Firming up pay equality

That women are paid less than men for the same job is unacceptable. Yet, as countries name and shame firms to tackle pay disparity, some caution is necessary. Firms are not countries, and equality within firms may imply little equality in the economy. For instance, contracting out low-paid jobs where female workers happen to be a majority (nurses versus doctors) will enhance pay equality for the remaining workers within the firm. But that would not tackle the root causes—we need more male nurses and female doctors. For the World Bank Group’s most ambitious look at how it pays its staff, we developed a new methodology that decomposes what looks like a large pay gap at the institution, highlighting the complex issues that firms will have to grapple with when they embark on similar analyses.

Understanding the problem

To fix ideas, consider a firm with two types of jobs: Type A jobs are paid $1,000 a year and Type B jobs are paid $2,000 a year. If the firm hires the same number of women and men in each job, promotes them at the same rate, and employees leave from each pool in the same manner, there will be no pay disparities at the firm.

Consequently, pay disparities within a firm can arise from one of four distinct reasons.

  1. Gender composition of hires: There may be more women hired in Type A relative to Type B jobs.
  2. Entry salaries: The entry salaries for women in Type A (or Type B) jobs may be different from that for men. Perhaps the women in Type A jobs are paid $800 and the men $1,200 when they join; recent research shows that men are more likely to negotiate when joining a new job, leading to higher entry salaries.
  3. Salary growth: Women may be promoted from Type A to Type B jobs at a lower rate.
  4. Attrition: Women from Type B jobs may be more likely to leave, relative to men in Type B jobs and women in Type A jobs.

Furthermore, every firm has a pipeline that introduces a temporal link between hiring yesterday and hiring today. In our example, suppose that most Type B employees are hired from within the firm, which takes 10 years to promote an employee from Type A to Type B. Then, even if hiring today is entirely fair, if there were more men hired as Type A employees yesterday, there will be more men in Type B jobs today—introducing a pay disparity inherited from historical hiring patterns.

The solution to each of these disparities is different—for instance, differential attrition requires the firm to retain talented women in high paying jobs, while disparities due to employee composition require recruiting women into top jobs.

A dynamic decomposition methodology

To evaluate the gender pay-gap, we developed a new dynamic decomposition method. We started with data on all World Bank Group (WBG) employees from 1987 to 2015, with anonymized information on salaries, job grade, and promotions. Taking 1987 as our starting year, we then predicted salaries in the next year following an (estimated) salary generating process, subtracted employees who left our “simulated WBG,” and added in new hires. Doing this year by year allowed us to predict the 2015 distribution of salaries, which was remarkably close to what we observed in the real data. This salary simulator then allowed us to shut down each source of pay disparities to decompose the gap we see today into its constituent parts.


At first glance, the WBG has a huge pay disparity problem. Women in the institution in 2015 earned 77 percent of what men earned, and though this is much better than the 52 percent we observed in 1987, it is far from ideal. But there have also been tremendous changes at the WBG since 1987 and Figures 1, 2, and 3 highlight how each of the dynamics outlined above affects our interpretation of this gap.

In 1987 (Figure 1), most women were hired into administrative positions (known as GA-GD positions) and most men into better paid technical and management positions (GE+). Since then, the fraction of women hired into GE+ positions increased quite rapidly (it is close to parity now), but the fraction of men hired into administrative positions barely budged. Even with parity in GE+ hiring today, historically low female hiring will translate into fewer women in higher positions today to the extent that such positions are driven by promotions among staff.

Figure 1: The fraction of new female entrants has risen among GE+ staff, but has declined slowly among GA-GDGlobal_figure 1_WBG pay

Note: Figure 1 shows the proportion of women among employees hired at different grade-levels over time. Gender equity has improved significantly as the fractions of men and women hired at different grade-levels has converged over time. Nevertheless, the fraction of women hired into lower grades remains high—primarily due to their over-representation in the GA to GD grades.

In contrast to the job composition differences, gaps in entry salaries and salary growth differ according to the grades that people are hired into. Figure 2 presents the difference in the salaries of women relative to men through their careers; to emphasize the differences we have “detrended” the salary of men who are represented by the flat blue line. We have also presented similar data for the “Part 1 ” and “Part 2” employees—think of these as roughly the countries that finance the WBG (Part 1) and those that the WBG lends to (Part 2). The career experience of staff depend on what grade they entered the WBG. For instance, after 10 years at the WBG, all staff who entered as Grade GF earn less than Part 1 men, although differences in entry salaries are small. For staff entering Grade GG, there are differences in entry salaries, but these differences do not compound over time—salary growth is similar for men and women.

Figure 2: The career experience of staff depends on entry grade

Global_figure 2_WBG pay
(Click image to enlarge)

Figure 3 shows that attrition at the WBG can be quite high, rising when new presidents come in (and often undertake a reorganization), and declining after they have been around a bit (presumably, the reform loses steam). If the women who leave are systematically different from the men who leave, the salaries of those who remain will show a gender gap, even if there are no differences in entry salaries or promotions.

Figure 3: Attrition at the WBG cycles, rising with new presidents and declining after a whileGlobal_figure 3_ staff who left WBG

Putting it all together in the dynamic decomposition setup (Figure 4) yields the startling conclusion that 76 percent of the existing pay disparity between men and women at the WBG is due to current and historical differences in job composition. Of the remaining 24 percent, 5.4 percent is due to salary growth and 7.3 percent due to entry salaries, with attrition playing a negligible role. The disparity in salary growth and entry salaries should not be trivialized, as they still amount to $3,400 each year. At the same time, fixing pay disparity at the WBG has less to do with salary negotiation or differential promotion—the majority is because women were historically not hired into higher paid jobs. This is something that has changed quite dramatically over the last 20 years and will continue to bring about a decline in pay disparities as the number of women in management positions increases. What will be harder is bringing more men into administrative positions—without that, the pay disparities will decline, but not to zero.

Figure 4: Decomposing the gender gap at the WBGGlobal_figure 4 decomposing WBG gender gap

Further thoughts On gender pay inequality

Small differences in entry salaries and salary growth are not surprising if the salary scale is flat with little room for performance incentives and decentralized negotiation. That is, there may be a tradeoff between rewarding performance and allowing conscious and unconscious biases to operate. To our surprise, we found considerable variation and a strong link between sustained performance and wages at the WBG. We also found little difference in performance ratings bestowed upon men and women. This means that pay disparities due to salary growth and entry salaries are low despite a culture of rewarding performance. Our research does not discuss how this was achieved, but we suspect that it speaks to the strength of multiple institutions within the WBG, including the staff association, the human resource unit, and a strong group working on diversity and inclusion. Combined with a reliance on empirical evidence to guide important decisions, the WBG is coming closer to attaining pay equity for its employees.

The broader message is that how firms report pay disparities matters. Employees and the broader public need to understand the sources of pay disparities that allow them to tailor solutions to the problems. Although the pay gap across men and women has been an area of fervent study at the level of the economy, it is only now that we are starting to look within firms and holding management responsible for the disparities that are becoming evident. This transition requires a parallel shift in the analytical tools that were developed for economy-wide analysis. We need to firm up C-suite analytics for pay equality.