Football was never intended to be a beauty pageant, and this study by West Virginia University researchers backs it up.
In fact, if you’re a handsome heartthrob coaching an NCAA Division I team, you might get a penalty flag when it comes to pay.
Using a face recognition and machine learning approach, a multidisciplinary team of WVU researchers found that head coaches with a more aggressive appearance earned a salary premium. More attractive football coaches, on the other hand, faced a salary discount.
And for WVU fans, just because you’re probably already wondering, Neal Brown ranked among the top quarter of attractive coaches.
Former WVU football coach Dana Holgorsen was ranked “substantially less attractive.”
“One explanation for the attractiveness discount and aggressiveness premium may stem from that fact that American football is a very aggressive sport, and an unattractive face might signal mental and physical toughness, viewed as a desirable characteristic in this market,” said Brad Humphreys, economics professor in the John Chambers College of Business and Economics.
In recent years, economics research has shown discrimination against unattractive workers. That outcome, called the “beauty premium,” refers to the idea that more attractive people earn a premium in labor markets, specifically in fields involving interpersonal interaction.
Humphreys wanted to see if this theory applied to college football coaches, who are, in many cases, the face of their university’s sports program.
Traditionally, in studies like this, volunteers are recruited to examine photos and rate each person’s physical attractiveness on a scale, such as from one to five, with five being the most attractive rating.
But as the old adage goes, “beauty is in the eye of the beholder,” and Humphreys wondered if there was a more objective way to conduct the research. He reached out to Guodong Guo, an associate professor at the Statler College of Engineering and Mineral Resources, who has extensive expertise in facial recognition and machine learning technologies. Guo has developed algorithms that can predict a person’s body mass index based on a photo.
“The key idea is to analyze the facial attributes by using computational techniques, so that the analysis can be performed in a large scale, avoiding the biases caused by human raters, as well,” Guo said.
The research team, including Humphreys, Guo and graduate students Yang Zhou and Mohammad Iqbal Nouyed, used data from three sources:
- Salary information for all NCAA Division I head coaches from 2014-2016.
- Photos of all NCAA Division I head coaches from 2014-2016.
- The Massachusetts Institute of Technology Adult Faces Database, which contains 10,168 facial photos from the adult U.S. population. This database was utilized to help identify and predict facial characteristics applicable to the coaches’ photos.
“We adopted a learning-based approach, which uses deep neural networks to learn feature extraction from a given set of face photos with the labeling of facial attributes, such as the attractiveness and aggressiveness,” Guo said. “After learning by the computer, the network can analyze a test face photo, and estimate the related attributes.”
In essence, a computer program measured attractiveness and aggressiveness based on certain facial features associated with those traits. The method identified 68 separate facial landmark points like eyebrows, eyes and lips.
The researchers have decided to not publicly release the full list, as it would likely drum up endless social media debate and controversy, Humphreys said.
But their working research paper mentions a few examples. Tracy Claeys, head coach at the University of Minnesota in 2015 and 2016, has among the lowest attractiveness scores in the sample. On the other end of the spectrum, Tony Levine, then-head coach of the University of Houston, has among the highest beauty scores.
On the aggressiveness scale, John Bonamego, formerly of Central Michigan University, ranks low while Jim Harbaugh, of the University of Michigan, ranks high.
The researchers contend that their study is the first to find evidence of an aggressiveness premium, which can extend economists’ understanding of observable factors influencing labor market outcomes.
The overall results surprised Humphreys, who has studied sports economics, marketing and sports gambling.
“I’ve been interested in this ‘beauty premium’ literature for some time,” Humphreys said. “You know, ‘it’s not whether you’re really good at your job. It’s just what you look like.’ That actually converts to higher earnings in some labor markets.
“With college football coaching, it’s a high-profile occupation. In most states, the head football coach is the highest paid public employee. And they go into the houses of 18-year-olds trying to convince them to come play football. You’d think being physically attractive would be helpful in those situations, but it’s not.”
A sign that this research is completely unbiased, Guo, the computer scientist, was asked if he knew where Brown and Holgorsen fit on the scale.
“To be honest,” Guo said, “I’m not familiar with those two people.”
CONTACT: Jake Stump, Director
WVU Research Communications
Follow @WVUToday on Twitter.