What polling can’t tell you

On Nov. 5, election officials across America will count more than 150 million ballots to answer a burning political question: Who will be president of the United States? Until then, the best signals we can get will mostly come from public opinion polls, which will be the fuel of endless debate on who has the lead, Democrat Joe Biden or Republican Donald Trump. But what do polls really tell us? Like a picture from a weather satellite, a poll can be pretty good at telling you what people think right now. It's much harder to tell what people will be doing weeks or months from now.

SAMPLE SIZE

Some polls ask a few hundred people what they think. Others ask several thousand. The number of people surveyed is called the sample size. A big sample is like having more pixels in a digital photo. The more people surveyed, the clearer the picture.

HOW TO READ BETWEEN THE LINES

Pollsters usually tell us how clear the picture is by giving a margin of error. A big margin might mean that the picture may be too fuzzy to draw clear conclusions. Also, the margin applies to each estimate in a poll. If one candidate has a lead over another, that lead should be about twice as large as the margin in order for the advantage to be meaningful.

The margin of error is tied to the number of people in a poll and the larger the sample size the smaller the margin of error. The margin serves to give a little wiggle room when extrapolating the results of a 1,000 person sample to the entire nation.

Pollsters usually describe the margin as having a 95% confidence interval. That means that if they did the poll 100 times, the results would be expected to fall within the margin 95 times.

WHO ARE WE POLLING AND WHY?

A lot of people don't vote. So as elections approach and people want to know who is going to win, pollsters try to focus on the people who will actually be casting ballots.

Many polls describe their sample as hundreds or thousands of registered voters, who are more likely to vote than those who aren't registered.

Other polls focus on "likely voters" which are based on complicated statistical models. Some pollsters guard their methodology as a trade secret. Their assumptions are important because even if the sample is large and the margin is small, poll findings might look very different from election results if the pollster is wrong in projecting who will cast a ballot.

Voters also change their minds. Pollsters often ask people who they would vote for if the election were today, even if that election is six months or a year away.

POLLING AND THE ELECTORAL COLLEGE

Many polls are based on responses gathered nationwide and give an idea of the share of the votes in the election each candidate might get. But the popular vote doesn't determine who will be president. The winner is determined by an arcane body known as the Electoral College, which divides 538 electors across U.S. states and the District of Columbia. A handful of states determine the course of the election, and the winner of the popular vote sometimes loses the Electoral College. That makes polls in states such as Pennsylvania and Arizona particularly important, though keep in mind there may not be many battleground state polls and their sample sizes could be small.

WEIGHTING, OR ADJUSTING THE DATA TO FIT AMERICA

How do we know a poll actually represents America? When pollsters go about surveying people, they try to make sure they are getting responses from all sorts of people. They ask people about their race, ethnicity, gender, education and where they live. Then they check that the mix of people in the poll matches the Census Bureau’s estimates on the mix of people that make up the whole country.

HOW TO MAKE SENSE OF IT ALL

It helps to look at lots of polls. There might be more polls done nationally than in the battleground states that will decide the election. Look at both. Remember that every poll has a margin of error so the results might look like they are whip-sawing. Averages of those polls will smooth out the trend.