Criticism of new mayoral poll prompts defense from polling company
Critics say Republicans were overrepresented in the SurveyUSA poll released Sept. 1 and that independent voters were underrepresented. Questions also were raised about the ratio of likely voters in the poll, which critics called too high.
The company that produced a new poll this week covering San Diego’s mayoral election and other local races is defending itself against criticism that the poll’s methodology was questionable and could have skewed the results.
Critics said Republicans were overrepresented in the SurveyUSA poll released Sept. 1 and that independent voters were underrepresented. Questions also were raised about the ratio of likely voters in the poll, which critics called too high.
Police reform measure looks strong; housing bond in trouble; height limit measure could be tight vote.
Those criticisms, if valid, could have played a role in Councilwoman Barbara Bry running 3 percentage points (37-34) ahead of state Assemblyman Todd Gloria in the poll, which showed Bry with stronger support among Republicans than Gloria.
Bry represents council District 1, which includes communities such as La Jolla, University City and Carmel Valley. Gloria’s 78th Assembly District ranges from Solana Beach to Imperial Beach and includes Pacific Beach.
The high ratio of voters the poll deemed likely to vote — 517 of the 534 registered voters who responded to the survey — is primarily because people who complete automated phone surveys about politics are typically people who vote, SurveyUSA President Ken Alper said Sept. 2.
“We’ve seen this for years and years,” Alper said. “In an automated survey, people who don’t plan to vote just hang up. They won’t even stay on the line.”
Alper said he believes recent political turmoil nationwide and locally is responsible for the relatively high participation of Republicans and the relatively low participation of independents in the poll.
Critics, including Paul Mitchell of Political Data Inc., said the poll did not include enough independents, based on recent San Diego elections.
“They’ve managed to come up with a sample where independents are only 18 percent of the likely voters, but these voters have been 29 percent, 32 percent and 32 percent of the votes cast in the last three elections,” Mitchell wrote on Twitter. “You can’t do a poll in San Diego and mess up this basic fact — independents matter there.”
Alper said his polls don’t ask people whether their official voter registration card shows them as a Democrat, Republican, independent or member of a smaller party. Instead, they ask how people identify themselves at the time of the survey.
“Your answer may depend on a whole lot of factors,” he said, adding that the recently concluded Democratic and Republican conventions and protests over racial issues have made many people more politically conscious. “Right now, it’s tough to find someone who doesn’t identify with one party or another.”
Alper said the relatively low number of independents in the poll likely explains the relatively high number of Republicans, with many people registered as independents identifying themselves as Republicans.
Alper said SurveyUSA previously has considered adjusting poll results to make the pool of respondents match up better with who has voted in previous elections, but he said the company has decided against it.
“We don’t want to put our thumb on the scale,” Alper said.
Another criticism of the new SurveyUSA poll is that it lacked a separate racial category for Blacks, instead lumping Black respondents with Asian voters.
Alper said that happened because only 16 of the respondents to the poll were Black, which is not a statistically reliable subgroup because it’s less than 5 percent of total respondents.
He said a separate subgroup for Blacks, if it had been included in the poll, would have featured asterisks because the results would not have been statistically reliable.
Blacks make up about 6.5 percent of San Diego’s population, so the 16 blacks who responded to the poll is less than half the roughly 34 or 35 who would have responded in a perfectly representative survey.
Mitchell levied another criticism of SurveyUSA that Alper declined to address. Mitchell said the lead that the poll shows for Democratic presidential candidate Joe Biden over Republican President Donald Trump of 60 percent to 36 percent is too small. He based that on results from the 2016 presidential election.
“This poll has Biden winning by 24 points — about the margin that [Hillary] Clinton won in the county. ... She has to have beaten Trump by 40 points in the city,” Mitchell said on Twitter.
Alper said he takes criticism of his polls in stride.
“It’s routine,” he said. “You want to take it seriously because there could be a legitimate problem, but in this case we stand by our polling.”
The poll, which was sponsored by The San Diego Union-Tribune and 10News, surveyed 534 registered voters Aug. 28-31 over landlines, smartphones, tablets and other electronic devices and has a margin of error of 5.3 percentage points.