Slate: How to Read Election Polls


William Saletan writes about how the simple headlines on election polls can be misleading — and how to interpret the data for yourself.

From Slate:

Early general-election polls have been released, and liberals are freaking out. Donald Trump is almost dead even with Hillary Clinton in theRealClearPolitics average of national surveys, upending assumptions that the race wouldn’t be close. Some Democrats are denouncing the polls as skewed. Others are sinking into panic or fatalism.

Relax. These surveys don’t forecast who’s going to win the election. They tell you how the public feels right now. They show you how different kinds of people are thinking about various questions—how conservative Democrats feel about Clinton, for example, and which qualities women ascribe to Trump. Most important, these polls illustrate how the race might shift, depending on what you and other people do. Here’s a guide to reading them.

1. Don’t whine about party composition. There’s been a lot of bitching on the left about the high percentage of Republicans included in some polls—notably, a Fox News survey taken May 14–17, in which Trump led Clinton 45 percent to 42 percent. In the 2012 election, 38 percent of the people who voted were Democrats; only 32 percent were Republicans. But in the Fox poll, 41 percent of respondents were Republicans; 40 percent were Democrats. That’s typical Fox bias, right?

Wrong. Read the poll’s methodology statement: “The Fox News Poll is not weighted by political party.” That means the poll aims for a random sample, but if more Republicans than Democrats answer the survey (or vice versa), Fox doesn’t tweak its numbers to even things out. (Fox says it does sometime adjust its numbers for imbalances in “age and race,” which can boost one party’s numbers—generally, Democrats.) So if the Fox poll shows a higher percentage of Republicans in May than it did in April—which it does—there’s a chance that the shift has to do with something going on in the electorate, such as rising interest among Republicans. You’d be a fool to close your eyes to that.

If you don’t trust the Fox poll, look at the nearly simultaneous CBS News/New York Times survey. The Times methodology statement says its “sample was adjusted to reflect the percentage of the population residing in mostly Democratic counties [and] mostly Republican counties.” Sure enough, Clinton wins that poll, 47 percent to 41 percent. But if you dig up the detailed tables—not available at or but posted by the CBS News political team at Scribd—you’ll find something that the Times didn’t report or include in its data sheet. Question 9 of the poll asked: “Compared to past presidential elections, how would you describe your level of enthusiasm about voting in the 2016 presidential election this year—are you more enthusiastic than usual, less enthusiastic, or about the same as usual?” Republicans were evenly split: 36 percent more enthusiastic, 36 percent less. But Democrats were more likely to say their enthusiasm was down (39 percent) than up (30 percent).

That’s a warning. When the Times “adjusts” its sample based on “the percentage of the population residing in mostly Democratic counties,” it’s betting that a partisan enthusiasm gap, such as the one detected in its own survey, won’t significantly change the balance of turnout between Republican and Democratic counties. That “adjustment” is prognostication disguised as polling.

Read the rest of the story here.