The authors offer a new methodology to help organizations diagnose gender inequities and point to next steps. Using this methodology, companies will be able to make evidence-based decisions about whether to invest their equity budget in, for example, anti-bias and inclusion training to remedy a culture problem or a comprehensive compensation audit to address pay discrimination. The analysis uses data in three stages — compensation, representation, and finally job satisfaction — to reveal hidden gender inequities. Similarly, the analysis could focus on racial equity instead of gender equity by examining differences between, say Black employees and white employees instead of between men and women.
A multinational company employing approximately 22,000 software engineers, mostly in the U.S. and India, had a problem: In the organization as a whole, women were paid on average 33% less than men: For every dollar men made, women earned only 67 cents. But, when comparing women to men with similar job titles, skills, and company tenure, this pay gap shrank to a mere 3% — still statistically significant but practically of much less consequence.
The company was performing an equity analysis — they wanted to ensure that employees were being treated and paid fairly— and so their next step was to look at their org chart. There they found, unsurprisingly, that there were relatively fewer women in higher-paying job positions than in lower-paying ones. But in digging deeper, with our help, they realized that simply promoting women into the upper ranks would not close their gap.
This company’s experience is unfortunately not unique. In our consulting practice, where we help companies bring a data-driven approach to their people management practices, we have encountered several executives who want to understand where their organizations stand in terms of gender equity.They sometimes have a sense of the problem, but aren’t sure what to measure or where to start.
To solve this problem, we have been using an analysis framework that uses internal data to identify and measure six distinct but interrelated dimensions of equity. Our focus here is on gender, but this framework can easily be applied to measure equity along other axes, including race or ethnicity. Our analysis begins with compensation, moves to representation, which we then account for by examining hires, promotions, and exits, and finally ends with job satisfaction. We’ve found that this methodology points to underlying causes in a way that traditional compensation analyses alone cannot. Here’s how it works.
Phase 1: Compensation
Standard analyses, especially of the gender pay gap, generally start by assigning each title to an ordered job level indicating where in the hierarchy a given position is located. The level corresponds to a specific pay band, with compensation increasing the higher up the hierarchy an employee moves. Tiers can then be assigned through the pyramid, with lower tiers representing entry-level positions and higher tiers representing the more senior positions. Plotting the employee headcount in each level reveals the approximate shape of the organization hierarchy, and a typical compensation discrimination analysis would ask whether women and men who are at the same level in the hierarchy are paid the same amount.
But our analysis goes one step deeper and looks at where in the hierarchy women and men tend to be working in the first place. This was how we confirmed that the large disparities in compensation in the multinational company we were working with were due to an overrepresentation of men in the upper echelons of the organization, where obviously the compensation was substantially greater. Still, the importance of women’s representation in higher job positions extends beyond the resulting discrepancies in compensation. An employee’s position in the hierarchy is also linked to their responsibilities, prestige, power, and access. Representation, therefore, is a critical component of any gender equity analysis.
Phase 2: Representation
The representation of women across the organizational hierarchy is measured at a specific point in time. But any given representation clearly comes about as a result of the past movements in and out of each level. Women could be underrepresented at the top because they are less likely to be promoted, because they are more likely to exit from the highest positions, or because fewer of them are hired into top-ranking jobs in the first place — or, of course, some combination of the three. We assess each of these possibilities in turn.
Given that discrepancies in the headcount may already exist, it’s not enough to compare promotion counts by gender. We need to track the promotion rates for employees who start in a given position. For men and women who start the year in the same job position, what fraction of each group ends up being promoted before the end of the year? The raw counts of women being promoted will of course be lower if women are already underrepresented in the headcount, but they may still be moving up at the same rate as their male counterparts. Indeed, that is what we found in the multinational company we studied. When we looked for discrepancies in turnover rates, we again found no statistically significant difference between the two groups. Once they were in a given job position, men and women were equally likely to leave the company.
Since the behavior and outcomes for men and women looked identical in terms of terminations and promotions, the gender discrepancies in the headcount must have arisen from discrepancies in hiring. In the multinational company we studied, we found that at the bottom tier, women represented 44% of hires made over the course of the year; at the top tier, women represented just 7% of hires. The proportion of women hired at each level nearly matched the proportion of women in the headcount — thus we confirmed that the gender composition of new hires was the primary source of underrepresentation of women at the top of the pyramid.
The final critical question of course was whether this discrepancy was due to discrimination in hiring (i.e., the company was less likely to make offers to equally qualified women) or was the result of fewer female applicants being sourced. If it was a sourcing problem, we would need to ask about the current supply of female technologists with at least a decade of experience in order to understand the specific actions required to remedy this issue. Can they source more female applicants externally now, or do they need to focus on developing women within their organization over the long term? With detailed records of the organization’s job applicants, which in this case we unfortunately did not have access to, we could begin to formulate answers to these questions and identify the right course of action.
Phase 3: Job Satisfaction
As one final check for gender equity, we also examine whether women’s sense of happiness in their jobs diverges from that of men. Based on a survey administered early in the year of our study, in which employees reported on a scale from 1 to 10 how happy they were in their careers at this company, we found no significant difference between the happiness levels of men and women, controlling for their geography and title.
Taken together, the evidence suggested that men and women were treated equitably in this organization once they were employed. The primary red flag for gender equity was that women were dramatically underrepresented in the more senior levels of the organization, where the source of this discrepancy was the gender composition of new hires.
Using the methodology
As we saw above in our analysis of the technology company, the framework outlined here helps us to diagnose gender inequities and point to next steps by zeroing in on the action-space. At each stage of the analysis — compensation, representation, and finally job satisfaction — the data reveals something about the nature of gender inequity. In the case of this multinational, our approach allowed us to identify the organization’s most important source of inequity as hiring. However, depending on the data, this methodology will clearly lead the analyst down different “rabbit holes” and suggest different conclusions. Similarly, the analysis could focus on racial equity instead of gender equity by examining differences between, say Black employees and white employees instead of between men and women.
Using this methodology, companies will be able to make evidence-based decisions about whether to invest their equity budget in, for example, anti-bias and inclusion training to remedy a culture problem or a comprehensive compensation audit to address pay discrimination. As more executives start to ask important questions about their organization’s progress on gender equity, we hope this framework will help them decide what to measure, where to start, and ultimately lead them to take high-impact actions that result in a more equitable labor market.