Tuesday, December 20, 2016

More On Polling and Probability

Shortly after the election, I wrote a post with my thoughts on what happened with the polls, and why they failed to predict that Trump would win. One thing I mentioned is the tenuous connection between behavioral intention (what you plan to do) and behavior (what you actually do). That is, people who said they were planning to vote for Clinton changed their minds and voted for Trump instead. And some new work by Dan Hopkins and Diana Mutz suggests this may have been the case:
Our October 2016 wave was conducted with nationally sampled adults over age 26 between Oct. 14 and Oct. 24, meaning that it ended soon after the third Clinton-Trump debate. At the time, Clinton was riding high in the polls — and 43 percent of our panelists in that wave expressed support for Clinton, as opposed to 36 percent for Trump. By way of benchmarking, this same group of panelists had gone for President Obama over Mitt Romney 46 percent to 39 percent in October 2012.

And while most people’s support remained the same, the changes we did observe were consequential. Consider the table below, showing panelists’ support in the October 2016 poll versus their support in the post-election poll, which took place from Nov. 28 to Dec. 7. Eighty-nine percent of the 1,075 American adults reported the same preference in both waves, whether it was for Clinton (38.0 percent), Trump (35.2 percent) or neither (15.8 percent). But among those who did move, Trump had the advantage. While no one moved from Trump to Clinton, 0.9 percent of our respondents moved from Clinton to Trump. Although that 0.9 percent isn’t a lot, those changes are especially influential, since they simultaneously reduce Clinton’s tally and add to Trump’s. If there were a comparable swing in the national electorate, 1.2 million votes would move to Trump.

In all, Trump picked up 4.0 percentage points among people who hadn’t been with him in mid-October, and shed just 1.7 percentage points for a net gain of 2.3 points. Clinton picked up a smaller fraction — 2.3 points — and shed 4.0 points for a net loss of 1.7 points. That’s certainly consistent with Trump gaining steam in the race’s final weeks. Seeing as the 2016 election was held on the latest possible day given the mandate to hold it on the first Tuesday after the first Monday in November, we might just add the 2016 leap year to the ever-growing list of reasons why Trump prevailed.

So what could have changed voters' minds at the last minute, causing them to shift their support? Hmm, I might have some ideas.

No comments:

Post a Comment