At the large North American retail chain that was the subject of Professor Kelly Shue’s new research, more than half of entry-level workers—56%—are women. But at each rung up the ladder, there are fewer and fewer women: women are 48% of department managers, 35% of store managers, and 14% of district managers—the glass ceiling at work. Why?
The study, co-authored with Alan Benson of the University of Minnesota and Danielle Li of MIT and based on assessment and promotion records for nearly 30,000 workers, finds that women are 14% less likely to be promoted at the company in each year, and that a major factor preventing women from being promoted is that they are consistently judged as having lower leadership potential than men. In a two-part annual assessment, according to the records, women’s performance at the company is rated higher than men’s, on average. But their potential is rated lower—a pattern than continues even when women exceed those expectations.
At the company in the study, managers use a “nine-box” grid—a commonly used tool at large organizations—to evaluate their employees, giving them a low, medium, or high score on their performance over the prior year and a low, medium, or high score on their potential for growth and development. The employees with high scores for both performance and potential—those in the upper right quadrant of the grid—are most likely to be promoted.
While managers can consider real-world metrics in evaluating performance, potential is more abstract—and that might make it more subject to bias.
“What is commonly talked about in terms of management and potential are characteristics such as assertiveness, execution skills, charisma, leadership, ambition,” Shue says. “These are, I believe, real traits. They’re also highly subjective and stereotypically associated with male leaders. And what we saw in the data is a pretty strong bias against women in assessments of potential.”
Specifically, Shue and her colleagues found that while women receive higher performance ratings—they are 7.3% more likely than men to receive a “high” rating in performance—their potential ratings are 5.8% lower. The authors estimate that lower potential ratings explain up to 50% of the gap in promotions.
Could managers be correct in their assessment that women at the company are excellent performers in their current roles but lack the skills to be successful at a higher level? To the contrary, the researchers found that managers consistently underestimate women’s ability to perform in the future. They identified women and men with similar performance and potential scores for a given evaluation period, then looked forward to the next period and found that women tended to have higher performance scores than men, whether or not they been promoted into a more senior role.
“It appears that they were held to a higher standard,” Shue says.
And even when women have exceeded expectations, they still don’t get the benefit of the doubt. Women continue to receive lower potential scores after they have demonstrated through their performance that the previous period’s potential score was inaccurate.
“Women get progressively lower potential scores relative to their actual future performance as we rise up the corporate ladder. So this is going to contribute to a stronger and stronger glass ceiling.”
This effect gets stronger for more senior positions, Shue points out. “Women get progressively lower potential scores relative to their actual future performance as we rise up the corporate ladder. So this is going to contribute, I think, to a stronger and stronger glass ceiling the higher up we go.”
Not surprisingly, being shut out of more senior roles also keeps women from earning as much as men. The authors estimate that 70% of the gender pay gap at the company is attributable to gender differences in job levels.
What can companies do to combat this kind of bias in their evaluation systems? One solution would be to remove potential scores from annual assessments altogether, and simply promote employees based on their performance. But, as Shue explored in earlier research, the best performers don’t always make the best managers. And, the authors point out, the data shows that despite the gender bias that they are subject to, potential scores do help the company in the study identify effective managers more accurately than performance scores alone.
The authors also wondered if having more women managers would help reduce bias in evaluating potential. “We thought the group that would be least likely to have these stereotypes of a male leadership in mind would be female managers,” Shue says. But they found that women also give lower scores for potential to their high-performing women subordinates.
Shue and her colleagues suggest that to avoid this type of bias, companies should pay close attention to their own internal data. For example, companies can use algorithms to look for systematic gaps between performance and potential ratings of the kind found in the study.
Or they can look beyond manager evaluations and identify other metrics that predict leadership talent, Shue says: “Look at other available data and see what variables actually forecast being a good manager.”