Abstract
Economic homogamy has important implications for gender inequality and for economic inequalities between households. However, the long-term association between spouses’ earnings is not well understood. This study reconceptualizes economic homogamy as a life course process rather than a static state of being that can be adequately captured at a single point in time. Using data from the National Longitudinal Survey of Youth 1979, I examine the association between spouses’ earnings trajectories over the course of 30 years of marriage to identify three distinct gender egalitarian earnings patterns among couples. 50% of couples follow a Dual earner pattern, in which spouses follow similar, stable earnings patterns over time, 6% of couples are Jointly mobile in that spouses’ earnings vary similarly and 5% follow an Alternating earner pattern. A large minority of couples follow patterns of long-term specialization, with 34% of couples following Male breadwinner patterns and 5% following Female breadwinner patterns. Multivariate analysis reveals that gender egalitarian earnings patterns are stratified by couples’ socio-economic status at marriage: while advantaged couples follow Dual earner patterns comprised of two stable earners, disadvantaged couples follow egalitarian earnings patterns characterized by joint earnings instability. By taking a long-term approach, this study provides insight into the varied ways gender equality in earnings manifests among married couples and reveals an important and understudied dimension of economic homogamy: the concentration of economic stability and instability within couples.
Keywords: Economic homogamy, Gender, Inequality, Life course
1. Introduction
Research has documented growing similarity in husbands’ and wives’ earnings over the past 70 years (Cancian and Reed 1999; Gonalons-Pons and Schwartz 2017; Schwartz 2010; Shen 2021), driven in part by increases in women’s educational attainment and labor force participation, as well as delays in marriage formation and childbearing (e.g., Lichter and Qian 2019). The degree of similarity (or difference) between husbands’ and wives’ earnings has important implications for both gender equality and for broader economic inequalities between households (Bloome et al. 2019; Möhring and Weiland 2021; Nutz and Gritti 2021). Although similarity in spouses’ earnings suggests greater gender equality in earnings within different-sex couples, it also signifies greater economic inequality between households as economic advantage and disadvantage become concentrated within couples.
To date, the study of economic homogamy has largely examined the cross-sectional association between husbands’ and wives’ earnings using correlational measures or loglinear models. However, husbands’ and wives’ earnings within marriage tend to develop and fluctuate in gendered ways over time in response to major family transitions, such as the transition to parenthood (Killewald and Garcia-Manglano 2016; Musick et al. 2020), and in response to changes in a partners’ employment circumstances and characteristics (Cha 2010; Cheng 2015). These patterns indicate that the cross-sectional association between spouses’ earnings is sensitive to when in the marital life course husbands’ and wives’ earnings are measured and may not sufficiently represent the association between spouses’ earnings in the longer term.
More broadly, cross-sectional measures of earnings homogamy provide only a narrow view of the association between spouses’ earnings that obscures its implications for gender and economic inequality. Snapshot measures capture how similar spouses’ earnings are at a particular point in time. However, gender equality in earnings may take a variety of forms over the course of long-term relationships that may not be captured using cross-sectional measures. In some couples, equality in earnings may mean that spouses maintain similar earnings throughout the relationship. In others, husbands and wives may alternate primary earnership, either deliberately or in response to events such as job loss or health declines (Becker and Moen 1999; Langner 2015). Furthermore, among couples following similar earnings patterns over time, whether spouses maintain similar, consistent earnings or whether their earnings are similarly unstable is consequential for economic inequalities between households. However, this dimension of economic homogamy cannot be observed using cross-sectional approaches. Understanding the association between spouses’ earnings within different-sex couples and its implications for broader social inequalities, therefore, requires a longitudinal, linked-lives perspective (Elder 1985; Moen and Hernandez 2009) that considers spouses’ entire marital earnings trajectories and their connection.
In this study, I use a long-term, group-based approach to examine the simultaneous development of husbands’ and wives’ earnings over the course of first marriage. Drawing on data from a cohort of late Baby Boomer men and women in the National Longitudinal Survey of Youth 1979, I use Dual Group-Based Trajectory Modeling (GBTM) to identify key earnings patterns among husbands and wives over up to 30 years of marriage and analyze how spouses’ long-term earnings patterns relate to one another within couples. The results reveal novel insight into the varied ways gender earnings equality (or inequality) manifests within couples over time. I identify six distinct long-term earnings patterns for wives and four for husbands. Based on the combinations of these patterns within couples, I identify five key long-term earnings constellations among couples. I find that 50% of couples followed a Dual earner pattern where both spouses maintained stable earnings over time. However, I also identify two alternative forms of long-term gender egalitarian earnings patterns that are not observable in cross-sectional approaches: 6% of couples followed Jointly mobile patterns in which spouses’ earnings increased, decreased or fluctuated similarly over time and 5% of couples followed Alternating earner patterns in which husbands and wives alternated which spouse was the primary earner. One-third of couples (34%) followed long-term Male breadwinner patterns while long-term Female breadwinner patterns were less common, accounting for 5% of couples.
I then conducted multivariate analysis to understand which couples are most likely to follow these distinct long-term earnings configurations, focusing on two key determinants of gender equality in the division of household labor identified in previous scholarship: wives’ absolute socio-economic status and wives’ relative status compared to their husbands’ (Becker 1981; Qian 2017; 2018; Usdansky 2011). I find that wives’ relative socio-economic status at the onset of marriage (in terms of earnings, education and employment status) is positively associated with their subsequent economic standing within the relationship, indicating that patterns of assortative mating have lasting consequences for wives’ long-term economic power within couples. However, the results also reveal that long-term gender egalitarian earnings patterns are stratified by couples’ socio-economic status. I find that couples who were most socio-economically advantaged at the start of marriage were most likely to follow a Dual earner pattern characterized by joint earnings stability over the marital life course. Couples with lower socio-economic status at marriage, by contrast, were more likely to follow Jointly mobile and Alternating earner patterns characterized by joint earnings instability.
Together these findings challenge previous scholarship that has debated whether gender equality in different-sex couples is reserved for high-socio-economic status couples (e.g., Esping-Andersen and Billari 2015; Goldscheider et al. 2015; Usdansky 2011), demonstrating instead that gender egalitarian earnings patterns look qualitatively different across the socio-economic spectrum. They also reveal an important dimension of economic homogamy that has been overlooked in cross-sectional approaches: the concentration of earnings stability and instability within couples.
2. Economic Homogamy as a Life Course Process
2.1. The Limitations of Cross-Sectional Estimates of Economic Homogamy
A substantial body of research has shown that husbands’ and wives’ earnings have grown more similar over recent decades in the U.S. and in many wealthy countries, (e.g., Esping-Andersen 2007). Much of the existing research documenting this increasing economic homogamy has relied on cross-sectional data to measure changes in the association between husbands’ and wives’ earnings over time, typically with the aim of estimating whether increasing marital homogamy has contributed to growing inequality between households (Blackburn and Bloom 1995; Burtless 1999; Cancian and Reed 1999; Schwartz 2010). This increasing association between spouses’ earnings is often interpreted as signifying a cultural shift in the dominant marriage model from one based on trading and efficient specialization (Becker 1981) to one based on greater equality and collaboration between partners (Oppenheimer 1988; Sweeney 2002). However, cross-sectional measures of economic homogamy offer limited insights into the long-term association between spouses’ earnings and its consequences for evolving economic inequalities within and between couples.
Recent scholarship leveraging longitudinal data demonstrates that the association between spouses’ earnings is dynamic rather than static (Choi and Denice 2023; Gonalons-Pons and Schwartz 2017; Qian 2017; 2018). Examining changes in the correlation coefficient between spouses’ earnings over 30 years of marriage, Gonalons-Pons and Schwartz (2017) showed that the correlation between spouses’ earnings follows a U-shaped pattern over time, with the strongest correlation observed in early marriage and the weakest mid-marriage, to varying degrees across cohorts. These findings suggest that, although cross-sectional measures of economic homogamy provide insight into the association between spouses’ earnings at a particular point in time, they may be subject to a life-cycle bias arising from the gendered development of husbands’ and wives’ earnings over the marital life course.
Spouses’ earnings in the U.S. are most similar at the start of marriage (Gonalons-Pons and Schwartz 2017). As marriage proceeds, however, husbands and wives respond to work and family events in distinctly gendered ways. For example, although the transition to parenthood is associated with sharp declines in women’s labor supply and earnings, it is associated with little change in men’s (Lundberg and Rose 2000; Musick et al. 2020; Sanchez and Thomson 1997). Similarly, wives’ employment behavior – but not husbands – is sensitive to changes in their partners’ employment circumstances (Cha 2010; Cheng 2015). Although wives’ employment and labor supply decline with husbands’ high wages and long working hours (Cha 2010; Shafer 2011) and increase following declines in husbands’ employment outcomes (Cheng 2015), husbands’ labor supply changes little in response to changes in wives’ wages, earnings or work hours (Cha 2010; Cheng 2015). These gendered patterns in the development of earnings over the marital life course indicate that the association between husbands’ and wives’ earnings measured at a particular time may not be representative of the association between spouses’ earnings in the longer term.
From a conceptual standpoint, snapshot measures of earnings homogamy have inherent limitations in illuminating the consequences of economic homogamy for inequality within households in the longer term. In measuring the degree of similarity between spouses’ earnings at a particular point in time, cross-sectional measures of earnings homogamy capture only a narrow view of gender equality in earnings within couples. However, as illustrated in Figure 1, gender equality in earnings may take a variety of forms over the course of long-term relationships. Husbands and wives may maintain similar levels of earnings throughout the relationship (as shown in panels A and C), or they may alternate over time which spouse is the primary earner, either by deliberately alternating the prioritization of one spouse’s earnings and career or in response to major events such as a spouses’ job loss or adverse health event (panel B) (Becker and Moen 1999; Langner 2015).
Figure 1.
Gender Egalitarian Earnings Patterns Over the Marital Life Course
Notes: Lines represent the earnings trajectories of two spouses over the course of marriage.
Furthermore, cross-sectional approaches often provide limited information regarding longer-term economic inequality between households. Rising earnings instability in the U.S. labor market, and particularly the high degree of earnings volatility among low-educated workers in the U.S. (Gottschalk and Moffitt 1994; Ziliak et al. 2011), highlight fluidity in economic statuses. A cross-sectional perspective thus may fail to capture the extent to which similarity in spouses’ earnings are produced through joint earnings stability or through joint earnings instability. For example, husbands’ and wives’ earnings may be highly correlated because spouses follow similar, consistent earnings patterns over time (as shown in Figure 1, panel A) or because their earnings fluctuate similarly over time (panel C). While these two manifestations of economic homogamy may appear similar in cross-sectional approaches, they likely have distinct consequences for the economic wellbeing of families.
2.2. The Long-Term Association Between Spouses’ Earnings
Understanding the association between spouses’ earnings and its implications for broader social inequalities, therefore, requires a life-course perspective that explicitly considers the connection between spouses’ long-term earnings patterns. However, the long-term association between husbands’ and wives’ earnings is not well-understood. Existing scholarship on the joint development of husbands’ and wives’ earnings and economic outcomes over time has primarily focused on the role of parenthood in intensifying and crystalizing couple-level gender inequalities in labor market outcomes among different-sex couples (e.g., Killewald and Garcia-Maglano 2016; Lundberg and Rose 2000; Musick et al. 2020). These studies, which typically focus on the years surrounding first birth, provide important insight into the mechanisms that produce gender inequalities during this pivotal point in the life course, revealing, for example that the emergence of gender inequalities in earnings within couples following the transition to parenthood is driven primarily by changes in wives’ labor supply rather than changes in husbands’ (Killewald and Garcia-Maglano 2016; Musick et al. 2020). However, the focus on the transition to parenthood provides little insight into how husband’s and wives’ earnings relate to one another in the longer term and tell us little about the association between spouses’ earnings among childless couples.
A small number of studies have examined spouses’ joint earnings patterns beyond the transition to parenthood. This line of research has highlighted the role of wives’ absolute characteristics and their characteristics relative to their husbands’ – particularly with respect to education – in shaping women’s economic standing within marriage. Using data from the National Longitudinal Survey of Youth 1979, Qian (2017; 2018) found that wives’ absolute and relative education at marriage were positively associated with their subsequent earnings and their likelihood of becoming a couple’s primary earner at some point during marriage. These findings suggest that wives’ absolute and relative socio-economic status likely play an important role in shaping the long-term association between spouses’ earnings. However, they shed little light on the association between spouses’ long-term earnings patterns itself or the varied ways that gender equality or similarity in spouses’ earnings trajectories may manifest within couples. In a subsequent study, Choi and Denice (2023) used Group-Based Trajectory Modeling to examine racial and ethnic variation in the association between wives’ relative education at marriage and their relative earnings trajectories, showing that the positive association identified by Qian (2018) is particularly strong for White women. Although this study provides greater insight into the long-term association between spouses’ earnings, its focus on relative earnings, rather than the association between spouses’ individual earnings trajectories, obscures our understanding of whether gender egalitarian long-term earnings patterns are produced through joint earnings stability or through joint earnings instability.
2.3. This Study: A Long-Term, Group-Based Approach to Examining Economic Homogamy
This study takes a long-term approach to measuring marital homogamy, examining husbands’ and wives’ earnings patterns over 30 years of marriage. Drawing on longitudinal data from a late Baby Boomer cohort of men and women, I use Dual Group-Based Trajectory Modeling (GBTM) to identify latent “ideal types” or groups of husbands and wives that follow qualitatively distinctive earnings patterns over time and examine the probabilities linking husbands’ and wives’ earnings groups (e.g., the probability that wives’ with decreasing earnings in marriage have husband with increasing earnings). This strategy allows me to evaluate the qualitative similarities (or differences) in husbands’ and wives’ entire marital earnings trajectories, rather than spouses’ earnings at a particular point in time, to uncover key underlying patterns in how spouses’ long-term earnings patterns relate to one another. In doing so, this study extends our understanding of marital homogamy by reconceptualizing economic homogamy as a life course process rather than a static characteristic that can be adequately captured at a single point in time.
I then examine how wives’ absolute socio-economic status at the time of marriage and their relative status compared to their husbands predict the long-term earnings configurations that couples ultimately follow to shed light on the implications of couples’ long-term earnings patterns for gender equality and economic inequalities between households. As a result of labor market polarization and the rise of economic precarity, earnings and employment outcomes are increasingly stratified in the U.S., with less-educated, lower-paid workers more likely to experience earnings instability and economic precarity (Gottschalk and Moffitt 1994; Kalleberg 2011). Wives’ own socio-economic status at the onset of marriage, therefore, is likely to be influential in shaping their own subsequent earnings trajectories (e.g., Becker 1994; Mincer 1974; Qian 2018), as well as the association between spouses’ earnings (Qian 2017).
Furthermore, couples across the socio-economic spectrum face differential constraints to achieving gender equality in the division of household labor, as well as different cultural orientations to gender egalitarian arrangements, which may affect the likelihood that spouses’ follow similar earnings patterns within marriage. One perspective argues that gender equality within different-sex couples is a privilege that is largely enjoyed by highly educated, high-SES couples (e.g. Esping Andersen and Billari 2015; Goldscheider et al 2015). This perspective suggests that higher-SES couples’ greater preference for gender egalitarian divisions of household labor (see Usdansky 2011), coupled with greater economic opportunities for women (Damaske 2011), better access to work-family policies to facilitate women’s employment continuity (see Chung 2020) and greater capacity to outsource domestic work and childcare (Gonalons-Pons 2015; Schneider and Hastings 2017) facilitate gender equality among high-SES couples in ways that may lead to greater similarity in spouses’ earnings over time. However, an alternative perspective argues that higher-SES couples may actually be less likely to follow gender egalitarian earnings patterns, despite their greater spoken support for gender equality, due to the long hours and strong ideal worker norms characteristic of professional occupations (Cha 2010; Maume 2006) and greater orientations towards intensive mothering (Hays 1996; Lareau 2003), each of which require substantial time commitments and lead couples to specialize in gendered ways (Usdansky 2011). This perspective suggests lower-SES couples, whose lower wages increase the need for two incomes and whose shift work or lack of work may facilitate a more gender equitable division of household labor (e.g., Gerstel and Clawson 2014), are more likely to follow gender egalitarian arrangements in practice.
Finally, the balance of spouses’ relative socio-economic status at marriage may also shape couples’ subsequent earnings configurations. Wives’ relative socio-economic status at marriage compared to their husbands’ reflects patterns of assortative mating, spouses’ comparative advantages in paid and unpaid work (Becker 1981) and the distribution of marital bargaining power (England and Kilbourne 1990). The balance of spouses’ relative characteristics, therefore, may set the tone for the economic roles each spouse is expected to fulfill in marriage. Wives with greater relative status at the start of marriage, for instance, may be more likely to maintain greater economic power over the marital life course (Choi and Denice 2023; Qian 2017; 2018).
3. Data, Measures and Methods
3.1. Data
The following analysis draws on data from the National Longitudinal Survey of Youth 1979 (NLSY79) from 1979 to 2018. NLSY79 follows a nationally representative sample of 12,686 men and women born between 1957 and 1964 who were first interviewed at ages 14–21 in 1979. The study collects detailed information for both sampled respondents and spouses on a variety of topics including demographics, marriage and relationships, work, education and income. Respondents were interviewed annually until 1994 and bi-annually thereafter. The large sample size, long-running data collection and availability of spousal information make NLSY79 and ideal data source for examining the long-term association between spouses’ earnings. However, beyond these advantages, NLSY79 respondents also represent a historically important cohort with which to examine couples’ long-term earnings patterns.
The last cohort of Baby Boomers, NLSY79 men and women came of age, married and entered the workforce during a remarkable period of social and economic change in the U.S. The reversal of the gender gap in college education and growing representation of women in male-dominated fields of study and occupations coupled with the mainstream availability of FDA-approved birth control pills, legalized abortion and the development of assistive reproductive technologies meant that NLSY79 women entered early adulthood with greater economic opportunity and greater reproductive freedom than women of previous generations (Goldin 2021). At the same time, NLSY79 men entered the workforce during a decades-long period of wage declines for men, driven by declines in manufacturing as well as technological change (e.g., Ruggles 2015). Taken together, these social and economic forces led to delays and declines in marriage among this cohort, and, among those who married, a substantial rise in the incidence of dual earner couples (Goldin 2021 and Ruggles 2015). This study considers how, in this historical context, married couples’ joint earnings trajectories unfolded.
My analytical sample focuses on 7,587 respondents in their first marriage and their different-sex spouses. Couples in this study consist of working-age (18–65) women and men whose marriage began in 1979 or later and where spouses were married between the ages of 18 and 50. I further restrict my sample to couples for whom earnings data is available for each partner for at least four waves to enable the estimation of both partners’ earnings trajectories, resulting in 5,491 couples. Those missing information on education or employment status at marriage were excluded from the analytic sample, resulting in a total of 5,354 couples. Couples were followed for up to 30 years of marriage, with an average of 16 waves for each spouse (minimum = 4, maximum = 24), resulting in 87,989 couple-wave observations.
Following previous research (e.g., Gonalons-Pons and Schwartz 2017; 2018), this study focuses on first marriages only for both substantive and practical reasons. Because remarriages take place later in the life course on average (Reynolds 2021) and are less bound by the clear gendered norms and expectations associated with first marriage (Cherlin 1978), spouses’ earnings trajectories and their association within remarriages may differ from those within first marriages. The relatively small sample size of second and subsequent marriages in NLSY79 (approximately 1,700), and the shorter marital duration observed among these remarriages in comparison to first marriages (16 years vs. 28 years at the median) preclude this type of comparison in this study. However, how partners’ long-term earnings patterns and their connections vary between first and subsequent marriages remains an important question for future study.
3.2. Measures and Methods
The following analyses proceeded in three steps. First, I used Dual Group-Based Trajectory Modeling to estimate husbands’ and wives’ earnings trajectories for up to 30 years of marriage. Second, I examined the relationship between spouses’ entire marital earnings trajectories using conditional and joint probabilities of husbands’ and wives’ group membership and classify each possible combination of husbands’ and wives’ trajectory groups into five broader couple types based on the qualitative characteristics of each spouses’ earnings trajectory. Finally, I used multinomial logistic regression to examine the factors associated with couples’ long-term joint earnings configurations.
3.2.1. Describing Husbands’ and Wives’ Earnings Trajectories
I used Dual Group-Based Trajectory Modeling to jointly estimate husbands’ and wives’ earnings trajectories over the course of up to 30 years of first marriage. Group-Based Trajectory Modeling (GBTM) is an application of finite mixture modeling that uses maximum likelihood estimation to identify clusters of individuals following approximately similar trajectories of an outcome over time (Nagin 2005). In contrast to more commonly used variable-centered longitudinal methods such as growth curve modeling, which estimates a single mean population trajectory of an outcome over time and examines how characteristics and events explain variation around that mean trajectory, GBTM assumes that the population is comprised of multiple latent, qualitatively distinct developmental trajectories or “ideal types” (see Song et al. 2022 for further discussion of these contrasting approaches). The primary aim of GBTM, as well as other person-centered longitudinal methods, is to identify and describe a range of key developmental patterns within the population using a data-driven approach that maximizes within-cluster similarity in longitudinal patterns and minimizes between-cluster similarities. As such, person-centered longitudinal approaches are particularly useful for identifying less-common developmental patterns that are obscured in variable-centered approaches that focus on population averages. An important advantage of GBTM and other formal person-centered methods over a theoretically-driven approach that defines trajectory groups based on a priori expectations is that such approaches offer no mechanisms for empirically falsifying the number of distinct trajectory groups or for assessing uncertainty in the assignment of individuals to groups (Nagin 2005).
GBTM has been used previously in sociological research to examine life course patterns of employment and unemployment (e.g., Damaske and Frech 2016; Damaske, Frech and Wething 2023; Garcia Manglano 2015; Weisshaar and Cabello-Hutt 2020). Its ability to model continuous outcomes, such as earnings, make it an ideal approach for the present study in comparison to other commonly-used person-centered methods, such as sequence analysis.
I used a dual extension of the Group-Based Trajectory Model, first developed by Nagin and Tremblay (2001), which allows for the simultaneous analysis of two separate but related outcomes and their connections – in this case, husbands’ and wives’ earnings. The key advantage of the Dual GBTM model over conducting separate univariate GBTM analyses is its ability to produce consistent estimates of the joint probabilities of group membership across outcomes and its ability to account for uncertainty in group assignment in producing these joint probabilities (see Nagin 2005 Chapter 8 for further discussion).
The outcome variables in this stage of analysis are the log of wives’ and husbands’ gross annual labor earnings, in 2018 USD. To address observations with zero earnings, I added a constant of one dollar to all inflation-adjusted values of earnings before taking the log.1 Time is measured as years since marriage, ranging from 0 to 30.
The likelihood function for the dual trajectory model is as follows. Earnings sequences from 0 to T years since marriage are given by Y1 = {y11, y12, …., y1T} for wives and by Y2 = {y21, y22, …., y2T} for husbands. The number of unique trajectory groups for wives and husbands are allowed to vary by gender and are represented by J and K, respectively. These parameters are not estimated directly in GBTM but are chosen by the researcher (discussed further below). For each couple, the likelihood of observing the earnings sequences {Y1, Y2} is given by:
where πjk is the joint probability of membership in trajectory group j for Y1 and group k for Y2. fj(Y1) denotes the probability of observing the earnings trajectory Y1, given wives’ membership in group j while hk(Y2) denotes the probability of observing the earnings trajectory Y2, given husband’s membership in group k.
Over time, T husbands’ and wives’ earnings trajectories are expressed as a joint probability for each spouse,
and
Where, within each group j and, k, conditional independence is assumed for income sequences Y1 and, Y2 respectively.
I modeled the development of earnings over marriage with a censored normal distribution to account for the left-censored nature of earnings. Given the dynamic nature of earnings over the life course, particularly for women, initial models assume that the log of husbands’ wives’ earnings’, Yit, is a cubic function of years since marriage in each trajectory group:
where the error term εit is assumed to be independently, identically distributed with a mean of 0 and a variance of σ2.
As discussed extensively in the GBTM literature (Nagin and Tremblay 2005a, 2005b; Nylund et al. 2007), determining the optimal number of unique trajectory groups depends on both the interpretability of results and goodness of fit to the data. Following the standard approach to Dual GBTM outlined in Nagin (2005), I tested nested models for husbands and wives separately, varying only the number of groups included in the models, and comparing the BIC and AIC across models. After selecting the number of groups included in the model, I refined the polynomial function of marital duration for each group based on the statistical significance of the parameter estimates. Finally, I estimated a dual trajectory model using the preferred number of groups and functional forms identified in the single trajectory models and assigned each husband and wife to the trajectory group for which they have the highest probability of group membership.
3.2.2. Associations between Husbands’ and Wives’ Earnings Trajectories: Developing Couple Types
Next, I examined the association between husbands’ and wives’ long-term earnings patterns by using the conditional and joint probabilities of husbands’ and wives’ trajectory groups. To examine broader underlying patterns in this association, I then classified each possible combination of husbands’ and wives’ earnings groups into five main couple types based on both the level and shape of predicted earnings in each trajectory group over time. Combinations of husbands’ and wives’ trajectory groups were classified as:
“Dual earner” where husbands and wives follow similar, consistent long-term earnings patterns;
“Jointly mobile” where husbands’ and wives’ earnings increase, decrease or fluctuate similarly over time;
“Alternating earner” where husbands and wives alternate which spouse is the primary earner over time;
“Male breadwinner” where husbands maintain higher earnings than wives; and
“Female breadwinner” where wives maintain higher earnings than husbands.
Each couple was then allocated to these five couple types based on spouses’ assigned earnings trajectory groups.
3.2.3. Factors Associated with Couple Types
Finally, I used multinomial logistic regression to examine what factors characterize couples’ long-term earnings configurations. The outcome variable in this analysis is the 5-category measure of couple type discussed above. Independent variables in the model included a series of indicators capturing wives’ absolute and relative socio-economic status at marriage.
Wives’ absolute socio-economic status at marriage:
I include measures of wives’ earnings, educational attainment and work hours in the year of marriage. Wives’ earnings are measured as a categorical variable indicating their earnings tertile within the analytic sample of wives, and distinguishes between wives with low, medium and high earnings at marriage. Educational attainment is a categorical measure based on years of education reported in the year of marriage, distinguishing between those with less than high school, a high-school-level degree, some college, and college or above. Work hours at marriage is a categorical measure derived from annual hours worked in the year of marriage, distinguishing between those working less than full time (0–1749 hours), full time (1750 – 2249 hours) and more than full time (2250 hours or more). Following Killewald and Gough (2013), where earnings, education or work hours are missing for the year of marriage, I imputed their values from the closest available wave within the first three years of marriage.2 After imputation, 46 couples (less than 1%) have missing information on earnings at marriage and are excluded from the multivariate analysis.
Relative socio-economic status at marriage:
I include measures of wives’ relative earnings, education and work hours at marriage to examine how couple-level power dynamics at the time of marriage relate to the long-term association between spouses’ earnings within couples. Relative earnings is a categorical measure of wives’ share of couple earnings at marriage, distinguishing between couples where wives earned less than 40%, 40–60%, and those where wives earned more than 60% of total couple earnings. Relative education is based on a comparison between husband’s and wives’ education levels at marriage (discussed above) and distinguishes between couples where wives had a lower, equal or higher level of education at marriage compared to husbands. Similarly, relative work hours distinguishes between couples where wives had lower, similar or greater labor supplies at marriage.
Controls:
I control for wives’ age at marriage (measured as a continuous variable), the age difference between spouses (measured by subtracting husband’s birth year from wives’) and the total number of children born to the couple during marriage. I also control for the gender and race of the NLSY79 respondent. Because spousal earnings data were collected by proxy in NLSY79, they may be subject to greater measurement error. Therefore, I include a binary control variable for whether the husband or wife was the respondent. Work-family trajectories and the division of household labor among couples in the U.S. vary by race and ethnicity (Cheng 2016; Choi and Denice 2023; Pessin and Pojman 2022). However, race information for spouses of NLSY respondents is only available since 2008, with respondents not interviewed in these waves missing data on spouses’ race. Therefore, I include a measure reflecting only respondents’ race. This measure distinguishes between Hispanic, Black and non-Black, non-Hispanic (referred to as White).
4. Results
4.1. Descriptive Results
Figure 2 shows mean annual labor earnings for husbands and wives in my analytic sample by years since marriage and the correlation between spouses’ earnings calculated in each year of marriage. Results corroborate prior research showing that long-term earnings patterns differ substantially between husbands and wives and that the association between spouses’ earnings varies over the marital life course (Gonalons-Pons and Schwartz 2017; Killewald and Garcia-Manglano 2016; Qian 2018). Husbands’ and wives’ earnings were most similar in the year of marriage, although the gender gap in earnings at marriage was still substantial. On average, husbands earned approximately $35,000 per year at marriage compared with around $21,000 among wives. In the years following marriage, wives’ average earnings increased much more slowly compared with husbands, resulting in a widening gap in mean earnings that peaked at over $50,000 per year 19 years after marriage when husbands’ means earnings reached $85,000 and wives’ earnings reached around $35,000. The correlation between spouses’ earnings followed a similar pattern over marriage. The correlation between husbands’ and wives’ earnings was highest in the year of marriage at 0.48 before declining thereafter, reaching a nadir of −.08 19 years after marriage before gradually re-increasing.
Figure 2.
Spouses’ Mean Annual Labor Earnings and Correlation Coefficient by Years Since Marriage
Notes: N=5,354; Husbands’ and wives’ mean annual earnings and the correlation between spouses’ earnings by years since marriage are shown in points. Trend lines are lowess smoothed. Source: NLSY79 1979–2018.
Table 1 reports additional descriptive statistics for couples in my sample. Across the sample, husbands and wives had similar levels of education at marriage, although wives were slightly more likely than husbands to have attended some college and were slightly less likely to have less than a high school degree. On average, wives earned less and worked fewer hours at marriage than husbands. These gender-specific patterns were largely reflected at the couple level as well. In over half of couples (54%), spouses had attained the same degree of education at marriage, while in 23% of couples the husband was more highly educated and in 24% of couples the wife was more highly educated. Husbands worked longer hours and earned more than wives did in 51% of couples while spouses had similar earnings and work hours in over a third of couples (34%) and women worked more and earned more than husbands in a minority of couples (14% and 15% respectively). On average, couples had 1.7 children over the period of marriage observed. Around two-thirds of NLSY respondents were White, 18% were Black and 16% were Hispanic. In 51% of couples the wife was the NLSY respondent.
Table 1.
Descriptive Statistics
| Variable | Wives | Husbands |
|---|---|---|
| Earnings at marriage* (mean(sd)) | 20,509 (22,923) | 34,623 (32,297) |
| Education at marriage* (%) | ||
| Less than HS | 12.9 | 14.7 |
| High School | 44.6 | 44.7 |
| Some college | 24.2 | 21.2 |
| BA or higher | 18.3 | 19.4 |
| Work hours at marriage* (%) | ||
| Less than FT | 56.2 | 28.4 |
| FT | 36.0 | 43.8 |
| More than FT | 7.8 | 27.9 |
| Age at marriage* (mean (sd)) | 23.9 (5.1) | 25.7 (5.5) |
| Relative earnings at marriage | ||
| Wife earned <40% | 51.0 | |
| Wife earned 40–60% | 34.3 | |
| Wife earned >60% | 15.0 | |
| Relative education (%) | ||
| Wife less educated | 22.7 | |
| Spouses equally educated | 53.6 | |
| Wife more educated | 23.8 | |
| Relative work hours at marriage (%) | ||
| Wife worked less | 51.5 | |
| Spouses worked similar hours | 34.3 | |
| Wife worked more | 14.3 | |
| Age gap (mean (sd)) | 1.9 (4.0) | |
| Total number of children born in marriage (mean (sd)) | 1.7 (1.2) | |
| Respondent race (%) | ||
| Black | 18.3 | |
| Hispanic | 16.0 | |
| White | 65.7 | |
| Wife is NLSY respondent (%) | 50.6 | |
Notes: N=5,354 couples; Asterisks indicate that gender differences are statistically significant at p<.05 using a χ2 test of independence for categorical measures and a two-tailed t-test for continuous measures. Source: NLSY79 1979–2018.
4.2. Describing His and Her Earnings Trajectories
I tested Group Based Trajectory Models with the number of trajectory groups ranging between two and eight and all groups specified using a cubic function of marital duration. Balancing goodness-of-fit to the data against interpretability of the results, I selected a six-group model for wives and a four-group model for husbands. BIC and AIC statistics (shown in Appendix A Table 1) for both husbands’ and wives’ models indicated that models with greater numbers of groups fit the data best. However, the additional groups generated by models with more than six groups for wives and more than four groups for husbands duplicated existing trajectory groups, indicating that the additional complexity added by increasing the number of groups in the model did not provide additional substantive insight (see Appendix A Figure 1 for an example). Furthermore, models with more than six groups for wives and more than four groups for husbands generated sparse trajectory groups comprising less than 5% of observations. After selecting the six-group model for wives and the four-group model for husbands, I tested models varying the polynomial function of marital duration for each group based on the statistical significance of the parameter estimates, ultimately reducing a cubic function to a quadratic function for one group of husbands and one group of wives. Additional model adequacy measures, including relative entropy and mean posterior probabilities, are available in Appendix A Table 2.
Figure 3 presents the results from the Dual GBTM visually, showing the estimated earnings trajectories for each group for up to 30 years of marriage (model coefficients are available in Appendix A Table 3).3,4
Figure 3.
Wives’ and Husbands’ Estimated Earnings Trajectory Groups
Notes: N=5,354; This graph shows the estimated earnings trajectories generated from a Dual Group-Based Trajectory Model with six groups for wives and four groups for husbands. Earnings are modeled as a cubic function of years since marriage for all groups except for the “Career climber” group for wives and the “Stable earner” group for husbands where earnings are modeled as a quadratic function of years since marriage. Each husband and wife in the sample was assigned to the trajectory group for which they had the highest estimated probability of belonging. Numbers in parentheses refer to the percentage of husbands or wives in each trajectory group. The thickness of trajectory lines is weighted according to the prevalence of each trajectory group in the sample. Source: NLSY79 1979–2018.
The results for wives (shown in the top panel of Figure 3) show substantial heterogeneity in earnings patterns over first marriage. Perhaps surprisingly, the most common long-term earnings pattern among wives was a Stable earner trajectory, describing 55% of wives in the sample. As the name suggests, this trajectory group is characterized by a high degree of earnings stability over the course of marriage. Additional descriptive statistics of earnings patterns across trajectory groups (available in Appendix A Table 4) show that, on average, wives in the Stable earner group maintained higher earnings on average than wives in other groups, with a marital average of nearly $36,000 per year. Wives in this group also spent little time out of the labor force throughout marriage, earning $0 per year in just 4% of marital years observed on average.
The second most common long-term earnings pattern for wives was a Career pause trajectory, describing the earnings patterns of 12% of wives in the sample. Wives in the Career pause group began marriage with relatively high earnings, but their earnings declined sharply within the first 10 years of marriage, remaining low for several years before eventually recovering to their initial level. On average, wives in this group spent over a third (38%) of marital years observed with zero earnings. The third most common trajectory group was a Career climber group (9% of wives), whose earnings were low at the start of marriage but increased steadily over time. Fourth was a Low earner group (9% of wives) characterized by very low levels of earnings throughout marriage. Average annual earnings were lowest among wives in this group at less than $2,000 per year, with wives spending most years (83% on average) not working. Fifth was a Late decreasing group (9%) in which wives experienced earnings growth in the first third of marriage but earnings declined thereafter. Finally, the least common trajectory group among wives was an Early exit group (7%). Like the Career pause group, wives in the Early exit group began marriage with relatively high earnings and experienced precipitous earnings declines in the first years of marriage, maintaining zero annual earnings for an average of 40% of marital years observed. Unlike Career pause wives, earnings in this group never fully recovered to their initial starting point following earnings declines.
The results for husbands (shown in the bottom panel of Figure 3) show less variation in earnings compared with wives – both in terms of the number of distinct long-term earnings patterns revealed and in terms of variation over time within groups. As for wives, the most common long-term earnings pattern identified for husbands was a Stable earner trajectory, describing the earnings patterns of 77% of husbands in the sample. Unsurprisingly, husbands in this group maintained the highest average annual earnings (nearly $59,000 per year) and spent the least amount of time out of the labor force (3% of marital years observed) compared with husbands in other groups. The second most common trajectory group for husbands was a Decreasing group, including 9% of husbands. Earnings among men in this group were relatively high at marriage but declined steadily over the marital life course. Third was a Late earner group (7%), in which husbands had little to no earnings at the start of marriage but whose earnings increased later on. On average, husbands in this group spent 80% of marital years observed with zero annual earnings and, as a result, they had the lowest average annual earnings of all husbands at nearly $7,500 per year. Finally, the least common trajectory type among husbands was a Career climber group (7%). Like the Career climber group for wives, husbands’ earnings in this group were relatively low at the start of marriage but increased sharply in the years following.
4.3. The Association between Husbands’ and Wives’ Earnings Trajectories
Table 2, presenting the conditional and joint distributions of husbands’ and wives’ earnings trajectories, shows a substantial degree of variation in the association between husbands’ and wives’ long-term earnings trajectories within couples. In many cases, the results show a high degree of congruence between husbands’ and wives’ earnings patterns within couples in terms of the level and shape of earnings trajectories. Stable earner husbands were more likely to be married to Stable earner wives (61%; top panel) than husbands in any other earnings group (20–45%). Similarly, Stable earner wives were more likely to be married to Stable earner husbands (86%; middle panel) than wives in the Career climber, Low earner and Late decreasing groups (55%, 56% and 58%, respectively). As shown in the bottom panel, the combination of a Stable earner husband and a Stable earner wife was the most common long-term earnings constellation by far, describing 47% of couples in the sample.
Table 2.
Conditional and Joint Probabilities of Earnings Group Membership
| Probability of wives’ earnings group conditional on husbands’ earnings group | |||||||
|---|---|---|---|---|---|---|---|
| Wives | |||||||
| Husbands | Stable earner | Career pause | Career climber | Low earner | Late decreasing | Early exit | Row total |
| Stable earner | 0.61 | 0.12 | 0.06 | 0.07 | 0.06 | 0.08 | 1.00 |
| Decreasing | 0.45 | 0.12 | 0.12 | 0.12 | 0.12 | 0.06 | 1.00 |
| Late increasing | 0.20 | 0.12 | 0.23 | 0.29 | 0.14 | 0.02 | 1.00 |
| Career climber | 0.35 | 0.08 | 0.21 | 0.11 | 0.22 | 0.03 | 1.00 |
| Probability of husbands’ earnings group conditional on wives’ earnings group | |||||||
| Wives | |||||||
| Husbands | Stable earner | Career pause | Career climber | Low earner | Late decreasing | Early exit | |
| Stable earner | 0.86 | 0.79 | 0.55 | 0.56 | 0.58 | 0.87 | |
| Decreasing | 0.07 | 0.09 | 0.11 | 0.12 | 0.13 | 0.08 | |
| Late increasing | 0.02 | 0.07 | 0.18 | 0.23 | 0.12 | 0.02 | |
| Career climber | 0.04 | 0.05 | 0.16 | 0.09 | 0.18 | 0.03 | |
| Column total | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | 1.00 | |
| Joint probability of wives’ and husbands’ earnings groups | |||||||
| Wives | |||||||
| Husbands | Stable earner | Career pause | Career climber | Low earner | Late decreasing | Early exit | Row total |
| Stable earner | 0.47 | 0.09 | 0.05 | 0.05 | 0.05 | 0.06 | 0.77 |
| Decreasing | 0.04 | 0.01 | 0.01 | 0.01 | 0.01 | 0.01 | 0.09 |
| Late increasing | 0.01 | 0.01 | 0.02 | 0.02 | 0.01 | 0.00 | 0.07 |
| Career climber | 0.02 | 0.01 | 0.01 | 0.01 | 0.02 | 0.00 | 0.07 |
| Column total | 0.55 | 0.12 | 0.09 | 0.09 | 0.09 | 0.07 | 1.00 |
Notes: N=5,354; Pearson χ2 test of independence between wives’ and husband’s earnings trajectories: χ2 (15 d.f.) = 686.4; p<0.00; Source: NLSY79 1979–2018.
Couples with earnings patterns that varied over time also tended to be more likely to pair with spouses with varying earnings. In some couples, changes in spouses’ earnings over time took place roughly in parallel. For example, wives with Late decreasing earnings were more likely to pair with husbands with Decreasing earnings than wives with other earnings patterns (13% vs. 7–12%; middle panel) and Late increasing and Career Climber husbands were most likely to pair with Career climber wives (23% and 21% vs. 6–12%; top panel). In other couples, changes in earnings patterns over time followed contrasting temporal patterns such that husbands and wives alternated which spouse was the primary earner over time. For instance, Career climber husbands were more likely than husbands with other earnings patterns to be married to wives with Late decreasing earnings (22% vs. 6–14%; top panel).
Finally, in some cases, the results also highlighted long-term economic specialization within couples. For example, the probability of being married to a Stable earner husband was highest among Early exit wives (87%; middle panel). And although this probability was second highest among Stable earning wives, as mentioned above, it was third highest among Career pause wives (79%).
To capture broader underlying patterns in husbands’ and wives’ long-term earnings patterns within couples, I categorized each combination of husbands’ and wives’ earnings groups into couple types based on the level of earnings and the shape of spouses’ earnings trajectories over the marital life course, shown in Figure 4. These couple types describe five main long-term earnings patterns among couples: Dual earner, Jointly mobile, Alternating earner, Male breadwinner or Female breadwinner.
Figure 4.
Couple Types based on Estimated Earnings Trajectories of Husbands and Wives
Notes: N=5,354; Lines show estimated earnings trajectories generated from a Dual Group-Based Trajectory Model with six groups for wives and four groups for husbands. Husbands’ earnings trajectories are shown in solid lines, wives’ earnings trajectories are shown in dashed lines. Each combination of husbands’ and wives’ earnings groups are classified into five couple types based on the level and shape of predicted earnings in each trajectory group. “Dual earner” couples are those where husbands and wives followed similar, stable earnings patterns over the course of marriage; in “Jointly mobile” couples, spouses’ earnings increase, decrease or fluctuate similarly over time; in “Alternating earner” couples, husbands and wives alternate which spouse is the primary earner over time; in “Male breadwinner” couples, husbands earnings tend to exceed wives’ in marriage; and in “Female breadwinner” wives’ earnings tend to exceed husbands. Each couple was allocated to a couple type based on spouses’ assigned earnings trajectory groups. Numbers in the bottom-left corner of each cell show the proportion of sample couples in each combination of husbands’ and wives’ earnings groups. Numbers in parentheses refer to the percentage of sample couples in each couple type. Source: NLSY79 1979–2018.
I find that over half of couples followed similar long-term earnings patterns over time. 50% of couples followed a Dual earner pattern, where husbands’ and wives’ earnings trajectories followed similar, consistent patterns over time, while 6% followed a Jointly mobile pattern, where spouses’ earnings increased, decreased or fluctuated similarly over time. In 4% of Jointly mobile couples, husbands’ and wives’ earnings patterns were predominantly upwardly mobile over the course of marriage (including: Career pause wives with Late earning husbands, Career climber wives with Late earning husbands, and Career climber wives with Career climber husbands) while 2% of couples were jointly downwardly mobile (Late decreasing wives with Decreasing husbands and Early exit wives with Decreasing husbands).5 5% of couples followed an Alternating earner pattern, with primary earnership alternating between spouses over time. These results show that although dual, stable earning couples constitute the majority of couples who follow homogamous or gender egalitarian earnings patterns, there is heterogeneity in how similarity or equality in earnings manifests within couples in the long term. Finally, 39% of couples followed patterns of long-term specialization over the course of marriage, with 34% following Male breadwinner patterns and 5% following long-term Female breadwinner patterns.
Additional descriptive statistics, presented in Table 3, demonstrate that the five long-term earnings patterns generated from GBTM earnings groups conform well to common approaches of measuring gender equality or similarity in earnings within couples used in prior scholarship. Drawing on approaches from the gender and family literature, I first considered how wives’ economic power within couples varies across couple types by calculating wives’ share of couple earnings averaged over all marital years observed. I find that wives’ average level of relative earnings within couple types aligns with the commonly-used definition of gender equality in economic power that considers couples where wives’ earn between 40–60% of couple earnings as gender egalitarian (see e.g., Raley, Mattingly and Bianchi 2006). On average, wives in Dual earner couples earned 40% of couple earnings, wives in Jointly mobile couples earned 47% of couple earnings, and wives in Alternating earner couples earned 44% of couple earnings over the course of marriage. By contrast, wives’ average share of couple earnings was 17% among Male breadwinner couples and 61% among Female breadwinner couples.
Table 3.
Relative Earnings by Couple Type
| Dual earner | Jointly mobile | Alternating Earner | Male breadwinner | Female breadwinner | Total | |
|---|---|---|---|---|---|---|
| Wives’ share of couple earnings (marital average) - mean | 0.40 | 0.47 | 0.44 | 0.17 | 0.61 | 0.33 |
| Proportion of years in which wives earned: - mean | ||||||
| <40% of couple earnings | 0.49 | 0.45 | 0.49 | 0.85 | 0.22 | 0.60 |
| 40–60% of couple earnings | 0.39 | 0.16 | 0.13 | 0.11 | 0.24 | 0.26 |
| >60% of couple earnings | 0.12 | 0.39 | 0.39 | 0.04 | 0.54 | 0.15 |
| Relative earnings tertile (marital average) - % | ||||||
| Wife lower | 12.8 | 5.8 | 11.3 | 53.2 | 2.5 | 25.6 |
| Spouses equal | 48.4 | 70.9 | 62.7 | 42.3 | 15.3 | 46.6 |
| Wife higher | 38.8 | 23.4 | 26.1 | 4.5 | 82.3 | 27.9 |
Notes: N=5,354; Wives’ share of couple earnings is calculated as the ratio of her average annual earnings during marriage by the sum of both spouses’ average marital earnings. Relative earnings tertile is calculated by comparing spouses’ gender-specific earnings tertile based on their average annual earnings during marriage. Source: NLSY79 1979–2018.
Recognizing that wives’ relative earnings may vary substantially over time (e.g., Winslow-Bowe 2006), I also considered the proportion of marital years in which wives earned less than 40%, 40–60% and more than 60% of couple annual earnings. I find that wives in gender egalitarian couple types maintained substantial economic power in a majority of years of marriage. Wives’ earnings were similar to or exceeded their husbands’ in 51% of marital years observed among Dual earner and Alternating earner couples and 55% of years among Jointly mobile couples. In comparison, wives in Male breadwinner couples earned less than 40% of couple earnings in 85% of marital years observed, while wives in Female breadwinner couples earned more than 60% of couple earnings in 54% of years observed.
Finally, drawing from the economic homogamy literature (e.g., Schwartz 2010; Shen 2021), I considered the degree of similarity in spouses’ locations within gender specific earnings distributions, comparing husbands’ and wives’ earnings tertiles within couples. This approach captures whether, for example, high-earning wives tend to pair with high-earning husbands within the context of gender-specific earnings distributions that differ substantially due to gender differences in labor force attachment, occupation and pay. The results show that wives’ earnings tertile was equal to or exceeded their husbands’ in the vast majority of Dual earner, Jointly mobile and Alternating earner couples. By contrast, in 53% of couples classified as Male breadwinner husbands’ tertile exceeded their wives’ and in 82% of Female breadwinner couples wives’ tertile exceeded their husbands. Taken together, these descriptive statistics confirm that, among couples classified as “gender egalitarian” in this analysis (i.e. Dual earner, Jointly mobile and Alternating earner couples), spouses’ earnings are similar both in terms of absolute earnings levels and in terms of spouses’ relative position within gender-specific earnings distributions.
4.4. Predictors of Couple’s Long-Term Earnings Patterns
Table 4 presents the results from the multinomial logistic regression, examining how wives’ absolute socio-economic status at marriage (measured in terms of earnings, education and employment status) and relative status compared with their husbands’ influence couples’ long-term earnings configurations6. To aid interpretation, all results are presented as average marginal effects on the probability of belonging to different couple types. For example, the 0.303 in the first cell of Table 4 indicates that the predicted probability of following a Dual earner long-term earnings pattern was 30 percentage points higher among couples where wives had medium-level earnings at the onset of marriage than where wives had low earnings. Full regression coefficients are available in Appendix A Table 5. Key results are summarized in Table 5.
Table 4.
Predicted Probability of Couple Type Membership – Average Marginal Effects
| Dual earner | Jointly mobile | Alternating earner | Female breadwinner | Male breadwinner | |
|---|---|---|---|---|---|
| Wife’s earnings tertile at marriage (ref = low) | |||||
| Medium | 0.303* | −0.093* | −0.074* | 0.037* | −0.174* |
| High | 0.383* | −0.108* | −0.096* | 0.030* | −0.209* |
| Wife’s education at marriage (ref = less than HS) | |||||
| High School | 0.053* | 0.005 | −0.005 | 0.012 | −0.064* |
| Some college | 0.134* | −0.011 | −0.023* | 0.015 | −0.114* |
| BA or higher | 0.183* | −0.028* | −0.037* | −0.005 | −0.113* |
| Wife’s work hours at marriage (ref = less than FT) | |||||
| FT | 0.110* | −0.009 | −0.028* | −0.026* | −0.047* |
| More than FT | 0.165* | −0.026 | −0.020 | −0.042* | −0.077* |
| Relative earnings at marriage (ref = wife earns < 40% of couple earnings) | |||||
| Wife earns 40–60% | −0.036* | 0.075* | 0.035* | 0.010 | −0.083* |
| Wife earns > 60% | −0.071* | 0.078* | 0.122* | 0.042* | −0.170* |
| Relative education at marriage (ref = wife less educated) | |||||
| Spouses equally educated | −0.003 | 0.003 | 0.017* | 0.005 | −0.021 |
| Wife more educated | 0.028 | −0.002 | 0.012 | 0.027* | −0.065* |
| Relative work hours at marriage (ref = wife worked less) | |||||
| Spouses worked similar hours | 0.015 | 0.001 | 0.011 | 0.010 | −0.037* |
| Wife worked more | −0.035 | −0.016 | 0.022 | 0.048* | −0.020 |
| Base rate | 0.501 | 0.055 | 0.053 | 0.052 | 0.339 |
Notes: N=5,308; Asterisks indicate where average marginal effects are statistically significant at p < 0.05. Model controls for wives’ age at marriage, the age gap between spouses, the number of children born in marriage and the race and gender of the NLSY respondent. Source: NLSY79 1979–2018.
Table 5.
Summary of the Association between Wives’ Absolute and Relative Socio-economic Statis at Marriage and Couples’ Joint Earnings Patterns
| Couple earnings’ patterns | |||||
|---|---|---|---|---|---|
| Dual earner | Jointly mobile | Alternating earner | Female breadwinner | Male breadwinner | |
| Wives’ absolute status | + | − | − | + | − |
| Wives’ relative status compared to husbands’ | − | + | + | + | − |
Notes: Socio-economic status is measured in terms of earnings, educational attainment and employment status. + indicates a positive association, - indicates a negative association.
The results for Dual earner, Jointly mobile and Alternating earner couples reveal a sharp socio-economic divide in how gender equality in earnings manifests within couples in the long term. The probability that couples followed Dual earner earnings trajectories was strongly, positively associated with wives’ greater absolute socio-economic status at the time of marriage (in terms of earnings, education and labor supply) but negatively associated with wives’ relative earnings at marriage compared with their husbands’. Couples with a high-earning wife at marriage, for example, were 38 percentage points more likely to follow Dual earner trajectories over the course of marriage than those with a low-earning wife. However, couples where wives brought in 60% of couple earnings or more at marriage were 7 percentage points less likely to follow a Dual earner pattern compared with couples where wives earned less than 40% of couple earnings. Taken together, these results indicate that the couples most likely to have two stable incomes over the course of marriage are those where wives were already socio-economically advantaged at marriage relative to other wives and where husbands were even more advantaged in terms of earnings.
By contrast, the couples most likely to follow alternative gender egalitarian earnings patterns over the course of marriage were those in which wives had lower absolute socio-economic status at the onset of marriage but greater status relative to their husbands. The probabilities of following Jointly mobile or Alternating earner patterns were negatively associated with wives’ earnings, education and labor supply at marriage. Meanwhile, couples in which wives earned 60% or more of couple earnings at marriage were 8 percentage points more likely to follow a Jointly mobile earnings pattern and 12 percentage points more likely to follow Alternating earner patterns compared with couples where wives brought in less than 40% of couple earnings at marriage. These findings suggest that among women with lower socio-economic status at marriage, greater marital bargaining power (in the form of greater relative earnings) increases the likelihood that couples follow gender egalitarian earnings patterns in the long term. However, reflecting broader educational inequalities in economic opportunities and earnings volatility in the U.S. (Autor 2014; Ziliak et al. 2011), these gender egalitarian joint earnings trajectories differ qualitatively from those followed by couples with higher socio-economic status and are marked by less earnings stability over the marital life course.
Considered alongside the results for Dual earner, Jointly mobile and Alternating earner patterns, the results for Female breadwinner and Male breadwinner couples highlight the importance of both absolute status and relative status in shaping the long-term association between spouses’ earnings. Like Dual earner couples, the probability that couples followed long-term Female breadwinner patterns was positively associated with wives’ absolute earnings at the start of marriage. However, in contrast to Dual earner couples, the likelihood of following Female breadwinner patterns was also positively associated with wives’ relative earnings and education at marriage compared with their husbands. Similarly, the likelihood that couples followed long-term Male breadwinner patterns was negatively associated with wives’ absolute earnings, education and labor supply at marriage, as observed among Jointly mobile and Alternating earner couples. However, Male breadwinner patterns were negatively associated with wives’ relative earnings and education compared with their husbands’.
5. Robustness Analyses
Marriages in my sample are disproportionately long-lasting, in part due to the requirement of earnings data for both spouses for at least four waves of data. 67% of marriages in the sample had not ended by the couples’ most recently available wave of data. 26% of marriages ended in divorce and the remaining 6% ended with the death of a spouse or for unknown reasons. Because husbands’ and wives’ earnings patterns within marriage are related to the risk of divorce (Kalmijn and Poortman 2006; Kalmijn et al. 2007), I examined whether the long-term earnings patterns identified in Figure 3 are biased by sample selection and differential attrition from the sample due to divorce.
In one sensitivity analysis, I reran my Dual GBTM separately on subsamples of couples whose marriages did not end and those that ended in divorce, truncating my analysis at 15 years of marriage because the large majority of divorcing couples had marriages lasting fewer than 15 years. The results from this analysis are available in Appendix C Figures 1 and 2 and show similar patterns across subsamples. Furthermore, the results for both divorced and enduring marriages broadly reflect the patterns identified in my main results in terms of the shape and distribution of trajectory groups, although, because the period of analysis is truncated at 15 years of marriage, the model cannot distinguish between the “Career climber” and “Late decreasing” groups for wives, resulting in 5 trajectory groups for wives in both subsamples (additional discussion is available in Appendix C).
In a second sensitivity analysis, I considered whether my Dual GBTM results are affected by attrition from the sample more broadly (whether due to divorce or not), by rerunning my analysis on subsamples of couples observed for fewer than 15 waves of data (34% of couples) and those observed for 15 waves of data or more. Results from these analyses are available in Appendix C Figures 3 and 4. As above, because of the truncated period of analysis, the model does not distinguish between the “Career climber” and “Late decreasing” groups for wives, resulting in 5 trajectory groups for wives in both subsamples. Although the results show some minor differences across subsamples in the timing of earnings declines and increases in some trajectory groups and in the proportion of husbands and wives in each group, the results show broadly similar patterns to my main results in terms of the shape of distinct trajectories identified and their distribution in the population. See Appendix C for further discussion.
6. Discussion
The association between husbands’ and wives’ earnings over the marital life course has important implications for gender inequality within different-sex marriages as well as for economic inequalities between households. Much of our understanding of economic homogamy is based on cross-sectional approaches. However, recent scholarship has demonstrated that association between spouses’ earnings varies substantially over the marital life course (Choi and Denice 2023; Gonalons-Pons and Schwartz 2017), highlighting the importance of adopting a life course approach. This study extends our understanding of economic homogamy by examining the association between husbands’ and wives’ entire marital earnings trajectories, rather than spouses’ earnings at a particular point in time, providing novel insight into the varied ways gender earnings equality (or inequality) manifests within couples over time and how wives’ absolute and relative socio-economic status at marriage predict the earnings configurations that couples ultimately follow.
Using Dual Group Based Trajectory Modeling, I identify six key long-term earnings patterns for wives and four for husbands, reflecting prior research that finds greater variation in wives’ earnings in marriage than husbands’ (e.g., Killewald and Garcia-Manglano 2016; Qian 2018). Despite the dominant focus of existing scholarship on explaining declines in wives’ earnings following marriage (e.g., Killewald and Gough 2013; Cheng 2016), I find that the most common long-term earnings pattern for both husbands and wives was characterized by consistent earnings over the marital life course. 55% of wives and 77% of husbands followed stable earnings patterns over time, with the rest following declining, increasing or low earnings patterns. These results reveal greater stability in wives’ marital earnings than has been emphasized in prior research. In demonstrating that nearly a quarter of husbands do not follow stable earnings patterns in marriage, this study also reveals greater heterogeneity in husbands’ earnings patterns in marriage than is often assumed. These findings are consistent with recent scholarship showing that a majority of women in the U.S. maintain steady, full-time working hours over the life course (Damaske and Frech 2016) while a sizable minority of men maintain intermittent labor force attachment (Frech et al 2023), and demonstrate the value of group-based approaches in uncovering patterns that are often obscured by more common variable-centered approaches that focus on population averages.
Examining how husbands’ and wives’ earnings trajectories combine within couples, I find a high degree of gender equality in earnings among a majority of couples. But I find that long-term gender equality in earnings takes a variety of forms among couples. 50% of couples followed a Dual earner long term earnings pattern over time where both spouses’ earnings were stable over the marital life course and 6% followed a Jointly mobile pattern, where spouses’ earnings varied in similar ways over time. A further 5% of couples followed an Alternating earner pattern where spouses alternated primary earnership over time, representing an alternative form of long-term gender equality within couples that is unobservable in cross-sectional estimates of earnings homogamy. These findings extend previous research on wives’ relative earnings trajectories over the life course (e.g., Choi and Denice 2023; Qian 2017) by revealing two distinct processes through which long-term gender egalitarian earnings patterns may be produced: joint earnings stability and joint earnings instability. Finally, I find that a large minority of couples followed patterns of long-term economic specialization, with 34% following a Male breadwinner earnings configuration and a much smaller proportion (5%) following a Female breadwinner pattern.
Consistent with prior research (e.g., Qian 2017; 2018), multivariate analysis indicates that patterns of assortative mating have lasting consequences for wives’ long-term economic standing within couples, with wives’ status at marriage relative to their husbands’ (in terms of earnings, education and labor supply) positively associated with their economic power within couples in the long run. However, this analysis also reveals sharp differences in how gender equality in earnings manifests within couples over time by couples’ absolute socio-economic status. I find that the couples most likely to follow Dual earner patterns – and therefore the most likely to have the benefit of two stable incomes over the marital life course – are those in which spouses were already more socio-economically advantaged at the onset of marriage. Couples with lower socio-economic status at marriage, by contrast, were more likely to follow Jointly mobile or Alternating earner patterns, which were characterized by less earnings stability over time.
This concentration of earnings stability and earnings instability within couples constitutes an important dimension of economic homogamy that has been overlooked in previous, cross-sectional studies, which may have implications for economic inequalities between married couple households. In addition to benefiting from two stable incomes over the course of marriage, Dual earner couples may also be more likely than Jointly mobile or Alternating earner couples to have access to employment benefits that accrue with job tenure, such as private retirement contributions, which may result in differential exposure to economic insecurity in retirement. Furthermore, because childhood exposure to income shocks and volatility are associated with children’s lower educational achievement (Hardy 2014), the concentration of earnings stability and instability within couples may also be consequential for inequalities in children’s life chances.
The results of this study also have important implications for our understanding of gender equality within different-sex couples and the stratification of gender equality. Scholarship has long debated to what extent gender equality within different-sex couples is a privilege largely reserved for highly-educated, high-socio-economic status couples (e.g., Esping-Andersen and Billari 2015; Goldscheider et al. 2015; Usdansky 2011). By taking a long-term approach to examining the association between spouses’ earnings, I show that long-term gender equality in earnings is not necessarily reserved for high-SES couples, but rather that gender egalitarian earnings patterns look qualitatively different according to couples’ socio-economic status. Furthermore, the results of this study suggest that cross-sectional studies may particularly underestimate gender equality in earnings among lower-socio-economic status couples, who are more likely to follow alternative gender egalitarian arrangements, like Alternating earner patterns, that are invisible in cross-sectional approaches.
This study is limited in several ways, raising important questions for future research. One key limitation of the Group-Based Trajectory Modelling approach used in the study is that it is an inherently descriptive approach that is ill-suited to addressing causal questions about why couples follow the long-term earnings configurations that they do. One reason lower-socio-economic status couples may be more likely to follow alternative, less stable gender egalitarian earnings patterns may be because lower socio-economic status husbands and wives are each, individually more exposed to economic precarity and earnings instability (e.g., Kalleberg 2011; Western et al. 2012). However, these earnings configurations may also reflect spouses’ responses to the economic instability of the other (DiPrete 2002). For example, Alternating earner patterns may arise when one spouse enters paid employment to compensate for the other’s job loss or health decline. Or, alternative gender egalitarian earnings configurations, like Alternating earner patterns, may constitute specific work-family strategies that lower-socio-economic status couples employ in the face of substantial structural constraints to balancing work and family in a liberal welfare state context with little social support for working families (Craig et al 2016; Landivar 2017). Testing these possible explanations is an important avenue for future research.
Second, as discussed above, my analysis relies on longitudinal data from a cohort of late Baby Boomer men and women (born between 1957 and 1964) and their spouses, who came of age and formed families at a time when women’s economic opportunities were expanding and men’s were contracting in the U.S. (Goldin 2021; Ruggles 2015). The results of this study, therefore, may not reflect the experiences of younger generations. On one hand, liberalizing gender norms, later ages at first marriage, and delays and declines in fertility may increase the incidence of female breadwinner and gender egalitarian earnings constellations among younger cohorts. On the other, growing labor market polarization and economic precarity (e.g., Autor 2014; Kalleberg 2011) may make Dual earner patterns rarer and more selected among younger cohorts than among the NLSY79 cohort. Future research examining data across birth cohorts would help shed light on whether and how spouses’ joint earnings patterns and their social patterning have changed over time.
Despite these limitations, this study offers novel insight into the long-term association between husbands and wives’ earnings in the U.S., revealing the variety of ways gender equality in earnings is realized over the course of marriage and the socio-economic stratification of gender equality in earnings among different-sex couples. By taking a long-term approach, this study also identifies an important and understudied dimension of economic homogamy: the concentration of economic stability and instability within couples. Together, the results from this study emphasize that the consolidation of economic advantage among high-socio-economic couples is likely to endure and solidify over the marital life course.
Acknowledgments
I am grateful to Xi Song, Pilar Gonalons-Pons, Jerry A. Jacobs and Rebecca Anna Schut for their helpful comments on earlier drafts of the paper. Earlier versions of this paper were presented at the 2022 Work Family Researchers Network Conference, 2023 Annual Meeting of the Population Association of America and the Family & Gender workshop in the Department of Sociology at the University of Pennsylvania. I am grateful to the participants for their feedback and suggestions.
Appendices for His and Hers Earnings Trajectories
Appendix A: Additional Results
Appendix A Table 1 shows the fit statistics for a series of nested Group-Based Trajectory models varying the number of latent trajectory groups included in each model. For husbands and wives separately, I tested models ranging from one latent trajectory group to eight groups. Given the dynamic nature of earnings over the life course, particularly for women, all initial models assumed that the log of husbands’ wives’ earnings’ is a cubic function of years since marriage (as shown in parentheses).
Greater BIC and AIC statistics indicate better fit to the data. However, in the model selection process, this must be balanced against interpretability of results (Nagin 2005). BIC and AIC statistics for both husbands’ and wives’ models indicated that models with greater numbers of groups fit the data best. However, the additional groups generated by models with more than six groups for wives and more than four groups for husbands duplicated existing trajectory groups, indicating that the additional complexity added by increasing the number of groups in the model did not provide additional substantive insight. Furthermore, models with more than six groups for wives and more than four groups for husbands generated sparse trajectory groups comprising less than 5% of observations.
For example, the results for models with seven trajectory groups for wives and five groups for husbands are shown in Appendix A Figure 1. For wives, the additional trajectory group generated in the seven-group model closely resembles the Early exit group, but with a steeper decline in earnings that occurs earlier in marriage than in the Early exit group. For husbands, the additional trajectory group generated in the five-group model closely resembles the Decreasing group, but with earlier declines in earnings. For both husbands and wives, these additional trajectory groups comprise 4% of observations, failing to meet the 5% threshold recommended in Nagin (2005).
After selecting the six-group model for wives and the four-group model for husbands, I tested models varying the polynomial function of marital duration for each group based on the statistical significance of the parameter estimates, ultimately reducing a cubic function to a quadratic function for one group of husbands and one group of wives.
Appendix A Table 2 presents the mean and median posterior probability of group membership for each trajectory group and a measure of relative entropy, calculated from my selected model. Posterior probabilities measure a specific wife’s (husband’s) likelihood of belonging to each of the model’s groups. GBTM uses a maximum-probability assignment rule to categorize individuals into the group in which their posterior membership probability is largest. Therefore, examining the average posterior probabilities of individuals assigned to each group provides an indication of how well the model corresponds to the data. The results from the dual model exceed Nagin’s (2005) recommendation that each group’s average posterior probability should exceed 0.70, with relative entropy exceeding 0.80, suggesting the model is a good fit to the data.
Appendix A Table 1.
Goodness of Fit Statistics for Nested Univariate Group Based Trajectory Models for Husbands and Wives
| Variables and no. classes | BIC (N= Person waves) | BIC (N= Couples) | AIC |
|---|---|---|---|
| Wives’ earnings (ln) | |||
| 1 (3) | −176886.9 | −176880.6 | −176864.2 |
| 2 (33) | −165104.5 | −165091.9 | −165059.0 |
| 3 (333) | −162482.5 | −162463.6 | −162414.2 |
| 4 (3333) | −159780.3 | −159755.1 | −159689.2 |
| 5 (33333) | −158218.0 | −158186.5 | −158104.2 |
| 6 (333333) | −157254.5 | −157216.7 | −157117.9 |
| 7 (3333333) | −156676.5 | −156632.4 | −156517.2 |
| 8 (33333333) | −156023.4 | −155973.0 | −155841.3 |
| 6 (332333) | −157450.9 | −157414.4 | −157318.9 |
| Husbands’ earnings (ln) | |||
| 1 (3) | −172044.6 | −172038.3 | −172021.9 |
| 2 (33) | −157909.0 | −157896.5 | −157863.6 |
| 3 (333) | −154050.8 | −154032.0 | −153982.7 |
| 4 (3333) | −155391.2 | −155366.2 | −155300.3 |
| 5 (33333) | −149590.4 | −149559.1 | −149476.8 |
| 6 (333333) | −148611.5 | −148573.9 | −148475.1 |
| 7 (3333333) | −148063.6 | −148019.8 | −147904.5 |
| 8 (33333333) | −147029.3 | −146979.2 | −146847.5 |
| 4 (2333) | −151117.9 | −151094.1 | −151031.6 |
Notes: The number of latent trajectory groups included in each model is indicated in each row; The numbers in parentheses indicate the functional form of each trajectory group; Final model selected is shown in bold. BIC = Bayesian information criterion; AIC = Akaike information criterion.
Appendix A Figure 1.
Wives’ and Husbands’ Estimated Earnings Trajectory Groups from Models with 7 Groups for Wives and 5 groups for Husbands
Notes: N=5,354; This graph shows the estimated earnings trajectories generated from Group-Based Trajectory Models with seven groups for wives and five groups for husbands. Earnings are modeled as a cubic function of years since marriage for all groups. Numbers in parentheses refer to the percentage of husbands or wives in each trajectory group. The thickness of trajectory lines is weighted according to the prevalence of each trajectory group in the sample. Source: NLSY79 1979–2018.
Appendix A Table 2.
Group Membership Probabilities for Final Dual Group Based Trajectory Model
| Group | Proportion | Mean group PProb | Median group PProb | Relative Entropy |
|---|---|---|---|---|
| Wives | 0.861 | |||
| 1 - Stable earner | 55.1 | 0.95 | 1.00 | |
| 2 - Career pause | 11.6 | 0.89 | 0.97 | |
| 3 - Career climber | 9.1 | 0.89 | 0.97 | |
| 4 - Low earner | 9.0 | 0.91 | 0.99 | |
| 5 - Late, decreasing | 8.5 | 0.83 | 0.90 | |
| 6 - Early exit | 6.7 | 0.90 | 0.99 | |
| Husbands | ||||
| 1 - Stable earner | 77.4 | 0.98 | 1.00 | |
| 2 - Decreasing | 8.7 | 0.93 | 1.00 | |
| 3 - Late increasing | 7.0 | 0.98 | 1.00 | |
| 4 - Career climber | 6.9 | 0.95 | 1.00 |
Notes: PProb = posterior probabilities of group membership. This is the probability of group membership in group j (k) for wives (husbands) assigned to group j (k). A posterior probability of 1.0 indicates zero classification error in group assignment. Therefore, greater mean and median posterior probabilities indicate a better fit of the model to the data. Relative entropy indicates the degree of classification accuracy of placing participants into a trajectory based on their posterior probabilities. As discussed in Nagin (2005), mean posterior probabilities exceeding 0.7 and relative entropies exceeding 0.8 indicate model adequacy.
Appendix A Table 3.
Dual Group Based Trajectory Model Coefficients
| Coef. | SE | ||
|---|---|---|---|
| Wives | |||
| Group 1: Stable earner | Intercept | 9.566** | 0.056 |
| Linear | −0.029 | 0.020 | |
| Quadratic | 0.006** | 0.002 | |
| Cubic | 0.000** | 0.000 | |
| Group 2: Career pause | Intercept | 9.477** | 0.153 |
| Linear | −2.663** | 0.060 | |
| Quadratic | 0.224** | 0.005 | |
| Cubic | −0.005** | 0.000 | |
| Group 3: Career climber | Intercept | −3.009** | 0.435 |
| Linear | 1.135** | 0.041 | |
| Quadratic | −0.026** | 0.001 | |
| Group 4: Low earner | Intercept | 1.486** | 0.461 |
| Linear | −1.754** | 0.130 | |
| Quadratic | 0.139** | 0.010 | |
| Cubic | −0.003** | 0.000 | |
| Group 5: Late decreasing | Intercept | 4.230** | 0.299 |
| Linear | 0.894** | 0.083 | |
| Quadratic | −0.036** | 0.009 | |
| Cubic | 0.000 | 0.000 | |
| Group 6: Early exit | Intercept | 10.324** | 0.167 |
| Linear | −0.694** | 0.082 | |
| Quadratic | −0.033** | 0.007 | |
| Cubic | 0.002** | 0.000 | |
| Sigma | 3.776 | 0.013 | |
| Husbands | |||
| Group 1: Stable earner | Intercept | 10.141** | 0.026 |
| Linear | 0.090** | 0.005 | |
| Quadratic | −0.003** | 0.000 | |
| Group 2: Decreasing | Intercept | 9.577** | 0.110 |
| Linear | −0.395** | 0.042 | |
| Quadratic | 0.013** | 0.004 | |
| Cubic | 0.000** | 0.000 | |
| Group 3: Late increasing | Intercept | 1.599** | 0.145 |
| Linear | −1.628** | 0.054 | |
| Quadratic | 0.160** | 0.005 | |
| Cubic | −0.003** | 0.000 | |
| Group 4: Career climber | Intercept | −0.813** | 0.174 |
| Linear | 1.837** | 0.058 | |
| Quadratic | −0.095** | 0.005 | |
| Cubic | 0.002** | 0.000 | |
| Sigma | 2.648 | 0.008 | |
| Person waves | 131838 | ||
| Couples | 5354 | ||
| BIC (person-years) | −308226.88 |
Source: NLSY79 1979–2018; N=5,354; Notes: Sigma is the standard deviation of the normally distributed residual term.
p < 0.05,
p < 0.01; two-tailed tests.
Appendix A Figure 2.
Wives’ and Husbands’ Estimated Earnings Trajectory Groups, with 95% Confidence Intervals
Source: NLSY79 1979–2018; N=5,354. Notes: This graph shows the estimated earnings trajectories generated from a Dual Group-Based Trajectory Model with six groups for wives and four groups for husbands. Earnings are modeled as a cubic function of years since marriage for all groups except for the “Career climber” group for wives and the “Stable earner” group for husbands where earnings are modeled as a quadratic function of years since marriage. Each husband and wife in the sample was assigned to the trajectory group for which they had the highest estimated probability of belonging. Numbers in parentheses refer to the percentage of husbands or wives in each trajectory group.
Appendix A Figure 3.
Wives’ Estimated and Observed Earnings by Trajectory Group, random sample of 20 wives per group
Source: NLSY79 1979–2018. Notes: Colored lines show the observed earnings trajectories for a random sample of 20 wives in each trajectory group. Black lines show the estimated earnings trajectories from a Dual Group-Based Trajectory Model using the full sample of 5,354 couples.
Appendix A Figure 4.
Husbands’ Estimated and Observed Earnings by Trajectory Group, random sample of 20 husbands per group
Source: NLSY79 1979–2018. Notes: Colored lines show the observed earnings trajectories for a random sample of 20 husbands in each trajectory group. Black lines show the estimated earnings trajectories from a Dual Group-Based Trajectory Model using the full sample of 5,354 couples.
Appendix A Table 4.
Additional Descriptive Statistics by Trajectory Group
| Mean | SD | Median | Min | Max | |
|---|---|---|---|---|---|
| Wives | |||||
| Stable earner (N=2952) | |||||
| Mean annual earnings ($) | 35883.61 | 26288.76 | 29847.37 | 1229.59 | 354329.30 |
| Prop. waves with zero earnings | 0.04 | 0.08 | 0.00 | 0.00 | 0.50 |
| Career pause (N=619) | |||||
| Mean annual earnings ($) | 12212.04 | 9288.34 | 10064.12 | 225.23 | 73400.43 |
| Prop. waves with zero earnings | 0.38 | 0.13 | 0.37 | 0.13 | 0.79 |
| Career climber (N=485) | |||||
| Mean annual earnings ($) | 7934.28 | 12818.01 | 4913.28 | 0.00 | 221396.70 |
| Prop. waves with zero earnings | 0.59 | 0.22 | 0.58 | 0.13 | 1.00 |
| Low earner (N=481) | |||||
| Mean annual earnings ($) | 1883.49 | 3201.01 | 670.70 | 0.00 | 25680.90 |
| Prop. waves with zero earnings | 0.82 | 0.16 | 0.83 | 0.33 | 1.00 |
| Late decreasing (N=456) | |||||
| Mean annual earnings ($) | 14651.76 | 13969.67 | 10953.20 | 343.09 | 124486.90 |
| Prop. waves with zero earnings | 0.32 | 0.15 | 0.30 | 0.06 | 0.76 |
| Early exit (N=361) | |||||
| Mean annual earnings ($) | 14931.61 | 10822.07 | 11778.85 | 1074.54 | 79592.32 |
| Prop. waves with zero earnings | 0.40 | 0.17 | 0.38 | 0.08 | 0.79 |
| Total (N=5354) | |||||
| Mean annual earnings ($) | 24339.42 | 24603.79 | 18742.56 | 0.00 | 354329.30 |
| Prop. waves with zero earnings | 0.25 | 0.29 | 0.14 | 0.00 | 1.00 |
| Husbands | |||||
| Stable earner (N=4142) | |||||
| Mean annual earnings ($) | 58659.05 | 41574.37 | 47814.60 | 3136.03 | 365313.50 |
| Prop. waves with zero earnings | 0.03 | 0.06 | 0.00 | 0.00 | 0.50 |
| Decreasing (N=467) | |||||
| Mean annual earnings ($) | 25935.81 | 19131.66 | 21013.69 | 0.00 | 136241.80 |
| Prop. waves with zero earnings | 0.27 | 0.14 | 0.24 | 0.00 | 1.00 |
| Late increasing (N=374) | |||||
| Mean annual earnings ($) | 7327.59 | 12552.25 | 1661.38 | 0.00 | 110840.90 |
| Prop. waves with zero earnings | 0.79 | 0.19 | 0.80 | 0.25 | 1.00 |
| Career climber (N=371) | |||||
| Mean annual earnings ($) | 28299.40 | 33787.81 | 20199.68 | 83.57 | 276366.80 |
| Prop. waves with zero earnings | 0.43 | 0.20 | 0.41 | 0.06 | 0.89 |
| Total (N=5354) | |||||
| Mean annual earnings ($) | 50115.32 | 41559.86 | 41752.71 | 0.00 | 365313.50 |
| Prop. waves with zero earnings | 0.13 | 0.24 | 0.00 | 0.00 | 1.00 |
Source: NLSY79 1979–2018; N=5,354.
Appendix A Table 5.
Multinomial Logistic Regression Results – Couple Types
| (Reference = Dual earner) | ||||
|---|---|---|---|---|
| Jointly mobile | Alternating earner | Female breadwinner | Male breadwinner | |
| Wife’s earnings tertile at marriage (ref = low) | ||||
| Medium | −2.263** (0.186) |
−1.997** (0.176) |
−0.003 (0.195) |
−1.341** (0.092) |
| High | −3.116** (0.284) |
−3.028** (0.283) |
−0.269 (0.243) |
−1.626** (0.125) |
| Wife’s education at marriage (ref = less than HS) | ||||
| High School | −0.090 (0.203) |
−0.256 (0.202) |
0.115 (0.270) |
−0.357** (0.119) |
| Some college | −0.691** (0.251) |
−0.906** (0.249) |
−0.013 (0.289) |
−0.762** (0.135) |
| College or higher | −1.316** (0.365) |
−1.452** (0.336) |
−0.528 (0.329) |
−0.875** (0.158) |
| Wife’s work hours at marriage (ref = less than FT) | ||||
| FT | −0.531* (0.225) |
−0.950** (0.223) |
−0.712** (0.179) |
−0.440** (0.100) |
| More than FT | −1.102* (0.511) |
−0.900* (0.385) |
−1.255** (0.306) |
−0.674** (0.182) |
| Relative earnings at marriage (ref = wife earns < 40% of couple earnings) | ||||
| Wife earns 40–60% | 1.540** (0.165) |
0.988** (0.177) |
0.330 (0.172) |
−0.106 (0.087) |
| Wife earns > 60% | 1.775** (0.231) |
2.023** (0.212) |
0.945** (0.197) |
−0.336* (0.131) |
| Relative education at marriage (ref = wife less educated) | ||||
| Spouses equally educated | 0.083 (0.173) |
0.374* (0.179) |
0.133 (0.181) |
−0.057 (0.090) |
| Wife more educated | −0.130 (0.226) |
0.172 (0.230) |
0.457* (0.202) |
−0.322** (0.113) |
| Relative work status at marriage (ref = wife worked less) | ||||
| Spouses worked similar hours | −0.024 (0.161) |
0.174 (0.168) |
0.193 (0.169) |
−0.174* (0.086) |
| Wife worked more | −0.258 (0.275) |
0.488* (0.246) |
0.885** (0.223) |
0.007 (0.143) |
| Number of children born in marriage | 0.164** (0.062) |
0.268** (0.062) |
0.048 (0.063) |
0.491** (0.033) |
| Wife’s age at marriage | 0.025 (0.018) |
0.071** (0.017) |
0.067** (0.014) |
0.059** (0.009) |
| Age gap (wife’s birth year - husband’s) | 0.016 (0.019) |
0.048** (0.018) |
0.084** (0.015) |
0.022* (0.010) |
| Wife was the respondent | 0.070 (0.141) |
−0.033 (0.141) |
0.147 (0.134) |
−0.209** (0.071) |
| Respondent race (ref = White) | ||||
| Hispanic | −0.019 (0.204) |
0.024 (0.191) |
0.226 (0.190) |
−0.113 (0.098) |
| Black | 0.627** (0.157) |
0.168 (0.169) |
0.616** (0.153) |
−0.348** (0.099) |
| Constant | −2.032** (0.490) |
−3.273** (0.482) |
−4.617** (0.486) |
−0.576* (0.255) |
Source: NLSY79 1979–2018; N=5,308;
Notes:
p < 0.05,
p < 0.01; two-tailed tests.
Appendix B: Supplementary Analysis – Predicting Husbands’ and Wives’ Trajectory Groups
Appendix B Tables 1 and 2 present the results from multinomial logistic regression models examining the individual, relative and family characteristics associated with wives’ and husbands’ earnings group membership. To aid interpretation, all results are presented as average marginal effects of individual, spouse and family characteristics on the predicted probability of earnings group membership.
Characteristics Associated with Wives’ Trajectory Groups
The results for wives, shown, in Appendix B Table 1, highlight contrasting distinguishing characteristics of membership in wives’ earnings trajectory groups. Although wives in the Stable earner, Career pause and Early exit groups began marriage with relatively high earnings compared with other wives, the probability that wives in those groups maintained their earnings was higher the greater wives’ absolute status at marriage (in terms of education and work hours), the higher their relative education at marriage compared with their husbands’ and the fewer children they ultimately had. Wives with a bachelor’s degree or higher, for example, were 18 percentage points more likely to belong to the Stable earner group than wives with less than a high school degree. Similarly, wives who were more highly educated their husband at marriage were 6 percentage points more likely to follow a Stable earner pattern. And each child born during marriage was associated with an 8-percentage point decrease in the likelihood that wives follow the Stable earner trajectory. By contrast, the probability that wives’ earnings declined sharply in the years following marriage (as observed among the Career pause and Early Exit groups) was higher the greater the number of children born in marriage and the lower wives’ relative earnings and education at marriage compared to their husbands.
The probability of membership in the Career climber, Low earner and Late decreasing groups was associated with wives’ lower absolute socio-economic status at marriage and, perhaps surprisingly, by wives’ equal or greater relative earnings and education compared with their husbands. However, the number of children born over the marital life course differed in its association with each earnings group: greater marital fertility was associated with greater likelihood of membership in the Low earner group but negatively associated with membership in the Late decreasing group.
Characteristics Associated with Husbands’ Trajectory Groups
The results for husbands (Appendix B Table 2) show that the probability of membership in the Stable earner group is characterized by husbands’ greater absolute socio-economic status (in terms of earnings, education and labor supply) at marriage. For example, husbands with a BA or higher at marriage were 11 percentage points more likely to follow Stable earnings patterns than those with less than a high school degree. However, the probability of following a Stable earner pattern was ambiguously associated with spouses’ relative status at marriage: the likelihood of membership in the Stable earner group was positively associated with wives’ relative education, contrasting previous research that finds little connection between educational assortative mating and husbands’ subsequent earnings (e.g., Qian 2018), but negatively associated with wives’ relative earnings at marriage.
In contrast to the Stable earner group, the probability of membership in the Decreasing group was strongly associated with husbands’ lower absolute education and labor supply at marriage. However, it was also associated with wives’ lower relative earnings at marriage.
Husbands in the Late increasing and Career climber groups started marriage with low earnings but experienced substantial earnings increases thereafter. Unsurprisingly, the probabilities of membership in these groups were lower the greater husbands’ earnings and labor supply at marriage. However, the probabilities of following each of these earnings patterns was greater the greater wives’ relative earnings at marriage.
Appendix B Table 1.
Predicted Probability of Wives’ Earnings Group Membership – Average Marginal Effects
| Stable earner | Career pause | Career climber | Low earner | Late decreasing | Early exit | |
|---|---|---|---|---|---|---|
| Wife’s earnings tertile at marriage (ref = low) | ||||||
| Medium | 0.316* | 0.047* | −0.181* | −0.126* | −0.098* | 0.042* |
| High | 0.396* | 0.021 | −0.198* | −0.151* | −0.118* | 0.050* |
| Wife’s education at marriage (ref = less than HS) | ||||||
| High School | 0.054* | 0.001 | −0.021 | −0.040* | 0.008 | −0.002 |
| Some college | 0.137* | −0.025 | −0.044* | −0.073* | −0.004 | 0.008 |
| BA or higher | 0.175* | −0.011 | −0.080* | −0.080* | −0.018 | 0.014 |
| Wife’s work hours at marriage (ref = less than FT) | ||||||
| FT | 0.101* | 0.008 | −0.018 | −0.053* | −0.030* | −0.008 |
| More than FT | 0.121* | −0.015 | 0.008 | −0.037 | −0.046* | −0.031* |
| Relative earnings at marriage (ref = wife earns < 40% of couple earnings) | ||||||
| Wife earns 40–60% | −0.034* | −0.027* | 0.049* | 0.035* | 0.005 | −0.028* |
| Wife earns > 60% | −0.021 | 0.014 | −0.019 | 0.035* | 0.038* | −0.047* |
| Relative education at marriage (ref = wife less educated) | ||||||
| Spouses equally educated | 0.006 | −0.025* | 0.010 | 0.005 | 0.021* | −0.017 |
| Wife more educated | 0.056* | −0.041* | 0.016 | 0.021 | −0.008 | −0.044* |
| Relative work hours at marriage (ref = wife worked less) | ||||||
| Spouses worked similar hours | 0.028 | 0.012 | −0.041* | −0.023* | 0.006 | 0.018* |
| Wife worked more | 0.006 | −0.019 | −0.031 | −0.003 | 0.011 | 0.038* |
| Number of children born in marriage | −0.077* | 0.042* | 0.004 | 0.024* | −0.013* | 0.021* |
| Wife’s age at marriage | −0.008* | −0.002 | 0.002 | 0.006* | 0.000 | 0.003* |
| Age difference (Wife’s birth year - husband’s) | −0.002 | −0.002 | 0.001 | 0.003* | 0.000 | −0.001 |
| Base rate | 0.551 | 0.116 | 0.091 | 0.090 | 0.085 | 0.067 |
Source: NLSY79 1979–2018; N=5,308; Notes: asterisks indicate that differences in predicted probabilities from the reference category are statistically significant at p < 0.05. Model controls for the race and gender of the NLSY respondent.
Appendix B Table 2.
Predicted Probability of Husband’s Earnings Group Membership – Average Marginal Effects
| Stable earner | Decreasing | Late increasing | Career climber | |
|---|---|---|---|---|
| Husband’s earnings tertile at marriage (ref = low) | ||||
| Medium | 0.237 * | −0.018 | −0.108 * | −0.111 * |
| High | 0.276 * | −0.038 * | −0.121 * | −0.118 * |
| Husband’s education at marriage (ref = less than HS) | ||||
| High School | 0.001 | −0.055 * | 0.039 * | 0.015 |
| Some college | 0.061 * | −0.070 * | 0.012 | −0.003 |
| BA or higher | 0.105 * | −0.096 * | 0.000 | −0.009 |
| Husband’s work hours at marriage (ref = less than FT) | ||||
| FT | 0.097 * | −0.018 | −0.036 * | −0.043 * |
| More than FT | 0.101 * | −0.036 * | −0.034 * | −0.031 * |
| Relative earnings at marriage (ref = wife earns < 40% of couple earnings) | ||||
| Wife earns 40–60% | −0.087 * | −0.032 * | 0.060 * | 0.059 * |
| Wife earns > 60% | −0.108 * | −0.046 * | 0.060 * | 0.094 * |
| Relative education at marriage (ref = wife less educated) | ||||
| Spouses equally educated | 0.030 * | −0.018 | −0.008 | −0.004 |
| Wife more educated | 0.041 * | −0.019 | −0.011 | −0.011 |
| Relative work hours at marriage (ref = wife worked less) | ||||
| Spouses worked similar hours | 0.076 * | −0.009 | −0.034 * | −0.033 * |
| Wife worked more | 0.079 * | 0.006 | −0.040 * | −0.045 * |
| Number of children born in marriage | 0.004 | 0.000 | −0.005 | 0.001 |
| Husband’s age at marriage | −0.004 * | 0.005 * | 0.000 | −0.001 |
| Age difference (Wife’s birth year - husband’s) | −0.006 * | 0.007 * | 0.000 | −0.001 |
| Base rate | 0.774 | 0.087 | 0.070 | 0.069 |
Source: NLSY79 1979–2018; N=5,308; Notes: asterisks indicate that differences in predicted probabilities from the reference category are statistically significant at p < 0.05. Model controls for the race and gender of the NLSY respondent.
Appendix C: Robustness Analysis – Divorce and Data Availability Comparisons
Marriages in my sample are disproportionately long-lasting, in part due to the requirement of earnings data for both spouses for at least four waves of data. 67% of marriages in my sample had not ended by the couples’ most recently available wave of data. 26% of marriages ended in divorce and the remaining 6% ended with the death of a spouse or for unknown reasons. Because husbands’ and wives’ earnings patterns within marriage are related to the risk of divorce (Kalmijn and Poortman 2006; Kalmijn, Loeve, & Manting, 2007) and because couples in this analysis are only observed while they are married, my main Group-Based Trajectory Model results may be driven primarily by couples with longer, enduring marriages. To examine whether my results are biased by sample selection and differential attrition from the sample due to divorce, I reran my Group Based Trajectory Models on a subsample of couples whose marriage ended in divorce and a subsample of couples whose marriage had not ended in the latest wave of data available. I truncated this analysis at 15 years since marriage because the large majority of couples whose marriages ended in divorce had marriages lasting fewer than 15 years.
The results from this sensitivity analysis are shown in Appendix C Figure 1. For both subsamples, I selected a model with 5 trajectory groups for wives and 4 trajectory groups for husbands based on goodness of fit to the data and interpretability of results (goodness of fit statistics for all models tested are available upon request). Comparing the results for both subsamples to my main results, a key difference is that I identify one fewer trajectory group for wives in the divorced and enduring marriage subsamples. However, this difference appears to be driven by the shorter window of analysis rather than substantive differences by marital outcome. Because this sensitivity analysis is truncated at 15 years since marriage, the Group-Based Trajectory Model cannot distinguish between the “Career climber” group and the “Late decreasing” group, which both demonstrated earnings growth in the first 15 years of marriage in my main results.
Beyond the difference in the total number of trajectory groups identified for wives, the results for both subsamples broadly reflect the patterns identified in my main results in terms of the shape and prevalence of trajectory groups. Analyzing how husbands’ and wives’ earnings trajectories combine within couples, in Appendix C Figure 2, I find similar results across divorcing and enduring couples.
Appendix C Figure 1.
Wives’ and Husbands’ Estimated Earnings Trajectory Groups, by Marital Outcome
Source: NLSY79 1979–2018; N=1,397 divorcing couples, 3,542 enduring couples. Notes: Full regression coefficients and goodness of fit statistics are available upon request.
Appendix C Figure 2.
Couple Types based on Estimated Earnings Trajectories of Husbands and Wives, by Marital Outcome
Source: NLSY79 1979–2018; N=1,397 divorcing couples, 3,542 enduring couples.
On average, couples included in my analytic sample are followed for 16 waves of data. However, there is considerable variation around this average, as shown in Appendix B Table 1, with the number of waves available for each couple ranging from 4 to 24. In a second sensitivity analysis, I considered whether my results are biased by attrition from the sample more generally (regardless of whether attrition is due to divorce) by rerunning my Group Based Trajectory Models on a subsample of couples with fewer than 15 waves of data available (34% of couples) and a subsample of couples with 15 waves or greater. To ensure comparability across subgroups, I truncated this analysis at 14 years since marriage in both subsamples.
Appendix C Table 1.
Number of Waves Available for Sample Couples.
| N waves observed | N | % |
|---|---|---|
| 4 | 14 | 0.26 |
| 5 | 87 | 1.62 |
| 6 | 161 | 3.01 |
| 7 | 186 | 3.47 |
| 8 | 168 | 3.14 |
| 9 | 213 | 3.98 |
| 10 | 176 | 3.29 |
| 11 | 229 | 4.28 |
| 12 | 234 | 4.37 |
| 13 | 188 | 3.51 |
| 14 | 179 | 3.34 |
| 15 | 213 | 3.98 |
| 16 | 205 | 3.83 |
| 17 | 208 | 3.88 |
| 18 | 262 | 4.89 |
| 19 | 523 | 9.77 |
| 20 | 574 | 10.72 |
| 21 | 498 | 9.3 |
| 22 | 523 | 9.77 |
| 23 | 508 | 9.49 |
| 24 | 5 | 0.09 |
| Total | 5354 | 100 |
Source: NLSY79 1979–2018.
The results from this sensitivity analysis are shown in Appendix C Figure 3. For both subsamples, I selected a model with 5 trajectory groups for wives and 4 trajectory groups for husbands based on goodness of fit to the data and interpretability of results (goodness of fit statistics for all models tested are available upon request). As in the sensitivity analysis for divorced and enduring marriages, the results of this analysis identify one fewer trajectory group for wives in both subsamples. Again, this appears to be driven by the shorter window of analysis with the “Career climber” and “Late decreasing” groups combined in one trajectory group.
The results from this sensitivity analysis reveal some minor differences across subsamples. The results for wives show that declines in earnings among the Early exit and Career pause groups occur later in the period of analysis for couples with 15 waves of data or more than for those with less than 15 waves of data. Similarly, earnings increases among Career climber wives occur later among couples with more waves of data than among those with fewer. Second, I find that Stable earner and Early exit patterns were more common among wives with fewer waves of data while Career climber/Late decreasing and Low earner patters were more common among wives with 15 waves of data or more. And finally, among husbands, I find that earnings decreases among husbands in the Decreasing group are less severe (at least as they are observed in 14 waves of data) for husbands with 15 years of data or more than they are for those with fewer than 15 waves. Despite these differences, the results for these subsamples broadly reflect the shape and prevalence of trajectory groups shown in my main results and demonstrate that husbands’ and wives’ earnings patterns do not differ substantially by the duration of data available.
Examining how husbands’ and wives’ earnings trajectories combine within couples, in Appendix C Figure 4, I find similar results across data availability subsamples.
Appendix C Figure 3.
Wives’ and Husbands’ Estimated Earnings Trajectory Groups, by Number of Waves of Data Available
Source: NLSY79 1979–2018; N=1,835 couples with less than 15 waves of data, 3,499 couples with 15+ waves. Notes: Full regression coefficients and goodness of fit statistics are available upon request.
Appendix C Figure 4.
Couple Types based on Estimated Earnings Trajectories of Husbands and Wives, by Number of Waves of Data Available
Source: NLSY79 1979–2018; N=1,835 couples with less than 15 waves of data, 3,499 couples with 15+ waves.
Footnotes
Results are robust to an alternative inverse hyperbolic spline transformation (results available upon request).
Analysis using a listwise deletion approach show similar results, which are available upon request.
Trajectories with 95% Confidence Intervals are shown in Appendix A Figure 2.
Spaghetti plots comparing husbands’ and wives’ observed earnings trajectories to model-based predicted trajectories are available Appendix A Figures 3 and 4.
Although joint upward mobility and joint downward mobility are likely to be experienced in very different ways within couples, small sample sizes preclude examining these earnings patterns separately in the multivariate analysis.
In supplementary analysis, presented in Appendix B, I examined the individual and couple-level factors that distinguish husbands’ and wives’ earnings trajectory groups. I find that, for both husbands and wives, stable earnings patterns were characterized by spouses’ higher absolute socio-economic status at marriage. But, reflecting prior research on the gendered effects of parenthood (e.g., Killewald and Garcia-Manglano 2016; Musick et al. 2020), wives were less likely to follow a stable earnings pattern the more children they had. Meanwhile, although declining earnings patterns were associated with lower absolute status at marriage for husbands, for wives declining earnings patterns were primarily associated with wives’ lower relative status at marriage compared to their husbands.
References
- Autor DH (2014). Skills, education and the rise of earnings inequality among the “other 99 percent”. Science, 344(6186): 843–851. [DOI] [PubMed] [Google Scholar]
- Becker G (1981). A Treatise on the Family. Cambridge, MA: Harvard University Press. [Google Scholar]
- Becker G (1994). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education, 3d ed. Chicago: University of Chicago Press. [Google Scholar]
- Becker PE, and Moen P (1999). Scaling back: Dual-Earner Couples’ Work-Family Strategies. Journal of Marriage and the Family, 61(4): 995–1007. [Google Scholar]
- Blackburn ML and Bloom DE (1995). Changes in the Structure of Family Income Inequality in the United States and Other Industrial Nations During the 1980s. In Polachek SW (Ed.), Research in labor economics (pp. 141–170). Greenwich, CT: JAI Press. [Google Scholar]
- Bloome DE, Burk D and McCall L (2019). Economic Self-Reliance and Gender Inequality between U.S. Men and Women, 1970–2010. American Journal of Sociology, 124(5): 1413–1467. [Google Scholar]
- Burtless G (1999). Effects of Growing Wage Disparities and Changing Family Composition on the U.S. Income Distribution.” European Economic Review, 43: 853–65. [Google Scholar]
- Cancian M and Reed D (1999). The Impact of Wives’ Earnings on Income Inequality: Issues and Estimates. Demography, 36: 173–184. [PubMed] [Google Scholar]
- Cha Y (2010). Reinforcing separate spheres: The effect of spousal overwork on men’s and women’s employment in dual‐earner households. American Sociological Review, 75: 303–329. [Google Scholar]
- Cheng S (2015). Unequal Origins, Unequal Trajectories: Social Stratification over the Life Course. PhD dissertation, Department of Sociology, University of Michigan, Ann Arbor, MI. [Google Scholar]
- Cheng S (2016). The accumulation of (dis) advantage: The intersection of gender and race in the long-term wage effect of marriage. American Sociological Review, 81: 29–56. [Google Scholar]
- Cherlin A (1978). Remarriage as an incomplete institution. American Journal of Sociology, 84(3): 634–650. [Google Scholar]
- Choi K and Denice P (2023). Racial/Ethnic Variation in the Relationship Between Educational Assortative Mating and Wives’ Income Trajectories. Demography, 60(1): 227–254. [DOI] [PubMed] [Google Scholar]
- Chung H 2020. “Company-Level Family Policies: Who Has Access to It and What Are Some of Its Outcomes?” In: Nieuwenhuis R, Van Lancker W (eds) The Palgrave Handbook of Family Policy. Palgrave Macmillan. [Google Scholar]
- Craig L, Perales F, Vidal S and Baxter J 2016. “Domestic Outsourcing, Housework Time, and Subjective Time Pressure: New Insights from Longitudinal Data: Domestic Outsourcing, Housework, and Time Pressure.” Journal of Marriage and Family, 78 (5): 1224–36. [Google Scholar]
- Damaske S 2011. For the Family? How Class and Gender Shape Women’s Work. New York: Oxford University Press. [Google Scholar]
- Damaske S and Frech A 2016. “Women’s Work Pathways Across the Life Course.” Demography, 53: 365–391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Damaske S, Frech A and Wething H 2023. “The Life Course of Unemployment: the timing and Relative Degree of Risk.” Work and Occupations, 0(0). [Google Scholar]
- DiPrete TA 2002. “Life Course Risks, Mobility Regimes, and Mobility Consequences: A Comparison of Sweden, Germany, and the United States.” American Journal of Sociology, 108(2), 267–309. [Google Scholar]
- Elder G (1985). Life course dynamics: Trajectories and transitions, 1968–1980. Itaca: Cornell University Press. [Google Scholar]
- England P and Kilbourne B (1990). Markets, Marriages, and Other Mates: The Problem of Power. In Friedland R and Robertson AF (Eds.), Beyond the Marketplace: Rethinking Economy and Society (pp. 163–88). New York: Aldine. [Google Scholar]
- Esping-Andersen G (2007). Sociological explanations of changing income distributions. American Behavioral Scientist, 50: 639–658. [Google Scholar]
- Esping-Andersen G and Billari F 2015. “Re-theorizing Family Demographics.” Population and Development Review, 41(1)1–31. [Google Scholar]
- Frech A, Lankes J, Damaske S and Ohler A (2023). The Myth of Men’s Stable, Continuous Labor Force Attachment: Multitrajectories of U.S. Baby Boomer Men’s Employment. Socius, 9: 1–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Garcia-Manglano J 2015. “Opting out and leaning in: The life course employment profiles of early baby boom women in the United States.” Demography, 52, 1961–1993. [DOI] [PubMed] [Google Scholar]
- Gerstel N, and Clawson D 2014. “Class Advantage and the Gender Divide: Flexibility on the Job and at Home.” American Journal of Sociology, 120 (2): 395–431. [DOI] [PubMed] [Google Scholar]
- Goldin C 2021. Career & Family: Women’s Century-Long Journey Toward Equity. Princeton: Princeton University press. [Google Scholar]
- Goldscheider F, Bernhardt E and Lappegård T 2015. “The Gender Revolution: A Theoretical Framework for Understanding New Family-Demographic Behavior.” Population and Development Review, 41(2): 207–39. [Google Scholar]
- Gonalons-Pons P 2015. “Gender and class housework inequalities in the era of outsourcing hiring domestic work in Spain.” Social Science Research, 52: 208–218. [DOI] [PubMed] [Google Scholar]
- Gonalons-Pons P, and Schwartz C (2017). Trends in economic homogamy: Changes in assortative mating or the division of labor in marriage? Demography, 54: 985–1005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Gottschalk P and Moffitt R (1994). The growth of earnings instability in the US labor market. Brookings Papers on Economic Activity 1994, no. 2, 217–272. [Google Scholar]
- Hardy B (2014). Childhood Income Volatility and Adult Outcomes. Demography, 51: 1641–1665. [DOI] [PubMed] [Google Scholar]
- Hays Sharon. 1996. The Cultural Contradictions of Motherhood. Yale University Press. [Google Scholar]
- Kalleberg AL (2011). Good jobs, bad jobs: The rise of polarized and precarious employment systems in the United States, 1970s to 2000s. New York: Russell Sage. [Google Scholar]
- Kalmijn M and Poortman A (2006). His or her divorce? The gendered nature of divorce and its determinants. European Sociological Review, 22: 201–214. [Google Scholar]
- Kalmijn M, Loeve A, and Manting D (2007). Income dynamics in couples and the dissolution of marriage and cohabitation. Demography, 44: 159–179. [DOI] [PubMed] [Google Scholar]
- Killewald A, and García‐Manglano J (2016). Tethered lives: A couple‐based perspectiveon the consequences of parenthood for time use, occupation, and wages. Social Science Research, 60: 266–282. [DOI] [PubMed] [Google Scholar]
- Killewald A, and Gough M (2013). Does specialization explain marriage penalties and premiums? American Sociological Review, 78: 477–502. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Landivar L 2017. Mothers at Work: Who Opts Out? Boulder: Lynne Rienner Publishers. [Google Scholar]
- Langner L (2015). Within-Couple Specialization in Paid Work: a long-term pattern? A dual trajectory approach. Advances in Life Course Research, 24: 47–65. [DOI] [PubMed] [Google Scholar]
- Lareau A 2003. Unequal Childhoods: Race, Class, and Family Life. University of California Press. [Google Scholar]
- Lichter DT and Qian Z (2019). The study of assortative mating: Theory, data, and analysis. In Schoen R (Ed.), Analytical family demography (1st ed., pp. 303–337). Cham, Switzerland: Springer International Publishing. [Google Scholar]
- Lundberg S, and Rose E (2000). Parenthood and the earnings of married men and women. Labour Economics, 7, 689–710. [Google Scholar]
- Maume DJ 2006. “Gender differences in restricting work efforts because of family responsibilities.” Journal of Marriage and Family, 68(4): 859–869. [Google Scholar]
- Mincer J (1974). Schooling, Experience and Wage. New York: Columbia University Press. [Google Scholar]
- Moen P and Hernandez E (2009). Social Convoys: Studying linked lives in time, context, and motion”. In Elder G Jr. and Giele J (Eds.), The Craft of Life Course Research (pp. 258–279). New York: Guilford Press. [Google Scholar]
- Möhring K and Weiland A (2021). Couples’ Life Courses and Women’s Income in Later Life: A multichannel sequence analysis of linked lives in Germany. European Sociological Review, 38(3): 371–389. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Musick K, Bea M and Gonalons-Pons P (2020). His and Hers Earnings Following Parenthood in the United States, Germany and the United Kingdom. American Sociological Review, 85: 639–674. [Google Scholar]
- Nagin D (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press. [Google Scholar]
- Nagin D and Tremblay R (2001). Analyzing Developmental Trajectories of Distinct but Related Behaviors. A Group-Based Method. Psychological Methods, 6: 18–34. [DOI] [PubMed] [Google Scholar]
- Nagin D and Tremblay R (2005a). Developmental trajectory groups: Fact or a useful statistical fiction? Criminology, 43(4): 873–904. [Google Scholar]
- Nagin D and Tremblay R (2005b). Further reflections on modeling and analyzing developmental trajectories: A response to Maughan and Raudenbush. The Annals of the American Academy of Political and Social Science, 602: 145–154. [Google Scholar]
- Nutz T and Gritti D (2021). Dyadic Employment Biographies and Within-Couple Wealth Inequality in Britain and Western Germany. Journal of Marriage and Family, 1–18. [Google Scholar]
- Nylund K, Asparouhov T, and Muthén B (2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14(4): 535–569. [Google Scholar]
- Oppenheimer V (1988). “A Theory of Marriage Timing.” American Journal of Sociology, 94: 563–91. [Google Scholar]
- Pessin L and Pojman EM (2022). Visualizing Racial-Ethnic Differences in the Division of Housework among Different-Sex Couples in the United States. Socius, 8:1–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Qian Y (2017). Educational assortative mating and female breadwinning trajectories: A group‐based trajectory analysis. In Scott C & Blair S (Eds.), Intimate relationships and social change (pp. 95–123). Bingley, West Yorkshire: Emerald. [Google Scholar]
- Qian Y (2018). Educational Assortative Mating and Income Dynamics in Couples: A Longitudinal Dyadic Perspective. Journal of Marriage and Family, 80:607–621. [Google Scholar]
- Raley S, Mattingly M, and Bianchi S (2006). “How dual are dual-income couples? Documenting change from 1970 to 2001.” Journal of Marriage and Family, 68, 11–28. [Google Scholar]
- Reynolds L (2021). Remarriage Rate in the U.S.: Geographic Variation, 2019 Bowling Green, OH, National Center for Family and Marriage Research, Bowling Green State University. Retrieved from https://www.bgsu.edu/ncfmr/resources/data/family-profiles/reynolds-remarriaage-US-geographic-variation-2019-fp-21-18.html. [Google Scholar]
- Ruggles S 2015. “Patriarchy, Power, and Pay: The Transformation of American Families, 1800–2015.” Demography, 52:1797–1823. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sassler S and Lichter D (2020). Cohabitation and Marriage: Complexity and Diversity in Union-Formation Patterns. Journal of Marriage and Family, 82: 35–61. [Google Scholar]
- Sanchez L and Thomson E (1997). Becoming Mothers and Fathers: Parenthood, Gender and the Division of Labor. Gender and Society, 11:747–772. [Google Scholar]
- Schneider D and Hastings O 2017. “Income Inequality and Household Labor.” Social Forces, 96 (2): 481–506. [Google Scholar]
- Schwartz C (2010). Earnings inequality and the changing association between spouses’ earnings. American Journal of Sociology, 115: 1524–1557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Shafer E (2011). Wives’ Relative Wages, Husbands’ Paid Work Hours, and Wives’ Labor-Force Exit. Journal of Marriage and Family, 73: 250–263. [Google Scholar]
- Shen Y (2021). The Nonlinear Linkage Between Earnings Homogamy and Earnings Inequality Among Married Couples. Demography, 58(2): 527–550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Smock P and Schwartz C (2020). The Demography of Families: A Review of Patterns and Change. Journal of Marriage and Family, 82: 9–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Song X, Zang E, Land K and Zheng B 2022. “Intergeneration Income Mobility Table Revisited: A Trajectory Group Perspective.” Research in Social Stratification and Mobility, 80: 100713. [Google Scholar]
- Sweeney M (2002). Two Decades of Family Change: The Shifting Economic Foundations of Marriage. American Sociological Review, 67(1): 132–147. [Google Scholar]
- Usdansky M 2011. “The Gender Equality Paradox: Class and Incongruity Between Work-Family Attitudes and Behaviors.” Journal of Family Theory & Review, 3(3): 163–178. [Google Scholar]
- Van Bavel J and Klesment M (2017). Educational Pairings, Motherhood and Women’s Relative Earnings in Europe. Demography, 54: 2331–2349. [DOI] [PubMed] [Google Scholar]
- Western B, Bloome D, Sosnaud B, and Tach L (2012). “Economic insecurity and social stratification.” Annual Review of Sociology, 38(1), 341–359. [Google Scholar]
- Weisshaar K and Cabello-Hutt T (2020). “Labor Force Participation Over the Life Course: The Long-Term Effects of Employment Trajectories on Wages and the Gendered Payoff of Employment.” Demography, 57: 33–60. [DOI] [PubMed] [Google Scholar]
- Winslow-Bowe S (2006). “The Persistence of Wives’ Income Advantage.” Journal of Marriage and Family, 68: 824–842. [Google Scholar]
- Ziliak J, Hardy B and Bollinger C (2011). Earnings Volatility in America: Evidence from matched CPS. Labor Economics, 18: 742–754. [Google Scholar]