Hi beloved guest/subscriber, you have discovered my amazing TH-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, your likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Hey, i am running xtabond2 on my stata and the estimator for my difference GMM is above the Poole OLS Estimator (Upper-bound). What should i do? Does that mean that the estimator is wrong?
@@Jing5231 The decision you have to make is very clear. It implies that the DGMM is upward biased, hence, the appropriate estimator for the model. May I know from where (location) you are reaching me?
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
Thank you so much for the efforts and resources put together to make this wonderful series. However, I have a below questions: I came across an article that uses GMM on a dataset of 42countries over 40years period. How is that possible considering the fact that the N and T are almost the same with regard to the challenge of large instruments?
Thanks for the explanation. Quick question regarding the article of Bond (2001). Can you please inform me of the title or DOI number of the article. I did not see you put it in your reference list at the last of your video. Appreciate your answer.
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
I am Dagim Tadesse from Ethiopia.First, i would like to thank you for your valuable GMM video series which helps me a lot in increasing my knowledge about dynamic panel data estimations. I would like to investigate determinants of income inequality for East Africa countries using panel date ( 2011-2016) for 7 countries. But, i faced some important problems to conduct my research using this estimation. First, to decide whether to use difference or system GMM i tried to follow the step you used but the estimation of difference GMM , both for the first and two step, omit the coefficient of lagged dependent variable. This is also the case when i estimate the model using one step and two step system GMM. What is the reason behind this? and do the sign of the coefficient has any implication on deciding the estimation method rather than the relation ship it shows as between the variables. For example, if the coefficient of the lagged dependent variable negative for the lower bound, positive for the upper bound and negative for the difference GMM. Eagerly waiting your kind response and thank you in advance!! sincerly
GMM estimation, if properly coded, MUST include the coefficient of then lagged depvar. You may need to watch this video again and follow the guide. Thanks.
Hello from Istanbul, First, I would like to thank you for your helpful tutorials. Then, I would like to know which model to choose if our lag dependent variable coefficients under difference GMM (one and two-step) are less than the ones under the fixed effect model, whereas coefficients of lag dependent variable under system are higher than the ones under fixed effect. Kind regards. Haman
Hi, I used 2 lags for estimated twostep GMM. In the first lag, coefficient FDGMM and SYSGMM are lower than FEM and lower than PLS. But, in the second lag, FDGMM and SYSGMM are higher than FEM and lower than PLS. So, what is that mean? Thank u
Hello Prof., I believe you are doing great by God's grace. I want to thank you for your helpful videos on GMM. I am using STATA 13 for my data analysis using GMM. My current challenge is that, anytime I execute the nodiffsargan command, I receive the following message from stata 13: Unrecognized command: nodiffsargan. I need your advice on what I can do to get this command to work, since I need it to run the difference and system GMM options. Thank you.
Hi Dennis, it appears you are yet to install the command. Type "help nodiffsargan" into the Command Window and follow the guide to installing the syntax.
Hello Prof, Please after typing the " help nodiffsargan" , Stata 13 told me that "help for nodiffsargan not found". What could be the problem with my Stata 13? Thank you.
Hi, thank you so much for serving the community with great knowledge. I am facing trouble in estimation. I am estimating two step difference GMM with 1 lag of my dependent variables which is not significant at 5%. However, values of Ar(2) and hansan test are acceptable. I tried increasing the lag of dependent variable along with lags of independent variables also. Still the significance level of dependent variable is not improving even tried with one step difference GMM. Does that mean I cannot use two step different GMM for estimation ? Please guide. Thanks
Hi Muhd, sometimes the explanatory variables may influence the significance of the lagged depvar. Tweak those around, change them if necessary and re-estimate.
First of all, I want to thank you for your knowledgeable videos. I have learnt a lot from these videos. However, whenever i visited your website for do files, these are now not freely available. Please do the needful so that we can have access to your do files. thanking you
Hi Mahesh, thanks for your encouraging feedback. Deeply appreciated. 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
Please answer me. looking at Rules-of-thumb1, If the dependent variable has a unit root (unstable), then difference GMM yields both a biased and inefficient estimate. So We use the system GMM to solve this. That is, if the dependent variable has a unit root, is it desirable to use SYSTEM GMM? Thanks
@@CrunchEconometrix okay. I'd really appreciate it if you could answer my one question. looking at Rules-of-thumb1, if the dependent variable has a unit root(persistent and close to being a random walk, is it desirable to use SYSTEM GMM??
Hi Fauzia, thanks for your enquiry. Kindly know that due to abuse and unethical conduct 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 The files don't cost much just a token to maintain my website. Thanks for your understanding and patronage.
Hello. How do you create the Year dummy variable please? How to solve the problem when our independent variables are not significant but instruments are OK based on Hansen result?
HI I am from Dubai. Your videos are amazing and has helped me a lot in doing my masters. I just have a question. What are the preliminary tests required to set up the panel data for conducting the GMM models. Thank you in advance.
Hi Sukh, thanks for the encouraging words and feedback. Deeply appreciated! All you need to do are laid out in the 9 GMM series. Follow the guide and you will be fine.
Dear Professor your tutorial videos help me in understanding and running GMM. It would be an honor for me if you would specify, why researchers use GMM instead of 2SLS?. somewhere at the STATA forum, I found they run the Hausman test to decide between 2SLS and GMM but I could not understand the output of the same. can you help me in this regard? Thanks and I am from Pakistan.
Hi Prof. Your vidoes are really helpful. I watch them frequently. My results of lag depvar are as follows
Fixed effects .0459216 Pooled OLS .0803209 Onestep diff gmm .6883005 Two step diff gmm .6883005 Onestep system gmm .1947425 Two step system gmm .1224242 Which of the four do you think it suggests: (i) one step system gmm (ii) two step system gmm (iii) first diff gmm (iv) two step diff gmm? Given the N=31, T=13. Another question in practice we use different models not one in our research, so should we use the same criteria on every model separately and what if two models qualify for two different gmm?
Hi Xubairxhan, thanks for the positive feedback. Deeply appreciated! I made very clear explanations in the video. You may need to watch again to take your decision. Thanks.
This is a general code for two step difference gmm, xtabond2 y L.y L2.y x1 L.x2 L(0/2).(x2 x3) i.year, gmm(L.y) iv(x1 L.x1 L(0/2).(x2 x3) i.year) nolevel twostep Please guide me if i have 7 (X variables) and 4 Control variables. how can i incorporate them in the above-mentioned code?
Hi Stephen, here is it: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
Dear Professor. Thank you a lot for the videos. They are very helpfull. I kindly ask you for an advice in my situation. Im trying to know if inequality and redistribution are harmfull for economic growth. Im using data of 18 countries and 7 five years periods. My dependent variable is 5 year growth rate (yt-yt-5), and independent variables are lag of the per capita product (yt-5), current and past net inequality (lag), redistribution, invesment (% GDP), public consumption (% GDP), trade opennes (% GDP), human capital, ln fertility, ln life expectations, rule of law and 7 year dummies (18 independent variables). Im following the steps of this video, and I got 0.014 as OLS coefficient of the lag of the dependent variable (not significat an any level); -0.43 (significat at 1%) in LSDV estimation, -0.92 in difference GMM one step (significant at 10%) and 0.111 (not signiticant an any level) in system GMM one step with the following specificaction: " xtabond2 dy ylag gdisp gdisplag absred i g xm h lnfert lnexp polright q1985 q1990 q1995 q2000 q2005 q2010 q2015 , gmm(ylag, collapse) iv(gdisp gdisplag absred i g xm h lnfert lnexp polright q1985 q1990 q1995 q2000 q2005 q2010 q2015, equation(level)) robust orthogonal small". With system GMM a lot of variables were significant, and the tests for instruments validity, overr restriction and AR(2) were aceptable. But instruments (20) >groups (18), even with collapse option. Considering this situation, the output I got from system GMM is not valid? what can i do to solve this situation? Thanks a lot for your attention.
Hi Nicholás, GMM technique is not a one-shot technique. You may need to engage several simulations before arrival at the final model that passes ALL diagnostics. Try this: reduce variables in the iv() set and re-estimate.
Thank you so much for the insightful video. I have a question: If the one-step and two-step GMM coefficients of the lagged dependent variable give conflicting results with regards to the fixed effects, for example if one of them (say 2-step gmm) has a coefficient (say 0.5) above that of the coefficient of the fixed effect (say 0.45) while the other (1-step gmm) has a coeffieicnt (say 0.43) below the coefficient of the fixed effect, how do you make your decision? That is the coefficient from the gmm estimates are not collectively below the coefficient of the fixed effect.Thank you.
Thanks Iddriss for the positive feedback, however, you're making it unnecessarily complicated. I was very clear on how to decide on which estimator to use. You are to make your decision on the basis of EITHER one- OR two-step GMM vs. FE. As a researcher, you can use any variants of either difference or system GMM. My videos are quite clear and straight forward on these. Thanks.
Hi, thank you so much for your videos professor. However, I still get confused with 3sls and GMM. Do 3sls and GMM (system GMM or difference GMM) are the same? because in my stata guidebook stated that " 3sls is a two step GMM estimator". Thank you in advance.
@@CrunchEconometrix i see, so if i am researching simultaneous equation, which one is better between 3sls or system/difference gmm sir? sorry my english is wiite bad because im from indonesia. Thank you so much sir
@@CrunchEconometrix Thank you Sir. However, I want to ask about the syntax of OLS Regression. If we are about to use GMM in the end, why do we need to robust our variable in OLS Regression while choosing between System/ Difference GMM? because I think GMM will also robust our variable eventually. Thank you so much sir
Hi, I followed your instructions as per the video to identify whether to use difference or the system GMM. As per the results, the system GMM can be run to analyze the data. Then I used system GMM to analyze the data. But P values are not significant for none of the independent variables or the control variables except the lag value of the dependent variable. Could you please advise on this matter. Thank you
Hello Prof. Just want to ask where can I find the exact article for rule of thumb-2. I tried to find it, unfortunately I cannot found it. To others who also have info on this, appreciate if you can share it with me. Thank you.
Hello.Great video! I am currently trying to run DGMM on an unbalanaced dynamic panel. My issue is that the arellano bond test for autocorrelation indicated that there is at least autocorrelation of order 7 in the lagged errors. Should I apply SGMM arguing that the process is very persistent ? If I try to correct including too many lagged values of the dependent variable as controls I am afraid I can loose too many observations. Many thanks
Hi CrunchEconometrix, again very helpful and informative video thank you so much! Quick question, if my one step difference GMM is close to the fixed effects estimate but my two step difference GMM is well within the bound of OLS and fixed effects, should I use two step difference GMM or system GMM? Thanks!
Thank you for the comprehensive video. Maybe one stupid question: if difference model doesn't work (Hansen test: 0,09) and system model works (all tests Ok, including Hansen: 0,144 ) is it necessary at all to conduct the deciding?
Hello professor, Thank you very much for your helpful videos. They are very clear and simple. I have two questions, example if we decide to use System GMM (comparing the coefficient of variables like in this video), after this what we have to test? One-step and two-step GMM?? Or just to do the commands: two-step robust orthogonal small (With and Without collapse option). The second question is when we do the OLS and FE should we do any test for them or not? Thnx again for your email and your videos. This is the first time that I am using the GMM but it was easy with your videos.
Hi Albulena, thanks for the positive feedback on my videos. Deeply appreciated! Decide the variant of GMM you want to engage and watch my videos on them on how to proceed. To use FE, the Hausman test will guide you. Watch my video on that too. For pooled OLS, watch my videos on LSDV for guidance.
@@CrunchEconometrix Professor thnx for your ans. I watched your vidoes :), but I have to chose to do ex. the SYS GMM, after I compare the coff. of lag variable, should I start from begning to do all stpes that you did for first and second SYS GMM. Sorry If I was not clear in first question.
Hi, Thank you so much for sharing your knowledge with us. These videos are very useful. I have a doubt to be clarified. In my analysis, there are three dependent variables and four independent variables. When I am deciding between Difference GMM and system GMM. Do I have to run models, i.e., OLS, FE & Difference GMM for variables mentioned above separately or is there another method for that. Could you please clarify this issue for me. Thank you :)
@@CrunchEconometrix Thank you so much for your kind reply. Is there a particular paper regarding this, which I can refer to and include in my thesis as a reference? If you have any, could you please share the name or the link? Thank you :)
@@neleemuthukumarana2492 I was taught by my econometrics professor while in the UK. Who knows you may find some references, aside from those I indicated in the video. Thanks!
@@CrunchEconometrix actually I study economics, PhD in Turkey but I am from Madagascar. I try to increase my knowledge on difference and system GMM in order to be able to give sufficient explanation in why I chose a specific model.
@@rmdrayfelix7157 Good, I hope you learnt a few hints from my videos. I will appreciate if you can help share my TH-cam Channel link with your friends, students and academic community in Turkey 🇹🇷 and Madagascar 🇲🇬 too. Thanks! 😊
Hi .i want to take your help for understanding Gmm.if in modal T is greater N .can i apply gmm modal in data set.but on the other side in my modal lag dependency here.
Hi Hamza, the basics of GMM are explained in the first video, advisable to watch it. Also, watch my video on "Basics of Panel ARDL" and you will be better guided. Thanks.
Thank you for your useful videos , i will estimate my model with fixed effect then with system GMM so stationarity check is important for system GMM and fixed effect? my model is already with first difference following Arellano and Bond (1991) and all variables are stationar in this difference i mean that all variables are integrated after first difference, but how to write that in my paper, all variables are integrated in I(o) or I(1) since my model is already with this difference? Thank you
Hi Samia, thanks for the positive feedback on my videos. Deeply appreciated. Due to the short-time span, stationarity of the variables is not required. Kindly watch ALL the GMM series again....may I know from where (location) you are reaching me?
Really helpful and informative video. Madam, if you can kindly provide the full reference of the Bond(2001) paper that you have mentioned in the video. I need to cite the paper on deciding system GMM against difference GMM in my thesis
Thanks Manish for the positive feedback on my videos. Deeply appreciated. Check the 1st video for selected references and if it's not there, you can do a Google search for it. Thanks.
@@CrunchEconometrix Thank you for your immediate response. I did check the first video, but I am afraid, I can't find it there. I did google as well but by citation alone without the title and other information, its really difficult to find the paper. Please provide the detail of the paper to substantiating my choice of following System GMM in my study. Thank You.
@@ManishKumar-cl7eg Manish, give the usual justifications for using system GMM and rest your case. No one will ask for references on choosing between difference or system GMM, at least I haven't come across such and I have 2 Scopus papers where I used system GMM.
I am looking for this reference also. I have searched the internet everywhere and cannot find this rule of thumb . Could you please provide the reference of the Bond (2001) paper so I can reliably include it in my thesis. Thank you.
Bond, S. R. (2002). Dynamic panel data models: a guide to micro data methods and practice. Portuguese economic journal, 1(2), 141-162. Just Check this out
Hi Yusibov, thanks for the encouraging feedback on my videos. Deeply appreciated! Unfortunately, my slides are not available to the public but you can get some notes from the video description. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix thank you very much for replying. Yes I think most of people may need some notes related to formulas and interpretation and brief steps to run models. I am Avaz Yusibov. I am from Azerbaijan
Thanks Avaz , my Channel is very different from others because I show the steps needed to carry out EVERY procedure. No other Channel come close because of the time I devote to making a video. So, kindly bear with my resource. I'm sure every student after watching any of my clips will be a lot better off than they were before (lol). Please tell your colleagues about my Channel and do stay safe.
Hi Nelee, your query shows you did not watch the prerequisite videos in this GMM series. Skipping videos will deny you of salient information. Watch all the videos in the arranged order and read Roodman (2009, 2014) for better understanding of the procedure. Regards.
Hey, i am running xtabond2 on my stata and the estimator for my difference GMM is above the Poole OLS Estimator (Upper-bound). What should i do? Does that mean that the estimator is wrong?
Hello Professor, Is it possible to run GMM if I have . N =16 and T=10 years ( however I changed my data from yearly to quarterly), so T=40 instead of 10.
Hi, my name is Fendi and im from Indonesia. Currently im witing a thesis about financial inclusion using GMM model. First of all i would like to say thank you for this GMM explanation. I have a question for you, my dataset has 120 observation, consist of 20 T and 6 N. my result in hansen test is 1, i follow your command to reduce the hansen test value but the result didn't change. Is this because my dataset has larger T than N or is there any way to decrease the hansen value? Thank you :)
In this you used the difference model with noleveleq option but i had obtained a better result without it keeping everything else in the model intact, so does that mean that I should use my model without nolevel equation for comparison?
@@CrunchEconometrix Just one more question, the y* you obtained, do i have to estimate it differently or its just the range of time dummies? yr_1-yr_14 = y* ? Is that it? Apologies for asking toomany questions :D
@@zuzia6021 Yes, dofiles are no longer free but available upon payment for a one-time download. I took the decision due to several attempts to hack into my Google drive.
Hi Crunch, I hope you are fine, I want to ask you about Sargan and hansan test when I want to decide between diff and sys. are they important or no? and what is the range of the sargan to accept? thank you in advance Ali
Hi there, from what I have gleaned from the literature when it comes to GMM, Hansen test is more relevant than Sargan. I emphasized these in my videos and will also refer you to Roodman (2009). Thanks.
@@CrunchEconometrix thanks madam for quick reply, when i used your code i just received individual year dummies and i didnot get any y* variable, so my questiin is that how can i develop y* variable for controlling all year dummies
Hi Agung, due to abuse dofiles are no longer free but available on my website upon payment of a token. Here's the link cruncheconometrix.com.ng/shop. Thanks 😊
Hi Crunch, Really I liked your videos your explaining of the Gmm, Honestly, I am a new in gmm and I wanted to choose between Sys and Diff. but I don't have good experience of that, However, I used your command in my stata but doesn't working with me I think it's my mistak, Could you help how to use it ? T=22 and N=50 Thank you in Avnance
Thanks for the positive feedback on my videos. Deeply appreciated! 💕 GMM may break down because the time span is long. This is because GMM is built for short panels where N is excessively larger than T and T is less than 20 years.
@@CrunchEconometrix Thank you for your answer, I removed one year. T=21 N=50, Is it necessary to use a dummy and lag? I already transformed my variables to Natural logarithm...what do you think?
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
@@CrunchEconometrix Dear prof, couple of additional questions: Do you use log_transformation of variables in order to reduce the effect of outliers or in order to ensure stationarity? How can I donate to your channel?
Hi Ksenija, yes that is one of the merits of using the log transformation...and thanks for the willingness to donate to my TH-cam Channel. I am in the process of getting approved for that from TH-cam to avoid sanctions. Kind regards.
Hi Linican, I don't know of any textbook that covered GMM but there are journal articles to guide you some of which I listed at the end of each GMM video.
@@sameenaftab8781 Awesome!!! Kindly share my videos with your, friends, students and academic community in Pakistan and on social media for awareness. They'll learn some usefull tips and skills too! :)
Case I have firms level panel data, financial in nature firms level data, like log of all the series and run OLS and FE and GMM system and difference my sargan test Pvalue is 0.0000 and hansen test pvlaue same 0.0000 AR(2) Pvalue again 0.000 how I can manage the situations that this pvalue should be good according to litrature
Hi Sarfraz, significant pvalues implies that the model is improperly identified which has to do with your choice of instruments. Kindly watch the series in sequential order to guide you on how to address this problem.
Hi beloved guest/subscriber, you have discovered my amazing TH-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, your likes, feedbacks and sharing my videos with your cohorts. Follow me on Facebook, Twitter and Reddit. Love you all, greatly!!!
Thank you most sincerely. This video helps me greatly. Waiting for more.
Hey, i am running xtabond2 on my stata and the estimator for my difference GMM is above the Poole OLS Estimator (Upper-bound). What should i do? Does that mean that the estimator is wrong?
@@Jing5231 The decision you have to make is very clear. It implies that the DGMM is upward biased, hence, the appropriate estimator for the model. May I know from where (location) you are reaching me?
@@CrunchEconometrix please share the complete reference for "Bond 2001" who set this rule of thumb
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
Excellent again! Eagerly waiting for the next videos.
Thanks for the encouragement, Moin! Keep watching, keep sharing...gracias! :)
Good information about GMMteaching
Thanks for the positive feedback, Gunim. Deeply appreciated! May I know from where (location) you are reaching me?
Thank you so much for the efforts and resources put together to make this wonderful series.
However, I have a below questions:
I came across an article that uses GMM on a dataset of 42countries over 40years period. How is that possible considering the fact that the N and T are almost the same with regard to the challenge of large instruments?
To be honest, I have no answer. The guides for estimating GMM are detailed in the literature.
Thanks, you're a good teacher!
You are welcome, Kiran...thanks!
Thank you very much for your videos. They are simply very helpful ......
U're welcome Mahabub...kindly share my Channel links with your students and colleagues :)
U're welcome Mahabub...kindly share my Channel links with your students and colleagues :)
Thanks for the explanation. Quick question regarding the article of Bond (2001). Can you please inform me of the title or DOI number of the article. I did not see you put it in your reference list at the last of your video. Appreciate your answer.
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
@@CrunchEconometrix Can you please state on which page exactly? I read the article but could not find it
@TheMarinho1 Keep reading from start to finish. You'll find it.
I am Dagim Tadesse from Ethiopia.First, i would like to thank you for your valuable GMM video series which helps me a lot in increasing my knowledge about dynamic panel data estimations. I would like to investigate determinants of income inequality for East Africa countries using panel date ( 2011-2016) for 7 countries. But, i faced some important problems to conduct my research using this estimation. First, to decide whether to use difference or system GMM i tried to follow the step you used but the estimation of difference GMM , both for the first and two step, omit the coefficient of lagged dependent variable. This is also the case when i estimate the model using one step and two step system GMM. What is the reason behind this? and do the sign of the coefficient has any implication on deciding the estimation method rather than the relation ship it shows as between the variables. For example, if the coefficient of the lagged dependent variable negative for the lower bound, positive for the upper bound and negative for the difference GMM. Eagerly waiting your kind response and thank you in advance!!
sincerly
GMM estimation, if properly coded, MUST include the coefficient of then lagged depvar. You may need to watch this video again and follow the guide. Thanks.
Hello from Istanbul,
First, I would like to thank you for your helpful tutorials. Then, I would like to know which model to choose if our lag dependent variable coefficients under difference GMM (one and two-step) are less than the ones under the fixed effect model, whereas coefficients of lag dependent variable under system are higher than the ones under fixed effect.
Kind regards.
Haman
Haman, kindly adapt the steps given in this video to your study to inform your decision. Thanks.
Hi, I used 2 lags for estimated twostep GMM. In the first lag, coefficient FDGMM and SYSGMM are lower than FEM and lower than PLS. But, in the second lag, FDGMM and SYSGMM are higher than FEM and lower than PLS. So, what is that mean? Thank u
I explained that in the video. You may need to watch it again, thanks.
an interesting video for GMM
Thanks for the positive feedback, Endalk. Deeply appreciated! 😊
Hi, if the result of hausman test is random effect, do we again implement system-GMM with random effect result?
Random effect model is a STATIC model while GMM is a DYNAMIC model. Yes, you can engage both analysis to observe if the results are consistent.
Hello Prof.,
I believe you are doing great by God's grace. I want to thank you for your helpful videos on GMM.
I am using STATA 13 for my data analysis using GMM. My current challenge is that, anytime I execute the nodiffsargan command, I receive the following message from stata 13: Unrecognized command: nodiffsargan.
I need your advice on what I can do to get this command to work, since I need it to run the difference and system GMM options. Thank you.
Hi Dennis, it appears you are yet to install the command. Type "help nodiffsargan" into the Command Window and follow the guide to installing the syntax.
@@CrunchEconometrix Well noted Prof. Thanks a million
Hello Prof,
Please after typing the " help nodiffsargan" , Stata 13 told me that "help for nodiffsargan not found". What could be the problem with my Stata 13? Thank you.
Post on Statalist org for constructive feedback. I don't use the syntax so can't help you with that.
Hi, thank you so much for serving the community with great knowledge.
I am facing trouble in estimation. I am estimating two step difference GMM with 1 lag of my dependent variables which is not significant at 5%. However, values of Ar(2) and hansan test are acceptable. I tried increasing the lag of dependent variable along with lags of independent variables also. Still the significance level of dependent variable is not improving even tried with one step difference GMM.
Does that mean I cannot use two step different GMM for estimation ? Please guide. Thanks
Hi Muhd, sometimes the explanatory variables may influence the significance of the lagged depvar. Tweak those around, change them if necessary and re-estimate.
First of all, I want to thank you for your knowledgeable videos. I have learnt a lot from these videos. However, whenever i visited your website for do files, these are now not freely available. Please do the needful so that we can have access to your do files. thanking you
Hi Mahesh, thanks for your encouraging feedback. Deeply appreciated.
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
Please answer me.
looking at Rules-of-thumb1,
If the dependent variable has a unit root (unstable), then difference GMM yields both a biased and inefficient estimate.
So We use the system GMM to solve this.
That is, if the dependent variable has a unit root, is it desirable to use SYSTEM GMM?
Thanks
Jun, so what is the issue?
So is it correct if the dependent variable has a unit root, is it desirable to use SYSTEM GMM??
@@CrunchEconometrix So is it correct if the dependent variable has a unit root, is it desirable to use SYSTEM GMM??
Jun, did you watch the clip? If yes, support that with reading references listed at the end of the video.
@@CrunchEconometrix okay. I'd really appreciate it if you could answer my one question.
looking at Rules-of-thumb1, if the dependent variable has a unit root(persistent and close to being a random walk, is it desirable to use SYSTEM GMM??
Where can I download your do file . please guide.
Hi Fauzia, thanks for your enquiry.
Kindly know that due to abuse and unethical conduct 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
The files don't cost much just a token to maintain my website.
Thanks for your understanding and patronage.
Hello. How do you create the Year dummy variable please? How to solve the problem when our independent variables are not significant but instruments are OK based on Hansen result?
Hi Bty, I covered these in my GMM series. Please find the time to watch all. Do not skip any. Thanks.
Hi, I would like to ask what if I got the smaller coefficient from pooled OLS than that of FE?
That is unlikely, Cao. But if it occurs perform the difference GMM.
HI I am from Dubai. Your videos are amazing and has helped me a lot in doing my masters. I just have a question. What are the preliminary tests required to set up the panel data for conducting the GMM models. Thank you in advance.
Hi Sukh, thanks for the encouraging words and feedback. Deeply appreciated! All you need to do are laid out in the 9 GMM series. Follow the guide and you will be fine.
Dear Professor your tutorial videos help me in understanding and running GMM. It would be an honor for me if you would specify, why researchers use GMM instead of 2SLS?. somewhere at the STATA forum, I found they run the Hausman test to decide between 2SLS and GMM but I could not understand the output of the same. can you help me in this regard? Thanks and I am from Pakistan.
Hi Abdul, kindly watch my introductory GMM video for insights about the technique. Well explained. Thanks.
Hi Prof.
Your vidoes are really helpful. I watch them frequently.
My results of lag depvar are as follows
Fixed effects .0459216
Pooled OLS .0803209
Onestep diff gmm .6883005
Two step diff gmm .6883005
Onestep system gmm .1947425
Two step system gmm .1224242
Which of the four do you think it suggests: (i) one step system gmm (ii) two step system gmm (iii) first diff gmm (iv) two step diff gmm? Given the N=31, T=13.
Another question in practice we use different models not one in our research, so should we use the same criteria on every model separately and what if two models qualify for two different gmm?
Hi Xubairxhan, thanks for the positive feedback. Deeply appreciated! I made very clear explanations in the video. You may need to watch again to take your decision. Thanks.
This is a general code for two step difference gmm,
xtabond2 y L.y L2.y x1 L.x2 L(0/2).(x2 x3) i.year, gmm(L.y) iv(x1 L.x1 L(0/2).(x2 x3) i.year) nolevel
twostep
Please guide me if i have 7 (X variables) and 4 Control variables. how can i incorporate them in the above-mentioned code?
Hi Sohaib, I gave clear and practical examples. Watch my GMM videos and adapt what I did.
Thanks for the videos. What is the full reference for Bond 2001? I searched but have not yet found it.
Hi Stephen, here is it: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
Dear Professor. Thank you a lot for the videos. They are very helpfull. I kindly ask you for an advice in my situation. Im trying to know if inequality and redistribution are harmfull for economic growth. Im using data of 18 countries and 7 five years periods. My dependent variable is 5 year growth rate (yt-yt-5), and independent variables are lag of the per capita product (yt-5), current and past net inequality (lag), redistribution, invesment (% GDP), public consumption (% GDP), trade opennes (% GDP), human capital, ln fertility, ln life expectations, rule of law and 7 year dummies (18 independent variables). Im following the steps of this video, and I got 0.014 as OLS coefficient of the lag of the dependent variable (not significat an any level); -0.43 (significat at 1%) in LSDV estimation, -0.92 in difference GMM one step (significant at 10%) and 0.111 (not signiticant an any level) in system GMM one step with the following specificaction: " xtabond2 dy ylag gdisp gdisplag absred i g xm h lnfert lnexp polright q1985 q1990 q1995 q2000 q2005 q2010 q2015 , gmm(ylag, collapse) iv(gdisp gdisplag absred i g xm h lnfert lnexp polright q1985 q1990 q1995 q2000 q2005 q2010 q2015, equation(level)) robust orthogonal small". With system GMM a lot of variables were significant, and the tests for instruments validity, overr restriction and AR(2) were aceptable. But instruments (20) >groups (18), even with collapse option. Considering this situation, the output I got from system GMM is not valid? what can i do to solve this situation? Thanks a lot for your attention.
Hi Nicholás, GMM technique is not a one-shot technique. You may need to engage several simulations before arrival at the final model that passes ALL diagnostics. Try this: reduce variables in the iv() set and re-estimate.
Thank you so much for the insightful video. I have a question: If the one-step and two-step GMM coefficients of the lagged dependent variable give conflicting results with regards to the fixed effects, for example if one of them (say 2-step gmm) has a coefficient (say 0.5) above that of the coefficient of the fixed effect (say 0.45) while the other (1-step gmm) has a coeffieicnt (say 0.43) below the coefficient of the fixed effect, how do you make your decision? That is the coefficient from the gmm estimates are not collectively below the coefficient of the fixed effect.Thank you.
Thanks Iddriss for the positive feedback, however, you're making it unnecessarily complicated. I was very clear on how to decide on which estimator to use. You are to make your decision on the basis of EITHER one- OR two-step GMM vs. FE. As a researcher, you can use any variants of either difference or system GMM. My videos are quite clear and straight forward on these. Thanks.
Hi, thank you so much for your videos professor. However, I still get confused with 3sls and GMM. Do 3sls and GMM (system GMM or difference GMM) are the same? because in my stata guidebook stated that " 3sls is a two step GMM estimator". Thank you in advance.
Jasmine, you just said you read it from a Stata article. What validation are you still requesting?
@@CrunchEconometrix i see, so if i am researching simultaneous equation, which one is better between 3sls or system/difference gmm sir? sorry my english is wiite bad because im from indonesia. Thank you so much sir
Your English is ok, Jasmine. My suggestion: review related studies to know the best technique to deploy.
@@CrunchEconometrix Thank you Sir. However, I want to ask about the syntax of OLS Regression. If we are about to use GMM in the end, why do we need to robust our variable in OLS Regression while choosing between System/ Difference GMM? because I think GMM will also robust our variable eventually. Thank you so much sir
Jasmine, the "robust" option controls for heteroscedasticity.
Hi, I followed your instructions as per the video to identify whether to use difference or the system GMM. As per the results, the system GMM can be run to analyze the data. Then I used system GMM to analyze the data. But P values are not significant for none of the independent variables or the control variables except the lag value of the dependent variable. Could you please advise on this matter. Thank you
Hi Nelee, if that is the case, use the approach that gives you a better result and give a brief explanation for it.
Hi, Could you please explain the rationale behind using l.lngdp as an independent variable in the Pooled OLS & Fixed Effect model. Thank you
Because the lagged depvar is a regressor in GMM. Hence, the reason for performing this test.
@@CrunchEconometrix Thank you :)
Hello Prof. Just want to ask where can I find the exact article for rule of thumb-2. I tried to find it, unfortunately I cannot found it. To others who also have info on this, appreciate if you can share it with me. Thank you.
Hi Diyama, do a Google search. Simply type in what you want into the search engine and you will get related articles.
Hello.Great video! I am currently trying to run DGMM on an unbalanaced dynamic panel. My issue is that the arellano bond test for autocorrelation indicated that there is at least autocorrelation of order 7 in the lagged errors. Should I apply SGMM arguing that the process is very persistent ? If I try to correct including too many lagged values of the dependent variable as controls I am afraid I can loose too many observations. Many thanks
What's your Hansen and AR(2) statistics?
@@CrunchEconometrix Hansen p value is = 0.065 and AR(2) p value = 0.005
@@0Luis318 Hansen looks ok at the 5% significance level. The problem is with autocorrelation. Use sys-GMM and compare the outcomes.
Hi CrunchEconometrix, again very helpful and informative video thank you so much! Quick question, if my one step difference GMM is close to the fixed effects estimate but my two step difference GMM is well within the bound of OLS and fixed effects, should I use two step difference GMM or system GMM?
Thanks!
Hi Luca, I'm surprised you still asked this question having indicated it on my slides and also demonstrated it. Use difference GMM.
Thank you for the comprehensive video. Maybe one stupid question: if difference model doesn't work (Hansen test: 0,09) and system model works (all tests Ok, including Hansen: 0,144 ) is it necessary at all to conduct the deciding?
You can skip the decider.
If the coefficient of the lagged dependent variable for difference GMM is greater than that in pooled regression, then what to do?
Hi Abhijit, I do not have the rule of thumb for that. You may want to check out other online resources. Thanks.
Hello madam, is it possible to do a hausman test in the GMM model? If so, how?
Hi Fabrice, Hausman test is not necessary for GMM analysis.
Hello professor,
Thank you very much for your helpful videos. They are very clear and simple.
I have two questions, example if we decide to use System GMM (comparing the coefficient of variables like in this video), after this what we have to test? One-step and two-step GMM?? Or just to do the commands: two-step robust orthogonal small (With and Without collapse option).
The second question is when we do the OLS and FE should we do any test for them or not?
Thnx again for your email and your videos. This is the first time that I am using the GMM but it was easy with your videos.
Hi Albulena, thanks for the positive feedback on my videos. Deeply appreciated! Decide the variant of GMM you want to engage and watch my videos on them on how to proceed. To use FE, the Hausman test will guide you. Watch my video on that too. For pooled OLS, watch my videos on LSDV for guidance.
@@CrunchEconometrix Professor thnx for your ans. I watched your vidoes :), but I have to chose to do ex. the SYS GMM, after I compare the coff. of lag variable, should I start from begning to do all stpes that you did for first and second SYS GMM. Sorry If I was not clear in first question.
Hi, Thank you so much for sharing your knowledge with us. These videos are very useful. I have a doubt to be clarified. In my analysis, there are three dependent variables and four independent variables. When I am deciding between Difference GMM and system GMM. Do I have to run models, i.e., OLS, FE & Difference GMM for variables mentioned above separately or is there another method for that. Could you please clarify this issue for me. Thank you :)
Hi Nelee, 3 depvars imply 3 models.
@@CrunchEconometrix Thank you so much for your kind reply. Is there a particular paper regarding this, which I can refer to and include in my thesis as a reference? If you have any, could you please share the name or the link? Thank you :)
@@neleemuthukumarana2492 I was taught by my econometrics professor while in the UK. Who knows you may find some references, aside from those I indicated in the video. Thanks!
I like your video. Thank you for sharing your knowledge with us. Please where can I find the do file for the command?
I got it from your website
cruncheconometrix.com.ng/stata-dofiles/
@@rmdrayfelix7157 Nice RM! 💕 😊 May I know from where (location) you are reaching me?
@@CrunchEconometrix actually I study economics, PhD in Turkey but I am from Madagascar. I try to increase my knowledge on difference and system GMM in order to be able to give sufficient explanation in why I chose a specific model.
@@rmdrayfelix7157 Good, I hope you learnt a few hints from my videos. I will appreciate if you can help share my TH-cam Channel link with your friends, students and academic community in Turkey 🇹🇷 and Madagascar 🇲🇬 too. Thanks! 😊
Hi .i want to take your help for understanding Gmm.if in modal T is greater N .can i apply gmm modal in data set.but on the other side in my modal lag dependency here.
Hi Hamza, the basics of GMM are explained in the first video, advisable to watch it. Also, watch my video on "Basics of Panel ARDL" and you will be better guided. Thanks.
Thank you for your useful videos , i will estimate my model with fixed effect then with system GMM so stationarity check is important for system GMM and fixed effect? my model is already with first difference following Arellano and Bond (1991) and all variables are stationar in this difference i mean that all variables are integrated after first difference, but how to write that in my paper, all variables are integrated in I(o) or I(1) since my model is already with this difference? Thank you
Hi Samia, thanks for the positive feedback on my videos. Deeply appreciated. Due to the short-time span, stationarity of the variables is not required. Kindly watch ALL the GMM series again....may I know from where (location) you are reaching me?
@@CrunchEconometrix Thank you dear professor for your rapid reply i am from Algeria
Awesome Samia!...my love to fellow Algerians and please share my TH-cam Channel link with your students and the academic community in Algeria. Thanks.
Really helpful and informative video. Madam, if you can kindly provide the full reference of the Bond(2001) paper that you have mentioned in the video. I need to cite the paper on deciding system GMM against difference GMM in my thesis
Thanks Manish for the positive feedback on my videos. Deeply appreciated. Check the 1st video for selected references and if it's not there, you can do a Google search for it. Thanks.
@@CrunchEconometrix Thank you for your immediate response.
I did check the first video, but I am afraid, I can't find it there. I did google as well but by citation alone without the title and other information, its really difficult to find the paper.
Please provide the detail of the paper to substantiating my choice of following System GMM in my study.
Thank You.
@@ManishKumar-cl7eg Manish, give the usual justifications for using system GMM and rest your case. No one will ask for references on choosing between difference or system GMM, at least I haven't come across such and I have 2 Scopus papers where I used system GMM.
I am looking for this reference also. I have searched the internet everywhere and cannot find this rule of thumb . Could you please provide the reference of the Bond (2001) paper so I can reliably include it in my thesis. Thank you.
Bond, S. R. (2002). Dynamic panel data models: a guide to micro data methods and practice. Portuguese economic journal, 1(2), 141-162.
Just Check this out
Hello. Your videos are excellent. Can you please share ppt that you use within explanation
Hi Yusibov, thanks for the encouraging feedback on my videos. Deeply appreciated! Unfortunately, my slides are not available to the public but you can get some notes from the video description. Please may I know from where (location) you are reaching me?
@@CrunchEconometrix thank you very much for replying. Yes I think most of people may need some notes related to formulas and interpretation and brief steps to run models.
I am Avaz Yusibov. I am from Azerbaijan
Thanks Avaz , my Channel is very different from others because I show the steps needed to carry out EVERY procedure. No other Channel come close because of the time I devote to making a video. So, kindly bear with my resource. I'm sure every student after watching any of my clips will be a lot better off than they were before (lol). Please tell your colleagues about my Channel and do stay safe.
@@CrunchEconometrix okay. Thank you very much for explaining well
Could you please also explain the reason for only using l.lngdp as the GMM variable? I.e. gmm(l.lngdp,collapse)
Hi Nelee, your query shows you did not watch the prerequisite videos in this GMM series. Skipping videos will deny you of salient information. Watch all the videos in the arranged order and read Roodman (2009, 2014) for better understanding of the procedure. Regards.
Thank you very much!!
U're welcome, Young. Please may I know from where (location) you are reaching me?
Hey, i am running xtabond2 on my stata and the estimator for my difference GMM is above the Poole OLS Estimator (Upper-bound). What should i do? Does that mean that the estimator is wrong?
Hi Jing, since the outcome is not in line with the assumptions, go ahead and use system GMM...may I know from where (location) you are reaching me?
Hello Professor,
Is it possible to run GMM if I have .
N =16 and
T=10 years ( however I changed my data from yearly to quarterly), so T=40 instead of 10.
Hi Jaber, I think I have responded to you on this.
Hi, my name is Fendi and im from Indonesia. Currently im witing a thesis about financial inclusion using GMM model. First of all i would like to say thank you for this GMM explanation. I have a question for you, my dataset has 120 observation, consist of 20 T and 6 N. my result in hansen test is 1, i follow your command to reduce the hansen test value but the result didn't change. Is this because my dataset has larger T than N or is there any way to decrease the hansen value? Thank you :)
Hi Fendiawan, GMM is applicable to N > T panels.
@@CrunchEconometrix Thank you for your answer. so i can't use GMM if my dataset has T > N?
But I responded to that. I expected that you watch the foundational GMM video to familiarize with the technique. Tx.
In this you used the difference model with noleveleq option but i had obtained a better result without it keeping everything else in the model intact, so does that mean that I should use my model without nolevel equation for comparison?
Of course. We show our best results.
@@CrunchEconometrix Just one more question, the y* you obtained, do i have to estimate it differently or its just the range of time dummies? yr_1-yr_14 = y* ? Is that it? Apologies for asking toomany questions :D
I looked at your website for the dofiles, it wasn't free as you had mentioned so I couldnt find how you estimated it
@@zuzia6021 y* represents the year dummies. Please may I know from where (location) you are reaching me?
@@zuzia6021 Yes, dofiles are no longer free but available upon payment for a one-time download. I took the decision due to several attempts to hack into my Google drive.
Hi Crunch, I hope you are fine, I want to ask you about Sargan and hansan test when I want to decide between diff and sys. are they important or no? and what is the range of the sargan to accept?
thank you in advance
Ali
Hi there, from what I have gleaned from the literature when it comes to GMM, Hansen test is more relevant than Sargan. I emphasized these in my videos and will also refer you to Roodman (2009). Thanks.
@@CrunchEconometrix thank you for your response..
@@economiclibya5069 No worries😊. Keep watching...keep sharing! 😄
hello, I have developed year dummy following your code but how you generate y* in your model
From the results, you'll see that y* represents the year dummies.
@@CrunchEconometrix thanks madam for quick reply, when i used your code i just received individual year dummies and i didnot get any y* variable, so my questiin is that how can i develop y* variable for controlling all year dummies
@@lutfaferdous8236 Please pay attention to my response. See my data editor and the code, then figure it out
can I have the do files?
Hi Agung, due to abuse dofiles are no longer free but available on my website upon payment of a token. Here's the link cruncheconometrix.com.ng/shop. Thanks 😊
Hi Crunch, Really I liked your videos your explaining of the Gmm, Honestly, I am a new in gmm and I wanted to choose between Sys and Diff. but I don't have good experience of that, However, I used your command in my stata but doesn't working with me I think it's my mistak, Could you help how to use it ? T=22 and N=50
Thank you in Avnance
Thanks for the positive feedback on my videos. Deeply appreciated! 💕 GMM may break down because the time span is long. This is because GMM is built for short panels where N is excessively larger than T and T is less than 20 years.
@@CrunchEconometrix Thank you for your answer, I removed one year. T=21 N=50, Is it necessary to use a dummy and lag? I already transformed my variables to Natural logarithm...what do you think?
@@economiclibya5069 Yes but better if year is between 10 to 15 years. Important to include year dummies. I explained why in the respective video.
@@CrunchEconometrix I agree with you my question can I do my estimation without dummy and lag?
@@economiclibya5069 No you can't. Please read Roodman (2009) for more insights on GMM estimation.
please share the complete reference for "Bond 2001"
Here is the full reference: Bond, Stephen R. “Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice.” Portuguese Economic Journal 1, no. 2 (2002): 141 -162.
Is there any difference in STATA codes in case we use unbalanced data?
Not at all, Ksenija. Same code applies to balanced and unbalanced data.
@@CrunchEconometrix Dear prof, couple of additional questions: Do you use log_transformation of variables in order to reduce the effect of outliers or in order to ensure stationarity? How can I donate to your channel?
Hi Ksenija, yes that is one of the merits of using the log transformation...and thanks for the willingness to donate to my TH-cam Channel. I am in the process of getting approved for that from TH-cam to avoid sanctions. Kind regards.
Hi dear, i'm LInican from Brazil and i have question for u. Would u recomend me some book to study S-GMM and D-GMM. Thanks for the help. I appreciate.
Hi Linican, I don't know of any textbook that covered GMM but there are journal articles to guide you some of which I listed at the end of each GMM video.
Hello can I have these all ppt slides related to GMM???
Hi Sameen, unfortunately not. But you can copy out the notes from the video description. May I know from where (location) you are reaching me?
@@CrunchEconometrix I'm from Islamabad (Pakistan )
@@sameenaftab8781 Awesome!!! Kindly share my videos with your, friends, students and academic community in Pakistan and on social media for awareness. They'll learn some usefull tips and skills too! :)
Case I have firms level panel data, financial in nature firms level data, like log of all the series and run OLS and FE and GMM system and difference my sargan test Pvalue is 0.0000 and hansen test pvlaue same 0.0000 AR(2) Pvalue again 0.000 how I can manage the situations that this pvalue should be good according to litrature
Hi Sarfraz, significant pvalues implies that the model is improperly identified which has to do with your choice of instruments. Kindly watch the series in sequential order to guide you on how to address this problem.