Instrumental variable estimation

[Lorenz Wagner]

Hi guys, I need your help. I am in the process of doing an instrument variable estimation. For this I watched a video and modeled it after this one. My dataset consists of data on wage (wage), years of education (education), work experience (experience), north (north), south (south), west (west), east (east) and the education fees (fees). I have included all variables in the model except fees and north. Now I want to take fees as the instrument variable for education. I have now followed the instructions from the video and came up with the following command.

reg education south east west experience fees, r
ivregress 2sls wage (education = fees) south east west experience, r

Now, however, the output says that I have more than one instrument variable. Therefore, do I really need to consider only the instrument variable for estimation in the first step? Would the following command be correct then?

 reg education fees, r

[Carlo Lazzaro]

Lorenz:
welcome to this forum.
1) why not considering the Stata built-in command - -ivregress-?
2) have you already ruled out that the latent variable -individual_ability- is not a further source of endogeneity?

[Maxence Morlet]

Great advice from Carlo.

In your situation a Heckman model would potentially be interesting as well; individuals self-select into education, and you will only observe education for individuals who have chosen to pursue educative programmes.

[Lorenz Wagner]

Thank you for your answers. My Problem is, that iThanks for your help. My problem is that I should "divide" the estimation on two steps (exercise for the university). Now I'm not sure if I should record the instrument and all others on the first step or only the instrument. Because if I record all of them it says after the second step that all of them were used as instrument. I hope you understand my concern.

[Carlo Lazzaro]

Lorenz:
1) if your question relates to an academic assignment, please see https://www.statalist.org/forums/help#adviceextras #4;
2) if what above is not true, please note that -ivregress- has a -first- option;
3) moreover, as you might have already read from -ivregress- entry in Stata .pdf manual, -ivregress- the independent variables that are not instrumented are all labelled as instruments, even though you did not intended to use all of them as instruments for your endogenous variable. Hence, no concern at all, it's only a matter of jargon/terminology.

[Nick Cox]

Cross-posted at https://www.reddit.com/r/stata/comme...le_estimation/

Please note our cross-posting policy, which is that you should tell us about it. Meanwhile, you should tell the good people at Reddit too, to stop anybody wasting their time saying the same things as here.