Econometric Question: Is it a problem if marginal estimates with IV are way higher than the marginal estimates without IV?

[Tanima Ahmed]

Dear All,

I am estimating a model with instrumental variable / exclusion variable.

I run the model with instrument/ exclusion variable: cmp (provider_group = outside_born $controls) (nn=i.provider_group $controls) if sample!=. [pweight = fp], ind(5 4) robust
The marginal estimates of the model with command below is:
margins, dydx(i.provider_group) predict(pr eq(lf))

But if I do not use the instrumental/exclusion variable approach I get the marginal estimate of: probit nn i.provider_group $controls if sample!=. [pweight = fp], robust
margins, dydx(i.provider_group)

I know in my model instrumental variable is required due to reverse causality between nn and i.provider_group. But I am a bit concerned about such a high estimate with IV.

My question is - is it ok to have such a high marginal estimate with IV as compared to the marginal estimates without IV? Or should I be concerned about any underlying problem with the model?

[Phil Bromiley]

You'll increase your chances of a helpful answer by following the fa

[Phil Bromiley]

FAQ on asking questions. Provide Stata code in code delimiters, readable Stata output, and sample data using dataex. We work better with results in the text rather than as pictures (which are often hard to read).

I can't fully understand your code - I don't use cmp. However, the reason for instrumental variables is that they change the results. So, in general, it is not a concern.