Who makes political polls and how they work

The election is keeping Pennsylvania in the spotlight, as polls continue to show narrow difference in who voters in the key battleground state want as the 47th president.

Puzzled by how close the margin seems — and having never been surveyed by a poll — a reader asked Curious Philly, The Inquirer’s forum for questions about the city and region: Who makes political polls and how do they work?

Mary Snow, a Quinnipiac University assistant poll director, said the first thing to remember is that “this is an inexact science.”

But this doesn’t mean the polls have no validity.

On the contrary, polls work like a social mirror. Results serve as an indicator of a population’s will because they reflect the general public sentiment about each candidate at the time of the poll.

» ASK US: Have something you’re wondering about the Philly region? Submit your Curious Philly question here.

They are based on public opinion.

When poll results say if the 2024 presidential election were held today, 50.4% of Pennsylvanians polled say they would vote for Vice President Kamala Harris. It isn’t predicting the outcome of the election, but rather reflecting what those surveyed told pollsters.

To get there, pollsters put together a set of multiple-choice questions that they feel will help them gauge public opinion about each candidate.

Open-ended questions are used seldom, Snow said, because they are longer to compose, harder to process, and can be harder to answer on the spot than direct questions.

Different teams at Quinnipiac University review the questions, which are ultimately distributed to poll askers on a script, she said.

Every asker gets the same script to read when they survey people and register answers, according to Snow.

“We don’t want to suggest an answer,” Snow said.

Having a script minimizes pollsters accidentally having a reaction that impacts the surveyee’s answer.

“They’re just there to ask the question, record the answer, and move on,” Snow said. “It’s just a very neutral process.”

Asking every single one of the more than nine million Pennsylvanians registered to vote, let alone the entire country, would be hard to achieve.

Instead, pollsters call a sample of the population using random digit dialing.

A software system that generates lists of random phone numbers, but not personal identifying information, Snow said. The responses will later be weighted to ensure they are a representative sample.

Sadly for those who wish to participate, polls are not conducted on a volunteer basis.

“The chances of being called are small, especially for national polls,” Snow said.

Phone number list and script in hand, poll workers begin ringing.

For Quinnipiac University polls, the school name will come up on the caller ID, Snow said.

“In this day and age that’s important because we get so much spam,” she added.

If folks pick up, surveyors identify themselves and explain what kind of poll they are going to conduct and the fact that Quinnipiac University polls are independent, meaning nonpartisan, before going through the script.

To avoid scams, Snow recommends asking who is financing the poll.

“You want transparency from a pollster,” Snow said. “Always ask who is this? Who do you represent and who’s paying for this poll? If they give you a name, look them up and make sure that they’re legitimate. You can say ‘Can you call me back?’”

If a pollster doesn’t get a hold of you, they will leave a voicemail with identification and information, Snow said.

Once results come in, they are posted on Quinnipiac University’s website, including the phrasing of the questions for transparency, Snow said.

But how can the responses of a small group represent the entire state or country’s opinion?

To explain it, Snow cites a theory from polling pioneer George Gallup.

Much like one only needs a spoonful of soup to know what the soup tastes like, a spoonful of public opinion can attest for the conglomerate.

A nationwide poll needs about 1,000 randomly chosen respondents for a valid result, according to Quinnipiac University polls.

That’s where the margin of error comes into play.

“The margin of error is a mathematical calculation and it’s used to determine how closely those answers by a small group match the views and characteristics of the larger population,” Snow said.

Once the data is in, pollsters make statistical adjustments to match Census data so that the findings can mirror the demographics of the population. That process is called weighting.

Different pollsters use different criteria for this process, but the larger the group surveyed, the smaller the margin of error.

“That’s why when we call, we hope they do participate because we want to hear from them,” Snow said.