Iowa has voted. Now what? I’ve got a few takeaways for those who wonder why the final result differed from the polling averages.

  1. Maybe the polls were wrong all along because they were counting too many Trump “unlikely voters”… This is a persuasive theory, in my view. In 2012, less than 6 percent of Iowa voters showed up to a caucus. And certainly, in 2016, the Republican side saw a 50% uptick in turnout (from roughly 120k caucus goers to over 180k). That’s still a very, very small slice of voters, and a far cry from the many public polls that assumed a third or more of Iowans were going to participate in the Republican caucus. Trump does best among unlikely voters, and those were the very folks who were likely being (over)counted in these media polls.
  2. …or maybe the polls were right at the time they were conducted but just missed a late breaking trend. This debate pops up every time there’s a prominent polling miss. When Eric Cantor’s primary defeat shocked the political world in 2014, Cantor’s pollster John McLaughlin explained the surprise as the result of a late shift toward Cantor’s opponent, Dave Brat. In the UK, polls missed the demise of Labour and the success of the Conservatives; the late shift theory was floated but mostly debunked. Of course, these two theories aren’t mutually exclusive. In this case, polls conducted during the very last week seemed to show a slight Cruz decline in the wake of his poor debate performance, though they did also begin to show signs of a Rubio surge. (Tweets from Scott Tranter who is a Rubio campaign analytics consultant suggest that their analytics were also showing some significant late movement.) Entrance polls showed that voters who made up their mind closest to caucus day tended to favor Cruz and Rubio over Trump. However, Trump’s loss may have always been baked into the cake, that he was never ahead or as far ahead as polls showed. We just don’t know. Both of these things could be right! Which brings me to…
  3. Beware headlines declaring “X stopped Trump/boosted Cruz/boosted Rubio in Iowa.” Part of the reason why the “late breaking voters” theory is so appealing is that it absolves public pollsters of responsibility for getting it right. Hey, we were probably right at the time, but people changed their mind! Not our fault! But the other part of the appeal is that it allows people to make shaky causal inferences about what caused the supposed shifts in support. These are the stuff of sexy headlines and breathless coverage of the consultants and tactics that supposedly “ushered so-and-so to victory.” For example: If you came out and said that Trump skipping the Fox News debate was a risky proposition, it can be really tempting to go on air and declare “See, I told you…Trump skipped the debate and LOOK! It killed him!” But nothing happens in a vacuum, and without a randomized control trial or a peek into the alternate universe where Trump didn’t skip the debate or where your ad didn’t air, we can’t know. So by all means, if your candidate exceeded expectations last night, celebrate! Enjoy it. But we should all skeptical of any one individual, consultant, or organization that tries to take credit for some late Cruz/Rubio surge or Trump decline.
  4. With that said, there is clear data suggesting one big choice paid off for Rubio and Cruz: expand the electorate by targeting sporadic, but nonetheless registered, voters. Lots of folks I admire were holding on to the hope that since Republican voter registrations in Iowa wasn’t seeing some enormous swell in new registrants, the Trump surge wouldn’t come to pass. This was part and parcel of the “high turnout favors Trump” theory that I admit I ascribed to, given how well Trump performed with those who had never caucused before in polls. First-time caucus goers did increase from 38 to 46 percent of all caucus goers, or from just under 50,000 voters in 2012 to over 80,000 this time around. And those voters slightly favored Trump. But there’s more than one way to grow turnout, and simply turning out new voters isn’t the only way. Growth in caucus participation yesterday didn’t just come from newbies, but at least half of the growth seems to have come from past caucus-goers who sat out in recent years. They knew how and where to caucus, but hadn’t bothered to do recently. I tweeted something to that effect (suggesting Trump should be trying to pursue these voters -ha!), and the aforementioned Rubio analytics guru simply tweeted a smiley face at me. That was their plan, and they executed on it. (As did the Cruz team, one suspects, given their controversial “social pressure” mailing that made headlines by shaming infrequent voters into getting out there on Monday.) This differs from the Obama 2008 strategy in that Obama changed the game by registering TONS of new vote. The Republican victors last night tweaked the formula by going after a slightly easier piece of the pie in their race to grow turnout: the already registered, but infrequent, voter.
  5. Digital analytics remain an interesting piece of intelligence. Here, I confess my own biases: my firm looks at things like search trends and social media discussion volume to supplement our understanding of “what’s happening” in an election. We don’t use them to replace polls, but do think they’re an interesting additional window into what people are thinking about. I’ve got another post coming shortly with my Echelon Insights co-founder Patrick about the methodological lessons from the Iowa caucuses, but one thing we did see was a decline in searches for Trump relative to Cruz and Rubio in the final few days. Again: not a replacement for polling, but nonetheless an intriguing data set to keep an eye on.