But that’s the whole point. It’s entirely meaningless for pollster to ask “Are you the kind of person who answers polls?” as the audience is 100% self-selected to agree. They have zero data from people who disagree, so anything they ascribe to the unpolled segment is entirely speculative.
At most, they can figure out the demographics of the people who answer polls, and come up with “95% Trumpy, 98% white,” etc., and then try to back-match the missing data to the results of some prior event, such as an election. But that’s no longer a poll, that’s just reporting the results of the most recent election. It’s completely invalid.
