If I strike up a conversation with a stranger, I generally ask if he or she will be voting by mail. People are happy to disclose how they will be voting and almost always, they will be voting by mail. But if I ask if they are - per chance - Democrat or Republican, they get suspicious and balky, and inevitably offended. I can pretty easily talk to people and glean information, but not when it comes to their voting preference. It's interesting.
Just the nature of what you are asking tells me that you have faith in the days gone by and are trying to apply yesterday's standards to today's information superhighway, and the two are vastly different. Landlines are becoming scarce, and people rarely answer unknown numbers because of scams. Cellphone wise, a person can live in Sacramento and have a Norwalk area code, can easily lie about their residence. But there again, people are not inclined to answer unsolicited or unfamiliar calls. In person is the only reliable method imho, because most people responding to a poll on the internet have an agenda, and are just interested in advancing it, while people who receive a poll in the mail are probably not going to sit down and answer all those questions unless they, too have an agenda, making it worth their while to make the effort.
In person right now is not as easy because of a little virus that has people less inclined to hang around a street corner answering questions, on top of which, people are busy. So it's likely that a significant amount of the respondents to polls are not reliable indicators of the voting public.
States that reveal mail-in ballot requests are a pretty good indicator of who is going to vote, but I don't measure those requests as you do, as an indicator of WHO they will vote for. I prefer to measure the ballot requests with the physical demographics of the respective state, ie age, race and the percentage of urban population. In my view, it provides a better indicator of political leanings, but there is absolutely no crystal ball. Thanks!!