Help figuring out a 2sls IV regression

[Luna Minerva]

Hello all.

For a university project on health and education I'm doing some IV regressions, using as a basis a paper where there is the following regression:

The key idea is to investigate the impact of an education reform on cognitive ability levels and similar measures. My dependent variable is different from theirs but it still is something akin to cognitive ability, hence I'm expecting results not too wildly different from those in the paper, and the dataset is the same. Now, as far as I can understand the regression I need to run - for the <19 and <16 subpopulations, and for different bandwidths - is

ivreg2 cognitiveability (educationdummy=reformdummy) runningvariable reformdummy#birthmonth height surveyyear

When I run it, though, I incur in the following error:

I imagine this is due to the fact that reformdummy is my instrumented variable, so I cannot have it both inside and outside the round parenthesis. But according to the notes below the table I attached above the authors do include the interaction between month of birth and the reform dummy. How is it possible?

What's wrong in my interpretation and how can I fix my code to get the same regression as the paper?

Thank you.

[Carlo Lazzaro]

Luna:
you may try to calculate the interaction by hand (which is usually not recommended), giving it a new name, such as:

g inter_ref_birth=reformdummy*birthmonth

and see if that does the trick.

As an aside, I interpret -educationdummy- as your instrumented variable (endogenous regressor).