Some musings early on election morning. The past several days have left almost all of America in suspense about who is going to be elected the country’s 47th president. Many scenarios remain in view after a turbulent campaign.
The surprising result of the highly regarded Selzer poll in Iowa, showing Vice President Kamala Harris leading former President Donald Trump 47% to 44% in a state he carried by 53% to 45% in 2020, suggests something like a 1980 scenario: A surge of late deciders in the last days of the campaign toss aside their indecision and opt for a candidate they had hesitated to support. Both Pat Caddell, Jimmy Carter’s pollster, and Richard Wirthlin, Ronald Reagan’s pollster, assured me after the fact that, following the candidates’ single debate on Tuesday, Oct. 28, their campaigns’ weekend polling showed a huge surge toward Reagan. Caddell had the unpleasant task of telling President Carter on Monday morning that he wasn’t going to be reelected. Reagan won the popular vote 51% to 41% and carried 44 states.
The analogy this year would be a massive swing in the counties outside the million-plus metropolitan areas and university cities away from Trump, whose victory in 2016 and near-victory in 2020 were due to massive majorities there, and toward Harris. Iowa, which is the largest state with no counties in a million-plus metropolitan area and whose two major university counties cast only 8% of its votes, could be a harbinger, as it was in 2016. But I haven’t seen, and I haven’t read accounts of anyone else seeing, any other signs of such a surge. The Selzer poll at this point looks, despite its pollster’s well-earned reputation, like an outlier: one out of 20 polls which, polling theory tells us, produce results well outside the margin of error.
A second possibility is a third consecutive election decided by a few thousand votes in a few of the target states. Trump won in 2016 by carrying Michigan, Wisconsin, and Pennsylvania by a total of 77,736 votes and lost in 2020 by falling short in Arizona, Georgia, and Wisconsin by a total of 42,918 votes. Analysts, none more scintillatingly than Nate Silver with his updated 538 model, are following these five states closely as well as Nevada, which Trump twice lost narrowly, and North Carolina, which he twice carried narrowly.
These seven target states, currently with 93 electoral votes, it is assumed, will determine the outcome, while the other 43 states and the District of Columbia will, by nontrivial margins, deliver 226 electoral votes for Harris and 219 for Trump — notwithstanding the single Selzer poll in Iowa and a single poll in New Hampshire with Trump ahead by the suspiciously precise margin of 50.2% to 49.8%. The RealClearPolitics average of recent polls shows Trump ahead in five of the seven states with a 0.8-percentage-point average lead but with a margin exceeding 2 points (still within the margin of error) in only one state (Trump in Arizona). The final edition of the more sophisticated Silver Bulletin, updated 30 minutes after midnight, similarly shows Trump ahead in five of the seven states, with a lead in only one state (Arizona, again, for Trump) that is over 1.2 points — in other words, again, well within the margin of error.
Intelligent readers, looking at these numbers, once again predict a close election, to be decided again by microscopically small percentages of the vote in one or more of the seven target states — especially Pennsylvania, the most closely divided state in each summary and with the most (19) electoral votes. That makes a certain sense. As Silver has pointed out, polling errors in 2016 and 2020 understated Trump’s eventual vote, but polling errors in other years have understated both parties’ nominees and in no particular way. If you look at the aggregated polling in target states and add 2 points to Trump’s numbers in each state, he ends up with over 300 electoral votes, and if you add 2 points to Harris’s numbers in each state, she does. The New York Times’s excellent polling analyst Nate Cohn has argued that the many pollsters who weight their results by respondents’ recalled 2020 vote may be overstating Trump’s support, but he has also argued that “there’s no reason to believe that pollsters have ‘fixed’ what went wrong in 2020.” He concludes: “I have no idea whether our polls (or any polls) may be ‘right’: too good for Harris or too good for Trump. No one does.” Take your pick.
But there’s something different in these numbers from 2016 and 2020, when national polls tended to show significantly larger margins than Democratic nominees actually were earned in target states. This year, they don’t: RCP has Harris with just a 0.1-point lead nationally, not far ahead of the 0.8-point lead it has given Trump in the target states, and Silver has Harris ahead by 1 point nationally while trailing by a 0.4-point average lead in the target states. This implies that the Republican popular vote margin in the 24 states counted as safe Republican would come close to matching the Democratic popular vote margin in the 19 states and Washington, D.C., counted as safe Democratic. That would be a sharp shift from 2020, when the popular vote margin for Democrats in their safe seats was 14.8 million, far ahead of the popular vote margin for Republicans in their safe seats (which was 7.9 million). Reported trends toward Trump among Latino voters, young black voters, and Jewish voters could reduce that gap but seem unlikely to come anywhere near to closing it.
When in doubt, you may want to consult Occam’s razor. Fundamentals usually matter. Repeated polling shows many more voters approve retrospectively of the performance of the Trump administration than approve of President Joe Biden’s administration, from whose acts Harris has declined to disapprove. Voters, none of whom under age 65 had experienced significant inflation before as adults, disapprove of Biden’s performance on the economy. Most voters disapprove of the immigration enforcement policy that has allowed perhaps 7 million illegal immigrants into the country. You may argue that those responses are unjustified, but they’re there. The last time voters were able to compare the performance in office of two successive presidents was in 1892, when they rejected the incumbent Benjamin Harrison and returned Grover Cleveland to office. My default expectation is that they’ll do something like that this year — in which case, the polls again will have understated Trump’s vote. But I can easily see how it could go the other way. If Nate Cohn has no idea who will win, and if Nate Silver puts the odds at 50% for a Harris win and 49.6% for a Trump win, I’m not going anywhere farther in the direction of predicting who will win than I already have.
