Monthly Archives: February 2016

How could pollsters get the Iowa vote so wrong? They didn’t

By Kent R. Kroeger (Source:, February 6, 2016)

We would love to brag that we were the only data-driven political forecasters to predict Cruz’ victory in the Iowa’s Republican caucuses.  We won’t.  Not that we don’t want to, but the truth is a bit more complicated than “we got it right and they didn’t.”

No, it would be wrong for us to highlight the logical flaws in J. Ann Selzer’s Des Moines Register’s Iowa Poll, who was deemed our nation’s most accurate pollster by Nate Silver, whose own forecast was similarly off the mark in Iowa’s Republican caucuses.  No, we will not boast because we would have to ignore the facts…and the facts actually support the pollsters and data modelers who predicted Trump would win the Iowa Republican caucuses.

What?  But, Cruz beat Trump 28 percent to 24 percent while the polls, summarized by RCPs rolling average, said Trump would beat Cruz 30 percent to 24 percent.  Selzer’s Iowa Poll had Trump winning 28 percent to 23 percent over Cruz, as of the weekend before the Monday caucuses.  The surface evidence indicates they missed.  But I would argue the pollsters were right given what they asked respondents and it was the caucus results that perverted the true sentiments among Iowa caucus-attending Republicans.

You may recall a similar argument about the Florida election results in the 2000 presidential election in which the final tallied results indicated George W. Bush beat Al Gore even though the exit polls indicated a majority of voters preferred Gore to Bush.  The exit polls were most likely correct; it was the vote tally that was wrong.

The argument that the caucus process perverts the true sentiments of Iowa caucus voters starts with my own experience at an Iowa Democratic caucus.  My wife and I moved to Des Moines, Iowa four years ago and she’s always been curious about how an Iowa caucus works.  While Iowans find the process a democracy-affirming experience, outsiders often find the process confusing and even undemocratic.  Further complicating things, the Republicans and Democrats have different caucus processes with the Republicans using a secret ballot and the Democrats using a multi-stage, non-secret balloting system.  The end result however makes the two processes more the same than different.  Though the Republicans use a secret ballot to cast caucus votes, the first step in the process has supporters for each candidate grouping together so they can select someone to speak on behalf of their preferred candidate before everyone casts their “secret” ballot.  The actual result is that each caucus voter discovers very quickly who their neighbors are voting for and that is start of the problem.

Back to my experience in the Democratic caucus.  I am kind when I say the process started out disorganized and devolved at each step.  The precinct chair’s first problem was the unusually large number of caucus-goers in attendance.  Over four hundred.  At least that is what the signed register indicated.  But even that official count of qualified voters was under suspicion when the first head-count of candidate preferences was taken after everyone moved to a specific section of the high school auditorium depending on their candidate preference.  Soon after the head-count, the precinct chair announced that the total number of first ballot voters exceeded the registered total by almost 30 voters.  Voters were counted that had not officially registered.  Did they forget to register?  Did they sneak in?  Sneaking in would not have been hard to do.  We were in a 600-seat auditorium with over eight separate doors to enter and exit and no security to speak of.  Yes, there were people standing at some of the doors, but they didn’t seem stop anybody from entering since there was no way to know if their eligibility had been validated and they had signed the register.  It was an honor system.

The register and vote total discrepancy was not a problem, said the precinct chair.  He asked that anyone who had failed to sign the register should do so before the next head-count vote.  We waited about 20 minutes while that process was completed and counted heads a second time.  We were off by only 20 or so people the second time.  Um.  A less trusting precinct chair would have read off the names from the register and kicked out anyone whose name wasn’t called.  But not this precinct chair.  Not on this night.  He declared the vote total, the one that exceeded the official register by over twenty voters, to be the final vote total and the Clinton and Sanders delegates were divided up proportionately, six for Sanders and four for Clinton.

Did I mention I, along with 27 other people, caucused for Martin O’Malley?  No matter.  We didn’t have enough O’Malley supporters to earn a delegate.  We weren’t viable.  Technically, we should have been allowed to select another “viable” candidate, but this precinct chair wasn’t about to introduce more complications into an already deteriorating situation.  We went with the 6-to-4 delegate result and walked out of the auditorium.  My wife had experienced her first Iowa caucus and as she left the auditorium wondering out-loud why Iowa should be given the honor of starting our country’s presidential selection process.  I told my wife that Stanford-educated elitists like herself are not in any position to judge our time-tested caucus system.  We also agreed not to talk about what we saw that night ever again.  We willingly participated so we felt somewhat responsible.  We felt unclean.

How does my experience relate to Cruz beating Trump?  Perhaps it doesn’t.  Maybe I just wanted to get it off my chest.  What I do know is that the Iowa caucus system is a disaster and more people, not just in Iowa, need to voice this opinion each time Iowans go through the process.  As to why Cruz beat Trump, we need to remember where Cruz drew his support — people who self-identified as “very strong conservatives.”  And these voters tend to be very religious and tend to live in rural areas.  Iowa’s rural communities are tight.  Every family knows every other family.  You know how much your neighbor owes on their house; you know how many cars they’ve owned; you know where they go to church on Sunday; and you know if they don’t.  Small Iowa communities exert a palpable pressure on its citizens not to deviate from local norms.  Of course, some do anyway.  They’re called Democrats.

The rest are Republicans and this leads us back to our original thesis:  The Iowa caucus process perverts the true sentiments among Iowa caucus-goers.  And so we get to caucus night, February 1, 2016.  Ted Cruz went from 23 percent in the Iowa Poll to 28 percent in the actual caucus results.  One theory is that conservative evangelicals in Iowa are far more organized in their get-out-the-vote efforts than other constituencies.  This is true.  But the Iowa Poll measures vote turnout intentions and its predicted evangelical turnout did not vary significantly from the actual turnout (for those details, go here).  No, the answer to the question “why did the polls get the result wrong?” lies not with differentials in socio-demographic turnouts and candidate ground-games.  The answer lies in the process itself.

People change their intended vote preferences when they see their neighbors (and their religious leaders) standing for a specific candidate.  This year it was Ted Cruz. Four years ago it was Rick Santorum.  In 2008, it was Rev. Mike Huckabee.  How do I know this?  Because I’ve seen it.  Not as much in the 2016 Democratic caucus I attended, but I saw it in 2012 when I caucused for Ron Paul and watched the Santorum contingent grow minute-by-minute as caucus voters walked in and saw neighbors and fellow parishioners standing together.  The arm-twisting was congenial but unmistakable.  “Are you supporting Santorum?”  In various forms, I heard the same question over and over.  It was not nefarious.  It is the caucus process.  It is not that much different from the peer-pressure dynamics in high school.  In fact, it is exactly the same.  Only in high school, it determines what clothes you wear, not who should be the next President of the United States.

The polls were correct leading into the February 1st caucuses. It was the Iowa caucus process that got things wrong.  The Iowa caucus system needs to end.