Authors: Charlotte O’Herron, Daniel Schneider, Kristen Harknett
Categories: Article, Gender, Income or wages, Inequality, Low-income families, Parenthood
Source: Journal of marriage and the family
Doi: 10.1111/jomf.13041
Authors: Charlotte O’Herron, Daniel Schneider, Kristen Harknett
We assess how the distribution of parents across firms contributes to parenthood wage gaps in a low-wage U.S. labor market and examine the role of understudied compensating differentials relevant to precarious work.
In the U.S., parenthood drives a wedge in wages, as mothers often earn less than women without children, whereas fathers typically earn more than men without children. Firms bear influence over setting wages and sorting workers, yet firms are largely omitted from research on parental wage gaps in the U.S.
We draw on novel employer-employee matched data on 74,086 hourly service-sector workers to decompose parental wage gaps into their within- and between-firm components. We leverage uniquely rich data on compensating differentials to test if they sort parents across firms.
We found that mothers are overrepresented in lower-wage firms, accounting for 68% of mothers’ wage gap. In contrast, fathers’ wage gap accrued within firms. We found limited evidence that compensating differentials, even schedule quality, produce parental wage gaps.
We show for the first time that in a major U.S. industry, mothers are segregated in low-paying firms compared to women without children, while fathers are paid more than men without children in the same firms. Our findings largely do not tell a story of parents voluntarily choosing between wages and job quality, instead calling for more research on firm practices.
Parenthood plays an important role in generating gender wage gaps in the United States and in other affluent countries, as mothers typically earn lower wages than women without children, whereas fathers typically earn more than men without children (Lundberg & Rose, 2000; Budig & England, 2001; Lundberg & Rose, 2002; England, 2005; Glauber, 2008; Gangl & Ziefle, 2009; Angelov et al., 2016; Glauber, 2018; Jee et al., 2019). A large body of research has exposed several sources of these parenting wage gaps, yet these explanations pertain largely to individual characteristics, such as human capital or productivity, or to occupational sorting (Cukrowska-Torzewska & Matysiak, 2020). This account omits a critical level of firms. Firms hire workers and, in combination with workers’ job searches and preferences, determine who works in certain organizations. Yet, we know relatively little about how the distribution of parents across firms contributes to parental wage inequalities. Nevertheless, a growing body of research centers between-firm inequality (Song et al., 2019) and suggests that parental wage gaps could be produced by mothers and fathers sorting into relatively lower-paying or higher-paying firms compared to women and men without children.
The U.S. context of limited labor standards, low minimum wages, and decentralized wage setting (Carré & Tilly, 2017) produces substantial heterogeneity in the wages that firms offer, even after accounting for occupation and worker characteristics (Groshen, 1991b, 1991a; Krueger & Summers, 1988; Lane et al., 2007). Against this backdrop, parental wage gaps could be produced if parents concentrate in relatively lower- or higher-paying firms compared to workers without children. One way that this inequality could come about is if working parents in the U.S. sacrifice higher wages in exchange for “family friendly” job amenities, a tradeoff referred to as compensating differentials. Examining firm segregation by parenthood status, then, has the potential to unveil an important part of the process by which parenthood wage gaps are produced.
In Canada and Europe, between-firm segregation accounts for a large portion of the motherhood wage gap and a smaller, more variable portion of the fatherhood wage gap (Beblo et al., 2009; Fuller, 2018; Petersen et al., 2014; Cooke & Fuller, 2018). However, this same evidence is lacking in the U.S., which differs from these international contexts in terms of greater wage variation between firms (Tomaskovic-Devey et al., 2020) and weaker policy supports related to worker benefits (McCrate, 2018), which could make between-firm sorting processes related to compensating differentials particularly important. Existing work on these dynamics in the United States has been limited both by the lack of employer-employee matched data in the U.S. that also measures parenthood, as well as by the lack of fine-grained measures of job amenities that vary between firms and that are valuable to working parents, such as predictable schedules (Carré & Tilly, 2017; Henly & Lambert, 2014; Ton, 2014).
To investigate parental between-firm sorting in the U.S. and the role of compensating differentials, we focus on the service sector, where nearly 20 percent of all U.S. workers are employed (Schneider & Harknett, 2020). This low-wage, U.S. context is a powerful site to conduct this research given that parents in these industries face serious work-family conflict (Williams et al., 2013) yet often lack work benefits, such as health insurance, paid sick leave, paid family leave, or schedule stability, which could relieve these work-family conflicts. Social policies to provide these benefits are meager compared with European-style welfare states, and also compared with Canada (Haley-Lock, 2011). Instead, U.S. employers are left to provide these benefits, but they do so unequally. Especially in low-wage industries, benefits and job quality vary significantly across firms (Carre & Tilly, 2017; Ton, 2014), and worker turnover is high (Lambert 2008, Choper et al., 2019). This landscape of limited public provision of benefits and job quality, voluntary but heterogenous firm provision, and high turnover rates may result in parents in the U.S. service sector being especially willing to work at a firm that offers lower wages in exchange for amenities such as paid vacation days or stable and controllable schedules. Further, given the high level of inequality in wages across firms in this labor market segment, the wage gains and losses associated with working for specific U.S. service sector firms are substantial (Osterman, 2018) and would be materially consequential for low-wage parents.
We draw upon a novel employer-employee linked dataset composed of surveys of 74,086 employees at 165 large food service and retail companies collected by the Shift Project. These data have the unique advantage of providing large-sample linked employer-employee data on wages, parenthood, demographics, and detailed job quality characteristics for a sample of parents and non-parents working in the low-wage service sector in the United States in the contemporary period. This allows us to decompose parental wage gaps into their within- and between-firm components and thoroughly assess the role of compensating differentials. We know of no other data source in the U.S. that meets these criteria. Using these novel data, we advance the study of gender and parenting inequality in the U.S. by centering firms in their role as wage-setters and worker-sorters.
Firms may play an important role in producing wage inequalities that are observed for mothers and fathers compared to women and men without children. Workers’ wages are determined not just by their position in a firm’s hierarchy, but also by how much the firm pays on average compared to more or less similar firms. In the U.S. and in many other countries, firms vary substantially in the average wages that they offer, even after adjusting for occupational and worker characteristics (Groshen, 1991b, 1991a; Krueger & Summers, 1988; Lane et al., 2007). In fact, much of the growth in income inequality in the U.S. can be explained by rising between-firm wage inequality (Barth et al., 2016; Song et al., 2019). Comparable firms may pay their employees differently for a wide variety of reasons, like market power or social goals (Osterman, 2018).
Because firms vary in the average wages that they offer, the unequal sorting of groups of workers across firms can contribute to wage gaps. A growing body of research focuses on between-firm inequality and finds that it contributes to wage gaps by gender (Bayard et al., 2003; Brick et al., 2023, p. 202; Card et al., 2016; Drolet & Mumford, 2012; Goldin et al., 2017; Javdani, 2015, 2020; Kaya, 2019; Li et al., 2020; Penner et al., 2022; Petersen & Morgan, 1995) and by immigration status and race (Arellano-Bover & San, 2020; Eliasson, 2013; Pendakur & Woodcock, 2010). A focus on between-firm inequality advances the established theory of firms as “inequality regimes” by moving beyond how inequalities emerge within organizations. It instead centers how workers’ ability to claim wages for themselves are shaped by their firms’ wage-setting power and decisions in relation to other comparable firms (Tomaskovic-Devey & Avent-Holt, 2019). Our paper applies this perspective to parenthood wage inequalities in the U.S. The next two sections develop arguments for how parents in the U.S. may be advantaged or disadvantaged by their positions across firms.
A small body of work from Canada and Europe has used linked employer-employee data to assess the role of between-firm segregation in producing the motherhood and fatherhood wage gaps. These studies estimate how much of the net parental wage gaps—meaning the gaps that remain after controlling for measured individual and job characteristics—are driven by “between-firm” segregation, or parents disproportionately working at higher- or lower- paying firms compared to observably similar workers without children. Overall, between-firm segregation appears to play a significant role in contributing to the motherhood wage disadvantage but a more modest and class-dependent role in contributing to the fatherhood advantage. Because of a lack of necessary employer-employee linked data, there is a gap in the literature on the extent to which within-firm vs. between-firm processes drive parental wage gaps in the U.S.
For mothers, their segregation in lower-paying firms accounted for 97% of the net motherhood wage gap in Canada (Fuller, 2018) and between 30 and 50% of the net motherhood gap in West Germany (Beblo et al., 2009) and Norway (Petersen et al., 2014). Thus, while mothers are consistently disadvantaged by their concentration in lower-paying firms, the extent to which this pattern contributes to the overall motherhood wage gap varies across studies.
For fathers, sorting into higher-paying firms accounted for 37% of the net fatherhood wage gap in Canada (Cooke & Fuller, 2018)—far less than the share of the motherhood wage gap that was driven by between-firm sorting in the same context (Fuller, 2018). These dynamics varied by class as only less educated and nonprofessional fathers were concentrated in higher-paying firms, which accounted for 47-60% of their net wage advantage over men without children (Cooke & Fuller, 2018), while more educated fathers and fathers with professional jobs earned more than men without children solely within firms. The pattern that more privileged fathers have a wage premium mostly or solely from within-firm gains aligns with findings on white-collar workers in Norway (Petersen et al., 2014).
Based on the literature on Canada and Europe, we hypothesize that firm sorting will contribute more to wage gaps for mothers than fathers in the U.S. service sector (Hypothesis 1).
While between-firm segregation that results in parental wage gaps can derive from multiple sources including queuing processes and information asymmetries, such segregation may be produced by workers responding to compensating wage differentials offered by firms. The neoclassical compensating differentials theory argues that when choosing jobs and firms, workers make trade-offs between wages and other aspects of job quality, such as benefits and work schedules, depending on their preferences (Rosen, 1986; Smith, 2003). Workers may pursue higher wages at the expense of non-wage amenities or better non-wage amenities at the expense of wages.
In line with this theory, mothers and fathers may make trade-offs between wages and “family-friendly” job amenities. Mothers may sacrifice higher earnings for family-friendly amenities (Budig & England, 2001) in the face of persistent gender inequalities in care work (Sayer, 2010; Bianchi et al., 2012; Barroso, 2021) and difficulty accessing affordable childcare (Ruppanner, 2020; Landivar, Ruppanner, et al., 2021; Landivar, Scarborough, et al., 2021). In contrast, to enact the male breadwinner ideal (Percheski & Wildeman, 2008), fathers may sacrifice family-friendly amenities or other dimensions of job quality for higher pay.
The empirical record on whether compensating differentials contribute to parental wage inequality within a low-wage labor market is mixed and incomplete. Most of the prior work on parental compensating differentials has been economy-wide and has not examined between-firm sorting. This work has returned mixed results. For example, part-time work appears to explain a portion of the motherhood wage gap (Waldfogel, 1997; Budig & England, 2001), with at least one exception (Glauber, 2012). Mothers seeking family-friendly supports (Heywood et al., 2007), feminized jobs or self-employment (Budig & Hodges, 2010), and the ability to work less at night and less stressful jobs (Felfe, 2012) also appear to explain part of the motherhood wage gap. However, other studies find that family-friendly work conditions do not explain the motherhood wage gap, using measures like flexible work arrangements, job support or ease, and benefits (Glass & Camarigg, 1992; Boushey, 2008; Glauber, 2012). For fathers, the evidence has been more consistent, finding that fathers pursue higher wages at the cost of work-life balance, such as by selecting jobs that require or reward long hours (Lundberg & Rose, 2002; Knoester & Eggebeen, 2006; Glauber, 2008).
Further, these studies generally use limited measures of potential compensating differentials. Schedule quality has been assessed as a compensating differential for parents in just a few studies (Boushey, 2008; Budig & Hodges, 2010; Fuller & Hirsh, 2019; Felfe, 2012), and some prior studies have accounted for health insurance (Budig & Hodges, 2010; Glauber, 2012), paid leave (Fuller, 2018), paid sick or vacation days (Glauber, 2012), and childcare assistance (Fuller, 2018) but rarely include them all together or account for other benefits such as dental insurance and tuition reimbursement. Prior research has also not distinguished between part-time work that is voluntary versus involuntary. Part-time work has long been assumed to be an “amenity” for parents, especially for mothers, for which they accept lower hourly and total wages. But, around a tenth of all U.S. workers (Golden & Kim, 2020) and a third of service sector workers (Schneider & Harknett, 2019b) work part-time involuntarily. Unless the data distinguish between voluntary and involuntary part-time, it is ambiguous whether part-time work operates as a compensating differential, in the case of voluntary part-time, or as an unwanted double disadvantage, in the case of involuntary part-time.
In addition to these limits on the measurement of compensating differentials, very few studies use employer-employee linked data to examine compensating differentials as a source of parenthood wage gaps.^1^ Having linked data allows for an important test of whether compensating differentials contribute to parenthood wage gaps between and within firms. Notable exceptions draw on data from Canada and show that part-time work (Fuller, 2018) and having schedule control (for less-educated mothers) (Fuller & Hirsh, 2019) appear to explain some of the between-firm motherhood wage gap. For fathers, there is little research on compensating differentials that leverages firm data. However, low-SES fathers in Canada sorted into lower-paying firms when they changed firms compared to similar men without children (Cooke & Fuller, 2018), which is consistent with low-SES fathers seeking better work-family balance at the expense of wages, among other explanations. This intriguing finding demonstrates how much there is yet to learn about how both mothers and fathers are matching with firms in low-wage labor markets.
What little we know of the role of between-firm segregation in generating parental wage gaps and of the role of compensating differentials in producing such segregation is limited to the European and Canadian contexts. We have no similar evidence on between-firm sorting or the influence of compensating differentials on sorting for parents in the United States.
Research on the role of compensating differentials in explaining between-firm and within-firm parenthood wage gaps in the U.S. has likely been limited due to the lack of U.S. datasets that link workers to employers in tandem with data on parenthood, demographics, human capital, occupations, and compensating differentials. One U.S. study, Yu and Hara (2021), shares a similar purpose to ours but uses the National Longitudinal Survey of Youth (NLSY), which contains longitudinal data but does not have firm identifiers or a firm-cluster design to obtain a significant number of workers within various firms. Yu and Hara (2021) find evidence of a motherhood wage penalty that is larger for women who change jobs following parenthood and a fatherhood wage premium that predominantly accrues to men who do not change jobs. But, because the NLSY data are not employer-employee linked, Yu and Hara (2021) cannot and do not provide estimates of the contribution of between-firm segregation to the motherhood penalty or fatherhood premium.
There is a similar lack of evidence in the U.S. on the role of compensating differentials in explaining parenthood wage gaps that accrue between and within firms. Yet, the role of compensating differentials is likely to be even more important in the U.S. context than in Europe or Canada. The U.S. lacks federal provision or regulation of many of the benefits and aspects of job quality that, in their absence, produce acute stress for working parents in the low-wage service sector, including a lack of national health insurance, paid family and medical leave, paid sick leave, paid vacations, and work schedule stability. Instead of federal guarantees, some of these benefits are provided voluntarily by firms, but this provision is unequal and heterogenous. Firms have a great deal of discretion over the benefits and other aspects of job quality that they offer (Ton, 2014; Carre & Tilly, 2017). Several U.S. states have implemented paid family, medical, and sick leave policies, which reach a subset of workers, but not all. In the context of this patchwork system, there is an opportunity for compensating differentials to play a prominent role in parenthood wage inequality.
The role of compensating differentials in contributing to parents sorting between firms may be especially pronounced for workers in low-wage sectors, such as the service sector. Low-wage firms are especially heterogeneous in benefits and job quality (Carré & Tilly, 2017). Many parents in low-wage jobs lack employer-sponsored benefits that are more common in higher-status sectors. For example, access to paid family leave, paid sick leave, and vacation days, which give parents the time they need to care for and bond with their children, as well as to health or dental insurance and childcare assistance, which allow parents to support their children in other ways, is highly varied in the service sector. Thus, low-wage parents are disproportionately affected by the lack of federal intervention in employee benefits.
Parents in the U.S. service sector face additional forms of work-family conflict that are distinct from their higher-status counterparts, especially related to scheduling (Williams et al., 2013). Many service sector workers face rigidity and a lack of control in their schedules, such that missing a shift or leaving early for a childcare crisis could be grounds for dismissal (Williams et al., 2013). They also face schedules that could be changed with little notice, making arranging childcare and fulfilling parenthood responsibilities difficult (Dodson, 2013; Henly & Lambert, 2005, 2014; Schneider & Harknett, 2019a). They also face the risk of involuntary part-time work, meaning that they desire more hours but their employer keeps them at part-time, likely for cost-saving reasons (Lambert, 2008). Therefore, “family-friendly” jobs and firms may be defined differently for parents in the low-wage service sector than they are for higher-status parents. Compensating differentials could then play a crucial, but rarely examined, role in contributing to between-firm parenthood wage gaps in this sector.
The Canadian context, for which we do have evidence on compensating differentials, between-firm sorting, and parenthood gaps (Fuller, 2018; Fuller & Hirsh, 2019; Cooke & Fuller, 2018), is similar in many respects to the U.S., such as both lacking government interventions to provide stable or controllable schedules (Haley-Lock, 2011; McCrate, 2005) or paid sick leave (with some state/provincial exceptions). However, there are also key differences between these country settings, especially in low-wage labor markets, that could lead to different findings.
Canada provides universal access to several non-wage amenities that the U.S. leaves up to states and firms, including universal health care, generous paid family leave and medical leave through the Employment Insurance (EI) program, and paid vacation days of at least two weeks plus at least six paid holidays, with some variation across provinces. Decoupled from specific employers, these benefits are much easier to obtain and maintain, as compared to in the U.S. These universal social policies improve the well-being and economic stability of low-wage workers in Canada compared to the U.S. (Card & Freeman, 1993; Chen, 2015; Haley-Lock, 2011; McCrate et al., 2019; Zuberi, 2018), and their absence in the U.S. could create opportunities for compensating differentials to unfold across firms, more so than they have in Canada. Thus, while the past research on compensating differentials is mixed, we hypothesize that given the variation in wages and job conditions in the U.S. service sector, compensating differentials will explain a portion of the parenthood wage gaps (Hypothesis 2).
Using a novel employee-employer linked dataset, we provide the first estimates of how between-firm sorting contributes to parenthood wage gaps in a low-wage, precarious American industry characterized by substantial variation in wages and job quality across firms (Osterman, 2018; Ton, 2014). This work advances our understanding of parenthood-based wage inequality among lower-status workers by identifying the part played by firms as wage-setters and worker-sorters. Around one-third of service sector workers are residential parents according to data from the 2017-2022 American Community Surveys (which do not capture non-resident parenthood), further emphasizing the importance of understanding how parenthood shapes their wages and workplaces.
We also provide new insight into the role of compensating differentials in between-firm sorting processes that give rise to parenthood wage gaps. As discussed, distinct job amenities are of concern for parents in low-wage sectors, including schedule stability and control, the ability to work full-time or long hours if desired, and certain benefits like health insurance or paid sick leave that are more ubiquitous in higher-status sectors. Yet, these important non-wage workplace amenities have largely gone unmeasured in this literature and could be driving residual parenting wage gaps, especially for the large number of workers in relatively low-wage jobs. We bring to bear uniquely detailed measures of these job quality dimensions to the illustrative case of the U.S. service sector, where workers are likely to particularly value this set of non-wage amenities and yet have little non-employer access to them.
The Shift Project survey has collected data twice annually since 2017 from workers employed at large firms in the retail and food service sectors in the United States. The data are employer-employee matched, meaning that they contain the name of the firm where each respondent works and have large numbers of workers per firm. This creates a large-scale linked employer-employee dataset, extremely rare in the U.S.
In this paper, we draw on surveys collected between September 2017 and May 2023 (we test robustness to excluding the COVID-19 pandemic period). The data are restricted to hourly workers. We use list-wise deletion and arrive at an analytic sample of 74,086 respondents (20,854 men and 53,232 women) that is used for all tables and figures. Of an initial sample of 211,745 respondents, 191,718 had non-missing wage data, 92,243 also had non-missing parenthood status and a gender identity of cisgender man or woman, and 74,086 also had non-missing data for all covariates. See Appendix A1 for more details on the cases excluded for missing data and A8 for a comparison of the analytic sample to the original sample and to the ACS. Respondents are employed at 165 large retail or food service firms, with 322 female respondents and 126 male respondents nested within each of these firms on average. We assess robustness to excluding firms with relatively small cell sizes in the sample.
The Shift Project sample, approved by the Harvard University Institutional Review Board, is constructed by delivering targeted advertisements on Facebook/Instagram to working-aged users in the U.S. who are employed at a given list of firms. The firms used are drawn from the top 50 largest retailers and restaurant chains, according to the National Retail Federation’s list of top 100 retailers and the Restaurant Business’ list of top 100 restaurant chains. The Shift Project rotates firms in and out of the list and also adds a purposive mix of retailers in the 50-100 rank of retailers and restaurants, as well as large firms in hospitality and logistics. The survey recruitment advertisements include employer names and a message, “Working at [company]? Take a Survey and Tell Us About Your Job!” The advertisement also includes an image of a worker in a setting that resembles the targeted workplace. Workers who click on the ad are routed to an online Qualtrics survey, which includes a digital consent form. Survey completers are entered into a lottery for a $500 gift card.
Because the universe of Facebook/Instagram users represents our sampling frame, this could raise concerns about coverage. However, Facebook coverage compares favorably to more traditional sampling frames such as random digit dialing or address-based sampling (Link et al., 2008; Couper, 2017). Further, recent estimates find that 80 percent of working-aged adults are active on Facebook or Instagram, and that usage is fairly uniform across class, race, and other demographic groups, though women are more active than men (Authors’ analysis of Pew Core Trends Survey Data, 2021).
Nevertheless, this is a non-probability sample, and there is risk of selection into the survey on observable and/or unobservable characteristics. To help assess these sources of bias, the Shift Project data have been validated through benchmarking against the Current Population Survey (CPS) and the NLSY (Schneider & Harknett, 2022). This work shows that the Shift Project sample skews female (Schneider & Harknett, 2022), but because we stratify by gender, this is less of a concern. We control for observables as opposed to weighting by them (Winship & Radbill, 1994).
As one of the few large-scale national data sources in the U.S. that link employees to employers, the Shift Project survey dataset is well-positioned to investigate the questions raised in this paper. The Longitudinal Employer-Household Dynamics data provide national employer-employee linked data that could be linked to point-in-time U.S. Census products to obtain additional individual-level variables, but this data source would still have limitations compared to the Shift Project data, such as lacking rich data for assessing compensating differentials.
Our dependent variable is logged hourly wages. It is derived from respondents’ self-reported hourly wage, excluding tips. We use the log functional form to produce estimates of wage gaps as percentages that are comparable with the motherhood and fatherhood wage inequality literatures, even though wages in the service sector are less right-skewed than they typically are in economy-wide data. We assess robustness to accounting for tips in the survey waves in which data on tips are collected but do not include them in the main analysis because hourly wage is standard in the literature (Leonard & Stanley, 2020) and because including tips is likely to introduce bias if mothers are more reliant on tips, while women without children and fathers may be more reliant on other dimensions of take-home pay, like overtime or bonuses, which are not measured in the Shift data.
Our primary independent variable is parenthood. Parenthood is a binary variable that reflects whether the individual reports having any children. It is derived from the question, “Do you have any children? These might be your biological children, stepchildren, adopted children, or foster children.” We assess robustness to defining parenthood as having children who are under 18 years old as well as under 6 years old.
We stratify the analyses by a binary variable for gender that reflects whether the individual self-identifies as male or female. We retain the 98.6% of respondents who identify as Male or Female and delete the 1.04% who identify as non-binary, prefer to self-describe, or prefer not to answer., For language efficiency and comparability within the literature, we refer to women who are parents as “mothers” and men who are parents as “fathers,” though we acknowledge that not all self-identified women and men who are parents also identify as mothers and fathers.
We construct control variables that we introduce in stages to isolate the net motherhood and fatherhood wage gaps. We first adjust for age, month, and year, then for demographics (English as a Second Language (ESL) status, state, and race/ethnicity, county rurality), then for human capital (education, firm tenure, and school enrollment status), then for partner/spouse information (accounts for marital status, cohabitation status, and partner or spouse’s employment status), then for occupational segregation (job title), then for compensating differentials (described below) and, finally, for the respondent’s main employer (firm fixed effects). Table 1 presents descriptive statistics by gender and parenthood for these variables.
Age is a continuous variable coded in years. We also use the squared value of age. Month and year are categorical variables indicating when the survey was completed.
ESL status is a binary variable that reflects whether a language other than English is spoken at home. State is a categorical variable that indicates the state in which the respondent’s workplace is located. Race/ethnicity is a categorical variable that reflects how individuals self-describe their race or ethnicity. They may choose one or more categories from White, Hispanic or Latino/Latina, Black or African-American, Asian or Pacific Islander, American Indian or Alaskan native, or Other. We collapse the responses into white/not Hispanic, Black/not Hispanic, Hispanic, and other or two or more races/not Hispanic. County rurality is a categorical variable based on the respondent’s county of residence that uses the 2010 U.S. Census’ classification of counties into “mostly urban” (< 50%), “mostly rural” (50-99.9%), and “completely rural” (100%), based on the population that lives in rural areas.
Education is a categorical variable that reflects the individual’s highest level of educational attainment from among no degree or diploma, high school diploma/GED, some college, Associate’s degree, Bachelor’s degree, or Master’s degree/Advanced degree. School enrollment status indicates if the respondent is currently enrolled in school. Firm tenure is the number of years worked at their current employer. It is top coded at 6 or more years.
Partner/spouse information is a categorical variable that combines information about the respondent’s marital status, cohabitation status, and partner or spouse’s employment status. It has five Not living with a spouse or partner; Married, living with spouse who is employed; Married, living with spouse who is not employed; Unmarried, living with partner who is employed; and Unmarried, living with partner who is not employed.
Job title reflects the respondent’s job title. The survey provides 15 possible Manager; Cashier or clerk; Salesperson; Customer service; Waiter/waitress/server; Host/hostess; Bartender; Barista; Cook; Baker; Butcher/meat cutter; Produce; Sandwich artist or other food preparation; Delivery person; Stocker/stocking/unloading; and Driver. 11 of these categories appear in the data.
We draw on 19 distinct survey items to construct four variables related to potential compensating schedule instability, schedule control, fringe benefits, and work hours. These capture in detail some of the most central dimensions of non-wage job quality in the low-wage service sector. We use these variables both at the individual-level and, by aggregating up to the firm-level, as time-invariant firm-level characteristics.
Schedule instability refers to service sector workers being unable to predict or rely on a regular working schedule due to their employers’ scheduling practices. We measure six possible exposures to schedule knowing one’s schedule usually less than 1 week in advance, having a variable schedule (one that changes day to day), having to be “on-call” in the last month, having a shift cancelled in the last month, having a scheduled shift’s timing or length changed in the last month, and having worked a “clopening” shift in the last month (working a closing shift and then working the very next opening shift with less than 11 hours off in between shifts). We construct an additive scale of these items.
Workers may also have limited control over when they work. To capture this experience, we create a count of the types of exposures to lacking schedule control that the respondent reports experiences. The three possible exposures include keeping one’s schedule open and available for the employer, starting and finishing times are decided fully by the employer or determined by things outside of the worker’s control and employer’s control, and disagreeing or strongly disagreeing that it’s easy to get time off when one needs it.
Fringe benefits are another key dimension of job quality that varies across firms in the U.S. service sector. We create a count of eight possible benefits that respondents indicate receiving as part of their job. These include a health plan or medical insurance, dental benefits, paid maternity or paternity leave, paid sick days, paid vacation days, paid retirement plan other than Social Security, tuition reimbursement, and childcare provision or subsidy.
Workers may seek part-time work or long hours, or they may have difficulty finding jobs that give them enough work hours. To assess this dimension of job quality, we create a categorical variable that reflects the respondents’ usual number of weekly hours worked and whether respondents are part-time involuntarily or voluntarily. From these two questions, we create a four-category Full-time (working 35 to 45 hours), Long hours (working 45 hours or more), Voluntary part-time (working < 35 hours and disagreeing or strongly disagreeing that they would like to work more hours at the employer), and involuntary part-time ( working < 35 hours and agreeing or strongly agreeing that they would like to work more hours at the employer).
Finally, we include fixed-effects for firm.
We stratify all analyses by gender, such that mothers are compared to women without children and fathers are compared to men without children. This approach allows for comparability with the literature on motherhood and fatherhood wage gaps. While stratifying by gender estimates the role of parental status, it obscures how gender and parental status together create large inequalities, especially between fathers and mothers. To provide benchmark indicators of how average wages are shaped by both gender and parenthood, we provide predicted hourly wages from a pooled model for the four gender-parenthood groups (fathers, men without children, mothers, and women without children), adjusted only for respondent age and the month and year of the survey response.
To provide benchmark indicators of how job quality indicators differ by parental status—the descriptive baseline for a compensating differentials argument—we provide predicted job quality indicators by parental status separately for men and women, also adjusted only for age, month, and year.
We estimate a series of ordinary least-squares linear regression models that predict log-hourly wages with parent as the main predictor variable. The final regression model ln(Yi)=β0+β1Parenti+β2Agei+μ+ω+β3Di+β4HCi+β5Pi+β6Occi+β7CDi+γf
Where our dependent variable is logged hourly wages, Y, for individual i, and β1 summarizes the motherhood wage disadvantage (for models estimated on our sample of women) or the fatherhood wage advantage (for models estimated on our sample of men). In the baseline model, we include only adjustments for age and age squared (Age) and fixed effects for survey month (μ), and survey year (ω). Across models 2-5, we add collections of covariates that capture “traditional” sources of parental wage gaps. In Model 2, we add key demographic and geographic measures (D), including state fixed effects, race/ethnicity, ESL status, and county rurality. In Model 3, we add human capital indicators (HC) of education, firm tenure, and school enrollment status. In Model 4, we add a measure that accounts for marital status, cohabitation status, and partner or spouse’s employment status (P). In Model 5, we include occupational fixed effects to account for occupational segregation (Occ). The parenthood wage gap literature commonly reports estimates from models similar to Model 5.
In Model 6, we further add the measures of compensating differentials (CD) at the individual-level: a categorical measure of typical weekly hours, schedule instability, lack of schedule control, and access to fringe benefits.
In Model 7, we add the same measures of compensating differentials as in Model 6 but aggregated to the firm-level. These measures are the percent of workers who work part-time voluntarily, the percent of workers who work long hours, the average number of employer benefits received, the average schedule instability scale, and the average schedule control scale for workers at each firm. Model 7 does not include compensating differentials at the individual level.
Finally, in Model 8, we introduce firm fixed effects by adding the variable that indexes respondents’ employers (γf). This model builds on Model 6 and so accounts for compensating differentials at the individual level but not at the firm level since these would be colinear.
We test whether accounting for compensating differentials at the individual level mediates the relationship between parenthood status and wages by using the suest procedure in Stata to compare the coefficient on parental status across Models 5 and 6 (Mize et al., 2019). Further, we test if accounting for firm-level compensating differentials (Model 7) attenuates the parental wage gaps, which would suggest that it is a source of between-firm sorting.
Model 6 and Model 8 together allow us to parse out the within- and between-firm contributions to the parenting wage gaps. To do so, we follow the steps that were first utilized by Javdani (2015) and Petersen et al. (2011, 2014), with a formal proof provided by Pendakur & Woodcock (2010).
The Model 8 parenthood coefficients reflect the “within-firm parenting wage gaps.” These reflect any wage differences between parents and observably similar non-parents within the same firms, accounting for all individual-level covaries including compensating differentials. To estimate the “between-firm parenting wage gaps,” we take the difference between β1 in Model 6 and Model 8 (β1_m6 – β1_m7). This wage gap is produced by any segregation of parents into relatively high- or low-wage firms compared to observably similar non-parents, accounting for all individual-level covaries including compensating differentials. We test the statistical significance of this component by using the suest procedure in Stata to compare the coefficient on parental status across these two models.
Finally, we divide the “within-firm” and “between-firm” wage gaps by the net parenting wage gaps from Model 6 to calculate the percent of the net parenting wage gaps that comes from each component. While our aims are similar to a decomposition analysis, our purpose is not to identify differences in returns to the same characteristics between parents and non-parents, as in decomposition.
While much of our analytic focus was on within-gender comparisons by parental status, we predicted wages for the four gender-parent groups from a single model, adjusted only for respondent age, month, and year, in order to highlight that within-gender parental disparities exist within a broader ecology of gender inequality. Figure 1 shows that fathers earned the highest wages, at 17.50 per hour. For women, this ordering was reversed, with childless women earning an average hourly wage of 15.14. The scope of mothers’ disadvantage in the U.S. service sector is evident. Mothers earned $3.60 less per hour than fathers, a 19% wage gap.
If compensating differentials are contributing to the within-gender parenthood wage gaps, we would expect that mothers would have better job quality and more voluntary part-time work compared to women without children, and fathers to have worse job quality and longer hours compared to men without children. Figure 2 shows predicted job quality indicators by parental status, stratified by gender, adjusted only for respondent age, month, and year. The theoretical prediction did not bear out in most cases, except that fathers did appear to work long hours at higher rates, and part-time voluntarily at lower rates, than men without children, in line with the expectation that they pursue higher wages at the expense of seeking longer hours. While mothers had slightly better schedule control, they appeared to have fewer fringe benefits and work long hours more often than women without children. In short, there was limited descriptive evidence of compensating differentials. These results are also included in Appendix Table A2.
Table 2 presents the coefficient associated with motherhood status. The coefficients for the control variables are presented in Appendix Tables A3/A4. In Model 1 Table 2, we show the motherhood wage gap adjusted only for respondent age and fixed effects for year and month. Mothers earned 8.1% less than women without children. In Model 2, we additionally accounted for demographic characteristics and found that the motherhood wage gap is attenuated to 6.3%. In Model 3, we further adjusted for human capital, as measured by educational attainment, firm tenure, and school enrollment, further reducing the wage gap to 4.3%. In Model 4, we adjusted for a five-category measure of marital status, cohabitation, and partner’s employment. Mothers were significantly more likely to be partnered, which was associated with higher wages, and we saw an increase in the estimated motherhood wage gap to 5.6%. In Model 5, we adjusted for occupation to account for occupational segregation and found a reduction in the motherhood wage gap to 3.2%. We take this estimate as the motherhood wage disadvantage net of standard adjustments for demographics, human capital, partnership, and occupational segregation.
Model 6 added four sets of measures that could potentially operate as compensating differentials. The motherhood wage gap was attenuated to 1.9%. Some of these variables were correlated with motherhood and wages in ways that attenuated the coefficient on motherhood, but few in the directions aligned with compensating differentials theory. Long hours, voluntary part-time, and schedule control traded off with wages (A3), but mothers appeared to work longer hours at higher rates, tracking against compensating differentials theory, and had only marginally more control over their schedules and no difference in part-time work rates (Fig. 3; A2). The variable that attenuated the motherhood penalty the most was fringe benefits, but mothers had fewer fringe benefits (Fig. 3; A2), which lowered their wage as benefits were associated with higher wages (A3). Nevertheless, a motherhood wage gap persisted even after accounting for these dynamics along with a wide array of other controls.
In Model 7, we added this same set of compensating differentials but aggregated to the level of the firm. We observed that this attenuated the motherhood wage gap relatively similarly as the individual-level job quality measures did (to 1.7%). At the firm level, only percentage of part-time workers traded off with average wages in the way that compensating differentials theory predicts (A3), but mothers were actually less likely to work at firms with larger shares of part-time workers (results not shown), again suggesting that compensating differentials played a limited role.
Finally, in Model 8, we introduced firm fixed effects. After accounting for between-firm segregation in this model, the motherhood gap was reduced to .6%. The difference between this within-firm gap from Model 8 and the net motherhood gap from Model 6 yields the between-firm gap, 1.3 percentage points, which itself was statistically significant (p < .001) and substantively meaningful, making up 68% of the net gap. Therefore, mothers appeared to be segregated in low-paying firms in an impactful way. In relation to the gross gap, which some papers also report (e.g., Jee et al., 2019), between-firm sorting net of individual and job characteristics made up 16% of the motherhood wage gap.
In Table 3, we present a parallel set of analyses for men. In Model 1, adjusting for age and month and year fixed-effects, we found a fatherhood wage advantage of 6.0%. In Model 2, after adjusting for demographic characteristics, we estimated a slightly larger fatherhood advantage of 7.1%, given that fathers lived in more rural areas, with lower wages. In Model 3, we further adjusted for human capital, leaving the wage gap largely unchanged. In Model 4, we adjusted for marital status and partner’s employment. Fathers were significantly more likely to be partnered, and the change in the wage gap was consistent with specialization, with the fatherhood wage advantage declining to 3.3%. In Model 5, we accounted for occupational segregation, reducing the gap to 2.3%
In Model 6, we introduced our individual-level measures of compensating differentials. The fatherhood wage advantage was slightly attenuated to 2.0%, but this change was not statistically significant. Therefore, we found little evidence of compensating differentials explaining fathers’ wage advantage.
In Model 7, we added this same set of compensating differentials but used firm averages of job quality. We see that this modestly attenuated the fatherhood wage advantage to 1.6%, driven by fathers being somewhat more likely to work at firms with fewer part-time and more long-hours workers (results not shown). Nevertheless, in Model 8, which introduced firm fixed effects, we found no evidence that between-firm segregation played a role in the fatherhood wage gap. The 2.2% gap in Model 8 represents how much fathers were paid more than men at their same firms on average (the within-firm gap), whilst virtually 0% was accounted for by firm sorting (the difference between model 6 and model 8). Thus, the entire net fatherhood wage gap occurred within firms, even though fathers were somewhat more likely to work at firms that allow or require longer hours.
In Figure 3, we plot the within- and between-firm components of the motherhood and fatherhood wage gaps. We formally provide these estimates in Appendix Tables A5/A6. There was a clear gender difference. In the low-wage, U.S. context, mothers were sorted into low-wage firms, and this accounted for the majority (68%) of the net motherhood wage gap. In contrast, fathers’ net wage gap occurred exclusively within firms. Men with children were not sorted into higher-paying firms but instead receive higher wages within firms compared to their observably similar co-workers without children. Between-firm sorting also explained the majority of the economy-wide motherhood wage gap in Canada (Fuller, 2018), but between-firm sorting contributed more to the fatherhood wage advantage for low-SES men in Canada than we observed in the U.S. (Cooke & Fuller, 2018).
We tested robustness to several different model specifications and present the results in Appendix Table A7. There may be variation in parenthood wage gaps over the stage of parenthood, so we tested robustness to defining parenthood as having children under 18 years old, and then as having children under 6 years old. In line with longitudinal research showing the potential for mothers’ wages to “bounce back” (Kahn et al., 2014), we found that mothers’ wage disadvantage was larger with younger children. Fathers’ wage advantage, however, was smaller when children are younger, aligned with findings that the fatherhood wage gap increases over the course of fatherhood (Gowen, 2023).
As another check, we dropped the height of the COVID-19 pandemic period (defined as March 2020 to August 2021) to ensure that our results were not driven by the pandemic’s immediate effects on work, schools, and families. We found that the overall parenthood wage gaps net of all controls were somewhat attenuated, but the prominent role of firm sorting in determining the motherhood wage gap still held, as did the prominent role of within-firm gains in determining the fatherhood wage gap.
Two other specifications did not change our excluding firms with small cell-sizes (defined as having fewer than 50 women or 50 men at the firm) and adding salaried workers to the sample. A very small portion of retail, food service and hospitality workers were salaried. We had data on 4,239 salaried workers to add to our sample to test if our results were robust to including them. We estimated an hourly wage for the salaried workers using salary divided by their reported usual weekly hours times 52 weeks. Including salaried workers did not change our results.
Finally, we assessed if our results were robust to accounting for tipping. We reconstructed our hourly wage variable to include tips by dividing workers’ weekly tip amount by the number of hours they usually worked per week and adding that amount to their hourly wage. While the fatherhood wage gap was unchanged, the net motherhood wage gap was meaningfully attenuated, which suggests that one contributor to mothers’ wage disadvantage is that they disproportionately rely on tips for their earnings as compared to women without children.
Parenthood plays an important role in contributing to gender economic inequality, with mothers experiencing a wage disadvantage and fathers often receiving a wage advantage. Yet, open questions remain about the social processes behind parental wage gaps. Prior research in the U.S. has focused on individual and occupational explanations and has yet to examine if or how parents accrue wage differentials by sorting into lower- or higher-paying firms. Foundational work on organizations (firms) as inequality regimes (Acker, 1992)) and recent work on how organizations produce categorical inequalities (Tomaskovic-Devey & Avent-Holt, 2019) motivate a focus on the role of firms as wage-setters and potential drivers of parental wage gaps not just within firms but also across them. Yet, the sorting of workers across heterogeneous firms is underexamined when it comes to U.S. parental wage gaps despite careful attention to it in other contexts (e.g., Fuller, 2018) and with regards to other categorical inequalities (e.g., Arellano-Bover & San, 2020).
In this paper, we home in on the service sector as a strategic site to examine the role of between-firm sorting in contributing to U.S. parental wage gaps. Emblematic of low-wage, precarious work, the service sector employs nearly 1 in 5 U.S. workers, and job requirements for hourly workers are similar, but wage levels and job quality vary widely. We take advantage of unique, employer-employee matched data in the U.S. service sector collected by The Shift Project from 2017 to 2019 to make two we provide the first examination of the role of between-firm sorting in producing parental wage gaps in the low-wage, U.S. context. We also advance a contested literature on whether trade-offs between wages and non-wage job quality, or compensating differentials, are a source of parental wage inequality for low-paid workers who face significant work-family challenges and scarce opportunities for better job quality.
In the U.S. service sector, we found a motherhood wage disadvantage and a fatherhood wage advantage, which persisted after accounting for individual and job attributes including demographics, human capital, partnership, and occupation. Net of these factors, hourly wages for mothers were around 3.2% lower compared to women without children, whereas wages were around 2.3% higher for fathers compared to men without children. Although these differences are not large in magnitude, they demonstrate that parental wage inequalities are entrenched in the low-wage labor market and highlight a persistent gendered double parenthood confers a wage advantage for fathers and wage disadvantage for mothers. Further, our research revealed a gender contrast in whether the parental wage gaps unfold within or across firms. Consistent with our hypothesis, we found that the net motherhood wage penalty was largely driven by between-firm inequalities rather than within-firm wage differences, whereas the net fatherhood wage premium was mostly persistent within firms. This suggests that firms operate as inequality regimes in different ways for mothers and fathers in low-wage, precarious mothers are disadvantaged by a sorting process into less-coveted firms that pay less, whereas fathers are rewarded more than otherwise similar men without children in the same firms.
In comparing our findings to a small body of similar, international work, we identified a stronger role of within-firm gains for low-SES fathers than research in Canada did (Cooke & Fuller, 2018). Our finding that most of the net motherhood wage penalty was between-firms is similar to findings from Canada (Fuller, 2018). These Canadian estimates are economy-wide, so our results show that even within industry and to some extent class, mothers sorting into relatively low-paying firms is a major source of wage disadvantage.
The theory of compensating differentials could potentially explain why mothers sort into lower-paying firms compared to their childless counterparts; namely, trading off hourly wages for family-friendly job amenities. We bring new evidence to bear on conflicting accounts of compensating differentials (e.g., Glauber, 2012; Heywood et al., 2007), using our uniquely rich data on understudied dimensions of job quality that are highly relevant to parents working in a low-wage context, and that we were able to measure at the individual and firm benefits, schedule stability, schedule control, and more weekly hours. Contrary to our hypothesis, we did not find evidence that compensating differentials contribute to parenthood wage gaps in the service sector. We generally found that mothers did not have more family-friendly job quality. Instead, the apparent contribution of these potential compensating differentials to the motherhood wage gap was largely due to associations that track against compensating differentials theory, like mothers having fewer benefits. While fathers were somewhat more likely to work long hours, this did not significantly contribute to the wage advantage in this sector. At the firm-level, we found no evidence that the motherhood wage gap was attributable to low-wage companies offering more family-friendly benefits, schedules, or hours. For fathers, we saw some hint that sorting into firms that offer or require more work hours may modestly contribute to their wage advantage, but the predominant finding was that fathers enjoy a persistent wage advantage relative to men without children employed at the very same firms. Overall, even in a context where firms are highly heterogeneous across wage- and non-wage amenities, compensating differentials did not emerge as the main driver of parental wage gaps within or across firms.
Therefore, our findings suggest that other processes drive the relationship between firm sorting and mothers’ lower hourly wages, such as hiring discrimination (Correll et al., 2007; Ishizuka, 2021; Weisshaar, 2018). Future audit studies, ideally focused on low-wage industries, should test if hiring discrimination toward mothers influences the type of company where mothers end up working, as a test of whether a “queue” of workers is formed in which mothers face barriers to accessing higher-paying firms and concentrate in lower-paying firms that offer no better job quality (Levanon et al., 2009; Reskin & Roos, 2009).Fathers benefit from within-firm pay-setting processes such as internal raise or promotion practices, though the reasons behind this remain underexplored. No audit study has found statistically significant discrimination in favor of fathers, but it is possible that fathers are favored within organizations, and distinguishing this process from selection represents an important subject for additional research.
The matched employer-employee data from the Shift Project survey have the advantage of allowing us to parse parenthood wage gaps into within- and between-firm components and test the role of compensating differentials in explaining parental sorting, but the data also have important limitations. First, our analyses are cross-sectional, and unobserved differences between parents and non-parents, like productivity differences, could bias the estimates of the net parental wage gaps. Studies that use longitudinal methods have yielded little evidence that unobserved differences are a driver of the motherhood penalty (see review Leonard & Stanley, 2020), but there is more substantial evidence behind selection as a pathway to the fatherhood premium (Icardi et al., 2022; Petersen et al., 2011). Future work should expand upon our findings by generating and using data that are both longitudinal and employer-employee linked.
Second, although we include firm tenure in our analysis, our data lack a measure of total number of job changes or accumulated work experience. While omitting number of job changes has more ambiguous implications, omitting work experience may result in overestimating the net between-firm motherhood wage gap (Gangl & Ziefle, 2009; Leonard & Stanley, 2020). However, studies that account for work experience or job breaks still find meaningful residual wage gaps (Gangl & Ziefle, 2009; Yu & Hara, 2021), and Budig and Hodges (2010) find that reduced experience accounts for almost none of the motherhood penalty for relatively low-paid women, a relevant group for our sample. If we take part-time work to signal less full-time work accumulated in the past, it does not alone meaningfully account for mothers’ concentration in lower-paying firms, which roughly suggests that reduced work experience is not a major factor in this context. Nevertheless, part of the residual motherhood wage gap may still have to do with interruptions in work tenure, and a limitation of our data is the lack of detailed measurement of prior work experience.
Third, our measure of parenthood does not distinguish among ages or life course stages of children. Prior research suggests that the motherhood wage disadvantage is largest when children are young and dissipates over time, whereas the fatherhood advantage grows and cumulates over time (Gowen, 2023; Kahn et al., 2014). When we do focus on parents with children younger than 6, we find evidence that is consistent with this pattern. Therefore, our overall estimates may understate the wage disadvantage that mothers with young children experience and overstate the wage advantage expected for fathers with young children. More generally, our use of cross-sectional data averages together parental wage gaps across a child and parent’s life course and does not allow an accounting of how parental wage gaps evolve over time. A fruitful direction for future research, but one with stringent data requirements, would be to examine the role of firms in parental wage inequalities longitudinally.
Fourth, our data represent a non-probability sample. The non-probability sample would be a concern if there were differential selection into the survey by the conjunction of parental status and hourly wages. Separate work has found that the Shift Project survey data yield comparable findings from the CPS and NLSY probability samples on wages, work scheduling characteristics, and wage returns to tenure (Schneider & Harknett, 2022) and thus lessens concerns about sample selection bias. We leverage the unique advantages of the Shift data advance the literature, but this work should be complemented by comparable probability-sample data should it become available.
Fifth and finally, the strategic focus on the service sector yields information on a sizeable share of the U.S. labor market, but these findings may not generalize to other labor market sectors, given prior evidence of classed sorting dynamics. Our findings likely apply to other low-wage industries with precarious working conditions, such as hospitality, transportation and logistics, and parts of the health care industry, but they likely do not apply to high-wage industries with better and more stable job conditions, which motivates future research in this area.
Our findings, and the data limitations noted above, provide some potential directions for future research. We found that firms play an important role in motherhood wage gaps and that compensating differentials were by and large not the explanation for between-firm wage differences. Therefore, questions remain about what attributes of firms lead to lower wage setting, and future research could incorporate measures of corporate ownership or management structure, market power and monopsony, and other potentially explanatory firm-level measures. In an ideal world, future research could also harness longitudinal, employer-employee linked data to address the role of productivity in between-firm motherhood wage gaps, though data limitations currently preclude this possibility. Another direction for future research, again contingent on addressing gaps in available data, is to capture more granular information on tips, overtime, and bonus pay to allow for a full accounting of guaranteed and discretionary compensation.
Parental wage inequalities in the U.S. exist against a backdrop of minimal labor protections and low minimum wages at the Federal level. In the absence of federal legislations to raise the floor on working conditions, a number of states and localities have acted to raise wages for workers locally. Raising the minimum wage has the potential to benefit mothers the most, given that their wages lag behind their counterparts. Examining whether the empirical evidence bears out this possibility represents yet another good direction for future research.
This paper advances a large body of research on parental wage gaps in the U.S. by bringing in a neglected level of the firm. It paints a picture of the different ways in which motherhood and fatherhood wage inequalities unfold across firms in an industry emblematic of low-wage and precarious work in the U.S. Our findings largely do not tell a story of parents voluntarily choosing between wages and job quality, instead calling for more research on firm practices, as mothers appear to involuntarily sort into lower-paying firms while fathers likely benefit from within-firm processes such as internal rewards. Our findings highlight how gender and parenthood interact to shape how workers are situated across heterogeneous firms in the contemporary period. Our work equips future research on U.S. parental wage gaps to consider within- versus between-firm social processes behind these inequalities.