Arrow-right Camera

Color Scheme

Subscribe now

This column reflects the opinion of the writer. Learn about the differences between a news story and an opinion column.

Spin Control: Once again, news media relied too much on polls to cover a presidential race

The CNN building is shown in Atlanta. AI chatbots have been found to copy articles published by the New York Times, Reuters and CNN.  (Elijah Nouvelage/Bloomberg)

After every presidential election, the soul-searching begins, not just by the losing candidates and political party, but by the news media that covered it.

The candidates and parties often make major changes. The news media, not so much.

For decades, the great minds of journalism (not an oxymoron, although one could argue the point) have scolded reporters and editors for their overreliance on polls. Reporters and editors have nodded their heads, murmured their general agreement and made promises to do better.

Four years later, national print and television news organizations generally try to do better by spending more on polls, not less. They start polling earlier and spend more time discussing them.

Their pollsters rewrite or at least tweak their algorithms to pick up more of the people they missed or collect fewer of the people they oversampled. They look at ways to compare and contrast the polls other organizations conduct, apparently on the theory more is better, even if that more has just as many problems as their own polling.

It has become standard for cable news talking heads to mention, when talking about their latest polling numbers, that polls don’t predict the outcome of the election. But that may be hard for viewers to grasp when a one-hour newscast spends more than half of its time dissecting poll numbers.

Probably the best – or is it the worst? – example of this overreliance on polling occurred last Sunday, when national news organizations went what might best be described as “coocoo for Cocoa Puffs” over a Des Moines Register poll that showed Vice President Kamala Harris running 3 percentage points ahead of former President Donald Trump in Iowa.

This result ran counter to common sense. A Democratic presidential candidate hasn’t won Iowa since 2012. It has a solidly Republican Legislature and hasn’t elected a Democratic governor in 14 years. Trump carried all but one county in the Iowa caucuses at the beginning of the year. No other poll suggested Harris was ahead, although not much polling had been done in Iowa because, well, Trump was expected to win handily.

But for much of last Sunday, pundits found ways to elevate those results, noting that the pollster has innovative methods, a good record and sometimes detects trends that others miss. Some even suggested the poll could be revealing a shift from Trump to Harris in other states outside of Iowa.

On Tuesday night, Iowa was called for Trump quickly. He topped Harris by more than 13 points or more than 220,000 votes.

For the pollster, this was likely what is called an outlier poll. All polls carry what’s known as a margin of error, a calculation that if the same survey were repeated at the same time with a similar sampling of voters, the results would be within a few percentage points 95 out of 100. But that means five other times, the results could be outside that range. Sometimes way outside.

Every political reporter, editor and pundit in the nation knows about outliers. Last Sunday, some even suggested the Iowa poll might be an outlier. But many spent far more time talking about why the poll might be right and what would happen if it is, than explaining, based on common sense, why the poll was likely wrong.

What was happening was an inherent problem with interpreting polls called confirmation bias. The poll allowed reporters or pundits, who thought Harris would win or the race was close, to give the poll more credence because the results lined up with their beliefs.

There are good and bad polls, and it’s probably not fair to pick on one. Good polls can be important for campaigns trying to assess the public mood over key issues. They are valuable to news organizations for pointing out what issues concern their readers or viewers.

But much of that information gets short shrift in favor of the head to head or “horse race” numbers. All the polls showed inflation and the economy were a major concern, but there was less in-depth coverage of the two candidates’ positions in the closing weeks than comparisons of the various Trump vs. Harris numbers.

If the national news organizations really want to make their coverage more meaningful, here’s an idea: Skip the horse-race numbers in the two weeks before the election because we’re going to find out the real results soon, anyway. Send more reporters into the field to talk to voters and listen to what they say. Poll on what voters wish they knew more about the two candidates, and report whether their policies or track records align.

People who really want to see horse race polls will find them elsewhere, and most of them already know how they’re going to vote.

More from this author