EING beautiful pays off. Economists have found that men with above-average looks are paid about 5 percent more than those with average appearance, while those who are below average in looks have wages 9 percent below the mean.

But is this because of discrimination or productivity?

Clearly, being attractive is part of the job description for some occupations — think of supermodels or television anchors. On the other hand, one might think that beauty would not have much effect on the productivity of, say, newspaper reporters or professors.

However, research by two economists, Daniel Hamermesh and Jeff Biddle, has found that beauty has a substantial payoff across a variety of occupations, even those where it doesn't seem to be inherently valuable.

Perhaps beauty contributes to productivity in subtle and indirect ways. Homely sales representatives might not be able to sell products as effectively as attractive ones, while good-looking reporters might have an easier time finding willing news sources.

On the other hand, it might be that bosses award beautiful people with pay raises independently of performance.

It's hard to disentangle these two hypotheses. If attractive people are able to work better with customers, colleagues and clients, thereby producing more value for employers, one might argue that it is appropriate to reward beauty.

But if there is little relationship between measures of job performance and beauty, then higher wages for attractive people would amount to discrimination based on irrelevant characteristics.

To decide between the productivity hypothesis and the discrimination hypothesis one needs data on job performance.

Recently Mr. Hamermesh, a labor economist at the University of Texas at Austin who has long studied beauty and labor markets, wrote a paper with an undergraduate economics major, Amy Parker, that investigates the effect of beauty on a particular measure of performance: teaching evaluations for college professors.

The economists collected teaching evaluations for 463 courses taught by 94 faculty members at the University of Texas at Austin, along with some characteristics of the instructors, like sex, race, whether they were on tenure track, and whether they were educated in an English-speaking country.

They asked six undergraduate students to rate the photographs of the professors on a 10-point scale and used the average measure as a beauty score. The student ratings on the beauty scale were highly correlated with one another, suggesting that they were measuring the same aspects of appearance.

According to the economists' statistical analysis, good-looking professors got significantly higher teaching scores. The average teaching evaluation was 4.2 on a 5-point scale. Those at the bottom end of the attractiveness scale received, on average, a teaching evaluation of about 3.5, while those on the top end received about 4.5.

Other variables that positively influenced teaching ratings were being a native English speaker, being a non-tenure-track faculty member, and being male. The first fact is hardly surprising, the second is explained by the fact that non-tenure-track faculty members are often selected primarily for their teaching ability, but the third is a bit surprising.

On closer investigation the economists found that good looks were significantly more important for men than women in producing high teaching evaluations. The same effect was found in earlier research relating wages to beauty: being good-looking, or at least not being bad-looking, is significantly more important for men than for women.

There are various ways these results might be biased. Suppose professors who are well organized and tidy put more care both into choosing their photos and in organizing their courses. Then better-looking photographs would be correlated with higher teaching ratings.

To check for this sort of problem, the authors separated out "formal" pictures (men with neckties, women with jacket and blouse) from informal ones. The results didn't change.

There are even more subtle effects that might be present. Maybe those who value beauty want to take art appreciation courses, and this love of beauty influences their subsequent ratings of their professors.

Luckily, several of the courses were taught by multiple instructors, so the authors could examine differences in ratings while controlling for the courses selected. Again, the results stayed essentially the same.

The bottom line is that better-looking professors get higher teaching scores, all else being equal. To the extent that teaching is an important component of job performance, and to the extent that teaching evaluations accurately measure performance, basing salary on measured teaching performance implies a significant wage premium for beauty, even for college professors.

The really tough question is whether being beautiful truly increases teaching performance — that is, helps the students learn more — or whether the students are just reacting to an irrelevant characteristic.

As Mr. Hamermesh and Ms. Parker put it, "What if students simply pay more attention to good-looking professors and learn more?"

This effect would be a true productivity effect. If a television anchor's looks allow him or her to attract more viewers, that makes that person more valuable to the network. If a professor's looks help students learn more, then we might expect that aspect of job performance should be rewarded.

Mr. Hamermesh and Ms. Parker lead off their paper with a quote from the supermodel Linda Evangelista: "It was God who made me so beautiful. If I weren't, then I would be a teacher."

But it appears that the same physical attributes that made Ms. Evangelista a successful model would also contribute to her success as a college professor. At least the students might stay awake in class.

Hal R. Varian is a professor of business, economics and information management at the University of California at Berkeley.