Polling experts on the red trickle: "The death of polling has been greatly exaggerated"

At the time of this writing, a number of prominent Republicans are refusing to accept their declared or potentially looming defeats.

From candidates for governor like Doug Mastriano in Pennsylvania, Lee Zeldin in New York and Kari Lake in Arizona to a former president with a lot on the line (Donald Trump himself), the party of Abraham Lincoln continues to buck the previously unbroken American tradition of conceding after losing an election. Yet after consulting several polling experts, Salon repeatedly heard the same view expressed over and over again, with some variation but little meaningful deviation. In the words of one of those experts, Joshua Dyck, who is a political science professor and Director of the Center for Public Opinion at the University of Massachusetts, Lowell:

"The death of polling has been greatly exaggerated."

W. Joseph Campbell — a professor at American University and author of the book "Lost in a Gallup: Polling Failure in U.S. Presidential Elections" — elaborated on exactly what happened in the 2022 contests.

"It may still be emerging, but as of this morning, I'd say polling performance overall was not bad," Campbell observed. "It certainly wasn't the embarrassment or the surprise that we saw in 2020. The presidential election polls collectively were well off the mark in terms of the Biden-Trump race, and also had problems down-ticket with certain Senate and gubernatorial races. This year, it's not quite as dramatic, at least as it stands now, although there were a number of pollsters who quite clearly overestimated Republican candidates and prospects."

David Barker, director of the Center for Congressional and Presidential Studies and a professor of government at American University, reflected to Salon that it was "somewhat surprising" to him that "the polls seemed to nail it." Initially describing Tuesday evening as "a good night for pollsters," Barker quickly revised his language.

"I guess I should say it was a quite a good night for traditional pollsters who use 'gold standard' traditional methods of polling [such as dialing people up on the phone], which most pollsters have kind of given up on because of incredibly low response rates and have moved to various other machinations that sometimes seem to work," Barker explained.

Barker's comments underscore the key issue involved in polling of all kinds – namely, the fact that the mechanics of actually conducting a poll are extremely complicated. There are so many variables involved that even competent pollsters face a not-insignificant chance of being totally wrong after the ballots are counted.

"For effectively predicting the result in advance, the main challenges for pollsters are identifying voters, reaching those would-be voters and then getting them to respond," Christopher Wlezien, a professor of government at the University of Texas in Austin, told Salon. "The first of these has been a problem for a long time but the latter two have grown in importance in recent years, and in ways that may be associated with candidate preferences, possibly the reaching aspect more than the other."

In 2022, that last variable did not seem to have been a factor, given that the polls aligned with the election results more closely than many pundits expected. Yet even if there had once again been noteworthy discrepancies between the polls and the results, that would not necessarily discredit the pollsters.

"Almost all polls, because of the cost and so on, are presented at the 95% confidence level," top election expert Larry Sabato of Sabato's Crystal Ball explained. "That is the margin of error applies at the 95% confidence level. That means 5% don't. That means one out of every 20 polls has results beyond the margin of error, the listed margin of error, simply for statistical reasons. They could be wrong even if they did everything right. Nobody ever mentions that."

While no one is arguing that the polls did everything right in 2022, they clearly came closer to hitting the mark than many skeptics believed would be the case. To be fair to those skeptics, however, historically the party which holds the White House loses a lot of seats during midterm elections. While exceptions have applied when a sitting president is particularly popular, Joe Biden's approval rating is stuck in the low 40s while he struggles with inflation and rising gas prices. During the normal course of events, the so-called "red wave" that Republicans had anticipated would have materialized.

The 2022 elections, apparently, were abnormal. A big reason why that happened (as Salon's Amanda Marcotte recently pointed out) is that the Supreme Court overturned abortion rights in America with its controversial ruling in Dobbs v. Jackson. By overturning the half-century of precedent that existed after Roe v. Wade, the Supreme Court's Republicans simultaneously galvanized voters who felt their rights were at stake and made it seem as if the Republican-controlled Supreme Court was the "in-party" while Democrats were the "out-party."

"That radically changed public policy in a particular direction, which is what you usually see out of a president or the Congress," Barker mused. "Therefore it created a psychology on the part of Democrats that made them feel like the out-party. It made them feel like they were losing something based on the current dynamics in Washington and wanted to get out there and try to do something to change that. In that case, that meant electing more Democrats."

Want more health and science stories in your inbox? Subscribe to Salon's weekly newsletter The Vulgar Scientist.

In addition to feeling as if the Supreme Court had rendered Democrats into the "out-party," voters seemed to have felt like the out-party because of Trump. Although the former president was not on the ballot this year, his repeated pushing of a Big Lie — namely, falsely claiming that he won the 2020 election — likely combined with the Jan. 6 insurrection and his many other scandals to keep him front of mind.

"We typically see punishment given out by voters to administrations that have had, at best, uneven economic performance," Campbell pointed out. "The Biden government has fared very poorly on those fronts, but the voters are not punishing the Biden administration for its failings. That encourages one to look elsewhere" for voting motivations. Noting that "not all midterms are alike," Campbell speculated that Senate Minority Leader Mitch McConnell may have "identified a real flaw" in Republicans' 2022 strategy when he commented weeks ago on low candidate quality, a veiled swipe at the many electoral duds endorsed by Trump during the primaries.

"Some of these candidates declared allegiance to Donald Trump and that does not seem to be going over well this cycle," Campbell observed. "Their ties to Trump are not necessarily going to help them in the general election. It may help them win the primary, but it doesn't seem to be translating into wide popular support in the general election. I think candidate quality may be a factor here as well, and then related to that is Trump's shadow."

Other factors may have also played a role in the midterm elections aligning with the polls instead of conventional wisdom. For instance, Generation Z turned out to vote in higher numbers than expected, and some pundits believe this surge could be attributed to Biden's student loan forgiveness program. Yet it is also important to recall that not all polls are made equal. If anything, the greatest shortcoming of the polls in 2022 was that so-called gold standard polls were mixed in with surveys by firms biased toward Republicans. By flooding the zone with polls skewed to make it seem like Republicans had massive advantages, these firms may have inadvertently caused their own side to make political miscalculations that could have been avoided if they had relied on neutral firms. Indeed, to the extent that polling averages veered from reality, the trend was in the direction of them overestimating Republican support because of those partisan polls.

"One side floods the zone, the Republicans, and affect the polling averages," Sabato told Salon. "You can't really trust them anymore." He specifically cited Trafalgar and co/efficient as examples of companies that produce these types of polls.

Dyck, who referred to the Republican poll-flooded environment as filled with "noise and garbage," described it as "either an overreaction to partisan non-responsiveness or to perceived right-wing bias in the polling industry" that can lead to "outward attempts by Republican and right-leaning firms to essentially create their own polling narratives." As one clue that this played a role in the surprising results, Dyck pointed out that the RealClearPolitics polling average and FiveThirtyEight polling average were different because the former gave pro-Republican polls more weight than the latter.

"RealClearPolitics essentially decided to include every single poll, regardless of which firm conducted it, and there's a little bit more of a correction for bias that exists in the FiveThirtyEight modeling," Dyck explained. As such, RealClearPolitics characterized races like the Senate contests in New Hampshire and Washington as close even though they wound up being won easily by the Democratic candidates. Ultimately RealClearPolitics predicted that Republicans would pick up three Senate seats, an outcome that is already impossible. Other polls, as Barker pointed out, offered ludicrous predictions such as Kari Lake blowing out her Democratic opponent in the Arizona gubernatorial election.

"Because they were the lion's share of polls this cycle, they actually had a heavy presence in these kind of polling averages that you see," Barker explained.

Finally, the experts who spoke to Salon all pointed out that presidential elections and midterm elections have fundamentally different dynamics. Indeed, while the 2016 and 2020 election polls were off, the 2018 midterm election polls wound up being pretty accurate. It could very well be that, for unknown reasons, the polling dynamics that added unpredictability to the two races where Trump ran for president did not apply in midterm election years. For that matter, even the dynamics behind the polling errors in the 2016 and 2020 contests were radically different.

"The 2016 error was a split," Dyck explained, meaning that some traditional polling methods worked and others fell short. By contrast, he added, "In 2020 we just had a messed up election because of COVID-19, and so you had differential partisan non-response bias that was caused by COVID-19. Pollsters were seeing higher response rates throughout the year because people were at home and bored, but there were differences between the people who were home and bored," with Democrats more likely than Republicans to answer pollsters. "That was not a Trump-caused phenomenon, as so many people attributed it to this 'Shy Trump voter' phenomenon. It had to do with actual underlying behavior about the COVID-19 pandemic; some people were going on as if life was normal, and so they had lower response rates and surveys."

It remains to be seen if the anti-Trump and pro-choice trends revealed in both the 2022 polls and elections will continue to shape politics or prove to be as anomalous as Trump's 2016 election victory. If nothing else, however, there is little doubt that the polls celebrated a big win on Tuesday night even if Democrats and Republicans continue to figure out the exact shape of what happened for them politically speaking.

As Sabato himself put it, "Nerds are good."