Why The Pollsters Got It Wrong In Iowa (Hint: Voter-Targeting Data)

Pre-caucus polls had Trump in the lead, but Cruz’s data-driven ground game won out in the end.

Why The Pollsters Got It Wrong In Iowa (Hint: Voter-Targeting Data)
[Photo: Flickr user Evan Guest]

Obviously, the pollsters got it wrong about Trump’s chances of winning the Iowa caucuses Monday night. One key reason may be that they didn’t account for the superiority of Ted Cruz’s data-driven ground game over Trump’s virtually non-existent one.


Poll results from just before the Iowa caucuses predicted that the longtime front runner Donald Trump would win the night. The widely cited Quinnipiac University poll had Trump ahead of Cruz 31% to 24%, expanding Trump’s lead from a week earlier. It had Marco Rubio at 17% and Ben Carson at 8%.

The Des Moines Register‘s final poll had Trump winning 28% to 23% over Cruz, with Rubio in third with 17%.

When the votes were all in on Monday night, however, Cruz had won in an upset, collecting 28% of the vote, to Trump’s 25% and Rubio’s 24%.

Photo: Flickr user Gage Skidmore

Among other things, the pollsters assumed that people who expressed support for Trump on the phone would actually come out to caucus.

The GOP caucuses usually draw around 120,000 people in Iowa. It was thought–even by people we talked to inside the Cruz campaign–that if more than 150,000 people showed up to caucus it would probably be a big night for Trump.

The demographic groups (white, younger, no college degree) that support Trump most fervently are also the ones who traditionally don’t make it out to vote. So high turnout numbers, people thought, would mean that many in those groups did.


But even this proved incorrect. The turnout for the GOP caucuses Monday night broke records, turning out large numbers of people who had never before participated in the caucuses. More than 180,000 Iowa Republicans voted, ABC News reports, compared to the 121,000 who showed up in 2012.

With their strong performances, Cruz (and Rubio) must have accounted for a large numbers of those new caucus goers.

In some ways the Cruz and Trump campaigns in Iowa couldn’t have been more different. In general, Cruz’s campaign was like an army of on-the-ground troops, while Trump’s flew over and dropped bombs from 30,000 feet.

Cruz had a well-peopled and well-organized ground game, Trump relied on his celebrity and his events to fire up supporters and move them to come out to caucus.

Cruz used data science to carefully target voters who were likely to come out to caucus, then used a high-touch communication strategy to move them to action. The Trump camp believed the constant television news coverage focused on the celebrity candidate would do the work of a finely tuned ground game.

Photo: Flickr user Max Goldberg

Only in October and November did Trump’s campaign begin getting serious about using voter data and analytics. The campaign licensed a voter database from the Washington-based company L2, and also hired a couple of ex-RNC operatives to work with the data. Several GOP data pros we talked to said the Trump data operation was rudimentary, and too little too late, for Iowa at least.


The lesson seems clear, at least in Iowa.

The low-touch celebrity campaign is enough to make supporters tell pollsters they support the candidate. But it may not be enough to get people off the couch on caucus night, drive to the caucus site (a school gymnasium or a church basement), and donate a few hours of their time to the cause.

Cruz’s approach of using data to know the demographic, psychographic, and even behavioral attributes of individual voters seems to have won the day. The Cruz campaign spent more than $10 million on data and analytics during 2015, far more than any other GOP candidate.

The Cruz campaign paid out $7.2 million in fees for data and analytics in the fourth quarter, while the Trump campaign paid out just $738,517.

A voter data expert told me last week that Iowa voters have become accustomed to campaigns reaching out to them personally several times before caucus night. That’s what Cruz did, by phone, and by meetings at the front door.

Perhaps more importantly, the Cruz campaign had lots of intel on which doors to knock on, or which numbers to call, to find the people who could most reliably be persuaded to vote Cruz.


In the end, that highly targeted, high-touch approach moved thousands to come out and vote for Cruz.

And, to be fair, polling in Iowa isn’t easy, as Nate Silver at FiveThirtyEight points out. Caucuses are harder to predict than regular primaries, especially when there are lots of candidates in the running, as there are this year in the GOP field.


About the author

Fast Company Senior Writer Mark Sullivan covers emerging technology, politics, artificial intelligence, large tech companies, and misinformation. An award-winning San Francisco-based journalist, Sullivan's work has appeared in Wired, Al Jazeera, CNN, ABC News, CNET, and many others.