Binary endogenous regressor with too many zero ( more than 99%) and few ones then what IV command

[Tariq Abdullah]

My binary endogenous regressor has more than 99% zeros and very few ones. What IV command should be appropriate in my panel data ?

Should I use ivprobit or ivregress ??

----------+-----------------------------------
          0 |     68,351       99.14       99.14
          1 |        593        0.86      100.00
------------+-----------------------------------
      Total |     68,944      100.00

[Jeff Wooldridge]

ivprobit is for the case when the outcome is binary and the endogenous explanatory variable is continuous. What is y in your case?

[Tariq Abdullah]

Professor, I'm still figuring out what IV I'm going to choose. I'm yet to try both - continuous and binary. Thanks for the feedback professor!

Now, I got clear explanation where should I choose ivregresss and where ivprobit is more appropriate !

[Jeff Wooldridge]

The nature of the IV doesn't matter. If y1 is the outcome and y2 is the endogenous explanatory variable, different methods are suitable depending on the nature of y1 and y2. Is y1 itself binary? Is it approximately continuous? If z2 is the IV for y2, it doesn't matter whether z2 is discrete or continuous.

[Tariq Abdullah]

Professor, as you know how difficult is contracting IV. Therefore, I've made two sets of IV , where one is binary and another one is continuous! Y1 is binary professor and Y2 is continuous. Z is binary in my first set of IV case.

In my alternative solution set of IV, z is continuous. Depending on getting the exogenous and relevant condition , I'd go with the one which meets all the criteria of validity of IV.

Before doing that wanted to know how to do the proper regression. Thanks for your invaluable feedback and precious time ! Meant a Lot !