Certainly- points taken, *however* the polls conducted over the 2016 election were all highly criticized, as obviously confounding factors skewed the results. Those confounding factors likely included heavy self-selection bias.
Secondly, the methodologies you describe that somewhat control for self selection were not the ones necessarily used by many of the polls. For example, providing internet access to survey participants who do not have it and otherwise could not participate is a much different methodology in both participant selection & survey administration.
I mean, this strays off-topic since we are discussing finer points of research, but telephone surveys definitely are now more prone to participant bias than ever before, so now neither form of polling is close to random. This unfortunate position in research methodology will hopefully be temporary, but it probably is one of the issues behind the 2016 polling inaccuracies.
I have not read all of the polls from 2016, but it would be interesting to do a meta analysis of those using in person random selection for accuracy. I bet the major problem with those polls was untruthful answers. Not sure how to control for that…
