There are two major showdowns coming to North Carolina this month: the Duke-UNC game on March 5, and the GOP Primary on March 15.
Propelled by his Super Tuesday victories in three states, Ted Cruz has decided to press forward in his campaign for the Republican nomination for the presidency, presumably to include the North Carolina Primary as a barbecue-laden pit stop on the path toward a Brokered Convention Hellscape.
Much has been made of Ted Cruz’s face this primary season: its undeniable backpfeifengesicht, its counter intuitive emote, his inability to smile naturally. The most flattering of these notes concerns his resemblance to Duke University basketball player Grayson Allen (with visual aids helpfully compiled by the esteemed Washington Post). Grayson’s political allegiances are unknown, and my sources indicate that he is by all accounts lovely and undeserving of this regrettable craniofacial happenstance.
But Duke’s guard-in-chief has provoked similarly inflamed rhetoric, albeit for offenses less grandiose than those of the Texas senator. Two of my favorite descriptors are “annoyingly competent” and “seemingly designed to make you angry,” followed promptly by “very, very, good, of course.”
Now: North Carolinians have plenty of meaningful matters to be concerned about, including racial gerrymandering and a confusing voter ID law. But in advance of the Duke-UNC game, I’d like to explore a harebrained hypothesis: that Cruz’s resemblance to Grayson Allen will deter some hardened Carolina fans, ultimately costing him some votes – and a delegate. And in North Carolina, proportional representation means that Ted Cruz’s electoral fate could hang in the balance.
(I want to emphasize, briefly, subtle differences between the two: Grayson is clearly capable of smiling, isn’t creepy-looking, & my intent is absolutely not to make fun of a 20 year old.)
Here’s where we’re going:
- An explanation of North Carolina’s primary
- An estimate of Cruz’s non-Allenized vote share in North Carolina, based on county-level demographic trends
- Computation and mapping of county-level Duke/UNC loyalty shares
- Incorporating some actual literature related to facial characteristics, subconscious biases, & political outcomes
- Projection of county-level vote shares and impact on Cruz’s delegates
North Carolina: A Semi-Closed Primary
North Carolina has 72 delegates, and allocates those delegates proportionally based upon vote totals. This means that each candidate gets a delegate for roughly every 1.4% of the vote they receive. The NC primary is semi-closed, meaning that party cross-over will be minimal – important in a purple state! – though unaffiliated voters may participate in either party (see limitations).
North Carolina, as many of you are probably aware, is also down with the College Sports. This expresses itself primarily through a fierce basketball rivalry between Duke and the University of North Carolina. Though the schools are separated by a scant 20 minutes (in traffic, tops), divided loyalties persist across the state.
How *Should* Ted Cruz Do in North Carolina?
We don’t really know. The most recent polls are from February 17 – before multiple candidates dropped out, and before this campaign cycle descended to never-before-seen ridiculous depths. Here’s what Public Policy Polling had then:
In lieu of up-to-date polling, I’ve decided to make my own estimate of Cruz’s vote distribution for the state based upon a demographic calculator thrown together by David Byler at Real Clear Politics. The calculator uses results and demographic characteristics of states that have already voted to draw out correlations between vote shares and county-level demographics. For Cruz, the biggest factors are
- % married households without children
- % married households with children
- % without a high school diploma
To translate the correlations into projected ’16 primary voteshare, I pulled county-level projections of the relevant factors for 2014 – the most recent year available, and multiplied them out with county population, Romney’s vote share in 2012 as an estimate for Republican electorate per county, and 2012 Republican primary turnout. I then used population-weighted averages to obtains statewide estimates. For good measure, I also checked these numbers against Feb ’16 voter registrations.
The most recent polls all had Ted Cruz hovering around 19%. Of Carson voters, 63% had favorable opinions of Cruz (versus 28% of Jeb voters), and 34% of Carson voters selected Cruz as their second choice. In lieu of recalculating Carson’s demographic share, I’ve added a quick 3% on to Cruz’s vote totals, giving him 1/3 of Carson’s potential votes
Displayed by county, here’s how it all works out:
Overall, my model is bullish on Cruz. Without the Carson bump, I had him at 20.9% statewide – slightly above PPP & folks. With Carson out of the race, this method projects he’ll take home 23.9% of the vote on the March 15 primary.
Duke vs. UNC: A State Divided
Okay, so: once we’ve got county-level Cruz estimates, how do we adjust them for a potential Duke-UNC loyalty split? Luckily for me, Public Policy Polling – which is based in Raleigh – has the foresight to ask about this sort of thing. According to their president: “A lot has changed in North Carolina during the decade we’ve been running a monthly poll on the attitudes of voters in the state, but one thing’s always a constant- pulling for the Tar Heels to beat the Blue Devils.” Speaking as a Duke grad: Fine. I like being the underdog, and I know this breed of resentment is ill-defined envy
College basketball team loyalties statewide are as follows, in descending order: UNC 27%, NC State 21%, Duke 16%, East Carolina 9%, Wake Forest 6%, and no preference 20%. Pared down to the two titans in a cross-tabulation, the UNC Tar Heels lead 40/30 with both Democrats and Republicans, and 34/25 with Independents (typical – what, did you guys go to Elon?).
To adjust potential Cruz voteshares based upon the Weird Grayson Allen Thing, I decided to go with geographic variation, for a few reasons:
- While Duke & UNC grads are concentrated in urban, relatively liberal cities, loyalties tend to vary throughout
- PPP also tabulated the results by extent of conservative ideology, and there was more regional variation in Duke/UNC support than between “somewhat conservative” and “very conservative”, where Cruz voters are concentrated (both 40% UNC
Unfortunately, geographic data was only available on the area code level, per PPP tabulation methodology and probably caller privacy. C’est la vie:
The Grayson Effect
(This is the part where I start making stuff up.)
Humans make snap judgments about others based upon facial characteristics all the time – including in the voting booth. Assessments of competence, based upon headshots, have been correlated linearly with vote share in actual U.S. Congressional and gubernatorial elections. A study out of Columbia built on this work, creating a model to predict individual-level vote choices based upon perceived facial competence and partisanship. Trustworthiness and dominance of political candidates have also been associated with these sorts of intuitive assessments. And per the “mistaken identity effect”, people also tend to overgeneralize similarities between individuals with similar facial characteristics, projecting the traits of those they are familiar with onto strangers – more so for men, and more so for people with more extreme traits and physical characteristics, both of which are relevant to Allen & Cruz.
Given all this, is it too much of a stretch to think that voters in North Carolina might project their assessments of Grayson Allen’s competence and trustworthiness onto his older, slimier twin?
The question is: how much? Could this actually impact vote totals? If Duke defeats UNC on March 5th, will the angst reverberating throughout the majority of the state be enough to allow other GOP candidates to snatch some delegates from Cruz?
I’m going to try to find out. Here’s the thinking:
- Duke fans are more likely to be subtly influenced in favor of Cruz due; NC fans are more likely to be subtly repulsed by Cruz.
- Duke and UNC fans both likely have entrenched ideas about Grayson Allen’s competence.
- The Columbia paper used incumbency, perceived facial competence, and partisanship to predict individual voter preferences. Because most voters in the primary will presumably share a party with Cruz, I’m going to replace political partisanship with Duke-UNC partisanship.
- With a statistically insignificant difference between political allegiance and team loyalty, let’s assume that Cruz-leaning Republicans will hold the team preferences characteristic of their respective geographic areas.
- The Columbia study found that a more competent challenger face increased probability of a challenger vote of 3.5% for independents, 2.5% for challenger-party voters, and 1.5% for incumbent party voters Studies are lacking about the extent of assumed transference due to resemblance are lacking, I’m going to assume for the purposes of this ridiculous exercise that those deviations hold, such that:
- Republican Duke fans are 1.5% more likely to vote for Cruz
- Republican UNC fans are 2.5% more likely to vote for some other candidate
Where this gets really interesting, however, is with the voters unaffiliated with a team. There are ten days between the Duke-UNC game and the primary, which should provide voters with ample time to reflect on the competence of Duke’s star guard, and to subconsciously associate this with his countenance.
- Half don’t pay attention.
- Those who do associate victory with competence, and are therefore 3.5% more likely to vote for Cruz in case of a Duke win, and 3.5% less likely to vote for Cruz in case of a UNC win.
Projecting onto the map, we’re left with this:
Poor Ted is out of luck to begin with: Tar Heels dominate the state, and in the few counties they don’t, subtle push factors are granted more percentage-strength than pull. But when associated with a victor, the unaffiliated voters in counties with appropriate demographic distributions might push some votes his way…
Results: The Delegate Count
The moment of truth has arrived. North Carolina’s 72 delegates mean there’s one delegate for every 1.38% of the vote.
I projected three outcomes against my based estimates for Cruz’s vote share:
A UNC win would risk detracting a delegate from Cruz, directly harming GOP chances of reaching a brokered convention. Put another way: a win for UNC is tantamount to an en-masse Trump endorsement, and the great state of North Carolina deserves better.
The data I compiled is available for download here. To loosely protect against the grubby hands of my least favorite Senator’s campaign staffers, the password is “repealHB2”. (I know he can’t do that, I just want to make his people type it.)
Numerical profligacy aside, isn’t that vote prediction paper wild? Back in 2004 I remember my mother saying she agreed with John Edwards on a lot of policy positions, but that his instant & overwhelming cad vibe was a dealbreaker. I was 13 and my creep-sensing abilities were not yet finely honed, but I do think this sort of thing registers instinctively and resonates deeply, and that exploring ways to blunt it might prove useful for elevating policy disagreements to the forefront of party debate (ha).
Thanks for reading! GTHC.
- While probably appropriate for a purple state, the semi-closed primary structure poses a slight problem for this projection: of the races so far, only Oklahoma and Massachusetts have hosted semi-open competitions, and Oklahoma only did so for Democrats. Given the general kookiness of this year’s turnout, I haven’t found a good way to deal with unaffiliated voters, so I ignored the voter category identities and allowed demographics to prevail when determining percentages. Consider this a major limitation.
- Republican turnout has been consistently higher than usual this election cycle. My lowball ’12 turnout assumptions make raw vote estimates extremely dubious, though I do think percentages will hold.
- This model assumes Jeb? votes were distributed equally.
- Postgraduate degrees were a huge selection factor against Cruz, but Durham & Orange Counties are so overwhelming liberal that I left this out.
- This model may neglect the impact of loyalists to NC State, Wake Forest, & East Carolina University. Their opinions may change year-to-year depending upon performance relative to both Duke and UNC.
- I’m a democrat from Texas and therefore derive a lot of schadenfreude from badgering Cruz, so this entire analysis is motivated by sour malice & sponsored by Charles Shaw. Assume transcription errors.
Atkinson, M., Enos, R., and Hill, S. Candidate Faces and Election Outcomes. Statistical Modeling, Causal Inference, and Social Science May 2007.
Ballew, C. and Todorov, A. Predicting Political Elections from Rapid and Unreflective Face Judgments. Proceedings of the National Academy of Sciences of the United States of America. November 2007.
Byler, David and Trende, Sean. What’s Really Going on With the Republican Race? Real Clear Politics. March 1, 2016.
Politico. Election Results, North Carolina. November 19, 2012.
Public Policy Polling. Trump, Clinton Continue to Lead By Double Digits in NC. February 17, 2016.
Rufianni, Paul. Demographics of the Iowa Caucus Results. Feburary 2, 2016.
Team Fix. Duke’s Grayson Allen Looks A Lot Like A Young Ted Cruz. The Washington Post. April 7, 2015.
Zebrowitz, L. and Collins, M. Accurate Social Perception at Zero Acquaintence: The Affordances of a Gibsonin Approach. Pers Soc Psychol Rev. August 1997.