Presidential Polls Are Confusing — Follow This Guide To Avoid Getting Fooled By The Numbers

There will be countless polls taken during the general election, and all of us will be hearing about them non-stop between now and November. But polls can be misleading, and it’s very, very easy to draw the wrong conclusion from a seemingly straightforward poll. Thankfully, there are just as many tricks to avoid getting fooled by the polls, which will ultimately make you a well-informed political consumer.

There are a number of reasons why polls can be deceptive. Sometimes, a poll is an outlier; other times, it suffers from bad methodology. In many cases, a poll itself is accurate but the conclusion that’s being drawn from it isn’t. But in other cases, a poll’s results are thrown off by a short-term event that temporarily excites the electorate. Regardless, savvy political followers have several tools at their disposal that can help avoid common polling pitfalls.

First of all, it’s very important to look at polling trends, not individual polls. Toward the end of the 2008 campaign, a lot of Democrats got into a tizzy when several polls showed John McCain overtaking Barack Obama in the popular vote. Obama had been leading McCain consistently for months, but during the month of September, McCain defeated or tied Obama in no less than 16 polls.

Win McNamee/Getty Images News/Getty Images

But Obama went on to win the election, as many historians have noted, and it shouldn’t have been a surprise. Despite those September polls that showed McCain ahead, Obama’s average polling numbers had been higher than McCain’s for a good five months straight. And even in September, when McCain briefly led in polling averages, there were plenty of individual polls that showed Obama ahead.

The point is that you should never put too much weight in any individual poll (such as the one from April that, contrary to almost every other general election poll, shows Donald Trump narrowly beating Hillary Clinton). If you want to have an idea of who’s really ahead, look at the trend over time. Websites like HuffPost Pollster and RealClearPolitics are great for this.

It’s also important to keep in mind that the outcome of presidential elections isn’t decided through a national popular vote, but by electoral college math, because America doesn’t have a national popular vote. And this has big implications as far as reading polls goes.

SAUL LOEB/AFP/Getty Images

Democrats almost always win certain states (California, Massachusetts) and Republicans almost always win others (Kentucky, Mississippi), and so the ultimate winner is usually determined by the vote in just a couple of borderline states, or swing states. This means that polls taken in these states are much more important than national polls.

But at this point in the cycle, there hasn’t been a whole lot of swing state polling, so it’s impossible to yet detect any trends one way or the other. That means it’s better to look at national polls for the time being, simply because there have been a whole lot more of them, and a bigger sample size is always better than a smaller one.

Once the election is closer, though, look at swing state polls: Florida, Ohio, Colorado, Wisconsin, and perhaps Pennsylvania and North Carolina. During the last week of the 2012 election, national polling showed Mitt Romney with a slight lead over Obama. But swing state polls showed Obama ahead, and Obama won that election decisively.

Lastly, remember that some polls simply shouldn’t be trusted. First and foremost among these are online polls, which are not methodologically sound at all and suffer from numerous biases. But there are also certain polling agencies that, for whatever reason, have a bad track record of predicting elections and aren’t worth looking at to begin with. FiveThirtyEight has a helpful ranking of polling firms.

In general, polls are usually accurate, but only if read with a discerning, cautious eye. Don't be fooled by the polling numbers this election season and be sure to check out a variety of polls to get a broader perspective on different datasets.