleye, thanks to reply, i am learning how to estimate dynamic panel model in stata. and i wanted to download your dofile to try it in my data. but your website didnt open
may I know why do you put year in dummies? Does it already include in your panel data when you set up xtset? Appreciate your reply
I explained the importance of year dummies in its dedicated video in the series. Kindly watch. Yes, time dummies are created in the dataset.
Hi Dr, hope you are doing well. I have a question on t-statistic GMM. Assuming I want to compare 2 values of t-statistic on A and B, how should I interpret it, if A t-statistic is 2.94 and B is 1.94. Thanks again for your kind help.
Hi Khairul, you can base your comparison on the level of statistical significance if they have the same sign.
CrunchEconometrix Thank you Dr. Your enthusiastic to help and explain is really appreciated. May God bless you always.
You too, Sir :)
Very nice, when will you upload the sytem GMM videos?
Thanks Alex for the compliments. I'm currently working on the system GMM videos, i a little while, they'll be out.
@@CrunchEconometrix Thanks! your videos are very helpful to me. :)
@@AlexvanZoelen Gr8! Please let others know too, I know you've helped greatly in sharing my videos....don't stop, keep sharing!!! gracias! :)
@@CrunchEconometrix I will definitely recommended your channel to my fellow colleagues ! Best regards from the Netherlands
Prof. Thanks for this great insight into Two step GMM. I am running a GMM estimator using a panel of 81 countries over 23 years. And in line with the work my study is modeled after, i used three years average data. the result of System GMM indicate that both Sargan and Hansen test are highly significant. Pls what do i do to solve this problem of over identification instrument even after including the collapse option.
Hi Ugo, thanks for the positive feedback on my video, deeply appreciated. From your query, 3-year averages reduces T to about 7years? That is good. GMM procedure can be quite naughty and you may have to try different simulations: reduce regressors, reduce instruments, estimate with and w/o year dummies. Try these and see the outcomes. Also, try use the 1-step system and the difference GMMs then compare results.
thanks very much for your response I sincerely appreciate. My understanding is that System GMM produces a better result than Difference GMM. What do you have to say about that?
@@ezeugonna5017 Not quite, only in terms of number of instruments used. Both difference and system GMM yield efficient results. One of the reasons I made this GMM series is to draw attention to the relevance of difference GMM. Do a Google to see several contemporary papers on difference GMM approach.
thanks so much Prof. Please I encountered another problem with my STATA while estimating the model. It started given this error message "Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
J(): 3900 unable to allocate real [163296,2016]
_Explode(): - function returned error
_ParseInsts(): - function returned error
xtabond2_mata(): - function returned error
: - function returned error
r(3900);" please do you have solution to this problem
@@ezeugonna5017 Honestly no idea. Perhaps, you can re-check the syntax if it's incorrectly stated.
Hello, first I will like to thank you for your work. I am working on the effect of remittances in the four pillars of food security, I have a data panel of N=34 and T=6, I was thinking in applying GMM, since then I performed a good model for one-step system but when I applied other types of GMM estimations, the regressors (except the lag dependent variable) become not significant, why do you think this is happening? The model has 111 observations (I don't know if that is enough) and due to the missing values at the end only use 25 groups, but there are 22 instruments, the Hansen is chi^2=0.112 and AR(2) z=0.438
Hi Francisco, thanks for reaching out. If one-step GMM give you the results you want then go for it. If you watch my videos you will see the measures involved in changing the status of coefficients.
@@CrunchEconometrix Thank you so much for your answer! I have another question, what do you suggest if the coefficient of the lag dependent variable in the GMM model is higher than in pooled OLS but lower than in LSDV. I also preformed ivhettest and showed a presence of homoskedasticity (as I understand GMM is recommended when you have a presence of heteroskedasticity).
thank you for your efforts can I ask why the following error appears when I apply Sys GMM :
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
/ / / / / / invalid.r(198);
PLEASE NEED HELP
THANKS IN ADVANCE
Hi HH, pardon my late response. You may have written the command wrongly.
Hi! Thank you for your video! I find it very helpful why didn't I find it sooner?!! Anyway, I tried to open your website to get the do file but I cannot open your website. Is there any other way for me to ask for your do file? I really need help with my thesis right now. Thank you so much!
Hi Novia, thanks for the encouraging feedback on my videos. Deeply appreciated! 🙏 ❤️ My website is accessible cruncheconometrix.com.ng/shop/. Kindly try again.
tell me answer as quickly as possible. I also subscribed your channel.
Umar, I have responded. Thanks for your subscription. Deeply appreciated! 😊
Hello thank you. I did not find two step GMM code yet. How can I get those codes
Mahmood, didn't I say all dofiles are on my website? I also displayed this info for all to see.
thank you for your video. please why resol the mention Warning: Uncorrected two-step standard errors are unreliable after the estimation
Hi Martin, please re-phrase your query. I don't quite get what you're saying.
@@CrunchEconometrix Hi, thank you for being interested in a question, but I will be clear and precise.
1) I would like to know what is the logic in this estimate of GMM, because I have the impression that it is a method that uses trial and error. unless I made a mistake.
2) How to play with the variables so that they are not significant in AR (1), AR (2) and Hansen, but also for the interpretation of the variables while passing by the possibility of the student must be equal to P
@@martinbitsie7270 I have responded to this on a different thread.
@@CrunchEconometrix hello Professor, thank you and congratulate you all for the work you do for science. I finally understood your method. But I have another concern, if I lead a study at the Aranese scale ie 54 countries, and later I would like to make an estimate on one of the regional economic communities so how am I to take when my base is made up of 54 countries? which command to execute or is there another method? Thank you in advance for your response and good luck to you.
Hello,
Thank you for your videos, they are really helpful.
I face a problem when I implement the GMM model. My problem is that my main independent variables is omitted!! Do you have any idea what would be the problem?
Can't say exactly why that is.
@@CrunchEconometrix when I remove the year dummies it works fine. Do you think it is better to remove the years dummy?
Many Thanks
It's advisable to include year dummies...so you must have good explanations for excluding them when asked by a Reviewer.
Dear Prof. I run the model for diff GMM but I get missing values for F test and its' probability. While I drop the option noleveleq and use system GMM I get results for F test. What would be the reason for that?
Hi Nursena, no idea. You may need to post your query on Statalist.org for more constructive feedback.
Thank you for the video, Prof! But I wonder why xtabond2 couldnt be load to my stata. Which version of stata you use in this tutorial?
Hi Ruri, it's in the video title. Stata13. Have you tried installing with "ssc install xtabond2" or "findit xtabond2" or "help xtabond2". May I know from where (location) you are reaching me?
@@CrunchEconometrix im from indonesia. No, i havent tried stata13 because mine is 14.2. Should I install the package or reinstall the apps?
@@rurialifia12 Install the package. It should work.
Hi, Ms, can you give an orientation please? I have a doubt: why when you use iv(variables), the xtabond and xtabond2 use the difference of this variables as instruments instead of levels?. I read arellano and bond (1991) and i don't find any reference, it is gonna makes me crazy.
Hi Gonzalo, I think it's the way the algorithms are programmed. I honestly can't say anything about it.
@@CrunchEconometrix thank you for your soon answer. Greetings
Hi thank you for your explanation! I have a question. I used xtabond2 y l.y l2.y x1 x2 x3 gmm(x1 x2, collapse) noleveleq twostep. But I got omitted value in x3 Could you give me solution for this?
Hi Yeseul, from what you specified that code is wrong. Adapt what I used and re-estimate. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix so I change code to xtabond2 y l.y l2.y x1 x2 x3 gmm(x1 x2,eq(diff) collapse) twostep and now I got omitted cons Could you help me how to fix it? Btw from korea😁
Constants are sometimes omitted due to the econometric constructs in GMM
@@CrunchEconometrix could you explain why constant is omitted often in difference GMM?
Kindly read Roodman (2009, 2014) for detailed info about GMM.
how to run the same exemple but panel-ARDL
Watch all my videos on panel ARDL and you'll know what to do.
Dear Professor, my dependent variable is ROA, i want to use two dummy variables (time periods ) as predictors in the main equation and as instruments other variables, the syntax is like : xtabond2 roa l.roa S2 P2 , gmm(roa ,eq(diff) collapse ) iv (inflation l.gdp pen corr tax l.size prem ) twostep small robust. N=24, T=15.Am I doing well ? or Am I completely in the wrong direction?
Hi Eva, there are several specifications to the GMM technique. Advisable to use that which you understand its configuration. Thanks.
@@CrunchEconometrix Thanks a lot dear professor. !!!!
how to find the control or predetermined variable?
That is up to you on variables classification. You make the assumptions.
Is these necessary to include?
In stata 13 help menu no need to introduce th predetermined variables only time dummy work is efficiently.
Dear Professor, I am trying to follow this method but i keep getting my instruments bigger than the number of groups. and also since my variables are only stationary in the first difference, so shall i be using them in the first difference as instruments?
Lema, I didn't perform any stationary test so I have no idea of what you intend doing. Follow my procedure.
@@CrunchEconometrix yes, that's true you didn't but since my T=20 and N=30, so shall I not do unit root test?
and my instruments are only reduced if I drop the year dummies. so can I do that?
@@lemakargar6356 You know you can't.
@@lemakargar6356 Yes, you should.
Can you send me the link for GMM videos using E views
No videos yet.
No videos yet.
brvoo... thanks
Thanks Lahmar. Kindly share this with your academic community and colleagues! 💕 😊
sorry does anyone what it means when the Hansen test =.
Hi John, this evidences that one or more of your variables has so many missing observations. Check your data to know which and change them with closer proxies having more data points. Hope this helps. May I know from where (location) you are reaching me?
@@CrunchEconometrix But I run the estimates already indicating if the sample==1 doesn't that take care of that issue?
@@johnaibinu683 No it doesn't.
Hi
Hello
thank you
U're welcome, Kihal...I'll be grateful if you can help share this with your academic community and colleagues. Thanks! 💕 😊