Surprising link found between physical attractiveness and income

In 2016, a dinner conversation between two researchers with seemingly different fields—one an expert in data science and engineering, the other in economic models—laid the groundwork for a groundbreaking study.

Stephen Baek, a data scientist, and Suyong Song, an econometrician, discovered a surprising intersection in their research interests, ultimately leading to an in-depth analysis of the “beauty premium.” This idea suggests that individuals considered more physically attractive tend to earn higher incomes.

Their collaboration culminated in a published journal article that sheds new light on the relationship between body shape and income, while also challenging previous research methods.

Baek, now an associate professor of data science at the University of Virginia, first met Song, an associate professor of economics and finance at the University of Iowa, when both were at Iowa. Baek’s expertise lies in modeling human body shapes for applications like product design and ergonomics, while Song’s work focuses on addressing errors in economic models.

What started as a casual idea sparked a five-year research project that would use advanced technology to explore human body data in a novel way.

Traditional studies examining the beauty premium often relied on simplified metrics such as height, weight, or BMI to estimate physical appearance. Baek and Song took a different approach by utilizing a detailed dataset from the 2002 Civilian American and European Surface Anthropometry Resource (CAESAR) project.

This dataset included not just demographic and income data but also 3D body scans and measurements from nearly 2,400 civilians. The 3D scans, in particular, provided far more nuanced insights into body shape than had previously been possible.

“The issue with previous works was that people were oversimplifying the parameters to describe body shape,” Baek explained. He emphasized that measurements like height, weight, and BMI fail to capture the complexity of human body shapes. This led them to develop a new method for analyzing body data, employing a machine-learning algorithm known as a “graphical autoencoder.”

This deep learning technique allowed them to analyze the 3D scans by breaking them down into numerical values representing essential body features. Once the data was processed, the algorithm could reduce the information into key components—such as stature or body proportions—that are more easily analyzed. Baek and Song used these features to study how body shape correlates with socioeconomic factors like family income.

For both men and women, stature and obesity were key features linked to income levels. Specifically, the researchers found that taller men tended to have higher family incomes, while women with greater obesity tended to have lower family incomes. Additionally, the waist-to-hip ratio, a feature more specific to women’s body shape, was identified as an important factor in the data.

Song’s expertise in economic models allowed the team to explore an additional layer of complexity—measurement error in previous body shape studies. By comparing self-reported height and weight to actual measurements in the CAESAR dataset, Song uncovered a consistent pattern: individuals often misreport their body size.

For example, those who weighed less tended to overestimate their weight, while heavier individuals often underestimated. These errors have skewed the results of past studies that relied on self-reported data.

“Many economists assume that these errors are negligible or average out to zero, but our study showed otherwise,” Song said. “In reality, these errors are correlated with true height and weight, which raises concerns about the accuracy of many studies using self-reported data.”

The implications of their research extend beyond the fields of economics and data science. Baek and Song’s findings highlight the limitations of relying on oversimplified or erroneous data in research on physical appearance and income. Their methods also open the door for more advanced approaches to studying human body shapes in disciplines such as computer science, biology, and social science.

The paper, titled “Body Shape Matters: Evidence from Machine Learning on Body Shape-Income Relationship,” was finally published in the journal PLOS One after three years of work. Both researchers hope their findings will raise awareness not only about the persistence of the beauty premium but also about the importance of addressing errors in previous studies.

By demonstrating how body shape correlates with income and exposing flaws in past research, Baek and Song have set the stage for future studies to take a more accurate and detailed approach to understanding the connections between physical appearance and socioeconomic outcomes.

Note: Materials provided above by The Brighter Side of News. Content may be edited for style and length.

Like these kind of feel good stories? Get The Brighter Side of News' newsletter.