displaying coefficients with reghdfe for instruments

[Zara Ri]

Hi,

I am running the following specification.

reghdfe homicide_articles_osorio 1.treatment (i.treatment#c.homicides = i.treatment#c.adj_homicides) if year>=2000, a(i.code_INEGI##c.homicides i.year##c.homicides) cluster(code_INEGI) old

Treatment is a binary [0,1] indicator. I'm having trouble displaying the coefficients for the i.treatment#c.homicides interaction for when the treatment variable is 1 rather than 0 (see photo below).

[Andrew Musau]

For a 0/1 indicator, you cannot estimate two coefficients. So there is not much gain in declaring the variable as categorical. If you do so, then specify the base levels so that you do not end up with different base levels. I am not familiar with this approach, but the problem arises as you have an interaction involving the DV on the right hand side (whereas the DV is not present).

(c.treatment#c.homicides = c.treatment#c.adj_homicides)

or

(1.treatment#c.homicides = 1.treatment#c.adj_homicides)

or

(ib0.treatment#c.homicides = ib0.treatment#c.adj_homicides)