Previous studies have suggested that male physicians earn higher salaries than their female counterparts, but the mechanisms underlying much of this difference remain poorly understood.1–9 Differences in the distribution of men and women into different specialties, work hours, and productivity have explained some of the observed difference; however, prior studies have indicated that a substantial proportion of the difference remains unexplained even after those variables are taken into account.1
In prior work, our group documented an unexplained gender difference in salary even within a relatively homogeneous population of midcareer physician–researchers.1 Given the extensive list of potential factors for which we controlled, including specialty, work hours, and productivity, we speculated that the difference observed might be rooted in gender-related differences in values or behaviors. For example, men might prioritize compensation more highly, either because of prevailing societal expectations of gender roles or the greater likelihood of a man serving as the sole breadwinner in a family. Similarly, men might negotiate more aggressively for salary. Employer attitudes may also play a role. Employers might value men’s contributions more than women’s. Alternatively, employers might view men as needing higher salaries (because of the notion of a “family wage”)10 if men are less likely to be in two-income households. However, like others, our prior work was limited by lack of information on the employment of the respondent’s spouse, precluding the ability to ascertain whether some of the gender effect on salary may have been mediated by spousal employment. Moreover, the sample of physician–researchers we previously considered had commenced their academic careers over a decade ago, and recent efforts to decrease inequities may have been successful with younger cohorts.
Therefore, we sought to evaluate gender differences in salary in a new population of physician–researchers who were similarly select and homogeneous, but who were early in their careers: physicians who received K08 and K23 awards (i.e., prestigious National Institutes of Health [NIH] mentored career development grants) between 2006 and 2009. We sought to evaluate whether the gender differences we previously observed in a midcareer population of elite physician–researchers would be apparent in this younger and more recently hired cohort at this earlier point in their career trajectories. In addition, we included questions eliciting spousal employment status (full-time, part-time, or not employed) and the perceived level of dependence of the family unit on the respondent for financial support in order to determine how much of any observed gender difference in salary might be mediated by spousal employment and gender roles within the family.
In 2010, using the NIH RePORTER database,11 we identified 1,719 researchers who received new K08 and K23 awards in 2006 through 2009. After receiving approval from the University of Michigan institutional review board (IRB), we conducted Internet searches and made telephone calls to obtain the current U.S. mailing addresses of these K award recipients. We obtained 1,708 valid U.S. mailing addresses (see also Figure 1).
Between August 2010 and February 2011, we sent a survey questionnaire and a $50 incentive to all 1,708 of these K award recipients. We enclosed a cover letter that explained that this was an IRB-approved research study investigating the experiences of individuals who received K08 and K23 awards from the NIH. The cover letter stated the voluntary nature of participation, our efforts to ensure confidentiality, the minimal risks involved (e.g., possible loss of confidentiality), and our source of funding for the study. It also included contact information for the IRB and the principal investigator. Following a modified Dillman approach (which employed an initial contact letter, a tailored questionnaire, and subsequent correspondence),12 we also sent a follow-up questionnaire to nonrespondents. On receipt of the completed questionnaires, we merged survey responses to data previously collected from RePORTER on K award type, year, and institution characteristics.
We designed the questionnaire after reviewing the relevant literature,1–9,12 considering instruments used in other research to determine outcomes of academic careers,3,13 and conducting detailed cognitive pretesting.14 Ultima tely, the questionnaire comprised 173 items that assessed demographics, education, time allocation, academic experiences, family responsibilities, and salary.
The principal dependent variable for the analysis was current annual salary, which we structured as a continuous variable rounded to the nearest thousand dollars. We also analyzed several independent variables as continuous variables: age, number of years on faculty, number of hours spent working (work hours), and percentage of time spent conducting research (research time).
As described in greater detail in previous studies,1,15 we grouped specialties into four categories based on their nature: (1) internal medicine and its subspecialties; (2) surgical specialties; (3) specialties related to the care of children, women, and families (family practice, obstetrics–gynecology, and pediatrics and its subspecialties); and (4) hospital-based specialties (emergency medicine, anesthesiology, pathology, and radiology). We also grouped specific specialties into four pay-level categories (low, medium, high, and extremely high) based on Association of American Medical Colleges data on median salary in that specialty in 2009, as described elsewhere.1 This additional grouping allowed for finer distinctions between subspecialties that are similar in nature but have different earning potential.
We grouped institutions such that all hospitals affiliated with a single university were considered to be a single institution. We then grouped the institutions employing the researchers into four tiers containing roughly equal numbers of K awardees, based on the amount of total NIH funding received (i.e., first tier = the institutions receiving the most NIH funding, and fourth tier = the institutions receiving the least NIH funding), as well as into categories for public or private.1,15 We grouped institution location into four categories based on region of country (Northeast, South, Midwest, and West).
We grouped the NIH institutes that funded respondents’ K awards (e.g., National Cancer Institute, National Institute for Mental Health) into three tiers of funding activity, based on the total dollar amount of R01 awards granted in 2000 (i.e., first tier = those granting the highest dollar amount of R01 grants, second tier = those in the middle, and third tier = those granting the least).1,15
We divided faculty as follows: by academic rank into five groups, by year of K award into four groups, by race (as self-reported in multiple-choice questions) into four groups, and by marital status into three groups (married or in domestic partnership, single, or divorced/widowed). We grouped spousal employment status into three categories (full-time, part-time, and not working). K award type, parental status, and possession of an additional PhD degree were binary variables, as was gender.
We also asked respondents how much their compensation depends on clinical volume or number of patients seen, as well as how much their compensation depends on amount of grant funding received. Another item asked respondents, “How dependent is your family upon your income to maintain an acceptable lifestyle?” We scored all of these items on a four-point response scale ranging from “not at all” to “very much.”
We performed statistical analyses using the SAS System, version 9.2 (SAS Institute Inc., Cary, North Carolina). We compared respondents with nonrespondents by those characteristics for which public data were available so as to evaluate for potential bias related to nonresponse. After comparing those who reported their salary with those who did not, we limited our sample to individuals holding MD degrees with academic positions in clinical specialties who reported their salary.
We described characteristics of this sample by gender and then constructed multiple variable linear regression models for salary. We began with the following respondent characteristics: gender, age, race, marital status, parental status, additional PhD degree, academic rank, number of years on faculty, specialty, specialty pay level, current institution type (public or private), current institution region, current institution NIH funding tier, K award type, K award funding institute tier, K award year, work hours, and research time. Most characteristics were categorical and modeled as indicator variables with a reference category. We centered continuous characteristics (e.g., age, work hours) at their means. We constructed both a full model using all covariates and a parsimonious model whereby we iteratively deleted variables from the model based on improvement in Akaike’s information criterion,16 using both forward stepwise and backward elimination approaches. We also explored pairwise interactions between gender and the other characteristics. These multivariable models offer estimates of the association between gender and salary, independent of the other variables included.
To explore the explanatory value of spousal employment within the married or partnered subset of our sample, we used the Peters–Belson approach. This approach allows for the decomposition of an observed gender difference in salary into two components: the component that is explained by gender differences in other measured characteristics, and the component that remains unexplained. Specifically, we developed a regression model using all measured characteristics for the men alone. We then applied the coefficients from that model to the characteristics for each woman to derive her expected salary, as if her gender were male, in order to quantify the proportion of the observed gender difference unexplained by the measured characteristics.17–21 We first conducted this exercise in the married/partnered subset without including spousal employment status and then repeated it after including spousal employment status, to measure the explanatory impact of that variable.
For statistical inference, we conducted two-tailed tests with test statistics, considering P values at or below 5% to be significant.
We received 1,275 completed questionnaires from the 1,708 individuals we contacted, for a response rate of 75% (see Figure 1). Our respondent sample did not differ significantly from nonrespondents by gender or K award year. A higher proportion of K23 recipients (645/831; 78%) responded than did K08 recipients (630/888 [71%], P = .002). Individuals at institutions with lower overall NIH funding were more likely to respond (322/401 [80%] from the lowest/fourth tier, 349/474 [74%] from the third tier, 353/486 [73%] from the second tier, and 236/340 [69%] from the top/first tier, P = .006). Of the 1,275 respondents, 1,055 (83%) held MD degrees—and of these, 1,046 (99%) held academic positions in clinical specialties. Finally, of these 1,046, we used the 1,012 (97%) who reported salary information to constitute the analytic sample.
The characteristics of the 419 female and 593 male K award recipients in the analyzed sample are detailed in Table 1. Women were more likely to be single (8.6% versus 4.2%, P = .01). Of those who were married/partnered, men were far more likely to have a spouse who was not employed (26.5% versus 7.5%, P < .001) or employed part-time (28.0% versus 6.4%, P < .001). Women were nearly twice as likely to be in the lowest-paying specialties (45.1% versus 24.1%, P < .001), more likely to hold K23 (rather than K08) awards (58.2% versus 35.4%, P < .001), and more likely to be funded by NIH institutes that awarded lower amounts of independent funding (38.0% versus 25.3%, P < .001). Women’s mean work hours were lower than men’s (54.0 versus 59.4, P < .001).
Overall, mean salary was $141,325 (95% confidence interval [CI] $135,607–$147,043) for women and $172,164 (95% CI $167,357–$176,971) for men in this sample. Table 2 presents the results of our bivariate analysis on the correlates of salary (i.e., the personal, family, professional, K award, and institutional demographics described in Method).
Table 3 presents multivariate models of salary in the sample: a full model including all theoretically selected covariates and a parsimonious reduced model. The gender effect was similar in both models (+$10,921 for men in the full model and +$10,663 for men in the reduced model). Of note, we observed one statistically significant interaction between gender and a modeled covariate. This significant interaction was between gender and specialty pay level (P < .001), and the interaction remained significant when modeled simultaneously with the main effects in the model, revealing the gender difference in salary in this sample to be larger in the higher-paying specialties.
Specifically, as depicted in Figure 2, the mean salary for women in low-paying specialties (e.g., pediatrics, family medicine) was $123,678, whereas for men in these specialties, mean salary was $132,058. The mean salary for women in medium-paying specialties (e.g., neurology, pathology) was $146,651 versus $152,622 for men in these same medium-paying specialties. The mean salary for women in the high-paying specialties (e.g., emergency medicine, gastroenterology) was $165,114, and the mean salary for men in these specialties was $195,771. Finally, the mean salary for women in extremely high-paying specialties (e.g., neurosurgery, radiology) was $264,636, and the mean salary for men in these specialties was $298,915.
Table 4 describes respondents’ perceptions regarding their salaries. Respondents were more likely to indicate that compensation depended heavily on grant funding than on clinical volume (P < .001, Stuart–Maxwell test). There were no statistically significant gender differences in response to questions asking how much the respondent’s compensation depended on clinical volume or number of patients seen (P = .13) or on grant funding received (P = .41). However, men were more likely to report that their families were “very much” dependent on their incomes to maintain an acceptable lifestyle than were women (77.5% versus 54.1%), and the difference was significant (P = .002) even after adjusting for spouse/partner employment status.
Peters–Belson analysis in the married/partnered subsample revealed that women earned less than what would be expected if they retained their other measured characteristics but were men. When we excluded spousal employment status from the Peters–Belson analysis, 17% of the total observed gender difference was unexplained by the other measured characteristics. When we did include spousal employment status, the proportion of the total gender difference in salary that remained unexplained was 10%. Thus, inclusion of a measure of spousal employment status explained only about a third of the previously unexplained gender difference.
In this cohort of elite, early-career physician–researchers who only recently commenced their faculty careers, we observed a substantial gender difference in salary that was not fully explained by specialty, academic rank, work hours, or even spousal employment. These findings suggest that salary disparities in academic medicine exist even in cohorts hired recently and that these disparities arise early in the course of a career. In a previous study of midcareer K award recipients, we observed similar gender differences in all specialties,1 but the gender difference in this study primarily existed in the higher-paying specialties (Figure 2). Thus, the salary gap appears to develop early in the career trajectory, especially for women in those specialties.
Scholars have noted that gender differences in salary that exist early in a career are likely to widen over time, and that the initial salary negotiation may merit particular attention.22,23 Some evidence suggests that women negotiate salary less aggressively than men do.24–28 Other, related research indicates that female academic physicians may need to prepare in advance for a conversation regarding salary in order to feel more comfortable being assertive during the negotiation and more self-confident afterwards.29 Additional research shows that women are judged more harshly than men for initiating negotiations.30–33
Workshops in negotiation for women faculty are an increasingly common intervention that offices dedicated to the support of women at various academic institutions are pursuing.34–36 Given these current findings, such programs should consider expanding eligibility and outreach to ensure that female residents and fellows experience negotiation training prior to their first faculty appointments. Even with such training, however, new junior faculty are hardly in a position to ensure their own salary equity. Those doing the hiring and setting the salaries need to be sensitized both to the corrosive impact of salary inequity on faculty morale and to the importance of working to avoid even small inequities early in women’s careers, particularly given evidence that such inequities grow over time.37 To that end, bias literacy workshops and other systematic educational interventions targeting department chairs, division chiefs, and medical school administrators merit further development and investigation.
In this study, we found that about one-third of the gender difference in salary that was unexplained by other factors could be explained by spousal employment. An unconscious influence of gender-linked beliefs about the “family wage” has been proposed as a mechanism underlying gender differences in salaries, despite the very high rates of women’s labor force participation nationally.10 Our findings are consistent with this speculation; that is, the idea of the family wage may partially explain salary inequity among physician–researchers. Employers may feel that men who are supporting a family deserve a higher salary than women whom they do not view (and who may not view themselves) as principal breadwinners. Given the large differences in family composition between men and women physicians (i.e., women generally have partners who are employed full-time, whereas men generally do not), salary setting may possibly be influenced by extraprofessional assumptions about gender, rather than by actual credentials or performance. Unobserved differences in activity at work (e.g., working a schedule that is equal in number of hours but more convenient for family life) not adequately addressed by control variables for work hours and research time are also possible explanations, although this seems less likely, given our selection of a relatively homogeneous and research-intensive population of academic medical faculty for this study. Future research, particularly employing qualitative methods, is necessary to explore further whether some of the observed salary differences result from differences in unmeasured aspects of job flexibility. After all, women may be more willing than their male colleagues to trade salary for flexibility; likewise, men may not perceive themselves to be as closely monitored at work as women do and are therefore more able to harness job flexibility without trading salary.
Even with the inclusion of spousal employment status in our model, an unexplained gender difference remained. Scholars of economics and of psychology have proposed various explanations for why gender differences in salaries may exist. Employers may exercise “statistical discrimination” when they set salaries, making inferences based on group rather than individual characteristics38; in other words, an employer might pay a woman who works long hours a lower salary because of an assumption that women in general work fewer hours than men do. Unconscious gender biases may also influence employers,39–42 particularly when considering employees who are mothers.43,44 To the extent that we found a substantial gender difference that is not explained by numerous theoretically selected covariates (i.e., factors such as specialty, rank, work hours, and research time), these explanations merit attention.
Of note, even some of the difference that was explained by covariates in our model may warrant concern and attention. As in our previous work,1 specialty was a key driver of the overall difference in salary. Whether salary differences related to gender differences in specialization are justifiable depends on whether women freely choose lower-paying specialties or whether they are discouraged from higher-paying specialties, and whether the feminization of a specialty itself leads to lower pay.5
This study has a number of strengths. We obtained a high response rate—from an elite and homogeneous population in whom gender differences in salary would not be expected. Our questionnaire included specific items measuring a large number and variety of mechanisms that might underlie gender differences in salary. Several limitations also merit acknowledgment. All survey studies must confront concerns about possible selection bias; in this case, it is reassuring that we obtained a high response rate and found no gender difference between the initially targeted population and respondents. In addition, our measures draw from self-report, making them vulnerable to recall or other biases. Nevertheless, we developed these measures with standard techniques of survey design, including cognitive pretesting,14 and the items have strong face validity.
In sum, this study suggests that gender differences in the compensation of physicians in academic medicine exist in cohorts hired recently who are still at the early stages of their careers. Some of the gender difference in salary appears to be explained by differences in spousal employment status, suggesting important mechanistic roles for differences in the behavior of physicians themselves and/or disparate treatment by employers. The residual unexplained gender difference suggests that other mechanisms are also important, including the possibility of conscious and unconscious bias. Efforts to ensure gender equity in physician pay should consider these findings and focus interventions accordingly, with particular attention towards transparent, consistent methods for determining pay at the institutional level.
Acknowledgments: The authors wish to thank the K award recipients who took the time to participate in this study.
1. Jagsi R, Griffith KA, Stewart A, Sambuco D, DeCastro R, Ubel PA. Gender differences in the salaries of physician researchers. JAMA. 2012;307:2410–2417
2. Ash AS, Carr PL, Goldstein R, Friedman RH. Compensation and advancement of women in academic medicine: Is there equity? Ann Intern Med. 2004;141:205–212
3. Kaplan SH, Sullivan LM, Dukes KA, Phillips CF, Kelch RP, Schaller JG. Sex differences in academic advancement. Results of a national study of pediatricians. N Engl J Med. 1996;335:1282–1289
4. Kehrer BH. Factors affecting the incomes of men and women physicians: An exploratory analysis. J Hum Resour. 1976;11:526–545
5. Wright AL, Schwindt LA, Bassford TL, et al Gender differences in academic advancement: Patterns, causes, and potential solutions in one US college of medicine. Acad Med. 2003;78:500–508
6. Ness RB, Ukoli F, Hunt S, et al Salary equity among male and female internists in Pennsylvania. Ann Intern Med. 2000;133:104–110
7. Weeks WB, Wallace TA, Wallace AE. How do race and sex affect the earnings of primary care physicians? Health Aff (Millwood). 2009;28:557–566
8. DesRoches CM, Zinner DE, Rao SR, Iezzoni LI, Campbell EG. Activities, productivity, and compensation of men and women in the life sciences. Acad Med. 2010;85:631–639
9. Lo Sasso AT, Richards MR, Chou CF, Gerber SE. The $16,819 pay gap for newly trained physicians: The unexplained trend of men earning more than women. Health Aff (Millwood). 2011;30:193–201
10. Barrett M, Mcintosh M. The “family wage”: Some problems for socialists and feminists. Capital Class. 1980;4:51–72
11. . National Institutes of Health. Research Portfolio Online Reporting Tools. http://projectreporter.nih.gov/reporter.cfm
. Accessed July 28, 2011
12. Dillman DA, Smyth JD, Christian LM Internet, Mail, and Mixed-Mode Surveys: The Tailored Design Method. 20093rd ed Hoboken, NJ John Wiley & Sons, Inc.
13. UM ADVANCE Program. . University of Michigan Survey of Academic Climate and Activities. http://www.advance.rackham.umich.edu/climatesurvey1.pdf
. Accessed July 19, 2013
14. Willis GB Cognitive Interviewing: A Tool for Improving Questionnaire Design. 2005 Thousand Oaks, Calif Sage Publications, Inc.
15. Jagsi R, Motomura AR, Griffith KA, Rangarajan S, Ubel PA. Sex differences in attainment of independent funding by career development awardees. Ann Intern Med. 2009;151:804–811
16. Burnham KP, Anderson DR. Multimodel inference: Understanding AIC and BIC in model selection. Sociol Methods Res. 2004;33:261–304
17. Peters CC. A method of matching groups for experiment with no loss of population. J Educ Res. 1941;34:606–612
18. Belson WA. A technique for studying the effects of a television broadcast. Appl Stat. 1956;5:195–202
19. Rao RS, Graubard BI, Breen N, Gastwirth JL. Understanding the factors underlying disparities in cancer screening rates using the Peters–Belson approach: Results from the 1998 National Health Interview Survey. Med Care. 2004;42:789–800
20. Oaxaca R. Male–female wage differentials in urban labor markets. Int Econ Review. 1973;14:693–709
21. Blinder AS. Wage discrimination: Reduced form and structural estimates. J Hum Resour. 1973;8:436–455
22. Gerhart B. Gender differences in current and starting salaries: The role of performance, college major, and job title. Ind Labor Relat Rev. 1990;43:418–433
23. Bowles HR, Babcock L, McGinn KL. Constraints and triggers: Situational mechanics of gender in negotiation. J Pers Soc Psychol. 2005;89:951–965
24. Stuhlmacher AF, Walters AE. Gender differences in negotiation outcome: A meta-analysis. Pers Psychol. 1999;52:653–677
25. Bear J. “Passing the buck”: Incongruence between gender role and topic leads to avoidance of negotiation. Negotiation Confl Manag Res. 2011;4:47–72
26. Small DA, Gelfand M, Babcock L, Gettman H. Who goes to the bargaining table? The influence of gender and framing on the initiation of negotiation. J Pers Soc Psychol. 2007;93:600–613
27. Babcock L, Laschever S Ask For It: How Women Can Use the Power of Negotiation to Get What They Really Want. 2008 New York, NY Bantam Books
28. Sambuco D, Dabrowska A, Decastro R, Stewart A, Ubel PA, Jagsi R. Negotiation in academic medicine: Narratives of faculty researchers and their mentors. Acad Med. 2013;88:505–511
29. Sarfaty S, Kolb D, Barnett R, et al Negotiation in academic medicine: A necessary career skill. J Womens Health (Larchmt). 2007;16:235–244
30. Bowles HR, Babcock L, Lai L. Social incentives for gender differences in the propensity to initiate negotiations: Sometimes it does hurt to ask. Organ Behav Hum Decis Process. 2007;103:84–103
31. Kolb DM. Too bad for the women or does it have to be? Gender and negotiation research over the past 25 years. Negotiation J. 2009;25:515–531
32. Tinsley CH, Cheldelin SI, Kupfer Schneider A, Amanatullah ET. Women at the bargaining table: Pitfalls and prospects. Negotiation J. 2009;25:233–249
33. Wade ME. Women and salary negotiation: The costs of self-advocacy. Psychol Women Q. 2001;25:65–76
34. . Johns Hopkins Medicine, Office of Women in Science and Medicine. Programs and lectures, information and conversation sessions. http://www.hopkinsmedicine.org/education/women_science_medicine/programs_lectures.html
. Accessed July 19, 2013
35. . Perelman School of Medicine, University of Pennsylvania. FOCUS on Health and Leadership for Women. Overview of FOCUS initiatives. http://www.med.upenn.edu/focus/user_documents/OverviewFOCUSInitiatives_Final_10.16.12.pdf
. Accessed July 19, 2013
36. . Brigham and Women’s Hospital Center for Faculty Development and Diversity. Office for Women’s Careers. http://www.brighamandwomens.org/Medical_Professionals/career/CFDD/OWC/Images/OWCExternalBrochure_2010reprint.pdf
. Accessed July 19, 2013
37. Gerhart B, Rynes S. Determinants and consequences of salary negotiations by graduating male and female MBAs. J Appl Psychol. 1991;76:256–262
38. Goldin C Understanding the Gender Gap: An Economic History of American Women. 1990 New York, NY Oxford University Press
39. National Academy of Sciences; National Academy of Engineering; Institute of Medicine, Committee on Maximizing the Potential of Women in Academic Science and Engineering. Beyond Bias and Barriers: Fulfilling the Potential of Women in Academic Science and Engineering. 2006 Washington, DC National Academies Press
40. Steinpreis RE, Anders KA, Ritzke D. The impact of gender on the review of the curricula vitae of job applicants and tenure candidates: A national empirical study. Sex Roles. 1999;41:509–528
41. Heilman ME, Eagly AH. Gender stereotypes are alive, well, and busy producing workplace discrimination. Ind Organ Psychol. 2008;1:393–398
42. Phelan JE, Moss-Racusin CA, Rudman LA. Competent yet out in the cold: Shifting criteria for hiring reflect backlash toward agentic women. Psychol Women Q. 2008;32:406–413
43. Benard S, Correll SJ. Normative discrimination and the motherhood penalty. Gend Soc. 2010;24:616–646
44. Correll SJ, Benard S, Paik I. Getting a job: Is there a motherhood penalty? Am J Sociol. 2007;112:1297–1338