Endogeneity tests

[@JOHNSMITH-ve3rq]

Totally just racked up a tonne of your playlist. Thanks for this wonderfully educational content.

[@mronkko]

You are welcome.

[@wandeezpso2114]

Thanks for the lecture It was very well explained.

[@mronkko]

You are welcome!

[@dr.kingschultz]

Hausmann test video is available? Your videos are very good!

[@mronkko]

Thanks. I do not have a general video about the Hausman test, but I talk about it in different context. You can take a look at this for example kzbin.info/www/bejne/hp-1pp6onLWhnLs

[@saynaislamdibasaynaislamdi8875]

Thank you

[@mronkko]

You are welcome

[@saynaislamdibasaynaislamdi8875]

@@mronkko can I use durbin -wu - hausman test for detecting endogeneity among the variables

[@mronkko]

@@saynaislamdibasaynaislamdi8875 Yes you can. I think I cover that on the video.

[@thesource8990]

Thank you, sir. Lovely lecture. Does the test of endogeneity hold good if the instruments are weak?

[@thesource8990]

@@mronkko Thank you very much for replying to my query. I would love to know how weak IV affects DWH test. I tried reading Hanh & Hausman (2013) at pubs.aeaweb.org/doi/pdf/10.1257/000282803321946912 but to no avail: the maths is way above my ability!

[@kussehaile158]

Thank you, Sir. Lovely presentation. How we test for endogeneity for probit model?

[@mronkko]

With questions like this, it is useful to start first by asking yourself if you really need a probit model or would a simpler linear probability model work? For some discussion on this, see doi.org/10.1177/1094428121991907 and the references therein. If you really need a probit model, I think that you can use any of the three approaches explained in the video. If you are unsure if these techniques are applicable, you can just put "probit endogeneity test" into Google scholar to see what kinds of approaches are recommended in the literature.

[@kabeldelices1016]

Please how can we apply Hausman test in STATA OR Eviews ? I am running Simultaneous Equation Model with 3sls estimation.

[@mronkko]

In Stata, you can use the hausman command.

[@kabeldelices1016]

@@mronkko i used as command "hausman" but i get msg telling that it's the old syntax... And i didn't find how. I also choose : statistics -> postestimation -> ..-> hausman specification test But i didn't understand by what i must fill the informations requested (as consistent, efficient...? (I am using reg3. simultaneous equation estimated by 3sls for a single country)

[@mronkko]

@@kabeldelices1016 I do not use Stata's menus, but the hausman command takes two sets of named estimates as its arguments. If you are using 3SLS, that would be the consistent estimator and if you want to test for endogeneity, the efficient estimator would be a similar estimator that assumes that endogeneity does not exist. So in this case you would use SURE. But you could also take a step back and ask yourself why you are using 3SLS in the first place. 3SLS is outdated or at least it is difficult to say what advantage it would have over more modern estimation approaches such as ML or GMM. Why not just use Stata's sem command for estimating the system and then doing a nested model test for testing for endogeneity?

[@kabeldelices1016]

@@mronkko Thank you for answering and advising me. I am a beginner in econometrics, that's why I based my choice on articles that dealt with approximately the same subject area as mine. I am trying to identify the direct and indirect link between the Exchange Rate on economic growth for only one country (indirect by the channel of Exports, imports, and Investments) while focusing on the interaction between macroeconomic variables. As I found at the theoretical issues, this model(Simultaneous Equation Model) is estimated by Indirect LS, 3SLS, 2SLS, SURE, OLS... depending on the order of identification of the equations composing the system. As my equations are over-identified I thought of using 3SLS or 2SLS. But as you just said, I'm not sure if I should use 3SLS, 2SLS or not I don't know if the GMM or ML method can be used in my work, as I just said I'm a beginner and I would like to have the opinion of an expert like you. Thank you.

[@mronkko]

@@kabeldelices1016 ML (sem command) and GMM are more general than 3SLS and can be applied always when 3SLS can be applied. If you are a beginner, I would recommend starting with sem because then you can learn one broadly applicable tool well.

[@selenium34]

should've gone to art school...

[@mronkko]

I do not see myself as much of an artist ;)

[@theodouwes4432]

@@mronkko selenium34 is referring to themselves