Sign up for our newsletter and get the latest big data news and analysis.

Addressing Demographic Pay Gaps with Data-driven Solutions

In this special guest feature, Dr. Margrét Vilborg Bjarnadóttir, Assistant Professor of Management Science and Statistics at the University of Maryland Robert H. Smith School of Business, suggests that despite massive cultural and societal pushes for gender equality, the gender pay gap in the workplace remains as strong as ever. Data analysis and visualization could play a vital role in helping to resolve this important issue. Dr. Margrét Bjarnadóttir graduated from MIT’s Operations Research Center in 2008, defending her thesis titled “Data Driven Approach to Health Care, Application Using Claims Data.” Dr. Bjarnadóttir specializes in operations research methods using large scale data; her research centers around data driven decision making, combining optimization modeling with data analytics.

Demographic pay gaps, including the gender pay gap, are the result of more complex factors than just a desire to minimize payroll expenses. They stem from unconscious biases and processes that are better suited to one group compared to another. And, as multiple executives have found out, good intentions and “mindfulness” are not enough to eliminate the gaps.

Data-driven decision support can help discover biases in the pay structure and can allow organizations to take concrete steps towards reducing and closing demographic pay gaps in a fair and efficient manner.

Why is it a complex issue?

The methodology of measuring the gender pay gap is well known and generally accepted. It is based off a regression model where wages (actually, the log of wages) are explained using bonafide determinants of pay such as job role, education, and performance. The resulting model highlights differences in pay that cannot be explained by anything else except the gender of the employee.

However, how best to close the pay gap has remained an open question. In collaboration with a team of academics, we studied each specific employee’s effect on the gap. And what we found is somewhat counter-intuitive: we can demonstrate that within organizations there may be women who, if given a raise, actually end up increasing the gap. Similarly, we can find men who, if given a raise, actually decrease the gap. In fact, one key finding from our research is that there is almost no correlation between the fairness of a raise and the impact of said raise on the gender pay gap. This means that an organization solely focused on cost efficiency in closing the gap may end up with a skewed and unfair compensation structure.

Data and optimization to the rescue

Given the complexity of the interactions between an employee’s impact on the pay gap and the notion of fairness, there is a great opportunity for the effective use of data-driven methods. Using companies’ data, we can build algorithmic approaches that are more fair and efficient than, say, extending a raise of the same percentage of salary to every female employee.

And data-driven approaches have additional benefits. For instance, they can highlight manifestations of unconscious bias in the pay structure. When we were working with our development partner, we saw that female top performers were not being compensated to the same extent as male top performers: while the pay for top performing females was above average, they were not at the top of the pay scale like the top performing males.

Data-driven methods deliver results

Our methodology and findings speak for themselves: Reykjavik Energy, our initial development partner, has driven down their gender pay gap to zero percent. After building a data-driven cloud solution, they can now test salary decisions – their impact on both the salary structure and the pay gap – on the fly before making them. This ensures that the gap remains closed. The drivers of demographic pay gaps may be complex, but with the right quantitative tools and buy-in from top-level management, we can eliminate them one raise at a time.

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: