Will something like the Bradley Effect play a role in this 2016 presidential election? Recall that the Bradley Effect is supposed to manifest itself in the tendency for at least some sizable percentage of voters to say that they’ll vote for a black candidate but to vote for the white candidate instead. I realize that it’s controversial in some circles whether we actually saw the Bradley Effect in recent presidential elections, much less have seen it in the election for which it was named. But such controversies don’t matter for my present purposes. I merely want to think about the possibility that
- (a) Voters are more likely to vote from Mr. Trump than tell pollsters that they’re going to vote for him.
- (b) Voters are less likely to vote from HRC than tell pollsters that they’re going to vote for her.
Short answer: We don’t have, as far as I can see, good reason to think that either (a) or (b) is true. That removes (baring contrary evidence currently unavailable) one source of skepticism about the accuracy of the polls.
Long answer: Let’s start with (a). The thought goes something like this: They are some voters who who recognize that Mr. Trump is seen by many as a racist, sexist, proto-fascist thug, who might well be, according to Fox News(!), mentally ill. Indeed, these voters might even think that Mr. Trump is some or all of the above. Hence, they do not want to be seen as supporting him. However, the theory goes, they prefer Mr. Trump to HRC because they share these repellent attitudes, because think that HRC would be even worse, or for other reasons I cannot quite fathom. Hence, these voters will say to pollsters, to their friends, to their neighbors, etc. (according to the theory) that they won’t vote for Trump, but they will do so anyway.
The LA Times discusses one poll that suggests a cautious “yes” to (a), while Red State expresses mild skepticism. I think that Red State gets the better of this disagreement since the dispositive value of a single poll is slight. But the idea behind the poll seems like it’s worth exploring further. But we’ll have to await the work of clever (and well-funded) pollsters. Finally, it’s worth remembering that Mr. Trump consistently under-performed his polls until last month. His recent tendency to over-perform his polls really dates, as I understand it, from his win in the New York primary.
What about (b)? I haven’t found any discussion of this possibility, and that fact surprises me. Perhaps the apparent failure of the Bradley Effect to surface in the 2008 and 2012 elections has soured people on this idea. But I’m not sure we should give up so quickly. It seems entirely possible that some people – especially some men – will want to sound right-thinking and enlightened by saying that they will vote for the first female president in US history while either refraining to do so or, quelle horreur, voting for Mr. Trump. Again, we haven’t seen much evidence for this. HRC has pretty steadily outperformed her polls, the opposite of what one would expect if the Bradley Effect were in operation.
 http://www.cnn.com/2016/04/15/politics/hillary-clinton-new-york-poll/ – This is true of primaries, but less so of caucuses.