One day before the Iowa caucuses, Emerson College released the final Iowa poll. Its results were a little narrower than most of its immediate predecessors, but it showed the same thing they did: Donald Trump, the blustery real estate magnate who was firmly ensconced as the Republican frontrunner despite his many departures from small-government conservatism, with a lead over Ted Cruz. Real Clear Politics, likely the best-known aggregator of polling data, had Trump with a 4.7-point lead over Cruz, and the renowned political statistics blog FiveThirtyEight’s most likely predicted outcome was a narrow Trump victory. A weighted average of the polls showed the same thing, and better still, pulled Trump outside the reported margin of error for the weighted estimate.

There was, perhaps, a little cause for alarm within Trump’s camp. Both of the most recent polls from Iowa showed Cruz narrowing the margin to within the poll’s error, but even so, neither predicted that Cruz would take the lead. Trump’s unconventional campaign focused on rallies and top-down momentum, rather than grassroots and legwork, appeared to have upset conventional wisdom and would notch up a victory before anyone else had a chance to go on the board.

The polls were wrong. Cruz not only won, he won with a record number of votes and by more than 3% – a commanding margin in a race with ten other candidates. The margin between Cruz and Trump was actually wider than the margin between Trump and the third-place finisher, Marco Rubio. Every poll showed Trump with a lead, many with a lead greater than the margin of error, and every poll was wrong.
The total breakdown of polling isn’t necessarily a new thing – caucuses are hard to predict – but it is worse this year than in the past. In 2008 a weighted average would have predicted the actual winner, Mike Huckabee, and even in 2012, when Rick Santorum pulled off one of the more shocking come-from-behind victories in recent memory, polls showed the surge, and a plot of the pre-caucus trajectory would have predicted the eventual outcome. This year neither the actual numbers nor the trajectory would have predicted the winner. The polls were simply wrong.

There will undoubtedly be many post-mortems trying to explain why the pollsters failed today, but there’s a simple enough explanation: Trump’s support is the political equivalent of a rice cake – crunchy, usually very noisy, air. He presented a platform as schizophrenic as his campaign, a muddled, angry caricature of what a New York-liberal might imagine to be conservatism. Pushing an isolationist view on the borders and a nationalist, even bloodthirsty, foreign policy, at the same time that he advocated positively socialist positions on issues like healthcare and a fascist stance on religious expression, Trump simultaneously attempted to appeal to the Republican conservative base and emphatically rejected crucial elements of a conservative worldview. His campaign preparation showed the same inattention to detail, eschewing hard work in favor of exciting rallies and passing over a well-developed organization in favor of slip-shod tweet-storms. He targeted a particular subset of voters, those who had not thought out their positions clearly enough to recognize the caricature.

It turns out, people too lazy to think through a consistent, coherent philosophy of government aren’t likely to be diligent enough to spend three hours casting a vote. Trump’s no idiot, although he clearly thinks his supporters are, and he’ll adapt. Future states will be primaries, which appeal to Trump’s brand of voters more than more time-consuming caucuses, and Trump will undoubtedly work to solidify his organization in the future. Tonight, though, it doesn’t matter. Trump’s “yuge” lead in the polls was finally put to the test, and it collapsed under pressure. There’s a long road ahead – Iowa only controls a little more than 1% of the total delegates – but first blood and the confidence to forge ahead regardless of the polls goes to the underestimated senator from Texas.