According to recent media commentaries on the midterm elections, Republicans are teetering on the edge of disaster. Where most observers had been predicting a big Republican win as recently as mid-July, many of the “experts” are now rushing to count the GOP out, even though the election is still over a month away.
Leading election analyst Nate Silver of FiveThirtyEight gave Democrats a 70 in 100 chance to hold the Senate in his recent and widely reported forecast. Other analysts are likewise now projecting growing Democratic strength. The impact of these dire predictions can have a significant impact on fundraising, volunteer enthusiasm, voter turnout, and resource allocation if party supporters accept as fact that the Senate is already lost.
But how reliable are conclusions made on elections when a major part of their basis is determined by polling data conducted two months before Election Day? Let’s take a look.
In 2020, media experts had all but written off Maine Sen. Susan Collins’s reelection bid two months out. By mid-September, a Quinnipiac poll had challenger Sarah Gideon up 12 points over the incumbent and the New York Times/Siena poll had Collins down by 5. In November, voters delivered Collins a comfortable 9-point victory.
In Iowa, incumbent Joni Ernst was considered very vulnerable. In mid-September, the Des Moines Register and Monmouth polls each had Theresa Greenfield leading by 3 points, and the New York Times/Siena poll had her up by just 2. In November, Ernst comfortably retained her seat by nearly 7 points.
Meanwhile, in South Carolina, polls were pointing to a very tight contest between Sen. Lindsey Graham and challenger Jaime Harrison. In August and September, Quinnipiac had the race tied and CBS News/YouGov, showed Graham clinging to a 1-point lead. Graham won reelection by 10 points.
Of course, 2020 wasn’t an isolated case. We saw similar misleading polling in 2018 and, famously, in 2016 as well. Since there are no compelling reasons to think this year’s polls will be more accurate than those in recent cycles, there is a significant probability that this year’s polls are likewise understating Republicans’ strength. Several commentators from the Left and Right have raised this issue in recent days, with many wishing there was a way to improve polling accuracy. At the Polling Monitor, we believe we have devised a very helpful methodology to address these concerns.
Our system is straightforward. We judge new surveys by each polling entity’s historical performance. Our comprehensive database contains the results of each pollster’s surveys over the past three election cycles in 2022 battleground states. Our analysis compares a pollster’s final pre-election polls in all presidential, Senate, and gubernatorial contests in those elections to the actual election results to determine how accurate each pollster has been in each state it has surveyed. Our database provides readers with the ability to see what the results of a new poll would be if we “adjust” those results by the historic performance of the pollster.
This methodology indicates overall that the current predictions of doom and gloom for Republicans are premature. For example, recent surveys in Pennsylvania and Georgia showing solid Democratic leads are, when adjusted by the pollsters’ track records, much tighter. In Pennsylvania, the Allentown Morning Call’s latest poll shows Republican Mehmet Oz down by 5 points, but over the last three cycles in Pennsylvania, Morning Call polls have overweighed Democratic candidates on average by 3.6 points. Adjusting for that historical bias, you end up with a very tight contest.
Similarly in Georgia, Quinnipiac has Democrat Ralph Warnock up by 6 points over Republican Herschel Walker. But in Georgia, Quinnipiac has overweighed Democrats by an average of 4.6 points since 2016, also suggesting a much closer contest.
Various other recent polls indicate results that are more favorable to Republican candidates when subjected to a historical analysis. Hence, the euphoria among those who hope for a Democratic victory this fall, largely based on polling that has been historically inaccurate in recent elections, is clearly premature.
Meanwhile, Republicans shouldn’t throw in the towel, panicked by polling data that are likely as inaccurate as polls from the past three cycles. This campaign is far from over.
Spencer Abraham is co-founder of the Polling Monitor (pollingmonitor.com) and served as an aide to the Senate Republican leader. Colin Pitts is co-founder of the Polling Monitor (pollingmonitor.com) and is a political consultant and pollster.
