SIOUX FALLS, S.D. (KELO) — How does a poll survey of 1,500 accurately reflect voting in South Dakota?

KELOLAND News has been releasing results from a new poll conducted from Oct. 19-21 by Emerson College in Massachusetts. Emerson’s polling has been working with KELOLAND’s parent company Nexstar since 2018.

Emerson Polling executive director Spencer Kimball said the organization will conduct nearly 100 polls this year. Many are political polls like the one completed in South Dakota.

“We do scientific research in that it’s a systematic observation of events. The polling that were doing in South Dakota, is similar to the polling that we’re doing Illinois, in Florida or that in Arizona, that we’re doing in New Hampshire,” Kimball said. “(We’re) using a sampling method and then being in able to pull the data using a weighting scheme and then being able to represent the general population from that methodology.”

So, how does a 1,500 sample of voters most likely to vote in the 2022 election accurately reflect the state?

Start with the methodology, Kimball said.

Remember when most surveys or polls were conducted by picking up a landline or even a cell phone and answering questions from a live person?

“You’d would have been getting live operator calling your house around 5:30 or 6 o’clock at dinner time because that’s when people were going to be home…,” Kimball said of the old days of poll surveys.

Emerson’s data collection still uses landline phones but in 2022, it also uses texts to cell phones that include survey links and emails that also include survey links. Kimball said those methods of data collection allow voters to respond when it best fits their schedule.

Kimball noted that South Dakotan voters were very responsive to the poll which resulted in 1,500 people in the sample. Typically, these types of polls result in a sample of 1,000, he said.

There needs to be a balance between landline phones, cell phones and email users to make sure the sample of voters is representative of the state, Kimball said. If the survey only used cell phones that would not fully represent South Dakota voters, he said.

Who makes up the sample for the survey?

The Emerson poll for South Dakota is a sample of voters most likely to vote in the 2022 election.

Most likely is key, Kimball said.

“It’s important to remember that a representative pre-election poll doesn’t necessarily represent the entire population it’s the population of people who are going to vote in the election,” Kimball said.

“Not 100% of registered voters will vote,” Kimball said.

The election has a historical turnout of about 60% of registered voters so Emerson knew that about 40% of registered voters were not likely to vote, he said.

The poll survey had questions to remove voters who were not registered and who were not very likely to vote.

Also, the poll couldn’t have 50% of the respondents be Democrat because only 25% of the registered voters are Democrats.

A poll that had 90% representation of men would also be inaccurate because that is not the demographic of the state.

Emerson watches for skewed demographics to ensure an accurate representation of the state and voters in the upcoming election, Kimball said.

The poll for South Dakota show more respondents aged 50 and older.

That’s because of accuracy, Kimball said. For example, voters 65 and older tend to vote at a higher rate than younger voters. Again, the poll focused on voters who are very likely to vote in the upcoming election.

The KFF said in 2020, 34,000 residents 18 to 24 voted in South Dakota in 2020. Another 53,000 in the 25 to 34 range. Those together do not equal the 138,000 in the 45 to 64 age range or the 106,000 who voted in the 65 and older age range.

Nationwide, the U.S. Census Bureau said in 2020, voter turned increased with age. Voter turnout Voter turnout was highest among those ages 65 to 74 at 76%, while the percentage was lowest among those ages 18 to 24 at 51.4%

What is a margin of error?

The Emerson poll has a plus or minus 2.4% of error.

“Each survey that’s done using a sampling method, carries a margin of error,” Kimball said.

A sample of 500 would have a 5 percentage point margin of error. The results could be plus or minus on either side of the score. If the poll shows a 55% to 45% lead for a candidate, that could actually be 50% to 50% because of the margin of error.

“As you increase that survey size to 1,000 people, now the margin of error drops to 3%,” Kimball said. The 55% to 45% score is outside the margin of error which makes the sample size result more precise, Kimball said.

The 55% could drop to 52% or the 45% could increase to 48%, but the original 55% would still be the leader because it falls outside the margin of error.

“This is an academic study that we use to test survey methodology. So to me, it doesn’t matter if the race is a 38-point race or a six-point race because we want to make sure the poll falls within the margin of error in any of these races to ensure the accuracy of public opinion,” Kimball said.

Polls are a chance for voter to tell lawmakers what issues are important to them, Kimball said.

The Emerson poll for South Dakota asked respondents to choose the issues that most determined their vote from a list of eight issues. Respondents were also asked about Medicaid expansion and the legalization of marijuana.

Getting it right and wrong

Fivethirtyeight, a website that focuses on politics, opinion analysis and related material, evaluates polling organizations. It gave Emerson polling an A- after the 2020 election cycle.

A Newsweek story from Nov. 6, 2020, cited Emerson’s accuracy in polls on the presidential race with voters in North Carolina and Maine.

Kimball said while organizations such as fivethirtyeight have given Emerson high marks for accuracy, it had to earn those high marks.

“We came into the arena with new methodology,” Kimball said. Eight years ago, Emerson got a C- rating. The rating has increased each year as Emerson hones its skills, Kimball said.

Emerson learns from its polls including how to fend off those who are determined to deliberately provide skewed answers to poll questions, Kimball said. The poll survey has built-in questions that help note discrepancies in answers or indicate false information so the respondent can be removed from the sample, he said.

Sometimes, a poll will get it wrong, he said. Even if the margin of error looks good, there can be what’s called an outlier poll, Kimball said.

Emerson had such a poll in 2017 in special U.S. Senate election in Alabama. It had Republican Roy Moore will a 10-point lead just days before the election in a pool with a margin of error of 3.9%. Democrat Doug Jones won with 50% of the vote to Moore’s 48%.

If Emerson does 100 polls, it’s likely 95 will be very accurate and five will outlier-type polls, Kimball said.

Also polling can also show a shifting landscape for candidates or issues, he said.

Kimball cited polling in New York which first showed one candidate with 12-point in the first poll only to shift to the other candidate with a 20-point lead in a third poll.