Political poll enters field as surveys face challenges

By Scott Benjamin

CROMWELL – Political polling provides an interesting contrast.

Connecticut campaign staffers and the political reporters hover around their computers at 6:30 a.m. the day that one of the installments in the noted Quinnipiac University poll is to be distributed for a gubernatorial or U.S. Senate race.

Near the end of the campaigns CTMirror posts a chart with the results of all of the polls from the campaign. Four different independent organizations conducted polling during the 2018 Connecticut gubernatorial race.

Seamus McNamee – a research director for GreatBlue Research in Cromwell, which conducted four Connecticut gubernatorial polls for Sacred Heart University in Fairfield and the Hearst Connecticut Media Group last fall - said that voters care about the polls.

“People always want to understand what are the likely outcomes,” he explained in an interview.

According to a 2017 Politico.com story, Connecticut’s Quinnipiac in Hamden has transformed its campus by becoming a brand name in polling over the last generation. Quinnipiac does polls not just in Connecticut but nationally and in battleground states.

Yet, Pew Research has reported that post-2016, 21 percent of the public gave polling an “A or B,” which was half of the percentage from 1988.

It isn’t just the outcry after Republican President Donald Trump’s surprise victory over Democrat Hillary Rodham Clinton in 2016. The late Pulitzer Prize-winner, William Safire, complained about the reliability in a 2000 New York Times column, noting that polling professionals had told him that they were typically getting a 35 percent response rate. Safire questioned whether with that kind of rate if pollsters could get a broad cross-section of the voters.

Pollsters today can only dream of a 35 percent response rate.

Jim Messina, who managed former Democratic President Barack Obama’s re-election campaign in 2012, told The Christian Science Monitor after that election that he had reservations about the reliability of polling, partly because of response rates as low as 10 percent.

McNamee said that “five or six years ago we were getting a response rate of one in every 10 calls and now it is one out of every 13 to 15.”

Anna Greenberg and Jeremy Rosner of the white shoe Greenberg Quinlan Rosner polling organization wrote in the Washington Post in 2017 that polling has faced “The rising cost of collecting high-quality data” from declining response rates and increased use of mobile phones, some from people who have moved and retain the area code from their previous destination.

Yet, National Affairs reported in 2018 that one extensive study of polling “acknowledged that the industry faces significant challenges, but concluded that there was no evidence to support the claims of a crisis in the accuracy of polling.”

“Where it’s become more difficult, technology has done a good job of keeping up,” McNamee said in response to criticism that pollsters have trouble finding cellular-exclusive voters who have retained their former area code but live halfway across the country.

He said that technology is now “great at specifying exactly where those cell phones are. There are ways to understand where the cell phones are based on their frequent use.”

McNamee said in most states about 35 to 40 percent of the residents are cellular-exclusive and even though Connecticut ranks in the bottom five nationally the figure is over 30 percent. In 2008 Associated Press reported that nationally only 14 percent of phone users were then cellular-exclusive.

He said for the four gubernatorial polls that GreatBlue Research conducted between mid-August and late-October, about 50 percent of the respondents were contacted on a landline phone and 50 percent on a cellular phone. He added that a computer system generates digits to provide non-published phone numbers.

Each of the Sacred Heart/Hearst Connecticut Media Group gubernatorial polls had about 500 respondents.

“We base our sample on recent Census data and demographic data,” McNamee explained. He added that the poll is designed to “create a representative sample on gender, affiliation and other factors.”

“We want to mirror the state and the electorate in a way that is reflective,” said McNamee.

The polling is done by 24 workers at GreatBlue Research’s own call center, which has two full-time supervisors. The calls for the gubernatorial polls were done on week nights and Saturdays.

For years GreatBlue has done polls and focus groups for organizations ranging from consumer products to utilities to golf courses from their call center in Cromwell.

“We train them to be polite and manage a conversation in a way that gets the answer,” he said regarding the attributes that GreatBlue Research seeks in its call center.

McNamee said since they didn’t begin polling until after the August gubernatorial primaries, GreatBlue did not conduct focus groups on the Sacred Heart University/Hearst Connecticut Media Group poll.

However, he said focus groups, which usually consist of eight to 12 people and are directed by one of GreatBlue’s three moderators have been valuable on other projects.

If we do a focus group, we usually do a focus group first,” said McNamee. “There is a value to having focus groups first in a lot of cases.”

McNamee said that focus groups can, for example, provide insight in how the questions should be worded.

He said that they can be utilized in a way “similar to our pre-test [of 25 respondents] in making sure that everything is clear.”

“It could be a consideration,” he said of using them before starting a poll for Sacred Heart/Hearst Connecticut Media Group. “It definitely could be of value.”

Greenberg and Rosner stated in their Washington Post column that, “The best polling has always been accompanied by directly listening to people, face to face, in their own words. . . Open listening can cast doubt on things that may have become conventional wisdom in a campaign. We have worked on many races where the ‘front-runner’ was actually quite weak, but that was more evident in focus groups than in standard survey measures of favorability or job performance.”

McNamee, who has a bachelor’s degree in Political Science and Religion Studies from Sacred Heart, said GreatBlue’s initial relationship with its two partners began with three polls on state issues, including the budget. GreatBlue Research, Sacred Heart and the Hearst Connecticut Media Group collaborated on four polls on the gubernatorial race between August and late October.

He said he believes the relationship will continue on projects to be determined.

The final Sacred Heart University/Hearst Connecticut Media poll, which was sampled between Oct. 29-31 and the final Quinnipiac poll, which was released two days earlier, both indicated the gubernatorial race between Democrat Ned Lamont of Greenwich, Republican Bob Stefanowski of Madison was too close to call. Former MetroHartford Alliance head Oz Griebel of Hartford was the other candidate.

The Sacred Heart University/Hearst Connecticut Media Group poll was the only one done during the campaign that ever showed Stefanowski, a longtime former business financial officer at both GE in Fairfield and UBS in Stamford with a lead. Their last poll had Stefanowski with an advantage of 2.4 percent over Lamont, a businessman from Greenwich, with a margin of error of 4.3 percent.

Just over 12 percent of the voters polled in the last survey said that they were undecided.

According to CNN, Lamont won the election by 49.4 percent to 46.2 percent for Stefanowski, a 3.2 percent margin, which appears to be consistent with the results of the final Sacred Heart/Hearst Connecticut Media polls when the number of undecided voters are considered.

McNamee noted that a poll is “a snapshot in time” of what is happening and not a predictor of eventual outcomes.

Regarding criticism of the 2016 presidential polling that didn’t accurately reflect Trump’s popularity, McNamee said a presidential election is not, in effect, a national poll since the electoral college is determined in each of the 50 states, thus making it more a poll of 50 individual states.

Some pollsters have noted that the surveys done in 2016 were accurate on Democratic nominee Hillary Rodham Clinton’s national popular vote percentage, but that in some stat’s where Trump had been slightly behind in the last poll, the eventual president scored a narrow victory.

McNamee said online polling , which has been used in recent years by some pollsters, is particularly effective in some consumer surveys, such as when the respondents can choice which of two can labels they like best.

He said there are greater “quality controls” in calling voters for a political poll since it ensures that they are randomly chosen since they don’t know that they will be a participant until the polling organization makes the call.

Former Hartford Courant political columnist Michele Jacklin has said if it was up to her all polls would be abolished since it shouldn’t matter to voters who is ahead in the race and they should instead focus on policy issues and character.

She added in a 1999 interview that after special interest money, political polling is probably the second most corrupting influence in American politics since if a candidate is trailing by a significant margin, the poll results can seriously hamper his ability to raise money, receive favorable news coverage and attract volunteers.

One former Republican congressional candidate in the Fifth District in Connecticut, who didn’t want to be identified so that he could be candid, said that “my experience is that poor poll numbers will do more than anything to dissipate enthusiasm among campaign volunteers.”

“We don’t have any horse in the race,” McNamee said regarding comments that independent political polls have a corrupting impact on elections. “We’re just doing good research and are trying to understand what’s going on” in the races.