A new study shows female managers are more than three times as likely as their male counterparts to underrate their bosses' opinions of their job performance.
The discrepancy increases with women older than 50, the study states.
"Women have imposed their own glass ceiling, and the question is why," said Scott Taylor, an assistant professor at the University of New Mexico Anderson School of Management who conducted the study.
Taylor will present his findings today in Chicago at the annual meeting of the Academy of Management, a 19,000-member organization devoted to research and teaching.
"It's pretty fascinating, actually. It's a different take on it," said Leanne Atwater, a management professor at the University of Houston.
Atwater has researched the standard management assessment tool that Taylor was examining when he discovered the gender difference.
In the study, 251 male and female managers from different industries nationwide rated themselves and requested ratings from supervisors, peers and subordinates. Each subject also was asked to predict the ratings made by others.
Taylor collected the data for the study in 2005 while a doctoral student at Cleveland-based Case Western Reserve University.
The ratings measured nine elements of emotional and social competence essential to leadership: communication ability, initiative, self-awareness, self-control, empathy, bond-building, teamwork, conflict management and trustworthiness.
The men who were studied slightly overestimated how their bosses would rate them, while the female respondents underestimated their ratings on average by about 11 percent.
Chelsea Walker, 52, an administrator for UNM's College of Pharmacy, participated in a similar exercise while taking Taylor's class and was shocked to find her results matched her professor's findings.
"I was very, very surprised by his responses," she said. "I guess that I just didn't think that he thought that highly of me, even though I thought pretty highly of myself."
The exercise was a confidence booster, Walker said.