Yesterday I suggested that we’re in for a bumpy landing in this final week of the presidential election. Polls are flying off the printers and are all over the map. The consensus seems to be that the race is tightening. And in talking about Trump fatigue, I asked whether voters are as fickle as we are being led to believe.
Right on cue comes this interesting insight from the pollsters at YouGov titled, “Beware the phantom swings: why dramatic bounces in the polls aren’t always what they seem.” By way of background, YouGov is one of the pollsters who have embraced a recent innovation in this field. They construct a sample of voters to interview over time rather than construct a new sample with every poll (as traditional polling has been done). Whether or not this gives more accurate results remains to be seen. But it does provide a contrast that tells us a lot about wild swings in polls in response to specific events. Here is how they describe the difference over time during this election season:
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. For example, after the release of the Access Hollywood video, Trump supporters were four percent less likely than Clinton supporters to participate in our poll. The same phenomenon occurred this weekend for Clinton supporters after the announcement of the FBI investigation: Clinton supporters responded at a three percent lower rate than Trump supporters (who could finally take a survey about a subject they liked).
I remember reading about this phenomenon back in 2012 when conventional wisdom decided that President Obama’s poor performance in the 1st debate nearly cost him the election. Kevin Drum looks at a recent paper by Andrew Gelman and three other researchers that points to what a few voices back then were suggesting was wrong with that analysis.
Remember how Obama did so poorly and plummeted in the polls? It turns out he didn’t, really. Obama fans just stopped responding to polls, producing the illusion of a 10-point collapse. In reality, he only dropped about 4 points.
When the ABC/Washington Post poll has Clinton with a 12 point lead in late October and then dropping to a one point lead this week, we are likely seeing the same thing. The bottom line is that the noise we’re currently seeing in the polls is probably more related to response rates – especially this late in the election season. Beyond that, looking at overall response rates in traditional polling, we can see how that might disproportionately affect the Democratic coalition.
“It now takes over 300 calls to get 1 response from a young Hispanic male & only 10 calls to reach 1 older white woman” via @CivisAnalytics pic.twitter.com/exrk9rhKCc
— Eileen Patten (@eileen_patten) November 2, 2016
The question becomes: how much time and money are pollsters willing to spend to reach that one young Hispanic male?
As a counterweight, YouGov’s results have Clinton with a consistent lead of 3-5 points. The issue with polling response rates doesn’t mean that YouGov is right about the size of Clinton’s lead. What it does mean is that voters aren’t as fickle as the current bumpy ride would seem to indicate.
