According to the Latest Poll . . .

For months now, polling organizations across the country have been scrambling to provide the most up-to-date picture of who Americans are expected to vote for in the presidential election next month.

For example, on the first day of September, Suffolk University/USA Today released a poll showing 48 percent of likely voters in favor of Democrat Hillary Clinton, and 41 percent in favor of Republican Donald Trump. 

Good news for Clinton, right? Not so fast. That same day, a Rasmussen poll found Trump leading by one point— 40 percent to Clinton’s 39 percent. Yet another poll, released a day earlier by Fox News, showed Clinton up 41 percent to Trump’s 39 percent.

Why did three polls released so close together arrive at such different results? In short: Polls are complicated.

The basic idea behind polling is that questioning a relatively small number of people can give a good idea of what an entire population is thinking. During elections, polls offer the public a sense of which candidate is ahead at any given time. In addition, they give candidates insights into how voters feel about specific issues.

As Election Day nears, it often seems like a new poll comes out every day. 

“Polls have never been more common and discussed as they are now,” says Christopher Borick, director of the Muhlenberg College Institute of Public Opinion in Pennsylvania. 

But it’s important to know the limits of polling. A poll tells us about the present, not the future: It’s not a crystal ball, but a snapshot of public sentiment at a particular moment—and not a perfectly sharp snapshot.

Most reputable polls rely on a method called “probability sampling.” That means selecting people at random from a whole population. This technique gives each person in a group an equal chance of being chosen, so polls that use it are likely to reflect broader public opinion.

Still, polls are only estimates. A typical survey of 1,000 people will usually have a “margin of sampling error” of plus or minus 3 percentage points. That’s important to know, since a few points can make a big difference.

For example, with a 3-point margin of error, the Fox News poll that showed Clinton leading Trump 41 to 39 percent actually means that anywhere from 38 to 44 percent of voters favored Clinton and anywhere from 36 to 42 percent favored Trump. So the poll could show Clinton ahead by eight points—or behind by four.

This illustrates a common misconception about polls: People tend to think of them as precise. But really, they can be pretty fuzzy. 

There are other possible sources of distortion as well. How questions are worded—and the order in which they’re asked—can sway results.

Studies have found that people largely tell the truth while being polled, but sometimes social pressures or other factors lead them to fib. Indeed, research suggests that people are more likely to respond truthfully about their views—say, whether they support a controversial candidate—when polls are conducted online—perhaps because they feel more anonymous.