Association of a Simulated Institutional Gender Equity Initiative With Gender-Based Disparities in Medical School Faculty Salaries and Promotions

Key Points Question What is the true magnitude of differences in salary, time to promotion, and accumulated wealth between male and female faculty, and how are these measures associated with a set of institutional gender equity initiatives? Findings In this quality improvement simulation study of salaries of 1481 faculty, the salary gap and time to promotion decreased after implementation of gender equity initiatives; however, small persistent differences were still associated with substantial disparities. Meaning Residual gender-based salary gaps may lead to substantial differences in accumulated wealth over the career course and into retirement, but institution-wide promotion of equity initiatives can slowly begin to narrow the disparity.


Introduction
2][3][4] From the available reports, compensation for men exceeds that for women by approximately 7% to 8% among US physician researchers, 2,3 nearly 20% among Japanese surgeons, 5 and 25% to 40% between scientists in public sectors in Europe 6,7 even after controlling for other factors such as years of experience and rank.
With increasing attention on such inequities faced by female scientists, the Johns Hopkins University School of Medicine (JHUSOM) reviewed institutional data and noted that there was a lag in promotion of women to full professorship despite the presence of a significant proportion of women at the lower faculty ranks.In response, the JHUSOM created the Committee on Faculty Development and Gender in 2002.The committee first investigated barriers to career promotion by (1) reviewing faculty representation, attrition, and promotion rates for women; (2) conducting faculty surveys and departmental director interviews to identify sources of differences in career progression; and (3) performing annual faculty salary analyses (FSAs) to track salary inequities.The committee then offered specific recommendations, summarized in the Box.
An oversight committee was created in 2006 to supervise implementation of these recommendations.In particular, leadership used annual FSAs to target remediation actions toward departments with higher salary gaps.Committees created templates that standardized criteria for promotion across departments.Financial and human resources were allotted for the recruitment and retention of women faculty.The Office of Women in Science and Medicine was created to increase pathways to leadership, and more than 600 women have participated in its programming to date.To increase transparency of efforts, JHUSOM continues to publish annual FSAs and interval reports on the status of women.
Collectively, these efforts-referred to as gender equity initiatives (GEI)-have narrowed salary and promotional gaps at the JHUSOM.However, a small overall gender-based salary difference of 1.9% persists.While our institutional goal is to achieve 0% difference by the 2018 fiscal year, others may argue that such single-digit residual gaps are relatively insignificant, thus negating the need for additional intensive measures to eliminate them completely.Yet when accounting for the cumulative effect of salary inequity on retirement and investment savings, compounded with issues such as increased time to promotion (TTP) and longer life expectancy for women, the total impact of even small gaps may have profound implications for lifetime earning potential and retirement for women.
To better illustrate the true costs of salary inequities, we have simulated the total disparity in lifetime wealth created by residual gender-based salary differences over the course of a career and into retirement.In doing so, we highlight the impact of more than a decade of work at JHUSOM and offer our GEI efforts as a potential framework for use across institutions with the goal of eliminating this gap.

Source Data
Annual FSAs were conducted for all JHUSOM faculty members for fiscal years 2005 to 2016, including information on department, self-reported gender, degree, rank, years in rank, and fiscal year salary.Data for both MD and non-MD faculty were considered in our evaluation.For the purposes of our analysis, annual salary for a given fiscal year included full-time equivalent base salary and supplemental salary for administrative, educational, or clinical roles.Bonus salary data were not available.All salary values were reported in US dollars.This study met the criteria for institutional review board review waiver.The authors followed Standards for Quality Improvement Reporting Excellence (SQUIRE) 2.0 reporting guidelines for reporting quality improvement.

Key Recommendations
Elevate the goal of gender equity to an essential mission of the institution Achieve salary equity through continued close review of faculty salary analyses, with leadership directing actions toward departments with salary disparities in an effort to narrow the gap Discourage meetings and conferences from being scheduled outside of normal business hours to promote inclusion of women Dedicate financial resources toward the recruitment and retention of women to senior faculty ranks Expand sexual harassment educational curriculum Interview departing women faculty to identify root causes for attrition Create a standing oversight committee to monitor issues related to gender equity Conduct a faculty survey in 3 years to assess progress Deidentified information from annual FSAs was analyzed by the Johns Hopkins Bloomberg School of Public Health and published internally to the JHUSOM community.Each FSA reported unadjusted mean annual salary by gender, subcategorized by faculty rank (assistant professor, associate professor, or professor) and degree (MD vs non-MD).Annual FSA reports also estimated gender-based difference in mean annual salary using multivariable linear regressions.These regressions applied a logarithmic transformation to salary to minimize the impact of outliers and adjusted for gender, department, department-specific rank, degree, and years in rank.As per model specifications, the resultant regression coefficients for gender corresponded to the estimated percentage difference in salary comparing women with otherwise similar men, with negative values indicating women earning less than men.Regression models available in the FSA included the overall gender difference and the rank-specific gender difference in salary.
Median TTP by gender was obtained from registrar records for 2 cohorts of faculty hired at the rank of assistant professor or associate professor.The pre-GEI era cohort reflects TTP for faculty hired between 1989 and 1990 (prior to GEI implementation) and was examined in November 2005.
The GEI era cohort was hired between 2005 and 2017 and was examined in January 2018.

Outcome Measures and Simulated Scenarios
Simulations were performed to model the association of gender-based differences in salary and TTP with estimated annual salary and additional accumulated wealth (AAW) over the course of a 30-year career for otherwise identical representative male and female full-time faculty members.In the simulation, the faculty was hired as an assistant professor and promoted to associate professor and then professor.Corresponding salary and TTP were estimated using gender-and rank-specific mean and median values for these variables from annual FSA and registrar cohort data, respectively.To demonstrate the impact of institutional efforts to decrease the compensation gap on AAW, we simulated 3 scenarios: 1. Pre-GEI: Gender-based differences in salary over the career course were estimated using the 2005 fiscal year FSA.Gender differences in TTP at each promotion were obtained using pre-GEI era cohort data.As such, this scenario simulates gender differences for lifetime AAW for a faculty member hired in pre-GEI conditions and for whom the pre-GEI salary and TTP inequities remained constant throughout the career course.
2. Post-GEI: Gender differences in salary over the career course were estimated using the 2016 fiscal year FSA.Gender differences in TTP were obtained using GEI era cohort data.As such, this scenario reflects the conditions faced by a faculty member hired in post-GEI conditions and thus subject to narrowed salary and TTP gaps throughout the career course.
3. Real-time GEI: Data from the 2005 fiscal year FSA were used to simulate baseline gender differences in salary for an assistant professor hired at that time.Gender differences in TTP were estimated using GEI era cohort data.At the time of promotion from assistant professor to associate professor (in x years for the man and y years for the woman) and from associate professor to professor (in x′ years for the man and y′ years for the woman), an increase in the gender-specific salary reflecting values reported in the corresponding year's FSA were applied as follows: baseline salary + (x, x′, y, y′).Thus, this scenario simulates the potential impact of salary and promotional differences for a faculty member hired in pre-GEI conditions but subject to the real-time efforts of the GEI.
To assess the differential effects of gender inequities by promotional rank, a secondary set of simulations were applied to each of the 3 scenarios described.In these simulations, the representative faculty members' careers took the following paths: (1) no promotion beyond assistant professor, (2) promotion only to associate professor, or (3) promotion through full professor (equivalent to the assumptions of the scenarios described).
For each scenario, we created a projection of both annual salary and AAW based on gender differences in salaries and TTP over the course of a 30-year career.

Annual Salary
In all 3 scenarios, the starting salary of the representative male assistant professor was set at $108 750, which is the average of the mean salaries for a male MD and male non-MD assistant professor from the 2005 FSA.To ensure that comparisons were made between otherwise similar men and women, we calculated the corresponding female assistant professor's salary using the rankspecific gender difference in salary for a given scenario, as estimated by the FSA's adjusted linear regression model.This approach was again repeated at each appropriate time for promotion to associate professor and to professor.Simulations of salary before and after the GEI assumed a 3% annual increase for cost of living for each year at a given rank until promotion. 8

Additional Accumulated Wealth
Additional accumulated wealth included estimated retirement savings and other salary-based investments accrued over time.A standard employer contribution to the pretax retirement account was set at 12% of gross income, reflecting our institutional practice.Each simulation included a pretax retirement account deposited with maximal employee contributions throughout the 30-year career.Initial maximum contributions were indexed to an inflation rate of 2% 9 and set at $18 000 through age 50 as per US federal regulations, with an additional 33% catch-up contribution after age 50.Take-home pay was estimated as monies received from salary after employee pretax retirement contributions and deduction of marginal tax rate of 33.45% (8% local, 24% US federal income, and 1.45% Medicare).The simulation assumed equivalent standard of living-and thus equivalent annual expenses-for both genders.For the simulation, we set the annual expenses equal to the lowest take-home pay between the genders.Any take-home pay in excess of annual expenses for a given gender was assumed to be invested in a long-term diversified portfolio.
The equity and bond composition used for retirement and salary-based investments is described in the eFigure in the Supplement.Expected rate of investment returns were 7% for equity markets and 3% for bond markets, based on an assumed risk-free rate of 2%, equity risk premium of 5%, and bond risk premium of 1%. 10,11

Postretirement Wealth
Using the real-time GEI scenario, we calculated postretirement AAW and income, starting with total AAW accrued at the time of simulated retirement by gender.Additional contribution of retirement savings to gross postretirement income was calculated based on a graduated annuity with a 20/80 equity/bond investment portfolio return and 2% inflation-adjusted income growth throughout retirement.We used US Social Security Life Tables 12 to estimate life expectancy and anticipate years in retirement by gender.Differences in postretirement wealth available for use over the course of remaining life were compared between genders.

Statistical Considerations
In addition to reporting outcomes for the JHUSOM, a goal of our study was to create an editable version of the simulation that external users could modify to reflect conditions faced at their own institutions.In light of this aim, our methodology was selected to optimize generalizability and use types of data most likely to be accessible.However, our ability to accurately estimate measures of variance or perform tests of significance for our own institutional data was compromised in order to meet this greater goal of generalizability.
For example, to protect faculty privacy, the annual FSA information analyzed at JHUSOM is not published at an individual level; instead, it is best considered as repeated cross-sectional data.We anticipated this to be the most likely level of available data for external users as well.However, the deidentified nature of the data prohibited us from direct longitudinal analysis of career trajectories for specific faculty over time.Moreover, subgroups used to report aggregate measures of central tendency, regression coefficients, and variance were not uniform within or across FSA reports.To enhance comparability, some subgroups (such as MD and non-MD faculty) were combined.Given these limitations, we determined that we could not accurately specify the large number of assumptions needed in order to capture variance and perform hypothesis testing for outcomes like AAW by gender.Microsimulation methods were considered but ultimately not used because the detailed procedures required to calibrate such simulations to our population would potentially compromise applicability to external users.
As per FSA analyses, the adjusted regression coefficients and standard errors reflecting the estimated overall and rank-specific gender difference in salary were reported, and 95% confidence intervals were specified.
Simulation projections were developed and displayed using Microsoft Excel 2016 and Shiny version 1.0.5 (RStudio).

Results
A total of 1481 faculty (432 women) in 2005 and 1885 faculty (742 women) in 2016 completed the FSA.Table 1 shows participant characteristics from the pre-and post-GEI cohorts.In 2005, 31% of full-time faculty were women, as compared with 39% in 2016.In an adjusted analysis, women's salaries were estimated to be a mean (SE) of −2.6% (1.2%) (95% CI, −5.6% to −0.3%) lower than

Simulated Association of Gender With Wealth in 3 Scenarios
Figure 1 demonstrates annual salary and AAW over the course of a 30-year career for an equivalent newly hired woman vs man in the pre-GEI, post-GEI, and real-time GEI scenarios.Salary and TTP data used for each scenario are outlined in Table 1.
Pre-GEI conditions (Figure 1A) resulted in an additional $501 416 AAW collected by the male faculty.This was composed of $35 000 from growth of employer retirement contributions and $466 415 from salary-based investments.Post-GEI (Figure 1B), the male faculty collected an additional $66 104 AAW, composed of $4831 from growth of employer retirement contributions and $61 273 from salary-based investments.Narrowed gaps in both TTP and compensation at associate professor and professor ranks in the post-GEI scenario corresponded to a reduction in the pre-GEI gender difference in AAW by 87%.
Real-time GEI simulation (Figure 1C) revealed the AAW gap was reduced to $210 829, composed of $15 883 from growth of employer retirement contributions and $194 946 from salary-based investments.The real-time GEI scenario was associated with a reduced pre-GEI gender difference in AAW by 58%.

Secondary Simulations for Differential Association by Rank
Figure 2 demonstrates secondary simulations performed to assess for differential effects of gender inequities by academic rank.Differences in AAW were consistent across all 3 scenarios for faculty remaining at the assistant professor rank throughout their careers.This is attributable to a starting salary gap of 2.9% to 3.4% across scenarios for this rank.Post-GEI gender differences in AAW were minimal for faculty promoted to associate professor or professor.For the real-time GEI scenario, Simulations reflect pre-gender equity initiatives (GEI) (A), post-GEI (B), and real-time (C) scenarios.Whereas gender differences in annual salary are small across time in all 3 scenarios, the cumulative effect of salary and promotion disparities results in a significant difference in additional accumulated wealth (AAW) in the pre-GEI and realtime scenarios of $501 416 and $210 829, respectively.The AAW narrows to $66 104 using post-GEI conditions, reflecting success of GEI efforts.persistent salary gaps correlated with larger gender differences in AAW compared with those in the post-GEI era despite similarly narrowed TTP gaps.

Association of Life Span With Retirement Income
Life expectancy in the United States for a woman now aged 35 years is 82.25 years, as compared with 78.22 years for a man aged 35 years.Thus, a woman can expect to spend 17.25 years in retirement past age 65 years as compared with 13.22 years for her male counterpart.To account for lower AAW and longer life expectancy, a woman would have to spend $0.60 for every $1 spent by her male counterpart in order for her resources to last through retirement (Figure 3A) in the pre-GEI scenario.
For the real-time and post-GEI scenarios, this corresponds to women spending $0.68 to $0.73, respectively, for every $1 spent by their male counterparts.Were a woman to spend her retirement funds at the same rate as her male counterpart, the woman would be expected to deplete her Variations of pre-gender equity initiatives (GEI), post-GEI, and real-time scenarios to simulate the difference in accumulated wealth (AAW) by gender that may be experienced had both the male and female faculty members taken the following 30-year career paths: (1) no promotion beyond assistant professor, (2) promotion only to associate professor, or as simulated in the manner described in the text, (3) progressing through the ranks to full professor.The magnitude of differences in AAW is smallest in the post-GEI scenario due to significant narrowing of gender-based salary and promotional gaps.Residual AAW differences are most pronounced when the male and female faculty are not promoted beyond assistant professor due to baseline salary gaps of 2.9% to 3.4% across scenarios.

Men Women
In this simulation of postretirement wealth and income by gender over the course of retirement years for the representative woman as compared with her male counterpart in the pre-gender equity initiatives (GEI), time from retirement to death was 17.25 years for the woman and 13.22 years for the man.Panel A shows the rate of spending required for each gender to use all postretirement wealth and income prior to death.Owing to the cumulative effect of salary and promotional disparities over her career course, a female faculty would be required spend her retirement savings at a rate 40% slower than her male counterpart to compensate for less retirement savings and longer life expectancy.Panel B demonstrates both genders spending postretirement wealth and income at an identical rate (set to the rate used by male faculty in A).Were she to spend her retirement wealth at the same rate as her male counterpart, her resources would be depleted 7 years prior to her death (represented by deficit of funds below the zero line).
their own institutional differences in salary by gender or other demographic characteristics of interest.
A second limitation is that we performed analyses with adjusted A+B data, as base salary (part A) was not reported separately.This could confound gender-based differences in salary if women are less likely to participate in or be compensated for administrative, educational, or clinical roles (part B).Moreover, we did not include bonus salary (part C) in our assessment.To address this, data on each separate component of salary will be acquired individually in future FSAs.
Our salary data were also averaged across MD vs non-MD faculty and adjusted for department.
Because faculty who are not MDs or who work in women-predominant specialties are often paid less, 2,19 our simulations may not fully characterize the association of gender inequities with wealth for these subgroups.However, review of unadjusted average salaries of non-MD faculty suggests a decrease in salary gaps over time that parallels implementation of GEI.In the pre-GEI era, unadjusted percentage differences in A+B salary for women vs men were −9.2%, −6.7%, and −7.5% for assistant professor, associate professor, and professor non-MD faculty, respectively.By 2016, these gaps changed to −3.8%, −7.5%, and −6.4%, respectively.Additionally, it is noted that available data were not stratified by race/ethnicity.Women of all major underrepresented minority groups in the United States earn less than white women, and women in general earn less than men of the same ethnic or racial group. 20Thus, the association of gender-based disparities with wealth might be compounded by race/ethnicity. 21This remains an underreported subject, and future studies must be aimed at understanding the added difficulties faced by women in science who are also members of underrepresented minority groups.
Although we have demonstrated that the effects of workplace gender disparities can persist past retirement, our analysis assumed that the representative male and female faculty began their careers equally without debt.In the United States, the median total debt among medical school graduates was $185 000 in 2015. 22Because repayment of educational debt generally occurs during the first 10 to 20 years of employment, this coincides with the time of greatest disparity between the salaries of women and men.Thus, a female faculty with existing debt may find herself less likely than her male counterpart to make maximum retirement contributions or other investments during repayment.As demonstrated in our simulations, these early gender-based differences could be expected to magnify with time.Again, the association may be even greater for women in underrepresented minority groups, whose educational debt is often higher than that of their white counterparts. 22Because educational debt is associated with lower job satisfaction 23 and physician burnout, 24,25 it may contribute to attrition rates of women in science.

Conclusions
We have demonstrated the association of a program to support women faculty with substantially narrowed gender-based salary and TTP differences.Future challenges include maintaining momentum to address small residual gaps and extending research to include other components affecting wealth such as baseline debt and differences by race and ethnicity.

Figure 1 .
Figure 1.Simulations of the Association of Gender-Based Inequities With Compensation and Promotion From Assistant to Full Professor

Figure 2 .
Figure 2. Simulations of the Association of Gender-Based Inequities With Compensation and Promotion in Different Career Paths

Figure 3 .
Figure 3. Simulated Postretirement Wealth and Income by Gender

Table 1 .
Descriptive Statistics of Faculty Cohorts Included in Pre-GEI and Post-GEI and Variables Used in Each of the 3 Simulation Scenarios 's salaries in 2005 across rank, with the overall gap narrowing to −1.9% (1.1%) (95% CI, −4.1% to 0.3%) in 2016.Salary differences generally narrowed with increasing academic rank.
f Abbreviations: FSA, faculty salary analysis; GEI, gender equity initiatives; NA, not applicable.aData from the 2005 FSA.b Negative values indicate that women earned less than men.cData from the 2016 FSA.dThe 2005 FSA mean salary for men at a designated rank, adjusted for a 3% annual increase to account for cost of living per year corresponding to time to male promotion.eData from 2009 FSA.fData from the 2014 FSA.men