Managers are increasingly asking their employees to rate each other’s work in a practice known as peer evaluation. How well those evaluations work, and whether bias plays a role, depends on a surprising factor, according to new Cornell research: when the peers evaluate each other.

When peers evaluate each other after an outside source, such as a client, has confirmed a team’s success, ingroup favoritism goes down. But when the team fails or gets no feedback at all, the peers’ biases tend to stick around – or even get worse, the study found.

Doing peer evaluations after a team’s success is confirmed can help reduce bias and give everyone a fair shot, according to the study “Peer Evaluations in Diverse Teams: How External Validation of Team Performance Influences Ingroup Favoritism,” published May 21 in Accounting Organizations and Society. Martin Wiernsperger, assistant professor at the Samuel Curtis Johnson Graduate School of Management, coauthored the study with Gerhard Speckbacher, professor at the Vienna University of Economics and Business.

“Peer evaluations aren’t perfect. Workers naturally favor others whom they relate to or feel similar to,” said Wiernsperger. “If you and your coworker are the same gender, went to the same school, or just ‘click,’ you might, without realizing it, give them a higher score than someone you don’t know as well or who’s different from you. This bias can be problematic if peer evaluations are used to decide bonuses, promotions or even job security.”

The study found that people are more likely to identify with the whole team – not just their ingroup – if the team is viewed as successful. When a manager, client or customer confirms achievement, it becomes easier for people to say, “We did well as a team,” and harder to cling to their subgroup identity. This broader sense of belonging could reduce the urge to rate only similar teammates more favorably, the researchers found.

But what about when a team fails? According to the researchers’ theory, people tend to distance themselves from failure. Instead of identifying with the team, they might cling even more closely to their subgroup and blame others, often outgroup members.

To test that hypothesis, Wiernsperger and Speckbacher conducted two experiments. In the first study, university students worked in four-person teams to create TV commercials. Teams had two men and two women. After the project, some participants were told how well their team performed before they were asked to evaluate their peers; others were told after.

The researchers found that if study participants were told their team had succeeded before they did peer evaluations, they showed less bias toward their ingroup peers. But when teams were told they failed, the bias increased, with participants rating similar peers more generously and others more harshly.

In the second experiment, students worked in two-person teams on a game involving emoji puzzles. This time, ingroup status was based on whether both teammates were students or if one came from an outside group. Again, the participants found out how their team performed, either before or after evaluating their peers.

This time, the results were a bit different. Participants who knew that the team succeeded reduced their ingroup favoritism, similar to the first study. But learning that the team failed didn’t make the bias worse; it stayed about the same. Across both experiments, external validation of team success makes people evaluate their peers more fairly, the researchers found.

“These findings have real implications for how companies conduct peer reviews,” Wiernsperger said. “Most organizations don’t think much about when these evaluations happen. Some companies ask for them throughout the year, while others wait until projects are finished.”

“Our research suggests that timing matters,” he said. “If peer evaluations happen after success is confirmed by an outside source, people are more likely to give fair assessments of their coworkers, regardless of group identity.”

Sarah Magnus-Sharpe is director of public relations and communications at the Cornell SC Johnson College of Business.