The polls were wrong, again

The polls were wrong, again

ORLANDO, Fla. — In 2016 the national polls were correct but the state polls were wrong; some by significant margins. Four years later, the national polls were wrong by a significant margin and the state polls were, at best, a mixed bag.

For an industry reliant on public trust, polling has come under increased scrutiny, most of it deserved.

READ: Trump campaign adviser Corey Lewandowski tests positive for COVID-19

Content Continues Below

“Some polls were very good, not that many to be perfectly frank with you, but the real problem was that the misses tended to push all in the same direction,” says Mike Binder, the Director of the Public Opinion Research Lab at the University of North Florida. “Most folks seemed to overestimate support for Joe Biden.”

Binder’s own final poll of Florida had Joe Biden up one point over Donald Trump. Even his own poll’s margin of error (3.3) doesn’t account for this miss.

“Because the errors of 2016 that we thought as an industry we had collectively solved appear not to have been solved, or at least the corrective measures we took did not solve the problem for 2020, and trying to figure what that missing piece is and what we can do is what the next couple of months will be devoted to by this industry,” says Binder.

Following 2016, the polling industry realized it had a problem by not weighting enough for education in its sample size. Most polls corrected for this in 2020, however, the results were just as scattered.

In states like Pennsylvania, North Carolina, Nevada, and Arizona, the average of final polls versus the final vote count, according to Real Clear Politics, was less than a point.

However, in states like Wisconsin, Ohio, and Iowa, the average was off by as much as seven points.

Meanwhile in states like Georgia and Florida, the polls were close but in the wrong direction.

“Listen there were some polls that were very, very good,” says Binder. “But if you look at the polls that had high ratings on 538, they’re missing by like 6 points and a lot of that was pushed towards Joe Biden, and if it were some on one side and some on the other, it would all average out, but the fact that it is all in one direction is really problematic.”

One of the big challenges facing pollsters, especially in Florida, is the state’s large number of independent and non-party affiliated voters; roughly one-third of the electorate. Pollsters say they can accommodate for shy voters on the Republican and Democratic side by simply calling more Republicans and Democrats, however, when NPA voters declined to participate, that can have an impact in the final results of a poll, especially if the bulk of those declining participation all favor one specific candidate.

This dynamic seems to disappear when polls are conducted on specific issues rather than candidates. In the final University of North Florida poll, pollsters got all six of Florida’s constitutional amendments correct most within two to three points of the final margin.