WASHINGTON, July 15, 2015 — This was a week of breathless headlines for Donald Trump:
“Donald Trump leads the GOP field in latest poll,” said Politico.
“Trump Towers Over Republican Rivals in New Poll,” quipped TIME.
Those came after a Suffolk University/USA Today poll showed the real estate magnate turned reality show host winning 17 percent support among prospective Republican primary voters.
Those headlines are wrong.
Trump is not the leading contender for the Republican nomination, at least not based on these polls. Whether serious or not, his candidacy offers news media consumers a valuable lesson: Don’t believe everything you read.
And that Suffolk/USA Today poll provides an excellent example of this lesson.
For starters, the poll is national in scope, pulling from various regions of the country approximately equal to population. That would make some sense for a truly national election, but the primaries are a series of smaller, statewide elections that happen over several weeks.
It’s no secret why New Hampshire, Iowa, and other early primary states become so important every four years: Winning attracts attention and money. A big win in an early primary means a bump in news coverage. Other campaigns are rightly focusing their energy and resources building up name recognition in these few select places.
Trump, in the Suffolk poll, has higher name identification than any other Republican candidate. That advantage might be mitigated when polling, say, New Hampshire, Iowa, or South Carolina, where other campaign operations are already working feverishly. But news organizations aren’t conducting state-by-state polls as frequently, even though they’re more relevant at this point than national polls.
On top of that, Suffolk’s modeling methodology, while good for a general opinion poll, makes for a poor predictor of electoral success. The poll’s demographic and regional quotas are based on 2010 census data — not 2012 Republican primary turnout, or 2014 participation, or any other model of who would actually show up to vote.
Further, the survey selected respondents by asking for the youngest adult in the household: in other words, the person who is demographically the least likely to show up on Election Day. And while the survey did ask respondents how likely they were to vote, self-identification is a notoriously unreliable way to determine likely voters.
This is not to pick on Suffolk and USA Today, or, for that matter, Trump. Other news and polling organizations do the same thing. The problem is with media outlets that report the results without critically examining their meaning. Trump’s recent edge in national polling, much like former Florida Governor Jeb Bush’s longer-term edge, could just as easily be a function of high name recognition.
Combine that with polls which unwittingly reach lower-information voters who aren’t even paying attention to the race yet, and you can see how skewed the results can be.
Giddy headlines which herald the rise of Trump based on polls like this are just plain silly. Sadly, it means plenty of attention from media figures who should know better.