Polls Versus Reality in 2016
Numbers: The Good and the Bad
Trying to pinpoint exactly what went wrong in the polling is difficult, but a clear argument can be made: polls under appreciated Trump’s level of impact with voters.
According to Pew Research Center, the fact that so many forecasts were off the mark was surprising, given the increasingly wide variety of methods being tested and reported via the mainstream media and other outlets.
Traditional telephone polls have been joined by increasing numbers of high profile, online probability and non-probability sample surveys, as well as prediction markets, all of which showed similar errors, according to Pew.
Now, keep in mind that National Polling was less of an issue than state polling. Most national polls had Clinton winning the popular vote by a fair margin, and that turned out to be fairly accurate.
In a video on pewresearch.org, Senior Research Methodologist Kyley McGeeney pointed out one hypothesis for the miscalculations: Nonresponse Bias.
McGeeney described this bias, saying, “the people who take your survey are different than the people who won’t take your survey.”
She went on to say, “We know from research that we’ve done in the past that people who have lower education, lower income, people who are less civically engaged, less likely to vote, less likely to contact and elected official… those people are all less likely to take your survey.”
Such a group of people being non-responsive to polling, but showing up on Election Day can present problems for pollsters, and evidently that was part of the problem. Those who didn’t respond to polls seemed to turn out to vote in much greater numbers than in previous elections.
Andrew Mercer, a Senior Research Methodologist at Pew Research Center explained a second issue with polling falters, the “Shy Trumper,” effect.
He explained the hypothesis, saying, “Voting for Trump is considered socially undesirable by some people… they don’t want to admit it either to an interviewer on the phone or to pollsters in general… and then when they get in to the voting booth, who they actual vote for is not the same person they told pollsters they were going to voter for.”
Additionally, Pew says there is a likely voter survey error, where voters tell a pollster they are going to vote, and they ultimately stay home on Election Day.
The questions that pollsters ask often determine whether they peg someone as a “likely voter,” and this can impact their percentages, when someone they saw as a “likely voter,” ends up not participating on Election Day.
The American Association for Public Opinion Research, according to Pew, has put together a task force of experts in the surveying industry, who will attempt to find out clearly, what went wrong with the polls and perhaps what improvements can be made in the system.
“If Democracy is the will of the people, then maybe somebody should go out and find out what that will is.” -George Gallup
Until then, polling companies perhaps need to go back to the drawing board and assess how they can best do their job in serving those who most need it.