Why Were The Polls So Wildly Off In Tuesday's Elections?

This is a key question that nobody's asking.

• The Real Clear Politics average for the New Jersey governor's race had Sherrill ahead by 2.7%, and many commentators had this within the margin of error. She won by 56.3% to Ciattarelli's 43.2%, a 13.1% margin.
• The RCP average for the Virginia governor's race had Spanberger ahead by 8.5%. She won by 57.2% to Earle Sears's 43.6%, a 13.5% margin.

I asked Chrome AI mode, "What is considered an accurate poll election prediction?" It answered,

An accurate poll election prediction is one where the actual election outcome falls within the poll's stated margin of error and confidence interval. However, polls often understate the full range of potential error, meaning the reported margin of error may need to be wider in practice to be truly accurate.

The RCP poll averages don't have a stated margin of error, because they are just a composite of individual polls with differing margins of error. But let's look at individual polls before those elections. According to Newsweek,

The latest AtlasIntel poll (October 25-30; 1,639 likely voters, =/-2 percent margin) gives Sherrill a narrow edge, 50.2 percent to 49.3 percent.

Meanwhile, a John Zogby Strategies poll, carried out among 1,205 people between October 31 and November 3, places Sherrill ahead (55 percent) of Ciattarelli (43.4 percent). There is a +/- 2.9 percentage point margin of error.

Research Co. also has Sherrill in the lead with 3 points more than Ciattarelli (52 percent to 48 percent) among 450 likely voters questioned between November 2 and 3—there is a =/- 4.5 percentage point margin of error.

Only the Zogby Poll, which had Sherrill ahead of Ciattarelli by 11.9 points vs the 13.1 point actual difference, was within the 2.9 point margin of error at 1.2 points. The others were wildly off. At the same link for the Virginia race,

Trafalgar Group has given [Spanberger] a 5.6 percentage point lead over Earle-Sears (49.8 percent to 43.3 percent).

Meanwhile, the latest Insider Advantage survey of 800 likely voters (questioned between November 2 and 3) has Spanberger at 50 percent and Earle-Sears at 40 percent. There is a +/- 3.46 percentage point margin of error.

Research Co.’s poll (November 2 and 3 among 450 likely voters) shows Spanberger with a lead of 54 percent over Winsome Earle-Sears’s 46 percent. There is a =/- 4.6 percent margin of error.

No margin of error is cited for the Trafalgar poll. The Insider Advantage poll had Spanberger ahead by 10%, with a margin of error at 3.46% Spanberger won by 13.5%, just outside the poll's margin of error. The Research Co poll had Spanberger ahead by 8 points, with a 4.6% margin of error. But Spanberger won by 13.5%, again outside the poll's margin of error. The Virginia polls were only slightly more accurate than the New Jersey polls, with the result for RCP that they got garbage in-garbage out in both races.

The effect of this sloppy work was to build an expectation that the Republicans had a chance of winning, especially in New Jersey. As far back as July, the conventional wisdom went,

Mikie Sherrill by a landslide?

Nobody believes that will be the result of New Jersey’s piping hot governor’s race come November. Not the Republicans. Not the Democrats. Not the campaigns. And not the Rutgers-Eagleton pollster who last week released a survey that says Sherrill, the Democratic nominee, has an early 20-point lead over Republican Jack Ciattarelli.

So somehow, the pollsters managed to create an inaccurate impression that Ciattarelli was going to come from behind and score an upset victory over Sherrill, when Sherrill nevertheless won by a landslide anyhow. Who benefitted from this?

On one hand, the media did, simply in terms of clicks and pageviews and eyeballs. It generated national interest in an otherwise blah and predictable local race, when the same Republican had already run and lost against a Democrat twice. The alt media also seized on the story as a potential Republican success story.

But let's recall what Rush Limbaugh said about the polls:

The polls are nothing more than a tool for making public opinion. The polls are not taken to reflect public opinion at all. The primary purpose of polls is to depress and dispirit Republican voters by making them feel constantly like they’re in the minority.

In this case, the polls and the media worked to create one "piping hot governor's race" that was never piping hot at all, while feeding completely unrealistic expectations that Republicans could win in Virginia. And when rhe optimistic buildup fell completely flat on election day, the media -- even the putative non-partisan commentators -- spun it as a disaster. Thus Sean Trende at Real Clear Politics:

The GOP went into the night with reasonably upbeat expectations. Polls showed a close race in New Jersey, and it looked as though Republicans might keep the attorney generals office in Virginia. Neither outcome happened.

. . . It's hard to see what good news there could possibly be for Republicans here. Incoming Virginia governor Abigail Spanberger won by 15 points, while even Jay Jones, the problematic attorney general candidate (see below) won by around 6 points, similar to the Democrats margins in 2017. In New Jersey Sherrill nearly equaled Gov. Phil Murphys 2017 margin. Counties like Bergen and Passaic, where Donald Trump made big gains among minorities last year, swung back toward Democrats.

Except that the Republican expectations were fed by the polls, whose purpose in life, at least according to El Rushbo, is to depress and dispirit Republican voters. But wait a moment -- here's Sean Trende, the public face of the Real Clear Politics polling averages, flat out admitting that it was the polls -- and the RCP averages -- that misled Republicans. Shouldn't he be asking himself some serious questions here? Apparently not. RCP is going to keep on getting a few good polls, mixing them in with a lot of garbage polls, and averaging it all out. the main effect is to depress and dispit Republicans, even when they're creating putatively optimistic trends out of pure moonshine.