ivprobit produces same coef and se for marginal effect

[James Park]

This problem seems to be present in any dataset but I will give an example from Stata official dataset.

use http://www.stata-press.com/data/r13/laborsup

probit fem_work fem_educ kids other_inc
margins, dydx(other_inc fem_educ kids) atmeans  post

ivprobit fem_work fem_educ kids (other_inc = male_educ)
margins, dydx(other_inc fem_educ kids male_educ) atmeans  post

If I run this, probit and margins after probit produce different coefficient and standard error, which is expected, and which is good. The fact that margins produce similar coef with LPM gives additional confidence.

However, ivprobit and margins after ivprobit produce the same coefficient and standard error. This seems highly unlikely and strange.

Why is this like this?

[FernandoRios]

If you are using Stata 14 or lower, you also need to indicate, when using margins, that you are interested in obtaining the marginal effects on the predicted probability:

use http://www.stata-press.com/data/r13/laborsup, clear

probit fem_work fem_educ kids other_inc
margins, dydx(other_inc fem_educ kids) atmeans  post

ivprobit fem_work fem_educ kids (other_inc = male_educ)
*This is what you did
margins, dydx(other_inc fem_educ kids male_educ) atmeans  
*This is what you need to do
margins, dydx(other_inc fem_educ kids male_educ) atmeans  predict(pr)

HTH
Fernando

[James Park]

If you are using Stata 14 or lower, you also need to indicate, when using margins, that you are interested in obtaining the marginal effects on the predicted probability:

use http://www.stata-press.com/data/r13/laborsup, clear

probit fem_work fem_educ kids other_inc
margins, dydx(other_inc fem_educ kids) atmeans post

ivprobit fem_work fem_educ kids (other_inc = male_educ)
*This is what you did
margins, dydx(other_inc fem_educ kids male_educ) atmeans
*This is what you need to do
margins, dydx(other_inc fem_educ kids male_educ) atmeans predict(pr)

HTH
Fernando

Thank you! This worked. Should I add predict(pr) for margins of ivprobit, but not necessarily for probit? I just confirmed that, for probit, margins will produce the same outcome whether or not I have predict(pr). But for ivprobit, predict(pr) will change the result.

Is there a situation where predict(pr) will produce different results even for just probit?

By the way, my Stata version is 15. Could it be that there is something wrong with my Stata 15?

[James Park]

Also, what does it tell us if I get very similar numbers for #1,2,4 below, but those for #6 is very different?

Also, #5 is very very statistically significant, whereas #6 is not. Is it a likely outcome to have huge significance for ivprobit beta while not getting any significance for ivprobit marginal effect? If yes when will that arise?

What will cause problem only in IV probit, while not causing problem in probit, linear probability model, and IV linear probability model?

1.
reg Y X
2.
ivreg Y X
3.
probit Y X
4.
margins, dydx(X) predict(pr)
5.
ivprobit Y X
6.
margins, dydx(X) predict(pr)

[FernandoRios]

Hi James
I think there is no situation that i know of where predict(pr) will change the results from just margins after probit.
For your 6 cases that you mention, i would first like to see what type of coefficients and changes you are observing before doing any other comment about it.
Best
Fernando