Estimating marginal effects for interactions in the second hurdle of double hurdle models

[Shutong He]

Hi there,

I am running double hurdle models with data from a multi-round modified dictator game, using the xtdhreg command in STATA16.0.

In the second hurdle, I include an interaction between a categorical variable (high vs low externality level) and a continuous variable (climate change concern). After the double hurdle model, I want to estimate the marginal effect of the categorical variable (high externality = 1) across the range of the continuous variable (climate change concern). The goal is to understand at which individual climate change concern levels, purchasing (with negative externalities) in the high-externality situation is significantly lower compared to a low-externality situation.

It seems that, by simply using the margins command without options and specifications (e.g. margins high, dydx(climate_concern),please ignore that the marginal effect I estimate here is for climate change concern rather than the externality level I described above), STATA calls the first hurdle function, i.e. the probit model. The output estimates are the same with the coefficient of climate_concern in the first hurdle.

I wonder if anyone knows how to estimate the marginal effects for the second hurdle. After reading the xtdhreg command document, I didn't find an answer. Maybe using the expression() option is the way to go, but I don't really know how to call the second hurdle with this option.

Thank you in advance for any help. Let me know if you need further clarification/information.

Many thanks
Shutong He

[Shutong He]

Sorry for the spamming. Right after posting this question, I found the solution in help xtdhreg.

For anyone who shares the same question, I copy the following from the help document:

"Users can calculate marginal effects in postestimation using margins. Stata treats the coefficients conditional on the first hurdle being passed as equation(above) or as equation(below) if up is chosen. Coefficients of the first hurdle are treated as equation(hurdle)."

Example: margins, dydx(*) predict(equation(above))