A new and particularly compelling study indicates that even when women and minorities receive the same starting salaries and performance ratings for doing the same job under the same supervisor, their merit increases are smaller than those awarded to their white male counterparts.
The study, conducted by Emilio Castilla, assistant professor at MIT’s Sloan School of Management and a visiting professor at New York University, analyzed the internal records on 8,898 support staff at a large high-tech service-sector company with a workforce of 20,000. The company sits at the cutting edge in research and information technology, prides itself on its diverse workforce and strives for best-company-to-work-for status.
But its merit pay plan produced gender- and race-based pay issues. "The disparities are small but very real," Castilla says. "And any difference is evidence of bias.
"When I presented the findings, the executives were shocked and immediately asked for recommendations to remedy the problem," Castilla recalls. "The company was full of good intentions, but completely unaware of the unintended consequences of its performance system."
"The company was full of good intentions, but completely unaware of the unintended consequences of its performance system."
—Emilio Castilla, assistant professor, Sloan School of Management, MIT
Although the company in the study separated the two stages, it still allowed unit heads some discretion to decide, for example, that an employee with a "five" rating was really a "high five" who warranted a larger increase than that given to others with the same rating.
"The unit heads were not accountable for the discretion they exercised within the small latitude of increases for each rating," Castilla says. "HR had to approve the merit increases recommended by the unit heads, but it basically rubber-stamped the decisions."
The company responded to Castilla’s findings by making the unit heads more accountable and assigning an exact increase for each rating. "You can still provide some difference and discretion, but the same formula must apply to all employees," Castilla says. "Some units, for example, may bring in more revenues, so within that unit, the range of salary increases might be higher and all top performers in that unit would receive a higher increase than top performers in other units."
Pay transparency across the workforce is necessary to control bias, Castilla says. "Companies need to communicate the salary increase associated with each rating. This is another mechanism for correcting pay disparities."
Castilla has presented his study at several conferences. "Compensation directors tend to minimize the findings," he says, "but the federal enforcement agencies have shown an interest." Given the size of the study and the tight controls Castilla used, the findings represent a clear warning for any employer using a merit pay plan.
Workforce Management, November 3, 2008, p. 36 -- Subscribe Now!