I want to understand the different impact of education and number of children on the probability of working for married and unmarried women. I am using the user-contributed -nehurdle- command. My code is

Code:
webuse womenwk
replace wage=0 if wage==.

nehurdle wage children education if married , trunc  
margins, dydx(*) predict(psel)                                        //(1)

nehurdle wage children education if !married , trunc
margins , dydx(*) predict(psel)                         //(2)
I want to test the difference between estimates in (1) and (2). What might be the way? I do not want to use interaction terms as the interpretation gets messy (in my opinion).