Two weeks ago, President Obama was crushing Republican challenger Mitt Romney in the polls. Then Romney graduated from likely to presumptive GOP nominee, Gallup started its general-election tracking poll, and suddenly Romney was up by two percentage points... even as a CNN poll still had Obama up by nine. Other polls showed Romney winning big among white males, but losing the female and Latino vote. And just about every day, political junkies are greeted by a flurry of new polls — often contradicting the surveys that were hyped mere hours or days earlier. As the polls come fast and furious, it's a lot to take in. Here, five tips on how to make sense of the numbers:
1. Don't obsess about any one poll
Instead of getting upset (or elated) about a poll's findings, remember that no one poll means very much by itself, says Ezra Klein at The Washington Post. And "individual polls that show unusual results are almost always wrong." You want to watch the averages of several polls, as compiled at places like RealClearPolitics or Pollster.com. If a poll looks like an outlier, it probably is. Also keep in mind, says Mark Blumenthal at The Huffington Post, that "in late October, polls will be highly predictive of the outcome," but 200 days out, they're no more predictive than a coin toss.
2. Ignore state polls — even for swing states
"Given how unreliable national polls are this early," the polls of individual states are completely worthless, says Jonathan Bernstein at A Plain Blog About Politics. "Really: Ignore state polls," and ignore speculation about electoral-college math, at least until Labor Day. Your best bet is still the national averages. "Why? Because generally, swings are national, not local, in nature," and the candidate who wins the popular vote almost always wins the electoral college.
3. Be aware of who's being polled
"It is worth looking at whether the poll is conducted among registered voters, likely voters, or all adults," says Nate Silver at The New York Times. In the past eight presidential elections, Republicans have done better in polls of likely voters that registered ones — about 2 percentage points better — and that's "probably not simply a statistical fluke." The groups that vote Republican, like older and wealthier voters, also tend to vote more reliably.
4. Figure out which pollsters you trust
All pollsters aren't created equal, and some tend to consistently produce results more favorable to one party — the pollster's "house effect." The most obvious example, says The Huffington Post's Blumenthal, is Rasmussen, whose "polls have shown a modest but consistent 'house effect' in Romney's favor compared to those of the other organizations." But don't just ignore or discount pollsters you don't like, says Ed Kilgore at Washington Monthly. They're usually just using a different methodology, and "polling averages will sort all that out."
5. Don't take history's lessons too seriously
"A final and more general point," says The New York Times' Silver: "Don't over-learn the lessons of history." There have only been 16 presidential elections in the era of modern polling, and "that simply isn't a lot of data." So anytime you read that "no president has ever been elected if X," or "no incumbent has ever lost if Y," take it with a sizable grain of salt.