Attempting to gage the intensity of discrimination is nearly impossible. However, a television game show called Weakest Link allowed author Steven D. Levitt to investigate discrimination through the voting of contestants against other contestants. Levitt finds which of two theories of discrimination, information-based and taste-based, were influential against which categorization of people (race, age). Surprisingly, Weakest Link showed little discrimination against women and blacks, groups many would presuppose to be discriminated against. However, Levitt found a taste-based bias against the old and an information-based bias against Hispanics.
Weakest Link is played by a number of contestants that answer trivia questions over a series of rounds, voting off one person until only two remain. The game show is constructed to incentivize contestants to vote in a manner that will maximize their own probability of winning. Questions are asked each round, and contestants can “bank” the current amount of money and start at zero, or aim for a higher reward for a right answer with the possibility of losing all the money with a right answer. The more sequential right answers, the more money the group earns.
In the early stages, most players’ strategy is to vote off those contestants who answer incorrectly because they decrease the money in the bank (the possible winnings). However, as the show nears its end, strategies switch, and contestants become more concerned with whether they can beat another contestant in the head-to-head final battle for all the money. Consequently, contestants start voting off strong opponents instead of weak ones.
The contestants only have a few moments to vote, so many quick judgments of others’ skill levels are probably made based on preconceived notions regarding their race, gender, style, or mannerisms. The authors refer to this as information based discrimination. In Weakest Link, this would appear by contestants voting against the group in earlier rounds because they perceived the group to be stupider and less likely to get the answers correct. In later rounds, contestants would be less likely to vote against that group because they want an easier head-to-head opponent.
Taste-based discrimination claims some people simply dislike a particular group of people and are willing to pay a financial cost to avoid any interactions. Applied to Weakest Link, taste-based discrimination reveals itself through continued systematic voting against the same group in both early and later stages. Unlike information-based, taste based discrimination has nothing to do with skill or ability, just dislike. Thus, contestants will vote against the group no matter what the strategic move would be if the contestant were worried about skill.
To see where information and taste based discrimination appeared, Levitt constructed a regression that controlled for race, gender, age, education level and region of residence. The data was divided into three groups: early rounds, middle rounds and the final round. Confirming Levitt’s theory on contestant strategy, poor performance lead to more votes in the early and middle round while this tendency to punish poor performers disappears in the final round. In addition, revenge emerged in voting patterns and a vote for an opponent would increase the likelihood of them voting for you in a later round.
Looking specifically for discriminatory patterns, Levitt looked at differences in race, education, gender and age. The most surprising finding was that little evidence existed showing racism towards blacks, not even when looking at white voting towards blacks. However, Levitt found significant evidence of an information based discrimination against Hispanics. Stereotyped as less intelligent, Hispanics were voted off in early rounds, but kept in later rounds.
Education showed an interesting finding that only doctors received an information bias, not those with college or graduate degrees. Contestants perceived doctors to be smarter. Also surprising was the lack of evidence supporting the discrimination of women. While women vote slightly more for men and visa versa, no significant evidence showed any sexism. The final surprise was that of age discrimination. Older contestants received taste bias, and received continued excess votes throughout all rounds.
Weakest Link offers a singular opportunity to look at discrimination in a somewhat controlled setting. I appreciated the unconventional method of looking at discrimination. The findings surprised me some, as the media often tells me, as a woman, that I will make less and be in a male-dominated business world, and as a white person, that people of my skin color discriminates against blacks. As Levitt discussed, the game show picked the contestants, so they were not representative of the entire US. This could explain the lack of discrimination against women and blacks.
Perhaps the data represents a more liberal, accepting group of people indifferent to ones race or gender, in terms of black or white, male or female. Yet, discrimination against Hispanics and older people isn’t a main stream concern as of yet and people may not be as conscious of this type of discrimination as they are other types. This data set could show us that even in a liberal, educated environment, discrimination still persists and those are against Hispanics and the old. It gives us insight into what people are unaware of and what discrimination continues in obscurity, if people educated enough to be on a trivia show still hold those ideas.
I am correlating education with accepting diversity in my above argument. I would be interested to see a regression that showed how different levels of education voted against other contestants.
Another point to bring up is the factor of omitted variables. Having read other papers on how attributes such as height, weight and attractiveness can have an impact on discrimination, I wonder if those personal characteristics could have played a role in this game show. As contestants see and interact with each other before voting, physical characteristics could have an impact on voting.
Otherwise, the paper was very well organized, easy to understand and intriguing. Steve D. Levitt has an unconventional method of looking at certain topics through strange channels. Most of those channels are closed.
Other Papers by Steven Levitt
Market Distortions when Agents Are Better Informed: The Value of Information in Real Estate