As Peggy Noonan infers in The Wall Street Journal this week, perhaps race isn't the only factor which can trigger misleading responses to pollsters. She cites:
The spirited remark of a Manhattan social figure who, when I asked in September if he knew who he would vote for, said he would be one of the 40 million people who would deny the day after the election that they voted for Mr. Trump, but had.Noonan implies we need to broaden our definition of the Bradley effect so it refers to the tendency of a significant number of responders to give pollsters a socially desirable misleading response when their intended voting action is socially undesirable. COTTonLINE concurs.
A famous fact of 2016 is that of almost 700 counties throughout the country that voted for Barack Obama twice, a solid third went this year for Trump. Many were Democrats. Very few saw it coming.