Everyone Makes Mistakes, Especially Pollsters

The last thing voters need in the runup to this year’s midterm election is another reason not to trust the results. Unfortunately, the primary season has been marred by polling scandals that are catnip for conspiracy theorists and bad for democracy.A polling outfit called the Public Sentiment Institute has come under scrutiny for allegedly manipulating survey data in last month’s Florida Republican gubernatorial primaryA polling outfit called the Public Sentiment Institute has come under scrutiny for allegedly manipulating survey data in last month’s Florida Republican gubernatorial primary to boost the prospects of James Fishback, the candidate who paid for the poll. A second firm, Patriot Polling, was also accused of publishing surveys in the Florida governor’s race to help Mr. Fishback without disclosing that it received funding from his campaign.Mr. Fishback lost the GOP primary to Rep. Byron Donalds, who will face Democrat David Jolly in the general election. The Public Sentiment Institute issued a statement to the Hill newspaper that accepted responsibility for “disclosure failure” and “methodological error” but denied deliberately manipulating polling data. The paper reported that Patriot Polling hadn’t responded to its inquiries, but both polling groups have subsequently been banned from popular poll-tracking websites.The Florida shenanigans followed a report in August that a third polling company, Median Strategies, was shutting down after admitting that it had published fake surveys on races in California, Nevada and Wisconsin as part of a “social experiment examining how purported polling information could enter and spread through the political information ecosystem without independent verification.”Among those fooled was the re-election campaign of Los Angeles Mayor Karen Bass. The poll purported to show the mayor “leading City Councilmember Nithya Raman by double digits—a surprising result given Bass’s low approval rating and her administration’s recent setbacks on homelessness,” CalMatters reported. “It also claimed that Bass had won support from a vast majority of voters who supported insurgent Republican Spencer Pratt in the primary election.” The Bass campaign trumpeted the results, which had been reported in several local news outlets, but it had to delete the social-media post after Median Strategies acknowledged the poll was a sham.It would be more comforting if we could blame all polling misses on frauds and pranksters, but the reality is that even the professionals have been way off in some of this year’s high-profile primary races. This fuels conspiracy theories and breeds general cynicism about the integrity of elections. In the Michigan Democratic Senate primary, reputable polls had progressive Abdul El-Sayed leading by double digits, yet he beat moderate Rep. Haley Stevens by a single point. In the Wisconsin Democratic race for governor, Democratic socialist Francesca Hong had a comfortable lead, according to the public opinion polls, but she lost to David Crowley.When someone well ahead in the polls ultimately loses, or wins by a much narrower margin than anticipated, people wonder. Are pollsters playing it straight, or are they trying to manipulate votes one way or another to advance a partisan agenda? Keep in mind that pollsters want to get hired in future election cycles, so they have little incentive to produce results that they know to be erroneous. It’s bad for business.A more plausible explanation for polling errors is that primary elections are more difficult to poll than general elections. For starters, turnout is less reliable. Typically, it’s a fraction of registered voters, and using past turnout as a predictor isn’t always reliable, especially if you have a candidate who is generating more excitement than usual and attracting support from people who normally don’t cast ballots in primaries. Enthusiastic voters tend to be the ones most eager to talk to pollsters. This might have led pollsters to oversample left-wing Democrats in the Michigan and Wisconsin primaries.Polling for the general election is likely to be more reliable, and right now that should worry Republicans more than Democrats. To win back control of the Senate, Democrats need to win a net four seats. According to polling data aggregator RealClearPolitics, GOP candidates are trailing in a least a half-dozen Senate races in battleground states, including Alaska, New Hampshire, Maine, Georgia, Ohio and North Carolina.The challenge for Republicans in November won’t be dastardly pollsters. It will be historical trends that show a high correlation between a president’s job-approval rating and his party’s performance in midterm elections. Since 1950, when a president’s job approval is above 50%, his party’s average loss of House seats has been 14, veteran Republican pollster Whit Ayres told me recently. When job approval is below 50%, the average loss has been 32 seats. Mr. Trump’s approval rating currently sits at about 39%. But that’s according to the polls.