Counterintuitive Ways to Close the Gender Pay Gap

SMITH BRAIN TRUST — On Women’s Equality Day last week at the White House, Apple, Facebook, Microsoft, General Motors and other big U.S. companies signed an “Equal Pay Pledge” to close the gender pay gap for their employees. On the same day, a pair of states passed laws for the same objective — to eliminate the often-cited statistic of U.S. women earning about 79 cents for every dollar earned by men.

Maryland expanded its equal pay law by prohibiting businesses from retaliating against employees for discussing or disclosing salaries. California passed a bill requiring firms, which do billions of dollars of business annually with the state, to have policies in place according to the California Fair Pay Act of 2015. The new law further requires contractors to provide the state with data on employee pay by gender and race.

Taking note of such developments, including tighter scrutiny on pay inequality in Europe, Margrét V. Bjarnadóttir, an assistant professor at the Robert H. Smith School of Business, has developed an algorithm designed to help companies close the gender pay gap — or any demographical gap — in the most efficient way possible. Together with her collaborators, David Anderson, a Smith School operations management PhD graduate now at the City University of New York, Cristian Dezső of the Smith School and David Ross from University of Florida, she has a provisional patent and has begun to test the market for the tool, whose logic (and limitations) she explains in a new working paper, “On a Firm’s Optimal Response to Pressure for Gender Pay Equity."

The first step in remedying a pay gap is controlling for legitimate reasons why women might earn less than men: They might have less experience, say, or choose to work fewer hours or to hold less-demanding positions. In the working paper, which used a company with 506 women and 266 men in 40 job categories as a test case, the authors found an absolute pay gap of 43 percent. But after applying controls, the unexplained "residual" pay gap dropped below ten percent — an amount that could be attributable to discrimination.

One approach to closing the pay gap would be to give every woman a raise of equal proportion, until the gap vanished. But that "naïve" method can cost more than twice as much as identifying and targeting the women whose pay most influences the gap. Those women tend to be more like their male colleagues in terms of credentials and career paths, Bjarnadóttir says. "In contrast," she says, "if you have a job group that is entirely female, raising the salary of those women does nothing to close the company's pay gap." That's because the standard is equal pay for equal work, and no gender gap can exist — by definition — when all employees doing equal work are the same gender.

In some rare cases, the measured pay gap can be reduced by giving raises to men. Meanwhile, giving raises to certain women can widen the gender pay gap. This relates to the challenges of defining “equal pay for equal work,” which requires an assessment of education, experience, performance and other traits. Consider an extreme example: A trait that only some men have — and no women — but actually has nothing to do with performance. If you increase the salaries of men with that trait, the trait will start to "explain" some of the gender pay gap, and therefore reduce the gap. “In our real-world examples, this is very rare,” Bjarnadóttir says. “We have seen males with small such effects but nothing large.”

In the company studied, closing the gap by targeting the women whose pay most influences the disparity costs just below 2 percent of total payroll. In contrast, closing the gap using the “everyone-gets-a-raise” method costs more than 4 percent of payroll. Some practical hurdles stand in the way of maximum efficiency, however. In the company studied, the algorithm recommended that five women get raises of 50 percent or more, including a 97 percent raise for one. "In reality, that wouldn't fly," Bjarnadóttir says. The algorithm can therefore take into account maximum allowable raises for both individuals and groups so that, for instance, no one gets a raise of more than 10 percent, or 15 percent, in a year.

The working paper is frank about the unintended consequences of the current methods of identifying and remedying the pay gap. The method is designed to help companies comply with the current standards for equal pay offered by third-party groups like the EDGE Certified Foundation. "Is this the same thing as pay equity?" Bjarnadóttir asks. "Maybe not." A law firm in which all the lawyers were men and all the secretaries women, for instance, might meet an "equal pay for equal work" standard, despite the deeply gendered structure of the workforce. But for companies trying to meet the tough new legal standards, the algorithm can provide critical insights during annual salary reviews and restructuring.

READ MORE: On a Firm’s Optimal Response to Pressure for Gender Pay Equity, by David Anderson, Margrét V. Bjarnadóttir, Cristian Dezső and David Gaddis Ross, is a working paper available for review on the Social Science Research Network.