I crunched the numbers! Exit polling sucks when it comes to measuring actual poor people on the GOP side; only 6 of the 20 states where exit polling was available had a big enough of a sample to crunch the numbers. For the other 14, I just used the statewide totals - because of that and due to the fact that there tends to be a slight lean toward Trump among poorer votes, take it with a grain of salt.
I did two: one with Trump vs Sanders where I just put their vote totals head-to-head in a two-way model, and one with Trump vs Sanders vs Clinton vs Cruz. The precise data is available below:
Full size imageFull-size image