The New York Times

June 30, 2005

Online Dating? Thin and Rich Works Here, Too


ONLINE dating is one of the most popular paid services on the Internet. A 2003 report by comScore Networks stated that 40 million Americans had visited an online dating site, and JupiterResearch reported early this year that industry revenue will reach $516 million in 2005.

Recently, three economists - Günter J. Hitsch, Ali Hortaçsu (both from the University of Chicago) and Dan Ariely (from the Massachusetts Institute of Technology) - examined the experiences of a sample of users of a major online dating service and subjected it to empirical scrutiny. Their paper, "What Makes You Click," is available on Mr. Hortaçsu's Web page.

Users who sign up for a dating service typically post a profile describing their age, income and other characteristics along with an optional photo. The researchers asked University of Chicago undergraduates to rate the users' physical attractiveness based on the photos, adding another variable to the mix.

The online service provided the researchers with information about which sites a user browsed, whether the user sent e-mail to other users or replied to them and whether the user exchanged phone numbers. What happened after that particular milestone was not recorded.

Start with the self-reported characteristics. There was a strong Lake Wobegon effect in the data, with only 1 percent of the population admitting to having "less than average" looks. Even so, only a third actually posted a photo. The reported weights of the women were substantially less than national averages and about 30 percent were blonde. The reported weights of the men were consistent with national averages and only about 12 percent were blond.

What are people looking for? The most important variable, for both men and women, is looks. Furthermore, posting a photo is a big help: women who post photos receive about twice as many e-mail messages as those who do not, even when they report that they have "average looks."

Having a lot of money is good for attracting e-mail messages, at least for men. Those men reporting incomes in excess of $250,000 received 156 percent more e-mail messages than those with incomes below $50,000. Women like men with a higher income than they have but men do not want to date women who earn more than they do.

The stated goals for using the service make a big difference in how many e-mail messages are received. Men who are "hoping to start a long-term relationship" receive substantially more e-mail than those who are "just looking/curious." The worst thing a man can say is that he is "seeking a casual relationship," receiving 42 percent fewer e-mail messages than he would otherwise. A woman, by contrast, gets 17 percent more e-mail messages by reporting this goal.

I would guess that none of these findings are terribly surprising. Everyone knows you can't be too rich or too thin.

But even if preferences are all too predictable, there is still the question of who matches with whom. Consider the case of the 90-year-old bachelor who was asked why he was still single. He replied that he was looking for the perfect woman.

"Never found her, eh," the interviewer asked.

"Oh, I found her all right," said the bachelor, "but she was looking for the perfect man."

In this study "matching" is defined as "exchanging phone numbers." The economists' analysis shows that the population tends to exhibit "assortative matching," meaning that people with similar characteristics tend to hook up. Age, income, looks, height, body mass and education are strongly correlated among couples.

But how well does the online market work in matching people up? Relative to what, one might ask. It turns out that there is a well-known algorithm, the Gale-Shapley algorithm, that can be used to find an "optimal" matching of partners.

The researchers used the initial e-mail inquiries to determine how the men and women ranked each other as potential partners, and then used the Gale-Shapley algorithm to compute a "stable" matching of men to women. In this context, a "stable" assignment means that there are no two people who would prefer each other to their assigned partner.

The researchers found that the outcome of the Gale-Shapley algorithm was very close to what was actually observed. This is consistent with my casual survey of single friends: they claim that online dating is no worse than other ways to meet people. In fact, I know one happily married couple who met via online dating.

It is perhaps disheartening that superficial attributes like looks, income and hair color are significant determinants of who succeeds in the dating game. But, as the researchers caution, these are the only attributes that are readily observable in the data. As they point out, "Factors such as personality traits apparently allow us to partly make up for deficiencies in good looks and wealth." Whew, that's a relief.

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