The Economic Foundations of Cohabiting Couples’ Union Transitions

Abstract

Introduction

In recent decades, cohabitation has become an increasingly important relationship context for U.S. adults and their children. The proportion of women aged 19–44 who had ever cohabited increased from 33 % to 45 % between 1987 and 1995 (Bumpass and Lu 2000). By 2009–2010, 60 % of women aged 19–44 had ever cohabited, along with 73 % of women aged 30–34 (Manning 2010). These changes have altered children’s family experiences. The share of children expected to spend time with a cohabiting mother rose from 37 % between 1990 and 1994 to 46 % between 1997 and 2001 (Kennedy and Bumpass 2008). By 2001, more than one-half of nonmarital births were to cohabiting mothers (Manlove et al. 2010).

Cohabiting unions in the United States are characterized by relatively short durations and high levels of instability. Estimates from 2006–2010 show that the median duration of first premarital cohabitation is less than two years, with 40 % ending in marriage and 27 % ending in dissolution within three years (Copen et al. 2013). In response to the growing prevalence of cohabiting unions and their high levels of instability, demographers have been increasingly interested in understanding why some couples marry but others separate.

Understanding why cohabiting unions end in marriage or dissolution is important because a large body of evidence indicates that both transitions have significant consequences for adult and child well-being. Marriage is associated with increases in economic well-being in addition to physical and psychological health (Burnstein 2007; Killewald and Gough 2013; Ribar 2015; Waite 1995; Waite and Lehrer 2003; Williams et al. 2008). Marriage also has potential benefits for child well-being, either by making unions more stable or by promoting joint investments in children (Lundberg and Pollak 2014; Osborne and McLanahan 2007). In contrast, union dissolution is linked to negative economic and psychological outcomes for adults and children (Brown 2004, 2006; Lee and McLanahan 2015; McLanahan et al. 2013; Tach and Eads 2015; Williams et al. 2008).

To understand why some cohabiting couples marry but others separate, researchers have drawn on theories of the economic determinants of marriage and divorce, including those emphasizing the benefits of economic specialization (Becker 1981), cultural expectations of male breadwinning (Bertrand et al. 2015; Oppenheimer 1988; West and Zimmerman 1987), low income as a source of relationship conflict and stress (McCubbin and Patterson 1983; Yeung et al. 2002), and rising economic standards associated with marriage (the marriage bar) (Edin and Kefalas 2005; Watson and McLanahan 2011). Although all of these theories focus on the importance of money and work, they differ substantially in the mechanisms through which economic resources are hypothesized to affect union transitions and the relative importance of men’s, women’s, and couples’ contributions. Each theory reflects a fundamentally different perspective on the economic foundations of family life.

Important gaps remain in our understanding of the economic determinants of marriage and dissolution among cohabiting couples. Prior research has been unable to reach clear conclusions due to data constraints, including a lack of data on partners’ earnings, length bias, recall bias, nonrandom sample attrition, and small sample sizes. Additionally, although qualitative studies have found that couples compare their own economic situation with a standard associated with marriage in assessing their marriage readiness, no study has used longitudinal data to test the predictions of marriage bar theory among cohabiting couples. As a result, we lack a clear understanding of whether and how cohabiting men’s, women’s, and couples’ economic resources influence union transitions.

In this study, I use data from a series of nationally representative panels of the Survey of Income and Program Participation (SIPP) from 1996–2013 to prospectively test alternative theories of the economic foundations of cohabiting couples’ union transitions. Given significant social, economic, and policy changes in recent decades (Corcoran et al. 2000; Goldin 2006; McLanahan 2004), shifting meanings of marriage (Cherlin 2004; Coontz 2004; Edin and Kefalas 2005; Goldscheider et al. 2015), and evidence of changing associations between economic resources and union transitions (Goldstein and Kenney 2001; Schwartz and Gonalons-Pons 2016; Sweeney 2002), it is particularly important to address these questions with recent, high-quality data.

Theory and Prior Research

Theories of the economic determinants of cohabiting couples’ union transitions were developed primarily as explanations of marriage among singles and divorce among married couples. However, it is important to consider cultural, institutional, compositional, legal, and structural differences among singlehood, cohabitation, and marriage that may alter the relationships between money, work, and union transitions. Although couples sometimes “slide” or “drift” from singlehood into cohabitation (Manning and Smock 2005), and cohabitation in principle shares many of the same benefits of marriage (e.g., coresidence, economies of scale, and resource pooling) (Oppenheimer 1988, 1994), important differences exist among singlehood, cohabitation, and marriage (Rindfuss and Vanden Heuvel 1990; Smock 2000; Thornton et al. 2007). This study considers how theories apply within the unique context of cohabitation.

Specialization and Economic Dependence Perspectives

According to specialization theory, individuals with complementary skills can take advantage of the gains to trade by forming a household and specializing in different activities (Becker 1981). Specialization allows each partner to invest more and become more efficient in their respective domains, which in turn increases household well-being (Becker et al. 1977). For example, having a partner available to take a sick child to a doctor’s appointment may free an employed partner to work a higher-wage job that requires longer and more continuous hours, thereby increasing family income (Kalmijn et al. 2007). Paid work may interfere with home and family life (Schieman et al. 2009), and dual-earner couples may feel especially acute time pressures (Jacobs and Gerson 2004).

But as Cherlin (2004) noted, specialization theory is designed to explain transitions from single-person to multiperson households—not between cohabitation and marriage. However, the benefits of remaining in a cohabiting union relative to separating theoretically depend on within-couple specialization. Although often overlooked, Becker’s (1981:14) theory is explicitly gender-neutral.¹ In principle, both married and cohabiting couples can benefit from specialization. However, given the high instability of cohabiting unions (Copen et al. 2013) and lack of legal protections in the event of dissolution (Brines and Joyner 1999), specialization is especially risky for cohabiting partners specializing in unpaid work.

A related theory of within-couple differences in money and work is the economic dependence perspective, which posits that within-couple income inequality reduces the risk of dissolution (Nock 2001). Within-couple income inequality may promote stability by increasing couples’ economic interdependence, commitment, and sense of mutual obligation (Nock 1995; Rogers 2004). Because cohabiting unions have shorter time horizons and lower commitment than marriage (Nock 1995), though, whether these processes apply similarly to cohabitation is unclear. Given that cohabitors tend to have egalitarian views about men’s and women’s economic roles (Clarkberg et al. 1995), equality rather than economic dependence may promote stability among cohabitors (Brines and Joyner 1999; Kalmijn et al. 2007).

Prior research on economic specialization or within-couple earnings inequality and union dissolution has produced positive, negative, and null findings. Consistent with specialization theory, Brines and Joyner (1999) showed that a greater difference in cohabiting couples’ work hours is associated with a lower dissolution risk. In a study combining married and cohabiting couples, Weisshaar (2014) found that heterosexual couples with unequal earnings are more stable than couples with equal earnings. Studies of married couples similarly have provided evidence that equal earnings increase the risk of divorce (Nock 2001; Rogers 2004). However, Brines and Joyner (1999) posited that within-couple economic inequality may actually stabilize marriage but destabilize cohabitation. Among cohabiting couples in the Netherlands, Kalmijn et al. (2007) found that within-couple income inequality is destabilizing. In contrast, other U.S.-based studies of cohabiting couples have found no association between within-couple earnings inequality and dissolution (Brines and Joyner 1999; Sassler and McNally 2003; Smock and Manning 1997).

The Male Breadwinner Perspective

Although specialization and economic dependence are theoretically gender-neutral (Becker 1981), the male breadwinner perspective emphasizes the symbolically gendered meanings of economic contributions (Sayer et al. 2011; Tichenor 1999; Zelizer 1997). Because of the historical association between men’s employment and family income (Oppenheimer 1988), the imperative to be a good provider is theorized to be stronger for men. Paid work is an important way that men and women “do gender” (Schneider 2012; West and Zimmerman 1987). Men who lack full-time employment or earn less than women violate male breadwinning norms (Bertrand et al. 2015; Killewald 2016; Sayer et al. 2011), threatening men’s gender identity and lowering marital satisfaction (Bertrand et al. 2015; Bittman et al. 2003; Brines 1994; Tichenor 1999). Men whose gender identity is threatened may overcompensate with extreme displays of masculinity (Munsch 2015; Willer et al. 2013).

Male breadwinning expectations, however, may be less consequential among cohabitors. Compared with singles who marry directly, cohabitors have more gender egalitarian attitudes (Clarkberg et al. 1995). Additionally, relative to marriage, cohabitation is an “incomplete institution” (Cherlin 1978) with fewer clear guidelines for men’s and women’s economic roles. As a result, what it means to “do gender” in paid work may be more flexible in cohabitation.

Prior research has provided partial support for the idea that gendered cultural expectations about money and work influence cohabiting couples’ union transitions. Studies have often found that only men’s earnings or employment increase the likelihood of marriage (Brown 2000; Lichter et al. 2006; Oppenheimer 2003; Sanchez et al. 1998; Smock and Manning 1997). Qualitative research has also suggested that men’s income and employment are seen as more important than women’s for marriage (Smock et al. 2005). Evidence that women outearning men deters marriage is inconsistent, however (Bertrand et al. 2015; Sassler and McNally 2003).

Previous research on union dissolution is also mixed. Some studies have shown that only men’s earnings are associated with a lower dissolution risk (Oppenheimer 2003; Sassler and McNally 2003), but other research has found null effects of both partners’ earnings or employment (Brown 2000; Lichter et al. 2006; Smock and Manning 1997). Some studies have found that couples in which women outearn men are more likely to separate than male breadwinner couples (Brines and Joyner 1999), whereas others have reported no differences (Sassler and McNally 2003). Despite more gender egalitarian attitudes among cohabitors, a lack of institutionalized economic roles, and evidence that equality promotes stability (Brines and Joyner 1999; Kalmijn et al. 2007), prior research has suggested that the male breadwinner norm may still affect union transitions.

The Family Stress Model

The family stress model emphasizes couples’ combined economic resources, highlighting the adverse effects of low and unstable income on family relationships (McCubbin and Patterson 1983; Yeung et al. 2002). Not having enough money appears to be an important source of stress and conflict for cohabiting couples (Hardie and Lucas 2010; Smock et al. 2005), which should increase union instability. Conversely, because income buffers couples from stressors associated with financial strain, higher-income couples should be more likely to marry. In this framework, income matters for its practical rather than symbolic value, regardless of the relative amount that men and women contribute to the total pool.

Despite strong theoretical reasons to expect couples’ combined income to matter, prior research in the United States has provided surprisingly limited support for the family stress model for cohabiting couples’ union transitions. Studies from Canada (Wu and Pollard 2000) and the Netherlands (Kalmijn et al. 2007) showed that higher earnings are associated with a lower risk of dissolution, but studies of couples in the United States reported either no association between couples’ earnings and marriage or dissolution (Sassler and McNally 2003) or even a positive relationship between earnings and dissolution (Brines and Joyner 1999). These findings are puzzling because cohabiting couples’ earnings are associated with lower levels of conflict (Hardie and Lucas 2010), and qualitative studies have found that not having enough money is an important source of stress and conflict (Smock et al. 2005).

The Marriage Bar

As marriage has become increasingly concentrated among socioeconomically advantaged couples (Lundberg and Pollak 2014), the economic standards associated with marriage—the marriage bar—may have become more difficult to achieve (Cherlin 2004; Edin and Kefalas 2005; McLanahan and Percheski 2008; Watson and McLanahan 2011). According to this theory, marriage is associated with norms, including achievement of a middle-class standard of living. The further couples are from reaching this standard, the more deviant they are from norms associated with marriage and the less likely to marry. What matters in this account are couples’ economic resources relative to a standard associated with marriage (Dixon 1978; Easterlin 1980). The marriage bar concept is most often applied to couples’ combined resources (Edin and Kefalas 2005; Gibson-Davis et al. 2005) but has also been applied to men’s earnings specifically (Watson and McLanahan 2011). The marriage bar is defined not only by earnings but also by wealth (Edin and Kefalas 2005; Gibson-Davis et al. 2005; Schneider 2011; Smock et al. 2005).

Increasing evidence suggests that middle-class economic standards affect marriage behavior. Qualitative studies have shown that low-income couples struggle to reach the high perceived economic prerequisites of marriage, including financial stability; enough savings for a big wedding; and having assets, such as a home and car (Edin and Kefalas 2005; Gibson-Davis et al. 2005; Smock et al. 2005). Quantitative studies have supported the idea that wealth and income relative to middle-class standards matter. In a study of single men and women, Schneider (2011) found that asset ownership is an important predictor of first marriage. Using census data, Watson and McLanahan (2011) defined the marriage bar as the median of full-time employed men’s earnings in a given metropolitan area, race and ethnicity, and education group. Focusing on the ratio of men’s earnings to the marriage bar, they showed that increases in men’s earnings below the bar significantly raise the probability of being married, whereas earnings increases above the bar have much smaller effects.

Hypotheses

Theories of the economic determinants of marriage and dissolution emphasize different mechanisms and measures. Table 1 lists hypotheses and key measures for each union transition and theoretical perspective.

Although specialization theory does not apply to transitions from cohabitation to marriage, specialization in work hours should increase couples’ well-being and lower the risk of dissolution. To examine how economic dependence influences union stability, I also consider within-couple earnings differences:

• Hypothesis 1: Couples with a greater difference in work hours and earnings should be less likely to dissolve.

However, if cohabitors value equality (Kalmijn et al. 2007), or if similar economic roles and equality promote commitment among cohabitors (Brines and Joyner 1999), specialization and earnings inequality may increase the risk of dissolution.

The male breadwinner perspective implies two predictions. First, although women’s employment is now normative and unlikely to violate gendered relationship norms (Killewald 2016), male breadwinning norms have been slower to change (Sayer et al. 2011). Therefore, although women’s employment should be unrelated to union transitions, men’s lack of full-time employment should decrease the risk of marriage and increase the risk of dissolution. Second, female breadwinner couples should be less likely to marry and more likely to dissolve:

• Hypothesis 2A: Women’s full-time employment will be unrelated to marriage and dissolution; men’s full-time employment will increase the risk of marriage and decrease the risk of dissolution.

• Hypothesis 2B: Female breadwinner couples will be less likely to marry and more likely to dissolve than equal-earner or male breadwinner couples.

Unlike the male breadwinner perspective, the family stress model does not differentiate between men’s and women’s economic contributions; what matters instead is the total pool of income. Higher income should promote marriage and union stability:

• Hypothesis 3: Increases in couples’ combined earnings should be associated with a higher risk of marriage and a lower risk of dissolution.

Marriage bar theory posits that couples compare their own economic well-being with a standard typical of married couples. For earnings, the measure that relates these concepts is the ratio of a couple’s earnings to the median of married couples’ earnings. Increases in earnings ratios below the marriage bar should increase the risk of marriage as couples move closer to achieving the standard. After couples have reached the marriage bar, however, further increases should have little additional effect. The functional form should be nonlinear, with a discontinuity after couples have reached the bar. Qualitative research has suggested that the economic prerequisites of marriage also include assets, such as a home and having enough savings for a big wedding (Edin and Kefalas 2005; Smock et al. 2005).

• Hypothesis 4A: Increases in couples’ earnings relative to the median of married couples’ earnings should increase the risk of marriage; increases in earnings below the marriage bar should have larger effects than increases above the bar.

• Hypothesis 4B: Having assets—such as a home, financial investments, and savings—should increase the risk of marriage.

Because each perspective emphasizes different mechanisms and measures, patterns for marriage and dissolution can provide important theoretical insights.

Gaps in the Literature and Study Contributions

Previous studies of the economic determinants of cohabiting couples’ union transitions in the United States, which relied on data from the National Longitudinal Survey of Youth (NLSY79), National Survey of Families and Households (NSFH), and Panel Study of Income Dynamics (PSID), have produced conflicting evidence and have important data limitations. First, studies using the NLSY79 have been limited in their ability to test theories because of a lack of data on partners’ earnings (Lichter et al. 2006; Oppenheimer 2003). Second, although the PSID has more extensive measures of economic characteristics, the two-year survey interval of the PSID since 1997 would fail to capture a large number of short-term unions and produce a length-biased sample. Brines and Joyner’s (1999) PSID study similarly oversampled more stable couples by restricting their analyses to couples cohabiting for at least two years. Third, for studies using the NSFH (Brown 2000; Sanchez et al. 1998; Sassler and McNally 2003; Smock and Manning 1997), the survey wave interval is approximately five years. Because cohabitation tends to be short-lived (Bumpass and Sweet 1989) and evidence suggests significant recall bias in retrospective cohabitation data (Hayford and Morgan 2008), analyses of NSFH data may introduce nonrandom measurement error. Another consequence of this five-year interval is nonrandom attrition, with conclusions about economic resources and union transitions being highly sensitive to how missing data are handled (Sassler and McNally 2003). Moreover, length bias arises because the NSFH samples couples that are already cohabiting at Wave 1. Finally, no prior study of cohabiting couples has used panel data to test the theory that wealth and income relative to the marriage bar shape marriage decisions.

Studying cohabiting couples’ union transitions has analytical and theoretical advantages. In contrast to studies of marriage among singles, information on both partners’ earnings and employment is important for differentiating between the influence of men’s, women’s, and couples’ economic resources on marriage. And unlike marriage, which involves longer durations, high exit costs, and greater opportunities for partners’ employment to change in anticipation of dissolution, cohabiting unions have a shorter time horizon, lower exit costs, and less opportunity to change employment and earnings in anticipation of dissolution. I consider how unique features of cohabitation may influence the relationship between money, work, and union transitions.

In this study, I use data from multiple panels of the SIPP over the period 1996–2013, merged with Current Population Survey (CPS) Basic Monthly data, to test alternative theories of the economic determinants of marriage and dissolution among cohabiting couples. This study is well suited to addressing the research questions for several reasons. First, because the SIPP is designed to measure individual and household economic well-being, it contains extensive and high-quality measures of both partners’ employment, earnings, and wealth, in addition to a range of potentially confounding factors that make the data ideal for differentiating between alternative theories of the economic determinants of union transitions. Second, because interviews are conducted every four months, the data enable me to capture short-term cohabitations that could be missed by surveys with longer interview intervals (Hayford and Morgan 2008). Third, the SIPP prospectively captures month-to-month change and stability in household composition, marital status, and economic resources, thus limiting recall bias and avoiding both length bias and left-censoring. Fourth, the sampling design of the SIPP allows me to make inferences using a series of recent, large, nationally representative panels. Compared with prior studies, the size of the SIPP panels significantly improves my ability to differentiate between null effects and statistical imprecision. Finally, this is the first study to use panel data to test the theory that wealth and income relative to the marriage bar influence cohabiting couples’ marriage transitions.

Data

The data are from the CPS Outgoing Rotation Group² (Flood et al. 2015) and the SIPP, and cover the period 1996–2013.³ I use the CPS to construct a measure of the marriage bar: the median earnings of married couples in each state and month over the study period. The SIPP data included in this study are from four continuous, nationally representative panels starting in 1996, 2001, 2004, and 2008. Each panel follows all members of initially sampled households and individuals residing with them for between three and five years, with interviews conducted every four months. The SIPP does not follow sample members who move abroad, move to military housing, or become institutionalized. SIPP interviews are conducted in person or by phone. I analyze union transitions for up to four years since union formation. Because the median duration of cohabiting unions is less than two years (Copen et al. 2013), the SIPP is especially well suited to studying cohabiting couples’ union transitions.

At the first wave of each SIPP panel, interviewers visit sampled addresses in person, creating a household roster and attempting to interview all household members aged 15 or older. All original household members, along with those who reside with them, are followed throughout the duration of the panel even if they move to new addresses. This design is important for capturing members of cohabiting couples who move after marrying or separating. For each household, the SIPP identifies a household reference person, who is generally an owner or renter of the housing unit. The household roster identifies relationships between the household head and all other household members in each month. A limitation of the SIPP is that it identifies only those cohabiting couples involving a household head, which misses an estimated 6 % of cohabiting couples (Kennedy and Fitch 2012). Although the first SIPP panel started in 1984, the SIPP explicitly started asking about cohabiting partners in the 1996 panel. Comparisons of inferred and direct cohabitation measures show that inferred measures produce different estimates of the prevalence and duration of cohabiting unions (Baughman et al. 2002). Therefore, I rely on the 1996, 2001, 2004, and 2008 panels—all of which use the direct cohabitation measure.

Measures

Cohabiting Couples

Using the household roster, cohabiting couples are identified as household heads and their unmarried partners. I exclude couples that were already cohabiting at the beginning of the first wave of each panel because their union duration is unknown. I prospectively follow all couples from the first month they are observed cohabiting until the month they experience a union transition or are censored. This exclusion results in a large reduction in the number of couples in the analytic sample but avoids the substantial bias that would be introduced by oversampling more-stable couples and analyzing couples with unknown union durations.

Union Transitions

Although SIPP interviews occur every four months, data are collected at each wave about each of the preceding four months, and union transitions are observed at the month level. Marriage occurs in the month in which both members of a cohabiting couple transition to being married and are living in the same household. Dissolution occurs when only one partner is observed in the month following the couple’s last cohabitation date. Couples still cohabiting at the last survey wave at which both partners are interviewed (either by loss to follow-up or because the panel ended) are censored.

Economic Characteristics

To adjudicate between model specifications, I compare Bayesian information criterion (BIC) model fit statistics. The BIC facilitates comparisons of nonnested models and favors more parsimonious specifications by penalizing models based on the number of parameter estimates (Raftery 1995). Initial model specifications included men’s and women’s earnings individually. However, BIC statistics indicated that couples’ combined earnings produce a better model fit than men’s and women’s earnings entered individually.

Control Variables

Based on prior research, I adjust for factors that may be associated with both economic resources and union transitions. Because multipartner fertility and shared biological children are associated with economic disadvantage (Guzzo and Furstenberg 2007) and union transitions (Carlson et al. 2004; Tach and Edin 2013), models differentiate between couples with no children present and couples with all shared biological children, some shared biological children, and no shared biological children. While recognizing that education is in part an indicator of long-term economic potential (Oppenheimer 2003; Smock and Manning 1997), this study focuses on how couples’ current economic resources are associated with union transitions, net of education. Therefore, I conceptualize education as a control variable in these analyses and include categorical measures of each partner’s educational attainment (less than high school graduate, some college, college graduate or more, with high school graduate as the reference category). Because cohabiting couples’ likelihood of marriage has declined (Kuo and Raley 2016) and union durations have increased (Copen et al. 2013) since the mid-1990s, I also include panel dummy variables (2001, 2004, 2008, with the 1996 panel as the reference category).

Studies conducted prior to welfare reform posited that eligibility for Aid to Families with Dependent Children (AFDC) incentivized cohabitation over marriage (Moffitt et al. 1998). Consistent with this argument, Lichter et al. (2006) found that women receiving AFDC were less likely to marry. However, changes in welfare eligibility theoretically alter the relationship between marriage and employment among low-income mothers (Bitler et al. 2004). Economic benefits may also provide resources to exit unions (Cancian and Meyer 2014). I include a measure of whether either partner receives any means-tested cash or in-kind transfers. Other control variables include age and age squared of each partner; region; race and ethnicity of the female partner; whether partners have a different race and ethnicity; whether each partner is a full-time student; and whether each partner was previously married. The full set of variables used in the analyses are summarized in Table S4 (Online Resource 1).

Methods

Descriptive Results

During the observation period, there are a total of 5,303 cohabiting couples in the analytic sample exposed to the risk of marriage and dissolution. I restrict analysis to the first observed cohabiting union for each originally sampled household.⁸ Among couples experiencing a union transition during the observation period, slightly more than one-half are dissolutions, with 1,104 marriages and 1,121 dissolutions. The majority (60 %) of right-censored couples are censored because the panel ended, with the remainder being lost to follow-up. Figure 1 plots the estimated proportion of couples that remain in a union over time, along with the proportion that have exited by marriage and dissolution separately. Consistent with prior research (Copen et al. 2013), descriptive findings show that most couples exit cohabiting unions quickly through either marriage or dissolution. Within two years, an estimated 57 % of cohabiting couples have experienced a union transition, with only 22 % still cohabiting by year 4.

Table 2 provides descriptive statistics for the first sample month observation for all cohabiting couples and separately for couples that eventually marry or dissolve. The third column indicates whether the mean or proportion for couples that marry differs from that of couples that do not marry. Descriptively, female breadwinner couples are not underrepresented among couples that eventually marry, and both female and male partners are more likely to be working full-time in couples that marry. In line with the predictions of the family stress model, couples that marry have significantly higher combined monthly earnings. Finally, consistent with marriage bar theory, the ratio of couples’ earnings to the marriage bar is significantly higher among couples that marry, and couples that marry are more likely to be homeowners, have financial assets, and have savings accounts.

The fourth column in Table 2 indicates whether the mean or proportion for couples that eventually dissolve differs from that of couples that do not dissolve. Contrary to specialization theory, couples that separate have a similar difference in paid work hours as couples that do not separate. Inconsistent with the economic dependence perspective, couples with unequal earnings are no more likely to dissolve than couples with equal earnings. Related to the male breadwinner perspective, couples that separate are less likely to have either a male or female partner employed full-time, and female breadwinner couples are not overrepresented among couples that separate. Last, consistent with the family stress model, couples that separate have significantly lower combined earnings.

Multivariate Results

Marriage

Given that economic measures may be correlated with each other and with noneconomic confounding factors, I run multivariate survival models. I first present tests of theories of the economic determinants of marriage with Cox proportional hazards models in column 1 of Table 3. The model describes how covariates affect the relative risk of marriage compared with continuing to cohabit. Couples that dissolve are censored. I focus primarily on the key economic predictors of union transitions relevant to each theory, and I present the full set of parameter estimates for marriage and dissolution in Table S5 (Online Resource 1).

The male breadwinner perspective emphasizes cultural norms about the primacy of men’s economic roles within families. I test whether men’s full-time employment is more important than women’s in raising the odds of marriage and whether women outearning men is associated with a lower risk of marriage. Contrary to the predictions of the male breadwinner perspective, couples are no less likely to marry when women are primary breadwinners, and neither men’s nor women’s full-time employment is associated with marriage. These results do not support the predictions of the male breadwinner perspective for marriage.

Couples’ earnings are hypothesized to promote marriage and union stability by buffering couples against stress and conflict. However, couples’ absolute earnings are not associated with marriage. Results instead largely support the predictions of marriage bar theory. Adjusting for couples’ absolute earnings, increases in couples’ earnings relative to the marriage bar are strongly and positively associated with marriage below the bar. Above the marriage bar, increases in earnings appear to be inconsequential. These findings are robust to more flexible specifications of couples’ absolute earnings, as well as more flexible specifications of the ratio of couples’ earnings to the marriage bar (see Online Resource 1, Table S2). Indicators of wealth also generally align with the predictions of marriage bar theory, with homeownership and receipt of interest from financial assets being strongly and positively associated with marriage. However, whether either partner has a savings account, another potential indicator of wealth, is unrelated to marriage.

Among other important covariates in the marriage model, college-graduate women have higher odds of marriage than high school graduates, and men with less than a high school diploma have lower odds than high school graduates. Additionally, couples with a female partner enrolled as a full-time student are less likely to marry. Inconsistent with the claim that means-tested benefits discourage marriage, however, receipt of cash or noncash benefits is unrelated to marriage. Finally, consistent with prior research (Kuo and Raley 2016), survey panel indicators show that couples’ odds of transitioning to marriage have declined, with the odds of marriage being 28 % lower in the 2008 panel relative to the 1996 panel.

Dissolution

Next, I estimate the relative risk of dissolution compared with remaining in a cohabiting union. Couples that marry are censored at the time of marriage. Because marriage bar theory does not apply to dissolution, I test predictions of the specialization and economic dependence perspectives, the male breadwinner perspective, and the family stress model.

Results for union dissolution do not support the predictions of specialization theory or the economic dependence perspective. Focusing on the difference between couples’ work hours, more-specialized couples do not have lower dissolution risks than less-specialized couples. Additionally, couples with greater earnings inequalities have higher odds of dissolution than couples with more equal earnings—a result consistent with prior studies finding that equality rather than economic dependence promotes union stability among cohabiting couples (Brines and Joyner 1999; Kalmijn et al. 2007). These results are robust to a number of alternative work hour specialization and earnings inequality measures.

Findings for union dissolution are consistent with the predictions of the family stress model but inconsistent with those of the male breadwinner perspective. Higher earnings promote union stability, which is consistent with the family stress model. However, both men’s and women’s full-time employment are statistically unrelated to dissolution. Additionally, female breadwinner couples have no higher odds of dissolution than equal earner or male breadwinner couples.

In line with prior research documenting an increasing duration of cohabiting unions between the mid-1990s and late-2000s (Copen et al. 2013), cohabiting couples’ odds of separating are 16 % lower in the 2008 panel relative to the 1996 panel. Consistent with neoclassical economic models of marriage (Becker et al. 1977), which posit that investments in shared assets—including children—promote union stability, both homeownership and shared biological children are negatively associated with dissolution. In contrast, the presence of children who are not biologically shared increases the odds of dissolution. Finally, receipt of means-tested benefits is associated with a higher dissolution risk, suggesting that benefits may provide low-income individuals with resources necessary to exit unions.

Limitations

This study has important limitations. First, the wealth measures are limited to homeownership, receipt of interest from financial assets, and possessing a savings account. These indicators capture important features of the marriage bar described in qualitative studies, but future research should measure other aspects of the marriage bar, such as car ownership, the amount of money in savings, the value of financial assets, and debt.

Second, because the SIPP identifies only cohabiting couples that involve the household head, these analyses miss subfamily cohabiting couples.⁹ Most subfamily cohabiting unions involve a child of the household head and tend to be especially economically disadvantaged (Kennedy and Fitch 2012). Cross-sectional comparisons of household head and subfamily cohabiting unions in SIPP topical module data demonstrate that subfamily couples are significantly more economically disadvantaged than cohabitations involving household heads. Given their high levels of disadvantage, these couples are especially unlikely to meet the economic prerequisites of marriage (Smock et al. 2005). Although these couples’ low incomes should increase their risk of dissolution, coresidence with parents—particularly for couples with children—may provide important resources (e.g., housing and childcare) that buffer such couples from stressors associated with low income.

Third, the data allow me to track couples for only up to four years. Although this window captures an estimated 78 % of cohabitations, I am unable to observe longer-term unions. Supplemental analyses that (1) interact economic variables with time and (2) test the assumption that the relative hazards for economic variables are proportional over time provide no evidence that the relationship between money, work, and union transitions differs over the first four years since union formation. However, these processes may differ in longer-term unions.

To understand how the joint risk of marriage and dissolution influences the composition of longer duration unions, I assess the risk of any exit from cohabitation. Results suggest that couples surviving to longer durations have more equal earnings, are less educated, and have higher odds of receiving means-tested benefits. Longer-duration unions may be heterogeneous with respect to marriage plans, with more-advantaged, egalitarian couples treating cohabitation as an alternative to marriage and more-disadvantaged couples striving to attain the economic prerequisites of marriage. Future research with longer observation windows should compare the economic determinants of union transitions among shorter- and longer-duration cohabitations.

Fourth, attrition from the SIPP may be related to both relationship outcomes and economic resources. However, analyses restricting the sample to couples over the first two years since union formation produce the same conclusions. Furthermore, tests of the proportional hazards assumption fail to reject the null hypothesis that the hazards are proportional over time. Finally, if the most economically disadvantaged couples are more likely to leave the panel, more likely to dissolve, and less likely to marry, this should produce estimates that are more conservative than those presented here.

Discussion and Conclusion

One of the most dramatic changes in the past century has been the increasing similarity of men’s and women’s labor market behavior (Goldin 2014). Women’s earnings have risen in absolute terms and relative to men’s (Blau 1998; Goldin 2006). Although these trends indicate that traditional gender specialization in paid work and within-couple earnings inequality have declined, specialization and economic dependence should theoretically promote union stability. This study finds that within-couple differences in paid work hours are not associated with dissolution, and within-couple earnings inequality increases couples’ risk of separating. Specialization and economic dependence may not promote stability because of unique characteristics of cohabiting unions. High instability and a lack of legal protections make specialization in unpaid work particularly risky for cohabitors and therefore unlikely to enhance commitment (Brines and Joyner 1999). Shorter time horizons in cohabitation relative to marriage also imply that economic dependence may not produce the same sense of mutual obligation and commitment observed for marriage. Equal earnings may increase stability within cohabitation either because cohabiting couples value equality in men’s and women’s economic contributions (Kalmijn et al. 2007) or because equal power promotes commitment by facilitating cooperation rather than bargaining (Brines and Joyner 1999).

Although studies from earlier periods and cohorts have provided evidence that gendered paid work expectations affect union transitions (Brines and Joyner 1999; Smock and Manning 1997), I find weak support for the male breadwinner perspective. I find no evidence that men’s or women’s full-time employment is associated with either union transition, and female breadwinner couples have similar risks of marriage and dissolution as equal earner and male breadwinner couples.

Social scientists have documented reversals of previously negative relationships between women’s socioeconomic resources and union transitions (Goldstein and Kenney 2001; Sweeney 2002) and have suggested the emergence of gender egalitarian relationship models (Esping-Andersen and Billari 2015; Goldscheider et al. 2015). Although wives who outearned their husbands in marriage cohorts from the 1970s and 1980s had increased divorce risks, female breadwinner couples formed in the 1990s no longer had a heightened risk of divorce (Schwartz and Gonalons-Pons 2016). Despite a recent stall, attitudes about men’s and women’s roles at work and at home have become more egalitarian since the 1970s (Cotter et al. 2011), and young adults in recent birth cohorts have gender egalitarian ideals for sharing paid and unpaid work (Gerson 2011; Pedulla and Thébaud 2015). Although the male breadwinner norm may be more consequential among single individuals and married couples (Bertrand et al. 2015; Killewald 2016), cohabitors appear to hold less-traditional attitudes about gender (Clarkberg et al. 1995).

Findings from this study describe diverging trajectories of family life for economically advantaged and disadvantaged couples, and align with research documenting a growing socioeconomic divide in marriage (DiPrete and Buchmann 2006; Goldstein and Kenney 2001; Kuo and Raley 2016; Lundberg and Pollak 2014). Because marriage and union dissolution are both associated with important economic and health outcomes (Avellar and Smock 2005; Burnstein 2007; Killewald and Gough 2013; Tach and Eads 2015; Waite 1995; Waite and Lehrer 2003; Williams et al. 2008), these findings suggest that inequalities in economic resources may amplify inequalities in cohabiting couples’ well-being. Because disadvantaged couples are more likely to separate and less likely to marry, these processes also have implications for inequalities in children’s experiences of family instability.

Why are economically advantaged couples more likely to marry and less likely to separate? For marriage, findings align with studies showing that marriage is increasingly a union status reserved for couples that have achieved a high economic standard (Cherlin 2004; Gibson-Davis et al. 2005; Watson and McLanahan 2011). Consistent with theory, increases in couples’ earnings below the marriage bar are far more important than subsequent increases above the bar. These analyses provide the first empirical evidence using panel data that couples’ earnings relative to a standard associated with marriage are consequential for marriage behavior. Analyses also show that wealth independently predicts marriage, with couples that own a home and receive interest from financial assets being more likely to marry.

For dissolution, results support the predictions of the family stress model. Although stress and conflict are not measured in this study, previous research has found that higher income is associated with lower levels of conflict among cohabiting couples (Hardie and Lucas 2010; Smock et al. 2005). Furthermore, analyses of SIPP topical module data show strong and consistent associations between cohabiting couples’ earnings and experiences of economic hardship. The stress and conflict pathway represents a plausible mediator of the association between income and dissolution that should be investigated more directly in future studies.

The rise of cohabitation and other major transformations that have occurred in the family with the unfolding of the second demographic transition have been highly uneven by socioeconomic status, reflecting a growing divergence in family life (McLanahan 2004). Declines in marriage prevalence and stability and increases in nonmarital fertility have been steepest for less-educated individuals (Ellwood and Jencks 2004; Lundberg and Pollak 2014; Rindfuss et al. 1996; Upchurch et al. 2002). Although cohabitation has become a common experience for men and women of all education levels, it remains concentrated among less-educated men and women (Copen et al. 2013). These trends in the family have occurred alongside the convergence of men’s and women’s economic roles, rising income inequality, and significant changes in welfare policy.

Findings from this study indicate that the economic foundations of cohabiting couples’ union transitions do not lie in economic specialization or only men’s ability to be good providers. Cohabitation operates, at least in part, on different economic principles than marriage. Equality in men’s and women’s economic contributions rather than economic dependence appears to hold cohabiting couples together. Yet, how men and women divide money and work appears to be less consistently important for cohabiting couples’ union transitions than disparities in earnings and wealth between couples. For both marriage and union dissolution, inequalities in economic resources are critical in producing inequalities in couples’ relationship outcomes.

Acknowledgments

This research received generous support from the Cornell Population Center and the Office of Population Research. I am grateful to Sara McLanahan, Kelly Musick, Viviana Zelizer, and the editors and anonymous reviewers for valuable comments and suggestions.

Notes

References