FAYETTEVILLE, Ark. — Using data from Nielsen Media Research ratings, a University of Arkansas labor economist studied the television show “American Idol” and found strong evidence of same-race preferences among viewers for show participants.

“Unfortunately, I found same-race preferences among viewers of this popular show,” said Jungmin Lee, assistant professor of economics in the Sam M. Walton College of Business. “This doesn’t mean these viewers are racist. But it does reveal the reality of same-race preferences and shows how the process works.”

Lee studied ratings of 82 individual shows from all four seasons, 2002 through 2005. The data showed aggregate ratings for total households and ratings for African American households, from which he could disaggregate the ratings. The disaggregated data demonstrated positive correlations between the race of viewers and contestants’ race. For example, a 10 percent increase in the share of black contestants attracted an additional 1.4 percent of black households. Excluding the first season, this same 10 percent increase in the share of black contestants decreased ratings among non-black households by 0.7 percent.

Lee also found correlations between viewership and the success of contestants based on race. For the first several shows of a particular season -- when the number of participants was more than six and ratings were comparatively low -- Lee did not find evidence of voting based solely on racial preference. But that changed with higher ratings.

“As expected, the race of contestants mattered in voting only if there were a relatively small number of contestants,” Lee said. “At the sixth week, when there were only six contestants left, race preferences heated up relative to voting.”

He found that when viewership and voting were taken into account simultaneously, the connection between black viewers and black contestants was particularly strong. With relatively more black viewers, black contestants were less likely to be eliminated. For instance, an increase in the number of black households by slightly more than half a million decreased by 32 percent the likelihood that a black participant would be voted off the show. Lee noticed this phenomenon had a multiplier effect, as more black viewers tuned in the following week.

“The findings suggest voting and viewing behaviors interacted dynamically, especially for higher rankings,” Lee said. “As more black contestants survive, more black viewers watched the show, and more black viewers, particularly those with strong same-race preferences, participated in voting.”

Lee said the findings do not mean that African Americans discriminate against whites or vice versa. Rather, the findings simply reveal the process and potential impact of same-race preference on consumer behavior. His study is part of a large body of research that demonstrates how preference based on race affects economic and labor markets. For example, previous research in the area of consumer discrimination has provided strong evidence that many people do not want to be served by someone of a different race. Moreover, another study showed that store owners in predominantly white neighborhoods tend to hire white employees, especially for businesses that require face-to-face interaction between employees and customers.

Several recent studies have used television viewing patterns because people select and watch programs based on their preferences. These studies revealed that the racial composition of athletic teams and broadcast news anchors had an effect on viewer ratings. Lee was motivated to examine “American Idol” after a recent study found no evidence of racial bias in voting among participants of the “Weakest Link,” another popular television show.

“The difference between the two shows has to do with anonymity,” Lee said. “On 'Weakest Link,’ show participants vote to eliminate other participants -- the weakest links. So, they have no anonymity. They’re out there in front of millions of American viewers, and they probably do not want to appear biased. On 'American Idol,’ though, viewers are the voters, and they are free to reveal their true preferences under anonymity.”

Lee said racial preferences are difficult to observe but can help explain racial discrimination in the labor market.

“It is still questionable whether these pure preferences among television viewers would translate into racial discrimination in the labor market where preferences should be reconciled with economic stakes,” he said. “On the other hand, it is true that social scientists have not found any reasonable cause for things like hate crimes.”

Lee’s study has been published as a working paper for the Institute for the Study of Labor, a prominent German institute of labor economics. The paper is also being reviewed by a major economics journal.