When bias hurts profits

Researchers have long known that bias can have an impact on hiring, but a new Harvard study suggests that it may also affect workplace performance.

Co-authored by Amanda Pallais, the Paul Sack Associate Professor of Political Economy and Social Studies, the study, which is based on data collected from a French grocery store chain, found that minority workers were far less efficient in a handful of important metrics when working with biased managers.

That drop in performance, Pallais and co-authors Dylan Glover and William Pariente say, can create a self-fulfilling prophecy among bosses biased against minorities. Since such managers believe that minority workers are worse employees, that bias can result in poorer performance by the affected minorities, and so managers assure themselves that the minority workers are worse employees.

“We were able to use metrics — how fast cashiers scan items, how much time they spend between customers, number of absences — to track their performance,” Pallais said. “But unlike many other jobs, the same worker works with different managers on different days, so we were able to compare the same worker’s performance on Monday with one manager to their performance on Tuesday with another manager.

“When we compared the same worker’s performance under biased and unbiased managers, we found two things,” she continued. “First, minorities perform much worse under biased managers. With biased managers, they perform on average at the 53rd percentile, but with unbiased managers, they perform at the 79th percentile. Second, there was no difference in the performance of other workers under biased or unbiased managers.”

To measure manager bias, Pallais and colleagues borrowed the implicit association test (or IAT) from psychology literature.

The test takes place over two rounds, during which participants sort words — in this case, French and North African names, and both good and bad adjectives — to either side of a computer screen. In the first round, the test takers sorted French names and good adjectives to the same side, and North African names and bad adjectives to the other side. In the second round, the rules are reversed, with French names and bad adjectives sorted together. A biased manager taking the test, she said, would be faster to sort “good” adjectives — punctual, reliable, hard-working — and French names together, and slower to the do the same with North African names.

One explanation for that change in performance could be that biased managers are simply poor ones, but Pallais said the evidence doesn’t support that. The workers who weren’t minorities on average performed about the same, regardless of whether managers were biased. Moreover, none of the negative effects that minorities experienced from working with biased managers could be explained by other managerial characteristics such as bosses’ age, gender, or ethnicity.

While the productivity changes can translate into lower profits, Pallais and colleagues found that the economic impact wasn’t limited to the stores. When minority employees are scheduled to work with biased managers, they are much more likely to be absent and less likely to stay late. Since workers are paid only for the time they spend on the clock, the minorities lose 2.5 percent of their pay when working with biased managers.

In their search for precisely why minority employees perform less efficiently when working with biased managers, Pallais and colleagues found a surprising answer.

“You might think that biased managers simply dislike minorities and treat them poorly, or assign them to the least pleasant jobs,” she said. “But we don’t find any evidence of that.”

Researchers surveyed workers about whether they liked their managers and whether they thought their managers liked them. Minorities didn’t say they disliked biased managers or that they thought biased managers disliked them. They did say they were less likely to interact with managers who were biased.

“It appears there’s evidence for what psychologists call aversive racism,” Pallais explained. “When you’re biased toward someone, you speak less to them, you’re more hesitant to speak to them, and you’re less friendly toward them because you’re uncomfortable with them.

“What it looks like is the biased managers just interact less with the minorities,” she continued.

Previous research has shown that manager interaction is a powerful predictor of performance, and in the case of grocery store clerks, Pallais said, “When workers know someone is watching, they put in more effort.”

It’s unclear, Pallais said, whether simply mandating more interaction between managers and workers might increase productivity among minority workers. Going forward, she hopes to explore what effect such policy changes might have.