O’Dwyer’s: Yes, Polling Still Matters

Written by

Matt George

Partner, Head of Research

View full profile


I recently had a conversation with the personification of “margin-of-error.” With a mischievous grin, a small business owner in North Carolina recounted with glee how he had personally thwarted the national polling industry back in 2016 and again in 2020. 

“One day, I’d answer that I was for Trump. The next, I’d be for the Democrat. It didn’t matter—I just said how I was feeling on any given day. It wasn’t about how I voted on election day. They asked me who I was going to vote for weeks in advance, and I flipped back and forth.” 

For someone who constructs, fields and analyzes thoughtful public opinion surveys, his answer infuriated me to my very core. He was unapologetic. But he was also not wrong.

Every two years, our country holds an election. And in between those two years, an entire industry works to build an infallible crystal ball that can predict the future with near certainty. 

And every two years, when that crystal ball invariably fails to be infallible, journalists, pundits and politicians sound a familiar cry, “Polling is broken. It doesn’t work.”

I’ll be the first to admit—if you expect polling to be 100 percent correct in its predictions, 100 percent of the time, then, yes, polling doesn’t work. Polls are a blunt instrument. They can’t predict the future. They’re simply a measure at a single given time. 

But in a culture that sees poll numbers as the goal rather than a metric, this basic framing gets lost in the noise. How many times have you seen a politician celebrate a poll that showed them up in their race? How many times have you seen pundits herald one poll as proof of one candidate’s dominance over another?

We need to rethink polls as means of unearthing valuable trends and currents of public opinion that can guide us to reasonable conclusions through careful contextual analysis. Polls can help us know where a candidate, company or organization is underperforming their potential, allowing them to adapt. 

For example, when Glen Youngkin’s Virginia gubernatorial campaign learned through trended data polling that the likely voter electorate was concerned about education, it pivoted, making the issue one of the foundations for their platform. This defining issue contributed significantly to his campaign victory. 

If you take the temperature today, you’ll have a single data point about what the weather is like at a single moment in time. But that data bears no real insights or predictive capability on what the temperature will be in an hour… or tomorrow… or a in a week. New day… New temperature. 

But what if you took the temperature every day for a week… or a month… and noticed that it was cold each day and getting colder over time. You could begin to identify trends and make reasonable conclusions about those trends—like that it might continue getting colder or that you might be entering a period of prolonged cold—Fall or Winter. Are there some hot days in Fall? Sure. But the overall trend provides you insight that you can use to plan appropriately. 

Polling can unearth those trends. Polling can measure momentum. 

In 2016, most polls predicted Hillary as the expected winner. But they also identified significant Trump momentum in late October after FBI director, James Comey’s letter to Congress. Which measure should have had more weight in our news cycles? The individual temperature readings which showed Hillary’s numbers as larger than Trumps… or the trend of Trump’s momentum? 

In 2020, preelection polling predicted a decisive Biden victory that turned out to be significantly less of a landslide than expected. Many polls in the Midwest were wildly off when compared to the results. But yet again, polling also identified Trump gaining momentum in the final weeks of the election, cutting into the decisive victory many Democrats had planned. 

In 2022, let’s learn our lesson and pay attention to the trends the polls are showing us.

As of late-October, polling in the lead-up to the 2022 Midterms looked familiarly favorable for some Democrats in high-profile Senate races, who were long expected to suffer electoral losses this year. For example, polls have shown Georgia Democratic incumbent Raphael Warnock leading Republican challenger Herschel Walker by as much as six points. But let’s remember that in 2016 and 2020, polls consistently underrepresented non-college educated voters, and had significant trouble eliciting accurate responses from Republican voters. Wouldn’t it be reasonable to consider that the trend of polls underrepresenting Republican turnout might continue? 

So, if polling sometimes gets the answer wrong, does it still matter? You bet. Especially when used to identify and harness momentum. 

It’s worth noting that polling can do so much more than make election predictions. It can uncover trends in how the public perceives the economy or detect critical changes in our culture. It can warn companies when their team members’ work engagement and morale might be slipping. Accurately measuring public opinion momentum can predict which products consumers will buy, which ideas voters will support, and which values we adhere to as a society. Finding the momentum of public opinion provides competitive advantages to companies and organizations by empowering them to adapt and remain nimble. 

The next time you read a political poll result, read it more as a single suggestion rather than a prediction. And then, keep looking to see if you can uncover the trends and momentum shifts that explain what people are really thinking.