Monday, June 1, 2009

Caveat for Employers

Customer-satisfaction surveys -- increasingly used by businesses to determine rewards for employees -- suffer from systematic prejudices, new research suggests.

Although many businesses think they are rewarding the best employees by relying on what customers say, that system effectively discriminates against women and minorities and can undermine business performance, according to a study to be published in the Academy of Management Journal.

"Right now, businesses think customer satisfaction surveys are highly reliable," said lead author David R. Hekman, assistant professor of management at the University of Wisconsin at Milwaukee. "They are highly reliable -- but they are reliably wrong."

Hekman and his colleagues evaluated 12,091 patient reports about 113 doctors working at a large HMO in the Pacific Northwest. They also studied objective data about the doctors. The regularity with which doctors place heart patients on certain drugs, for example, is a good measure of the quality of their care. The number of e-mails doctors send patients is a measure of their accessibility. And the number of questions doctors ask patients during checkups is a measure of their diligence.

In all these domains, however, Hekman found that these objective measures of performance correlated with patient satisfaction reports only when the doctors were white men. For women and minorities, extra quality, accessibility and diligence not only did not result in better evaluations by patients -- they produced worse evaluations.

"E-mails make patients happier only if you are a white male doctor," Hekman said, noting that 6 percent of the doctors' income was based on the patient ratings. "It does not make sense -- working harder seems to be counterproductive for women and minorities."

In a related experiment involving bookshop employees, volunteers were shown two videotaped interactions between a customer and a sales clerk and were told to imagine they were customers and rate the shop's service. Some were shown a white male sales clerk, while others were shown a black male clerk or a white female clerk. All the clerks were actors -- and everything else in the videos was identical, down to the script.

Those shown the white male clerk rated the service provided 19 percent higher than volunteers shown the woman or the black man. They also rated stores with white male clerks as being cleaner.

Hekman also studied the satisfaction levels of 3,600 golfers at 66 clubs nationwide. Clubs that employed higher numbers of Latinos were rated more poorly than clubs employing fewer minorities -- even when they performed identically on objective measures.

Hekman said the point was not to suggest that because white men generate higher customer satisfaction, companies act "rationally" when they give larger bonuses to white men than to women and minorities, but to alert customers about potential biases and businesses about the limitations of customer-satisfaction reports.

-- Shankar Vedantam

Do you remember Mr. Ed -- the talking horse on the popular 1960s television show? Well, no one has found a horse or any other animal yet that can talk. But scientists in Germany may have taken a small step toward the Dr. Doolittle world of talking animals.

Researchers at the Max Planck Institute for Evolutionary Anthropology in Leipzig have created a genetically engineered strain of mice that they are calling partially "humanized" because the animals have been endowed with a gene thought to play a key role in the human ability to speak.

The goal of the research was not to make mice speak but to try to get a better understanding of how human speech evolved. Humans and chimpanzees have versions of the gene, known as FOXP2, but the human form is slightly different, with two minor differences in its amino acid building blocks.

"Changes in FOXP2 occurred over the course of human evolution and are the best candidates for genetic changes that might explain why we can speak," said Wolfgang Enard, who led the research published Friday in the journal Cell. "The challenge is to study it functionally."

Enard and his colleagues spliced the human version of the gene into mice, which have a similar gene of their own, and studied the resulting pups. Although the mice seemed normal in most ways, they showed subtle but important changes in a part of their brains known as the basal ganglia, which plays a central role in human speech, the researchers reported. Nerve cells in the basal ganglia of mice with the human form of the gene appeared to form more and stronger connections.

In addition, the pups with the gene that were removed from their mothers made sounds that were slightly different from regular mice in subtle ways, such as having a slightly lower pitch.

But further study will be needed, the researchers said, to understand the significance of the differences and exactly how much of a role the single gene plays in human speech.

"More studies will be needed to clarify to what extent mouse vocalizations can model aspects of human speech evolution," the researchers wrote.

-- Rob Stein