For about a month, reporters and pundits have been heralding surges of support for outsider candidates like Senator Bernie Sanders, Senator Ted Cruz, and everyone’s favorite bombastic billionaire, Donald Trump. Based on the flood of news stories predicting insurgent victories, readers might believe that Trump and Cruz are the only Republicans left and that Sanders is about to deliver the knockout punch to Clinton’s glass jaw in Iowa and then New Hampshire.
But the truth is considerably murkier. With only a week to go before the Iowa caucuses, the political landscape remains as tumultuous and unsettled as ever, and these surge story lines may ultimately hurt the outsider candidates in the long road to the nomination for a number of reasons.
This morning, Andrew Prokop from Vox gave an excellent breakdown of the Iowa caucuses and why they matter. The piece explores the history of the caucuses, the difficulty of the process in caucusing, and the outsized impact that Iowa has in the nomination process.
In addition to providing a fantastic overview, Prokop touches on one of the most salient issues about the early voting states: Winning them matters less than meeting or exceeding expectations.
As Prokop notes, Santorum’s and Huckabee’s surges in Iowa weren’t important because they won, but because both candidates defied expectations and changed the story. By cruising to surprising victories, both candidates were able to capture more news coverage, convince donors they could win states, and challenge the established narrative.
The opposite is true for failing to meet expectations. Underperforming in Iowa (and later New Hampshire) can shake donors’ faith, lead to negative—or worse, less—news coverage, and ultimately result in a drop in voter support in later states.
Right now, there are no expectations for a clear winner in Iowa on the Republican side. Most pundits predict that Trump or Cruz will win, but that the winning candidate will win narrowly. Regardless of who wins, if this is the case, neither campaign is really hurt, and predictions of a Trump or Cruz nomination remain on track. But if either candidate underperforms, then they’ll face significant challenges.
The same is true for Bernie Sanders. If Sanders takes a close second or even wins the state, he’s upset expectations and probably created a much stronger road to the nomination. However, due to news stories citing a late surge for Sanders, a distant second—as was predicted as little as six weeks ago—would be seen as a huge setback to the senator.
The irony is that even a resounding second place finish by Cruz three months ago would have been seen as a boon for his campaign. The same is true for Sanders, who was once down twenty points. But polling numbers changed the narrative.
Over the past six months, Iowans phones have been ringing constantly as polling firms desperately try to get a snapshot of what the electorate is thinking. These polls, especially over the past month, have driven the Cruz/Trump victory narrative, as well as the Sanders upset story.
The problem is that polling in Iowa is notoriously unreliable for a number of reasons, and things may be even less accurate on the Republican side of the field.
I won’t get into the nitty gritty of P values and confidence indices (especially since I have only a baseline understanding of them myself), but a significant number of the surveys are working off very small sample sizes (sometimes as low as 266 voters), and posting margin of errors as high as 8.9 points. That means that, even if we accept the surveys, the spread between two candidates could be off by as much as 18% (although most MoE are in the 6 point range, suggesting a max gap of 12 points).
Normally repetitive polling helps negate this effect: If different polls are popping out the same number week after week, there’s a good chance they’re reliable. However, this hasn’t been the case, especially in the Republican field. Over similar polling periods, there have been results that show Cruz with a 2-point lead over Trump and polls with Trump leading Cruz by 11 points. These are notable swings, and an 11 point win by Trump would be a huge deviation from the Cruz/Trump horse race that so many reporters have trumpeted.
But the situation gets even murkier, because different pollsters have different methodologies. Many of the polls that are used in news stories do a mix of web and phone surveys. Web surveys require participants to opt in instead of being contacted randomly, meaning that pollsters are no longer working with completely random samples. Thus, despite the fact that many pollsters are publishing margins of error, it’s hard to see how they came to those conclusions, as the process requires a random sample.
On top of that, the threshold for including respondents is different for each survey. Some surveys have much looser definition of a likely voter, meaning they may be capturing the opinion of individuals who are leaning toward caucusing, but haven’t decided to actually caucus.
These problems plague both the Republican and Democratic polls in Iowa, but Republicans face another unique issue, which is the Trump effect.
There is no consensus on what exactly the Trump effect is, but there are two basic interpretations: Trump’s numbers are artificially low in polls because respondents are embarrassed to say their voting for him or Trump’s numbers are artificially high because of a variety of reasons (increased fervor, including name recognition, Democratic trolling, and general disdain for pollsters or the establishment).
Those who say Trump’s numbers are artificially low cite the fact that, in online and hybrid polls, Trump does considerably better, suggesting that Trump respondents are hiding their support due to a social desirability bias. The argument is that Trump supporters don’t want to admit they support Trump, and thus they are answering less honestly when they answer live pollsters than online survey. Others have pointed out that the gap could also be due to the zeal of Trump supporters, who may be more willing—eager even—to jump into an online poll and show their support for their candidate.
Regardless, there’s data to suggest that Trump’s numbers may not be entirely accurate. Even if they are, there are significant doubts about whether or not those poll numbers will turn into votes. The wider parameter for likely voters is one such reason, but another (connected) issue is that it’s hard to get people to turn out to caucus. All candidates face this problem, but particularly those who rely on “somewhat” likely voters.
The Difficulty of Caucusing
Caucusing is a unique political environment, one that spurns all but the most die-hard supporters. This is because caucuses are not like primaries: They are not standard primary elections where individuals vote over the course of a day on a secret ballot, tally the votes, and distribute delegate votes accordingly.
Caucuses are held at specific times, at specific places, and, in the case of the Iowa Democratic caucuses, are not a secret ballot (more on this in a bit). All of this limits the number of participants in either caucus, usually only to diehard voters who repeatedly vote in the caucuses and other party elections. This means that campaigns that rely on nontraditional voters but that lack a phenomenal ground game may lack the ability to get voters to the caucus locations on time.
For Ted Cruz, who has built an incredibly strong field team in Iowa, this may be a good thing. Trump is at an organizational disadvantage in the number of offices and volunteers he has in Iowa, it’s possible that Trump isn’t able to turn out the voters who support him to beat Cruz.
Hillary Clinton also enjoys an organizational advantage, but has the added benefit that the Democratic caucuses are open ballot. Individuals are encouraged to discuss the candidates, debate the choices, and ultimately choose delegates. Sanders’ supporters—according to the nebulous polls—are more likely to be open to changing their mind.
All of this means that the Iowa caucuses are a lot less clear than the media is letting on, and that the narrative that evolves from any electoral situation may not be what pundits think it is now. Trump, Cruz, and Sanders are riding high press stories, stories that may well garner them additional support in Iowa and get them the wins or narrow losses they need to continue to dominate the narrative.
However, those same stories have set up perverse expectations for outsider candidates in a state that holds a nomination process tailored to diehard partisans and well-organized, traditional campaigns. Furthermore, those expectations are based on a flawed polling process that may not accurately report the voting public’s perception.
So what happens with the Iowa caucuses? Given the complete uncertainty in the polling data combined with the traditional difficulty of accurately gauging the state anyway, I think it’s fair to say no one knows. Trump or Cruz may post a dominant showing, or both may stagger, opening the path to Rubio, Kasich, or Cruz. Sanders may beat or narrowly lose to Clinton, or may get a beat down, sending him limping into New Hampshire a week later.
One thing is for sure though: Expectations have been raised in Iowa for all the outsider candidates. Poor showings may not be the death knell—especially for Trump and Sanders, who may do well in New Hampshire—but it may be the beginning of a new media storm that, yet again, leaves the outsiders with the noses pressed against the glass.
 In a nutshell: Because everyone says they matter. Politicians give it attention, so media gives it attention, which makes politicians give it attention, creating a feedback loop until Iowa and New Hampshire dominate the news cycle for six months, despite their relatively low delegate counts.
 I’m not going to even get into the complications that come from having 12 different candidates in the field.
 Social desirability bias suggests that people are less likely to answer honestly to live pollsters if they think the pollster may judge them. This has been demonstrated in issues affecting race and sexual orientation.