That’s the thing, I don’t. I know number of male and female voters and number of votes candidate a & b got, as in the table above.
One possible way to model is defining p as a function of male/female ratio. But what I described earlier feels like a more natural generative process.