Swing State Tracker: Hot Pollster-on-Pollster Controversy!

Do we have a scam poll problem?

A person walks in the rain toward a building with a sign out front that says "Voting Ballot Drop-Off"
Early voting in Allentown, Pennsylvania, on Thursday. Angela Weiss/Getty Images

Welcome to the final edition of Swing State Tracker. Over the last six weeks, we’ve made a lifetime of memories together. What will you cherish most? The time in September when Biden had a stable lead, or that one week in October when Biden’s lead was stable? Let’s all stay in touch!

If only we could just continue with a few more paragraphs of dumb jokes to ease up the tension in these last few days of the race. But unfortunately we have some business to discuss.

We’ve been using RealClearPolitics polling averages for our tracker. RCP simply mixes together all the polls without adjusting for “house effects,” or pollster bias, like FiveThirtyEight’s average does. We’ve tried to avoid scouring the crosstabs of each and every poll to look for reasons their results might be “skewed,” and to just trust the averages. But RCP’s method has a vulnerability: It can allow partisan pollsters of questionable repute, who release polls prolifically, to bend averages away from what higher-quality, nonpartisan pollsters are finding.

Trafalgar is a Republican pollster that tries to capture what it believes is a significant “shy Trump” voter element—i.e., Trump voters who theoretically feel uncomfortable telling pollsters that they support Trump—that other pollsters are missing. It has consistently found Trump’s support about 6 points better than most other pollsters across the board. Fine. Throw it in the average. But until recently, Trafalgar hadn’t been releasing crosstabs of many of its polls. And now that we’ve seen them, they’re really … something else. A Trafalgar Florida poll, for example, that has Trump leading overall by 3 points, shows Trump winning 25 percent of Black voters and leading among 18-to-24-year-olds. A Michigan poll that has Trump leading by 2 points—the other most recent Michigan polls show Biden leading by 10, 8, 8, 9, and 7—finds Trump leading among 18-to-24-year-olds by 14 points, winning 25 percent of Black voters, and winning independents by 28 points. These figures are in a separate universe from what other pollsters, private and public, are showing. Other pollsters, like Susquehanna and InsiderAdvantage, have been showing similar top-line results. They’ve been commissioned by a pro-Trump entity called the “Center for American Greatness.”

It’s a whole controversy.

Maybe we should have been more skeptical from the start about using polling averages from a website that dangerously overheats our computer whenever we load it. But we’re not about to switch out methodologies in the last week. Maybe RCP’s averages are correct! Who’s to say?

What both FiveThirtyEight and RealClearPolitics broadly show, though, is a notable drop-off in Biden’s polling leads after Pennsylvania, the tipping point state. That’s where Trump has the opportunity to pull an inside straight. If the polls are systemically underestimating Trump by a few points, he could collect Georgia, North Carolina, Florida, Ohio, Iowa, and Arizona in short order. Then it’s just a battle over Pennsylvania and, unfortunately, a battle that will be litigated by a revamped federal judiciary that doesn’t really care about ensuring votes are counted.

The FiveThirtyEight forecast gives Biden an 89 percent chance of winning, and Trump 10 percent. There is a 1 percent chance of a tie.

If you have not voted, vote in person or drop your ballot off at a drop box. Do not put your ballot in the mail at this point.

Chart showing swing states Trump and Biden need to win
Source: RealClearPolitics polling averages, Oct. 30. Slate