REAL CLEAR POLITICS: The Current State of U.S. Political Polling.

The final Monmouth University poll had New Jersey Gov. Phil Murphy 11% over GOP challenger Jack Ciattarelli – in a race eventually decided by just under 3 points. Monmouth wasn’t alone: Farleigh Dickinson and Stockton University both had the race at 9% in their final poll, while Rutgers Eagleton had it at 8%.

Patrick Murray, director of the Monmouth University Polling Institute, was candid afterward. “I blew it,” he wrote in a Newark Star-Ledger mea culpa. The Monmouth poll, he added, “did not provide an accurate picture of the state of the governor’s race.”

So why are pollsters getting things so wrong? And which polls – and which races – should Americans keep an eye on in the final week of the 2022 midterms?

Let’s take the first question first. American University professor W. Joseph Campbell invokes Leo Tolstoy’s famous admonition in “Anna Karenina” about families: “All happy families are alike; each unhappy family is unhappy in its own way.” In other words, there are many ways to make polling mistakes. Here are seven:

GIGO: In computer science, the acronym stands for “garbage in, garbage out,” a principle that organizations that average polls (RealClearPolitics and 538 are the most prominent) have been contemplating lately. But the same principle applies to individual polls themselves. A basic problem when trying to extrapolate the voting preferences of millions of people based on a sample of some 600 is that the polling sample can turn out to be unrepresentative of the electorate. Too many Democrats, for example, or too few. The challenge has grown exponentially because of modern technology. Cord-cutters, mobile phones, call screening devices, an explosion in the number of polls – all these factors and more have helped produce very low response rates.

Pollsters themselves were the first to recognize the problem, but it’s proven a tricky one to fix. Steve Koczela noted after the 2020 shortcomings that pollsters thought they’d addressed the big problem of 2016, which was underestimating – and therefore undercounting – Trump’s appeal among white voters without college degrees. Yet it happened again four years later.

“We all made our best attempt to solve all of the methodological issues that we identified after 2016,” he said. They apparently didn’t weight their samples enough. Meanwhile, new problems arose, including not detecting Trumpism’s ability to pump up turnout among working-class Hispanics.

Here’s video of Campbell himself appearing on C-SPAN: When Polls Goes Bad: American University Professor Joseph Campbell taught a class on public opinion and election forecasting. He spoke about some of the most significant polling misses in American politics. American University is located in Washington, D.C.