Modeling problem with Multinomial Logit

Hi,

I have a problem modeling a multinomial regression.
This is the tentative function:
multinom(resp ~ age + sex + ... +
I(log(wage)),data=work)

resp is a factor with 4 levels.
My problem is that the variable wage should only go
into the equation for 
certain levels of resp and should be zero for the
others.

I would be very grateful if somebody could tell me how
I can model this in R.

Thanks a million,
   Werner