# Endogenous Interaction panel IV

I want to estimate the following:

$$Y = b_0 + b_1D_1 + b_2X_1 + b_2X_1D_1 + b_3X_3....+error$$

where $Y$ = outcome, continuous
$D_1$ = endogenous, binary
$X_1$ = exogenous, binary
$X_1D_1$ = interaction of the two
$X_3, X_4$ = exogenous variables

I have a panel data. From what I read so far, though not finding something that answers directly my question, I seem to be needing to do something like:

• run an xtlogit $D$ $X_3$ $X_4$ etc
• predict $\hat{d}$, pr
• use as instruments for the interaction $\hat{d}$ and $Z\times \hat{d}$ , where $Z$=instrument, continuous.

Any comments or ideas about this?

• Why do you need a panel ordered logit if the endogenous variable is is binary? Jan 10, 2017 at 23:21
• Sorry there was a typo. It was just a logit not an ordered one. Jan 11, 2017 at 12:42