Iowa polls, “expectations” were a pale shadow of the truth

Graphic: New York Times

For months, the political media have had nothing to talk about but polls and expectations. Yesterday reality crashed the party in Iowa, revealing the vacuity of the polls and the “expectations game.” Let’s examine how poll/expectations-based “analysis” differs from reality-based analysis.

The polls were totally off

Let’s compare two recent Iowa polls, the Quinnipiac poll done Sunday and the Des Moines Register poll just before that, with what actually happened.

Quinnipiac Republican Poll: Trump 31%, Cruz 24%, Rubio 17%.
Quinnipiac Democrat Poll: Sanders 49%, Clinton 46%

Des Moines Register Republican Poll: Trump 28%, Cruz 23%, Rubio 15%
Des Moines Register Democrat Poll: Clinton 45%, Sanders 42%.

Actual Republican Caucuses: Cruz 28%, Trump 24%, Rubio 23%
Actual Democratic Caucuses: Clinton 50%, Sanders 50%

The polls accurately predicted that Trump, Cruz, and Rubio would be the top three Republicans with the rest far behind. They overestimated Trump by 4-7%, underestimated Cruz by 4-5%, and underestimated Rubio by 6-8%. The polls differed by 6% on the Democratic result, each of them off in opposite directions from the actual result (a virtual tie).

Nate Silver explains that Iowa is hard to poll. But since polls are the only horserace information that the media has, there’s endless dissection of who is surging or receding. In fact, we were only looking at a pale shadow of the truth. The polls were bullshit.

Expectations are irrelevant, let’s analyze reality

Before the election, there were many articles that projected the future based on exceeding or falling short of expectations. Expectations come from media, which come from polls. So that’s a bullshit narrative. My news search today revealed hardly anyone talking about expectations — what happened to all that supposedly significant analysis?

Compare the expectations narrative with the reality-based narrative based on what happened yesterday.

Expectations: Donald Trump underperformed in Iowa, which hurts him.
Reality: Donald Trump lost Iowa. Even if he wins New Hampshire, his issue is that so many people hate him. This caps his chances. Expectations are irrelevant.

Expectations: Ted Cruz did better than expected, which boosts his candidacy.
Reality: Ted Cruz did great with religious conservatives in Iowa. He’s about to get a lot scrutiny on his views, for example, that God comes before country. That won’t play in New Hampshire. With Trump and Rubio splitting Republican votes, there might be enough religious conservatives to get Cruz the nomination, but they won’t win him the general election.

Expectations: Marco Rubio did better than expected, which boosts his candidacy.
Reality: Marco Rubio consolidated the votes of people scared of Cruz and Trump. If he can keep doing that, he’s got a path to the nomination.

Expectations: The rest of the Republican field did poorly, which will hurt them.
Reality: The rest of the Republican field spent an awful lot of money for minuscule vote totals. It looks doubtful any of them will change that any time soon.

Expectations: Bernie Sanders did better than expected, earning a tie in the first state to vote.
Reality: Bernie Sanders has to win in southern states to have a shot.

Expectations: Hillary Clinton didn’t decisively win Iowa. She was a loser there in 2008. This is her Waterloo.
Reality: Hillary Clinton now faces an actual challenge, but is still the most likely to win the nomination because of her appeal across racial lines.

Reality-based predictions

Here are predictions based, not on polls, but on what actually happened.

The other Republican pretenders will drop out. Bush spent $2800 per vote in Iowa and came in sixth with 3%. Ben Carson is tanking after putting the electorate to sleep. Rand Paul’s issues aren’t what people are worried about right now. Kasich has little appeal with Southerners and must fight with Trump for Northeastern Republicans. It’s now about Trump, Cruz, and Rubio. The nomination depends on whether Rubio can roll up all the support from Republicans horrified by Trump and Cruz.

On the Democratic side, Bernie Sanders has proven he can excite white people in Iowa. And he will win New Hampshire. The race depends on whether he can excite non-white people in the rest of the country. Fivethirtyeight.com now projects a 96% chance that Clinton will win South Carolina. That’s not a good sign for Bernie.

Final note for all the polemicists who may comment: my analysis is based on reality, not who I like. If you disagree, cite evidence, not passion. Vote based on passion; analyze based on facts.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.


  1. I would term your observations lame. The purpose of polls are to give a snapshot of people’s thinking at the moment. This has obviously been a very fluid election campaign. I was doing surveying last night at a solidly conservative precinct. It was amazing the number of people who were making up their mind the day of the caucus or even after they arrived at the caucus site. Polls are an easy target for overly simple analysis, but I would term those analysts as not having a grasp of reality. To suggest otherwise is diogenes waste.

    1. Great to hear from an actual caucus-attendee.

      We seem to agree that polls are bullshit and observations based on analyzing poll results are lame.

      We’ll see soon whether my analysis is lame or note. That’s what I aim for — verifiable predictions, not “on-the-one-hand, on-the-other-hand” blather.

      1. Josh, have you read Superforecasting? Making verifiable predictions is one of the first recommendations. The book is based on the author’s research on effective forecasting for IARPA–it’s a good read.

  2. Even with last night’s results revealing the vacuity of the polls, I think one of the fundamental problems of the media coverage we’ve seen so far has been the general lack of perspective. Far too many stories have implied that early victories in February will create an unstoppable juggernaut candidate who will coast to the nomination. That’s akin to assuming a team that scores three runs in the top of the first inning is automatically going to win a baseball game. There are still eight innings left to play and any number of things can happen. Momentum can swing, errors can happen and cold hitters can suddenly get hot.

    Let’s focus on the Republican side, for example. In the January 26 Wall Street Journal, Reid Epstein lays out a scenario in which no Republican candidate will have an insurmountable lead heading into the July convention. (http://www.wsj.com/articles/contested-republican-convention-is-possible-1453828686)

    There are 1,237 delegates needed to win the nomination. Epstein points out that the four February contests in Iowa, New Hampshire, South Carolina and Nevada total 133 delegates. He further points out that about 600 delegates are at stake on March 1 in Texas, Virginia and nine other states. All of these early states award their delegates proportionally based on the percentage of votes the candidates receive. Epstein notes we won’t hit the first winner-take-all states until March 15 in Florida (99 delegates) and Ohio (66).

    “Even if one candidate wins 40% of the delegates awarded by March 15, he or she would have less than one-half of the necessary delegates to win the nomination,” Epstein writes. Doing the math, it is not outrageous to think that if a candidate has the money and organizational strength to withstand a few early third, fourth or even fifth place finishes, he could still have a chance at the nomination.

    Clearly, the American voting public is unpredictable. Early momentum and continued media coverage of vacuous polls could easily sway many voters. But looking at the math, the numbers show that in this election year, just like in a baseball game, no lead is automatically insurmountable.

  3. My observation from the data you highlighted is that while the two polls differed by about 6% on the total percentage of the vote the top three would receive (72% vs 66%), distribution among the three candidates is nearly the same. Trump 43% vs 42%, Cruz 33% vs 35%, Rubio 24% vs 23%. Both missed the mark on the total of 75% though Quinnipiac came close at 72%. But both were off on the distribution among the three with Trump at 32%, Cruz at 37% and Rubio at 31%. It suggests to me that, first, quite a few Trump supporters did not make it to the caucuses. That had been predicted by a number of people. And second, that a number of the supporters of “other” candidates shifted to vote for either Cruz or Rubio. Shifts in actual voting from non-competitive candidates is common, and it makes some sense as it is easy to imagine someone supporting one of the others to see Cruz or Rubio as their second choice rather than Trump.

  4. as the man said ………rigged! , the freakin polls were bullshit . come on , everything pointed to clinton as the winner , I even thought that ,but the people have spoken …….so obay!!!!!!!!!!1

    1. as the man said ………rigged! , the freakin polls were bullshit . come on , everything pointed to clinton as the winner , I even thought that ,but the people have spoken …….so obey!!!!!!!!!!1