Dissecting the Iowa Caucus Results
On Friday and Saturday of this past week, we polled Iowa ahead of the GOP caucus on Monday; (we ran out of time to poll the Democratic caucus). You can find the full results of the Iowa caucus survey here. Needless to say, most of those who polled the caucuses saw relatively poor results, compared to the event’s outcome. That said, along with Emerson College, we managed to detect Trump’s fall and Rubio’s rise, calling the election an effective dead heat. That turned out to be largely true, though Cruz managed to top all expectations with a 3.3% lead over Donald Trump.
In the interest of posterity, let’s take a look at some of the numbers from the GOP caucus, wherein there were plenty of surprises. Below is a table of the candidates with columns representing a) our poll percentages for the horse race; b) the actual percentages of the election (from the AP); c) the spread between each candidate and the candidate immediately preceding, using data from our poll, in the order of the actual election result; and d) the same spread for the actual election numbers.
|OS Poll (%)||Actual Result||OS Poll (spread)||Actual Result (spread)|
Surprise #1: Rand Paul
While the numbers were trickling in on Friday, I saw a massive bump for Rand in the 18-29 (mostly male) demographic. This seemed strange until I considered two facts: 1) Rand managed to attract insanely vocal supporters to events and debates and 2) Iowa has been good for the elder Paul in the past. So, while it seemed unlikely that he would tie with Carson, it wasn’t totally outside the realm of possibilities. At first, I thought we may have sampled too many men in the 18-29 demographic who, along with their counterparts in the 45-64 demographic, I like to refer to collectively as the “Army of Trolls.”*
These are guys who are sometimes well-informed, but often overstate both their support for one candidate, as well as their likelihood to vote. Now, Rand did come in fifth, but by a slimmer margin than we expected (to some degree). This was also true across the board for candidates outside of the top 4, so it seems more likely that folks switched their vote, rather than an over-sampling of 18-29 year-old men. I would normally aim for far fewer respondents within the troll demographic, but given Iowa’s bizarre dynamics, I left it alone.
*I am a part of the 18-29 male demographic, so please – don’t troll me for this.
Surprise #2: Cruz
On Friday, the initial results were clear: Cuz was way, way ahead. Again, I was very skeptical of this; Trump had a significant lead in all of the other polls without any major news breaking (apart from skipping the debate). Rubio was also ahead of Trump in the raw numbers, but weighting would have put Trump ahead by a few points.
On Saturday night, Cruz’s numbers began to erode, while Rubio’s skyrocketed and Trump’s held steady. At this point, it could have been any of the three; it all came down to weighting. This brings us to the next surprise…
This shouldn’t have been mind blowing, but self-described evangelicals turned out in droves. Based on entrance polling from the New York Times, white evangelical voters made up 62% of caucus-goers. That is an astronomical figure – one which no one seemed to expect. Originally, our model accounted for 56% evangelicals, which would have put Cruz 2 points ahead of Trump. We decided to change this to 47.5%, based on many different factors that we hedged on. This turned out to be the wrong call, but in reality, the results would have shown an identically close match. The chart below shows the breakout of evangelicals by first vote choice, which shows that while Cruz snatched the largest percentage, Trump and Rubio weren’t far behind. The biggest surprise, in my opinion, is the relatively poor showing for Carson within this demographic.
Not a Surprise: Rubio
As mentioned, we saw this coming Saturday night. Rubio’s surge might have been the result of Cruz’s voter scorecard debacle, but as results poured in, it was an astounding transformation. Looking at the NYT entrance poll, it’s no wonder why Rubio was underestimated by other polls: of the voters who made up their minds either within the last few days, or on the day of the caucus itself (35%), Rubio won 31% and 28%, respectively. No mystery there – late polls will undoubtedly have an advantage (unless you’re Quinnipiac).
Also Not a Surprise: Santorum
He came dead last (not counting Gilmore), to absolutely nobody’s surprise. Now, apparently, he’s calling it quits.
Good Stuff: Screening
We decided to use several checkpoints for screening potential caucus-goers. First, we used a voter file, which did not in itself exclude first-time caucusers. Instead, this selected folks who had at least voted previously, either in the 2014 primary and general, or in the 2012 primary and general. Next, we asked respondents whether they intended to caucus, and excluded all but “definitely” – fairly standard stuff. Finally – and this was the crafty bit – we asked respondents the following:
Where will your caucus be held?
a. At a school
b. At a Church
c. At a community center
e. I don’t know
We eliminated all “I don’t know” responses. The hypothesis here is that if you don’t know where your caucus will be held two nights before it starts, then you’re less likely to attend. While I can’t confirm the hypothesis outright, I do believe that this helped to screen some of the ‘wishful’ voters, for lack of a better term. It seemed to pay off.
I was happy with our results in Iowa, but there’s no doubt that this could have been a much better round of polling. I had never polled Iowa before this election; (this was our second Iowa poll, the first conducted nearly a year ago, when Walker was still a thing). I have to say – it’s a very strange place, with an even stranger method of choosing a presidential candidate. The key here seems to be polling as late as possible, but in reality, the key is not polling Iowa at all. Not only is Iowa not representative of the rest of the country, but the caucuses are designed to disrupt voters’ beliefs when they enter the room. I’ll go one further too: Iowa should not be the first state to hold its presidential nomination process. But I digress…
No polling from us for New Hampshire, but we will certainly see you closer to home in South Carolina.