2022 polling: The bad, the ugly and the really ugly

Is a Republican wave building for Election Day? Possibly, but that wave may already be here, and it may have been coming for weeks. Pollsters may be heading for a beating, perhaps even worse than the Democrats.

In 2016, Trump’s win shocked the media and the polling community. But the fact is Trump was always within the margin of error in the national polls, and Hillary Clinton never got above 50 percent. The state polls were problematic (to say the least), but the national polls were not that far off. The issue was the media refused to believe Trump could win and would not report objectively.

The polling was much worse in 2020. The polls undercounted the Trump margin badly in several states, less so nationally. That Trump lost covered up the errors. Worse, the polls whiffed at the state level. Senate races in Maine and North Carolina were called wrong, while margins in Alaska, Iowa, Montana and South Carolina were much larger for Republicans.

It is perfectly legitimate for polling estimates to be wrong. Polls are a sample of the larger electorate, and any sample can and often is off by at least a little bit — that’s what “margin of error” is. Legitimate polling should be wrong in a nonpartisan fashion — it should call as many races incorrectly for Republicans as it would for Democrats. The problem is that the polls have been wrong in one direction, favoring the Democrats — and that is systemic bias.

The pollsters claim they have improved, but it doesn’t look like it. I looked at four prominent university polls (Quinnipiac, Marist, Suffolk and Siena) and they all look problematic.

Schools of misinformation?

Each of the university polls consistently overestimated Democratic results in 2020, and generally they report above-average numbers in the current term. Marist and Quinnipiac tilt toward the Democrats the most. Using their last poll for each race, Quinnipiac was a disaster in 2020, calling the wrong winner in Maine and the Georgia special (Sen. Jon Ossof ran behind in the general election but won the run-off). Marist did less polling but missed in North Carolina and overestimated Michigan for Senate and president.

In pre-election polling, all candidates should generally have a lower percentage than their election results, as each will pick some undecideds (and some voters drop out). Rarely do voters switch candidates late in a race. Both Quinnipiac and Marist overestimated the final percentage for every Democratic candidate they polled, except in the Texas Senate race (Quinnipiac).

Siena performed better but missed on North Carolina and significantly underestimated Republican margins in Senate races in Alaska, Montana and Iowa. Suffolk polled few states in 2020, and their missed margins were narrow, but all favored the Democrats. Suffolk has a bigger problem in that they only sample 500 voters — an inexcusably low number that raises their margin of error.

The 2022 elections are not looking better — with Quinnipiac, Marist and Suffolk appearing to be as biased as ever; Siena seems a bit better. Quinnipiac has only polled Georgia of late, but their lead for Democratic Sen. Raphael Warnock was higher than any other poll, except SurveyUSA, which had Warnock up a not-credible 12 points this month. Marist had Democratic Arizona Sen. Mark Kelly up 10 points in late September, much higher than any other poll, and had Democrat John Fetterman leading in Pennsylvania in September by 7 points, higher than any credible poll since the end of July. Marist is the only poll since September not showing Republican  J.D. Vance leading in Ohio. Siena is the only poll that shows GOP Sen. Ron Johnson trailing since Labor Day and the only poll with Democratic Rep. Tim Ryan leading Vance.

But it is the Suffolk polling that stands out. Suffolk has the margin for Kelly in Arizona higher than any of the last 10 polls, Democratic Sen. Maggie Hassan in New Hampshire higher than any of the last six polls. Only Suffolk has Democratic Sen. Catherine Cortez-Masto leading in Nevada and has the biggest post-Labor Day lead for Fetterman, except for the Marist poll. Suffolk has Republicans leading in Florida, North Carolina and Ohio, but by the smallest margins of recent polling.

Dangerous times for Democrats

Until election results come in, nobody will know for sure whether these and the wider batch of polls were biased. But recent history and the current record of the academic polling organizations, which should be less biased, should be a major concern. Not only does it look like a repeat of 2020, but the environment for accurate polling is getting worse.

Response rates are down significantly. With a response rate of just 0.4 percent, a telephone survey will have to make 250,000 attempts just to get to a sample size of 1,000. But it’s worse than that, as response rates are different by partisanship and demography. In order to get representative sample, a poll needs to reach quotas by age, ethnicity, gender, party, education and income. For any sub-group with lower response rates, it could be impossible to get a representative sample in Congressional districts or smaller states.

Compounding the problem is a cratering trust in the media among Republicans and independents. According to Gallup, only 14 percent of Republicans and 27 percent of independents have “confidence” in the media. With Republicans, that is a slight improvement from the nadir of 10 percent in 2020, but for independents, it is an historic low and down from 42 percent trust as recently as 2018.

Low confidence likely feeds into a suspicion of polling and may be severely increasing the difficulty of reaching GOP voters and — now — independents. If pollsters are desperately scraping for voters outside the Democratic Party, one has to question whom they are actually reaching to fill out their quotas and whether they are failing to include enough Republicans and independents at all.

Yet, Democrats and their media friends still can’t believe they might lose. The recent Siena poll showed significant movement toward Republicans, particularly with independent women. While it is true that the smaller the poll, the larger the error — and that splicing a poll into smaller demographic groups makes any conclusion less certain, the fact is that most polling has been biased in favor of Democrats. Thus, the true results might be even worse.

The bottom line is that recent polling has detected a trend toward GOP candidates with the big issue of the economy dominating with voters — as it normally does. Republicans are looking more and more likely to gain control of the Senate and not by a small margin. The House has been locked in for them for over a year.

It may appear now to be a sudden change, but the real story of 2022 may end up not that the Democrats collapsed into a rout but that they were always behind, the polls just missed it.

Keith Naughton, Ph.D., is co-founder of Silent Majority Strategies, a public and regulatory affairs consulting firm. Naughton is a former Pennsylvania political campaign consultant. Follow him on Twitter @KNaughton711.