Since the late 1970s, gender has been shown to be associated with lower incomes among U.S. physicians, even after adjusting for work effort.1 More recent studies that also adjusted for physician age and specialty2–5 revealed similar income disparities, although the only study of income differences attributable to gender among obstetrician–gynecologists found that the 13–17% observed disparity in incomes was accounted for by differences in productivity attributable to gender, suggesting that male and female obstetrician–gynecologists' incomes are essentially equivalent.6 However, that study was limited by use of data from a single year, 1998, and failed to incorporate into the analysis important practice and provider characteristics, including provider race. Admittedly, little is known about the influence of race on physicians' incomes. In 1972, black physicians were reported to have different practice characteristics than their white counterparts7 and in 1977 a plea for analysis of geographic and functional distribution of black physicians was made,8 but analyses of differences between black and white physicians' incomes have not been published.
Whether income disparities among physicians are attributable to race or gender is of interest for at least two reasons. First, blacks and women represent an increasingly large proportion of medical students,9,10 residents,11 and the practicing physician work force.10,12,13 Second, because black primary care physicians have been shown to be more likely to care for the underserved14,15 as well as medically indigent and sicker populations,16 it stands to reason that their annual incomes might suffer.
Therefore, we wanted to explore the influence of race and gender on the incomes of black and white obstetrician–gynecologists, after adjusting for work effort, practice characteristics, and provider characteristics that are likely to influence physician incomes.
MATERIALS AND METHODS
Between 1992 and 2001, the American Medical Association (AMA) conducted regular telephone surveys of physicians and collected a broad variety of individual physician-level data, including weeks and hours of practice, number of patient visits seen, provider characteristics, practice characteristics, and physician incomes.17–24 The survey was designed to provide representative information on the population of all actively practicing, nonfederal physicians who spend the greatest proportion of their time in patient care activities; weights for each respondent were calculated to correct for potential bias created by unit nonresponse and survey eligibility and to ensure physician responders reflected the national distribution of physicians.
Each year, the telephone-administered survey was conducted on a random sample of the AMA Masterfile that are eligible for the survey. The following physicians were excluded: doctors of osteopathy, foreign medical graduates with temporary licensure, inactive physicians, physicians who were sampled during the last 5 years, physicians who are on the “do not contact” list, physicians not practicing in the United States, and physicians who have no license. In addition, after initial screening, federally employed physicians and physicians who spent less than 20 hours each week in patient care activities were excluded.
The following field procedures were developed to minimize nonresponse bias: two weeks before data collection, advance letters were sent describing the process and the survey, many specialty organizations provided endorsement letters, and summaries of the type of expense questions to be asked were provide in advance of the survey. In addition, a minimum of four callbacks to respondents was made before abandoning interview efforts, letters encouraging participation were sent to physicians who initially refused participation, and refusal conversion attempts were made by select interviewers.24
Survey weights were derived by first dividing the AMA Physician Masterfile population and survey respondents into 200 cells defined by specialty, years since the respondent received a doctor of medicine degree, AMA membership status, and board certification status. Unit response rates were constructed as the ratio of the number of physicians in the population to the number of respondents in each cell. Second, an eligibility correction was used, as only nonfederal patient care physicians—excluding residents—are eligible. The eligibility correction divides the subset of the population for which eligibility is known into 40 cells (according to years in practice, AMA membership status, gender, and board certification) and calculates the proportion of physicians in each cell who are eligible. This defines the eligibility weight. The overall weight applied for a given respondent is the product of the unit response weight and the eligibility weight.24
Although the survey had been conducted for much longer, we limited our analysis to data collected between 1992 and 2001 because these were the most recent data available and therefore likely to be the most relevant to the currently practicing physician workforce.
To ensure that all the physicians analyzed were comparable, that variables critical to the analysis were available for each subject, and that extreme outliers did not drive results, we used a sequential process of elimination of survey respondents (Fig. 1). First, we included only self-identified black or white physicians who were identified as practicing in an “office-based practice” in the study, thereby eliminated the minority of physicians who worked primarily doing research, as medical educators, as administrators, or in hospital settings. In addition, only respondents who provided information on key variables were included, and extreme outliers in annual patient visits and net incomes were excluded.
From the AMA data set, we extracted three types of independent variables that were likely to influence the dependent variable—net annual income:
- Physician work effort. While it has been demonstrated that hours worked is an important variable in analysis of physician incomes,2–5,25 the number of visits a physician sees each year may influence annual incomes. While private practice physicians typically bill based on patient visits, employed physicians are likely to have either quotas or incentive based production bonuses associated with patient visit volumes such that compensation methods are unlikely to be related to use of health services per person.26
- Provider characteristics. When making gender comparisons of physician incomes, age has commonly been used as an adjustment factor.2–5 Over the working lifetime, incomes demonstrate an “inverted-U” pattern27 that typically peaks near age 55 for primary care physicians,28,29 or after 20 to 25 years of practicing medicine. To dispel a concern that race or gender might influence the age at which a physician entered medical school, and therefore bias results, we incorporated the number of years that respondents had been practicing medicine into the analysis instead of physician age. In addition, because practice arrangements, such as having an ownership interest in the practice, have been associated with differences in annual income among physicians,30 we included whether the physician was an employee, as opposed to a full or partial owner of the practice, in the analysis. Finally, because board certification has been associated with higher incomes,31 we also included board certification status as an independent variable in the analysis.
- Practice characteristics. Physicians who live in different U.S. Census regions have been shown to have modestly different annual incomes17–24; therefore, we collected information on the U.S. census region in which the practice was located. In addition, because physicians who live in sparsely populated settings have been shown to have both lower32 and higher33 incomes, we classified responding physicians' county codes into three categories of metropolitan settings (fewer than 50,000, between 50,000 and 500,000, or greater than 500,000). Finally, because black physicians disproportionately serve the medically indigent and those with relatively poor insurance, factors which have been hypothesized to decrease physicians' incomes,16 we incorporated variables likely reflect those factors into the analysis: whether the practice provides Medicare services and the reported proportion of patients in the practice who are on Medicaid.
We used the consumer price index to adjust reported net annual income to constant 2004 dollars. Using a methodology that we have previously used,5,25,28,29,34–36 we multiplied the reported number of weeks worked in the last year by the total number of hours worked in the last week and the total number of visits seen in the last week to calculate the annual number of hours worked and the annual number of visits seen, respectively. Although for some respondents, the most recent week was atypical and therefore may not reflect an annual work effort, we could not construct an argument that would suggest that this methodology of estimating annual work effort would systematically bias results for any particular group examined. Further, an estimate of work effort in the previous week may be more accurate than a broader estimate of work in the past year. Because of the inverted-U relationship between number of years practicing medicine and annual incomes, we constructed dummy variables37 that reflected the categorization of years practicing medicine into 5-year increments, from 0 to 5 years practicing through 40-plus years practicing. While we used these dummy variables in the regression analysis, we aggregated them into 10-year increments through 30-plus years practicing for the purposes of demographic comparisons.
We used linear regression modeling to estimate the influence of race and gender on physicians' incomes, after adjustment for practice and provider characteristics. Within the regression model, we used dummy variables for each race–gender combination to calculate regression coefficients and 95% confidence intervals (CIs) in a model that used the independent variables detailed above and used consumer price index adjusted annual incomes as the dependent variable. We used SPSS 11.5 (Chicago, IL) and survey weights for all analyses. Because some variables, such as income and hours worked, were not normally distributed, we repeated our analyses using log-transformed data and found the same results. To aid in interpretation of results, we present all data and coefficients using nontransformed data. In addition, we performed diagnostic tests that examined our regression models for multi-colinearity and did not find that variables were colinear. This study was approved by Dartmouth Medical School's Committee for the Protection of Human Subjects, Hanover, NH (CPHS # 17707).
The sequential process of elimination of survey respondents left 733 white male, 41 black male, 164 white female, and 24 black female obstetrician–gynecologists available for analysis. Using survey weights, these respondents represented 709 white male, 40 black male, 162 white female, and 26 black female obstetrician–gynecologists (Fig. 1).
After adjusting only for inflation, white male obstetrician–gynecologists reported mean net annual incomes of $289,764 (Table 1). Compared with white men, black men reported mean annual inflation adjusted incomes that were $67,633 (23%) lower, white women had incomes that were $72,694 (25%) lower, and black women had incomes that were $29,300 (10%) lower. Compared with white males, black male obstetrician–gynecologists reported completing 5% more visits and working 18% more hours while white women reported completing 18% fewer visits and working 10% fewer annual hours. Indeed, among the study sample, there was a modest linear relationship between inflation-adjusted annual physician incomes and annual patient visits seen (r=.44, P<.001) and annual hours worked (r=.23, P<.001). Work effort for black women was comparable to that for white men.
White and black female obstetrician–gynecologists had practiced medicine for fewer years than their male counterparts: very few women who responded to the survey reported having practiced more than 20 years. Among the study sample, the number of years practicing medicine was highly correlated with age (r=.88, P<.001). Women of either race were more likely than men to be employees, as opposed to having an ownership interest in the practice. White and black females were less likely than their male counterparts to be board certified. Black obstetrician–gynecologists of both genders were more likely to live in the Southern U.S. Census region and less likely to live in the Western U.S. Census region than their white counterparts. Black obstetrician–gynecologists were more likely to practice in settings with a population greater than 500,000. Compared with their white counterparts, a higher proportion of black obstetrician–gynecologists' patients were on Medicaid, much more so for black men. The large majority of all groups provided Medicare services.
The regression model accounted for 27% of the variance in annual incomes and had strong face validity (Table 2). Higher numbers of annual visits were strongly associated with higher incomes; higher number of work hours, less so. The anticipated inverted-U lifetime earnings curve was reflected in the model. While board certification was associated with higher incomes, being an employee (as opposed to an owner of the practice), caring for a larger proportion of Medicaid patients, providing Medicare services, and living in less-populated setting were associated with lower incomes. Obstetricians practicing in the Western U.S. Census region earned lower incomes than those practicing elsewhere in the United States. After adjustment for these variables, black men's mean annual income was $210,859, $78,905 (27%) lower than that for white men (95% CI $120,082 lower to $37,729 lower, P<.001); white women's was $242,721, $47,043 (16%) lower (95% CI $70,127 to $23,958 lower, P<.001); and black women's was $246,355, $43,409 (15%) lower (95% CI $92,296 to $5,478 higher, P=.08). Adjusted incomes with 95% CIs for each group are presented as a proportion of white men's adjusted annual incomes in Figure 2.
While using our regression model to adjust for differences in work effort, provider characteristics, and practice characteristics somewhat mitigated the initial income difference found for white women, that was not the case for black men and women (Fig. 3).
We examined provider and practice characteristics that were likely to be associated with obstetrician–gynecologists' annual incomes, revealed differences attributable to race and gender in those characteristics, adjusted net annual incomes for observed differences, and found that race and gender independently contributed to lower net annual incomes among office-based obstetrician–gynecologists. The expected reduction in annual income for black male and white female obstetrician–gynecologists was substantial and statistically significant; the reduction for black women approached statistical significance.
This analysis revealed a strong association between higher annual incomes and work effort, particularly the number of patient visits seen for obstetrician–gynecologists. This finding is intuitive: physician reimbursement is largely based on the volume of patients seen. In addition, we found a strong association between having an ownership interest in the practice and having a higher mean annual income. This finding also has strong face validity: employed physicians might not be motivated as those with an ownership interest in the practice to see additional patients.
The finding that black physicians of either gender have a much larger proportion of Medicaid patients in their practices is consistent with previous findings that black physicians are more likely than whites to care for the underserved and medically indigent.14–16 The hypothesis that providing services to a large proportion of Medicaid enrollees might adversely influence physicians' incomes16 was borne out in the regression analysis. Undoubtedly, the association between lower annual incomes and serving a greater proportion of Medicaid patients reflects the low reimbursement rates provided in general by Medicaid funded health care services.
The association between higher annual incomes and board certification is consistent with findings from the early 1980s.31 This association might be explained in part by a propensity for provider organizations to require board certification for employment, by requirements by third-party payers that providers be board certified, or by market forces that use board certification as a marker for quality that is indirectly reimbursed. The modest difference between male and female obstetrician–gynecologists' rate of board certification was surprising. The reason for this difference is difficult to explain. Their shorter durations of practice might limit their ability to have completed the board certification process; alternatively, female obstetrician–gynecologists may be disinclined to pursue board certification or may be dissuaded from taking the exams by their high costs.
After correcting for differences in provider and practice characteristics, our finding that black and white female obstetrician–gynecologists should expect annual incomes that are so heavily discounted compared with that of white male practitioners was disconcerting. While limited by the small number of respondents, the 27% anticipated income differential between white and black male obstetricians is daunting. The anticipated 15–16% reduction in annual incomes found for female obstetrician–gynecologists was similar to that found in other studies that compared work-effort adjusted female with male physicians incomes in other specialties2–5; however, those analyses did not take into account the plethora of provider and practice variables that we examined. Indeed, the only previous study of the relationship between provider gender and incomes among obstetrician–gynecologists concluded that the income differences between male and female physicians' incomes could be attributed to differences in productivity.6 While we found similar productivity differences between white male and female obstetrician–gynecologists in this study, and greater productivity as measured by patient visits is clearly associated with higher incomes, productivity differences did not account for the race and gender differences that we found.
Our analysis has several limitations. First, the number of black survey respondents was small. Unfortunately, in their conduct of the survey, the AMA did not over-sample the black physician population. A larger sample of black and female physicians would improve confidence in these findings. In particular, our small sample size may have caused us to conclude incorrectly that some examined variables were not significant contributors to the regression model. In particular, that black female obstetricians' incomes were not statistically significantly different at the P=.05 level than white males' should not be misinterpreted—there was a very strong trend suggesting that black female obstetrician–gynecologists had substantially lower incomes than their white male counterparts. Second, the study was limited by the methods of the survey, which, although an established survey of physicians, experienced a survey response rate that ranged from 57% to 70% among obstetrician–gynecologists and demonstrated substantial year-to-year variation in number of respondents during the time period examined. In addition, we were unable to calculate response rates specific to each of the groups examined. However, the ability to combine 10 years of data strengthened the study and offered a much more robust data set than would have been the case had fewer years of data been available. Although we tried to ensure comparability by limiting our analysis to those who shared important characteristics such as work setting, range of work hours, and data availability, our findings may not generalize to obstetrician–gynecologists who do not have those characteristics. In addition, the survey data are self-reported; we were not able to validate reported incomes, work effort, provider characteristics, or practice characteristics. It is important to note, however, that, to influence our findings, any misrepresentation of these variables would need to be specific to race and gender, done in such a way that anticipated the regression model that we used, and systematically and consistently performed by survey respondents. The same can be said of nonresponse bias. Both seem highly unlikely.
Third, while incomes were adjusted to constant dollars and were adjusted for regional practice setting and urbanization level, we were not able to adjust for differences in purchasing power parity across those settings—differences that were shown to mitigate constant dollar income differences among rural and urban physician practices.33 While we were not able to adjust for any differences in other important variables, such as hours worked and visits seen, that might have been associated with changes in practice patterns over the time period examined, our previous investigation suggested that there were not dramatic differences in work effort for obstetrician–gynecologists during the time period examined.25 Fourth, the study was inherently limited by the data available from the AMA survey. Although it would have been interesting to explore alternative explanations for the race- and gender-based income disparities that we found, such as the type of practice chosen (ie, more versus less surgically oriented), proportion of charity care, respondents' educational debt burden, clinicians' levels of satisfaction with their practices, and rates of different procedures performed by obstetrician–gynecologists, those data were not available.
Fifth, we were not able to examine differences in the quality of care provided by white and black, male and female obstetrician–gynecologists. This is an important limitation—higher incomes might be justified for providers who provide higher-quality care or are in greater demand, and several studies suggest that female obstetrician–gynecologists provide higher quality of care38,39 and are in greater demand40 than their male counterparts. These studies suggest that, to the degree that quality is associated with higher income, our findings likely underestimated the income difference associated with gender among obstetrician–gynecologists.
Despite these limitations, the results of this study suggest that black race and female gender are independently associated with lower annual incomes among obstetrician–gynecologists. These findings should be contextualized, however. Foremost, the anticipation of financial returns should not drive the choice to enter the medical profession; the results presented here are therefore unlikely to dissuade blacks or women from entering obstetric and gynecology. In addition, physicians derive many nonfinancial benefits from their roles, including prestige, the ability to serve their communities, and the opportunity to model for others of similar backgrounds the advantages of pursuing higher education—benefits that are likely to be highly motivating regardless of physician gender or race.
Nevertheless, we believe our findings have implications for current and future obstetric and gynecologic practice. First, policy makers should explore gender equity in medicine more carefully and over a broader range of specialties. Second, black and female obstetrician–gynecologists who are seeking employment may want to examine the workload and incomes of their white male counterparts when negotiating their salaries. Finally, policy makers should be wary of the cyclical reasoning inherent in adjusting for practice arrangement, subspecialty practice, and academic rank when comparing incomes of male and female or black and white physicians—differences in these factors may themselves be indicators of discrimination.
Regardless of the implications of these findings, we can think of no reason why black or female obstetrician–gynecologists should be categorically underpaid. Although additional study of race and gender differences in income is warranted, immediate examination and remediation of salary inequities are imperative. Black and female obstetrician–gynecologists have achieved the same level of education, have made the same time commitment to training, and have experienced the same direct and opportunity costs required of such commitment28 as their white male counterparts. They should be paid equitably.
1. Langwell KM. Differences by sex in economic returns associated with physician specialization. J Health Polit Policy Law 1982;6:752–61.
2. Ohsfeldt RL, Culler SD. Differences in income between male and female physicians. J Health Econ 1986;5:335–46.
3. Carr P, Friedman R, Moskowitz M, Kazis L, Weed H. Research, academic rank, and compensation of women and men faculty in academic general internal medicine. J Gen Intern Med 1992;7:418–23.
4. Baker L. Differences in earnings between male and female physicians. N Engl J Med 1996;334:960–4.
5. Wallace AE, Weeks WB. Differences in income between male and female primary care physicians. J Am Med Womens Assoc 2002;57:180–4.
6. Pearse WH, Haffner WH, Primack A. Effect of gender on the obstetric-gynecologic work force. Obstet Gynecol 2001;97:794–7.
7. Thompson T. Selected characteristics of black physicians in the United States, 1972. JAMA 1974;229:1758–61.
8. Gray LC. The geographic and functional distribution of black physicians: some research and policy considerations. Am J Public Health 1977;67:519–26.
9. Health, United States, 2004. Table 107. Hyattsville (Maryland): National Center for Health Statistics. 2004.
10. AAMC Data Book: Statistical information related to medical schools and teaching hospitals. Washington, DC: The Association of American Medical Colleges. 2000.
11. Brotherton SE, Rockey PH, Etzel SI. U.S. graduate medical education, 2004-2005. JAMA 2005;294:1075–82.
12. Pasko T, Seidman B, Birkhead S. Physician Characteristics and Distribution in the U.S.: 2000–2001. In: Division of Survey and Data Resources. Chicago (IL): American Medical Association. 2000.
13. Corbie-Smith G, Frank E, Nickens H. The intersection of race, gender, and primary care: results from the Women Physicians' Health Study. J Natl Med Assoc 2000;92:472–80.
14. Komaromy M, Grumbach K, Drake M, Vranizan K, Lurie N, Keane D, et al. The role of black and Hispanic physicians in providing health care for underserved populations. N Engl J Med 1996;334:1305–10.
15. Xu G, Fields SK, Laine C, Veloski JJ, Barzansky B, Martini CJ. The relationship between the race/ethnicity of generalist physicians and their care for underserved populations. Am J Public Health 1997;87:817–22.
16. Moy E, Bartman BA. Physician race and care of minority and medically indigent patients. JAMA 1995;273:1515–20.
17. Gonzalez M. Physician Marketplace Statistics, 1992. Chicago (IL): American Medical Association, Center for Health Policy Research. 1992.
18. Gonzalez M. Physician Marketplace Statistics, 1993. Chicago (IL): American Medical Association, Center for Health Policy Research. 1993.
19. Gonzalez M. Physician Marketplace Statistics, 1994. Chicago (IL): American Medical Association, Center for Health Policy Research. 1994.
20. Gonzalez M. Physician Marketplace Statistics, 1995. Chicago (IL): American Medical Association, Center for Health Policy Research. 1995.
21. Gonzalez M. Physician Marketplace Statistics, 1996. Chicago (IL): American Medical Association, Center for Health Policy Research. 1996.
22. Gonzalez M. Physician Marketplace Statistics, 1997/8. Chicago (IL): American Medical Association, Center for Health Policy Research. 1998.
23. Zhang P, Thran S. Physician Socioeconomic Statistics, 1999-2000. Chicago (IL): American Medical Association, Center for Health Policy Research. 1999.
24. Wassenaar J, Thran S. Physician Socioeconomic Statistics, 2000–2002. Chicago (IL): American Medical Association, Center for Health Policy Research. 2001.
25. Weeks WB, Wallace AE. Time and money: a retrospective evaluation of the inputs, outputs, efficiency, and incomes of physicians. Archives of Internal Medicine 2003;163:944–8.
26. Conrad DA, Maynard C, Cheadle A, Ramsey S, Marcus-Smith M, Kirz H, et al. Primary care physician compensation method in medical groups: does it influence the use and cost of health services for enrollees in managed care organizations? JAMA 1998;279:853–8.
27. Polackek SW, Siebert WS. The Economics of Earnings. Cambridge (NY): Cambridge University Press. 1993.
28. Weeks WB, Wallace AE, Wallace MM, Welch HG. A comparison of educational costs and incomes of physicians and other professionals. N Engl J Med 1994;330:1280–6.
29. Weeks WB, Wallace AE. The more things change: revisiting a comparison of educational costs and incomes of physicians and other professionals. Acad Med 2002;77:312–9.
30. Ness R, Ukoli F, Hunt S, Kely S, McNeil M, Richarson V, et al. Salary equity among male and female internists in Pennsylvania. Ann Intern Med 2000;133:104–10.
31. Becker ER, Culler S, Ohsfeldt R. Impact of board certification on physician practice characteristics. J Med Educ 1985;60:9–15.
32. Weil TP. Attracting qualified physicians to underserved areas. Part 2. Pay physicians more to practice in underserved areas. Physician Exec 1999;25:53–63.
33. Reschovsky JD, Staiti AB. Physician incomes in rural and urban America. Issue Brief Cent Stud Health Syst Change 2005;92:1–4.
34. Weeks WB, Wallace AE. Financial returns on specialty training for surgeons. Surgery 2002;132:795–802.
35. Weeks WB, Wallace AE. Long-term financial implications of specialty training for physicians. Am J Med 2002;113:393–9.
36. Weeks WB, Wallace AE. Return on educational investment in geriatrics training. J Am Geriatr Soc 2004;52:1940–5.
37. Kleinbaum DG, Kupper LL, Muller KE, Nizam A. Applied regression analysis and multivariable methods. 2nd ed. Pacific Grove (CA): Duxbury Press. 1997. p. 317–60.
38. Pham HH, Schrag D, Hargraves JL, Bach PB. Delivery of preventive services to older adults by primary care physicians. JAMA 2005;294:473–81.
39. Turner BJ, Amsel Z, Lustbader E, Schwartz JS, Balshem A, Grisso JA. Breast cancer screening: effect of physician specialty, practice setting, year of graduation, and sex. Am J Prevent Med 1992;8:78–85.
40. Chandler PJ, Chandler C, Dabbs ML. Provider gender preference in obstetrics and gynecology: a military population. Mil Med 2000;165:938–40.