Economics of Beauty: Measurement (Part 1)

This is the first of 3 posts on the Economics of Beauty. Post 2 and 3 will focus on “Beauty and the Labour Market” and “Beauty and the Social Market”. All these posts are summarizing Dan Hamermesh’s fantastic book *Beauty Pays. To incorporate looks into a model of labour or social interaction, one must first introduce the variable of beauty. Here I will discuss how beauty is measured in academic literature and what problems arise through such measurement. I will also summarize some mitigations regarding systematic measurement errors.

Since the 1971 national representative survey by the University of Michigan, including the variable of beauty, the 5-1 scale is the gold standard in beauty literature. Here the person surveying has to rate the survey participant as either:

5. Strikingly Handsome or Beautiful

4. Good-looking

3. Average

2. Quite plain

1.Homely

It turns out that this measurement is far more uniform among surveyors than a 10-1 score. Uniform, in this case, means that different surveyors, rating the same person, are extremely likely to be in close agreement. If 4 people are asked to rate one human about 70% of the time, 3 of their rating will only differ by one point. 14% of the time they are in complete agreement. In only 0.1% all four surveyors had a different rating for a person.

It turns out we agree on beauty after-all. But this measure comes with some problems. We give different average scores to males and females, and to different age groups. Women are generally rated slightly more favorable than men, even when women do the rating. Also, young people are rated more favorable on average. Older women are worse perceived than older men. But it's not only the averages that differ but it’s also the dispersion. Women are rated more at the extremes, either homely or beautiful. Men are more likely to be rated somewhere in the middle. Interestingly African Americans are also not rated at the extremes. They are more likely to get a 2., 3., or4., than a 1. or 5., compared to whites. One must point out that surveyors are mostly whites, which might explain the low variance in African American ratings.

What we observed until now is that there is a pretty uniform measure of people (the 5-1 scale), that is universally skewed. That means that on average people agree on ratings of other people, but they are affected by age, gender, and race. This is pretty good news. This now means that one can adjust those measurements knowing those biases. Artificially correcting for those heuristics means: rank up elder people, disperse African American ratings, and decrease women’s ratings (or vice versa for each).

We now have a measure of beauty, scaling from 1-5 that people agree on, adjusted for age, gender, and race, that is ready to be implemented in all kinds of models. Tomorrow we will discuss how this variable affects income and/or productivity, and how this is measured.

*Evidence for all these numbers can be found in Hamermesh’s book. This blog post refers to chapters "How do we measure Beauty?", "Do Observers Agree on Beauty?", and "Does Beauty Differ by Gender, Race, or Age?"I do recommend you read it. It’s a fantastic summary of a strange and beautiful field of economics. Hamermesh is one of the leading researchers on economics of beauty. He taught Economics of Life at my uni for undergrad. A lot of motivation for this blog comes from his inspiring lectures.