Reading The Polls This Election Year ... So They Make Sense

No one wants to look at a poll number, and say, well, the numbers show a candidate with a lead of 51 to 49 — so therefore the race is tied.

Sometimes the practice of journalism drives me crazy.

No one wants to look at a poll number, and say, well, the numbers show a candidate with a lead of 51 to 49 — so therefore the race is tied.

Tied? How can that be? Someone must be leading. Well, yeah, they are leading that poll. But it does not mean they are ahead.

There is a small number at the bottom of every poll that reflects the “margin of error.” This number means that the pollster is confident that the poll represents a demographic balance of say a thousand people. But if you polled another random thousand people with the same demographics, then the answer would be different. The margin of error is based on a mathematical formula, one that’s based on the size of the sample group (the more people, the smaller the margin of error) and the level of confidence. (An aside: This is why there is no poll about the Washington NFL team that reflects American Indian and Alaska Native viewpoints. The sample size would have to be ginormous to be statistically relevant.)

Mark Trahant

Mitt Romney probably gets this better than most politicians. On November 5, 2012, the final Gallup Election Survey reported the Republican nominee with a slight lead, 50 to 49 percent for Romney. But the margin of error was plus or minus 2 percentage points. The final tally was Obama 51.1 percent, Romney 47.2 percent.

It’s important to remember this background when you look at various 2014 polls. The oddest one is from Alaska. The CBS News/New York Times/YouGov poll shows Republican challenger Dan Sullivan leading incumbent Mark Begich by six points. That’s a huge lead. But then there’s that margin of error: Six points. Turns out the sample size is small, 412 voters. Another older poll shows Sullivan with a two-point lead — and a margin of error of plus or minus 4 points. And a third poll, even older, shows Begich with a 4-point lead and a margin of error of 3.8.

Add these three polls up and it proves we know nothing. We don’t have enough information. End of story.

Another way to look at polls — and this won’t help with Alaska — is to look at the average of all polls. Real Clear Politics does a great job of that.

In Minnesota, for example, several polls show Sen. Al Franken with a 50 percent lead, and the support that is larger than the margin of error. That’s solid information.

So why do politicians care what the polls say? Shouldn’t the only poll that matters be the one counted on Election Day? Simple: People want to associate with a winner. When people think a candidate is likely to win, an undecided voter is more likely to vote with the majority.

It’s also important to note that not all forecasters use polls alone. Some look at fundamentals, such as the popularity of the president, the state of the economy, that sort of thing. This accounts for a really diverse set of projections, ranging from Sam Wang’s prediction that Democrats will keep the Senate to Nate Silver’s opposite conclusion.

As I have pointed out before, Nate Silver’s misses in 2012 all involved Indian country. It’s really hard to measure a motivated base of American Indian and Alaska Natives. And that’s probably a good thing. We like the element of surprise.

Mark Trahant holds the Atwood Chair at the University of Alaska Anchorage. He is an independent journalist and a member of The Shoshone-Bannock Tribes. For up-to-the-minute posts, download the free Trahant Reports app for your smart phone or tablet.