Andrew Kohut, who runs the Pew Center, has an interesting take on the "Bradley Effect" today and how the polls coulud have gotten it wrong.
His main argument: it’s not that people who were polled lied about their intent, but rather that pollsters weren’t able to reach the people for whom race might be a factor:
Poorer, less well-educated white people refuse surveys more often than
affluent, better-educated whites. Polls generally adjust their samples
for this tendency. But here’s the problem: these whites who do not
respond to surveys tend to have more unfavorable views of blacks than
respondents who do the interviews.
Kohut notes that Hillary beat Barack by 12 points among those with incomes under $50,000.
Also: more analysis from pollster Mark Blumenthal here.