Real wage growth in the U.S. health workforce and the narrowing of the gender pay gap

The division of labor in healthcare reinforces professional boundaries with occupational closure strategies that can limit the earnings of those with only vocational qualifications [10, 15]. Cross-nationally, occupational hierarchies distinguish jobs requiring significant educational training, credentials, and licensing (high status) from those having fewer employment preconditions (low-status) [18, 19]. Higher wages for professionals reflect the derived demand for healthcare services and the greater bargaining power and status associated with their occupation. While increased requirements can raise salaries for highly experienced registered nurses, such credentials do little for improving the pay of certified nursing assistants [11]. During 2005–2015, wage inequality increased between professional and non-professional workers in healthcare [10]. And certifications and other closure strategies did not benefit women's wages as they had for men [1].

Disparities in earnings among men and women employed in the same healthcare occupations are not unique to the U.S. Between 2006 and 2014, across lower and upper-middle-income countries, the gender wage gap in healthcare increased while remaining relatively constant in higher-income countries [6]. Gender and health system hierarchies are embedded with institutional practices attached to social characteristics such as race, ethnicity, class, or caste. Fair and accurate professional productivity evaluations are hidden in compensation policies, allowing bias to determine standards and who meets them [19].

Cross-national explanations for the gender wage gap in healthcare include women’s concentration in specific jobs, specialty choices in medicine, labor market discrimination, the devaluation of caring skills in pay structures, and depressed bargaining power among women [20‐23]. Care work by female nurses is often regarded as less valuable than the medical interventions performed by predominantly male physicians [24]. Furthermore, immigrant women and women of color overwhelmingly inhabit the lower-waged care jobs in the U.S. and lack regulatory protections or union representation [1, 3, 25]. Consequently, in the highest-skilled physician and nurse occupations and non-professional healthcare support jobs, pay gaps by gender, race, and nativity are evident [3, 10, 20, 21].

The American Community Survey (ACS) public use microdata sample (PUMS) for 2001, 2004, 2008, 2013, and 2017 were used to derive hourly wages. Five years of ACS sample data were concatenated, and five datasets were created for each occupational group. The ACS is the largest, individual-level data set capable of tracking healthcare occupations in the U.S. [26]. Conducted annually by the U.S. Census Bureau, it reaches roughly 295,000 U.S. addresses per month. Households are required by law to respond, and response rates are generally about 95%. The 2001 and 2004 ACS had approximately 1.1 to 1.2 million observations per year. By 2017, this increased to nearly 3.5 million individuals. ACS statistics uniquely provide estimates of consistently defined demographic and economic variables.

Occupations were identified using the Census Bureau's 2010 ACS classification scheme [26]. This scheme organizes homogeneous occupations into clusters. Examples of jobs found within each category and their relevant Census code follow physician and surgeon (3060: (hospitalists, urologists)); nurse (3000–3540: (registered nurse, nurse practitioners); other healthcare practitioner and technical (3000–3540: (chiropractors, dentists)), and healthcare support (3600 -3650: (orderlies, dental assistants)). A fifth category was compiled to include jobs not coded as healthcare occupations in the ACS [27]. This cluster is not homogeneous. The first six jobs demand specific qualifications, while the remaining two have minimal educational or licensing requirements. Direct patient care jobs include medical/health services managers (0350), social/community service managers (0420), psychologists (1820), social workers (2010), counselors (2000), various community and social services specialists (2020), medical/dental/ophthalmic laboratory technicians (8760) and personal/home care aides (4610).

A median hourly wage was determined for all full-time (30 or more hours a week), full-year (51 weeks) workers aged 18–75, or for physicians and surgeons, subset to aged 35 years and older. The ACS wage value was divided by usual hours and weeks worked. Median wage growth rates by occupation were calculated. Then, using a quantile regression model, the natural logarithm of the CPI-adjusted, trimmed hourly wage was adjusted for demographic, industry, and hours worked differences. Results from the 50th percentile of the hourly wage distribution are reported.

Independent variables included in the regression analyses were gender (female/male), age (18–34, 35–40, 41–49, 50–59, 60–75), race/ethnicity (non-Hispanic Black/African-American, Hispanic/Latino, non-Hispanic White, Asians/others), nativity (native-born, naturalized citizen, non-citizen), education (high school or lower, some college, Bachelor's degree or higher), and hours of work (30–34, 35–40, 41–50, 51 or more). Four healthcare industries, including hospitals, ambulatory care (physicians' offices, outpatient care centers), and long-term care (home healthcare, nursing care facilities), were included. The fourth industry consists of "other" locations where healthcare employees reported they worked. Largely comprising the government and education sectors, it also included community/social service organizations and retail pharmacies. The listed variables do not capture everything that could affect median hourly wages, so the state of residence and fixed-year effects were included to control for observed and unobserved influences.

The multivariable regression analyses predicted the log median hourly wage for each occupation using weighted, pooled cross-sections of the ACS so that results are nationally representative. Quantile regression modeling was performed using the R computer language. Robust standard errors for the estimated intercept and coefficients were made using the Huber Sandwich estimator [28].

Additional file 1: Table S1 provides the pooled-sample characteristics by year for the sample used in the wage regression analyses. For each of the explanatory variables, percentages by gender are also reported. The data confirms the health workforce has grown, and occupational shares have changed over time. Healthcare support’s share declined between 2001 (19%) and 2017 (16.8%). Nurses increased their share between 2013 (20.7%) and 2017 (21.6%) after relative flatness between 2001 and 2013. Direct patient care generally trended upward, reaching 28.9% of the health workforce in 2017. Healthcare practitioners experienced a decrease in share from 2001 to 2013, followed by a substantial increase from 27.5% to 28.9% between 2013 and 2017. The proportion of physicians and surgeons generally trended downward, constituting 5.5% of the total workforce in 2001 but only 5.2% by 2017.

Healthcare support and physicians were the only two occupations that lost workforce share between 2001 and 2017. Declines in the physician share are attributed to retirements, whereas low pay and high turnover in healthcare support increase the difficulty of recruitment to this occupation [3, 29]. All other occupational shares expanded due to rapid growth in healthcare demand [5, 9, 16].

Women were 74.1% of the workforce in 2001, and this increased to 76.5% by 2017. Their representation rose in direct patient care, healthcare practitioners, and physicians between 2001 and 2017. The percentage of non-Hispanic Whites working in healthcare declined, as did the number of native-born workers. Latinos and Asians, along with naturalized citizens, grew their workforce share, while Black workers saw their employment share decline after 2008, only to rebound by 2017.

Workers aged 50–59 and 60 years or older increased their share between 2001 and 2017, as did workers aged 18–34. The percentage of workers with higher education credentials increased significantly. Women exhibited higher levels of educational attainment by 2017. The proportion of the health workforce located in the four healthcare industries was relatively stable. However, women saw job declines in hospitals, while their share of employment in ambulatory and the "other" categories grew between 2001 and 2017. The percentage of female workers reporting working 35 to 40 or 41 to 50 hours a week increased between 2001 and 2017. Among all workers, the percentage working 51 hours or more declined.

Table 1 reveals that wage growth rates in all occupations were highest between 2001 and 2004. From 2004 to 2008, healthcare support, direct patient care, and healthcare practitioner workers experienced negative wage growth. Nurses and physicians also registered slower wage growth, with physicians having a rate of 1.3% compared to their 2.47% rate for 2001–2004. By 2008–2013 healthcare support workers continued to have negative wage growth, and nurses saw real wage declines. Direct patient care and healthcare practitioners experienced less than 1% annualized wage growth between 2008 and 2013, while physicians realized an improved rate of 2.16%. Between 2013 and 2017, workers in direct patient care and healthcare practitioner jobs suffered wage declines during this period of economic recovery. Healthcare support and nurses had growth of less than 1%. In contrast, physicians saw wages grow at 2.45%.

Overall, between 2001 and 2017, professional workers registered the fastest cumulative median wage growth. Specifically, physicians' wages grew almost four times faster than nurses' (37.6% versus 9.92%) and over six times faster than the rate of 5.68% in healthcare practitioner occupations. Paraprofessional workers in healthcare support had a cumulative wage growth rate of less than one percent. In contrast, the mixed professional direct patient care category experienced a negative wage growth rate of −8.4%.

Table 2 reports the conditional quantile regression coefficients taken at the 0.50 percentile of the wage distribution. The gender dummy coefficients in Table 2 are estimates of the occupational wage penalty for women in 2001, after adjustments for the other covariates. Being a woman decreased the median wage in 2001 (the reference year) by approximately (e^−.1452)–1 × 100) 16% in healthcare support, by 4% in Nursing, 13% in direct patient care, 15% in healthcare practitioners, and 23% in the physician surgeon category.

In every occupation except physicians, workers younger than 41 made significantly less than employees aged 60–75 (the omitted category). Nurses aged 60–75 years of age earned more than other age groups. Physicians aged 41–59 earned higher median wages than those aged 60–75. African Americans earned less than non-Hispanic Whites (the omitted category) except in healthcare support. Latinos realized wage penalties in every domain. Non-Hispanic Whites made more than Asians/others in healthcare support and direct patient care but earned less in nursing.

Naturalized citizens had higher median wages than native-born workers in healthcare support, healthcare practitioner, and nursing occupations. Non-citizens always earned less than native-born workers. Individuals without a Bachelors' Degree (the omitted category) were paid less than those who had completed college.

Jobs located in hospitals (the omitted industry) paid significantly higher median wages when compared to the other three sectors. Long-term care workers in healthcare support, nursing, and direct patient care had 19%, 17%, and 38% lower wages, respectively, than their hospital-based colleagues. Healthcare practitioners and physicians faced their most significant wage penalty in the "other" sector, earning 11% and 14% less than their peers in hospitals.

Those working 30–34 hours a week in healthcare support, nurse, and healthcare practitioner jobs had significantly lower wages than workers in the omitted 35–40 hours category. Laboring 41–50 hours a week lowered wages in the healthcare support, nurse, and healthcare practitioner jobs. And wage penalties were evident for working 51 or more hours across occupations.

Year dummy variable coefficients estimated the difference between the year intercept and the 2001 omitted intercept and accounted for potential changes in wages over time. Table 2 reveals that when compared to 2001, there was a negative wage trend beginning in 2008 for direct patient care and healthcare support laborers. The gender–year interaction term calculated the difference between the effect of gender on median wages in a specific year compared to its impact in 2001. In 2004, women's earnings in healthcare support and physicians increased, while their wages fell in the healthcare practitioner occupation. But the significant 2004 gender interaction coefficient for physicians is suspect as 72.8% of male and 52.9% of female physicians had top-coded earnings that year. Thus, the median wage for all physicians was understated, but the underestimation was more significant for male physicians. In 2017, women in healthcare support and physicians’ made wage gains of 7% and 10%.

Table 3 provides the predicted median wages for men and women by year and occupation. The log regression coefficients from Table 2 were used to predict wages, and an adjusted gender wage gap was calculated. The wage gap narrowed between 2001 and 2017 in every occupation except for nurse. As a percentage of men's, women's median wages in nursing declined from 95.7% in 2001 to 94.3% in 2017 but still reflected the greater pay parity found in this job. The gender wage gap in the physician occupation appears drastically cut between 2001 and 2004, with women losing ground between 2004 and 2008. However, these findings are likely an artifact of top coding [see Additional file 2].

In the healthcare support and direct patient care occupations, both men and women were negatively impacted by the great recession, as evidenced in the annualized percentage wage change columns for 2004–2008 and 2008–2013. Additionally, male median wages grew more slowly than women's during 2008–2013 in the nurse, healthcare practitioner, and physician professions. During the recovery years of 2013–2017, both men and women in nurse, direct patient care, and physician jobs realized improved wage growth. Alternatively, all healthcare support and healthcare practitioner employees experienced negative wage growth.

Overall, men in healthcare support registered a negative cumulative wage change, while women experienced wage stagnation. As a group, direct patient care workers experienced cumulative negative wage growth. Women in the healthcare practitioner and physician occupations had higher overall wage growth rates than men, except in nursing.