I don't know if voters and non-voters have the same political leanings. It isn't something I've ever looked into. My observation was merely that measures of statical confidence assume random samples. Extrapolating from a non-random sample can give odd results. But this isn't a research paper, so it doesn't much matter.
You are reading too much into it. If I study runners, I should presume the study will apply to those who don't run should they become runners, unless I have evidence otherwise. All the more since many runners were once non-runners. It's not obviously a confounding factor, that would need to be demonstrated. And as I and others have already said, the actual studies indicate the results would have been the same in this election.
