Politicians, media and research groups spend scads of money each election to poll the public about whom they will vote for. Despite all the money and technology used, sometimes the results are a surprise.
Americans may read too much into polls, according to one expert. “If you are expecting polls to say, ‘This person wins by 0.1 percentage point,’ they don’t do that,” said John Zogby, pollster and founder of John Zogby Strategies.
Polls give a snapshot of a particular moment in time. News that sways voters might break after a poll is taken, and when voters have a change of heart, it will not be factored in. (It happened in the 2016 presidential campaign, according to Courtney Kennedy, director of survey research at Pew Research Center.)
But even before those things might affect an opinion poll, the pollster must figure out who’s going to show up on Election Day. About 56 percent of voting-age Americans actually voted in the last presidential election. Unless a pollster surveys that group, his results are not valuable.
Polling has gotten harder over time, and it is tougher in the United States than in other countries, said Donald P. Green, a political science professor at Columbia University. Potential voters in the U.S. are less likely than they used to be to take a survey. Only about 1 in 7 of those asked to participate will, Green says.
Zogby calls the participation rate in telephone surveys “minuscule.” Many Americans have stopped answering their phones due to recent surges in telemarketing calls.
Younger voters and minorities, in particular, are less likely to answer calls to their cellphones. And many older voters won’t take online polls.
So how can pollsters get a broad cross section of voters?
They weight the opinions of respondents they do get. For every real person polled, they assign a weight in the overall results. It’s how they adjust results to represent a realistic cross section of race, gender and geography. (Nevertheless, in 2016, more non-college-educated whites voted than pollsters expected, particularly in the Upper Midwest.)
In U.S. national elections, national polling doesn’t mirror Electoral College results. Because presidents are selected by the electors from each state, not the national popular vote, pollsters watch swing states, where the electoral votes might go either way, more closely than states where a candidate appears to be way ahead or behind.
Pollsters like Kennedy are optimistic because the error rate in national polls was at historic lows in 2016 and pollsters did well again in 2018 midterm election polling. She points out that problems that cropped up in 2016 with state polls are fixable by weighting the results so that, for instance, college-educated voters are not over-represented.
Still, Zogby said at a recent briefing for foreign correspondents, “This is like walking barefoot on hot charcoal.”