Gender Differences in Earnings of Early- and Midcareer Pediatricians

We analyzed cross-sectional survey data collected in 2016 from PLACES.²⁶  Participants are surveyed twice each year: a longer survey covering several domains (eg, work characteristics, satisfaction, work-life balance, life experiences) and a shorter survey with more in-depth questions about a single topic, such as earnings, which was selected by participants for the 2016 survey.

Pediatricians were recruited in 2012 to participate in PLACES. Participants were randomly selected from an AAP database that includes all pediatricians who completed US residency programs, both AAP members and nonmembers. In recruitment, 2 cohorts were targeted: those who completed residency in 2009–2011 and those who completed residency in 2002–2004. Four in 10 pediatricians (1925 out of 4677) invited to participate in PLACES initially signed up for the longitudinal study; of these, 1804 (93.7%) completed the first survey and were thereafter considered PLACES participants. Study participants are more likely to be female, AAP members, and graduates of US medical schools than target samples of pediatricians; detail has been described previously.²⁶ 

The AAP Institutional Review Board approved the study.

We linked data from 2 PLACES surveys conducted in 2016 and demographic information collected during study recruitment in 2012. Questions were developed on the basis of review of other studies and input from the AAP PLACES advisory committee.

Participants answered a question on their annual income adapted from other research^5,27  (“What is your estimated annual income from your professional activities before taxes? For employees, please include salary, bonus, and profit sharing contributions. For owners, please include earnings after tax-deductible business expenses but before income tax.”). Participants also answered questions about their demographic, job, and work-family characteristics, many of which were adapted from previous research.^5,7,28–32  We hypothesized that these characteristics would be associated with earnings.

For analysis, we grouped the measured variables into 3 key groups: labor force, physician-specific job, and work-family characteristics, which are described in more detail below and in Table 1, with further information in Supplemental Table 3.

For labor force characteristics, we included gender, years since residency training completion (proxy for age and work experience), race and ethnicity, medical school location, region of country, hours worked per week, and primary job or specialty because these are key variables in the US Census Current Population Survey used to examine income^11,34,35  and important conventional human capital characteristics.^36,37  For race and ethnicity, because of small numbers, we combined Hispanic, African American or black, American Indian, and other race into one group. We included region of country because earnings and cost of living might vary across regions³⁸  and hours worked because part-time hours are common among female pediatricians,³⁹  and human capital models indicate work hours are a major source of gender earning gaps.⁴⁰  Specialty (eg, general pediatrics, hospitalist, smaller subspecialty, and larger subspecialty, which is described in detail in Supplemental Table 3) was included as a labor force characteristic because it further defines how occupation and earnings among different pediatric subspecialties might vary.⁴¹  We categorized subspecialty as larger (neonatology, cardiology, critical care, emergency medicine, gastroenterology, and hematology-oncology) and smaller (all other subspecialties) on the basis of fellowship size reported by the American Board of Pediatrics.³³ 

For physician-specific job characteristics, we included variables that provided more detail on the nature of the pediatrician’s current job, including number of years in current job, workload (number of patients seen per week), practice ownership, work setting (area, academic), and compensation method. In previous research, authors controlled for work load,¹³  practice ownership,^13,42,43  work area,^13,44  and academic setting⁴⁵  in studies on earnings.

For work-family characteristics, we included marital status, number of children, choices made in job for their own children, choices made in job for their work environment, and periods of work reduction (reported working part-time and/or not working on one or more surveys over the last 5 years). These variables provide information often missing in analyses of income disparities: choices made for their own families (further described in Supplemental Table 3). For example, women may be seeking flexible work schedules and/or take on more responsibilities at home.⁴⁶ 

Annual earnings self-reported by pediatricians in 2016 were analyzed in several ways. First, we examined the distribution and overall mean of earnings by gender. Second, for the overall sample and separately for women and men, we used t test and 1-way analysis of variance to examine pediatrician earnings by each labor force, physician-specific job, and work-family characteristic. Third, to estimate the adjusted difference in the log of pediatrician earnings between men and women, we used ordinary least squares regression. We conducted 4 regression models. Model 1 examined gender, unadjusted; model 2 controlled for labor force characteristics; model 3 controlled for both labor force and physician-specific job characteristics; and model 4 controlled for labor force, physician-specific job, and work-family characteristics (all characteristics are described in Table 1, Supplemental Table 3). We converted the logged regression gender coefficient back to the original scale to help interpret the findings and report the geometric means. We also used the likelihood ratio test to compare the model estimates for model 1 and model 2, model 2 and model 3, and model 3 and model 4.

For bivariate tests, continuous variables (years since residency graduation, work hours in a typical week, years in current position, number of patients seen in a typical week, number of children) were categorized for presentation purposes. For multivariable tests, these variables were included as continuous variables, centered at their means.

The number of cases in each analysis varied slightly because of missing values for specific questions. All analyses were conducted by using Stata Version 15 (Stata Corp, College Station, TX).

Sixty-seven percent of PLACES participants completed both 2016 surveys (1213 out of 1801 = 67%), which was similar for women and men (P = .99). The analytic sample was restricted to participants who completed the question on income and worked in general pediatrics (n = 507), hospitalist care (n = 118), or subspecialty care (n = 373) in 2016 (total n = 998). The 215 participants who were excluded from analysis were trained in a surgical specialty or internal medicine-pediatrics, working in a nonclinical and/or non-US position, in a fellowship training position, not working in 2016, and/or had missing data on income and/or specialty. Overall, 73.5% of the analytic sample are women.

Overall pediatrician-reported mean annual income in 2016 was $189 804. For women, the unadjusted mean was $176 738 (median = $170 000, minimum = $9000, maximum = $750 000, and skewness = 1.76). For men, the mean was $226 133 (median = $210 000, minimum = $80 000, maximum = $700 000, and skewness = 1.83).

Across characteristics, female pediatricians consistently reported lower mean earnings than male pediatricians (Table 2). For example, for those in general pediatrics, women reported earning $171 047 and men reported $223 472. Also, a disproportionate number of women were in the lowest-paid group of several characteristics. For example, 418 of 734 women (56.9%) and 89 of 264 men (33.7%) reported working in general pediatrics, and 271 of 727 women (37.3%) and 35 of 261 men (13.4%) reported working <40 hours per week.

Overall earnings were associated with many of the characteristics (Table 2). Among labor force characteristics, pediatricians in the larger subspecialty group (eg, neonatology, cardiology, critical care, emergency medicine, gastroenterology, and hematology-oncology) and those working ≥60 hours reported the highest incomes, reporting overall mean earnings of $231 930 and $216 818, respectively. Among physician-specific characteristics, pediatricians who reported more years at their current job, more patients per week, being an owner, working in a nonacademic work setting, and being paid by a mix of salary and billings or performance measures reported higher earnings. For example, the overall mean income of owners was $228 861 and employees was $182 060, P < .001. Among work-family characteristics, large differences in mean unadjusted earnings were found. For example, the overall mean for pediatricians who reported making choices in their earned income for their children was significantly lower than those not reporting such choices ($167 884 vs $205 183, P < .001).

In Figure 1, we present the male gender coefficient and R² for 4 models using the log of annual earnings. Without any adjustment, male pediatricians earned $51 319 more than female pediatricians, P < .001 (model 1). Adjusting for labor force characteristics, men earned $25 566 more than women, P < .001 (model 2). Adjusting for both labor force and physician-specific job characteristics, men earned $14 240 more than women, P < .001 (model 3). Finally, when adjusting for labor force, physician-specific job, and work-family characteristics, men earned $7997 more than women, P < .05 (model 4). The model R² increased across the 4 models, from 0.08 in the unadjusted model 1 to 0.48 in model 4. We used likelihood ratio tests to assess whether each subsequent model led to a significant increase in model fit; across models, each increase in explained earnings variation was statistically significant (P < .001). The female-to-male earnings ratio increased across models, from 0.76 in model 1, to 0.87 in model 2, 0.91 in model 3, and 0.94 in model 4. The full model results are presented in Supplemental Table 4.

We analyzed self-reported 2016 annual earnings among early- to midcareer pediatricians and found women earn less than men. Before any adjustment, female pediatricians earn 76% of what male pediatricians earn, or ∼$51 000 less. Adjusting for common labor force characteristics, such as demographics, experience, work hours, and specialty, female pediatricians earn 87% of what male pediatricians earn, or ∼$26 000 less. Adjusting for both labor force and physician-specific job characteristics, female pediatricians earn 91% of male pediatrician earnings, or ∼$14 000 less. Finally, adjusting for a comprehensive set of labor force, job-specific, and work-family variables, female pediatricians earn 94% of male pediatrician earnings, or ∼$8000 less.

Our study is comparable to gender differences found in other studies of US workers. In our unadjusted model, we found the female-to-male earnings ratio was 0.76. Among full-time US workers, the female-to-male unadjusted annual earnings ratio was 0.81 in 2016.⁴⁷  A US government report by the Joint Economic Committee summarized the persistent gender earnings gap across diverse occupations and education levels, with the largest pay gap among those with professional degrees and worse for mothers and part-time workers.⁴⁸ 

Our findings compare with studies of other physicians in which researchers found large unadjusted gender differences in income and smaller differences when adjustments are made, ranging from women earning 8% to 38% less than men.^{9,13,17–19,49–51}  Because we controlled for labor force, job-specific, and work-family variables in our study, the findings support previous reports in which authors suggested earnings gaps may be related to mothers who might prioritize job flexibility¹⁹  and that periods of part-time work and not working are related to earnings.¹² 

The role of employment negotiations and discrimination in the earnings gap has been discussed.⁴⁴  Women and men may make different choices, negotiate differently, or experience different pressure from themselves or others regarding their own children and families. Women, including physicians, spend more time on household and child care responsibilities than men.^46,52,53  They may accept lower salaries for nonmonetary benefits, such as fewer work hours,⁵⁴  flexibility of work hours, and location or community of work place.⁴⁴  Our study is unique because our focus is on one specialty, and we adjust for a comprehensive set of work-life characteristics that other researchers have suggested are important.^12,18–20  Besides labor force variables (eg, work hours and specialty), some of the largest differences in earnings were among the work-family characteristics. Still, even when controlling for these characteristics, we found that early- to midcareer women in pediatrics earned $8000 less than men.

The gender pay gap is complex and multifaceted. Economists have discussed the issue extensively and propose explanations such as education, occupation choice, experience, work place flexibility and hours, negotiation skills, competitiveness, parenting, and time away from work.^55,56  They suggest that, in part, women earn less because of long work hours and inflexible conflict with home responsibilities.⁵⁵  In a study of master of business administration graduates, it was found that women and men had similar income and work hours at the start of their careers, but over time, both income and work hours increasingly diverged.⁵⁵  Researchers also reported that work interruptions and differences in hours worked explained much of the gender gap in earnings and having children was related to more work interruptions, fewer hours worked, and less income for women but not men. Economist Claudia Goldin suggests that to obtain convergence of pay, changes to the structure and renumeration of jobs are needed (eg, more-flexible work schedules without reduction in opportunities and pay).⁵⁷ 

US academic medical institutions are working toward policies to achieve pay equity.⁵⁸  In a study of an academic pediatrics department, gender salary inequalities were found and an intervention strategy was designed that focused on several areas, including salary corrections and annual monitoring of salaries.¹⁵  Postintervention, gender salary differences were no longer significant. Their intervention might be helpful to other physician employers.

Our study has strengths, including a national sample, focus on a specialty with a majority of women, and examination of a comprehensive set of physician characteristics. Limitations include that data are self-reported by early- and midcareer pediatricians and many different pediatric subspecialties were combined because of insufficient numbers to examine separately. The generalizability to other specialties or career stages and whether our findings differ from studies of specialties with lower portions of women is unclear. The difference in earnings might be more substantial as pediatricians progress through their careers. Although the response among PLACES participants who completed the study surveys was high, the initial project sign-up rate was lower, at ∼40%, and women, US medical school graduates, and AAP members were more likely to sign up. We could not adjust for all possible characteristics, so it remains unclear whether it is only gender that explains the remaining income gap between men and women. Because of small numbers of Hispanic and African American pediatricians in our sample, we combined these pediatricians with those who specified “other race” and were not sufficiently powered to examine differences in earnings for specific groups. Finally, the data are cross-sectional and therefore do not explore how pediatrician earnings and the gender gap progress over time, which will be possible with future PLACES data.

In this study, we found gender differences in annual earnings among early- and midcareer pediatricians, with women earning less than men. This difference persists after adjustment for important labor force, physician-specific, and work-life characteristics. With more female physicians in the workforce, the findings can be important to understanding overall physician earnings. In future work, researchers should use longitudinal analysis and further explore periods of part-time work, family obligations, and work flexibility.

We thank the pediatricians participating in PLACES who are all giving generously of their time to make this project possible. This research was supported by the AAP.

AAP

American Academy of Pediatrics

PLACES

Pediatrician Life and Career Experience Study