What polls can tell us: They’re not meant to predict winners, but they’re still useful tools – Winston-Salem Journal

That’s about as much as the national polls can tell us right now, even if it looks like Democrat Kamala Harris is down in one poll or Republican Donald Trump is up in another.

That’s good; although polls are sometimes treated as predictions, they are not intended to tell you who is likely to win.

Tracking changes in voter intent is difficult to do with a survey, especially when the number of truly persuasive voters is relatively small. Voters’ opinions can and often do change before Election Day.

Horse racing surveys can only capture people’s perspectives at one point in time. Even then, a margin that looks like one that could decide an election — say, one candidate has 48 percent support and the other has 45 percent support — may not be a real difference at all.

The Associated Press believes that the focus on pre-election polls may overstate the significance or reliability of these numbers, so horse racing data is not central to its coverage.

Election-year polls are still useful, especially when trying to gauge what the public thinks about candidates or the state of the country. For example, they told us that many Americans want Democratic President Joe Biden to drop out of the race in 2024.

But polls are not the same thing as an election result. Even a poll taken just before Election Day still reflects opinion before all the ballots are cast.

Surveys are useful tools, but it’s important not to overdo their accuracy. After all, an election organization cannot talk to every single person in the country. They rely on a sample to make a statistically valid estimate of the views of all adults.

Although polls can provide a reasonable approximation of the views of the larger group, the question is how much each finding can vary. The margin of error that all high-quality pollsters will share with their results helps capture some of that uncertainty.

In a poll with a margin of error of plus or minus 3 percentage points, a finding that 47% of voters say they will support a particular candidate actually means that there is a very good chance that somewhere between 50% and 44% of voters will support that candidate. If the other candidate has 45% support, which could really be 42% to 48%, the 2 percentage point difference is not statistically significant.

Therefore, AP will only say that a candidate leads if that candidate leads by more than twice the margin of error.

The fewer people interviewed, the greater the margin of error. This means that state-level polls, or polls that measure the views of a subgroup such as women, men, Hispanics, or black Americans, are subject to even more error than national findings.

Sampling error is not the only source of research error. It is simply the only one that can be quantified using established statistical methods.

There are other factors as well. The wording and order of questions can affect how people answer. The skill of the interviewer can have an effect. Even with high-quality surveys, some respondents are less likely to respond, meaning their views may be underrepresented.

National polls measure how voters across the country feel about elections, but that’s not how we elect presidents.

The Electoral College system means that presidential elections are functionally decided by a small number of states. So in some ways, looking at the polls in those states is a better way to gauge the state of the race.

State-level surveys have their challenges. They are not conducted as often as national polls, and some states are polled more often than others.

The number of people surveyed for state surveys is often smaller than for national surveys, meaning the margin of error is wider.

Some media or organizations publish averages or aggregates that combine the results of multiple surveys into a single estimate. Some organizations create polling averages or models that try to determine which candidate is leading in the general polls.

Averaging survey results does not eliminate survey error and may create additional problems. Survey averages contain their own methodological decisions, such as which surveys are included or given more weight. Some include other factors, such as the state of the economy, to turn estimates into predictions.

While poll averages can provide a general picture of the state of a race, it’s important not to overestimate the accuracy of the average or expect it to be a crystal ball for an election outcome.

Sometimes individual results from multiple different surveys can give a better idea of ​​the potential range of results than the average value reduced to a single number.