HUFFPOLLSTER: Poll Instability Could Be Deceiving

Some shifts in the polls may come down to how much voters feel like answering a survey.

There’s a lot of talk about how “differential nonresponse” affects polls ― here’s what that means. Undecided voters don’t seem likely to break for Trump or Clinton. And most Americans aren’t actively involved in campaigns. This is HuffPollster for Wednesday, November 2, 2016.

POLLSTERS WARN AGAINST ‘PHANTOM SWINGS’ - Benjamin Lauderdale and Doug Rivers: “We did not see any shifts after the release of the Access Hollywood video, the second or third presidential debates, or the reopening of the FBI investigation into Clinton’s emails. When the same people were reinterviewed, almost all said they were supporting the same candidate they had told us they were supporting in prior interviews. The small number who did change their voting intentions shifted about evenly toward Clinton and Trump so the net real change was close to zero. Although we didn’t find much vote switching, we did notice a different type of change: the willingness of Clinton and Trump supporters to participate in our polls varied by a significant amount depending upon what was happening at the time of the poll: when things are going badly for a candidate, their supporters tend to stop participating in polls….We consider it almost certain that Clinton was never as far ahead as many published polls suggested at the high points of the campaign, and equally that she has not lost as much by recent events as some published polls suggest. The truth is more boring: real change mostly happens slowly, and the impact of campaign events is much less than the media makes out.” [YouGov]

That doesn’t necessarily mean that the polls will be wrong - YouGov’s findings don’t suggest that survey data this year is fundamentally skewed, but that the true state of the race could be more stable than it seems. Lauderdale and Rivers note that this year’s pattern in non-response is similar to 2012, when Barack Obama’s internal polling showed a far less volatile race than many public surveys. But the phenomenon they describe, known as “non-response bias,” is different from claims that polls this year are systematically underestimating support for either candidate. 

UNDECIDED VOTERS ARE UNLIKELY TO BREAK FOR EITHER CANDIDATE - Mark Blumenthal: “[T]ruly undecided voters are few and far between. SurveyMonkey’s nationwide Election Tracking, for example, finds just 2 percent of likely voters who are still totally undecided about a choice between Trump, Clinton and third party candidates Gary Johnson and Jill Stein. That said, there are far more people - 7 percent of likely voters - who choose a candidate but also say (on a follow-up question) that their chance of voting for that person is just ‘50-50’ or less. Let’s call them uncertain voters….A closer analysis of 4,007 likely voters who are either undecided or uncertain - including 1,049 who are totally undecided - finds no evidence of a ‘hidden vote’ lurking that might upend the race….Whether we focus on the totally undecided 2 percent or the 9 percent that also includes the uncertain, these potentially persuadable voters divide nearly evenly in terms of their partisan leanings and impressions of Clinton and Trump.” [HuffPost]

DONALD TRUMP HAS A DIFFICULT PATH TO WINNING THE ELECTORAL COLLEGE - HuffPollster: “The HuffPost presidential forecast gives Clinton a 98 percent chance of winning. That runs counter to the narrative that polls are closing in and is very different from other forecast models ― most notably the 71 percent probability she has in FiveThirtyEight’s model. One reason for that is the HuffPost model relies almost entirely on state-level polling data, which shows Trump still has a huge Electoral College deficit, rather than national-level data.State polls show that Clinton is maintaining her lead in the key states she needs to win: Colorado (+5 points), New Hampshire (+5 points), Wisconsin (+6 points), Pennsylvania (+6 points) and Michigan (+7 points). Those five states, plus all of the Democratic strongholds, get her to 273 electoral votes. Clinton also narrowly carries North Carolina (+2 points), Florida (+2 points), Ohio(+1) and Nevada (+2). That’s a total of 341 electoral votes... Trump, on the other hand, faces a steep uphill battle. He only has 164 electoral votes from states he leads by more than 5 percentage points, plus another 33 from Georgia, Arizona and Iowa, where he has a narrower lead. That’s still only 197 electoral votes. He needs to turn 73 votes over in order to win ― no small task.” [HuffPost]

Why Nate Silver thinks Trump more of a chance - Nate Silver: “Trump remains an underdog, but no longer really a longshot: His Electoral College chances are 29 percent in our polls-only model — his highest probability since Oct. 2 — and 30 percent in polls-plus…. From a set of simulations the polls-only model ran earlier this evening, I pulled the cases where Clinton won the national popular vote by 3 to 5 percentage points. In other words, we’re positing that the national polling average is about right, and seeing how the results shake out in the states… Trump’s chances are slim-to-none in this scenario. His odds are 10 percent or below in all of the Clinton firewall states except for Maine and New HampshireThe question is how robust Clinton’s lead would be to a modest error in the polling, or a further tightening of the race. So here’s a second set of simulations, drawn from cases in which Trump or Clinton win the national popular vote by less than 2 percentage points… This isn’t a secure map for Clinton at all. In a race where the popular vote is roughly tied nationally, Colorado and New Hampshire are toss-ups, and Clinton’s chances are only 60 to 65 percent in Wisconsin, Michigan and Pennsylvania. She has quite a gauntlet to run through to hold her firewall, and she doesn’t have a lot of good backup options.” [538]    

The difference seems to be interpreting the possibility of poll failure - The HuffPost forecast model focuses on state polls, assuming that they will reflect national polling changes as appropriate. So if state polls do indeed begin closing, Clinton’s probability of winning will come down a bit. The model also assumes that the polls ― more specifically, the polling aggregates ― will generally be correct. There is some additional uncertainty in the model for systematic polling error, but it only amounts to about a 2 percent chance (the probability of Trump winning right now). In general, the HuffPost model assumes polls will be correct, while the FiveThirtyEight model allows for more uncertainty. That contrast highlights the fact that we don’t really know how much total error polls have ― a bigger topic for all of us in the polling community to address. The bottom line is that if the polls are wrong, the HuffPost model will be wrong too.   

FORECAST UPDATE - HuffPost’s models give Hillary Clinton a 98.1 percent chance of winning the presidency. Democrats have a 50 percent chance of winning the Senate outright, and Republicans a 15 percent chance of holding on, with a 35 percent chance of a tie that would be decided by the winning vice presidential candidate. Accounting for the strong probability that the Clinton/Kaine ticket wins the presidency and vice presidency, Democrats have an 84 percent chance of taking the Senate majority. [Presidential forecast, Senate forecast]

Josh Katz illustrates how Clinton’s odds of winning have changed over time in forecast models

MOST AMERICANS AREN’T GETTING ACTIVELY INVOLVED IN THE CAMPAIGN - HuffPollster: “With less than two weeks until the election, the last-minute flurry of campaigning ― big rallies, boatloads of yard signs and armies of volunteers, not to mention candidates’ attempts to tout all those things as signs of their voters’ enthusiasm ― can seem unavoidable….A new HuffPost/YouGov poll, though, makes two things clear: one, most Americans aren’t going to rallies, volunteering or even publicly displaying their support for either for the presidential candidates, making it a poor gauge for voters’ excitement, and two, neither Clinton nor Trump has an especially clear advantage on any of those metrics….Eight percent of Americans said in the survey that they’d been to a rally or event this year for Clinton, and 6 percent that they’d attended one in support of a Democratic congressional candidate in their state. Six percent had been to a Trump rally, and 4 percent to an event for a GOP congressional hopeful….But the vast majority, 81 percent, of Americans haven’t gone to any events or rallies this year, either for Trump, Clinton, or any other presidential or congressional candidates. Eighty-five percent haven’t volunteered or donated, and 79 percent haven’t put up yard signs or bumper stickers. About two-thirds haven’t done any of those things.” [HuffPost]

HUFFPOLLSTER VIA EMAIL! - You can receive this daily update every weekday morning via email! Just click here, enter your email address, and click “sign up.” That’s all there is to it (and you can unsubscribe anytime).

WEDNESDAY’S ‘OUTLIERS’ - Links to the best of news at the intersection of polling, politics and political data:

-Voters are split on which presidential candidate is better on the issues, but give Hillary Clinton a clear advantage on having presidential qualities. [Gallup]

-Bloomberg finds Clinton narrowly leading among political independents. [Bloomberg]

-Donald Trump’s campaign says it’s disputing the size of its pollster’s bill, not refusing to pay it. [WashPost]

-Libby Nelson reports on a study looking at the role of sexism in support for Trump. [Vox]

-Stuart Rothenberg sees an “uncomfortably cloudy” forecast for Senate. [WashPost]

-Orange County could vote Democratic this year for the first time since 1936. [LA Times]

-Political scientist Mo Fiorina doesn’t think voters are becoming increasingly polarized. [Vox]