Nonmarital Relationships and Changing Perceptions of Marriage Among African American Young Adults

Abstract

Cohabitation has become increasingly widespread over the past decade. Such trends have given rise to debates about the relation between cohabitation and marriage, in terms of what cohabitation means for individual relationship trajectories and for the institution of marriage more generally. Using recent data from a sample of almost 800 African Americans and fixed effects modeling procedures, in the present study the authors shed some light on these debates by exploring the extent to which cohabitation, relative to both singlehood and dating, was associated with within-individual changes in African Americans’ marital beliefs during the transition to adulthood. The findings suggest that cohabitation is associated with changes in marital beliefs, generally in ways that repositioned partners toward marriage, not away from it. This was especially the case for women. These findings suggest that, for young African American women, cohabitation holds a distinct place relative to dating and, in principle if not practice, relative to marriage.

Cohabitation has become increasingly widespread in the United States over the past decade. In fact, cohabiting unions are now the modal route to marriage and a common experience in the lives of young people (Cherlin, 2010; Smock, 2000). Such trends have given rise to debates about the relation between cohabitation and marriage. These debates have centered around not only what cohabitation might mean for individual marriage experiences and trajectories (e.g., Manning & Cohen, 2012) but also, more broadly, what the increased prevalence of cohabitation might mean for the future of marriage as an institution (Cherlin, 2004; Heuveline & Timberlake, 2004; The National Marriage Project, 2010; Wilcox & Cherlin, 2011).

In her model of marriage entry, McGinnis (2003) shed some light on these debates by arguing that cohabitation, by affecting the costs and benefits associated with marriage, “appears to significantly change the context in which decisions about marriage are made in romantic relationships” (p. 105). Stanley, Rhoades, and Markman (2006) made a similar argument about the potential for cohabitation to change relational partners’ standpoint with respect to marriage. In particular, their inertia perspective argues that the constraints associated with cohabitation versus dating increase the difficulty of ending a relationship, hence “tipping the scale toward staying together and, for some, marriage” (Stanley et al., 2006, p. 504).

Although these perspectives draw on different theoretical frameworks, both suggest that cohabitation repositions romantic partners with respect to marriage. In the current study we further explored this possibility by examining how cohabitation, relative to both dating and singlehood, is associated with changes in marital beliefs among young African Americans. These marital beliefs—perceived marital costs, perceived marital benefits, the general importance of marriage, and marital salience (the relative importance placed on marriage at the current point in the life course)—tap into several of the multiple dimensions of “marital paradigms” highlighted by Willoughby, Hall, and Luczak (2013), including beliefs about “being married” and beliefs about “getting married.” We addressed two primary questions: (a) To what extent is cohabitation, relative to dating and singlehood, associated with intraindividual change in marital beliefs? and (b) To what extent is the effect of cohabitation gendered?

It is important to note that we tackled these questions using a recent, all–African American sample of young people during the transition to adulthood. As we discuss further below, marriage politics in the United States intersect heavily with racial politics in ways that call for a deeper understanding of marriage, in practice and in principle, in the lives of African Americans. Although such a sample has limitations, it allows for a nuanced investigation of marital perceptions among a population that is often at the center of debates surrounding the perceived declining importance of marriage. Compared to Whites, African Americans are more likely to cohabit rather than marry as a first union, and they are less likely to transition from cohabitation to marriage (Copen, Daniels, & Mosher, 2013; Manning & Smock, 1995). Furthermore, African American cohabiting partners are more likely than White cohabiting partners to share children (Lofquist, Lugaila, O’Connell, & Feliz, 2012). Such indicators suggest that cohabitation might serve as an alternative to marriage more so for African Americans than for Whites (Heuveline & Timberlake, 2004). Although group comparisons were beyond the scope of this study, understanding the extent to which cohabitation is distinctly associated with changes in the marital beliefs of young African Americans will enable a deeper understanding of the meaning and place of cohabitation and of marriage today.

Cohabitation: Increasing the Appeal of Marriage?

As others have noted (e.g. Cherlin, 2010; Smock, 2000; Copen et al., 2013), cohabitation is a relatively common experience in the United States today. By age 25, roughly 55% of young women in the United States (51% of African American women) have experienced a cohabiting union, and by age 30 almost three quarters have done so (Copen et al., 2013). Furthermore, adolescents have begun to place cohabitation in their life course: Roughly one third of adolescents in Toledo, Ohio, indicated that they saw themselves “probably” or “definitely” cohabiting in the future (Manning, Longmore, & Giordano, 2007). As indicated above, the increased prominence of cohabitation has led to both scholarly and popular interest about the future of marriage. Some of this interest concerns the extent to which cohabitation may serve as an alternative to marriage, thereby weakening the institution of marriage (Heuveline & Timberlake, 2004; Wilcox & Cherlin, 2011). Others focus on what, if anything, cohabitation might mean for individual experiences of marriage. For instance, several studies have found a positive association between cohabitation and marital instability or distress (e.g. Jose, O’Leary, & Moyer, 2010; Stanley, Rhoades, Amato, Markman, & Johnson, 2010). This “cohabitation effect,” however, has recently been called into question (Kuperberg, 2014; Manning & Cohen, 2012).

Nonetheless, two models have been put forth that are relevant to debates concerning cohabitation’s effect on both the institution of marriage and individual experiences of marriage. Both of these models, although drawing on different theoretical insights, suggest that the experience of cohabitation, instead of serving as an alternative to marriage, may reposition relational partners with respect to marriage in ways that dating does not. In other words, cohabitation is not an alternative to marriage but makes marriage more salient and probable.

The first of these models is one of marriage entry, put forth by McGinnis (2003). This model posits that cohabitation changes the perceived costs and benefits of marriage, thereby affecting partners’ intentions and expectations to marry and ultimately increasing the likelihood of marriage. In particular, McGinnis argued (and found support for her argument) that cohabitation, relative to dating, reduces both the costs and benefits of marriage (e.g., the number of ways respondents thought their lives would change for the worse/better if they were to get married). Because, she argued, the lower costs perceived by cohabitors should outweigh the greater benefits perceived by daters, cohabitation increases the likelihood of marriage.

Stanley et al. (2006) argued that, instead of affecting the costs and benefits of marriage, cohabitation, more so than dating, orients partners toward marriage simply by affecting the costs associated with ending the relationship. In other words, by virtue of the added constraints associated with cohabitation, cohabitors may be more likely than daters to maintain a relationship, even one of poorer quality, and, perhaps, to marry. Although McGinnis (2003) did not tie her model of marriage entry into the literature on the cohabitation effect, Stanley et al. argued that the inertia of cohabitation may help explain any purported effect of cohabitation, in particular pre-engagement cohabitation, on future marital troubles and instability. Nonetheless, both models contrast sharply with the alternative-to-marriage framework by suggesting that cohabitation repositions relational partners with respect to marriage in ways that make marriage more probable. It is important to note that this reorientation toward marriage is presumed to operate independently of relationship quality.

Race, Gender, and the Appeal of Marriage

Both the inertia model put forth by Stanley et al. (2006) and the model of marriage entry put forth by McGinnis (2003) are general models, yet they have been tested on largely White samples. Although the current study was not a direct test of either model but simply of their shared assertion that cohabitation nudges partners toward marriage, it is important to consider the extent to which this general process might be applicable in the lives and relationships of young African Americans. As mentioned above, research on family behavior indicating the infrequent transitions to marriage among African American cohabitors (Copen et al., 2013; Manning & Smock, 1995), and the higher likelihood of children present in cohabiting unions (Lofquist et al., 2012), suggests that cohabitation might serve as a general alternative to marriage for African Americans (Smock, 2000). If this is indeed the case, we would expect cohabitation to be associated with no change (or even a negative change) in the appeal of marriage.

Contrary to the alternative-to-marriage framework, however, scholarship highlighting the racialized politics of marriage in the United States today suggests that marriage remains distinct from cohabitation, and perhaps even especially so for African Americans. As Cherlin (2009, 2013) has pointed out, the symbolic value of marriage has increased over time, with marriage today now serving as a “symbol of successful self-development” (Cherlin, 2009, p. 140). Critical race and gender scholars have argued that, for African Americans, marriage also serves as a symbol of respectable citizenship (Collins, 1998; Jenkins, 2007; Moon & Whitehead, 2006; Onwuachi-Wilig, 2005); that is, given the historic pathologization of African American sexuality and African American families (Collins, 1990, 1998, 2005; Jenkins, 2007), heterosexual marriage is imbued with substantial symbolic significance for African Americans (see Collins, 1998, 2005; Moon & Whitehead, 2006 for reviews), given that it provides a means by which to “claim the mantle of Black respectability” (Collins, 2005, p. 253).

These respectability politics surrounding marriage are not only racialized but also gendered. For instance, both Chasteen (1994) and Sharp and Ganong (2007, 2011) have shown that women often problematize singlehood and that marriage, or at least the prospect of it, may afford women a sense of normalcy. Chasteen argued that this is the case because men provide women the symbolic capital of simply “looking less ‘out of place’ to others,” of “appear[ing] ‘normal’ and appropriate” in their everyday environments (p. 322). Similarly, Huang, Smock, Manning, and Bergstrom-Lynch (2011) reported that women, but not men, noted the social disapproval of cohabitation and viewed marriage as the more legitimate and legitimating relationship status.

Taken together, this work on the gendered and racialized symbolic significance of marriage suggests that, although cohabitation might appear an alternative to marriage among African Americans based on studies that have examined cohabitation and marital behavior, marriage remains a distinct and desirable status relative to cohabitation. This work leads one to expect that the general claim made by McGinnis (2003) and Stanley et al. (2006) that cohabitation repositions partners toward marriage should be applicable not only to Whites but also to young African Americans, if not more so. This work also suggests, however, that the effects of cohabitation on perceptions of marriage may be gendered. Given the greater symbolic significance of marriage for women than for men, cohabitation may enhance the appeal of marriage for women more so than for men.

Summary and Hypotheses

As mentioned previously, in this study we did not assess the move toward marriage in terms of marital behavior but in terms of four marital beliefs: (a) marital costs, (b) marital benefits, (c) general marital importance, and (d) marital salience. Hence, the increased appeal of marriage would be indicated by an increased perception of marital benefits, importance, and salience and a decreased view of marital costs. If cohabitation serves as an alternative to marriage for young African Americans, then entering a cohabiting union should not be associated with increased marital appeal. If, however, cohabitation nudges partners toward marriage, as McGinnis (2003) and Stanley et al. (2006) have claimed, one would expect that cohabitation will enhance the salience and favorability of marriage. Cohabitation should be associated with reduced perceptions of marital costs and increased perceptions of marital benefits, importance, and salience. This may be especially the case for women compared to men. Formally stated, these hypotheses are as follows:

• Hypothesis 1: Cohabitation, relative to both dating and singlehood, will be associated with a negative change in perceived marital costs and a positive change in perceived marital benefits, marital importance, and marital salience for young African Americans.

• Hypothesis 2: Cohabitation will increase the appeal of marriage (i.e., more favorable marital perceptions) more strongly for young African American women than for young African American men.

A caveat is in order at this point. Hypothesis 1 is, by and large, consistent with Stanley et al.’s (2006) notion of inertia and McGinnis’s (2003) model of marriage entry. An exception, however, is that in the current study we hypothesized a positive association between cohabitation and perceived marital benefits, whereas McGinnis hypothesized and found a negative association between cohabitation and perceived marital benefits. The current hypothesis concerning marital benefits diverges from McGinnis’s largely because the benefits being assessed are of a different kind. Whereas the National Survey of Families and Households (NSFH; Sweet & Bumpass, 2002) used by McGinnis asked respondents to report how they expected that their lives would change if they were to get married, the Family and Community Health Study (FACHS), the data we used in this study, asked respondents to report about the general benefits of marriage (e.g. the extent to which they agreed that “marriage leads to a happier life”). Hence, the benefits attended to here capture the more general, symbolic benefits of marriage rather than the tangible, practical benefits (e.g., changes in economic security and standard of living) attended to by McGinnis.

In addressing these hypotheses, we expand on and overcome some of the limitations inherent in our current understanding of how cohabitation might reposition one with respect to marriage. First, with the use of fixed effects models, we assessed intraindividual change in marital beliefs over time, something McGinnis (2003) was unable to do. On a related note, the data we used contained relationship quality indicators across multiple waves and for respondents in both dating and cohabiting relationships. This allowed for the examination of the cohabitation effect independent of relationship quality, something Stanley et al. (2006) argued is essential. Finally, given recent evidence that the link between cohabitation and marriage may be gendered (e.g., Huang et al., 2011), we attended to the potentially gendered effects of cohabitation on marital beliefs.

Method

Data

We used data from the FACHS to test the above hypotheses. The FACHS is a multisite, longitudinal research study of more than 800 African American youth (the target respondents) and their family members living in Iowa and Georgia at the study’s initiation. Targeted youth were in the fifth-grade public school system at the time of recruitment in 1997. Unlike more commonly used studies containing significant numbers of African Americans (e.g. Fragile Families; see Reichman, Teitler, Garfinkel, & McLanahan, 2001), the FACHS was designed to capture the diversity of African American families and the variety of communities in which they live. Hence, youth and their families were drawn from school rosters across a variety of communities that differed in racial composition and economic level within each state (sampling and data collection procedures have been described in much greater detail elsewhere; e.g., Simons et al., 2002).

At the first wave of the FACHS (1997–1998), data were gathered on 889 target children (average age at Wave 1 = 10.5) and their family members. Subsequent waves were completed every 2 to 3 years thereafter, with the sixth and most recent wave of data collection occurring more than a decade after the study’s initiation, in 2010 and 2011. At this latest wave, 699 target respondents, now in their early-to-mid 20s, participated in the study (78.6% of the original sample; average age at Wave 6 = 23.6). In its entirety, then, the FACHS captures the experiences of African American youth from late childhood through the early years of the transition to adulthood. Given the current interest in romantic relationship experiences during this transition, the latest three waves of FACHS data, beginning at Wave 4 in late adolescence (collected in 2007–2008; average age = 18.8), were used in this study. Because respondents who did not participate in one wave of data were still contacted for participation at later waves, 793 of the original 889 target respondents (89.2%) participated in at least one of the later three FACHS waves. Across the course of the FACHS, there is some evidence of selective attrition with respect to gender and delinquency: Participants in the later waves were more likely than Wave 1 participants to be female, and they showed slightly less childhood delinquency, than their nonparticipating peers. There has been little evidence of selective attrition, however, with respect to other factors, such as community characteristics, family structure, and parenting practices.

Although not nationally representative, respondents in the FACHS sample are similar to national estimates of African American young adults on measures of fertility (U.S. Census Bureau, 2011b) and educational attainment (U.S. Census Bureau, 2011a, 2011c). This is, perhaps, because the FACHS was intended to capture heterogeneity among African American youth, their families, and the communities in which they live. Furthermore, the FACHS provides several advantages over more representative data sets for an examination of relationship status and marital beliefs among African Americans. First, unlike in the NSFH the data were collected relatively recently, providing an assessment of cohabitation and marital beliefs among a cohort of young African Americans for whom cohabitation is normative. Given rapidly changing family patterns (Cherlin, 2010; Smock & Manning, 2010), associations found in more representative yet older data sets might not be relevant to young African Americans today (Smock, Casper, & Wyse, 2008). Second, the FACHS data were collected every 2 to 3 years, allowing for a more nuanced analysis of change in marital beliefs throughout the transition to adulthood than other data sets that are based on waves separated by much longer periods (e.g., Add Health; see Harris et al., 2009). Finally, the FACHS examines multidimensional aspects of both relationship quality and marital beliefs for respondents across all relationship statuses. Like other representative studies, the NSFH, for instance, contains extensive relationship quality measures for married and cohabiting respondents but not for dating respondents, contains cost/benefit measures only for those in relationships, and lacks a measure of marital salience.

Our original sample, drawn from the most recent three waves of data, included 2,102 observations (person-years) from 793 respondents. Excluding person-years with missing values for key study variables and person-years in which respondents reported being married, the final sample consisted of 1,989 observations from 780 unmarried respondents. As noted below, however, this analytic sample varied slightly across outcomes.

Measures

Dependent Variables

Perceived benefits of marriage

The perceived benefits of marriage were assessed at each wave via a two-item index indicating the degree to which respondents think marriage brings “happiness” and a “fuller life.” Potential responses ranged from 1 (strongly disagree) to 5 (strongly agree). Items were summed to form a measure of perceived marital benefits. Cronbach’s alphas across waves ranged from .69 to .87.

Perceived costs of marriage

The perceived costs of marriage were assessed at each wave via a four-item index indicating the degree to which respondents associate marriage with a loss of friends, a loss of freedom, a worse sex life, and a harder life. Potential responses ranged from 1 (strongly disagree) to 5 (strongly agree). Cronbach’s alphas ranged from .67 to .77.

General marital importance

The general importance of marriage at each wave was assessed with one item, which asked, “How important is it to you to have a good marriage?” Potential responses ranged from 1 (not at all important) to 5 (extremely important).

Marital salience

The current salience of marriage at each wave was assessed with one item that inquired about the extent to which respondents agreed that “Getting married is the most important part of my life.” Potential responses ranged from 1 (strongly disagree) to 5 (strongly agree). Unlike the costs, benefits, and general marital importance measures, current marital salience addresses what Willoughby et al. (2013) identified as “beliefs about getting married” rather than “beliefs about being married.” The current marital salience item tapped the relative importance of marriage during the transition to adulthood, which is an essential aspect in the recently developed marital horizon theory (Carroll et al., 2007). This item, then, is thought to be distinct from more general, culturally prescribed marital attitudes indicative in beliefs about being married, and, in particular, in general marital importance. Hakim (2003) referred to this distinction as one between “personal choice” and “public morality” attitudes. We discuss the validity of this distinction further below.

Independent Variable: Union Type

At each wave, union type was assessed via an item that asked respondents to best describe their current relationship status. Those who reported currently living with a partner but not being married to that partner were coded as cohabiting, those who reported being in a nonmarital romantic relationship but not living with their partner were coded as dating, and those who reported that they were not romantically involved with someone on at least “a regular basis” were codes as being single. Using a one-wave lagged indicator of engagement, we excluded respondents who reported being engaged to help rule out the possibility that respondents altered their views on marriage, became engaged, and then cohabited.

Time-Varying Control Variables

Relationship quality

Given that the cohabitation effect on one’s orientation toward marriage is expected to operate independently of relationship quality, it was necessary to control for quality. We used three subscales—relationship satisfaction, partner warmth, and partner hostility (reverse coded)—to assess relationship quality for both cohabiting and dating respondents. Relationship satisfaction was assessed with two questions about respondents’ overall satisfaction and happiness with their romantic relationship. Because potential responses varied across these two items, we standardized the items and then averaged them to form an index of relationship satisfaction. Partner warmth was assessed via three questions that asked how often in the past month one’s romantic partner had acted “loving and affectionate,” helped the respondent do something that was important to him or her, and expressed appreciation. These three items were averaged to form the index of relationship warmth. Finally, partner hostility was assessed via five items that asked how often in the past month one’s romantic partner had been verbally or physically abusive (e.g. shouted at, insulted or cursed at, slapped or hit). These five items were first reverse coded and then averaged to form the index of (lack of) relationship hostility. The satisfaction, warmth, and reverse-coded hostility subscales were then standardized and summed to create an index of relationship quality at each wave. Using Nunnally’s (1978) formula for the linear combination of measures, the reliability of this composite index was .82 at Wave 4, .86 at Wave 5, and .84 at Wave 6.

The measure of relationship quality was then recoded into an “internal moderator” by first standardizing the composite index to have a mean of 0 and a standard deviation of 1 and then assigning single respondents a score of 0. This process is not equivalent to assigning missing values the mean score; instead, it results in a relationship quality index that shows variation among partnered respondents and no variation among single respondents. In essence, the coefficient for relationship quality indicates the extent to which relationship quality conditions the effect of being in a relationship. This coefficient is irrelevant to respondents who are not in a relationship but allows one to make comparisons that conventional coding schemes do not allow (Frech & Williams, 2007; Mirowsky, 1999).

Additional controls

In addition to relationship quality, the following analyses included several time-varying control variables that, if left unattended to, may confound the associations between cohabitation and marital beliefs. These variables included school enrollment (in school = 1); parental status (parent = 1); current employment (full-time employment [≥ 35 hours/week] = 1); and a relatively crude measure of economic distress, recent unemployment (experienced unemployment in past year = 1). The indicators of current employment and recent unemployment are not mutually exclusive. We also controlled for whether or not respondents’ single status was the result of a breakup between waves (breakup between waves = 1). This control variable helps reduce the likelihood that any association between cohabitation/dating and changing marital beliefs is due to changes that occur from exiting a relationship. Furthermore, given evidence of heterogeneity among cohabitors with respect to the timing and status of engagement (Stanley et al., 2006), we controlled for a lagged indicator of engagement. Finally, we controlled for the fixed effects of time by entering dummy variables for wave into the models (reference period = Wave 4). These time effects are necessary when special events (e.g., the Great Recession) may affect the outcome variable, or when there is significant variation in the outcome by time period. Preliminary models suggested that the variation by time period was significant, and hence a control for time was necessary. As we describe further below, fixed effects models examining intraindividual change were used, making time-invariant control variables (e.g. religion, gender, family background, state recruited from, etc.) unnecessary.

Plan of Analysis

The above hypotheses were tested via a series of fixed effects models (performed in Stata 12). Fixed effects models are “designed to study the causes of changes within a person [or entity]” (Kohler & Kreuter, 2009, p. 245) and therefore assess the relation between changes in relationship status (single, dating, cohabiting) and changes in marital beliefs within each individual. This focus on intraindividual change attempts to rule out omitted-variable bias and potential selection mechanisms by controlling for observed and unobserved time-invariant factors; that is, given that all time-invariant factors (e.g., gender and family background) are constant across waves, changes in the dependent variable over time are attributed to changes in the time-varying factors, such as relationship status (Stock & Watson, 2007). Because time-invariant factors are held constant in fixed effects models, the direct effect of these factors cannot be estimated. It is possible to test for the interaction between time-varying variables (e.g., relationship status) and time-invariant variables (e.g., gender) by entering the main effect of the time-varying predictor and the interaction effect between the time-varying and time-invariant predictors into the model (Allison, 2009). This was the approach we took to examine whether changes in relationship status were associated with changing marital beliefs differently for young men and women.

Fixed effects models for continuous variables, like our measures of marital benefits and costs, are fairly straightforward. The conventional unconditional maximum likelihood estimator with robust standard errors was used. The coefficients for relationship status in these fixed effects models indicate the intraindividual unit change in marital beliefs that accompanies a change in relationship status across the transition to adulthood. Fixed effects models for ordinal variables, like our measures of general marital importance and marital salience, are more computationally demanding. Because neither the maximum likelihood estimator nor the conventional conditional likelihood estimator used for fixed effects logit models is appropriate for ordinal outcomes, we used the “blow-up and cluster” (BUC) estimator, proposed by Mukherjee, Ahn, Liu, Rathouz, and Sánchez (2008) and developed by Baetschmann, Staub, and Winkelmann (2011). The BUC estimator has been shown to provide reliable fixed effects estimates on ordered outcomes in small samples, to be less sensitive to the number of panel waves, and to perform better than the conventional approach of dichotomizing an ordinal variable at only one cutpoint (Baetschmann et al., 2011). When exponentiated, the coefficients produced by this estimator can be interpreted as proportional odds ratios, with numbers greater than 1 indicating a positive change in the dependent variable and numbers less than 1 indicating a negative change associated with a unit change in the independent variable. It is important to note that fixed effects ordered logit models, like fixed effects logit models in general, can be performed only on data from respondents whose outcome is not constant across waves. Hence, as will become evident in our presentation of the results, the analytic sample for the ordered logit models was smaller than that for the linear regression models.

For each of the four outcomes, the analyses proceeded in several steps. First, we considered the main effects of relationship status by entering the time-varying cohabiting and dating indicators, along with the time-varying controls, into the model. Because this first model compared cohabiting and dating to singlehood, differences between cohabiting and dating statuses were assessed via an incremental likelihood-ratio (LR) chi-square test. We conducted this test by comparing the unconstrained model in which the cohabiting and dating coefficients were allowed to vary with an alternative, constrained model in which cohabiting and dating statuses were combined into one partnered status, thus constraining their coefficients to be equal. A significant test statistic indicates that the cohabiting and dating coefficients significantly differed from one another, or that the unconstrained model was a better fitting model. After determining main effects for relationship status, we assessed the moderating role of gender by entering its interaction with relationship characteristics into the model. For outcomes in which there was evidence of gendered effects, the models were separated by gender and are discussed as such.

Results

Descriptive Statistics and Preliminary Findings

The descriptive statistics for all study variables are presented in Table 1. As shown in the table, women made up slightly more than half of the sample at each wave. The percentage of respondents enrolled in school declined across waves, from about two thirds at Wave 4 to about half at Wave 6, while parenthood and full-time employment increased across waves. By Wave 6, 42% of respondents were parents, and 44% held full-time jobs. The percentage of respondents who experienced unemployment in the past year held relatively steady across waves at about 50%. Consistent with the work of Barr, Culatta, and Simons (2013), relationship-related variables, including relationship status, quality, and marital beliefs, were not consistently linear in their changes across time.

Table 1.

Descriptive Statistics for All Study Variables

Variable	Wave 4				Wave 5				Wave 6
	M	SD	Min	Max	M	SD	Min	Max	M	SD	Min	Max
Gender (1 = female)	.56	.50	.00	1.00	.57	.50	.00	1.00	.58	.49	.00	1.00
School enrollment (1 = in school)	.66	.47	.00	1.00	.50	.50	.00	1.00	.51	.50	.00	1.00
Parental status (1 = parent)	.17	.38	.00	1.00	.28	.45	.00	1.00	.42	.49	.00	1.00
Full-time employment (1 = ≥ 35 hr per week)	.27	.44	.00	1.00	.42	.49	.00	1.00	.44	.50	.00	1.00
Recent unemployment	.43	.49	.00	1.00	.51	.50	.00	1.00	.50	.50	.00	1.00
Recent breakup					.17	.37	.00	1.00	.17	.37	.00	1.00
Engagement (t − 1)					.02	.15	.00	1.00	.06	.24	.00	1.00
Cohabitation status (1 = cohabiting)	.06	.23	.00	1.00	.14	.35	.00	1.00	.11	.32	.00	1.00
Dating status (1 = dating)	.47	.50	.00	1.00	.38	.49	.00	1.00	.41	.49	.00	1.00
Relationship qualitya	0.00	2.23	−11.53	3.71	−0.01	2.28	−10.07	3.48	−0.03	2.31	−8.49	3.63
Perceived benefits of marriage	7.08	1.55	2.00	10.00	7.47	1.98	2.00	10.00	7.22	1.89	2.00	10.00
Perceived costs of marriage	11.58	2.64	4.00	20.00	11.27	3.17	4.00	20.00	11.10	2.97	4.00	20.00
General marital importance	4.31	0.97	1.00	5.00	4.39	1.03	1.00	5.00	4.13	1.11	1.00	5.00
Marital salience	2.89	1.05	1.00	5.00	2.98	1.19	1.00	5.00	2.66	1.13	1.00	5.00

Note. Average at was 18.8 at Wave 4, 21.6 at Wave 5, and 23.6 at Wave 6. Min = minimum; Max = maximum.

^aFor descriptive purposes, relationship quality was not coded as an internal moderator and is representative of those in relationships only.

In general, respondents’ perceptions of marital benefits were high (M = 7.2 on a scale ranging from 2 to 10), whereas their perceptions of marital costs were more neutral (M = 11.3 on a scale ranging from 4 to 20) and declined over time. Furthermore, respondents averaged a score of 4.3 on a scale ranging from 1 to 5 on general marital importance (between “very important” and “extremely important”) but only a 2.8 on current marital salience (between “disagree” and “neutral or mixed”). The four marital beliefs were significantly, but only weakly to moderately, correlated with one another (correlation matrix available on request). The strongest correlation (r = .48) was between general marital importance and marital benefits, whereas general marital importance and marital salience were correlated only .34, supporting the theoretical and empirical distinction between these two items and, more broadly, between beliefs about being married and beliefs about getting married. It is also worth noting that although women perceived fewer benefits to marriage than did men, they also perceived fewer costs and reported lower marital salience. Gender was not significantly associated with general marital importance.

Fixed Effects Models

The results of the fixed effects regression models predicting the four marital beliefs are displayed in Table 2. For all outcomes except marital salience, the models are separated by gender because interactive models indicated that results were weaker for men than for women. Given space constraints, we forego a lengthy discussion of all control variables and focus only on the relationship quality indicator. As can be seen in Table 2, relationship quality tended to enhance the favorability of marriage across all outcomes. Given the coding of this variable as an internal moderator, this meant that the estimated relationship status effects we discuss below are generally enhanced by the quality of the relationship. Keeping this in mind, we turn now to a discussion of these relationship status effects directly related to our hypotheses.

Table 2.

Fixed Effects Regression Models Predicting Marital Salience and Marital Costs From the Relationship Experiences of Unmarried Respondents

Predictor	Marital saliencea (full sample)	General marital importancea		Marital benefitsb		Marital costsb
		Women	Men	Women	Men	Women	Men
	Model 1 (n = 525)	Model 2 (n = 244)	Model 3 (n = 196)	Model 4 (n = 430)	Model 5 (n = 350)	Model 6 (n = 430)	Model 7 (n = 350)
Time-varying relationship status
Cohabiting	1.96*c	5.61***c	2.11†	0.29c	0.36	−0.57c	−0.34
	(0.27)	(0.40)	(0.45)	(0.24)	(0.27)	(0.40)	(0.48)
Dating	1.44*c	2.51***c	2.30**	−0.08c	0.45*	−1.14***c	−0.41
	(0.18)	(0.27)	(0.28)	(0.16)	(0.19)	(0.24)	(0.31)
Time-varying controls
Quality (IM)	1.19†	2.19***	1.41*	0.14†	0.31**	−0.49***	−0.52*
	(0.10)	(0.20)	(0.14)	(0.08)	(0.10)	(0.10)	(0.20)
Recent breakup	1.08	1.58	0.88	−0.18	−0.05	−0.40	0.27
	(0.22)	(0.34)	(0.39)	(0.22)	(0.28)	(0.37)	(0.42)
Engaged (t − 1)	1.23	1.41	0.24*	0.19	−0.16	−0.08	0.82
	(0.34)	(0.60)	(0.66)	(0.31)	(0.39)	(0.48)	(0.80)
In school	0.69**	1.11	0.87	0.01	−0.20	−0.24	0.32
	(0.14)	(0.21)	(0.24)	(0.14)	(0.19)	(0.20)	(0.28)
Parent	1.19	1.02	0.92	0.03	−0.32	−0.04	0.47
	(0.17)	(0.28)	(0.31)	(0.17)	(0.24)	(0.24)	(0.40)
Full-time employment	0.80†	1.07	0.67	−0.16	−0.18	0.36†	0.14
	(0.14)	(0.22)	(0.26)	(0.13)	(0.17)	(0.21)	(0.27)
Unemployment in past year	0.99	1.09	0.83	−0.00	−0.17	0.11	0.17
	(0.13)	(0.22)	(0.24)	(0.12)	(0.15)	(0.20)	(0.21)
Wave 5 (ref. = Wave 4)	1.15	1.71*	0.92	0.47***	0.35*	−0.61**	0.21
	(0.12)	(0.23)	(0.22)	(0.12)	(0.16)	(0.19)	(0.25)
Wave 6 (ref. = Wave 4)	0.57***	0.60**	0.66	0.26*	0.16	−0.55**	−0.23
	(0.13)	(0.20)	(0.26)	(0.12)	(0.16)	(0.19)	(0.28)

Note. Numbers in parentheses are robust standard errors. IM = internal moderator; ref. = reference category.

^aExponentiated regression coefficients (proportional odds ratios) are presented.

^bUnstandardized regression coefficients are presented.

^cDating and cohabiting coefficients significantly differ from one another at p < .05.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001 (two-tailed).

As shown in Model 1, cohabitation and dating relative to singlehood were associated with increased marital salience. With all other variables held constant, the odds of marriage being more salient nearly doubled (an increase of 96%) in person-years spent cohabiting versus being single. As indicated by the significant dating coefficient, the salience of marriage also differed between dating and single statuses, but alternative comparisons via LR chi-square tests indicated that the effect of cohabitation on marital salience was greater than the effect of dating (LR χ² = 3.69, p < .05); That is, cohabitation was associated with an increase in marital salience relative to both dating and singlehood. It is important to note that these estimated effects of relationship status were independent of relationship quality and did not differ by gender.

For general marital importance, marital benefits, and marital costs, estimated cohabitation effects proved stronger for women than for men, and hence in Table 2 the models are separated by gender. Regarding general marital importance, Models 2 and 3 of the table showed a significant and positive effect of both dating and cohabiting (vs. singlehood) for both women and men. In other words, being in any romantic relationship appears to increase the general importance of marriage. For women (Model 2), cohabiting unions were associated with nearly a six-fold increase, and dating unions nearly a threefold increase, in general marital importance (e^{^b}_cohabiting = 5.61, p < .001; e^{^b}_dating = 2.51, p < .001). For men (Model 3), cohabiting and dating unions both about doubled the odds of placing greater general importance on marriage (e^{^b}_cohabiting = 2.11, p < .10; e^{^b}_dating = 2.30, p < .01). Post hoc tests revealed that the cohabiting and dating coefficients did indeed differ significantly for women (LR χ² = 8.35, p < .01), but they did not differ significantly for men (LR χ² = 0.07, p > .10). In other words, for men, cohabiting and dating relationships equally enhanced general marital importance, but for women cohabiting relationships enhanced general marital importance more so than dating relationships.

These gendered distinctions between the estimated effects of cohabitation and dating were also found for perceived marital benefits. As shown in Model 4 (see Table 2), for women, neither cohabitation nor dating was associated with a change in marital benefits compared to singlehood. Cohabitation did, however, significantly enhance women’s perceived marital benefits relative to dating (LR χ² = 5.67, p < .05). This was not the case for men (Model 5), given that cohabitation and dating did not significantly differ in their estimated effects on men’s perceived marriage benefits (LR χ² = 0.23, p > .10).

With the exception of gendered effects, the results for the three outcomes discussed thus far have proven fairly consistent. With few exceptions, cohabitation has been associated with enhanced marital salience, general marital importance, and greater marital benefits relative to both dating and singlehood, and this has been especially the case for women. In predicting marital costs, however, Models 6 and 7 of Table 2 revealed a somewhat unique pattern of results. For women, dating was associated with a reduction in perceived marital costs relative to singlehood, and this reduction was also significantly greater than that of cohabitation (LR χ² = 5.69, p < .05). For men, perceived marital costs were unaffected by changes in relationship status: Dating and cohabiting unions did not differ from singlehood or from one another (LR χ² = 0.05, p > .10) in their effect on perceived marital costs.

In sum, for women, cohabitation was distinct from dating across all outcomes, but its estimated effects were not consistent in that it increased not only the salience, importance, and benefits of marriage but also the perceived costs of marriage. For men, cohabitation was distinct from dating only in its greater effect on marital salience. Such results provide at least partial support for both study hypotheses.

Post Hoc Propensity Models

Although they use data across multiple waves and examine within-person effects, fixed effects models require the assumption of a particular causal ordering. Guided by theory, the presentation of our results thus far has presumed the causal order to be one in which cohabitation led to changes in marital beliefs. Although we took several steps to rule out reverse causation (e.g., controlling for the relationship exits), we used a propensity score approach to evaluate our causal ordering further. To do so, Waves 4 and 5 were used (N = 541) and a variable was constructed that indicated whether respondents had entered a cohabiting union between these two waves. This variable was coded 1 if a respondent reported not cohabiting at Wave 4 but cohabiting at Wave 5. It was coded 0 otherwise. Respondents who reported cohabiting at Wave 4 or being married at Wave 4 or 5 were removed from the analysis so that those who entered a cohabiting union were being compared and matched with those who either entered no union or entered a nonmarital and noncohabiting union. These treatment and control groups were successfully matched (via nearest neighbor matching with a caliper of .05) on a number of variables at Wave 4, including prior levels of respective marital beliefs, age, education level, gender, school enrollment, employment status, parental status, and whether they were involved in a dating relationship. Entering a cohabiting union was associated with more than a twofold increase in current marital salience (e^{^b} = 2.12, p < .001, one-tailed), nearly a twofold increase in general marital salience (e^{^b} = 1.67, p < .05, one-tailed), and about a half-point increase in perceived marital benefits (b = .41, p < .05, one-tailed). It was not significantly associated with a change in marital costs (b = −.024, p > .10, one-tailed). Although these propensity models could not attend to gender differences and relationship quality effects given sample size limitations, the results supported the causal ordering assumed in our fixed effects regression models.

Discussion

In the current study, we sought to examine the extent to which cohabitation might reposition African American relationship partners toward marriage by affecting their perceptions of marital costs and benefits as well as the general importance they attribute to marriage and its salience during the transition to adulthood. In general, with the exception of marital costs for women, the results supported the notion that cohabitation changes marital beliefs in ways that enhance the salience of marriage and accentuated, rather than minimized, its potential benefits. Of note is that this increased appeal of marriage took place regardless of the quality of the relationship, although higher quality relationships tended to enhance its effect. Such findings imply that, instead of serving as a deterrent to or a replacement for marriage, cohabitation appears to direct relational partners’ focus toward marriage.

These effects of cohabitation were gendered in ways that are largely consistent with past research. With the exception of marital salience, cohabitation was indistinguishable from dating in its effects on men’s marital beliefs. Both of these statuses, however, were generally distinguishable from singlehood, indicating that, for men, it may be the experience of simply having a romantic partner, rather than the type of relationship, that produces a shift toward more favorable perceptions of marriage. For women, however, cohabitation and dating were consistently distinguishable in their effects on marital beliefs. These specific gendered findings are not surprising given evidence both that the legitimating value, the symbolic capital, of marriage is greater for women than men (Chasteen, 1994; Sharp & Ganong, 2007), a symbolic capital apparently not provided to women by cohabitation (Huang et al., 2011). It is important to note that, despite these gendered findings, cohabitation tended to enhance the current salience of marriage equally for men and women. Furthermore, the young African American men in the FACHS sample actually scored slightly higher than young African American women on the measure of current marital salience. Such patterns cannot be overlooked, given that they contradict popular depictions and academic accounts of the prospect of marriage mattering very little to African American men (Boswin, 2008; Edin & Nelson, 2013). How it matters and how it matters differently for African American women and both women and men in other racial groups needs further exploration.

Although not a core aspect of the study, the second finding to which more attention should be paid is that, across all outcomes for both men and women, relationship quality, in both cohabiting and dating unions, was positively associated with more favorable marital perceptions. Consonant with the work of Simons and colleagues (Simons, Simons, Lei, & Landor, 2011), relationship quality, in addition to relationship status, appears to be an essential element in understanding changes in marital beliefs.

Not only are such findings relevant for understanding the link between cohabitation and marriage, but they also shed important insights on broader sociological debates about the relationship between attitudes and behavior. They highlight the changing nature of marital beliefs across the life course, as well as the multidimensionality of these beliefs. As Willoughby (2010) pointed out, research that assesses only the predictive power of marital beliefs in shaping future behavior implicitly assumes that such beliefs are relatively unchanging. Willoughby refuted this assumption by showing that marital beliefs change significantly across adolescence. The current findings present another challenge to this implicit assumption by revealing that marital beliefs continue to be responsive to social conditions throughout the transition to adulthood for African Americans. This was the case across multiple dimensions of marital beliefs, including marital salience, general marital importance, and marital costs and benefits. It is important to note, however, that although the general pattern of change was consistent across these different marital beliefs there were some noteworthy differences by outcome. Current marital salience, for instance, was the only marital belief measured in the present study that was responsive to changes in nonrelationship factors, such as respondents’ schooling situation. This finding is consistent with Carroll et al.’s (2007) contention about the utility of distinguishing between more general marital beliefs (i.e. “public morality”; Hakim, 2003) and more life course–specific (i.e. “personal choice”; Hakim, 2003) marital beliefs across the life course. The current findings support this multidimensionality of marital beliefs, but it is worth noting that even general, or public morality, beliefs about marriage were subject to change across the relatively short time span we analyzed.

The multidimensionality of marital beliefs also is illustrated in the finding that, for young women, cohabitation, relative to dating, tended to enhance the benefits, salience, and importance of marriage while also enhancing the perceived costs. These findings regarding costs and benefits counter those of McGinnis (2003) yet still support her general claim underlying her model: that cohabitation repositions partners toward marriage in ways that dating does not. The differences in cohabitation’s effects on the perceived costs and benefits of marriage may be attributable to the fact that cohabitation and costs/benefits of marriage were measured at only one time point in McGinnis’s model, making an analysis of change impossible. Likewise, as mentioned earlier, measures of costs and benefits differed across studies. Whereas McGinnis assessed respondents’ perceptions about how their lives would change upon getting married, we assessed the more general costs and benefits associated with marriage. These elements capture more generalized “beliefs about being married” (Willoughby et al., 2013) rather than expected life changes that might occur with marriage. Nonetheless, the fact that cohabitation, compared to dating, enhanced both the perceived benefits and perceived costs of marriage for women in the FACHS sample suggests that these young women related cohabitation to marriage differently depending on the dimension being assessed; that is, although cohabitation might remain distinct from marriage in terms of its perceived benefits, this may not be the case in terms of its perceived costs.

Although the above findings are interesting in their own right, they are especially remarkable given that we used data from an all–African American sample of young people. Although comparisons of the effect of cohabitation on marital beliefs by race/ethnicity cannot be drawn, one might have expected cohabitation to matter very little in changing marital beliefs for young African Americans. This is so because scholars have postulated that cohabitation may serve as more of an alternative to marriage for African Americans than it does for Whites (see Smock, 2000, for a brief review), and African Americans have consistently been shown to be much less likely than Whites to convert cohabiting unions into marital unions (Copen et al., 2013; Manning & Smock, 1995). Given these patterns, the model of marriage entry extended by McGinnis (2003) and the process of inertia identified by Stanley et al. (2006) may have been thought to be relatively inapplicable to African Americans. Although such models, with their ultimate focus on marital behavior, may still prove to be less applicable to African Americans than to Whites, the general argument of these models—that cohabitation reorients relational partners’ toward marriage—holds true in the current sample of African American young adults, at least when this reorientation is measured by changing marital beliefs. It is imperative that future work examine this paradox and its implications for cohabiting couples. The most obvious question may be why, if cohabitation causes a reorientation toward marriage, do fewer than one third of first cohabiting unions among African American women transition to marriage (Copen et al., 2013)? Material resources, which prove important in explaining the gap between marital attitudes and behavior in general (e.g., Gibson-Davis, Edin, & McLanahan, 2005), undoubtedly play a large role in answering this question. Perhaps a more interesting line of inquiry, then, might be how the added salience of marriage brought on by cohabitation affects the well-being of cohabiting partners and the stability of their unions, in particular among those who lack the real or perceived resources deemed necessary for marriage by young people today (Cherlin, 2009, 2013; Cherlin, Cross-Barnet, Burton, & Garrett-Peters, 2008).

Our findings and their implications are not without limitations. Although the all–African American sample we used in this study was relevant given racialized demographic trends and politics surrounding marriage and family formation behaviors, this sample necessarily restricts the study’s generalizability. Given its sampling design, however, the FACHS did allow us to capture heterogeneity among African American young people not only in marital attitudes but also in community context, family background, relationship experiences, and personal resources. This heterogeneity has often been obscured in popular discourses about cultural values. Second, given that respondents averaged about 24 years of age by Wave 6 of the study, marriage was an uncommon phenomenon, and hence questions about the ways in which marital beliefs might mediate the link between cohabitation and marriage, marital quality, and marital stability could not be attended to here. Third, given the small number of respondents who were engaged to be married, we could not attend to more nuanced analyses examining the potentially differential effects of cohabitation on marital beliefs by engagement status. Larger sample sizes would allow for a more in-depth investigation of this heterogeneity among cohabitors with respect to changing marital beliefs. Fourth, the data used here were individual rather than dyadic. Given that relationship processes are inherently dyadic, and that recent research suggests that both partners’ orientation toward marriage is predictive of relationship outcomes (Willoughby, Carroll, & Busby, 2012), future research assessing the relational nature of marital beliefs and their effects is needed. Additional research using multi-item indices of marital salience and general marital importance also is needed. Finally, although we attempted to establish causal ordering between relationship status and marital beliefs via several avenues, as well as relying on recent research indicating that favorable marital attitudes may actually deter cohabitation for women (Huang et al., 2011), changes in relationships and marital beliefs are likely not abrupt but gradual and co-occurring. Data with shorter time spans between waves may help to explicate this nuanced process, which likely includes reciprocal effects unable to be captured here.

Nonetheless, by examining intraindividual change in marital beliefs over time, the current study indicates that cohabitation among African American young adults tends to reposition relational partners—in particular, women—toward marriage rather than away from it. Although Stanley et al. (2006) argued that such repositioning may result in partners “sliding” into marriage, the transition from cohabitation to marriage appears to be the exception rather than the norm for African Americans (Copen et al., 2013). Other implications of this repositioning remain to be seen. Inquiring about them holds promise for enhancing our understanding about the meaning and role of marriage in the lives of young people today.