WHERE DID YOU GET THIS NUMBER?

The inside dope on how polls work—and don’t work—from CBS’s News Director of Elections and Surveys.

When the phone rings and someone asks for your opinion on a political matter, writes Salvanto, kindly take the call and give an answer. Polls are imperfect measures, but perhaps less imperfect than we think—especially, writes the author, if we disagree with the result. They differ, of course, and they can be errant in giving an impression of certainty when they indicate only probability; think of all the polls showing that Donald Trump had no chance of winning the last election. There is a vast difference between certainty and possibility, and while a good poll will indicate a range of possibilities and not a single outcome, we dislike guesses. “The very idea of expressing things in probabilistic terms is to express uncertainty,” Salvanto writes, “but too often everyone just wants to express things in quite the opposite fashion: as either yes or no.” Of those yes-and-no matters, there are many. The author looks closely at polls of gun owners and other putatively single-issue voters to find many points of agreement (“background checks, to rule out criminals and terrorists, find nearly universal favor, in principle, because they take action against bad actors rather than the weapon”) but also extremely broad areas of disagreement that often devolve into hatred. Conservatives agree on many points with liberals, but they’ll tell you that liberals are evil all the same, and vice versa. Salvanto notes that polling indicates that those who are most committed to a political party are most likely to characterize their opponents as enemies, even as, say, conservatives step away from traditional Republican ideology to say that government should do more to help solve economic problems.

A revealing look at the numbers, how they’re derived and interpreted, and how they sometimes fail us. Timely reading for the coming midterm elections.