In multilevel regression and poststratification (MRP), state and race are typically included as key predictors in the model. However, given the substantial variation in racial composition across states, should an interaction between state and race also be incorporated into the MRP model?
Provided you have enough data to fit the model, that seems sensible, to me