Election polls are flawed. They’re also the best tool we have.

Political ViewpointsThe Bangor Daily News publishes first-person viewpoints from experts to provide comprehensive coverage of the ideas shaping Maine politics. Send your pitch to the politics editor here.
Traverse Burnett is a research director at Digital Research Inc., a Portland firm that conducts surveys for private-sector clients as well as regular political polls. The Bangor Daily News has previously paid the company for both types of surveys.
Election years can be very good for those of us in the polling business. We’re typically asked to do more surveys and focus groups, and those projects get more attention than almost any of our other work. 
Most of us in the consumer insights field make our living providing businesses and other private clients with information about public and consumer opinion — helping them make decisions that can improve their bottom line or solve a particular challenge.
I’d like to think we’re good at that. Our clients can measure our accuracy based on outside measures of sales, market share and customer satisfaction, and the better we are at finding the trends and gauging underlying currents in consumer sentiment, the better our business is. 
But predicting an election is an entirely different matter. 
This is especially true for the toss-up 2024 presidential election between Vice President Kamala Harris and former President Donald Trump. Many talented groups are trying to predict the outcome of the election and are only able to reaffirm the race is “close” and “within the margin of error.” Frustrating as it is for the pollster and the public, it’s likely the truth. 
Even in the best of times, pollsters have to make a number of judgment calls in the methodology, sample design and conclusions that shape our predictions of key races. Slight shifts in any of these components of a survey have an outsized impact on our overall predictions.
Survey researchers are haunted by the failure of 2016, when disengaged voters weren’t well represented in polls yet proved decisive for Trump at the ballot box. The predictions since then have generally come much closer to the actual outcome, but the fact that most polls of national and swing-state voters are incredibly close suggests differences in methodology aren’t dramatically impacting the predictions this time around.  
It’s probably also worth noting that Maine presents a special case on its own. For one, we’re highly engaged, with some of the highest rates of voter registration and participation in the nation. Since the voters we survey in the state are highly likely to actually vote, predicting the outcome of an election can be easier here than in other states.
At the same time, Maine allows ranked-choice voting for national elections, which can drastically change how accurate a poll may be. Even a well-run poll may put one candidate across the threshold needed to avoid a ranked-choice count, meaning we wouldn’t incorporate a runoff in our prediction. But things can change dramatically if voter turnout comes in slightly differently than we estimated or if just a few voters unexpectedly move toward one of the candidates. 
That sort of small shift may not matter much in determining the outcome of a standard election, but a runoff can change the outcome of a race in ways no poll can predict. That’s what happened in 2020, when most polls pointed to Democrat Sara Gideon winning against incumbent U.S. Sen. Susan Collins in a ranked-choice runoff. 
In the end, Collins was able to meld her Republican base with unexpected swing voter support to win an outright majority and avoid a runoff. Even if estimates of voter preference weren’t drastically different from reality in that race, predictions of the results definitely were.  
There’s also a technical difficulty involved in asking voters to rank multiple candidates sequentially for a poll in a ranked-choice election. It’s not so much of a challenge with an online survey, but it does have the potential to constrain or impact the accuracy of a telephone poll — no longer the gold standard it once was in polling methodology. 
Elections are complicated, and we will never be able to see everything before a race is over. And while polling is flawed, it remains the best way to objectively understand and estimate where things stand in the political and business world. Even if the election predictions aren’t spot-on in Maine, swing states or the country as a whole, they still tell us something.
What we know they’re telling us now is that America is evenly divided between two dramatically different paths, although you probably knew that already.