tag:blogger.com,1999:blog-2474498300859593807.post5360263746306195605..comments2024-03-08T02:53:17.969-05:00Comments on Econometric Sense: A Toy Instrumental Variable ApplicationMatt Bogardhttp://www.blogger.com/profile/10510725993509264716noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-2474498300859593807.post-9545303962391148682013-09-03T10:18:38.445-04:002013-09-03T10:18:38.445-04:00I think all of your results are correct. The probl...I think all of your results are correct. The problem is that my notation in the post was a little sloppy. I forgot to include intercepts in the notation, but should have. All of my regressions in R include intercepts, and you should get the same results in E-Views by including an intercept. As you suggest we are not interested in interpreting the intercept, but it is necessary to get the correct estimates for the treatment effects we are after. I apologize for the confusion, and will correct the post as soon as soon as time permits. Keep in mind, the data is synthesized to get a feel for IV estimates, so don't pay so much attention to R-square or p-values in this case. I'm only trying to show the mechanics of IVs. Thanks for reading my blog and pointing out the issue with my notation!Matt Bogardhttps://www.blogger.com/profile/10510725993509264716noreply@blogger.comtag:blogger.com,1999:blog-2474498300859593807.post-28880982632679058072013-08-22T14:46:10.596-04:002013-08-22T14:46:10.596-04:00Dear Prof. Bogard,
This post was really helpful i...Dear Prof. Bogard,<br /><br />This post was really helpful in understanding the IV estimation.<br />It is a nice post.<br /><br />To understand it fully, I run all the regressions as you have suggested.<br /><br />My coefficients for model (2) and (3) are same as you have listed above.<br /><br />In models (5) and (6), why should not we include intercept? I think since we are interested to find out the impact of Z on CAMP and RET respectively, we should not include intercept.<br /><br />so when ran regressions (5) and (6) and i got beta 1 = 0.545455 and beta 2 = 0.636364.<br /><br />Following model (7) if we divide beta 2/beta 1, we get 1.166<br /><br />But you found out .3839<br /><br />(Note, i ran all regression in EViews)<br /><br />But when i included intercept terms in model (5) and (6), my beta 1 and beta 2 are 0.263158 and 0.526316 respectively (and respective intercept terms are 0.282297 and 0.526316)<br /><br />Now, following model (7) if we divide beta 2/beta 1, we get 1.166 we get 0.3839 which is what you have reported.<br /><br />But here problem is that my R-sqr are .07 and .01, which is really a problem.<br /><br />Further, when i tried to get IV estimates form 2SLS (model 8), i get IV estimate exactly 1.166 which is what i found when i tried to extract IV estimate from model (5) and (6).<br /><br />And my R-squared looks terrible -0.703003<br /><br />I know my question is bit lengthy. <br />So kindly help.<br /><br />Thanking you in advance.<br />-------<br />skp <br />Anonymousnoreply@blogger.com