Earth to polling industry: Polling error wasn't just about THIS election - polls totally missed the coming blowout in '17 VA House of Delegates also. All signs point to bad voter modeling assumptions & methods 1/4 https://www.politico.com/news/2020/11/06/polling-industry-blows-it-again-434591
For example, assuming a voter turnout universe that's misses a large group that often skips certain elections when the likely voter model is built out for sampling or adjustment or 2/4
that all White or Black Men or "Latinos" should be adjusted uniformly for purposes of making adjustments when data samples are short in that group or not asking respondent education level so adjustments can be more accurate, and finally 3/4
Using telephones as a sole source of data gathering skews poll respondents given that phone answering rates are radically different in different groups - seniors are probably the only ones who pick up the phone consistently. Change your methods 4/4