A Brief Lesson in the Futility of Small, Early Presidential Polls

Attention, attention: Mike Huckabee is set to win the 2016 Republican nomination. Sort of.

Attention, attention: Mike Huckabee is set to win the 2016 Republican nomination. Sort of. That's only if winning Iowa means he'll win the nomination. And if a new poll from Suffolk University released on Wednesday showing him with a slight lead is to be believed. And if you ignore the massive margin of error on that poll.

We can't fault Suffolk for releasing the survey. They were polling people on Iowa races, including the upcoming gubernatorial and Senate races. And, because they had people on the phone, they also asked about 2016. Democrats were clear: Hillary Clinton leads in the state by a wide margin. The Republican field, however, was "muddled," in the words of McClatchy's David Lightman.

Nonetheless, there's that good news for Huckabee. "Iowa survey in 4 races shows Braley, Branstad, Clinton and Huckabee leading," the Des Moines Register trumpets. Huckabee's lead is razor-thin, though, as the full slate of results show.

If you go look at the Suffolk poll, though, it's worth scrolling down to the bottom of the page. There, a footnote. "The margin of error is +/-8.7 percent for the 127 GOP caucus-goers," the report says, "and +/-8.4% for the 135 Democratic caucus-goers surveyed." Which means that the graph above isn't quite right. Including error bars, it should look like this.

So the actual results for any of the candidates could fall anywhere along the red line. For more than half the candidates, the margin of error is larger than the vote total! What the Suffolk poll tells us is that literally any of these candidates could be in the lead. It is muddled. Huckabee is not leading.

Which of course sets aside the facts that 1) none of these people is officially running for president yet and 2) no one has campaigned at all for it. The poll would be relatively unhelpful if it were conducted the night before the caucuses; a sample size of 127 is so small as to be unhelpful in any context. But this far out? It tells us literally nothing useful about Iowa — much less the rest of the country.

So why cover it? Because people love love love presidential races. Love love love love them. Suffolk knew that more people would be interested in their work if, alongside the good news for Terry Branstad, they offered up the name "Hillary Clinton." (Her 51-point lead is well outside that 8.4 percent margin of error.) And that's fine. Just recognize what they're actually telling you: There is no possible way to tell who will win in 2016 at this point. That's actually useful data.