第 11 页

answer trivia questions and compete for a single cash jackpot. But the player

who answers the most questions correctly isn’t necessarily the player who

advances. After each round, every contestant votes to eliminate one other

contestant. A player’s trivia-answering ability is presumably the only

worthwhile factor to consider; race, gender, and age wouldn’t seem to matter.

But do they? By measuring a contestant’s actual votes against the votes that

would truly best serve his self-interest, it’s possible to tell if discrimination is at

play.

The voting strategy changes as the game progresses. In the first several rounds, it

makes sense to eliminate bad players since the jackpot grows only when correct

answers are given. In later rounds, the strategic incentives are flipped. The value

of building the jackpot is now outweighed by each contestant’s desire to win the

jackpot. It’s easier to do that if you eliminate the other good players. So, roughly

speaking, the typical contestant will vote to eliminate the worse players in the

early rounds and the better players in the later rounds.

The key to measuring the Weakest Link voting data is to tease out a contestant’s

playing ability from his race, gender, and age. If a young black man answers a lot

of questions correctly but is voted off early, discrimination would seem to be a

factor. Meanwhile, if an elderly white woman doesn’t answer a single question

correctly and is still not voted off, some sort of discriminatory favoritism would

seem to be at play.

Again, keep in mind that all of this is happening on camera. A contestant knows

that his friends, family, and co-workers are watching. So who, if anyone, is

discriminated against on The Weakest Link?

Not, as it turns out, blacks. An analysis of more than 160 episodes reveals that

black contestants, in both the early and late rounds of the game, are eliminated at

a rate commensurate with their trivia-answering abilities. The same is true for

female contestants. In a way, neither of these findings is so surprising. Two of the

most potent social campaigns of the past half-century were the civil rights

movement and the feminist movement, which demonized discrimination against

blacks and women, respectively.

So perhaps, you say hopefully, discrimination was practically eradicated during

the twentieth century, like polio.

Or more likely, it has become so unfashionable to discriminate against certain

groups that all but the most insensitive people take pains to at least appear fair-

minded, at least in public. This hardly means that discrimination itself has

ended—only that people are embarrassed to show it. How might you determine

whether the lack of discrimination against blacks and women represents a true

absence or just a charade? The answer can be found by looking at other groups

that society doesn’t protect as well. Indeed, the Weakest Link voting data do

indicate two kinds of contestants who are consistently discriminated against: the

elderly and Hispanics.

Among economists, there are two leading theories of discrimination.

Interestingly, elderly Weakest Link contestants seem to suffer from one type,

while Hispanics suffer the other. The first type is called taste-based

discrimination, which means that one person discriminates simply because he

prefers to not interact with a particular type of other person. In the second type,

known as information-based discrimination, one person believes that another

type of person has poor skills, and acts accordingly.

On The Weakest Link, Hispanics suffer information-based discrimination. Other

contestants seem to view the Hispanics as poor players, even when they are not.

This perception translates into Hispanics’ being eliminated in the early rounds

even if they are doing well and not being eliminated in the later rounds, when

other contestants want to keep the Hispanics around to weaken the field.

Elderly players, meanwhile, are victims of taste-based discrimination: in the

early rounds and late rounds, they are eliminated far out of proportion to their

skills. It seems as if the other contestants—this is a show on which the average

age is thirty-four—simply don’t want the older players around.

It’s quite possible that a typical Weakest Link contestant isn’t even cognizant of

his discrimination toward Hispanics and the elderly (or, in the case of blacks and

women, his lack of discrimination). He is bound to be nervous, after all, and

excited, playing a fast-moving game under the glare of television lights. Which

naturally suggest another question: how might that same person express his

preferences—and reveal information about himself—in the privacy of his home?

In a given year, some forty million Americans swap intimate truths about

themselves with complete strangers. It all happens on Internet dating sites. Some

of them, like Match.com, eHarmony.com, and Yahoo Singles, appeal to a broad

audience. Others cater to more specific tastes: ChristianSingles.com, JDate.com,

LatinMatcher.com, BlackSinglesConnection.com, CountryWesternSingles.com,

USMilitarySingles.com, PlusSizeSingles.com, and Gay.com. Dating websites are

the most successful subscription-based business on the Internet.

Each site operates a bit differently, but the gist is this: You compose a personal ad

about yourself that typically includes a photo, vital statistics, your income range,

level of education, likes and dislikes, and so on. If the ad catches someone’s

fancy, that someone will e-mail you and perhaps arrange a date. On many sites,

you also specify your dating aims: “long-term relationship,” “a casual lover,” or

“just looking.”

So there are two massive layers of data to be mined here: the information that

people include in their ads and the level of response gleaned by any particular

ad. Each layer of the data can be asked its own question. In the case of the ads,

how forthright (and honest) are people when it comes to sharing their personal

information? And in the case of the responses, what kind of information in

personal ads is considered the most (and least) desirable?

Two economists and a psychologist recently banded together to address these

questions. Ali Horta.su, Günter J. Hitsch, and Dan Ariely analyzed the data from

one of the mainstream dating sites, focusing on roughly 30,000 users, half in

Boston and half in San Diego. Fifty-seven percent of the users were men, and the

median age range for all users was twenty-six to thirty-five. Although they

represented an adequate racial mix to reach some conclusions about race, they

were predominantly white.

They were also a lot richer, taller, skinnier, and better-looking than average. That,

at least, is what they wrote about themselves. More than 4 percent of the online

daters claimed to earn more than $200,000 a year, whereas fewer than 1 percent

of typical Internet users actually earn that much, suggesting that three of the four

big earners were exaggerating. Male and female users typically reported that

they are about an inch taller than the national average. As for weight, the men

were in line with the national average, but the women typically said they

weighed about twenty pounds less than the national average.

Most impressively, fully 70 percent of the women claimed “above average”

looks, including 24 percent claiming “very good looks.” The online men too were

gorgeous: 67 percent called themselves “above average,” including 21 percent

with “very good looks.” This leaves only about 30 percent of the users with

“average” looks, including a paltry 1 percent with “less than average” looks—

which suggests that the typical online dater is either a fabulist, a narcissist, or

simply resistant to the meaning of “average.” (Or perhaps they are all just

realists: as any real-estate agent knows, the typical house isn’t “charming” or

“fantastic,” but unless you say it is, no one will even bother to take a look.)

Twenty-eight percent of the women on the site said they were blond, a number

far beyond the national average, which indicates a lot of dyeing, or lying, or both.

Some users, meanwhile, were bracingly honest. Eight percent of the men—about

1 in every 12—conceded that they were married, with half of these 8 percent

reporting that they were “happily married.” But the fact that they were honest

doesn’t mean they were rash. Of the 258 “happily married” men in the sample,

only 9 chose to post a picture of themselves. The reward of gaining a mistress

was evidently outweighed by the risk of having your wife discover your

personal ad. (“And what were you doing on that website?” the husband might

bluster, undoubtedly to little avail.)

Of the many ways to fail on a dating website, not posting a photo of yourself is

perhaps the most certain. (Not that the photo necessarily is a photo of yourself; it

may well be some better-looking stranger, but such deception would obviously

backfire in time.) A man who does not include his photo gets only one-fourth the

volume of e-mail response of a man who does; a woman who doesn’t include her

photo gets only one-sixth the response. A low-income, poorly educated,

unhappily employed, not-very-attractive, slightly overweight, and balding man

who posts his photo stands a better chance of gleaning some e-mails than a man

who says he makes $200,000 and is deadly handsome but doesn’t post a photo.

There are plenty of reasons someone might not post a photo—he’s technically

challenged or is ashamed of being spotted by friends or is just plain

unattractive—but as in the case of a brand-new car with a for-sale sign,

prospective customers will assume he’s got something seriously wrong under

the hood.

Getting a date is hard enough as it is. Fifty-seven percent of the men who post

ads don’t receive even one e-mail; 23 percent of the women don’t get a single

response. The traits that do draw a big response, meanwhile, will not be a big

surprise to anyone with even a passing knowledge of the sexes. In fact, the

preferences expressed by online daters fit snugly with the most common

stereotypes about men and women.

For instance, men who say they want a long-term relationship do much better

than men looking for an occasional lover. But women looking for an occasional

lover do great. For men, a woman’s looks are of paramount importance. For

women, a man’s income is terribly important. The richer a man is, the more e-

mails he receives. But a woman’s income appeal is a bell-shaped curve: men do

not want to date low-earning women, but once a woman starts earning too

much, they seem to be scared off. Men want to date students, artists, musicians,

veterinarians, and celebrities (while avoiding secretaries, retirees, and women in

the military and law enforcement). Women do want to date military men,

policemen, and firemen (possibly the result of a 9/11 Effect, like the higher

payments to Paul Feldman’s bagel business), along with lawyers and financial

executives. Women avoid laborers, actors, students, and men who work in food

services or hospitality. For men, being short is a big disadvantage (which is

probably why so many lie about it), but weight doesn’t much matter. For

women, being overweight is deadly (which is probably why they lie). For a man,

having red hair or curly hair is a downer, as is baldness—but a shaved head is

okay. For a woman, salt-and-pepper hair is bad, while blond hair is very good. In

the world of online dating, a headful of blond hair on a woman is worth about

the same as having a college degree—and, with a $100 dye job versus a $100,000

tuition bill, an awful lot cheaper.

In addition to all the information about income, education, and looks, men and

women on the dating site listed their race. They were also asked to indicate a

preference regarding the race of their potential dates. The two preferences were

“the same as mine” or “it doesn’t matter.” Like the Weakest Link contestants, the

website users were now publicly declaring how they felt about people who

didn’t look like them. They would act on their actual preferences later, in

confidential e-mails to the people they wanted to date.

Roughly half of the white women on the site and 80 percent of the white men

declared that race didn’t matter to them. But the response data tell a different

story. The white men who said that race didn’t matter sent 90 percent of their e-

mail queries to white women. The white women who said race didn’t matter sent

about 97 percent of their e-mail queries to white men.

Is it possible that race really didn’t matter for these white women and men and

that they simply never happened to browse a nonwhite date that interested

them? Or, more likely, did they say that race didn’t matter because they wanted

to come across—especially to potential mates of their own race—as open-

minded?

The gulf between the information we publicly proclaim and the information we

know to be true is often vast. (Or, put a more familiar way: we say one thing and

do another.) This can be seen in personal relationships, in commercial

transactions, and of course in politics.

By now we are fully accustomed to the false public proclamations of politicians

themselves. But voters lie too. Consider an election between a black candidate

and a white candidate. Might white voters lie to pollsters, claiming they will vote

for the black candidate in order to appear more color-blind than they actually

are? Apparently so. In New York City’s 1989 mayoral race between David

Dinkins (a black candidate) and Rudolph Giuliani (who is white), Dinkins won

by only a few points. Although Dinkins became the city’s first black mayor, his

slender margin of victory came as a surprise, for pre-election polls showed

Dinkins winning by nearly 15 points. When the white supremacist David Duke

ran for the U.S. Senate in 1990, he garnered nearly 20 percent more of the vote

than pre-election polls had projected, an indication that thousands of Louisiana

voters did not want to admit their preference for a candidate with racist views.