“Polling problems appeared to be greater at times when Mr. Trump was on the ballot and that, disproportionately, voters whom Mr. Trump activated didn’t participate in polls.” We slightly overrepresented voters who support government intervening “in the economy and people’s lives.”...The pandemic may have made Democrat voters more available to pollsters because they were “more likely to hold jobs that allowed them to work from home and had the greater ability to answer pollster calls.”
4/13/21, “Pollsters Say They Undercounted Republicans, Trump Supporters in 2020," Wall St. Journal, Aaron Zitner (Print ed., April 14)
“Self-assessments by two polling groups [including Global Strategy Group, GBAO Strategies, ALG Research, Garin-Hart-Yang Research Group, Normington Petts, and Pew Reserach] in recent days conclude that they undercounted Republicans and Trump supporters in 2020 and that there is no firm answer yet for how to fix the problem.
The examinations come after many pollsters significantly understated Republican strength in the 2020 presidential and congressional elections, and the findings suggest that a variety of polling methodologies came up short. One report, released Tuesday by five Democratic polling firms, finds that their surveys undercounted voters who viewed former President Donald Trump as “presidential.’’ Another, by the nonpartisan Pew Research Center, concluded that obtaining the proper representation of Republicans in surveys has become more difficult.
While the reports suggest several possible sources for the underrepresentation of Republicans and Trump supporters, one leading idea from the groups suggests that the problem will be hard to resolve: Even when pollsters in 2020 reached Republican voters or those in GOP-leaning groups, such as white voters without college degrees, those who chose to answer surveys didn’t represent a full range of views. Those who participated in surveys tended to be less supportive of Mr. Trump, while those who declined surveys tended to more firmly back the former president….
What’s unclear is why surveys are underrepresenting Trump voters. Pollsters who conducted the surveys said that even though they are reaching demographic groups that leaned toward Mr. Trump, they may not be talking to a sufficient number of his supporters. For example, when pollsters set out to survey voters who don’t have four-year college degrees, they may be getting responses from those who work in offices but not at construction sites. Those two types of workers may hold different views of the president.
Alternatively, more–conservative voters may be refusing to participate in surveys because of increased distrust of pollsters, of the media that often sponsor polls and of other institutions—a distrust fueled by Mr. Trump….
That finding suggests that pollsters can’t fix their problem merely by including more voters from any one party or demographic group but will have to find new ways to access the hardest-to-reach voters within several demographic groups.
Jill Normington, a principal with Normington, Petts & Associates, which participated in the Democratic study, said the five firms concluded that even when a survey sample properly mirrored the characteristics of the U.S. voter pool, “We were getting the wrong people, at the end of the day.’’
Underestimating Trump Support
“The Republicans who are saying yes to pollsters don’t seem to be a representative mix,’’ said Courtney Kennedy, director of survey research at the Pew Research Center, though this finding was more an inference than one supported by hard data in the center’s self-examination.
Preliminary work from the Democratic firms suggests that other factors were at play. With voter turnout at a modern record, pollsters were surprised by the large number of ballots cast by “low-propensity’’ voters, or those who vote occasionally or never before. Those voters in many places tilted substantially to the GOP.
At the same time, the coronavirus pandemic may have exacerbated the pollsters’ challenge. Democratic-leaning voters may have been more available to pollsters than were Republicans, because they were more likely to hold jobs that allowed them to work from home and had the greater ability to answer pollster calls.
The studies by the two polling groups show that survey problems appeared across many polling methods. The Pew Research Center, for example, draws its respondents from a panel of 10,000 randomly selected adults, who have been recruited to the panel in a manner that mirrors the broader U.S. population. They answer survey questions online. The Democratic pollsters tended to reach randomly selected voters by cellphone and landline, but most were also experimenting with additional ways to reach respondents, like text messages inviting voters to answer a survey online, Ms. Normington said….
No matter which factor is the bigger contributor to error, “one of several challenges that the polling industry faces is making sure that we’re able to contact anyone who is eligible to vote,’’ said Fred Yang of Garin-Hart-Yang Research Group, which participated in the Democratic study.
The Democratic firms said polling problems appeared to be greater at times when Mr. Trump was on the ballot and that, disproportionately, voters whom Mr. Trump activated didn’t participate in polls.
“Our initial analysis has discovered, for example, that we underrepresented voters who considered Trump to be `presidential.’ We also slightly
overrepresented voters
who support the government taking certain actions to intervene
in the economy and people’s lives,’’ their report said.
An open question is whether polling will be more accurate in the 2022 midterm elections, when Mr. Trump won’t be on the ballot, than they were in 2020.
Firms in the Democratic study included Global Strategy Group and GBAO Strategies, in addition to ALG Research, Garin-Hart-Yang Research Group and Normington Petts.
In their reports, both the Democratic firms and Pew laid out steps to test ways to improve their surveys—including old-school techniques, like going home-to-home and knocking on doors as well as turning to traditional mail.
The Pew Research Center said it would take several steps intended to make sure that its panel of respondents reflected the broader population more accurately. Those included removing several thousand participants from overrepresented groups and adding adults who initially prefer taking surveys by mail rather than online.”
………………..
Copyright ©2020 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8
Appeared in the April 14, 2021, print edition as ‘Pollsters Undercounted Republicans in 2020.'”
No comments:
Post a Comment