In the last 20 years, more and more female physicians have entered the specialty of obstetrics and gynecology, creating substantial changes in the process. Women represented just 30% of obstetrics and gynecology residents in 1980, 47% by 1990, and 76% by 2005.1,2 Consequently, while only 12% of obstetrician gynecologists were female in 1980, that number reached 40% by 2005, and it continues to rise.3 As these new physicians enter practice, they are transforming the face of the specialty of obstetrics and gynecology.
This entry of women into obstetrics and gynecology, combined with the frequent patient preference for being treated by a female physician, raises a number of questions about the implications for the specialty and for individual physicians. It also leads inexorably to a broader discussion of the nature of gender equity or inequity within the specialty.4–6 In particular, one question of great interest is the relative productive and economic performance of female and male obstetrician–gynecologists. Do men and women practice the specialty in the same manner, or are there differences in practice patterns and productivity? Do women still earn relatively lower incomes in obstetrics and gynecology, as they do in most occupations and in most medical specialties, or have they gained ground? In an environment in which patient preference increasingly favors female physicians (at least to some extent), are female obstetrician–gynecologists faring better over time or better relative to male obstetrician–gynecologists? How have gender differences evolved in the last decade? This paper seeks to provide insight into these questions.
The heart of the matter is to estimate to what extent differences in income can be explained by differences in productivity, professional characteristics, or human capital. One earlier study reported that female obstetrician–gynecologists were 15% less productive and that these productivity differences corresponded reasonably well with income differences.7 Two other studies found that unexplained income differences of 14–18% remained even after controlling for productive characteristics.8,9 This paper focuses on young obstetrician–gynecologists and uses multiple years of detailed data to examine gender differences in obstetrics and gynecology in greater detail. The paper will investigate the size of these differences, their evolution over the 1990s, and the extent to which income differences can be explained by differences in practice style, productivity, or other provider characteristics.
MATERIALS AND METHODS
Over the past 15 years, the American College of Obstetricians and Gynecologists (ACOG) has collected a wide range of detailed physician-level data pertaining to all aspects of obstetrics and gynecology practice with four Socioeconomic Surveys of Fellows. The Socioeconomic Surveys were designed by ACOG staff in collaboration with Princeton Survey Research Associates and were conducted by Princeton Survey Research Associates in each of the survey years 1991, 1994, 1998, and 2003. The samples were stratified random samples of the ACOG membership, excluding residents, those serving in the active military, and those not involved directly with patient care. The data were collected by mail surveys over a period of 2 months in each survey year, and ACOG members sampled were provided with one advance letter, a packet of survey materials, and three follow-up contacts. The materials emphasized the value of the information to ACOG and its members, the random nature of the sample, and the complete confidentiality of survey responses. The resulting sample size was approximately 1,200 per year, with response rates of 67% in 1991, 56% in 1994, 42% in 1998, and 49% in 2003. To adjust for the small variations in response rates across geographic and demographic groups, and thereby to make the sample representative of the ACOG membership, Princeton Survey Research Associates calculated data weights. These weights will be used in the analysis below. Princeton Survey Research Associates also investigated the overall quality of the data, concluding that the data are generally of very good quality for a long and complex self-administered survey of professionals. Item nonresponse was below 5% for most nonfinancial variables.
The analysis in this paper used survey data on physicians’ personal demographics, professional demographics, practice arrangements, work effort, productivity, and income. Existing research has confirmed important interrelationships among these variables.8–11 Personal demographics included gender, race, and age. Professional demographics included subspecialty (if any), years in medical practice, and fellowship category. (Fellowship category was used as a proxy for board certification because ACOG regulations state that all Fellows are board-certified and all Junior Fellows are not board-certified unless they have become so in the past year.) Practice arrangements were described by whether the physician was salaried and was in private practice, and the practice environment was described by indicator variables for the geographic region and the size of the physician’s county. Physician work effort and productivity were measured as weeks worked, hours worked, patients seen, and procedures performed. It is important to note that, because the surveys were designed specifically for obstetrician–gynecologists, the data are exceptionally well-suited to analyzing obstetrics and gynecology practice, even including detailed counts of hours spent in specific activities, patients seen in different settings, and types of procedures performed.
Physician income was measured as net medical income in the previous year, adjusted to 2002 dollars using the Consumer Price Index.12 Because physicians worked varying numbers of weeks, a measure of weekly income was used, calculated as annual income divided by weeks worked. Furthermore, to account for the nonnormal distribution of incomes (assessed with the Shapiro-Wilk test), income values were transformed into logarithms for the primary regression analysis. (An added advantage of this transformation is that coefficients in the log-linear specification reflect the income return to any characteristic as a percentage of income.) Thus, results are reported by using the logarithm of weekly income, although results using other income measures were not significantly different.
Data cleaning proceeded as follows. Observations were retained in the sample if they contained values for income, weeks, hours, patients (except in 1991), procedures, age, and sex. Furthermore, observations were retained if 1) age was less than 80 years, 2) annual income was between the 2nd and 98th percentiles ($38,000 and $675,000), 3) weeks worked was at least 26 (except for those who had just entered practice, for whom smaller numbers were acceptable), 4) hours worked and patients seen were between the 1st and 99th percentile (18–130 hours, 10–220 patients), and 5) procedures performed was less than the 99th percentile (682 per year).
Some analysis was performed on a sample restricted to only physicians aged 40 years or younger. The career path and experience of older male obstetrician–gynecologists is likely to have been so different from that of younger obstetrician–gynecologists (male or female) that they are not an appropriate reference group when comparing income or productivity. Indeed, because older obstetrician–gynecologists are overwhelmingly male (80% in the ACOG data, 81% in American Medical Association [AMA] data covering this time period),3 age and gender covary systematically and regression results may be biased. It is not sufficient to merely control for age in a regression, nor is it sufficient to include a set of flexible dummy variables to allow for a nonlinear life cycle pattern of income. Gender differences can only be assessed accurately by focusing on a sample of similarly situated young obstetrician–gynecologists.
The analysis begins with a comparison of the practice characteristics and productivity of male and female obstetrician–gynecologists. Significance of gender differences was assessed with t tests (unpaired two-sample t tests allowing unequal variances) for means and χ2 tests for proportions. The main regression analysis estimated the influence of gender, race, practice characteristics, and physician activity on physician income:
Log weekly income=α1female+α2age+α3I(black)+α4I(Asian)+α5I(other race)+α6I(subspecialist)+α7I(fellow)+α8Years in practice+α9I(salaried)+α10I(private practice)+α11Deliveries+α12Surgeries+α13Hours+α14Patients+α15I(census region)+α16I(County population less than 50,000)+α17I(County population more than 100,000)+ϵ
This analysis was performed by using the sample of all physicians and the sample of young physicians. It was also performed on all years of data together (in which case, indicator variables for years were included as independent variables), and as separate regressions on each year of data (allowing all coefficients to vary across years). The omitted demographic group (regression control) was a white male obstetrician–gynecologist who was not a subspecialist, not a full fellow, not salaried, not in private practice, and who was located in the northeastern United States and in a county with population between 50,000 and 100,000. Regression results were estimated as weighted least squares and are reported using the years to which the income data refer rather than the years in which the data were collected. Lastly, the gender differential in each year was decomposed into two portions: 1) the portion explained by differences in practice characteristics and physician activity (salaried practice, hours worked, procedures performed), and 2) the portion which cannot be explained by observable factors other than gender. This Oaxaca Decomposition is a standard technique in economic analyses of gender and race differentials and was first described by Ronald Oaxaca in 1973.13,14 The computer program STATA/SE 8.2 for Windows (StataCorp, College Station, TX) was used for analysis. This research has been approved by the Institutional Review Board at the National Bureau of Economic Research (Cambridge, MA).
The data selection process described above produced a sample of 3,698 physicians. Of these, 1,124 (31%) were female, and 1,501 (41%) were aged 40 years or younger. The female share rose over time: 22% of the sample was female in 1990, compared with 40% in 2002. This female share was also higher and rose faster among young physicians, from 39% in 1990 to 60% in 2002. Over all years, the younger physicians in the sample were equally split by gender (779 men, 722 women, 48% women).
Table 1 shows a gender comparison of income, characteristics, and physician activity. In the full sample, female obstetrician–gynecologists’ annual incomes averaged $196,000, which is 24% less than the $257,000 income reported by male obstetrician–gynecologists (P<.001). When considering weekly income, the observed gender gap was 22%, whereas for young physicians the gender gaps were similar at 23% for annual income and 20% for weekly income. Female obstetrician–gynecologists also differed on their basic demographic characteristics: they were younger (average age 40 years, compared with 48 years for men), had been in practice for fewer years (8 versus 17), and were slightly more likely to be nonwhite. The age and experience differences were much smaller when the sample was restricted to young physicians.
There was also evidence that male and female obstetrician–gynecologists chose different practice characteristics and settings. Male obstetrician–gynecologists were more likely to be Fellows (rather than Junior Fellows), even among the younger group. They were also more likely to subspecialize, a difference that was more pronounced among the younger group (in which men were 50% more likely to be subspecialists). Female obstetrician–gynecologists were substantially more likely to be in a salaried position and less likely to be in private practice.
The most interesting gender differences are visible in physician productivity. Female physicians worked about 1.5 fewer weeks, most of which can be attributed to maternity leave. Among all obstetrician–gynecologists, women worked about 5% fewer hours per week (P<.001), but this gap was larger (10%) among young obstetrician–gynecologists (P<.001). Women also saw fewer patients per week: 8% among physicians of all ages, and 9% fewer among young physicians (P<.001 for both samples). The most striking difference was in procedures performed: overall, female physicians performed 18% fewer procedures annually and 14% fewer weekly (P<.001). Among young physicians, women performed 21% fewer procedures annually and 14% fewer weekly (P<.001). This gap was present for both deliveries and surgical procedures although it was more pronounced for surgical procedures. Hence, based on these measures, female obstetrician–gynecologists were about 85–90% as productive as male obstetrician–gynecologists. (Further tabular analysis of the ACOG survey data is available in two reports written in 2004 by the Directors of the Department of Health Economics at ACOG: Rebecca Kelly, “Profile of Ob-Gyn Practice” and James Scroggs, “Financial Trends in Ob-Gyn Practice, 1990–2002.” These are available online at www.acog.org.) Lastly, the evidence in Table 1 suggests that younger male obstetrician–gynecologists are more productive than older male obstetrician–gynecologists, and this is confirmed by further statistical analysis comparing these groups. It is not, however, clear whether this represents a change from one cohort to the next or is merely an artifact of observing these individuals at different points in their careers.
Table 2 shows an abbreviated gender comparison for each individual year of data, focusing on only young physicians. The data show that the raw income gap between young male and female obstetrician–gynecologists stayed relatively high throughout the period 1990–2002. It neither compressed nor widened uniformly, but rather widened from 20% to 26% between 1990 and 1993, and then narrowed to 20% by 1997 and 19% by 2002. The gender gap in weekly income showed a similar pattern. The table shows nonlinear trends in many of the other variables as well. By 2002, female obstetrician–gynecologists were relatively less likely to be subspecialists, even more likely to be in salaried positions, and remained less likely to be in private practice. The productivity gaps in hours and patients stayed stable at about 10%. In contrast, the gap in procedures performed widened considerably: while young female obstetrician–gynecologists performed 18% fewer procedures than their male counterparts in 1990, by 2002 that gap had grown to 28%. The gap in surgeries performed was particularly notable: women reported performing 31% fewer surgeries in 2002.
Regression analysis allows a more systematic consideration of gender gaps in income, their evolution over time, and the importance of differences in productivity and practice characteristics. These results are shown in Table 3. Among obstetrician–gynecologists of all ages, the regression estimates that female obstetrician–gynecologists had weekly incomes that were 22.4% lower than male obstetrician–gynecologists (P<.001). Furthermore, this gap narrowed over time (from 31% in 1990 to 19% in 2002), although not monotonically. Turning to obstetrician–gynecologists younger than 40 years old (column 3), the results indicate that women earned 19% less per week (P<.001). This gap stayed relatively constant over the 1990s, changing insignificantly from 21% in 1990 to 18% in 2002. (It is useful to note that the falling gender gap shown in column 1 is primarily attributable to the shift of the age and gender composition of the specialty rather than to a change in relative performance.)
Controlling for practice characteristics and productivity (columns 2 and 4) provides additional insight. Once controls were included, the estimated gender gap was much smaller: among the whole sample it was 9% (P<.001), and among the sample of young obstetrician–gynecologists it was 7% (P=.02). For physicians of all ages, the unexplained gender gap remained relatively stable over time (increasing somewhat, although insignificantly, between 1990 and 1993). For young physicians, however, the unexplained gender gap narrowed substantially over time: in the 1990s the point estimates ranged between 8.4% and 11.1%, all of which were significant, but by 2002 the estimated gap fell to 1.5%, which is insignificant and fairly precisely estimated (P=.76, standard error 4.9%.) By 2002, gender per se is an insignificant factor in explaining the income of young obstetrician–gynecologists. Analysis of alternate measures of income (annual income, weekly income, or the logarithm of annual income) yielded similar results. Gaps in annual income were slightly higher but evolved in a similar fashion.
The Oaxaca decomposition reaffirms these results, showing that the share of the gender gap that cannot be explained by differences in practice characteristics and productivity remained stable at about half in the 1990s (0.50 in 1990, 0.47 in 1993, and 0.46 in 1997), but fell to near zero by 2002 (0.04 in 2002). The results can be seen from another perspective in Figure 1, which shows female income as a percentage of male income over time, both unadjusted and adjusted for differences in productivity and practice characteristics. The unadjusted share was relatively stable at 75–80% over the entire time period. In contrast, the adjusted share was slightly higher and relatively stable in the 1990s (aside from a decrease for annual income between 1990 and 1993), but rose considerably between 1997 and 2002. By 2002, female weekly income very closely approaches 100% of male weekly income when adjusted for characteristics and productivity.
To consider the returns to other characteristics, Table 4 shows the full regression results. Age was not significantly related to income in this linear specification. (An alternate specification including a quadratic in age to fit the inverted U relationship did yield significant coefficients for age and its square, but the other results remained unchanged.) Race was negatively related to income: black and Asian physicians earned slightly lower incomes, although these effects were less apparent in the more homogeneous sample of young physicians. Subspecialization appears to have been associated with higher income, although not among young physicians, possibly because the additional years of training mean that they had entered practice more recently. (Detailed analysis of specific subspecialties revealed the expected heterogeneity in their relationship to income, but these effects were orthogonal to gender.) Being a full Fellow (a proxy for board certification) showed a large effect on weekly income, rendering weekly income 45% higher in the full sample and 18% higher for young physicians. Years in medical practice had a significant effect on income only for young physicians, increasing income by 2.6% for each year in practice.
Turning to productivity, a 10% increase in each of the productivity measures was associated with a percentage increase in income: 1.5–1.9% for deliveries, 0.7–1.1% for surgeries, 0.4–1.2% for hours, and 3% for patients. (Although the unavailability of patient counts in 1991 meant that this variable was not included in the regressions in Table 4, results were obtained using only the other years of data.) Finally, the last three rows of Table 4 show that inflation-adjusted weekly incomes for all obstetrician–gynecologists fell since 1990, although not in a precisely monotonic fashion. Comparing 2002 with 1990, real incomes were 19.5% lower for obstetrician–gynecologists of all ages, and 13.3% lower for younger obstetrician–gynecologists. These changes are statistically and practically significant and are largely similar to income trends reported by the AMA.15,16
The above results indicate that female obstetrician–gynecologists receive incomes that are substantially lower (20–25% lower) than their male counterparts. However, the gender gap in income is reduced significantly when differences in practice characteristics and physician productivity are taken into account. In fact, among young physicians the share of the gender income gap that could not be explained by differences in characteristics or productivity fell so much around the turn of the century that it was statistically insignificant by 2002. This is the most striking conclusion of the present analysis: according to the most recent available data, male and female obstetrician–gynecologists who practice in the same manner appear to receive the same incomes. Gender does not matter. It is only when female obstetrician–gynecologists choose less financially rewarding practice arrangements or do less (see fewer patients, work fewer hours, perform fewer procedures) that they earn lower incomes.
This analysis is unusual in its use of detailed data that was collected specifically on obstetrics and gynecology practice and that spans a long time period. The results above show that there are large gender differences in detailed measures of physician productivity, such as the numbers of deliveries and surgeries, and that accounting for these differences is crucial in properly understanding gender differences in obstetrics and gynecology practice. The long time period spanned by the data is also important, given that women physicians entered obstetrics and gynecology in increasing numbers in the 1990s. Moreover, the paper’s analysis of the most recent available data has proved to be important, uncovering intriguing new information about the specialty in recent years.
The nature of the data also means that the sample sizes were large enough to permit regression analysis that controlled for numerous factors and that was performed on several different subsamples. Restricting samples to a single year of data permitted investigation of changes over time. Restricting samples based on age was even more crucial: since older and younger physicians exhibit many important differences and there are so few older female obstetrician–gynecologists, it is imperative to make gender comparisons using a more homogenous sample of young physicians. Otherwise, the male-female comparison is contaminated by the old-young comparison.
The major potential weakness of this analysis derives from possible nonresponse bias in the ACOG surveys. As discussed above, response rates were around 50%, and were closer to 40% when considering only responses that included valid data on all important variables. Princeton Survey Research Associates, which designed and executed the surveys, reported on the quality of the data and came to the conclusion that the data were generally of good quality and that there was no substantial reason to suspect significant nonresponse bias. The response rates are also comparable to those in similar mail surveys of physicians.17 Furthermore, the surveys provide the only extensive, detailed, and recent data on obstetrics and gynecology practice. Given these facts, it seems appropriate to make use of the unique and valuable information gathered by the ACOG surveys while acknowledging that the data were gathered through self-administered mail surveys with moderate response rates.
The results herein concord well with previous published results and also contribute new insights. One earlier study reported a raw gender gap in income of 22%, of which 12% remained unexplained by differences in characteristics, while another reported a raw gap of 33%, of which 14% remained unexplained.8,18 The current study finds similar raw gender gaps but is able to explain more of the gaps by focusing on young obstetrician–gynecologists and including more detailed variables on physician characteristics and workload. In addition, Pearse et al7 noted that gender gaps in income and productivity were of roughly the same magnitude (15%), and the current analysis comes to a similar basic conclusion. Most recently, Weeks and Wallace9 used AMA data from the period 1992–2001 and concluded that there was an unexplained income gap of 18% throughout the 1990s. The current analysis reaches a different conclusion: that there is indeed a substantial raw gender gap, but that over time it has increasingly been explained by differences in professional activity between male and female obstetrician–gynecologists.
So what do these results mean for physicians and for the specialty of obstetrics and gynecology? First, it appears that obstetrics and gynecology is one of the few professions in which men and women can earn similar incomes for doing the same work. There is substantial evidence to indicate that this is not the case in most other occupations, even professional occupations, and even in other medical specialties.10,11,19,20 Second, this gender equality within obstetrics and gynecology is a relatively new phenomenon: as recently as 1997, women earned significantly less even if they did the same work. The gender equality goal of “equal pay for equal work” appears to be a reality in obstetrics and gynecology, but a fairly recent one. It is possible that a “tipping point” was reached around the turn of the century, as the presence and influence of women in the specialty reached some critical threshold. Third, this does not mean that men and women have equal annual or weekly incomes within obstetrics and gynecology: women do less (10% fewer hours and patients, 28% fewer procedures), and earn less (21% less per week). So, from one perspective, incomes are lower, but there is a good reason for it. On the other hand, there are a number of reasons why women doing less may not entirely explain women earning less.
How might this appearance of equality in fact belie an underlying reality of gender inequity? First, it is important to recall that the gaps in hours and patients are smaller than the income gaps: if the amount of work is measured by time spent on the job, women are in fact underpaid. Indeed, it may be the case that men achieve their higher incomes by focusing more heavily on more financially rewarding aspects of obstetrics and gynecology practice (deliveries and surgeries, subspecialization, or private practice). If the financial rewards to different aspects of practice change over time such that the “male” practice style generates more income, that may itself be a source of unjustified gender inequity. In short, the results support a conclusion of gender equity only if the differences in practice style are truly by choice and the differential returns do not change over time in a way that favors men. Furthermore, patient preference for female obstetrician–gynecologists creates market pressure that could be expected to increase the incomes and employment opportunities of female obstetrician–gynecologists. If this type of pro-female bias does indeed increase incomes and opportunities, it can only be serving to offset a prevailing pro-male bias. It certainly does not appear to generate a net advantage for female obstetrician–gynecologists.
As one attempt in understanding this history, I draw on the analysis reported herein (as well as more detailed analysis of the same data) to propose the following conjecture. It is possible that, as women’s visibility and status in obstetrics and gynecology increased, not only were women more able to choose a practice style that better matched their preferences (such as doing fewer procedures), but they were also able to eliminate any remaining gender discrimination in income. Under this scenario, women’s incomes would have fallen because they started doing less and risen because they were treated more equally: the net result would have been that the raw income gap remained stable while the unexplained gap diminished. I stress that this is only a hypothesis, consistent with the data, but requiring confirmation by additional research.
Future research will thus seek to investigate these questions in more depth, further exploring the economic structure of obstetrics and gynecology practice and the anatomy of gender differences within obstetrics and gynecology. Sorting out the exact nature of the differences in practice style, the role these play in explaining income differences, and the changes that took place recently will be of particular interest. There are also many less tangible differences in practice style, such as communication styles, health screening, or procedure choice, that may be of great consequence but which lie beyond the scope of the current study. A broader question that arises is whether all obstetrician–gynecologists—male and female—earn less because obstetrics and gynecology is increasingly regarded as a female-dominated specialty.21 It is well known that there is a strong negative correlation between the share of physicians in a medical specialty who are female and the average income (and prestige) in that specialty. If the specialty of obstetrics and gynecology is in the process of transitioning from a male specialty to a female specialty, the average income in the specialty may be falling for obstetrician–gynecologists of both genders.
1. Rowley BD, Baldwin DC Jr, McGuire MB. Selected characteristics of graduate medical education in the United States. JAMA 1991;266:933–43.
2. Graduate medical education. JAMA 2005;294:1129–43.
3. American Medical Association. Physician characteristics and distribution in the US, 2006 edition. Chicago (IL): American Medical Association Press; 2006.
4. Lyon DS. Where have all the young men gone? Keeping men in obstetrics and gynecology. Obstet Gynecol 1997;90:634–6.
5. Lyon DS. Graduate education in women’s health care: where have all the young men gone? Curr Womens Health Rep 2002;2:170–4.
6. Ossorio PN. No boys allowed? J Gend Specif Med 1999 2:34–8.
7. Pearse WH, Haffner WH, Primack A. Effect of gender on the obstetric-gynecologic work force. Obstet Gynecol 2001;97:794–7.
8. Kelly RW, Pereles SA. Gender influences on earnings of obstetrician-gynecologists. Obstet Gynecol 1995;86:112–8.
9. Weeks WB, Wallace AE. The influence of physician race and gender on obstetrician-gynecologists’ annual incomes. Obstet Gynecol 2006;108:603–11.
10. Baker LC. Differences in earnings between male and female physicians. N Engl J Med 1996;334:960–4.
11. Langwell KM. Differences by sex in economic returns associated with physician specialization. J Health Polit Policy Law 1982;6:752–61.
13. Oaxaca R. Male-female wage differentials in urban labor markets. Int Econ Rev 1973;14:693–709.
14. Borjas GJ. Labor economics. 3rd ed. Boston (MA): McGraw-Hill/Irwin; 2005.
15. American Medical Association. Socioeconomic characteristics of medical practice 1993. Chicago (IL): American Medical Association Press; 1993.
16. American Medical Association. Physician socioeconomic statistics, 2000–2002. Chicago (IL): American Medical Association Press; 2001.
17. Bettes BA, Strunk AL, Coleman VH, Schulkin J. Professional liability and other career pressures: impact on obstetrician-gynecologists’ career satisfaction. Obstet Gynecol 2004;103:967–73.
18. Reyes JW. Do female physicians capture their scarcity value? The case of OB/GYNs. NBER Working Papers series no. 12528. Cambridge (MA): National Bureau of Economic Research, Inc; 2006.
19. Blau FD, Kahn LM. Gender differences in pay. J Econ Perspect 2000;14:75–99.
20. Ohsfeldt RL, Culler SD. Differences in income between male and female physicians. J Health Econ 1986;5:335–46.
21. Curtis MG. A guest editorial: is “male ob/gyn” a new oxymoron? Obstet Gynecol Surv 2001;56:317–21.