- Email: [email protected]
Trends in the earnings gender gap among dentists, physicians, and lawyers
ORIGINAL CONTRIBUTIONS
Trends in the earnings gender gap among dentists, physicians, and lawyers Thanh An Nguyen Le, PhD; Anthony T. Lo Sasso, PhD; Marko Vujicic, PhD
A
mong the many dramatic changes in the labor market over the last several decades, perhaps none has been more important than the increase in women’s labor force participation, particularly in the professional ranks. Women constituted only 3% of dentists in 1982. The percentage of female dentists increased to 22% in 2004 and is projected to be 28% to 30% by 2020.1-4 In 1985, women made up 24% of dental students; the number increased to 42% in 2004 and 47% in 2014.5 A similar pattern has occurred in medical schools. In 1975, women represented 22.7% of all medical school applicants; by 2011, women represented 47.3% of all medical school applicants.6 Similarly, among first-year law students in 1975, women represented 27% of the total; by 2012, women represented 47% of the total.7 However, sex differences in earnings are still apparent in these professions. Despite the larger role of women in dentistry, female dentists consistently earn less than do their male colleagues.8,9 For physicians, the earnings gap between men and women has been documented since the mid1970s,10 and a rich literature has developed since. A goal of this line of research has been the effort to explain the observed differences in earnings between women and men in the same profession. Differences in professional characteristics including choice of specialty, choice of
ABSTRACT Background. The authors examined the factors associated with sex differences in earnings for 3 professional occupations. Methods. The authors used a multivariate BlinderOaxaca method to decompose the differences in mean earnings across sex. Results. Although mean differences in earnings between men and women narrowed over time, there remained large, unaccountable earnings differences between men and women among all professions after multivariate adjustments. For dentists, the unexplained difference in earnings for women was approximately constant at 62% to 66%. For physicians, the unexplained difference in earnings for women ranged from 52% to 57%. For lawyers, the unexplained difference in earnings for women was the smallest of the 3 professions but also exhibited the most growth, increasing from 34% in 1990 to 45% in 2010. Conclusions. The reduction in the earnings gap is driven largely by a general convergence between men and women in some, but not all, observable characteristics over time. Nevertheless, large unexplained gender gaps in earnings remain for all 3 professions. Practical Implications. Policy makers must use care in efforts to alleviate earnings differences for men and women because measures could make matters worse without a clear understanding of the nature of the factors driving the differences. Key Words. Salary; earnings disparities; professionals; Blinder-Oaxaca decomposition. JADA 2017:148(4):257-262 http://dx.doi.org/10.1016/j.adaj.2017.01.005
Copyright ª 2017 American Dental Association. All rights reserved.
JADA 148(4) http://jada.ada.org
April 2017 257
ORIGINAL CONTRIBUTIONS
entrepreneurial or salaried practice, working hours, productivity, and experience have been identified as contributing explanations for the gender gap in professional earnings. However, results still are mixed as to whether such observable characteristics fully account for differential earnings. For physicians, Baker11 concluded that earnings of female and male physicians converge after controlling for observable characteristics, though investigators in other studies found an unexplained gap varying from 12% to 21%.12-15 The aim of this study is to further the understanding of the gender gap in earnings by studying dentist earnings in relation to earnings of other high-skill occupations with large fractions of women: physicians and lawyers. Legal, dental, and medical professions are useful to compare and contrast because all 3 fields have experienced a striking influx of women during the past few decades and at least at a broad level they represent alternative career paths for high-achieving young adults. We examine the extent to which the gender gap in earnings is linked to changes in characteristics of male and female professionals over a 20-year period and how these changes differ across dentists, physicians, and lawyers. We use an approach common in labor economics to decompose the earnings gap over time. In particular, we explore the extent to which sex differences in earnings can be accounted for by observable factors such as age, hours worked per week, weeks worked per year, and self-employment and the extent to which the remaining unexplained portion of the salary difference changes over time. METHODS
Data source. We used data from the Integrated Public Use Microdata Series census microdata for 1990 and 2000 and the 5-year American Community Survey sample for 2007-2011 (hereafter referred to as 2010).16 We identify physicians, lawyers, and dentists in the data on the basis of reported occupation. In addition to having large numbers of observations, allowing us to identify a sufficient sample of men and women in the 3 occupations of interest, census data contain a number of important characteristics that likely are related to salary. Among the variables available in the census in addition to income, sex, and occupation are age, race or ethnicity, marital status, number of children, hours worked per week, weeks worked per year, and whether the respondent is a business owner. With these variables, we constructed a basic model of earnings for the purposes of decomposing sex differences by using the techniques described here. Analysis. A common method to study outcomes differences across groups is the Blinder-Oaxaca decomposition method by Blinder17 and Oaxaca.18 Broadly speaking, the Blinder-Oaxaca approach allows identification of the difference in earnings of men and women that is attributable to differences in their respective characteristics, including age, work hours, and other
258 JADA 148(4) http://jada.ada.org
April 2017
factors—that is, how much of the observed difference in earnings can be explained if women and men have the same characteristics? Any remaining difference in earnings is considered the residual or unexplained portion. The unexplained portion is always conditional on the characteristics observable to the analyst; however, it could be the case that other important characteristics not available to the analyst, such as family commitments, preferences, and other job characteristics, could account for the difference. The Blinder-Oaxaca method uses a simple earnings model and is assumed to be linear and separable by observable and unobservable characteristics: Yg ¼ X bg þ εg
[1]
in which Yg represents annual (log) earnings of the 2 groups (g), men and women, and X is a vector of observable characteristics that are likely to be related to earnings. The b terms represent the coefficient estimates that express the relationship between the characteristics, X, and the outcome, Y. Separate regressions are run for male (M) and female (F) professionals. The difference in mean earnings between men and women can be written as: b b b b X b Y F Y M ¼ XF XM b þ M F F M
[2]
Variables with bars over them represent means, and variables with hats on them represent regression estimates. The first term on the right-hand side of equation [2] is the explained component of differential earnings that is due to the average differences in observable characteristics of women and men; the difference in earnings that might result, for example, from men being older (thus more experienced and earning more) on average than women. The second term is the unexplained component, which represents the differences in the relationship between a given characteristic (such as age or work hours) and women’s and men’s earnings; for example, an additional year of age might be associated with a greater increase in earnings for men than for women, all else constant. Hence, the second term on the right-hand side of equation [2] shows differences in the coefficient estimates, which in an earnings regression represent how each variable affects earnings. For presentation purposes, the explained difference and the unexplained difference are each summed and presented as a percentage of the total difference. We conducted the Blinder-Oaxaca decomposition by using the Oaxaca command in Stata 13.1 (StataCorp).19 RESULTS
Unadjusted sex differences among dentists, physicians, and lawyers. Tables 1 through 3 display unadjusted differences in mean earnings over the 3 periods for
ORIGINAL CONTRIBUTIONS
TABLE 1 each profession. The results indicate Dentist characteristics over time, according to sex, that in 1990 men 1990-2010.* made approximately twice as much as DENTIST CHARACTERISTIC 1990 2000 2010 women in all 3 proMale Female Male Female Male Female fessions. The largest gap Income, $ 143,874 65,744 192,145 105,766 185,192 120,475 was in 1990 among Age, y, Mean 47.22 39.02 49.99 39.16 53.63 42.30 dentists, with the No. of Children 1.11 0.81 1.01 1.05 0.88 1.09 average male dentist White, % 92.76 82.01 89.32 71.02 87.21 66.85 earning $143,900 and Black, % 2.35 4.93 2.39 5.77 2.11 4.79 the average female Other Race, % 4.89 13.05 8.30 23.22 10.68 28.36 dentist earning $65,700 Married, % 85.53 68.86 84.82 70.79 84.98 75.96 (all dollar amounts are Self-Employed, % 79.66 44.71 78.11 50.83 72.87 50.52 in 2011 dollars). SimiNo. of Hours Worked per Week 38.73 34.74 37.63 36.77 35.79 35.66 larly, male physicians Weeks Worked per Year, %† on average earned 1-13 1.89 4.67 1.89 3.07 1.89 2.57 $187,000, whereas fe14-26 2.77 6.90 2.81 4.75 1.41 2.57 male physicians on 27-39 2.24 6.16 1.98 4.37 2.37 3.57 average earned $97,000; 40-47 8.11 10.08 8.92 11.58 9.98 9.71 male lawyers earned 48-49 14.92 11.78 14.85 11.96 12.31 10.33 $138,600 and female 50-52 70.06 60.40 69.54 64.26 72.04 71.24 lawyers earned $73,600 No. of Participants 7,110 973 6,582 1,318 7,297 2,130 in 1990. Over the * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American ensuing 2 decades, the Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for the difference between previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. average earnings † Percentages do not total 100% because of rounding. decreased in relative terms for all 3 occupaTABLE 2 tions. By 2010, male dentists earned $185,200 Physician characteristics over time, according to sex, versus $120,500 for 1990-2010.* women. Male physicians 1990 2000 2010 earned $234,200, and fe- PHYSICIAN CHARACTERISTIC Male Female Male Female Male Female male physicians earned $145,100 in 2010. For Income, $ 187,027 97,020 227,700 127,728 234,201 145,061 lawyers in 2010, men Age, y, Mean 46.71 40.81 48.33 41.74 50.76 43.33 earned $171,300, and No. of Children 1.13 0.88 1.09 0.96 1.01 0.98 women earned $108,800. White, % 87.12 78.37 80.56 71.31 78.94 68.86 In our later analyses, Black, % 2.56 5.10 3.55 6.38 3.12 6.23 we will examine system- Other Race, % 10.32 16.52 15.90 22.31 17.94 24.91 atically the extent to Married, % 84.40 66.88 83.17 69.59 83.98 71.01 which changes in the Self-Employed, % 49.95 26.52 38.75 20.89 34.17 19.94 characteristics of male No. of Hours Worked per Week 51.06 44.21 50.92 45.88 50.07 46.43 and female professionals Weeks Worked per Year, %† over time potentially can 1-13 1.89 2.93 1.90 2.17 1.72 2.21 explain the narrowing of 14-26 3.09 6.56 3.10 6.04 1.57 2.38 the gender gap. Never27-39 2.05 5.08 2.16 3.98 1.86 3.07 theless, it is useful to 40-47 5.66 7.51 6.16 7.50 5.32 6.31 highlight a few descrip48-49 8.35 7.67 8.43 7.65 5.31 5.18 tive trends in key 50-52 78.96 70.25 78.25 72.66 84.22 80.86 explanatory variables. No. of Participants 23,646 6,210 26,679 9,689 31,811 15,551 Table 1 indicates that * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American male dentists were Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real generally several years dollars. older than female † Percentages do not total 100% because of rounding.
JADA 148(4) http://jada.ada.org
April 2017 259
ORIGINAL CONTRIBUTIONS
TABLE 3
each other), physicians exhibited more diversity over the 20-year period as the fraction reporting white race decreased for LAWYER CHARACTERISTIC 1990 2000 2010 both men and women by Male Female Male Female Male Female approximately 10 percentIncome, $ 138,553 73,632 167,740 96,627 171,284 108,831 age points. Age, y, Mean 43.60 36.10 46.80 39.77 50.63 43.24 For lawyers, we again No. of Children 0.98 0.61 0.95 0.78 0.86 0.79 observe many similar White, % 95.97 91.29 94.13 87.41 92.96 85.34 patterns: like dentists, Black, % 2.26 5.03 2.41 5.87 2.63 6.41 male lawyers were Other Race, % 1.77 3.67 3.46 6.72 4.41 8.25 approximately 7 years Married, % 76.39 54.67 77.41 60.69 79.10 62.52 older, and the gap perSelf-Employed, % 47.20 19.51 47.78 24.58 39.68 20.81 sisted over time. Male No. of Hours Worked per Week 45.33 40.11 46.27 41.09 44.15 40.16 lawyers were more likely Weeks Worked per Year, %† to be self-employed, but 1-13 3.30 6.41 2.06 3.81 2.88 3.89 that gap also narrowed 14-26 3.31 7.06 2.39 4.81 1.91 3.20 over time. Male lawyers 27-39 2.20 6.25 1.79 3.54 2.22 3.89 also were more likely to 40-47 3.59 7.24 3.19 5.71 3.11 4.61 work more hours per 48-49 4.10 4.56 3.87 4.18 2.74 2.78 week and more weeks per 50-52 83.51 68.48 86.70 77.95 87.15 81.63 year, but in both cases the No. of Participants 28,234 9,335 30,607 13,056 40,755 21,126 gap narrowed but was not * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American eliminated. Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for Decomposition of the the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. earnings gap among † Percentages do not total 100% because of rounding. dentists, physicians, and lawyers. Table 4 reports dentists, and the difference (approximately 8 years in summary Blinder-Oaxaca decomposition results con1990) grew over time to more than 11 years in 2010. Male trolling for age (and its square), race, marital status, dentists were also much more likely to be self-employed number of children, self-employment status, and hours than were female dentists, but the gap narrowed per week and weeks worked per year. We used the modestly as men became less likely to be self-employed Blinder-Oaxaca method to estimate how much of the and women became more likely to be self-employed. male–female professional earnings differential is Male dentists also tended to work more hours per accounted for by differences in the observable characweek and more weeks per year in 1990, but both differ- teristics included in the model and what fraction remains unexplained. Table 4 indicates that, for dentists, nearly ences were almost completely eliminated by 2010. We observed convergence between men and women in 38% of the earnings differential in 1990 can be accounted marital status and the number of children over the for by differences in characteristics between male and period. Although both male and female dentists were less female dentists. Put differently, if male and female denlikely to report white race over the period, female tists hypothetically were to have the same characteristics, dentists became much more racially and ethnically female dentists would have earned nearly $30,000 more diverse relative to men. in 1990 or approximately $95,700. In 2010, the hypoFor physicians (Table 2), many of the same trends thetical female earnings after removing differences in observed for dentists apply. Like dentists, male physiobservable characteristics would have been cians were generally several years older than female approximately $143,800, compared with male dentist earnings of $185,200. However, there nevertheless rephysicians, and the difference (approximately 7 years) mains an unexplained difference in earnings not did not change much over time. Male physicians were accounted for by the characteristics measured in the also more likely to be self-employed than were female model. The remaining 62% of the salary difference in physicians, although at a much lower rate than that of dentists, but the gap narrowed as both sexes became less 1990 cannot be accounted for by factors in the model. likely to be self-employed. Male physicians also tended to Moreover, although the overall earnings gap narrowed work more hours per week and more weeks per year, but over time, the fraction unaccounted for by observable characteristics remained approximately constant at 64% both differences narrowed over time. In terms of varito 66%. ables that did not change over time (at least relative to
Lawyer characteristics over time, according to sex, 1990-2010.*
260 JADA 148(4) http://jada.ada.org
April 2017
ORIGINAL CONTRIBUTIONS
TABLE 4
Summary Oaxaca-Blinder decomposition of earnings differentials from census and American Community Survey data, 1990-2010.* EARNINGS CHARACTERISTIC
DENTISTS
PHYSICIANS
LAWYERS
1990
2000
2010
1990
2000
2010
1990
2000
2010
Male Earnings, $
143,874
192,145
185,192
187,027
227,700
234,201
138,553
167,740
171,284
Female Earnings, $
65,744
105,766
120,475
97,020
127,728
145,061
73,632
96,627
108,831
Overall Earnings Difference (Male in Relation to Female), %
118.8
81.7
53.7
92.8
78.3
61.5
88.2
73.6
57.4
Explained, %
37.6
33.6
36.1
45.6
47.9
42.7
65.7
60.8
55.0
Unexplained, %
62.4
66.4
63.9
54.4
52.1
57.3
34.3
39.2
45.0
7,720
7,557
8,981
28,152
33,611
44,268
35,922
41,273
57,342
No. of Participants
* Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American Community Survey.16 Each column represents results from 9 separate Oaxaca-Blinder decomposition regressions of log earnings differences between men and women. Variables accounting for differences in the explained row (via regression analysis) are age, age squared, number of children, hours worked per week, race (white, black, other), married indicator, self-employment indicator, and weeks worked per year (1-13, 14-26, 27-39, 40-47, 48-49, 50-52). Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars.
We observed a similar pattern for physicians. As seen previously, the overall earnings gap narrowed over time, but the ability of the observable characteristics to account for the earnings difference was higher for physicians: 43% to 48% of the earnings gap is explained by differences in observable characteristics. This means that in 2010, if women had had the same characteristics as men, they would have earned approximately $38,000 more, or $183,100—closer, but still less than male earnings of $234,200. For lawyers, we observed a slightly different pattern. We observed the same narrowing of the earnings gap, but the ability of the characteristics to explain the difference diminishes over time, from nearly 66% in 1990 to 55% in 2010, leading to a greater unexplained salary difference for lawyers in 2010. Detailed Blinder-Oaxaca regression results are available in the eTable (available online at the end of this article). The detailed results indicate the influence of each individual variable in the regression model in explaining observed differences in earnings. In sum, age differences and differences in hours worked per week between men and women in all 3 professions are large and significant explanatory factors. Weeks worked per year is significant but accounts for only a modest amount of the difference. Self-employment mattered most for dentists, mattered relatively little for physicians, and was associated with lower earnings for lawyers. In the latter case, having lower rates of self-employment contributed to women lawyers making more than their male counterparts. Marital status and the number of children had only small effects. DISCUSSION
The analysis shows that sex differences in earnings have been shrinking over time for physicians, dentists, and lawyers. The reduction in the earnings gap was driven
largely by the fact that there was a general convergence between men and women in the observable characteristics used in our study over time. For physicians, observable characteristics explained 46% of the earnings difference in 1990, 48% in 2000, and 43% in 2010. For dentists, observable characteristics explained 38% of the earnings difference in 1990, 34% in 2000, and 36% in 2010. For lawyers, observable characteristics explained 66% of the earnings difference in 1990, 61% in 2000, and 55% in 2010. In terms of observable characteristics, age and hours worked per week were significant factors in explaining differences in earnings. However, there is still a large component of the earnings gender gap for all 3 professions that is not explained by differences in observable characteristics. For physicians, the unexplained difference in earnings for women was 54% in 1990, 52% in 2000, and 57% in 2010. For dentists, the unexplained difference in earnings for women was approximately constant over the period at 62% to 66%. For lawyers, the unexplained difference in earnings for women exhibited the most consistent growth over time despite being the smallest of the 3 professions: 34% in 1990, 39% in 2000, and 45% in 2010. Our findings are at least partly consistent with the findings of Seabury and colleagues20 who used a different method and different data but a similar time period; they documented persistent unexplained sex differences in salary across a range of medical and other occupations and pointed to the need for further study. Our results raise questions about the source of unexplained earnings differences between the sexes across high-skill professions. It is clear that we cannot control for specialty in any of the professions; for example, female physicians are much more likely to work in lower-paying primary care specialties.15 There are many unobservable characteristics that could explain the persistent earnings gap. Although our models control for
JADA 148(4) http://jada.ada.org
April 2017 261
ORIGINAL CONTRIBUTIONS
age, we do not have the ability to measure employment history and years spent on the job. If women were more likely to spend time out of the labor force in the past, their earnings could differ later. There is long-standing literature on labor economics that characterizes the often challenging allocation decisions women disproportionately face in terms of time devoted to formal labor market production versus home production.21 Moreover, we do not observe other important characteristics relevant for all the professions in our study, including the type of practice, size of the practice, geographic location of the practice, leisure activities of both male and female professionals, time away from work because of illnesses, hours per week (per year) treating patients or working with clients, type of school attended, and on-the-job education or training (continuing education). A more difficult attribute to measure that is often mentioned as a factor is negotiating ability.22,23 Another limitation of our work and of the BlinderOaxaca method generally is that the approach is only as good as the variables available to the analyst. Put differently, our use of the method is limited by the available variables contained in the census data; the unexplained differences we observed by definition represent unmeasured factors. A richer source of data undoubtedly would reduce the unexplained variation we observe. In addition, the Blinder-Oaxaca approach assumes a linear relationship between the explanatory variables and income; it could be that the relationship is more complex. CONCLUSION
Our findings have important implications for future research and policy. Better measurement of attributes, both of the individual person and of jobs, remains key. Policy makers must use care because measures intended to help alleviate earnings differences for men and women could make matters worse, particularly absent a clear understanding of the nature of the factors driving the differences. n SUPPLEMENTAL DATA
Supplemental data related to this article can be found at http://dx.doi.org/10.1016/j.adaj.2017.01.005. Dr. Nguyen Le is a lecturer, International University, Vietnam National University - Ho Chi Minh City, Ho Chi Minh City, Vietnam. Dr. Lo Sasso is a professor, University of Illinois at Chicago, 1603 W. Taylor St, MC 923, Chicago, IL 60612, e-mail [email protected]. Address correspondence to Dr. Lo Sasso. Dr. Vujicic is the chief health economist and the vice president, Health Policy Institute, American Dental Association, Chicago, IL.
262 JADA 148(4) http://jada.ada.org
April 2017
Disclosure. None of the authors reported any disclosures. 1. Munson B, Vujicic M. Number of practicing dentists per capita in the United States will grow steadily. Health Policy Institute Research Brief. American Dental Association. June 2016 (Revised). Available at: http:// www.ada.org/w/media/ADA/Science%20and%20Research/HPI/Files/ HPIBrief_0616_1.pdf?la¼en. Accessed August 3, 2016. 2. US Census Bureau. Statistical abstract of the United States: 2006. Available at: http://census.gov/library/publications/2005/compendia/ statab/125ed.html. Accessed August 3, 2016. 3. American Dental Association. Dental Workforce Model: 1996-2020. Chicago, IL: Health Policy Institute; 1998. 4. Solomon ES. Dental workforce. Dent Clin North Am. 2009;53(3):435-449. 5. American Dental Association. Survey of Dental Education Series, Report 1: Academic Programs, Enrollment, and Graduates. Chicago, IL: Health Policy Institute; 2014. 6. American Association of Medical Colleges. The changing gender composition of U.S. medical school applicants and matriculants. Available at: https://www.aamc.org/download/277026/data/aibvol12_no1.pdf. Accessed July 14, 2016. 7. American Bar Association. First year and total J.D. enrollment by gender 1947-2011. Available at: http://www.americanbar.org/content/dam/ aba/administrative/legal_education_and_admissions_to_the_bar/statistics/ jd_enrollment_1yr_total_gender.authcheckdam.pdf. Accessed July 14, 2016. 8. Brown LJ, Lazar V. Differences in net incomes of male and female owner general practitioners. JADA. 1998;129(3):373-378. 9. Vujicic M, Wall TP, Nasseh K, Munson B. Dentist income levels slow to recover. Health Policy Institute Research Brief. American Dental Association. February 2013. Available at: http://www.ada.org/w/media/ADA/ Science%20and%20Research/HPI/Files/HPIBrief_0213_1.pdf?la¼en. Accessed August 3, 2016. 10. Kehrer BH. Factors affecting the incomes of men and women physicians: an exploratory analysis. J Hum Resour. 1976;11(4):526-545. 11. Baker LC. Differences in earnings between male and female physicians. N Engl J Med. 1996;334(13):960-964. 12. Ohsfeldt RL, Culler SD. Differences in income between male and female physicians. J Health Econ. 1986;5(4):335-346. 13. Hampton MB. Physician’s income and the percent female of the physician’s specialty. Econ Lett. 1991;36(4):425-428. 14. Weeks WB, Wallace TA, Wallace AE. How do race and sex affect the earnings of primary care physicians? Health Aff. 2009;28(2):557-566. 15. 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(2):193-201. 16. Ruggles S, Genadek K, Goeken R, Grover J, Sobek M. Integrated Public Use Microdata Series: Version 6.0 [dataset]. Minneapolis, MN: University of Minnesota, 2015. Available at: http://doi.org/10.18128/ D010.V6.0. Accessed February 6, 2017. 17. Blinder A. Wage discrimination: reduced form and structural estimates. J Hum Resour. 1973;8(4):436-455. 18. Oaxaca R. Male-female wage differentials in urban labor markets. Int Econ Rev (Philadelphia). 1973;14(3):693-709. 19. Jann B. The Blinder-Oaxaca decomposition for linear regression models. Stata J. 2008;8(4):453-479. 20. Seabury SA, Chandra A, Jena AB. Trends in the earnings of male and female health care professionals in the United States, 1987 to 2010. JAMA Intern Med. 2013;173(18):1748-1750. 21. Altonji JG, Blank RM. Race and Gender in the Labor Market. Amsterdam, The Netherlands: Elsevier; 2009. 22. 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(1):84-103. 23. Bowles HR, Babcock L, McGinn KL. Constraints and triggers: situational mechanics of gender in negotiation. J Pers Soc Psychol. 2005; 89(6):951-965.
ORIGINAL CONTRIBUTIONS
eTABLE
Full Blinder-Oaxaca decomposition of earnings differentials from census and American Community Survey data, 1990-2010.* CHARACTERISTIC
DENTISTS
PHYSICIANS
1990
2000
2010
1990
2000
2010
7,720
7,557
8,981
28,152
33,611
44,268
Overall Difference, Mean (Standard Error)
2.469† (0.0796)
2.022† (0.0601)
1.634† (0.0408)
2.006† (0.0273)
1.727† (0.0208)
1.660† (0.0173)
Explained
1.435† (0.0290)
1.288† (0.0216)
1.203† (0.0179)
1.380† (0.0129)
1.301† (0.0098)
1.247† (0.0083)
Unexplained
1.721† (0.0530)
1.570† (0.0472)
1.359† (0.0341)
1.453† (0.0176)
1.327† (0.0148)
1.331† (0.0128)
Age overall
1.166† (0.0134)
1.112† (0.0133)
1.071† (0.0109)
1.163† (0.0071)
1.167† (0.0065)
1.177† (0.0064)
Age
2.136† (0.1310)
2.144† (0.1570)
1.928† (0.1170)
2.273† (0.0711)
2.338† (0.0605)
2.513† (0.0628)
Age squared
0.546† (0.0295)
0.519† (0.0357)
0.556† (0.0326)
0.511† (0.0141)
0.499† (0.0114)
0.468† (0.0103)
No. of Participants
Detailed Explained, Mean (Standard Error)
§
‡
†
†
†
†
Race overall (white )
1.009 (0.0040)
1.032 (0.0063)
1.033 (0.0057)
1.007 (0.0013)
1.012 (0.0014)
1.003¶ (0.0011)
Black
1.005‡ (0.0022)
1.008¶ (0.0027)
1.006¶ (0.0022)
1.003† (0.0008)
1.004† (0.0009)
1.003† (0.0008)
1.004 (0.0032)
1.024† (0.0055)
1.027† (0.0052)
1.005† (0.0009)
1.008† (0.0010)
1.000 (0.0006)
Married
1.020† (0.0048)
1.013¶ (0.0043)
1.010† (0.0027)
1.016† (0.0024)
1.012† (0.0018)
1.009† (0.0016)
No. of children
1.015† (0.0032)
1.000 (0.0007)
0.997 (0.0016)
1.019† (0.0017)
1.011† (0.0014)
1.008† (0.0012)
Self-employed
1.053† (0.0081)
1.063† (0.0077)
1.034† (0.0052)
1.055† (0.0029)
1.021† (0.0020)
1.000 (0.0013)
Hours worked per week
1.035† (0.0055)
1.017† (0.0037)
1.021† (0.0040)
1.028† (0.0023)
1.022† (0.0018)
1.023† (0.0017)
Weeks worked per year (1-13§)
1.083† (0.0120)
1.026† (0.0067)
1.022† (0.0066)
1.049† (0.0043)
1.032† (0.0031)
1.015† (0.0024)
Other race
14-26
0.997 (0.0042)
0.998 (0.0020)
0.998 (0.0012)
1.004 (0.0018)
1.001 (0.0016)
0.997† (0.0008)
27-39
0.983¶ (0.0053)
0.989¶ (0.0042)
0.990¶ (0.0037)
0.994† (0.0018)
0.996† (0.0012)
0.992† (0.0014)
40-47
0.986 (0.0092)
0.976¶ (0.0086)
1.000 (0.0098)
0.986† (0.0029)
0.993¶ (0.0023)
0.989† (0.0029)
‡
‡
‡
‡
‡
48-49
1.026 (0.0116)
1.024 (0.0097)
1.025 (0.0102)
1.006 (0.0030)
1.003 (0.0026)
1.004 (0.0026)
50-52
1.092† (0.0191)
1.040¶ (0.0133)
1.010 (0.0148)
1.061† (0.0069)
1.038† (0.0051)
1.035† (0.0054)
* Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American Community Survey.16 Each column represents results from 9 separate Oaxaca-Blinder decomposition regressions of log earnings differences between men and women. Coefficient estimates are relative risk ratios. Variables accounting for differences in the explained row (via regression analysis) are age, age squared, number of children, hours worked per week, race (white, black, or other), married indicator, self-employment indicator, and weeks worked per year (1-13, 1426, 27-39, 40-47, 48-49, and 50-52). Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. † P < .01. ‡ P < .1. § Reference category. ¶ P < .05.
JADA 148(4) http://jada.ada.org
April 2017 262.e1
ORIGINAL CONTRIBUTIONS
eTABLE (CONTINUED)
LAWYERS 1990
2000
2010
35,922
41,273
57,342
1.795† (0.0198)
1.578† (0.0156)
1.469† (0.0128)
1.453† (0.0117)
1.313† (0.0082)
1.233† (0.0068)
†
1.236 (0.0116)
1.202 (0.0112)
1.191† (0.0096)
1.188† (0.0053)
1.153† (0.0045)
1.124† (0.0036)
1.654† (0.0308)
1.448† (0.0243)
1.343† (0.0190)
†
†
0.718 (0.0117)
0.796 (0.0125)
0.837† (0.0113)
1.008† (0.0011)
1.011† (0.0013)
1.010† (0.0012)
1.006† (0.0010)
1.008† (0.0010)
1.009† (0.0010)
†
1.002 (0.0005) †
†
†
1.001‡ (0.0006)
†
1.003 (0.0007)
1.026 (0.0025)
1.026 (0.0020)
1.022† (0.0017)
1.016† (0.0017)
1.013† (0.0013)
1.007† (0.0008)
1.001 (0.0024)
0.978† (0.0022)
0.969† (0.0017)
†
1.063 (0.0032) †
†
1.063† (0.0029)
†
1.024† (0.0023)
1.064 (0.0031)
1.093 (0.0049)
1.042 (0.0028)
0.996† (0.0012)
0.999 (0.0010)
1.000 (0.0005)
0.986† (0.0017)
0.996† (0.0009)
0.994† (0.0009)
0.977† (0.0024)
0.987† (0.0017)
0.990† (0.0014)
0.996 (0.0024)
0.997 (0.0017)
1.000 (0.0013)
1.145† (0.0079)
1.064† (0.0049)
1.041† (0.0037)
262.e2 JADA 148(4) http://jada.ada.org
April 2017
Trends in the earnings gender gap among dentists, physicians, and lawyers Thanh An Nguyen Le, PhD; Anthony T. Lo Sasso, PhD; Marko Vujicic, PhD
A
mong the many dramatic changes in the labor market over the last several decades, perhaps none has been more important than the increase in women’s labor force participation, particularly in the professional ranks. Women constituted only 3% of dentists in 1982. The percentage of female dentists increased to 22% in 2004 and is projected to be 28% to 30% by 2020.1-4 In 1985, women made up 24% of dental students; the number increased to 42% in 2004 and 47% in 2014.5 A similar pattern has occurred in medical schools. In 1975, women represented 22.7% of all medical school applicants; by 2011, women represented 47.3% of all medical school applicants.6 Similarly, among first-year law students in 1975, women represented 27% of the total; by 2012, women represented 47% of the total.7 However, sex differences in earnings are still apparent in these professions. Despite the larger role of women in dentistry, female dentists consistently earn less than do their male colleagues.8,9 For physicians, the earnings gap between men and women has been documented since the mid1970s,10 and a rich literature has developed since. A goal of this line of research has been the effort to explain the observed differences in earnings between women and men in the same profession. Differences in professional characteristics including choice of specialty, choice of
ABSTRACT Background. The authors examined the factors associated with sex differences in earnings for 3 professional occupations. Methods. The authors used a multivariate BlinderOaxaca method to decompose the differences in mean earnings across sex. Results. Although mean differences in earnings between men and women narrowed over time, there remained large, unaccountable earnings differences between men and women among all professions after multivariate adjustments. For dentists, the unexplained difference in earnings for women was approximately constant at 62% to 66%. For physicians, the unexplained difference in earnings for women ranged from 52% to 57%. For lawyers, the unexplained difference in earnings for women was the smallest of the 3 professions but also exhibited the most growth, increasing from 34% in 1990 to 45% in 2010. Conclusions. The reduction in the earnings gap is driven largely by a general convergence between men and women in some, but not all, observable characteristics over time. Nevertheless, large unexplained gender gaps in earnings remain for all 3 professions. Practical Implications. Policy makers must use care in efforts to alleviate earnings differences for men and women because measures could make matters worse without a clear understanding of the nature of the factors driving the differences. Key Words. Salary; earnings disparities; professionals; Blinder-Oaxaca decomposition. JADA 2017:148(4):257-262 http://dx.doi.org/10.1016/j.adaj.2017.01.005
Copyright ª 2017 American Dental Association. All rights reserved.
JADA 148(4) http://jada.ada.org
April 2017 257
ORIGINAL CONTRIBUTIONS
entrepreneurial or salaried practice, working hours, productivity, and experience have been identified as contributing explanations for the gender gap in professional earnings. However, results still are mixed as to whether such observable characteristics fully account for differential earnings. For physicians, Baker11 concluded that earnings of female and male physicians converge after controlling for observable characteristics, though investigators in other studies found an unexplained gap varying from 12% to 21%.12-15 The aim of this study is to further the understanding of the gender gap in earnings by studying dentist earnings in relation to earnings of other high-skill occupations with large fractions of women: physicians and lawyers. Legal, dental, and medical professions are useful to compare and contrast because all 3 fields have experienced a striking influx of women during the past few decades and at least at a broad level they represent alternative career paths for high-achieving young adults. We examine the extent to which the gender gap in earnings is linked to changes in characteristics of male and female professionals over a 20-year period and how these changes differ across dentists, physicians, and lawyers. We use an approach common in labor economics to decompose the earnings gap over time. In particular, we explore the extent to which sex differences in earnings can be accounted for by observable factors such as age, hours worked per week, weeks worked per year, and self-employment and the extent to which the remaining unexplained portion of the salary difference changes over time. METHODS
Data source. We used data from the Integrated Public Use Microdata Series census microdata for 1990 and 2000 and the 5-year American Community Survey sample for 2007-2011 (hereafter referred to as 2010).16 We identify physicians, lawyers, and dentists in the data on the basis of reported occupation. In addition to having large numbers of observations, allowing us to identify a sufficient sample of men and women in the 3 occupations of interest, census data contain a number of important characteristics that likely are related to salary. Among the variables available in the census in addition to income, sex, and occupation are age, race or ethnicity, marital status, number of children, hours worked per week, weeks worked per year, and whether the respondent is a business owner. With these variables, we constructed a basic model of earnings for the purposes of decomposing sex differences by using the techniques described here. Analysis. A common method to study outcomes differences across groups is the Blinder-Oaxaca decomposition method by Blinder17 and Oaxaca.18 Broadly speaking, the Blinder-Oaxaca approach allows identification of the difference in earnings of men and women that is attributable to differences in their respective characteristics, including age, work hours, and other
258 JADA 148(4) http://jada.ada.org
April 2017
factors—that is, how much of the observed difference in earnings can be explained if women and men have the same characteristics? Any remaining difference in earnings is considered the residual or unexplained portion. The unexplained portion is always conditional on the characteristics observable to the analyst; however, it could be the case that other important characteristics not available to the analyst, such as family commitments, preferences, and other job characteristics, could account for the difference. The Blinder-Oaxaca method uses a simple earnings model and is assumed to be linear and separable by observable and unobservable characteristics: Yg ¼ X bg þ εg
[1]
in which Yg represents annual (log) earnings of the 2 groups (g), men and women, and X is a vector of observable characteristics that are likely to be related to earnings. The b terms represent the coefficient estimates that express the relationship between the characteristics, X, and the outcome, Y. Separate regressions are run for male (M) and female (F) professionals. The difference in mean earnings between men and women can be written as: b b b b X b Y F Y M ¼ XF XM b þ M F F M
[2]
Variables with bars over them represent means, and variables with hats on them represent regression estimates. The first term on the right-hand side of equation [2] is the explained component of differential earnings that is due to the average differences in observable characteristics of women and men; the difference in earnings that might result, for example, from men being older (thus more experienced and earning more) on average than women. The second term is the unexplained component, which represents the differences in the relationship between a given characteristic (such as age or work hours) and women’s and men’s earnings; for example, an additional year of age might be associated with a greater increase in earnings for men than for women, all else constant. Hence, the second term on the right-hand side of equation [2] shows differences in the coefficient estimates, which in an earnings regression represent how each variable affects earnings. For presentation purposes, the explained difference and the unexplained difference are each summed and presented as a percentage of the total difference. We conducted the Blinder-Oaxaca decomposition by using the Oaxaca command in Stata 13.1 (StataCorp).19 RESULTS
Unadjusted sex differences among dentists, physicians, and lawyers. Tables 1 through 3 display unadjusted differences in mean earnings over the 3 periods for
ORIGINAL CONTRIBUTIONS
TABLE 1 each profession. The results indicate Dentist characteristics over time, according to sex, that in 1990 men 1990-2010.* made approximately twice as much as DENTIST CHARACTERISTIC 1990 2000 2010 women in all 3 proMale Female Male Female Male Female fessions. The largest gap Income, $ 143,874 65,744 192,145 105,766 185,192 120,475 was in 1990 among Age, y, Mean 47.22 39.02 49.99 39.16 53.63 42.30 dentists, with the No. of Children 1.11 0.81 1.01 1.05 0.88 1.09 average male dentist White, % 92.76 82.01 89.32 71.02 87.21 66.85 earning $143,900 and Black, % 2.35 4.93 2.39 5.77 2.11 4.79 the average female Other Race, % 4.89 13.05 8.30 23.22 10.68 28.36 dentist earning $65,700 Married, % 85.53 68.86 84.82 70.79 84.98 75.96 (all dollar amounts are Self-Employed, % 79.66 44.71 78.11 50.83 72.87 50.52 in 2011 dollars). SimiNo. of Hours Worked per Week 38.73 34.74 37.63 36.77 35.79 35.66 larly, male physicians Weeks Worked per Year, %† on average earned 1-13 1.89 4.67 1.89 3.07 1.89 2.57 $187,000, whereas fe14-26 2.77 6.90 2.81 4.75 1.41 2.57 male physicians on 27-39 2.24 6.16 1.98 4.37 2.37 3.57 average earned $97,000; 40-47 8.11 10.08 8.92 11.58 9.98 9.71 male lawyers earned 48-49 14.92 11.78 14.85 11.96 12.31 10.33 $138,600 and female 50-52 70.06 60.40 69.54 64.26 72.04 71.24 lawyers earned $73,600 No. of Participants 7,110 973 6,582 1,318 7,297 2,130 in 1990. Over the * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American ensuing 2 decades, the Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for the difference between previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. average earnings † Percentages do not total 100% because of rounding. decreased in relative terms for all 3 occupaTABLE 2 tions. By 2010, male dentists earned $185,200 Physician characteristics over time, according to sex, versus $120,500 for 1990-2010.* women. Male physicians 1990 2000 2010 earned $234,200, and fe- PHYSICIAN CHARACTERISTIC Male Female Male Female Male Female male physicians earned $145,100 in 2010. For Income, $ 187,027 97,020 227,700 127,728 234,201 145,061 lawyers in 2010, men Age, y, Mean 46.71 40.81 48.33 41.74 50.76 43.33 earned $171,300, and No. of Children 1.13 0.88 1.09 0.96 1.01 0.98 women earned $108,800. White, % 87.12 78.37 80.56 71.31 78.94 68.86 In our later analyses, Black, % 2.56 5.10 3.55 6.38 3.12 6.23 we will examine system- Other Race, % 10.32 16.52 15.90 22.31 17.94 24.91 atically the extent to Married, % 84.40 66.88 83.17 69.59 83.98 71.01 which changes in the Self-Employed, % 49.95 26.52 38.75 20.89 34.17 19.94 characteristics of male No. of Hours Worked per Week 51.06 44.21 50.92 45.88 50.07 46.43 and female professionals Weeks Worked per Year, %† over time potentially can 1-13 1.89 2.93 1.90 2.17 1.72 2.21 explain the narrowing of 14-26 3.09 6.56 3.10 6.04 1.57 2.38 the gender gap. Never27-39 2.05 5.08 2.16 3.98 1.86 3.07 theless, it is useful to 40-47 5.66 7.51 6.16 7.50 5.32 6.31 highlight a few descrip48-49 8.35 7.67 8.43 7.65 5.31 5.18 tive trends in key 50-52 78.96 70.25 78.25 72.66 84.22 80.86 explanatory variables. No. of Participants 23,646 6,210 26,679 9,689 31,811 15,551 Table 1 indicates that * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American male dentists were Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real generally several years dollars. older than female † Percentages do not total 100% because of rounding.
JADA 148(4) http://jada.ada.org
April 2017 259
ORIGINAL CONTRIBUTIONS
TABLE 3
each other), physicians exhibited more diversity over the 20-year period as the fraction reporting white race decreased for LAWYER CHARACTERISTIC 1990 2000 2010 both men and women by Male Female Male Female Male Female approximately 10 percentIncome, $ 138,553 73,632 167,740 96,627 171,284 108,831 age points. Age, y, Mean 43.60 36.10 46.80 39.77 50.63 43.24 For lawyers, we again No. of Children 0.98 0.61 0.95 0.78 0.86 0.79 observe many similar White, % 95.97 91.29 94.13 87.41 92.96 85.34 patterns: like dentists, Black, % 2.26 5.03 2.41 5.87 2.63 6.41 male lawyers were Other Race, % 1.77 3.67 3.46 6.72 4.41 8.25 approximately 7 years Married, % 76.39 54.67 77.41 60.69 79.10 62.52 older, and the gap perSelf-Employed, % 47.20 19.51 47.78 24.58 39.68 20.81 sisted over time. Male No. of Hours Worked per Week 45.33 40.11 46.27 41.09 44.15 40.16 lawyers were more likely Weeks Worked per Year, %† to be self-employed, but 1-13 3.30 6.41 2.06 3.81 2.88 3.89 that gap also narrowed 14-26 3.31 7.06 2.39 4.81 1.91 3.20 over time. Male lawyers 27-39 2.20 6.25 1.79 3.54 2.22 3.89 also were more likely to 40-47 3.59 7.24 3.19 5.71 3.11 4.61 work more hours per 48-49 4.10 4.56 3.87 4.18 2.74 2.78 week and more weeks per 50-52 83.51 68.48 86.70 77.95 87.15 81.63 year, but in both cases the No. of Participants 28,234 9,335 30,607 13,056 40,755 21,126 gap narrowed but was not * Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American eliminated. Community Survey.16 Income indicates each respondent’s total pretax personal income from all sources for Decomposition of the the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. earnings gap among † Percentages do not total 100% because of rounding. dentists, physicians, and lawyers. Table 4 reports dentists, and the difference (approximately 8 years in summary Blinder-Oaxaca decomposition results con1990) grew over time to more than 11 years in 2010. Male trolling for age (and its square), race, marital status, dentists were also much more likely to be self-employed number of children, self-employment status, and hours than were female dentists, but the gap narrowed per week and weeks worked per year. We used the modestly as men became less likely to be self-employed Blinder-Oaxaca method to estimate how much of the and women became more likely to be self-employed. male–female professional earnings differential is Male dentists also tended to work more hours per accounted for by differences in the observable characweek and more weeks per year in 1990, but both differ- teristics included in the model and what fraction remains unexplained. Table 4 indicates that, for dentists, nearly ences were almost completely eliminated by 2010. We observed convergence between men and women in 38% of the earnings differential in 1990 can be accounted marital status and the number of children over the for by differences in characteristics between male and period. Although both male and female dentists were less female dentists. Put differently, if male and female denlikely to report white race over the period, female tists hypothetically were to have the same characteristics, dentists became much more racially and ethnically female dentists would have earned nearly $30,000 more diverse relative to men. in 1990 or approximately $95,700. In 2010, the hypoFor physicians (Table 2), many of the same trends thetical female earnings after removing differences in observed for dentists apply. Like dentists, male physiobservable characteristics would have been cians were generally several years older than female approximately $143,800, compared with male dentist earnings of $185,200. However, there nevertheless rephysicians, and the difference (approximately 7 years) mains an unexplained difference in earnings not did not change much over time. Male physicians were accounted for by the characteristics measured in the also more likely to be self-employed than were female model. The remaining 62% of the salary difference in physicians, although at a much lower rate than that of dentists, but the gap narrowed as both sexes became less 1990 cannot be accounted for by factors in the model. likely to be self-employed. Male physicians also tended to Moreover, although the overall earnings gap narrowed work more hours per week and more weeks per year, but over time, the fraction unaccounted for by observable characteristics remained approximately constant at 64% both differences narrowed over time. In terms of varito 66%. ables that did not change over time (at least relative to
Lawyer characteristics over time, according to sex, 1990-2010.*
260 JADA 148(4) http://jada.ada.org
April 2017
ORIGINAL CONTRIBUTIONS
TABLE 4
Summary Oaxaca-Blinder decomposition of earnings differentials from census and American Community Survey data, 1990-2010.* EARNINGS CHARACTERISTIC
DENTISTS
PHYSICIANS
LAWYERS
1990
2000
2010
1990
2000
2010
1990
2000
2010
Male Earnings, $
143,874
192,145
185,192
187,027
227,700
234,201
138,553
167,740
171,284
Female Earnings, $
65,744
105,766
120,475
97,020
127,728
145,061
73,632
96,627
108,831
Overall Earnings Difference (Male in Relation to Female), %
118.8
81.7
53.7
92.8
78.3
61.5
88.2
73.6
57.4
Explained, %
37.6
33.6
36.1
45.6
47.9
42.7
65.7
60.8
55.0
Unexplained, %
62.4
66.4
63.9
54.4
52.1
57.3
34.3
39.2
45.0
7,720
7,557
8,981
28,152
33,611
44,268
35,922
41,273
57,342
No. of Participants
* Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American Community Survey.16 Each column represents results from 9 separate Oaxaca-Blinder decomposition regressions of log earnings differences between men and women. Variables accounting for differences in the explained row (via regression analysis) are age, age squared, number of children, hours worked per week, race (white, black, other), married indicator, self-employment indicator, and weeks worked per year (1-13, 14-26, 27-39, 40-47, 48-49, 50-52). Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars.
We observed a similar pattern for physicians. As seen previously, the overall earnings gap narrowed over time, but the ability of the observable characteristics to account for the earnings difference was higher for physicians: 43% to 48% of the earnings gap is explained by differences in observable characteristics. This means that in 2010, if women had had the same characteristics as men, they would have earned approximately $38,000 more, or $183,100—closer, but still less than male earnings of $234,200. For lawyers, we observed a slightly different pattern. We observed the same narrowing of the earnings gap, but the ability of the characteristics to explain the difference diminishes over time, from nearly 66% in 1990 to 55% in 2010, leading to a greater unexplained salary difference for lawyers in 2010. Detailed Blinder-Oaxaca regression results are available in the eTable (available online at the end of this article). The detailed results indicate the influence of each individual variable in the regression model in explaining observed differences in earnings. In sum, age differences and differences in hours worked per week between men and women in all 3 professions are large and significant explanatory factors. Weeks worked per year is significant but accounts for only a modest amount of the difference. Self-employment mattered most for dentists, mattered relatively little for physicians, and was associated with lower earnings for lawyers. In the latter case, having lower rates of self-employment contributed to women lawyers making more than their male counterparts. Marital status and the number of children had only small effects. DISCUSSION
The analysis shows that sex differences in earnings have been shrinking over time for physicians, dentists, and lawyers. The reduction in the earnings gap was driven
largely by the fact that there was a general convergence between men and women in the observable characteristics used in our study over time. For physicians, observable characteristics explained 46% of the earnings difference in 1990, 48% in 2000, and 43% in 2010. For dentists, observable characteristics explained 38% of the earnings difference in 1990, 34% in 2000, and 36% in 2010. For lawyers, observable characteristics explained 66% of the earnings difference in 1990, 61% in 2000, and 55% in 2010. In terms of observable characteristics, age and hours worked per week were significant factors in explaining differences in earnings. However, there is still a large component of the earnings gender gap for all 3 professions that is not explained by differences in observable characteristics. For physicians, the unexplained difference in earnings for women was 54% in 1990, 52% in 2000, and 57% in 2010. For dentists, the unexplained difference in earnings for women was approximately constant over the period at 62% to 66%. For lawyers, the unexplained difference in earnings for women exhibited the most consistent growth over time despite being the smallest of the 3 professions: 34% in 1990, 39% in 2000, and 45% in 2010. Our findings are at least partly consistent with the findings of Seabury and colleagues20 who used a different method and different data but a similar time period; they documented persistent unexplained sex differences in salary across a range of medical and other occupations and pointed to the need for further study. Our results raise questions about the source of unexplained earnings differences between the sexes across high-skill professions. It is clear that we cannot control for specialty in any of the professions; for example, female physicians are much more likely to work in lower-paying primary care specialties.15 There are many unobservable characteristics that could explain the persistent earnings gap. Although our models control for
JADA 148(4) http://jada.ada.org
April 2017 261
ORIGINAL CONTRIBUTIONS
age, we do not have the ability to measure employment history and years spent on the job. If women were more likely to spend time out of the labor force in the past, their earnings could differ later. There is long-standing literature on labor economics that characterizes the often challenging allocation decisions women disproportionately face in terms of time devoted to formal labor market production versus home production.21 Moreover, we do not observe other important characteristics relevant for all the professions in our study, including the type of practice, size of the practice, geographic location of the practice, leisure activities of both male and female professionals, time away from work because of illnesses, hours per week (per year) treating patients or working with clients, type of school attended, and on-the-job education or training (continuing education). A more difficult attribute to measure that is often mentioned as a factor is negotiating ability.22,23 Another limitation of our work and of the BlinderOaxaca method generally is that the approach is only as good as the variables available to the analyst. Put differently, our use of the method is limited by the available variables contained in the census data; the unexplained differences we observed by definition represent unmeasured factors. A richer source of data undoubtedly would reduce the unexplained variation we observe. In addition, the Blinder-Oaxaca approach assumes a linear relationship between the explanatory variables and income; it could be that the relationship is more complex. CONCLUSION
Our findings have important implications for future research and policy. Better measurement of attributes, both of the individual person and of jobs, remains key. Policy makers must use care because measures intended to help alleviate earnings differences for men and women could make matters worse, particularly absent a clear understanding of the nature of the factors driving the differences. n SUPPLEMENTAL DATA
Supplemental data related to this article can be found at http://dx.doi.org/10.1016/j.adaj.2017.01.005. Dr. Nguyen Le is a lecturer, International University, Vietnam National University - Ho Chi Minh City, Ho Chi Minh City, Vietnam. Dr. Lo Sasso is a professor, University of Illinois at Chicago, 1603 W. Taylor St, MC 923, Chicago, IL 60612, e-mail [email protected]. Address correspondence to Dr. Lo Sasso. Dr. Vujicic is the chief health economist and the vice president, Health Policy Institute, American Dental Association, Chicago, IL.
262 JADA 148(4) http://jada.ada.org
April 2017
Disclosure. None of the authors reported any disclosures. 1. Munson B, Vujicic M. Number of practicing dentists per capita in the United States will grow steadily. Health Policy Institute Research Brief. American Dental Association. June 2016 (Revised). Available at: http:// www.ada.org/w/media/ADA/Science%20and%20Research/HPI/Files/ HPIBrief_0616_1.pdf?la¼en. Accessed August 3, 2016. 2. US Census Bureau. Statistical abstract of the United States: 2006. Available at: http://census.gov/library/publications/2005/compendia/ statab/125ed.html. Accessed August 3, 2016. 3. American Dental Association. Dental Workforce Model: 1996-2020. Chicago, IL: Health Policy Institute; 1998. 4. Solomon ES. Dental workforce. Dent Clin North Am. 2009;53(3):435-449. 5. American Dental Association. Survey of Dental Education Series, Report 1: Academic Programs, Enrollment, and Graduates. Chicago, IL: Health Policy Institute; 2014. 6. American Association of Medical Colleges. The changing gender composition of U.S. medical school applicants and matriculants. Available at: https://www.aamc.org/download/277026/data/aibvol12_no1.pdf. Accessed July 14, 2016. 7. American Bar Association. First year and total J.D. enrollment by gender 1947-2011. Available at: http://www.americanbar.org/content/dam/ aba/administrative/legal_education_and_admissions_to_the_bar/statistics/ jd_enrollment_1yr_total_gender.authcheckdam.pdf. Accessed July 14, 2016. 8. Brown LJ, Lazar V. Differences in net incomes of male and female owner general practitioners. JADA. 1998;129(3):373-378. 9. Vujicic M, Wall TP, Nasseh K, Munson B. Dentist income levels slow to recover. Health Policy Institute Research Brief. American Dental Association. February 2013. Available at: http://www.ada.org/w/media/ADA/ Science%20and%20Research/HPI/Files/HPIBrief_0213_1.pdf?la¼en. Accessed August 3, 2016. 10. Kehrer BH. Factors affecting the incomes of men and women physicians: an exploratory analysis. J Hum Resour. 1976;11(4):526-545. 11. Baker LC. Differences in earnings between male and female physicians. N Engl J Med. 1996;334(13):960-964. 12. Ohsfeldt RL, Culler SD. Differences in income between male and female physicians. J Health Econ. 1986;5(4):335-346. 13. Hampton MB. Physician’s income and the percent female of the physician’s specialty. Econ Lett. 1991;36(4):425-428. 14. Weeks WB, Wallace TA, Wallace AE. How do race and sex affect the earnings of primary care physicians? Health Aff. 2009;28(2):557-566. 15. 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(2):193-201. 16. Ruggles S, Genadek K, Goeken R, Grover J, Sobek M. Integrated Public Use Microdata Series: Version 6.0 [dataset]. Minneapolis, MN: University of Minnesota, 2015. Available at: http://doi.org/10.18128/ D010.V6.0. Accessed February 6, 2017. 17. Blinder A. Wage discrimination: reduced form and structural estimates. J Hum Resour. 1973;8(4):436-455. 18. Oaxaca R. Male-female wage differentials in urban labor markets. Int Econ Rev (Philadelphia). 1973;14(3):693-709. 19. Jann B. The Blinder-Oaxaca decomposition for linear regression models. Stata J. 2008;8(4):453-479. 20. Seabury SA, Chandra A, Jena AB. Trends in the earnings of male and female health care professionals in the United States, 1987 to 2010. JAMA Intern Med. 2013;173(18):1748-1750. 21. Altonji JG, Blank RM. Race and Gender in the Labor Market. Amsterdam, The Netherlands: Elsevier; 2009. 22. 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(1):84-103. 23. Bowles HR, Babcock L, McGinn KL. Constraints and triggers: situational mechanics of gender in negotiation. J Pers Soc Psychol. 2005; 89(6):951-965.
ORIGINAL CONTRIBUTIONS
eTABLE
Full Blinder-Oaxaca decomposition of earnings differentials from census and American Community Survey data, 1990-2010.* CHARACTERISTIC
DENTISTS
PHYSICIANS
1990
2000
2010
1990
2000
2010
7,720
7,557
8,981
28,152
33,611
44,268
Overall Difference, Mean (Standard Error)
2.469† (0.0796)
2.022† (0.0601)
1.634† (0.0408)
2.006† (0.0273)
1.727† (0.0208)
1.660† (0.0173)
Explained
1.435† (0.0290)
1.288† (0.0216)
1.203† (0.0179)
1.380† (0.0129)
1.301† (0.0098)
1.247† (0.0083)
Unexplained
1.721† (0.0530)
1.570† (0.0472)
1.359† (0.0341)
1.453† (0.0176)
1.327† (0.0148)
1.331† (0.0128)
Age overall
1.166† (0.0134)
1.112† (0.0133)
1.071† (0.0109)
1.163† (0.0071)
1.167† (0.0065)
1.177† (0.0064)
Age
2.136† (0.1310)
2.144† (0.1570)
1.928† (0.1170)
2.273† (0.0711)
2.338† (0.0605)
2.513† (0.0628)
Age squared
0.546† (0.0295)
0.519† (0.0357)
0.556† (0.0326)
0.511† (0.0141)
0.499† (0.0114)
0.468† (0.0103)
No. of Participants
Detailed Explained, Mean (Standard Error)
§
‡
†
†
†
†
Race overall (white )
1.009 (0.0040)
1.032 (0.0063)
1.033 (0.0057)
1.007 (0.0013)
1.012 (0.0014)
1.003¶ (0.0011)
Black
1.005‡ (0.0022)
1.008¶ (0.0027)
1.006¶ (0.0022)
1.003† (0.0008)
1.004† (0.0009)
1.003† (0.0008)
1.004 (0.0032)
1.024† (0.0055)
1.027† (0.0052)
1.005† (0.0009)
1.008† (0.0010)
1.000 (0.0006)
Married
1.020† (0.0048)
1.013¶ (0.0043)
1.010† (0.0027)
1.016† (0.0024)
1.012† (0.0018)
1.009† (0.0016)
No. of children
1.015† (0.0032)
1.000 (0.0007)
0.997 (0.0016)
1.019† (0.0017)
1.011† (0.0014)
1.008† (0.0012)
Self-employed
1.053† (0.0081)
1.063† (0.0077)
1.034† (0.0052)
1.055† (0.0029)
1.021† (0.0020)
1.000 (0.0013)
Hours worked per week
1.035† (0.0055)
1.017† (0.0037)
1.021† (0.0040)
1.028† (0.0023)
1.022† (0.0018)
1.023† (0.0017)
Weeks worked per year (1-13§)
1.083† (0.0120)
1.026† (0.0067)
1.022† (0.0066)
1.049† (0.0043)
1.032† (0.0031)
1.015† (0.0024)
Other race
14-26
0.997 (0.0042)
0.998 (0.0020)
0.998 (0.0012)
1.004 (0.0018)
1.001 (0.0016)
0.997† (0.0008)
27-39
0.983¶ (0.0053)
0.989¶ (0.0042)
0.990¶ (0.0037)
0.994† (0.0018)
0.996† (0.0012)
0.992† (0.0014)
40-47
0.986 (0.0092)
0.976¶ (0.0086)
1.000 (0.0098)
0.986† (0.0029)
0.993¶ (0.0023)
0.989† (0.0029)
‡
‡
‡
‡
‡
48-49
1.026 (0.0116)
1.024 (0.0097)
1.025 (0.0102)
1.006 (0.0030)
1.003 (0.0026)
1.004 (0.0026)
50-52
1.092† (0.0191)
1.040¶ (0.0133)
1.010 (0.0148)
1.061† (0.0069)
1.038† (0.0051)
1.035† (0.0054)
* Data from 1990 and 2000 Integrated Public Use Microdata Series census microdata and 2007-2011 American Community Survey.16 Each column represents results from 9 separate Oaxaca-Blinder decomposition regressions of log earnings differences between men and women. Coefficient estimates are relative risk ratios. Variables accounting for differences in the explained row (via regression analysis) are age, age squared, number of children, hours worked per week, race (white, black, or other), married indicator, self-employment indicator, and weeks worked per year (1-13, 1426, 27-39, 40-47, 48-49, and 50-52). Income indicates each respondent’s total pretax personal income from all sources for the previous calendar year. Amounts are expressed as they were reported to the interviewer, adjusted to 2011 real dollars. † P < .01. ‡ P < .1. § Reference category. ¶ P < .05.
JADA 148(4) http://jada.ada.org
April 2017 262.e1
ORIGINAL CONTRIBUTIONS
eTABLE (CONTINUED)
LAWYERS 1990
2000
2010
35,922
41,273
57,342
1.795† (0.0198)
1.578† (0.0156)
1.469† (0.0128)
1.453† (0.0117)
1.313† (0.0082)
1.233† (0.0068)
†
1.236 (0.0116)
1.202 (0.0112)
1.191† (0.0096)
1.188† (0.0053)
1.153† (0.0045)
1.124† (0.0036)
1.654† (0.0308)
1.448† (0.0243)
1.343† (0.0190)
†
†
0.718 (0.0117)
0.796 (0.0125)
0.837† (0.0113)
1.008† (0.0011)
1.011† (0.0013)
1.010† (0.0012)
1.006† (0.0010)
1.008† (0.0010)
1.009† (0.0010)
†
1.002 (0.0005) †
†
†
1.001‡ (0.0006)
†
1.003 (0.0007)
1.026 (0.0025)
1.026 (0.0020)
1.022† (0.0017)
1.016† (0.0017)
1.013† (0.0013)
1.007† (0.0008)
1.001 (0.0024)
0.978† (0.0022)
0.969† (0.0017)
†
1.063 (0.0032) †
†
1.063† (0.0029)
†
1.024† (0.0023)
1.064 (0.0031)
1.093 (0.0049)
1.042 (0.0028)
0.996† (0.0012)
0.999 (0.0010)
1.000 (0.0005)
0.986† (0.0017)
0.996† (0.0009)
0.994† (0.0009)
0.977† (0.0024)
0.987† (0.0017)
0.990† (0.0014)
0.996 (0.0024)
0.997 (0.0017)
1.000 (0.0013)
1.145† (0.0079)
1.064† (0.0049)
1.041† (0.0037)
262.e2 JADA 148(4) http://jada.ada.org
April 2017