Is healthcare the new manufacturing?: Industry, gender, and “good jobs” for low- and middle-skill workers

Is healthcare the new manufacturing?: Industry, gender, and “good jobs” for low- and middle-skill workers

Social Science Research 84 (2019) 102350 Contents lists available at ScienceDirect Social Science Research journal homepage: www.elsevier.com/locate...

2MB Sizes 0 Downloads 29 Views

Social Science Research 84 (2019) 102350

Contents lists available at ScienceDirect

Social Science Research journal homepage: www.elsevier.com/locate/ssresearch

Is healthcare the new manufacturing?: Industry, gender, and “good jobs” for low- and middle-skill workers

T

Janette Dilla,∗, Melissa J. Hodgesb a b

The University of Minnesota, Division of Health Policy & Management, USA Villanova University, Department of Sociology and Criminology, USA

ARTICLE INFO

ABSTRACT

Keywords: Low-wage work Health care workforce Feminized occupations Job quality

Using the 2004 and 2008 panels of the Survey for Income and Program Participation (SIPP), we examine whether the heavily feminized health care industry produces “good jobs” for workers without a college degree as compared to other major industries. For women, we find that jobs in the health care industry are significantly more likely than the food service and retail industries to provide wages above $15 per hour, health benefits, fulltime hours, and job security. Jobs in the health care industry are not “good jobs” for low- and middle-skill men in terms of wages, relative to the industries of construction and manufacturing, but health care jobs can provide men with greater job security, and in comparison to construction, a higher probability of employer-based health insurance. That said, the findings emphasize that because men and women are differentially distributed across industries, access to different forms of job quality is also gendered across industries, with important implications for gender dynamics and economic strain within working class families.

1. Introduction Male-dominated industries like manufacturing and construction have traditionally been sources of “good jobs” for workers with less education, or jobs with decent wages and access to health benefits, regular hours and job security (Kalleberg, 2011). In other words, manufacturing jobs and other historically highly gender segregated industries (Charles and Grusky, 2005; Stainback et al., 2010), have provided a pathway into the middle class for low- and middle-skill men. However, in recent decades, manufacturing employment in the U.S. has declined, while health care organizations have become increasingly recognized as centers of stable employment and contributors to economic development and job growth (Adams, 2003; Bartik and Erickcek 2007; Parrillo and de Socio 2014). Case studies of metropolitan areas have documented how health care organizations have become pillars of employment in many cities where manufacturing industries have collapsed (De Socio, 2012, 2007; Nelson and Wolf-Powers, 2009). But, do health care jobs, which tend to employ greater numbers of women, provide the same opportunities? Clearly health care organizations provide “good jobs” for more educated workers, such as physicians, pharmacists, and registered nurses. However, many workers in these organizations – if not the majority – do not have high skill levels (Schindel et al., 2006). In this study, we explore the availability of “good jobs” for low- and middle-skill workers across different industries, with a focus on the highly feminized health care sector. Because industry and occupation account for a substantial proportion of the overall



Corresponding author. E-mail address: [email protected] (J. Dill).

https://doi.org/10.1016/j.ssresearch.2019.102350 Received 8 October 2018; Received in revised form 16 July 2019; Accepted 28 August 2019 Available online 03 September 2019 0049-089X/ © 2019 Elsevier Inc. All rights reserved.

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

variation in earnings (Autor et al., 2006; Dwyer, 2013; Massey and Hirst, 1998; Wright and Dwyer, 2003), we target our investigation on the comparison between industries that employ the greatest percentages of low- and middle-skill workers, including food service, retail, manufacturing, construction, and jobs in the health care sector. Following Kalleberg (2011) work, we develop minimum thresholds of a “good job” for low- and middle-skill workers along four of dimensions, including wages, benefits, hours, and job security. Throughout this paper, we argue that persistent patterns of occupational and industry sex segregation among the working class (Cotter et al., 2004; Dwyer, 2013; Sutton et al., 2016) produce gender differences in access to different forms of job quality. Despite the growth of the care sector and other traditionally female-concentrated jobs for low- and middle-skill women, these jobs fall short in replacing “good” working class jobs for men (Autor, 2015; Bergmann, 2011; England, 2010). 2. Navigating the New Economy without a college degree Workers without a college degree are not a small segment of the workforce; in 2015, only 32.5% of Americans above the age of 25 had a four-year bachelor's degree (Ryan and Bauman 2016). We refer to workers without a four-year bachelor's degree as low- and middle-skill workers. We define low-skill jobs as those that require a high school degree or less, while “middle-skill” jobs are those that require some education and/or training beyond high school. Postsecondary education or training requirements can include associate's degrees, vocational certificates, significant on-the-job training, previous work experience, or some college, but less than a bachelor's degree (Holzer and Lerman, 2009, 2007). In the mid-twentieth century, the United States economy had a strong manufacturing sector, and organized labor and union activity secured good wages and employment benefits for many workers without a college degree (Abraham et al., 2010; Kalleberg, 2011; Osterman, 2014). However, in recent decades, the United States and many western economies have experienced a decline in the share of manufacturing in overall employment, with a concurrent rise in the share of services (Pilat et al., 2006). The decline of blue-collar factory jobs has been associated with a reduction in pay for less educated workers and has contributed to growing US earnings inequality (Mishel et al., 2012), as the service sector jobs that have emerged in the New Economy do not have the same level of job quality as jobs in the manufacturing sector (Autor and Dorn, 2009, 2013; Meisenheimer, 1998). There is ample evidence that the shift to a service-based economy has been difficult for low- and middle-skill workers, particularly men (Carnevale et al., 2013; Silva, 2013). For example, real hourly wages for the bottom twentieth percentile of male workers dropped from $12.23 to $10.16 (a 14.9 percent decrease) between 1973 and 2011 and low-wage women experienced stagnant wages over the same time (Mishel et al., 2012). Employer-benefits for low- and middle-skill workers have also declined. The percentages of recent high school graduates in 1973 that received employer-based health insurance and pensions were 63.3 percent and 36 percent, respectively; in 2010 these figures were 22.8 percent and 16.3 percent (Mishel et al., 2012). Finally, reliable hours and job security have also declined for these workers (Lambert, 2008; Even and Macpherson, 2018). These changes in our economy were in many ways magnified during Great Recession, which began at the end of 2007 and has been identified as the worst financial situation in the United States since the Great Depression of the 1930s (Grusky, Western, and Wimer 2011). The majority of jobs lost during the Great Recession were in male-dominated industries, leading many commentators to describe this period as the “Man-cession” (Thompson, 2009), while many service-sector industries, including health care, maintained current job-levels or even increased. For example, employment in education and health services was up 3.3 percent between 2007 and 2009, while nearly all other industries, including male-dominated industries like construction and manufacturing, had significant declines in employment over the same time period (−19.8% and −14.6%, respectively) (Goodman and Mance, 2011). What is a “good job” for low- and middle-skill workers in today's economy? A number of definitions and dimensions have been used to conceptualize job quality, from pay and benefits, to skill, autonomy, and overall job satisfaction (Kalleberg, 2011; Green, 2006; Handel, 2005; McGovern et al., 2004). Kalleberg (2011) argues that most people agree that compensation and benefits are central components of job quality and a sense of dignity at work (Bolton, 2007; Hodson, 2001). In this paper, we focus exclusively on extrinsic job rewards, including whether a job pays at least $15 per hour, provides health benefits, provides sufficient, stable work hours and protects workers against being laid off. We discuss these different definitions of a “good job” below. First, we have chosen a threshold of $15 per hour as a “good wage” for low- and middle-skill workers because of the focus on the $15 minimum wage movement over the past four years (Greenhouse, 2016) and the success the movement has had to establishing $15 as a threshold for a dignified wage and a pathway into the middle class (Meyerson, 2014). However, $15 per hour also indicates that an individual's wages fall near the middle of the wage distribution for workers without a college degree. When we divide wages of men and women without a college degree into quintiles, men and women who earn $15 both fall in the third wage quintile.1 Second, we conceptualize a good job as one that provides health benefits for workers. Health benefits are especially important in the US because individuals are primarily dependent on employers for health insurance coverage (Gautié and Schmitt, 2010). And finally, important characteristics of “good jobs” are steady hours and protection against layoff. In the New Economy, low-wage and less skilled workers are often underemployed and cannot obtain as many hours of work as they would prefer (Golden and Gebreselassie, 2007; Kalleberg, 2007), and rates of work hour instability increased during the recession and recovery period (Finnigan, 2018). Low-

1 Based on the authors' calculations. The wage quintiles were calculated using IPUMS-USA data (Ruggles et al., 2017) for the years that correspond to the years in the current study (2004–2013). In supplemental analyses, we also consider differences in wage thresholds above $15 per hour.

2

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

and middle-skill workers are also at greater risk of being laid off from their jobs, particularly in declining industries like manufacturing and production (Holzer et al., 2011). We consider these job qualities to be the minimum threshold for a “good job” in the U.S. That said, we want to emphasize the relative nature of “good jobs” (Kalleberg, 2011). For example, a “good job” for someone with a high school degree may not be a “good job” for a college graduate. We argue that although jobs that meet these minimum thresholds may be “good jobs” in in today's economy, job quality characteristics may vary across industries that employ different proportions of women and men, thus what qualifies as a “good job” for women in the working class may not be the same “good jobs” for men. 3. Are health care jobs “good jobs?” Most low- and middle-skill occupations in the health care sector are female-dominated occupations. In general, occupations with a higher percentage of women pay less than occupations with fewer women, even after adjusting for measurable differences in education requirements, skill levels and working conditions (Cohen and Huffman, 2003; Kilbourne et al., 1994; Levanon et al., 2009). Further, research has shown that workers in occupations involving care work experience a greater “wage penalty” as compared to other feminized jobs of similar skill levels that do not involve care (England et al., 2002; Budig et al., 2019). Many health care jobs for low- and middle-skill workers have what we would call “bad job” characteristics (Kalleberg, 2011), particularly those occupations that have the highest percentage of women and are most strongly associated with “women's work” (Duffy, 2011; Folbre, 2012). For example, for certified nursing assistants (CNAs), a common low-skill health care occupation, the median yearly earnings in 2016 was $26,590, a rate that falls well below the median rate for women across all occupations ($38,948) (Bureau of Labor Statistics, 2018), and for most remain low throughout their careers (Baughman and Smith, 2012; Ribas et al., 2012). Indeed, Dwyer (2013) argues that care work jobs – a category which includes all health care occupations – are contributing to job polarization in today's economy by creating good, well-paid jobs for highly skilled workers, and low-wage, poor quality jobs for lowskill workers. However, there is some evidence that the health care sector provides some “good jobs” for low- and middle-skill workers. For example, many mid-level health care occupations that require an associate's degree or less, such as respiratory therapists, surgical technicians, and ultrasound technicians, are in high demand, pay good wages, and provide solid employment benefits (Dill et al., 2016; Holzer and Lerman, 2009; Autor, 2015). Andersson, Holzer, and Lane's (2005) study of social mobility among low-wage workers found that the health care sector is one of the few industries that provides consistent upward career mobility. There is also evidence that health care organizations are motivated to provide jobs with more benefits to improve recruitment and retention of high quality workers (Dill et al., 2014; Fitzgerald, 2006). Besides gender composition, the health care sector differs from manufacturing – and many other sectors – in some notable ways. First, while unionization rates across industries have been declining over time (Kalleberg, 2011; Brady et al., 2013)), credentialing has increased across all sectors in the United States (Kleiner, 2006). Health care has among the highest rates of licensure and other credential requirements (Redbird, 2017). The occupational closure of credentialing and licensure – and the attending educational requirements needed to obtain credentials – may work to increase demand and raise wages for workers (e.g., sharp increases in wages during the nursing shortage of the 1990s and early 2000s) (Kleiner, 2006; Spetz, 2016; Weeden, 2002). However, the widespread use of credentials may also make it difficult for low-level health care workers to move up within health care organizations (Baughman and Smith, 2012; Ribas et al., 2012; Glazer, 1991). Further, Redbird (2017) has shown that occupational licensure may actually increase the competition within an occupation by creating an established career pathway, potentially driving wages down and limiting pay scales compared to those in manufacturing and construction. Another notable feature of the health care sector that affects workers' pay is the role of insurance and the ways in which health care services are funded. Federal and state governments play a large role in setting prices in health care through reimbursement for services; in 2015, Medicare and Medicaid accounted for about 37 percent of total national health expenditures (U.S. Centers for Medicare and Medicaid Services 2018). Consequently, health care organizations may be constrained in setting wages by the level of reimbursement that they receive for their services. On the other hand, the fact that the government covers medical care to a reasonable standard at least for much of the population, including the elderly (Medicare), the poor (Medicaid), and veterans (though Veteran's Affairs), means that the industry has more guaranteed demand for services than would otherwise be the case, which may raise wages. The purpose of this paper is to examine gender differences in the prevalence of “good jobs” for low- and middle-skill workers across industries. Using minimum thresholds of a “good job” along four of dimensions, including wages, benefits, hours, and job security, we examine how job quality in the health care industry compares to job quality in other industries, with a focus on industries that have traditionally provided “good jobs” for low- and middle-skill workers, such as manufacturing and construction. We argue that because of gender segregation patterns across industries such that women are more likely to work in health care than men, the growth in the health care sector has produced more “good job” options for working class women. These gender segregation patterns are linked to differences in job quality across industries, such that “good jobs” for women are not necessarily the same “good jobs” for men, especially in terms of wages.

3

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

4. Methods 4.1. Data The data used in this analysis come from the 2004 and 2008 panels of the Survey of Income and Program Participation (SIPP), administered by the U.S. Census Bureau. The SIPP universe includes the non-institutionalized resident population living in the United States. Individuals were interviewed every four months over the survey panel; each interview is called a wave. The panels cover four to five years and include twelve to fourteen waves of data per person.2 The strengths of SIPP include its longitudinal design, detailed three-to four-year information on individuals, and its unique identification of respondents' employers. The analytical sample includes observations for individuals that 1) had an educational level that was less than a college degree during the survey panel, and 2) were between the ages of 18 and 65. Individuals who had no observations where they were employed during the survey period were not included in the sample. The sample includes 24,036 women (184,543 observations) and 24,236 men (187,548 observations) in the 2004 cohort, and 22,003 women (220,540 observations) and 24,415 men (228,817 observations) in the 2008 cohort.3 Sample characteristics are included in Table 1. 4.2. Measurement 4.2.1. Dependent variables We use a number of dependent variables for a comprehensive look at job quality. We first look at compensation, including the natural log of hourly wages (inflation-adjusted to 2013 dollars).4 We then use a series of dichotomous measures aimed at understanding whether a job is a “good job.” First, we measure whether a job provides a wage of at least $15 per hour. We also consider hourly wage thresholds of $20 and $25 dollars per hour in supplementary analyses. Our second dichotomous measure of job quality is whether an individual adopts employer-based health benefits. We include a measure of participation in rather than availability of employer-based health insurance because past research has shown that many workers do not take-up employer-based health insurance because it is too expensive and not a viable option (Baicker et al., 2012). As proxies of job stability, we look at whether respondents work fulltime (35 + hours per week),5 and the whether a respondent was on layoff at any point during the survey period.6 4.2.2. Independent variables The key independent variables are the industry in which the individual is employed. The industry categories are based on the 2002 U.S. Census industry codes and defined by the North American Industry Classification System (NAICS).7 We focus on the five industries that employ the greatest percentage of working adults without a college degree: 1) Health Care and Social Assistance (reference), 2) Retail Trade, 3) Accommodation and Food Services, 4) Manufacturing, 5) Construction. We also include a category entitled All other industries, which indicates if an individual is an industry other than one of the five listed above. The Health Care category is the reference category in all models because it is industry of primary interest in our analysis and the industry to which we want to compare all other industries. To measure gendered differences in job quality across industries, we indicate if an individual is female or male (reference), and we include interaction terms between gender and industry categories. Additional demographic variables included in the analyses are age and age squared, marital status (married = 1, else = 0), and whether the respondent has a child under the age of 18 (yes = 1, no = 0). We include four racial/ethnic categories: white (reference), 2 The 2004 panel includes 12 waves during 2004–2007. The 2008 panel includes 14 waves during 2004–2013. Although the full 2008 panel of the SIPP includes 16 waves, we include only Waves 1–14 to focus the 2008 panel on the recession and recovery and examine panels of roughly the same length of time. 3 For our analyses of wages, earning at least $15 per hour, and fulltime hours, the sample is restricted to observations where an individual has wage, industry, and other variable observations. This brings our analytic sample for these models down to: 23,097 women (128,441 observations) and 24,253 men (134,526 observations) in the 2004 cohort, and 22,111 women (144,401 observations) and 23,283 men (149,645 observations) in the 2008 cohort. For the models of employer-based health insurance and layoff, we include lagged measures for those characteristics from the previous wave. This restricts our sample considerably because both of those models require that individuals have observations from the previous wave. This eliminates all observations from the first wave (because there is no previous observation for the lagged variable) and any observations where any industry or health insurance is missing in the previous wave. 4 Hourly wages are inflation-adjusted to 2013 dollars, the last year of data collection. We used the Consumer Price Index (CPI) calculator to adjust wages, available on the Bureau of Labor Statistics website: https://www.bls.gov/data/inflation_calculator.htm. The SIPP includes a measure of both hourly wages and monthly earnings for both hourly and salary workers. Where hourly wages were missing, we calculated hourly wages based on monthly earnings and the number of reported hours per week. 5 Working part-time may mean that full-time employment is not available, but it also could result from a supply-side need or preference for working part-time. That said, there is evidence that full-time work is increasingly unavailable to low-skill workers and that there is growth in involuntary part-time employment (Even and Macpherson, 2018; Kalleberg et al., 2000; Lambert, 2008). To measure whether an individual works fulltime, we use a variable that indicates if the individual has worked less than 35 h during the any weeks during the last wave. If they indicated that they had worked less than 35 h in any weeks during the last wave, the observation was coded as not working fulltime (0). If they indicated that they had worked at least 35 h per week during all weeks during the last month, the observation was coded as working fulltime (1). 6 We indicate an individual was on layoff during the survey period using a variable where individuals identify reasons for why they are not employed. Individuals were coded as being on layoff if they indicated that the reason they were not working was layoff. 7 A list of the 2002 U.S. Census industry codes and major industry categories can be found here: https://www.bls.gov/tus/census02iocodes.pdf.

4

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Table 1 Summary statistics for men and women, by cohort.

Dependent variables Hourly wage Wage at least hour Employer insurance Fulltime hours Laid off during survey period Selected industries Health care Retail Food service Manufacturing Construction All other Personal demographics Age Married Children under 18 Race/ethnicity White Black Latino Other minority Educational level High school degree or less Some college Associate's degree Region Northeast Midwest South West N

2004 Women

2008 Women

2004 Men

2008 Men

$13.20 28.6% 59.4% 65.0% 1.1%

$13.22 28.4% 56.2% 63.3% 1.9%

$16.36* 43.8%* 58.8%* 82.1%* 1.2%

$16.05* 43.1%* 55.0% 78.8%* 2.5%*

21.2% 13.4% 16.9% 9.1% 1.6% 37.8%

22.7% 14.3% 16.9% 7.9% 1.4% 36.8%

3.9%* 9.9%* 14.9%* 18.8%* 15.9%* 36.6%*

4.4%* 10.9%* 15.1%* 16.5%* 15.2%* 37.9%*

36.5 44.1% 50.2%

36.9 42.0% 47.4%

35.4* 46.7%* 42.3%*

35.7* 44.4%* 40.6%*

67.1% 15.1% 11.3% 6.5%

63.6% 14.9% 14.5% 7.0%

67.8%* 11.7%* 14.4%* 6.1%

63.8%* 12.1%* 17.2%* 6.9%

49.6% 38.2% 12.2%

47.7% 38.8% 13.5%

56.1%* 34.4%* 9.5%*

54.9%* 35.4%* 9.7%*

15.3% 25.8% 36.9% 21.9% 23,036

17.0% 24.9% 35.7% 22.4% 22,003

15.0% 26.0% 36.0% 22.9% 24,236

16.7% 24.5% 35.1% 23.7% 23,415

Data source: SIPP. Notes: Mean values include the first observation for each respondent in the 2004 and 2008 cohorts of the SIPP *Indicates significant differences by gender in corresponding cohort (p < .05).

Black, Latino, and other minority. Educational attainment level is included as a time-varying categorical variable: high school graduate or less (reference), some college (but no degree), or an associate's degree. Finally, in our model of whether an individual receives employer insurance, we control for whether an individual had other insurance in the previous quarter.8 We include control variables that indicate the region in which the respondent lives: Northeast (reference), South, Midwest, and West. We include dummy variables that indicate the calendar year of data collection (not shown in tables). Finally, we separate our analyses by cohort, indicating if an individual is the 2004 SIPP panel or the 2008 SIPP panel. A summary of the sample characteristics is shown in Table 1.9 4.3. Analyses The analyses in this study use longitudinal data in which there are multiple observations for individuals over time. Therefore, with the data clustered by individual, we use random effects models to estimate random intercepts that account for the non-independence of observations over time. Random effects models allow us to capture both within- and between-individual variation, and they can allow for both time-varying (e.g., industry transitions, changing personal characteristics such as marital and parental status) and time-invariant characteristics of individuals (e.g., gender, race). That said, the random-effects models are limited in that they do not account for unobserved heterogeneity and bias caused by omitted variables. Thus, we include fixed effects models in the Appendix 8 Individuals may opt out of employer-based health insurance if they have insurance from another source, such as a spouse's insurance, Medicaid, or health care through Veteran's Affairs. To control for this selection, we include a lagged variable in the model that indicates whether or not a person has health insurance from a source other than their employer in the previous quarter. We lag the variable because having health insurance from another source in the same observation predicts employer-based insurance perfectly; whether an individual has health insurance from another source indicates the probability that an individual will switch to employer-based health insurance. 9 We do not include work characteristics (e.g., tenure, union member, salaried) in our analysis because we are primarily interested in the total effect of industry on job quality, and work characteristics may be related to the industry in which an individual is working and act as intervening variables between industry and the outcomes in our analyses. However, we do include models with work characteristics in the Appendix (Section 3, Tables 4 and 5).

5

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

(Section 2) that account for stable characteristics of the individual and capture within-individual variation; this more conservative technique allows us to examine how a change in industry is associated with a change in wages and job quality, net of time-varying observed characteristics and time-invariant characteristics. In our initial analyses, we use random effects linear regression models where the natural log of inflation-adjusted wages is the dependent variable. We then calculate the predicted wages of men and women by industry and cohort. Second, we conduct a series of analyses that measure the probability of having “good job” qualities (with binary dependent variables) using random effects logit models. The four binary outcome variables are: earning at least $15 per hour, having employer-based health insurance, working fulltime hours, and experiencing layoff during the survey period. In the model of layoff, the industry and educational variables are lagged by one quarter so they indicate the industry and education level of the individual before they were on layoff. We then calculate the predicted probability of having these “good job” outcomes for men and women by industry and cohort. All statistical analyses were conducted using Stata 14. For all models, we test for significant differences in the marginal effects for men and women by industry using the contrast option with the margins command in Stata. 5. Findings 5.1. Descriptive findings We present summary statistics for men and women by cohort in Table 1. The average inflation-adjusted wage for low-and middleskill women in our sample is about $13.20 in both the 2004 and 2008 cohorts. Low- and middle-skill men have significantly higher wages than low- and middle-skill women, but they experienced a substantial decline in wages between 2004 and 2008 (from $16.36 per hour to $16.05 per hour). This pattern is consistent across all of our measures of job quality, including whether an individual earns $15 per hour, employer-based health insurance, working fulltime hours, or being laid off during the survey period. Women did not experience the same degree of decline between 2004 and 2008 in terms of job quality; for example, women's average wages dropped just $0.02 between 2004 and 2008. In sum, men have better job quality across these measures, but they fared worse than women during the recession. Table 1 includes the distribution of men and women across the selected industries in our analyses, including health care, retail, food service, manufacturing, construction, and all other industries. These results reveal clear gender segregation across industries: Women are heavily concentrated in health care (23%), food service (17%) and retail (14%), with very few women in manufacturing (8%) or construction (1%) (percentages from 2008 cohort). Men, on the other hand, are most concentrated in manufacturing (16%), construction (15%), and retail (15%) (percentages from 2008 cohort). About 11% of low- and middle-skill men work in food service, while only about 4% of men in our sample work in health care. In Fig. 1a and b, we have included a larger spectrum of U.S. Census industry categories (15 categories) than those we will focus on in our analysis to give a more comprehensive picture of where workers are located in today's economy. Fig. 1a shows the percent of women employed by industry. The health care industry employs the largest percent of female workers in our sample of low- and middle-skill workers, and the percentage increases over time. In 2004, about 24% of women in our sample worked in the health care sector, and by 2013, 28% of female workers were in the health care sector. Fig. 1b shows the percent of men employed by industry. The manufacturing industry employs the largest percentage of low- and middle skill men, but this percentage falls over time. In 2005, the manufacturing sector employed 22% of men, but that had fallen to 18% by 2013. Construction is the second-largest employer of men in our sample, but the percentage of men employed in construction has also fallen over time, from about 15% in 2004, to just above 12% in 2013. Retail also employs a substantial percentage of lowand middle-skill men, and the percentage of men in retail increased over time, rising from about 13% in 2004 to about 14% in 2013. Men's employment in health care increased slightly over this period (from just under 4% in 2004 to 5% in 2013).10 5.2. Wage analysis Table 2 presents the results for random effects models of inflation-adjusted logged hourly wages; we run separate models for the 2004 and 2008 cohorts. In our models, we include interaction terms between gender and the selected industries in the models; men in health care are the omitted category, and the main effect for gender captures the effect for women in health care as compared to men in health care. As shown in Table 2, the total effects of industry and gender are significant (p < .001), as are all of the interaction terms between industry and gender (p < .001). Across both the 2004 and 2008 cohorts, our findings show that men in the food service and retail industries earn lower wages as compared to health care (p < .001), and men in manufacturing and construction industries earn higher wages as compared to health care (p < .001). The interaction terms measure differences in the gender wage gap across industries; we find that the gender wage gap is significantly different across all industries as compared to the gender gap in the in the health care sector (p < .01). We conducted sensitivity tests to directly test differences in wages within women across industry by running models where female was the reference group; we found that women in food service, retail, and other industries had significantly lower wages as compared to women in health care (p < .001), while women in manufacturing and construction had significantly higher wages as compared to women in health care (p < .001). Results available upon request. The explained variance within individuals in both the 2004 and 2008 cohorts is about 2%, suggesting that industry and the other 10

For context, Appendix Table 1 shows substantial differences in the three highest frequency occupations by industry for men and women. 6

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Fig. 1. a and b. Percent of women and men employed by industry. Data source: SIPP.

demographic variables included in the model do not provide a strong explanation of individual wage change over time. The explained variance between individuals, however, is much higher; industry and other demographic variables explain about 32% of variation between individuals. The overall explained variance of the full model is 28%. The predicted wages for men and women by industry are shown in Fig. 2. In the random effects models shown in Table 2, we use the natural log of inflation-adjusted wages as the dependent variable to normalize the distribution of wages. Thus, the predicted wages from the random effects models are logged, but we used the exponential function in Excel to transform the logged predicted wages to non-logged wages for ease of interpretation. The highest paying industry for both men and women is construction, where men earn about $17 per hour and women earn about $14 per hour (in both the 2004 and 2008 cohorts). The construction industry is followed by manufacturing, where men earn just under $16 per hour in 2008 (which reflects a substantial decrease from 2004, which we discuss below), and women earned around $13.50 in 2008. However, it is important to remember that the construction and manufacturing industries employ only a small 7

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Table 2 Coefficients from random-effects models predicting logged inflation-adjusted hourly wages, by cohort. 2004 Cohort

Selected industries Health care Food service Retail Manufacturing Construction All other Gender Female Male Interaction terms Health care*Female Food service*Female Retail*Female Manufacturing*Female Construction*Female All other*Female Personal characteristics Age Age squared Married Parent of kids under 18 Race/ethnicity White Black Latino Other minority Education High school degree or less Some college Associate degree Region Northeast Midwest South West Constant R2 within R2 between R2 overall Observations Individuals

2008 Cohort

Coef

SE

Coef

SE

Reference −0.174*** −0.029*** 0.110*** 0.139*** 0.051***

(0.009) (0.008) (0.008) (0.009) (0.008)

Reference −0.177*** −0.040*** 0.110*** 0.182*** 0.053***

(0.008) (0.008) (0.008) (0.008) (0.007)

−0.109*** Reference

(0.009)

−0.084*** Reference

(0.008)

Reference −0.121*** −0.078*** −0.101*** −0.069*** −0.077***

(0.010) (0.010) (0.011) (0.015) (0.009)

Reference −0.110*** −0.077*** −0.069*** −0.108*** −0.079***

(0.010) (0.009) (0.010) (0.015) (0.008)

0.050*** −0.001*** 0.057*** 0.003

(0.001) (0.000) (0.003) (0.002)

0.041*** 0.000*** 0.063*** −0.002

(0.001) (0.000 (0.003) (0.002)

Reference −0.088*** −0.123*** −0.081***

(0.006) (0.006) (0.008)

Reference −0.077*** −0.125*** −0.081***

(0.006) (0.005) (0.007)

Reference 0.070*** 0.191***

(0.003) (0.005)

Reference 0.071*** 0.155***

(0.003) (0.005)

Reference −0.067*** −0.104*** 0.007 1.610*** 0.023 0.316 0.277 262,967 47,262

(0.006) (0.005) (0.006) (0.017)

Reference −0.070*** −0.088*** 0.028*** 0.000 0.022 0.322 0.285 294,046 45,394

(0.006) (0.005) (0.006) (0.000)

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Dummy variables for year of survey period are included but not shown.

percentage of low- and middle-skill women (7.9% and 1.4% in the 2008 cohort, respectively), so only a small percentage of women are gaining access to the higher wages available in these industries. For women, wages in the health care industry ($13.06 in 2008) – the industry of primary interest in this paper – are relatively competitive with wages in manufacturing and construction industries. Further, women's wages in health care are substantially higher than wages in the two industries (other than health care) that employ large percentages of women: retail and food service ($11.62 and $9.81, respectively, in the 2008 cohort). While Fig. 2 shows that low- and middle-skill men consistently earn higher wages than women across all industries, it also shows that men's wages declined more during the recession as compared to women's wages. For low- and middle-skill men, their predicted wages declined – on average across all industries - $0.44 per hour between the 2004 and 2008 cohorts. Predicted wages for women declined just $0.06 per hour over the same time period, reflecting our descriptive finding that women fared better in terms of wages during the recession as compared to men. Men experienced the largest declines in retail (-$0.66 per hour) and manufacturing (-$0.58 per hour), but they experienced losses across all industries with the exception of construction, where wages went up slightly ($0.12 per hour). Women also experienced wage losses across all industries between the 2004 and 2008 cohorts, with the exception of manufacturing ($0.29 wage gain per hour), but again, their wage losses were much lower. For example, in the health care industry, predicted wages for women were $0.14 lower in the 2008 cohort as compared to the 2004 cohort. Because low- and middle-skill men experienced greater declines in wages during the recession, the wage gap between men and women declined in the 2008 cohort. In the food service, retail, manufacturing, and health care industries, we see substantial declines 8

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Fig. 2. Predicted wages for men and women by industry and cohort. Data source: SIPP, Notes: We used the natural log of inflation-adjusted wages to normalize the distribution of wages. Thus, the predicted wages from the random effects models are logged, but we used the exponential function in Excel to transform the logged predicted wages to non-logged wages for ease of interpretation.

in the 2008 cohort as compared to the 2004 cohort. For example, the wage gap in food service in 2004 was $2.54 per hour but dropped to $2.09 in the 2008 cohort; in the health care sector, the wage gap dropped from $1.52 per in the 2004 cohort to $1.14 in the 2008 cohort. Notably, the wage gap is smallest in the health care sector in the 2008 cohort. 5.3. Industry as a predictor of “good jobs” In this section, we examine industry as a predictor of earning wages of at least $15 per hour, the receipt of employer-based benefits, full-time hours, and layoff during the survey period. Full random effects models of earning wages of at least $15 per hour, the receipt of employer-based benefits, full-time hours, and layoff results for the 2004 and 2008 cohorts are presented in the Appendix (section 4, Table 6). In all models, we include interaction terms between gender and the selected industries in the models. In this section, we discuss the predicted probabilities based on our models of the four job quality measures for men and woman across industry and cohort. First, we examine the probability of earning at least $15 per hour. Fig. 3a shows the predicted probability of earning at least $15 per hour for men and women by industry and cohort. In the health care sector – the industry of primary interest in this paper – we find that men have about a 42% probability of earning at least $15 per hour and women have a 24% probability in the 2008 cohort. Women's probability is much higher in the health care sector as compared to the other two sectors that employ large percentages of low- and middle-skill women, namely retail (10% probability in the 2008 cohort) and food service (4% probability in the 2008 cohort). For men, the probability of earning at least $15 per hour is much higher in manufacturing (61% in the 2008 cohort) and construction (61% in the 2008 cohort) as compared to health care (48% in the 2008 cohort). These results are significantly different for men and women across all industries in both the 2004 and 2008 cohorts (p < .001). Second, we examine the likelihood of having employer-based health insurance. As described above, we include a measure of participation in rather than availability of employer-based health insurance because past research has shown that many workers do not take-up employer-based health insurance because it is too expensive (Baicker, Congdon, and Mullainathan 2012). The predicted probability of having employer-based health insurance for men and women across industry and cohort are shown in Fig. 3b. The health care industry ranks second among our selected industries in terms of providing employer-based health insurance, with both men and women had a 62% predicted probability of having employer-based health insurance in 2008. The manufacturing sector has the highest rate of workers with employer-based health insurance at 66% for both men and women in 2008. However, we see a substantial difference between cohorts. The probability of having employer-based health insurance fell for both men and women across all industries between 2004 and 2008 with an average decline of 6% between 2004 and 2008 across gender and industry. These results are significantly different for men and women across all industries in the 2004 cohort (p < .05), but in the 2008 cohort, there is not a significant difference between men and women in the receipt of employer-based health insurance in the health care, manufacturing, or construction industries. 9

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Third, we examine the probability of having fulltime hours; we use fulltime hours as a measure of job quality because there is evidence that full-time work is increasingly unavailable to low-skill workers and that there is growth in involuntary part-time employment (Even and Macpherson, 2018; Kalleberg et al., 2000; Lambert, 2008). The results are shown in Fig. 3c. Manufacturing provides highest probability of working fulltime for both men and women (91% and 88%, respectively), in the 2008 cohort. The health care industry and construction are next in terms of providing fulltime hours; the predicted probabilities of fulltime hours for men and women in the health care sector in the 2008 cohort are 78% and 67%, respectively, and 76% and 68%, respectively, in the construction sector. We see declines in the probability of fulltime hours across all industries during the recession, and similar to the pattern we see in predicted wages, the decline was more substantial among men. Across industries the decline for men was 5% on average, while the decline for women was 3% on average. The results are significantly different for men and women across all industries in both the 2004 and 2008 cohorts (p < .001). Finally, we examine the probability of an individual being on layoff during the survey period; being on layoff includes individuals who indicated that they were unemployed because of a layoff; the industry and education variables are lagged by one wave to reflect the industry and educational status of the individual before they were on layoff. As shown in Fig. 3d, we found that the health care sector has the lowest predicted probabilities of layoff for both men and women across both the 2004 and 2008 cohorts; the probability of being on layoff for men and women in the 2008 cohort was 0.94% and 0.70%, respectively. Across all industries for both men and women, there was an increase in the probability of layoff in the 2008 cohort, reflecting the impact of the recession. Notably,

Fig. 3. a–3d. Predicted probabilities from random-effects logit models predicting of “good job” qualities, for women and men. Data source: SIPP. 10

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Fig. 3. (continued)

there was a large increase in the probability of layoff for men in the construction industry in the 2008 cohort (2.41%). While the construction industry provides the highest wages for men and women (without large declines during the recession), there was also more instability in this sector during the recession. These results are significantly different for men and women in manufacturing and “all other” industries in the 2004 cohort (p < .05), but in the 2008 cohort, men have a significantly higher probability of layoff across all industries (p < .05). 5.4. Supplemental analysis In addition to measuring the probability of earning $15 per hour, we measure job quality using two higher wage thresholds: whether an individual earns $20 or $25 per hour. We examine the wage thresholds of $20 and $25 per hour to provide additional context for how many low- and middle-skill men have wages high enough not only to pull them out of poverty but provide a pathway into the middle class. Fig. 4a shows the predicted probability of earning at least $20 per hour for men and women by industry and cohort, and Fig. 4b shows the predicted probability of earning $25 per hour. When we set the wage threshold at these higher levels, the differences between industries and gender become more pronounced. For women, the predicted probability earning $20 per hour in the health 11

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Fig. 4. a. Predicted probabilities from random-effects logit models predicting earning at least $20 per hour, for women and men., Data source: SIPP, Figure 4b. Predicted probabilities from random-effects logit models predicting earning at least $25 per hour, for women and men.

care sector is just 3%, but this is higher than the food service and retail sectors, where the probability of earning $20 per hour in almost 0%. The predicted probability of earning $20 per hour in the health care sector is similar to manufacturing for women (about 3%), but slightly below construction (around 6% in 2008). When we look at the probability of earning $25 per hour, women have a predicted probability of roughly 0% across all industries. However, the health care industry actually provides the highest probability of earning $25 per hour for women, at 0.51% probability in 2008. For men, the manufacturing and construction industries emerge as providing far higher probabilities for earning $20 or $25 per hour as compared to the health care sector. The health care sector provides men with a 10% probability of earning $20 per hour, while manufacturing provides an 18% probability and construction provides a 27% probability (in 2008). The predicted probability of earning $25 per hour is much lower across all industries (just 2% in health care and 3% in manufacturing), but construction continues to provide clear advantages for low- and middle-skill men in the labor market. The full random effects models used to produce the predicted probabilities are not shown but are available upon request. 6. Discussion The health care sector has become an increasingly important employer in today's economy, as industries that used to dominate the U.S. economy, such as manufacturing and production, have declined (Frogner, 2018; Himmelstein and Venkataramani 2019). The

12

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

purpose of this study was to assess the health care sector as a source of “good jobs” for low- and middle skill workers by comparing measures of job quality across industries that employ different proportions of men and women. The health care industry is the largest employer of low- and middle-skill women (see Fig. 1a), and the percentage of women employed in the health care industry increases over time (from about 24% in 2008 to about 28% in 2013). Manufacturing and construction remained the top employers of men in the sample, but the percentage of men employed in these industries decreased over time, which is consistent with characterizations of the Great Recession as a “Man-cession” (Thompson, 2009). It is notable that we find, consistent with past research, that the recession hurt working class men more across all job quality measures as compared to women (Card and Mas, 2015). That said, the growth in the health care sector as an employer of low- and middle-skill workers – and the decline of industries like manufacturing – may have significant gendered implications for work and family life in the working class, something we will discuss more below. Based on our minimum thresholds of a “good job” along the dimensions of wages, benefits, hours, and job security, are health care jobs “good jobs” for workers without a four-year college degree? We find a gendered story: for low and middle-skill women, health care jobs are relatively good jobs, compared to alternatives in food service and retail – the two industries that employ the highest percentages of low- and middle-skill women behind health care - both of which have wages and earnings that are significantly lower than the health care sector. Gender segregation by occupation and industry creates unequal access to different forms of job quality for low and middle-skill women and men. Retail trade and food service are significantly less likely than health care jobs to provide wages that are at least $15 per hour and fulltime hours for women. Consistent with recent research (Budig et al., 2019), health care is a better option compared to other forms of care and service work available to women with less education. When women are able to access work in manufacturing and construction industries, their predicted wages are only slightly higher than wages in the health care sector (see Fig. 2), indicating that women do not benefit to the same degree as men in male-dominated occupations in manufacturing and construction. For men, however, there are substantial differences in terms of wages between jobs in the health care sector and jobs in manufacturing and construction, making the health care industry seem a less likely source of “good jobs” for low- and middle-skill men. Men working in manufacturing and construction earn wages of around $16 and $17 per hour, respectively, compared to wages just above $14 per hour in health care jobs (see Fig. 2). About 59% of men in manufacturing and 66% of men in construction earn wages above $15 per hour, compared to just 40% of men in health care, in the 2008 cohort (see Fig. 3a). Manufacturing also provides a higher rate of employer-based health insurance for men and women compared to health care and a higher probability of fulltime hours. That said, we found that the health care sector provided other benefits for low- and middle-skill men, namely greater job security. Men in the health care sector were less likely to report that they were on layoff during the recession compared to men in all industries that we identified, including food service, retail, manufacturing, construction, and other industries. The greater job stability that we see among health care workers likely reflects the greater demand for workers in the health care sector, particularly in comparison to industries that had substantial declines during the recession, including manufacturing and construction (Frogner, 2018; Goodman and Mance, 2011). We also find that men in health care had a higher probability of employer-based health insurance as compared to the construction industry. For example, only 54% of low- and middle-skill men in the construction industry had employer-based health insurance in the 2008 cohort, compared to 62% of men in the health care sector. Despite the advantages of employer-based health insurance and greater job security, the percentage of low- and middle-skill men in the health care sector is very low and increased only slightly during the recession (to about 5% of low- and middle-skill men; see Fig. 1b). Our findings are consistent with past research that has shown that men are reluctant to enter female-dominated health care occupations, many of which are strongly stigmatized as “women's work” (Gauchat et al., 2012; Jacobs, 1993; Simpson, 2005). Men are also likely reluctant to enter health care jobs because of lower wages compared to wages available in male-dominated industries such as manufacturing and construction (Levanon et al., 2009). That said, in line with past research (Price-Glynn and Carter, 2012), we do find evidence of a “glass escalator” for men in the health care sector and obtain upward wage mobility given that low- and middle-skill men's wages are higher in the health care sector as compared to women's wages (see Fig. 2). Although we do not measure it directly in the current study, past research has found that this may be especially true for racial minorities who may lack access to other pathways to upward mobility (Dill et al., 2016). The high demand for health care workers and the decline of many alternative job options traditionally held by men (e.g., manufacturing and production) may also be contributing to a narrowing of the wage gap between these industries (Bodenheimer et al., 2009; Schindel et al., 2006). Higher wages in manufacturing and construction are in part rooted in a history of labor organization in these sectors, which has worked to raise wages and secure benefits like health insurance and job security (Kalleberg, 2011). Declining rates of unionization have contributed to falling wages and fewer benefits for men in these industries (Brady et al., 2013). However, although generally low across industries, unionization rates in our sample remain substantially higher in male-dominated industries, at 15% and 19% respectively for manufacturing and construction as compared to only 8% in health care (see the impact of unionization and other job quality measures on wages in Appendix Table 4 and Appendix Table 5). In the health care sector, meanwhile, professions have largely used other closure strategies of licensure and certification to obtain higher wages and better job quality (Kleiner and Krueger 2010).11 We found some evidence that closure strategies help to raise wages; for example, health care workers have substantially better job quality compared to food service work, another form of low-skill, reproductive care work. However, by establishing higher certification requirements, these strategies are also limited in securing higher wages for workers, especially women and minorities 11 For context, 23% of workers in our sample in the health care industry have associate's degrees, while only 11% of workers in manufacturing and 8% of workers in construction have associate's degrees.

13

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

(Redbird, 2017; Budig et al., 2019). Importantly, our research supports Weeden (2002) finding that women do not benefit in the same way as men from closure strategies.12 In part, this may be the result of funding structures in the health care sector since many organizations that have low reimbursement rates – such as skilled nursing units, home health agencies – are constrained in their ability to raise wages, despite credentialing efforts on the part of workers and workforce development organizations. Although working class women earn lower wages in health care compared to men in male-dominated industries, the health care sector provides better job quality and stability compared to other service-sector options. Consequently, given the continued decline in blue-collar, male-dominated industries, the health care sector may increasingly replace manufacturing as a source of steady and reliable work for working class families. Thus, working class men may increasingly face a trade-off between higher wages available in manufacturing and construction or greater job stability in the health care sector. This shift in labor market opportunities towards feminized care work may have significant implications for gender expectations between men and women within working class families. In particular, we may see women - and especially women in the health care sector – making increasingly greater financial contributions in working class households as it becomes more difficult for men to secure wages that are sufficient for a “primary breadwinner” role. However, our findings show that even these “good jobs” for women fall short of blue-collar “good jobs” for men, contributing to growing hardship among working class families (Damaske, 2011). This study has a number of limitations. We do not capture workers' full career trajectories; the SIPP panels only provide a three-to four-year window into a worker's career. In addition, like other studies of career transitions, the outcomes we examine might be affected by unmeasured factors, such as dual employment, relocation, and leaves of absence. Finally, an examination of variation in the results by race/ethnicity and other individual demographics is beyond the scope of this study. Future research can expand on our work by modeling these outcomes separately by racial/ethnic categories and other characteristics to assess differences across groups. 7. Conclusion The health care sector has not replaced male-dominated industries as a new source of “good jobs” for all low- and middle-skill workers as manufacturing and construction industries continue to provide higher earnings for men. For women, however, the health care sector is providing opportunities for better jobs than what is available in other sectors. In other words, because of gender segregation across industries, job quality is gendered by industry, particularly in the working class where there are high levels of sex segregation across occupations and sectors (Cotter et al., 2004; Dwyer, 2013; Sutton et al., 2016). While working class women may find more opportunity and higher wages in the health care sector as compared to retail or food service, these “good jobs” are still gendered jobs with higher entry barriers, and are not equal to the “good jobs” in terms of wages that are available to men in manufacturing and construction. Acknowledgements The authors would like to thank Arne Kalleberg, Paula England, Michelle Budig, Nancy Folbre, Mignon Duffy, and Joann Xi for feedback on earlier drafts of this manuscript. This paper was presented at the American Sociological Association's 2017 annual meeting in Montreal, Canada. Appendix 1. Occupations by industry In Appendix Table 1 below, we include the highest frequency occupations by industry for men and women, and there are substantial differences by gender. For example, the top occupations for women are health care occupations, including nursing, psychiatric and home health aides, followed by occupations in retail and food service. For men in the healthcare industry, after nursing, psychiatric, and home health aides, the second most frequent occupation is janitors and building cleaners, an occupation within the health care industry but not a health care occupation itself, which falls in line with prior work that men are more likely to be employed in reproductive care jobs (such as food preparation and cleaning services), rather than nurturant care jobs (Duffy, 2007; Yavorsky et al., 2012). Similarly, in the construction industry, the most frequent occupations for men are construction laborers and carpenters, but the most frequent occupations for women in construction are secretaries and administrative assistants, and bookkeeping, accounting and auditing occupations. It is notable that the average wages for the largest occupations within health care are greater for women than men in Appendix Table 1, with the exception of “Nursing, psychiatric, and home health aides,” but men's average wages in health care are higher than women's average wages in Fig. 2 (in the main manuscript). This results from a few lower frequency but higher paying occupations in the health care sector (e.g., general and operations managers, health services managers, etc.) that employ higher numbers of men in the health care sector and bring up men's average wages. Median wages for men and women in the health care sector are fairly similar

12 An exception here is the occupation of nursing, which has experienced substantial wage gains over the last two decades. Licensure requirements for nurses – combined with increasing demand for health care services and a shortage of registered nurses – have worked to increase pay in this heavily feminized occupation. Moreover, unionization may also play a significant role in increasing wages (Coombs et al., 2015); rates of unionization among registered nurses has steadily increased since the 1970s to about 15% today.

14

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

($13.53 for women and $13.77 for men), but when we look at the 70th-90th percentiles, we see a growing distance in wages for men as compared to women. For example, hourly wages in the 90th percentile for women are $32.82 but $34.89 for men. Appendix Table 1

Highest frequency occupations by industry for women and men. Women Occupation Health care 1. Nursing, psychiatric, and home health aides 2. Registered nurses 3. Medical assistants and other health care workers Retail 1. Retail salespersons 2. Cashiers 3. First-line supervisors of retail salespersons Food service 1. Waiters and waitresses 2. Cashiers 3. Cooks Manufacturing 1. Miscellaneous assemblers and fabricators 2. Production workers, all other 3. Sewing machine operators Construction 1. Secretaries and administrative assistants 2. Bookkeeping, accounting and auditing 3. First-line supervisors of office and administrative support workers

Men Obs

Wage

11,517 8294 3854

$12.25 $25.57 $14.47

9856 11,089 5742

$10.85 $9.83 $14.16

8368 3501 3340

$6.63 $8.58 $9.22

2384 1692 1339

$13.71 $13.88 $10.71

1022 602 280

$15.44 $14.87 $21.74

Occupation Health care 1. Nursing, psychiatric, and home health aides 2. Janitors and building cleaners 3. Personal and home care aides Retail 1. Retail salespersons 2. First-line supervisors of retail salespersons 3. Stock clerks and order fillers Food service 1. Cooks 2. Waiters and waitresses 3. Food preparation workers Manufacturing 1. Production workers, all other 2. Miscellaneous assemblers and fabricators 3. First-line supervisors of production and operating workers Construction 1. Construction laborers 2. Carpenters 3. First-line supervisors of construction trades and extraction workers

Obs

Wage

1126 824 628

$13.96 $11.72 $11.03

7530 5368 4274

$13.97 $19.33 $10.92

6096 2603 1577

$10.20 $7.90 $9.64

3803 3404 3010

$16.01 $16.17 $22.66

6538 4979 3298

$16.26 $18.45 $24.17

Data source: SIPP

2. Fixed effects analysis In Appendix Table 2 below, we present fixed effects models predicting inflation-adjusted hourly wages and our dichotomous measures of job quality for men and women, by cohort. Fixed effects models reveal how changes in the dependent variable are a function of changes in the independent variables. Thus, since individuals serve as their own controls, the models reduce omitted variable bias that results from unmeasured, stable personal characteristics (Allison 2009), such as early career expectations. In our first analysis, we use fixed effects linear regression models to measure changes in logged inflation-adjusted hourly wages. We interpret a change in the dependent variable – inflation-adjusted wages - based on changes in the independent variables. In other words, fixed effects models measure how industry transitions are related to changes in inflation-adjusted wages. Our second set of analyses presents the results of fixed effects models of our dichotomous measures of job quality, including earning at least $15 per hour, employer-based health insurance, working full-time hours, and layoff. We do not include predicted probabilities for these models because calculating predicted probabilities based on fixed effects logit models can be problematic.13 Because fixed effects models look at changes in the independent and dependent variables, observations where no change occurs are not included in the model. This is reflected in the sample sizes of the fixed effects models presented below; there is substantial variability in the sample sizes between models, depending on the degree of variation within individuals in terms of each outcome. Finally, fixed effects models do not allow the inclusion of time-invariant variables, such as gender. To address this limitation, we ran fixed effects models in two ways: 1) we ran separate models for men and women by cohort, and 2) we ran models interacting industry with gender but omitting the main effect for gender. The results of the FE models that were run separately by gender and cohort are presented below; the FE models with interaction terms but no main effect for gender are available upon request. We found similar patterns across all models, suggesting that the results shown below are robust. The findings from our fixed effects models of inflation-adjusted wages (shown below in Appendix Table 2) are very similar to our random effects models presented in the main manuscript (Table 2 and Fig. 2). The predicted wages for women and women by industry and cohort (shown in Appendix Fig. 1 below) are slightly higher in the fixed effects models as compared to the random effects models in the main manuscript. For example, predicted wages for women in the fixed effects models are $13.45 per hour (see Appendix Fig. 1 below) and $13.06 in the random effects models in the main manuscript (see Fig. 2 in the main manuscript). However, the overall pattern of the distribution of wages across industries and gender is the same. The results of our fixed effects models of dichotomous job quality measures are shown in Tables 3a through 3d below. Comparing

13 For a discussion of why predicted probabilities can be problematic after using “xtlogit fe” see: https://www3.nd.edu/~rwilliam/stats3/ Panel03-FixedEffects.pdf and https://www.statalist.org/forums/forum/general-stata-discussion/general/1304704-cannot-estimate-marginal-effectafter-xtlogit.

15

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

the results of the fixed effects logit models of “good job” qualities to the random effects logit models of “good job” qualities in the main manuscript is limited because we cannot calculate predicted probabilities from the fixed effects models. Further, the structure of fixed effects models shown here are different (with men and women modeled separately). That said, the odds ratios in the fixed effects models tell a similar story. The retail and food service industries lag behind the health care sector across all measures of job quality for both men and women (p < .001), while construction and manufacturing provide higher odds of earning $15 per hour as compared to health care, especially for men (p < .001). Manufacturing provides a higher odds of employer-based health insurance as compared to health care (p < .001), both men and women have lower odds of having employer-based health insurance in construction as compared to health care (p < .001). Men in all industries have higher odds of layoff during the recession as compared to the health care industry (p < .001). Appendix Table 2

Fixed effects models predicting inflation-adjusted hourly wage for women and men, by cohort. 2004 Women

Selected industries Health care Food service Retail Manufacturing Construction All other Personal characteristics Age Age squared Married Parent of kids under 18 Education High school degree or less Some college or associate degree College degree or more Region Northeast Midwest South West Constant R2 within R2 between R2 overall Observations Individuals

2008 Women

2004 Men

2008 Men

Coef

SE

Coef

SE

Coef

SE

Coef

SE

−0.268*** −0.094*** 0.012 0.049*** −0.027***

0.006 0.006 0.008 0.014 0.005

−0.262*** −0.104*** 0.050*** 0.060*** −0.025***

0.006 0.005 0.007 0.014 0.005

−0.155*** −0.038*** 0.077*** 0.101*** 0.021*

0.011 0.010 0.010 0.010 0.009

−0.164*** −0.050*** 0.072*** 0.143*** 0.021*

0.010 0.009 0.010 0.010 0.009

0.049*** 0.000*** 0.012* −0.005

0.003 0.000 0.005 0.004

0.031*** 0.000*** 0.030*** −0.002

0.002 0.000 0.005 0.004

0.047*** −0.001*** 0.027*** 0.004

0.003 0.000 0.006 0.004

0.041*** 0.000*** 0.017*** −0.006

0.002 0.000 0.005 0.004

−0.009 0.071***

0.007 0.011

−0.004 0.035***

0.006 0.009

0.005 0.052***

0.007 0.012

0.007 0.018

0.006 0.011

−0.049* −0.074*** −0.023 1.578*** 0.03 0.2256 0.1908 128,441 23,097

0.025 0.021 0.024 0.067

0.027 −0.040 0.067*** 1.932*** 0.0272 0.2273 0.1959 144,401 22,111

0.022 0.020 0.022 0.054

−0.008 −0.066*** 0.036 1.746*** 0.0197 0.2838 0.2362 134,526 24,209

0.023 0.020 0.023 0.060

0.057* 0.046* 0.111*** 1.808*** 0.0208 0.2584 0.2128 149,645 23,352

0.025 0.022 0.024 0.053

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Race/ethnicity is not included in the models because fixed effects models do not allow for time invariant variables. Dummy variables for year of survey period are included but not shown.

Appendix Fig. 1. Predicted wages for men and women by industry and cohort (based on FE models presented in Table 2), Data source: SIPP, Notes: Predicted probabilities were estimated using logged wages and converted post-analysis to dollars..

16

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Appendix Table 3a

Fixed-effects logit models predicting the odds of earning at least $15 per hour, for women and men. 2004 Women

Selected industries Health care Food service Retail Manufacturing Construction All other Personal characteristics Age Age squared Married Parent of kids under 18 Education High school degree or less Some college or associate degree College degree or more Region Northeast Midwest South West Observations Individuals

2008 Women

2004 Men

2008 Men

OR

SE

OR

SE

OR

SE

OR

SE

Ref 0.232*** 0.303*** 0.900 1.424 0.703***

0.026 0.028 0.100 0.263 0.050

Ref 0.443*** 0.722* 2.020*** 2.210*** 1.243

0.067 0.094 0.261 0.290 0.150

Ref 0.210*** 0.410*** 1.714*** 1.798*** 0.928

0.023 0.036 0.183 0.321 0.061

Ref 0.415*** 0.647*** 2.292*** 3.468*** 1.379**

0.056 0.078 0.269 0.414 0.150

1.914*** 0.994*** 1.143 0.731***

0.098 0.001 0.099 0.053

2.032*** 0.992*** 1.372*** 0.992

0.083 0.000 0.107 0.060

1.657*** 0.994*** 1.236** 0.991

0.062 0.000 0.094 0.063

1.755*** 0.994*** 1.380*** 0.802***

0.058 0.000 0.095 0.044

Ref 1.075 3.223***

0.146 0.621

Ref 1.154 1.558*

0.132 0.278

Ref 0.910 1.564***

0.113 0.237

Ref 1.228** 1.334

0.122 0.202

Ref 0.605 0.500* 1.156 34,761 4733

0.205 0.141 0.404

Ref 0.510* 0.453*** 0.754 42,996 4900

0.140 0.103 0.197

Ref 1.182 0.714 1.408 43,094 6020

0.391 0.200 0.502

Ref 1.424 1.468 1.797 51,883 6135

0.439 0.396 0.556

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Race/ethnicity is not included in the models because fixed effects models do not allow for time invariant variables. Dummy variables for year of survey period are included but not shown.

Appendix Table 3b

Fixed-effects logit models predicting the odds of having employer-based health insurance, for women and men. 2004 Women

Selected industries Health care Food service Retail Manufacturing Construction All other Personal characteristics Age Age squared Married Parent of kids under 18 Education High school degree or less Some college or associate degree College degree or more Other insurance (lagged) Region Northeast Midwest South West Observations Individuals

2008 Women

2004 Men

2008 Men

OR

SE

OR

SE

OR

SE

OR

SE

Ref 0.524*** 0.696*** 1.223* 1.020 0.820***

0.045 0.054 0.126 0.198 0.054

Ref 0.552*** 0.785*** 1.217* 0.694* 0.762***

0.044 0.057 0.120 0.122 0.046

Ref 0.535*** 0.762* 1.175 0.600*** 0.773*

0.073 0.096 0.152 0.079 0.092

Ref 0.597*** 0.858 1.307* 0.720** 0.788*

0.070 0.094 0.146 0.082 0.080

1.034 0.999 2.109*** 1.005

0.048 0.000 0.174 0.069

1.308*** 0.997*** 1.778*** 0.989

0.042 0.000 0.115 0.055

1.108* 0.999* 1.854*** 1.121

0.046 0.000 0.158 0.070

1.381*** 0.996*** 1.299*** 1.085

0.041 0.000 0.088 0.057

Ref 0.809* 0.802 0.483***

0.073 0.125 0.018

Ref 0.947 0.712** 0.582***

0.080 0.088 0.021

Ref 0.897 0.745 0.442***

0.085 0.120 0.016

Ref 1.073 0.939 0.721***

0.086 0.118 0.025

Ref 0.507* 0.348*** 0.256*** 35,661 5471

0.160 0.094 0.077

Ref 1.234 1.039 0.849 51,489 6490

0.320 0.257 0.242

Ref 1.045 0.644 1.280 38,241 5854

0.311 0.165 0.375

Ref 1.150 0.645 0.848 55,454 7063

0.362 0.175 0.259

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Race/ethnicity is not included in the models because fixed effects models do not allow for time invariant variables. Dummy variables for year of survey period are included but not shown. 17

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Appendix Table 3c

Fixed-effects logit models predicting the odds of working fulltime hours, for women and men.

Selected industries Health care Food service Retail Manufacturing Construction All other Personal characteristics Age Age squared Married Parent of kids under 18 Education High school degree or less Some college or associate degree College degree or more Region Northeast Midwest South West Observations Individuals

2004 Women

2008 Women

2004 Men

2008 Men

OR

SE

OR

SE

OR

SE

OR

SE

0.440*** 0.628*** 1.370*** 1.103 0.841***

0.028 0.037 0.108 0.169 0.042

0.418*** 0.523*** 1.416*** 1.064 0.792***

0.025 0.029 0.110 0.158 0.037

0.484*** 0.723*** 1.648*** 0.916 0.888

0.052 0.073 0.171 0.096 0.085

0.424*** 0.643*** 1.598*** 0.867 0.936

0.041 0.059 0.151 0.081 0.080

1.448*** 0.996*** 1.006 0.790***

0.046 0.000 0.060 0.037

1.265*** 0.997*** 0.985 0.905*

0.030 0.000 0.051 0.038

1.421*** 0.996*** 1.112 0.932

0.043 0.000 0.072 0.045

1.365*** 0.997*** 1.025 0.805***

0.032 0.000 0.056 0.034

1.123 1.654***

0.085 0.192

0.906 1.278**

0.061 0.121

0.858* 1.289

0.065 0.169

0.865* 1.178

0.057 0.129

0.603* 0.637* 0.524* 72,557 10,230

0.154 0.142 0.131

1.321 1.381 1.114 87,047 10,175

0.294 0.280 0.248

0.404*** 0.600* 0.520** 68,394 9775

0.102 0.137 0.131

1.061 1.283 0.625 84,929 10,161

0.282 0.303 0.166

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Race/ethnicity is not included in the models because fixed effects models do not allow for time invariant variables. Dummy variables for year of survey period are included but not shown.

Appendix Table 3d

Fixed-effects logit models predicting the odds of layoff, for women and men. 2004 Women

Selected industries Health care Food service Retail Manufacturing Construction All other Personal characteristics Age Age squared Married Parent of kids under 18 Education High school degree or less Some college or associate degree College degree or more Region Northeast Midwest South West Observations Individuals

2008 Women

2004 Men

2008 Men

OR

SE

OR

SE

OR

SE

OR

SE

0.980 1.314 1.749** 2.338* 1.166

0.166 0.207 0.353 1.014 0.165

1.178 1.137 1.810*** 1.542 1.313*

0.162 0.150 0.297 0.508 0.149

1.255 1.112 1.474 1.421 1.248

0.350 0.287 0.389 0.379 0.303

1.665* 1.903*** 2.158*** 2.441*** 1.890***

0.332 0.376 0.429 0.484 0.346

1.130 0.999 0.851 0.695*

0.146 0.001 0.175 0.118

0.939 1.000 1.034 0.902

0.069 0.001 0.163 0.112

1.159 0.998 1.092 0.891

0.124 0.001 0.250 0.136

0.820*** 1.001* 0.968 0.999

0.049 0.001 0.134 0.098

0.953 1.848

0.272 0.846

0.802 0.649

0.149 0.181

0.951 1.528

0.231 0.665

1.041 1.085

0.158 0.298

1.769 1.307 0.809 7991 1239

1.106 0.665 0.482

1.339 0.456 1.362 14,097 1937

0.776 0.267 0.855

0.958 0.537 0.267* 8589 1310

0.643 0.295 0.173

0.904 0.876 0.820 18,964 2601

0.517 0.432 0.431

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Race/ethnicity is not included in the models because fixed effects models do not allow for time invariant variables. Dummy variables for year of survey period are included but not shown.

18

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

3. Random effects analysis with the inclusion of work-related characteristics In the analyses included in the main manuscript, we do not include work-related characteristics in our models (e.g., tenure, hours, salaried job, union member). We did not include these characteristics because we are primarily interested in the total effect of industry on job quality; work characteristics may be related to the industry in which an individual is working and act as intervening variables between industry and the outcomes in our analyses. However, in Appendix Tables 4 and 5 we provide models with work characteristics included for comparison. In Appendix Figure 4 below, we present predicted wages based on the random effects models presented in Appendix Table 4. We measure individual's tenure in their current job (in months, divided by 100) and whether an individual works fulltime (yes = 1, no = 0). We also control for whether the respondent is in a salaried position as compared to a job that pays an hourly rate (1 = salaried job, 0 = not salaried),14 and whether an individual is in a union (yes = 1, no = 0). We include the type of organization for which an individual works: 1) private, for profit (reference), 2) private nonprofit, or 3) public. Work characteristics are largely significant predictors of inflation-adjusted wages, as shown in Appendix Table 4. For example, individuals who are fulltime have wages that are 3.4% higher than those that are not fulltime (p < .001), while those who work for nonprofit firms have wages that are 2.1% lower than those that work for private, for profit firms. That said, the inclusion of work characteristics in our models has very little effect on the industry and gender coefficients and odds ratios in our analyses (see Table 2 in the main manuscript). The predicted wages shown in Appendix Fig. 2 are nearly identical to the predicted wages shown in the main manuscript. In Appendix Table 5 we present the results of random effects models of “good job” characteristics, with the inclusion of work characteristics. Again, the work characteristic variables – tenure, hours, whether a job is salaried, and whether an individual is a union member – are all strongly significant across the models. However, the patterns of industry and gender as predictors of “good job” characteristics are the same as compared to models without the inclusion of work characteristics. In other words, the effect of industry and gender on the probability of having these “good job” characteristics cannot be explained by work characteristics. The predicted probabilities for the models shown in Appendix Table 5 are available upon request. Appendix Table 4

Coefficients from random-effects models predicting logged inflation-adjusted hourly wages, by cohort, with the inclusion of work-related characteristics. 2004 Cohort

Selected industries Health care Food service Retail Manufacturing Construction All other Gender Female Male Interaction terms Health care*Female Food service*Female Retail*Female Manufacturing*Female Construction*Female All other*Female Personal characteristics Age Age squared Married Parent of kids under 18 Race/ethnicity White Black Latino Other minority Education High school degree or less Some college or associate degree

2008 Cohort

Coef

SE

Coef

SE

Reference −0.177*** −0.031*** 0.085*** 0.138*** 0.039***

0.009 0.008 0.008 0.009 0.008

Reference −0.180*** −0.046*** 0.087*** 0.171*** 0.040***

0.008 0.008 0.008 0.008 0.007

−0.086*** Reference

0.009

−0.067*** Reference

0.008

−0.124*** −0.083*** −0.103*** −0.074*** −0.077***

0.010 0.010 0.010 0.015 0.009

−0.114*** −0.076*** −0.073*** −0.105*** −0.080***

0.010 0.009 0.010 0.015 0.008

0.045*** 0.000*** 0.050*** 0.005

0.001 0.000 0.003 0.002

0.036*** 0.000*** 0.057*** 0.000

0.001 0.000 0.003 0.002

Reference −0.095*** −0.118*** −0.071***

0.005 0.005 0.007

Reference −0.079*** −0.118*** −0.070***

0.005 0.005 0.007

0.073***

0.003

0.073***

0.003

(continued on next page)

14 The SIPP includes a measure of both hourly wages and monthly earnings for both hourly and salary workers. Where hourly wages were missing, we calculated hourly wages based on monthly earnings and the number of reported hours per week.

19

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Appendix Table 4 (continued) 2004 Cohort

College degree or more Work characteristics Tenure (divided by 100) Always fulltime Salaried job Union member Employer type Private, forprofit Private, nonprofit Public Region Northeast Midwest South West Constant R2 within R2 between R2 overall Observations Individuals

2008 Cohort

Coef

SE

Coef

SE

0.193***

0.005

0.158***

0.004

0.087*** 0.034*** 0.026*** 0.129***

0.002 0.002 0.002 0.004

0.088*** 0.032*** 0.015*** 0.123***

0.001 0.001 0.002 0.003

Reference −0.022*** −0.004

0.005 0.004

Reference −0.021*** 0.012***

0.004 0.003

Reference −0.058*** −0.086*** 0.023*** 1.661*** 0.0286 0.3873 0.3419 261,214 46,965

0.005 0.005 0.006 0.017

Reference −0.065*** −0.070*** 0.037*** 1.718*** 0.0310 0.3938 0.3495 291,898 45,041

0.005 0.005 0.005 0.016

*p < .05, **p < .01, ***p < .001. Data source: SIPP Notes: Dummy variables for year of survey period are included but not shown.

Appendix Fig. 2. Predicted wages for men and women by industry and cohort (based on RE models presented in Table 3), Data source: SIPP, Notes: Predicted probabilities were estimating using logged wages and converted post-analysis to dollars. .

20

Selected industries Health care Food service Retail Manufacturing Construction All other Gender Female Male Interaction terms Health care*Female Food service*Female Retail*Female Manufacturing*Female Construction*Female All other*Female Personal characteristics Age Age squared Married Parent of kids under 18 Race/ethnicity White Black Latino Other minority Education High school degree or less Some college Associate degree Work characteristics Tenure (divided by 100) Always fulltime Salaried job Union member Employer type Private, forprofit Private, nonprofit Public Other health insurance Region Northeast

21

0.01 0.01 0.03

0.15 0.81

0.06 0.04 0.05 0.29

0.05 0.06

Ref 0.21*** 0.18*** 0.36***

Ref 3.48*** 12.59***

2.89*** 1.63*** 1.85*** 5.76***

Ref 0.77*** 1.18**

Ref

0.02 0.00 0.07 0.03

0.06 0.03 0.03 0.10 0.04

0.05

0.03 0.08 0.35 0.58 0.17

1.94*** 0.99*** 1.87*** 0.94

Ref 0.35*** 0.22*** 0.23*** 0.49** 0.36***

0.38*** Ref

Ref 0.20*** 0.72* 2.99*** 4.86*** 1.58***

Ref

Ref 0.96 1.40***

2.45*** 1.67*** 1.58*** 4.41***

Ref 2.95*** 7.60*

Ref 0.30*** 0.20*** 0.40***

1.73*** 0.99*** 2.16*** 0.90***

Ref 0.37*** 0.35*** 0.38*** 0.32*** 0.43***

0.41*** Ref

Ref 0.21*** 0.62*** 3.57*** 7.85*** 1.73***

0.06 0.07

0.04 0.04 0.04 0.21

0.12 0.44

0.02 0.01 0.04

0.02 0.00 0.08 0.03

0.06 0.05 0.05 0.06 0.05

0.04

0.03 0.07 0.38 0.85 0.17

Std Err

Ref

Ref 1.47*** 1.79*** 0.20***

2.94*** 2.03*** 1.60*** 2.89***

Ref 2.05*** 3.09***

Ref 0.36*** 0.14*** 0.42***

1.06*** 1.00*** 3.05*** 1.04

Ref 1.02 0.85 1.08 3.16*** 1.22

0.99 Ref

Ref 0.32*** 0.82 1.54*** 0.27*** 0.61***

Odds Ratio

0.09 0.09 0.01

0.07 0.05 0.05 0.15

0.08 0.18

0.02 0.01 0.03

0.01 0.00 0.12 0.03

0.14 0.11 0.15 0.63 0.14

0.11

0.04 0.09 0.17 0.03 0.06

Std Err

Odds Ratio

Odds Ratio

Std Err

2004

2008

2004

Ref

Ref 1.24*** 1.93*** 0.38***

2.15*** 2.10*** 1.59*** 2.35***

Ref 2.21*** 2.92***

Ref 0.38*** 0.14*** 0.45***

1.02* 1.00** 2.48*** 0.93*

Ref 0.79* 0.81 0.88 1.51* 0.92

1.24* Ref

Ref 0.42*** 0.97 1.67*** 0.42*** 0.75***

Odds Ratio

2008

Employer-based health insurance

$15 per hour

0.06 0.08 0.01

0.04 0.04 0.04 0.11

0.08 0.15

0.02 0.01 0.03

0.01 0.00 0.08 0.03

0.09 0.09 0.11 0.28 0.09

0.12

0.04 0.09 0.16 0.04 0.06

Std Err

Ref

Ref 0.67*** 0.86***

1.51*** – 2.2s8*** 1.29***

Ref 0.83*** 1.04

Ref 1.37*** 1.80*** 1.01

1.38*** 1.00*** 1.18*** 0.82***

Ref 0.93 0.82* 0.96 1.11 0.80**

0.40*** Ref

Ref 0.36*** 0.67*** 1.86*** 0.86 0.86*

Odds Ratio

2004

0.03 0.03

0.05 0.04

0.02

0.02 0.04

0.05 0.07 0.05

0.01 0.00 0.03 0.02

0.09 0.07 0.09 0.15 0.07

0.03

0.03 0.05 0.15 0.07 0.06

Std Err

Fulltime hours

Random-effects logit models predicting of “good job” qualities for women and men, with the inclusion of work characteristics

Appendix Table 5

Ref

Ref 0.78*** 1.00

1.49*** – 2.51*** 1.26***

Ref 0.90*** 1.15***

Ref 1.17*** 1.47*** 1.01

1.40*** 1.00*** 1.15*** 0.83***

Ref 0.89 0.68*** 0.82* 1.10 0.65***

0.54*** Ref

Ref 0.35*** 0.68*** 2.02*** 0.83** 0.99

Odds Ratio

2008

0.03 0.03

0.05 0.04

0.02

0.02 0.04

0.05 0.05 0.05

0.01 0.00 0.03 0.02

0.08 0.06 0.08 0.15 0.05

0.04

0.03 0.05 0.15 0.06 0.07

Std Err

Ref

Ref 0.83 0.63***

0.45*** 0.84*** 0.64*** 0.78**

Ref 0.88** 0.72***

Ref 1.71*** 1.03 1.36***

1.05*** 1.00*** 0.63*** 1.00

Ref 1.07 1.25 1.60* 0.99 1.16

0.84 Ref

Ref 1.36 1.13 1.09 1.53* 1.17

Odds Ratio

2004

Layoff

Ref

Ref 0.74*** 0.57***

0.45*** 0.92* 0.69*** 0.77***

Ref 0.84*** 0.75***

Ref 1.41*** 0.95 1.04

0.99 1.00 0.75*** 1.00

Ref 0.98 1.18 1.22 0.82 1.07

0.79 Ref

Ref 1.38* 1.15 1.64*** 2.39*** 1.43**

Odds Ratio

0.06 0.04

0.02 0.03 0.03 0.05

0.03 0.04

0.07 0.05 0.07

0.01 0.00 0.03 0.04

0.16 0.19 0.20 0.20 0.16

0.11

0.19 0.16 0.22 0.32 0.18

Std Err

(continued on next page)

0.09 0.06

0.02 0.04 0.04 0.07

0.04 0.06

0.11 0.07 0.12

0.01 0.00 0.03 0.05

0.23 0.26 0.35 0.32 0.23

0.15

0.26 0.21 0.20 0.28 0.21

Std Err

2008

J. Dill and M.J. Hodges

Social Science Research 84 (2019) 102350

0.53*** 0.40*** 1.17* 0.00*** 264,634 47,203

0.04 0.03 0.08 0.00

0.42*** 0.43*** 1.14* 0.00*** 295,840 45,329

0.03 0.03 0.08 0.00

1.04 0.72*** 0.72*** 0.58** 225,994 42,413

Odds Ratio 0.06 0.04 0.05 0.12

Std Err 0.94 0.67*** 0.80*** 0.60** 259,955 40,728

Odds Ratio 0.05 0.04 0.05 0.11

Std Err 0.85*** 1.57*** 0.93 0.01*** 264,634 47,203

Odds Ratio

2004

0.03 0.06 0.04 0.00

Std Err

Fulltime hours

0.88*** 1.39*** 0.83*** 0.00*** 295,840 45,329

Odds Ratio

2008

0.03 0.05 0.03 0.00

Std Err 0.98 0.82** 0.90 0.01*** 218,378 40,458

Odds Ratio

2004

Layoff

0.07 0.06 0.07 0.00

Std Err 1.03 0.97 1.14* 0.03*** 251,498 38,587

Odds Ratio

2008

0.06 0.05 0.07 0.01

Std Err

22

Selected industries Health care Food service Retail Manufacturing Construction All other

Appendix Table 6

Ref 0.18*** 0.76* 4.56*** 5.59*** 2.16*** 0.03 0.09 0.55 0.69 0.25

Ref 0.18*** 0.63*** 4.74*** 9.61*** 2.24*** 0.02 0.07 0.52 1.08 0.23

Std Err Ref 0.27*** 0.71** 1.78*** 0.27*** 0.69***

Odds Ratio

0.03 0.08 0.19 0.03 0.07

Std Err

Odds Ratio

Odds Ratio

Std Err

2004

2008

2004

Ref 0.35*** 0.85 1.81*** 0.45*** 0.82*

Odds Ratio

2008

Employer-based health insurance

$15 per hour

Random-effects logit models predicting of “good job” qualities, for women and men.

0.04 0.08 0.17 0.04 0.07

Std Err

Ref 0.40*** 0.75*** 2.30*** 0.93 1.03

Odds Ratio

2004

0.03 0.06 0.18 0.07 0.08

Std Err

Fulltime hours

Ref 0.37*** 0.74*** 2.29*** 0.86* 1.15*

Odds Ratio

2008

0.03 0.05 0.17 0.06 0.08

Std Err

Ref 1.45* 1.13 0.93 1.53** 0.95

Odds Ratio

2004

Layoff

Ref 1.46** 1.34* 1.80*** 2.60*** 1.41**

Odds Ratio

0.18 0.16 0.21 0.30 0.15

Std Err

(continued on next page)

0.23 0.17 0.14 0.23 0.13

Std Err

2008

Appendix Table 6. below includes the random effects models of job quality measures used to calculate the predicted probabilities presented in Fig. 3a–d in the main manuscript.

4. Random effects models of earning wages of at least $15 per hour, the receipt of employer-based benefits, full-time hours, and layoff

*p < .05, **p < .01. Data source: SIPP Notes: R2 statistics are not available for random effects logit models. The models of employer-based health insurance, we include whether an individual has health insurance from a source other than their employer, and this variable is lagged by one wave. In the models of the likelihood of experiencing a layoff, all industry and education variables are lagged by one wave to indicate the industry and education of individuals prior to layoff. Dummy variables for year of survey were included in the models but are not shown in table.

Midwest South West Constant Observations Individuals

Std Err

Odds Ratio

Odds Ratio

Std Err

2004

2008

2004

2008

Employer-based health insurance

$15 per hour

Appendix Table 5 (continued)

J. Dill and M.J. Hodges

Social Science Research 84 (2019) 102350

23

Std Err

0.01 0.01 0.03

Ref 0.18*** 0.12*** 0.26***

Ref 0.38*** 0.26*** 0.89 0.00*** 262,967 47,262 0.03 0.02 0.07 0.00

0.21 1.40

0.03 0.00 0.11 0.04

2.34*** 0.99*** 2.40*** 0.92*

Ref 4.24*** 19.27*** –

0.07 0.03 0.03 0.14 0.05

0.02

Ref 0.36*** 0.20*** 0.22*** 0.65* 0.37***

0.19*** Ref

Ref 0.34*** 0.31*** 1.00 0.00*** 294,046 45,494

Ref 3.38*** 9.76*** –

Ref 0.26*** 0.14*** 0.31***

1.97*** 0.99*** 2.57*** 0.91***

Ref 0.39*** 0.33*** 0.43*** 0.40*** 0.45***

0.24*** Ref

0.02 0.02 0.08 0.00

0.16 0.63

0.02 0.01 0.03

0.02 0.00 0.10 0.03

0.06 0.05 0.06 0.09 0.05

0.03

Ref 0.93 0.61*** 0.57*** 0.30*** 228,402 42,828

Ref 2.10*** 3.42*** 0.20***

Ref 0.38*** 0.12*** 0.36***

1.16*** 1.00*** 3.54*** 1.03

Ref 1.05 0.83 1.03 2.41*** 1.21

0.77* Ref

Odds Ratio

0.06 0.04 0.04 0.06

0.08 0.20 0.01

0.02 0.01 0.03

0.01 0.00 0.14 0.03

0.14 0.10 0.14 0.47 0.14

0.09

Std Err

Odds Ratio

Odds Ratio

Std Err

2004

2008

2004

Ref 0.86* 0.58*** 0.69*** 0.32*** 262,681 41,130

Ref 2.28*** 3.17*** 0.37***

Ref 0.37*** 0.12*** 0.40***

1.10*** 1.00*** 2.75*** 0.92**

Ref 0.84 0.79* 0.96 1.32 0.97

1.00 Ref

Odds Ratio

2008

Employer-based health insurance

$15 per hour

0.05 0.03 0.04 0.06

0.08 0.17 0.01

0.02 0.01 0.03

0.01 0.00 0.09 0.03

0.10 0.09 0.12 0.24 0.10

0.10

Std Err

Ref 0.78*** 1.50*** 0.84*** 0.01*** 262,967 47,262

Ref 0.87*** 1.14*** –

Ref 1.31*** 1.72*** 0.96

1.43*** 1.00*** 1.29*** 0.81***

Ref 0.94 0.80* 0.96 1.32 0.80***

0.36*** Ref

Odds Ratio

2004

0.03 0.06 0.04 0.00

0.02 0.04

0.05 0.07 0.05

0.01 0.00 0.03 0.02

0.09 0.07 0.09 0.19 0.07

0.03

Std Err

Fulltime hours

Ref 0.80*** 1.31*** 0.77*** 0.00*** 294,046 45,494

Ref 0.95 1.30*** –

Ref 1.11** 1.39*** 0.94

1.45*** 1.00*** 1.25*** 0.82***

Ref 0.90 0.65*** 0.85 1.26 0.68***

0.49*** Ref

Odds Ratio

2008

0.03 0.05 0.03 0.00

0.02 0.05

0.04 0.05 0.05

0.01 0.00 0.03 0.02

0.08 0.05 0.08 0.18 0.05

0.04

Std Err

Ref 1.01 0.79*** 0.90 0.01*** 227,810 42,080

Ref 0.72*** 0.57*** –

Ref 2.34*** 1.31*** 1.70***

1.02* 1.00*** 0.50*** 0.99

Ref 1.04 1.34 1.55* 1.15 1.43*

0.80 Ref

Odds Ratio

2004

Layoff

0.07 0.05 0.06 0.00

0.03 0.04

0.14 0.08 0.14

0.01 0.00 0.02 0.04

0.19 0.24 0.29 0.30 0.23

0.12

Std Err

Ref 1.07 1.02 1.37*** 0.02*** 267,899 40,813

Ref 0.75*** 0.58*** –

Ref 1.82*** 1.17*** 1.19*

1.02 1.00*** 0.57*** 1.05

Ref 1.17 1.18 1.17 0.79 1.22

0.74** Ref

Odds Ratio

2008

0.06 0.05 0.08 0.00

0.03 0.03

0.10 0.06 0.09

0.01 0.00 0.02 0.04

0.17 0.16 0.17 0.17 0.15

0.09

Std Err

*p < .05, **p < .01. Data source: SIPP Notes: R2 statistics are not available for random effects logit models. The models of employer-based health insurance, we include whether an individual has health insurance from a source other than their employer, and this variable is lagged by one wave. In the models of the likelihood of experiencing a layoff, all industry and education variables are lagged by one wave to indicate the industry and education of individuals prior to layoff. Dummy variables for year of survey were included in the models but are not shown in table.

Gender Female Male Interaction terms Health care*Female Food service*Female Retail*Female Manufacturing*Female Construction*Female All other*Female Personal characteristics Age Age squared Married Parent of kids under 18 Race/ethnicity White Black Latino Other minority Education High school degree or less Some college Associate degree Health insurance from other source Region Northeast Midwest South West Constant Observations Individuals

Appendix Table 6 (continued)

J. Dill and M.J. Hodges

Social Science Research 84 (2019) 102350

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

References Abraham, Katharine G., Spletzer, James R., Harper, Michael, 2010. Labor in the New Economy. University of Chicago Press. https://doi.org/10.7208/chicago/ 9780226001463.001.0001. Adams, Carolyn, 2003. The meds and Eds in Urban economic development. J. Urban Aff. 25 (5), 571–588. Andersson, Fredrik, Holzer, Harry J., Lane, Julia I., 2005. Moving up or Moving on: Who Gets Ahead in the Low-Wage Labor Market? Russell Sage Foundation, New York. Autor, David H., 2015. Why are there still so many jobs? The history and future of workplace automation. J. Econ. Perspect. 29 (3), 3–30. Autor, David H., Dorn, David, 2009. Inequality and Specialization: the Growth of Low-Skill Service Jobs in the United States*. NBER Working Paper Series. Autor, David H., Dorn, David, 2013. The growth of low-skill service jobs and the polarization of the US labor market. Am. Econ. Rev. 103 (5), 1553–1597. https://doi. org/10.1257/aer.103.5.1553. Autor, David H., Katz, Lawrence F., Kearney, Melissa S., 2006. The polarization of the U.S. Labor market. Am. Econ. Rev. 96 (2), 189–194. https://doi.org/10.1257/ 000282806777212620. Baicker, K., Congdon, W., Mullainathan, S., 2012. Health insurance coverage and take‐up: lessons from behavioral economics. Milbank Q. 90 (1), 107–134. Bartik, Timothy J., Erickcek, George, 2007. Higher Education, the Health Care Industry, and Metropolitan Regional Economic Development: what Can" Eds & Meds" Do for the Economic Fortunes of a Metro Area's Residents? Upjohn Institute Working Paper. Baughman, Reagan A., Smith, Kristin E., 2012. “Labor mobility of the direct care workforce: implications for the provision of long‐term care. Health Econ. 21 (12), 1402–1415. Bergmann, Barbara R., 2011. “Sex segregation in the blue-collar occupations: women's choices or unremedied discrimination? Comment on England. Gend. Soc. 25 (1), 88–93. Bodenheimer, T., Chen, E., Bennett, H., 2009. Confronting the growing burden of chronic disease: can the US health care workforce do the job? Health Aff. 28 (1), 64–74. Bolton, Sharon, 2007. Dimensions of Dignity at Work. Routledge. Brady, David, Baker, Regina S., Ryan, Finnigan, 2013. When unionization disappears: state-level unionization and working poverty in the United States. Am. Sociol. Rev. 78 (5), 872–896. https://doi.org/10.1177/0003122413501859. Budig, Michelle J., Hodges, Melissa J., England, Paula, 2019. Wages of nurturant and reproductive care workers: individual and job characteristics, occupational closure, and wage-equalizing institutions. Soc. Probl. 66, 294–319. Bureau of Labor Statistics, 2018. Occupational Outlook Handbook: Nursing Assistants and Orderlies. ” U.S. Department of Labor. https://www.bls.gov/ooh/ healthcare/nursing-assistants.htm. Card, David, Mas, Alexandre, 2015. Introduction: the labor market in the aftermath of the Great recession. J. Labor Econ. 34 (S1), S1–S6. https://doi.org/10.1086/ 682829. Carnevale, Anthony, P., Rose, Stephen J., Cheah, Ban, 2013. The College Payoff: Education, Occupations, Lifetime Earnings. The Georgetown University Center on Education and the Workforce. Charles, Maria, Grusky, David B., 2005. Occupational Ghettos: the Worldwide Segregation of Women and Men, vol. 71 LIT Verlag Münster. Cohen, Philip N., Huffman, Matt L., 2003. “Individuals, jobs, and labor markets: the devaluation of women's work. Am. Sociol. Rev. 443–463. Coombs, Christopher K., Newman, Robert J., Cebula, Richard J., White, Mary L., 2015. The bargaining power of health care unions and union wage premiums for registered nurses. J. Labor Res. 36 (4), 442–461. Cotter, David, Hermsen, Joan, Vanneman, Reeve, 2004. Gender Inequality at Work. Russell Sage Foundation and Population Reference Bureau, New York. Damaske, Sarah, 2011. For the Family?: How Class and Gender Shape Women's Work. Oxford University Press. De Socio, Mark, 2012. “Regime network restructuring in akron, Ohio, 1975–2009: a longitudinal social network analysis. Growth Chang. 43 (1), 27–55. De Socio, Mark, 2007. Business community structures and urban regimes: a comparative analysis. J. Urban Aff. 29 (4), 339–366. Dill, Janette, Chuang, Emmeline, Morgan, Jennifer C., 2014. “Healthcare organization–education partnerships and career ladder programs for health care workers. Soc. Sci. Med. 122, 63–71. Dill, Janette, Price-Glynn, Kim, Carter, Rakovski, 2016. “Does the ‘glass escalator’ compensate for the devaluation of care work occupations? The careers of men in low-and middle-skill health care jobs. Gend. Soc. 30 (2), 334–360. Duffy, M., 2007. Doing the dirty work: gender, race, and reproductive labor in historical perspective. Gend. Soc. 21 (3), 313–336. Duffy, M., 2011. Making Care Count: A Century of Gender, Race, and Paid Care Work. Rutgers University Press. Dwyer, Rachel E., 2013. The care economy? Gender, economic restructuring, and job polarization in the U.S. Labor market. Am. Sociol. Rev. 78 (3), 390–416. https:// doi.org/10.1177/0003122413487197. England, Paula, 2010. The gender revolution: uneven and stalled. Gend. Soc. 24 (2), 149–166. England, Paula, Budig, Michelle, Folbre, Nancy, 2002. Wages of virtue: the relative pay of care work. Soc. Probl. 49 (4), 455–473. Even, William E., Macpherson, David A., 2018. The affordable care act and the growth of involuntary part-time employment. Finnigan, Ryan, 2018. Varying weekly work hours and earnings instability in the Great recession. Soc. Sci. Res. Fitzgerald, Joan, 2006. Moving up in the New Economy: Career Ladders for US Workers. Cornell University Press. Folbre, N., 2012. For Love or Money: Care Provision in the United States. Russell Sage Foundation. Frogner, Bianca K., 2018. The health care job engine: where do they come from and what do they say about our future? Med. Care Res. Rev. 75 (2), 219–231. Gauchat, Gordon, Kelly, Maura, Wallace, Michael, 2012. Occupational gender segregation, globalization, and gender earnings inequality in US metropolitan areas. Gend. Soc. 26 (5), 718–747. Gautié, Jérôme, Schmitt, John, 2010. Low-Wage Work in the Wealthy World. Russell Sage Foundation. Glazer, Nona Y., 1991. Between a rock and a hard place’: women's professional organizations in nursing and class, racial, and ethnic inequalities. Gend. Soc. 5 (3), 351–372. Golden, Lonnie, Gebreselassie, Tesfayi, 2007. Overemployment mismatches: the preference for fewer work hours. Mon. Labor Rev. 130, 18. Goodman, Christopher J., Mance, Steven M., 2011. “Employment loss and the 2007–09 recession: an overview. Mon. Labor Rev. 3–12. https://www.bls.gov/opub/ mlr/2011/article/employment-loss-and-the-2007-09-recession-an-overview.htm. Green, Francis, 2006. Demanding Work: the Paradox of Job Quality in the Affluent Economy. Princeton University Press. Greenhouse, Steven, 2016. How the $15 Minimum Wage Went from Laughable to Viable. The New York Times. Grusky, David B., Bruce, Western, Christopher, Wimer (Eds.), 2011. The Great Recession. Russell Sage Foundation, New York. Handel, Michael J., 2005. Trends in perceived job quality, 1989 to 1998. Work Occup. 32 (1), 66–94. Himmelstein, Kathryn EW., Venkataramani, Atheendar S., 2019. Economic vulnerability among US female health care workers: potential impact of a $15-per-Hour minimum wage. Am. J. Public Health 109 (2), 198–205. Hodson, Randy, 2001. Dignity at Work. Cambridge University Press. Holzer, Harry J., Lerman, Robert I., 2007. American's Forgotten Middle-Skill Jobs: Education and Training Requirements in the Next Decade and beyond. The Workforce Alliance, Washington, D.C. Holzer, Harry J., Lerman, Robert, 2009. The Future of Middle-Skill Jobs. The Brookings Institute, Washington, D.C. Holzer, Harry J., I Lane, Julia, Rosenblum, David B., Andersson, Fredrik, 2011. Where Are All the Good Jobs Going?: what National and Local Job Quality and Dynamics Mean for US Workers. Russell Sage Foundation. Jacobs, Jerry A., 1993. “Men in female-dominated fields: trends and turnover.” men in female-dominated fields: trends and turnover. In: Williams, C.L. (Ed.), Research on Men and Masculinities Series, Vol. 3. Doing "women's Work": Men in Nontraditional Occupations. Sage Publications, Inc, Thousand Oaks, CA, US, pp. 49–63. https://doi.org/10.4135/9781483326559.n4.

24

Social Science Research 84 (2019) 102350

J. Dill and M.J. Hodges

Kalleberg, Arne L., 2007. The Mismatched Worker. WW Norton & Company, New York. Kalleberg, Arne L., 2011. Good Jobs, Bad Jobs: the Rise of Polarized and Precarious Employment Systems in the United States, 1970s–2000s. Russell Sage Foundation, New York. Kalleberg, Arne L., Reskin, Barbara F., Hudson, Ken, 2000. Bad jobs in America: standard and nonstandard employment relations and job quality in the United States. Am. Sociol. Rev. 65 (2), 256–278. https://doi.org/10.2307/2657440. Kilbourne, Barbara Stanek, England, Paula, Farkas, George, Kurt, Beron, Weir, Dorothea, 1994. Returns to skill, compensating differentials, and gender bias: effects of occupational characteristics on the wages of white women and men. Am. J. Sociol. 100 (3), 689–719. Kleiner, Morris M., 2006. Licensing Occupations: Ensuring Quality or Restricting Competition? WE Upjohn Institute. Kleiner, Morris M., Krueger, Alan B., 2010. The prevalence and effects of occupational licensing. Br. J. Ind. Relat 48 (4), 676–687. Lambert, Susan J., 2008. Passing the buck: labor flexibility practices that transfer risk onto hourly workers. Hum. Relat. 61 (9), 1203–1227. Levanon, Asaf, England, Paula, Allison, Paul, 2009. Occupational feminization and pay: assessing causal dynamics using 1950–2000 US census data. Soc. Forces 88 (2), 865–891. Massey, Douglas, Hirst, Deborah, 1998. From escalator to hourglass: changes in the U.S. Occupational wage structure 1949–1989. Soc. Sci. Res. 27 (1), 51–71. https:// doi.org/10.1006/ssre.1997.0612. McGovern, Patrick, Deborah Smeaton, Hill, Stephen, 2004. Bad jobs in britain: nonstandard employment and job quality. Work Occup. 31 (2), 225–249. Meisenheimer, Joseph R., 1998. The services industry in the good versus bad jobs debate. Mon. Labor Rev. 121, 22. Meyerson, Harold, 2014. The Seeds of a New Labor Movement. The American Prospect. Mishel, Lawrence, Bivens, Josh, Gould, Elise, Shierholz, Heidi, 2012. The State of Working America. Cornell University Press. Nelson, Marla, Wolf-Powers, Laura, 2009. Chains and ladders: exploring the opportunities for workforce development and poverty reduction in the hospital sector. Econ. Dev. Q. 24 (1), 33–44. https://doi.org/10.1177/0891242409347721. Osterman, Paul, 2014. Securing Prosperity: the American Labor Market: How it Has Changed and what to Do about it. Princeton University Press. Parrillo, Adam J., de Socio, Mark, 2014. Universities and hospitals as agents of Economic Stability and growth in small cities: a comparative analysis. Ind. Geogr. 11, 1–28. Pilat, Dirk, Cimper, Agnès, Olsen, Karsten Bjerring, Webb, Colin, 2006. The changing nature of manufacturing in OECD economies. http://www.oecd-ilibrary.org/ science-and-technology/the-changing-nature-of-manufacturing-in-oecd-economies_308452426871. Price-Glynn, Kim, Carter, Rakovski, 2012. Who rides the glass escalator? Gender, race and nationality in the national nursing assistant study. Work Employ. Soc. 26 (5), 699–715. Redbird, Beth, 2017. The new closed shop? The economic and structural effects of occupational licensure. Am. Sociol. Rev. 82 (3), 600–624. Ribas, Vanesa, Dill, Janette S., Cohen, Philip N., 2012. Mobility for care workers: job changes and wages for nurse aides. Soc. Sci. Med. 75 (12), 2183–2190. Ruggles, Steven, Genadek, Katie, Goeken, Ronald, Grover, Josiah, Sobek, Matthew, 2017. Integrated Public Use Microdata Series: Version 7.0. University of Minnesota, Minneapolis, MN. https://doi.org/10.18128/D010.V7.0. Ryan, Camille L., Bauman, Kurt, 2016. Educational Attainment in the United States: 2015. U.S. Department of Commerce, Economics and Statistics Administration. https://www.census.gov/content/dam/Census/library/publications/2016/demo/p20-578.pdf. Schindel, Jennifer, O'Neal, Edward, Iammartino, Brian, Solomon, Kim, Cherner, David, Santimauro, Janine, 2006. Workers Who Care: A Graphical Profile of the Frontline Health and Health Care Workforce. Robert Wood Johnson Foundation. https://folio.iupui.edu/bitstream/handle/10244/540/workers_who_care.pdf. Silva, Jennifer M., 2013. Coming up Short: Working-Class Adulthood in an Age of Uncertainty. Oxford University Press. Simpson, Ruth, 2005. Men in non‐traditional occupations: career entry, career orientation and experience of role strain. Gender Work Organ. 12 (4), 363–380. https:// doi.org/10.1111/j.1468-0432.2005.00278.x. Spetz, Joanne, 2016. The nursing profession, diversity, and wages. Health Serv. Res. 51 (2), 505–510. Stainback, Kevin, Tomaskovic-Devey, Donald, Skaggs, Sheryl, 2010. Organizational approaches to inequality: inertia, relative power, and environments. Annu. Rev. Sociol. 36. Sutton, April, Bosky, Amanda, Muller, Chandra, 2016. Manufacturing gender inequality in the new economy: high school training for work in blue-collar communities. Am. Sociol. Rev. 81 (4), 720–748. Thompson, D., 2009. It's Not Just a Recession. It's a Man-Cession. The Atlantic. U.S. Centers for Medicare & Medicaid Services, 2018. National health expenditure fact sheet. https://www.cms.gov/research-statistics-data-and-systems/statisticstrends-and-reports/nationalhealthexpenddata/nhe-fact-sheet.html. Weeden, Kim A., 2002. Why do some occupations pay more than others? Social closure and earnings inequality in the United States. Am. J. Sociol. 108 (1), 55–101. Wright, Erik Olin, Dwyer, Rachel E., 2003. The patterns of job expansions in the USA: a comparison of the 1960s and 1990s. Soc. Econ. Rev. 1 (3), 289–325. https:// doi.org/10.1093/soceco/1.3.289. Yavorsky, J., Cohen, P., Qian, Y., 2012. Man Up, Man Down: Race–Ethnicity and the Hierarchy of Men in Female-Dominated Work. Sociol. Q. 57 (4), 733–758.

25