Monday, October 17, 2016

The "Bradley Effect" and Why Polls Measuring Trump Support May Actually Overstate His Election Performance

Much has been written about the well-known, at least to political scientists, "Bradley Effect" and the role that a similar phenomenon could play in this election. Ballotpedia explains:
The Bradley effect, sometimes called the Wilder effect, is a concept that attempts to explain discrepancies between voter opinion polls and outcomes in elections where white candidates campaign against minority candidates. Adherents of the Bradley effect believe that some voters will tell pollsters that they are undecided or likely to vote for a minority candidate but will vote against the minority candidate on Election Day. It was named for [former Los Angeles Mayor]Tom Bradley, an African-American candidate who lost the 1982 California gubernatorial race despite having a lead in the polls going into the election.
Of course, this presidential election doesn't involve a black candidate versus a white one. But could
Former Los Angeles Mayor Tom Bradley
something similar be going on, namely that people are lying to pollsters as to which candidate they will support on Election Day?  Breitbart reported on the theory back in August, including a quote from an Emerson College Professor Gregory Payne:

“I think with Trump .... [m]any people are saying to maybe their friends while they’re having a sip of Chardonnay in Washington or Boston, ‘Oh, I would never vote for him, he’s so – not politically correct,’ or whatever, but then they’re going to go and vote for him. Because he’s saying things that they would like to say, but they’re not politically courageous enough to say it and I think that’s the real question in this election.”
Prof. Payne is right to consider the theory, but he offers no evidence to back up his analysis.  Certainly we have had numerous examples of polls v. election results during the nomination phase of this presidential election.  If the secret Trump support were a real thing, wouldn't Trump have outperformed his polls consistently in those state GOP contests?  Not only did that not occur the exact opposite happened, i.e. Trump consistently underperformed the polls in the primaries.  I wrote about that phenomenon back in March:
Twenty state contests had a Real Clear Politics poll average before the election, or if no RCP average, a recent poll that RCP deemed credible enough to publish the result.  I looked at Trump's margin of victory compared to those poll results, or if he lost the state, the margin between Trump and the winner,   I found that when it came to the margin, Trump underperformed his poll numbers in 15 of 20 elections, by an average of -7.84%.  It has also increased over time. In the fourteen contests March 1st or earlier, Trump's underperformance was -4.47%.  In the 6 contests since then, the underperformance was -15.7%
Only in the late primaries did Trump's performance at the polls finally begin to match or exceed his poll numbers.

In addition to that primary history of Trump underperforming the polls, I would also point to general election polling which consistently show people view Trump as unqualified to be president.  In the latest Fox News poll, 56% of respondents viewed Trump as unqualified.  A September Quinnipiac poll pegged that number at 62%, while 61% thought Hillary Clinton was qualified to be President.

Obviously many people who are telling pollsters that Trump is unqualified to be President are also expressing support for him over Hillary Clinton.  But could it be that those anti-Clinton people are, in fact, lying to pollsters, using the opportunity to vent their extreme displeasure regarding the Democratic nominee to pollsters without the real life consequences of casting a ballot?  When those same poll respondents enter the voting booth, are they going to vote for someone they see as unfit for office over someone they might view as qualified but who they for, good reason, do not like?

I am guessing that many of them will not.  On Election Day, I believe Trump will considerably underperform the polls, especially the state polls.  My crystal ball says Trump will fall short of 200 electoral votes and it will be the worst GOP presidential election loss since 1996 when former GOP Kansas Senator Bob Dole only received 159 electoral votes. 

1 comment:

Glen Gallo said...

I find it interesting being a Californian and an independent my entire voting life and 1982 was my first chance at voting.

What is lost in history for whatever reason is that Bradley most likely suffered from the Brown Effect and Deukmajian benefited from the Reagan Effect.

Exiting Jerry Brown was not popular. He had a Republican Lt Governor Mike Curb. Governor Brown was dating Linda Ronstadt at the time and would go on trips. Mike Curb would swing into action and vetoed bills issue executive orders and other political moves that for whatever reason Brown would still leave. This sense that the democrats just did not care I believe spilled over to Bradley whom was not a great public speaker. Capable but not great.

The Duke had Reagan campaigning for him. He was a very good public speaker. He had a Democratic party that really was crumbling in popularity.

This claim that you do something in public and turn around and do something different in private is so full of holes it is laughable.

1)Polls were not in public they call a sample at home
2)Bradley carried SF and LA (no surprise there) and the polls probally called those two area codes more

That being said these two elections have little in common and I think polls are better today.