Margin Of Error: Why Narratives – Not Facts – Often Drive Political Coverage And Public Perception
On the MSNBC News program "Morning Joe" last month, host Joe Scarborough drove the conversation about a poll he said showed former Vice President Joe Biden had fallen to fourth place in Iowa.
"Politics is all about momentum," he said, "and you can look at Biden and see that his candidacy has really struggled."
Scarborough's co-host, Willie Geist, casually interjected that the poll's margin of error indicated it was "statistically" a four-way tie.
The Quinnipiac University survey of 698 likely Democratic caucus-goers in Iowa showed Biden in fourth place with 15% support and U.S. Sen. Elizabeth Warren in first place with 20%.
But the poll had a margin of error of plus or minus 4.5 percentage points. That means the survey showed Warren's support could be as low as 15.5%, and Biden's as high as 19.5%. In other words, the margin of error applies in both directions to each candidate's estimated support.
“The reason we have a margin of error is because we are measuring the attitudes of some population, but from a sample of that population," says Joshua Blank, research director of the Texas Politics Project at UT Austin. "There’s a certain amount of uncertainty inherent in those estimates."
Media reports of public opinion surveys, especially in broadcast news, often ignore or downplay margins of error to the detriment of audiences and the frustration of some pollsters. And it's not limited to any specific news outlet; the practice is widespread.
"Both media organizations and pollsters must do a better job in explaining the results and other uncertainties of a poll," Spencer Kimball, the director of the Emerson College Polling Society, wrote in Media Ethics magazine about the 2016 presidential election. Kimball pointed out that polling data is all too often presented as a "precise figure."
Blank says he's troubled by how media coverage of polling can affect the very public opinion pollsters are trying to measure.
"The polling affects the coverage and then the coverage affects what people hear about the candidates, and that in turn affects the polling again and that becomes a feedback loop," he said.