What is polling really telling us this year?
By Matthew Gagnon
One of the stories — some would say scandals — of both the 2016 and 2020 campaigns for president was how broken polling was, particularly as it related to Donald Trump.
In state after state, particularly the swing states, polling underestimated Trump’s expected performance against both Hillary Clinton and Joe Biden, versus how he really did on election day. This polling error in 2016 ultimately produced wildly inaccurate predictions of an almost guaranteed win for Clinton. In 2020, the polls showed Joe Biden with a comfortable lead in most swing states, yet he only won the three “tipping point states” of the election by 0.23 percent (Georgia), 0.31 percent (Arizona) and 0.63 percent (Wisconsin).
Something has been very wrong in polling, yet interestingly it seems to be particularly pronounced when Trump is on the ballot, and less so when he is not. Polling was halfway decent in the 2018 and 2022 midterms, for instance. Theories abound for what has caused the so-called “Trump effect,” in polling, with “nonresponsive bias” — the historically low response rate of Trump supporters who pollsters attempt to reach — being the most persuasive explanation.
And yet, that particular problem was evident, and much talked about after the 2016 failure, and caused most pollsters to re-evaluate how they conducted their polls in an attempt to get things right. After doing well in 2018, it looked like they may have figured it out. But then 2020 happened and it was obvious that all of their efforts were for naught.
This has left pollsters once again asking themselves what they can do to correct for the mistakes, and get things right this time. This has led them to change how data is collected, weight the polling responses differently, and attempt to employ new tactics to sample particular groups.
But have these changes fixed polling? Or will the polls be wrong again in 2024?
As you look at current swing state polling, it is an important question to ask. If they haven’t fixed polling, Trump looks poised to win a fairly easy election in two weeks. If they have indeed fixed things, this will be yet another very close election.
Let’s review what’s going on.
As of Oct. 23, the RealClearPolitics aggregate polling average in the state of Pennsylvania shows Trump with a 0.8 percent lead in the state. He has led six out of the last seven polls of the Keystone state, and seven of the last 10. Still, a 0.8 percent lead would suggest that it is very close, and we may be in for a repeat of 2020’s results.
Yet if you go back in time, on Oct. 23 of 2016, Hillary Clinton was leading Trump by 5.6 percent in the RealClearPolitics average. In 2020, Joe Biden was leading Trump by 5.1 percent. In both years, Trump massively overperformed the polls, winning Pennsylvania in 2016 and losing in 2020 by only 1.17 percent.
If the polls have been corrected in Pennsylvania, then all that we are seeing is an accurate reflection of public opinion in a state that has been remarkably close in the last two presidential election cycles. If, however, there remains some kind of “Trump effect,” making the polls somewhat weighted toward the Democrats, then a shift toward Trump will have occurred.
This same story is apparent in every single swing state.
At this point in 2016, Clinton was leading by 11 points in Michigan according to the RealClearPolitics average, while in 2020 Biden was leading by 8.1 percent. Currently, Trump has a 1.2 percent edge. In Wisconsin, Clinton was up 6.2 percent and Biden was up 5.4 percent, while Trump is now ahead by 0.4 percent. Arizona showed Clinton up 1.5 percent and Biden up 2.4 percent, and now shows Trump up by 1.8 percent. Nevada had both Clinton and Biden up 5.2 percent, while Trump is now up 0.9 percent.
You also see this in North Carolina, where both Clinton and Biden were leading, but Trump is now up 0.4 today. It is also true in Georgia, where Trump is currently up 2.5 percent.
In every single swing state right now, Trump has some kind of lead, usually quite narrow. The question remains: Have the pollsters “fixed” the polls, and is what we see now accurate? Or is there still some kind of lingering problem that has led pollsters to underestimate Trump’s performance.
The answer to that question — and the implications it has on the outcome — is unknowable until after election day.
Less than two weeks to go.
Gagnon of Yarmouth is the chief executive officer of the Maine Policy Institute, a free market policy think tank based in Portland. A Hampden native, he previously served as a senior strategist for the Republican Governors Association in Washington, D.C.