This study use an administrative dataset of annual pay rates of US DHHS employees, coupled with use of the SSA first name database and a database of the State of Iowa employee salaries, to assess the gender pay gap among the federal governmental public health workforce. We find a narrowing but persisting gender pay gap, after controlling for location, job title, pay plan, and job grade. Compared to the raw and adjusted pay gap of 11% and 4% in 2007 identified by the GAO for the federal workforce, the gender pay gap among DHHS employees has narrowed significantly, with raw and adjusted pay gaps at 8.3% and 1%. The estimated trend of a narrowing gender pay gap observed in this study is consistent with earlier research examining different components of the federal workforce [11,12,13] and the trend across the high-income countries [5]. Even though this study and two earlier publications have examined different parts of the federal government, the results on the female-to-male pay ratio over time reveal a consistent trend of a narrowing gender pay gap (Fig. 1).
Our analysis of the data from FedsDataCenter.com provides a close approximation of the DHHS administrative record. Our estimated proportion of females in the DHHS workforce is slightly less than 60%, which is consistent with the proportion reported in earlier studies using administrative data in 2006 [19] or employee survey data in 2004 [30]. This may indicate a willingness of female workers to self-select into health-related fields [7]. The estimated gender pay gap (a female-to-male pay ratio of 91.8% in 2018) is similar to what is published on Fedscope.com, which shows an average pay of $98,409 for female and $106,723 for male among DHHS employees in 2018, resulting in a female-to-male pay ratio of 92% [31].
The narrowing gender pay gap can be explained by the increase in education and experience as well as an increased entry of females into occupations that had been dominated by males [1, 11,12,13, 19, 32]. The GAO reported has identified a contributing factor that may be both encouraging and concerning—the clerical positions that were dominated by female employees have been diminishing in the federal government [13]. The diminishing clerical positions is encouraging because it may indicate that technological and social changes have enabled and empowered females to enter a broad array of fields. However, the federal government's use of contractors may lead to more females being hired as contractors, whose pay will not appear in the database thus unaccounted for in the calculation of gender pay gaps [33].
Grade, or the position in the bureaucratic hierarchy, is an essential element of pay for federal employees. The GAO report, though acknowledged the narrowing gender pay gap among the federal workforce, has also pointed to an important concern—the portion of the gender pay gap that cannot be explained by occupation, experience, and education increased from 2 cents in 1988 to 7 cents in 2007. Our study suggests that occupation and pay plan explained about 8 cents (the difference between − 0.053 and − 0.131) of the gender pay gap in 2010, and 5 cents in 2018. Including the pay grade explains an additional 2.5 cents in 2010 and 2018. However, because pay grade is a proxy of promotion, it may have components that can be explained by education, years of government service, as well as an unexplained portion. Similar to what has been observed among the employees of the Mexican National Institutes of Health, [7] the distribution of female employees across different grades of the GS has been uneven, with the smallest portion of female employees reaching the highest grade, GS-15, and higher percentages of female employees at lower grade levels (e.g., GS-13). Given a grade level, females earn slightly more than males at lower grade levels, but the advantage starts to disappear at grade level 13 (Table 2). This, coupled with the higher female average pay, may indicate that female employees tend to stay longer on the higher end of the 10-step band of lower grade levels, being passed on promotion opportunities. To summarize, the unexplained portion of the gender pay gap may lie within the interval determined by models 2 and 3, i.e., between 1.8 and 5.3 cents in 2010 and between 1.0 and 3.5 cents in 2018.
Competing theories exist in explaining the unexplained portion of the gender pay gap, e.g., childcare and wage structure [1, 15, 16], gender difference in psychological attitude [18], unpaid overtime [8], and the lack of role models [19]. Employee viewpoint survey or focus group interviews may help to pinpoint the exact causes of the remaining gender pay gaps.
Policies and interventions to reduce the gender pay gap include a one-time pay raise for female employees as the University of Essex did [34], and the Denmark and UK mandates for institutions to publish gender pay gaps. Such mandates have been shown to have narrowed the gender pay gap in the UK and Denmark [35,36,37]. Our results provide support for the proposed Pay Check Fairness Act (H.R.7, 117th US Congress) to improve transparency, to protect female employees’ right to challenge pay discrimination, and to hold employers accountable.
Our study does send a message to the international community: a high-income, well-developed civil society may not be immune from gender pay disparities. Continued efforts to eliminate gender disparities in pay, promotion, and leadership roles are needed anywhere and everywhere.
Limitations
This study has several limitations. First, FOIA requests are handled within a set timeframe, often at the price of quality control. Agency might also have different policies as to what information to be withheld over time. We have seen the variation in our sample size, which dipped after 2014. The direction of the impact of such variation on our estimated gender pay gap is unclear. However, if the quality control or information withheld does not vary systematically across the gender groups, our results will not be critically impacted, which seems to be the case as we have seen the consistent trend in the estimated gender pay gap. Second, we have more than 10% of the first names that cannot be assigned a gender. Some of the names are gender-neutral names, such as "Drew" or "Robin", that neither gender had been used at least 95% of the time during 1940–2000. Though individuals with those names are slightly more likely to be female, as shown in the SSA data, thus our estimated gender pay gap may bias upward as those who cannot be assigned a gender have a higher average salary. Some of the other names that cannot be assigned gender are from foreign languages, which is difficult to determine the direction of the biases from dropping those names. However, the resulted bias may be negligible because the Fedscopes.com data provide strong support for our results. Third, the job titles provided through FOIA requests are often unstructured texts with many different forms of abbreviations and variations. This leads to issues where multiple fixed effects are estimated for the same occupation. However, the additional degree of freedom is trivial and should not impact our point estimates, given the large sample size of this study. Fourth, we do not have information on experience, education, and race and ethnicity. Meanwhile, pay grade may capture some of the differences in experience and education because specific requirements on education and experience exist for progressing to a particular pay grade. However, we acknowledge that our estimates may not reveal the important pattern of gender pay gaps by race and ethnicity due to the data limitation.