Relationship Quality in Midlife: A Comparison of Dating, Living Apart Together, Cohabitation, and Marriage

Abstract

Objective:

This study compared the relationship quality of U.S. midlife adults in dating, living apart together (LAT) relationships, cohabitation, and marriage.

Background:

Unmarried partnerships are gaining ground in midlife but how these partnerships compare to each other and to marriage is unclear. From an incomplete institutionalization perspective, those in unmarried relationships, especially LAT relationships but also cohabitations, face challenges due to unclear relationship norms and expectations which may eventuate in poorer relationship quality than that of the married. Alternatively, cohabitation and, by extension, LAT relationships, offer flexibility and autonomy and thus may function as an alternative to marriage marked by comparable relationship quality.

Method:

Data were drawn from the 2013 Families and Relationships Survey, a nationally representative survey of U.S. adults. The analytic sample was composed of adults aged 50–65 in a partnership (N=2,166). Multivariable models compared the associations between relationship type (dating, LAT, cohabiting, and married) and relationship quality (happiness, support, commitment, disagreement, and instability).

Results:

The incomplete institutionalization perspective was supported for LATs, who tended to report poorer relationship quality than marrieds. For cohabitors, this perspective received mixed support. Although cohabitors reported less happiness and commitment than marrieds, which aligned with the incomplete institutionalization perspective, the groups did not differ on relationship support, disagreement, or instability, supporting the cohabitation as an alternative to marriage perspective. Dating, LAT and cohabiting relationships were remarkably alike.

Conclusion:

This study has implications for understanding the shifting landscape of relationships in midlife which in turn may shape individual health and well-being.

The retreat from marriage is evident across the life course. Even among midlife adults, a declining share is married these days. One in three U.S. Baby Boomers was unmarried in 2009 compared with just 20% of midlife adults in 1980 (Lin & Brown, 2012). Yet, being unmarried does not necessarily mean going solo. Many unmarried midlife and older adults are partnered, whether through cohabiting, dating, or living apart together (LAT) relationships. In the U.S., the number of cohabiting adults aged 50 and older has more than quadrupled since 2000 (Stepler, 2017) and dating is actually more common than cohabitation (Brown & Shinohara, 2013). LAT relationships appear to be gaining ground, but empirical estimates are lacking (Benson & Coleman, 2016; Connidis, Borell, & Karlsson, 2017; Strohm et al., 2009). Increasingly, midlife adults who are partnered are not married.

Despite the growth in unmarried partnerships during midlife, it remains unclear how these various relationship types operate. The rise in unmarried partnerships among midlife adults is a recent phenomenon and thus these partnerships are incompletely institutionalized, meaning the norms and expectations governing these relationships are fuzzy (Cherlin, 1978; Nock, 1995). The uncertainty surrounding these relationship norms foretells poorer relationship quality for daters, LATs, and cohabitors than marrieds. Alternatively, unmarried partnerships are desirable precisely because of their flexibility, which may bolster their quality (Connidis et al., 2017; Lewin, 2017). In fact, there is mounting evidence that cohabitation in later life may operate as a long-term substitute for marriage (Brown & Wright, 2017; King & Scott, 2005) with cohabiting and remarried older adults enjoying largely comparable relationship quality (Brown & Kawamura, 2010; Wright, 2019). But the relationship quality dynamics characterizing dating and LAT relationships and how these compare with cohabiting and marital unions are largely unknown (Lewin, 2017).

Our goal is to examine the relationship quality of midlife partnered adults differentiating among those in dating, LAT, cohabiting, and married relationships. The incomplete institutionalization perspective leads us to expect that those in unmarried relationships—whether dating, LAT, or cohabiting—have poorer relationship quality than married individuals because unmarried relationship types remain relatively rare and ill-defined (Benson & Coleman, 2016; Upton-Davis, 2012). LAT relationships may be marked by the poorest relationship quality because they are arguably the least well-defined type of partnership. In fact, a central feature of LAT relationships according to Connidis et al. (2017) is ambivalence, which reflects the tension between maintaining autonomy and being in a committed union. Alternatively, the independence afforded to partners in unmarried relationships, particularly LAT relationships, may help to minimize partner disagreement and strain (Lewin, 2017). Likewise, cohabitation in later life appears to be akin to marriage (King & Scott, 2005) and prior research documents similar relationship quality for the two groups (Brown & Kawamura, 2010; Wright, 2019). To the best of our knowledge, our study is the first to address the full spectrum of partnered relationships, explicitly recognizing the diversity of partnership options available today. Consequently, it informs conceptual work on the meaning and purposes of these various relationship types by illuminating the extent to which they are characterized by comparable quality.

Background

The shrinking share of married midlife adults coincides with the emergence of an array of unmarried partnership options. It is also likely an artifact of the deinstitutionalization of marriage, which appears to be affecting the family behaviors of not just young adults, but middle-aged and older adults, too. The norm of lifelong marriage is waning (Wu & Schimmele, 2007), particularly during midlife as the rate of divorce among this age group has doubled since 1990 (Brown & Lin, 2012). This trend portends growth in unmarried adults who are eligible to form either a nonmarital union or a remarriage. Although the remarriage rate has remained stable among midlife adults over the past few decades, the proportion cohabiting has climbed, signaling a growing preference for nonmarital unions among midlife adults who may be reluctant to make a legally binding tie to their partners (Brown, Bulanda, & Lee, 2012; Vespa, 2013).

Between 2000 and 2016, the percentage of unmarried midlife U.S. adults (ages 50–64) who were cohabiting doubled from 7% to 14% (authors’ calculations using CPS data) and the number of cohabiting adults aged 50 and older skyrocketed from less than 1 million to 4 million (Stepler, 2017). The majority of older adult cohabitors have been previously married (Stepler, 2017), underscoring the role that the gray divorce revolution may play in the growth in cohabitation during the second half of life.

Apart from cohabitation, midlife adults may choose to form a non-co-resident partnership, whether dating or living apart together (LAT). About 18% of unmarried U.S. adults aged 57–64 had a romantic or sexual non-co-resident partner, according to data from 2005–2006 National Social Life, Health, and Aging Project (NSHAP) (Waite, Laumann, Levinson, Lindau, & O’Muircheartaigh, 2014). The proportions dating differed considerably by gender with 27% of unmarried men versus just 7% of unmarried women aged 57–64 reporting a dating relationship (Brown & Shinohara, 2013). How these contemporary levels of midlife dating compare to earlier decades is unknown.

Likewise, estimates of the prevalence of LAT relationships are hard to come by as prior research in the U.S. has conflated dating and LAT relationships (Lewin, 2017; Strohm et al., 2009). We conceptualize LAT relationships as distinct from dating (Connidis et al., 2017; Duncan & Phillips, 2011, 2010; Upton-Davis, 2012). Dating relationships are at an early stage of relationship progression and may transition into either cohabitation or marriage. Or, the dating couple could break up. In contrast, LAT relationships during the second half of life are longer-term, more committed unions that typically represent a conscious decision not to cohabit or marry (Wu & Brown, 2021).

The landscape of midlife partnerships is varied, ranging from marriage to cohabitation, LAT, and dating relationships (Brown & Wright, 2017). Demographic trends, such as the rise of gray divorce (Brown & Lin, 2012), point to sustained growth in unmarried partnerships in the coming decades as fewer midlife adults are married (Cooney & Dunne, 2001). Cohabitation continues to accelerate among adults over age 50 (Stepler, 2017). At the same time, LAT relationships are particularly appealing to those in the second half of life (Connidis et al., 2017) and dating is common among unmarried older adults (Brown & Shinohara, 2013). Although a growing literature addresses cohabitation during the second half of life, the literature on LAT and dating relationships is comparatively sparse. The shift away from marriage towards unmarried partnerships in midlife raises new questions about the quality and dynamics of these various partnership types.

The Incomplete Institutionalization Perspective

The incomplete institutionalization perspective has guided much of the research on relationship quality and stability differentials. This perspective, which Cherlin (1978) first introduced four decades ago to explain the higher instability of remarriages relative to first marriages, also has been applied to cohabiting unions (Nock, 1995). Remarriage and cohabitation are arguably incomplete institutions that lack clearly defined norms and expectations for the relationship. Unlike first marriage, for which there are widely shared expectations for spousal roles, couples in remarriage and cohabitation must actively negotiate and construct these roles. Establishing relationship norms and expectations can be contentious, ultimately undermining couple relationship quality and stability.

Yet, as cohabitation has diffused across the population, it has become necessarily less selective and arguably more institutionalized, which may help to solidify cohabiting relationships (Liefbroer & Dourleijn, 2006). Moreover, marriage has become deinstitutionalized, according to Cherlin (2004), blurring the boundaries between cohabitation and marriage. Recent comparisons of the relationship quality of cohabitors and marrieds reveal considerable similarity, with comparable quality evident among the two largest segments: cohabitors with plans to marry and marrieds who premaritally cohabited. Individuals who married directly report relatively high relationship quality whereas cohabitors without plans to marry are in unions of relatively poor quality (Brown, Manning, & Payne, 2015).

Cohabitation as an Alternative to Marriage

The resemblance between cohabitation and marriage is pronounced during the second half of life. Research focusing specifically on older adults shows that cohabitors and marrieds enjoy similar relationship quality (Brown & Kawamura, 2010; Wright, 2019). The two groups do not differ in their reports of relationship satisfaction, pleasure, openness, and time spent together. Cohabitors do report lower levels of relationship happiness than their married counterparts, on average. At the same time, cohabitors report less criticism by and fewer demands from their partners than do married individuals, although these differentials are accounted for by sociodemographic factors (Brown & Kawamura, 2010). Thus, older cohabitors and married individuals fare similarly across multiple dimensions of both positive and negative relationship quality.

In fact, the consensus among researchers is that during later life, cohabitation operates as a substitute for marriage (Brown & Wright, 2017; King & Scott, 2005). Cohabiting unions tend to be of long duration, averaging nine or ten years, and are unlikely to eventuate in marriage or to break up. Rather, dissolution most often occurs through partner death (Brown et al., 2012). Cohabitation offers couples many of the benefits of marriage, including a close, intimate co-resident partnership, without the constraints. Individuals who want to preserve their financial autonomy or who may lose material benefits such as a pension or Social Security through marriage are likely to find cohabitation particularly attractive (Chevan, 1996). Women may prefer cohabitation over marriage to avoid the gendered caregiving obligations that marriage has traditionally entailed (McWilliams & Barrett, 2014; Talbott, 1998; Watson & Stelle, 2011). Older adults are nearly as likely to form a cohabiting union as they are to marry (Brown et al., 2012; Vespa, 2012).

LAT and Dating Relationships

A new frontier in partnered relationships is living apart together (Connidis et al., 2017). If cohabitation was once construed as radical because it involved living together outside of marriage, LAT relationships widen the boundaries of partnership still further by removing the element of co-residence. Couples do not have to be married or even live together to achieve a long-term, committed union. Now, couples can be together for the long haul while maintaining their own residences, preserving their freedom and carving out their own levels of partner involvement and interdependence. The LAT relationship is itself an endpoint, and as such is unlikely to eventuate in either cohabitation or marriage (Benson & Coleman, 2016; Haskey & Lewis, 2006; Regnier-Loilier, Beaujouan, & Villeneuve-Gokalp, 2009; Wu & Brown, 2021).

LAT relationships are clearly incompletely institutionalized. Partners often struggle to define and agree upon the ambiguous boundaries of their relationships to achieve their desired balance between intimacy and autonomy (Benson & Coleman, 2016). Some LAT partners are ambivalent about their relationship arrangements and still others actively oppose them, preferring instead to live together but remaining in a LAT configuration due to their partner’s unwillingness to co-reside (Benson & Coleman, 2016; Liefbroer, Poortman, & Seltzer, 2015). However, many other LAT partners strongly favor their arrangement, valuing the flexibility it affords. Much like cohabitation, LAT relationships are likely an alternative to marriage for midlife adults, allowing them to avoid the responsibilities and obligations that accompany co-residence and marriage (Connidis et al., 2017; Liefbroer et al., 2015). LAT couples can arguably retain greater autonomy in their unions than can either cohabitors or marrieds. Benson and Coleman (2016, p. 808) concluded that “LAT afforded partners the ability to be in a committed relationship by choice rather than due to obligation.” LAT partners did not subscribe to a til-death-do-us-part mentality, but they insisted they were highly committed to their partners. Many expressed difficulties reconciling high commitment with low obligation, a scenario that is at odds with a traditional, marital-based system to intimate partnerships (Benson and Coleman, 2016). The ambivalence that is emblematic of LAT relationships (Connidis et al., 2017) is a manifestation of its incomplete institutionalization, which may create strain for some partners as they attempt to establish and embrace their roles as LATs.

Alternatively, LATs may experience similar or possibly even lower levels of relationship strain than cohabitors and marrieds because they prefer their flexible relationship circumstances even if they are marked by ambivalence. In a recently published examination of relationship quality among older adults aged 57–84, Lewin (2017) compared a group she classified as LATs to first married, remarried, and cohabiting individuals. Her measure of LATs effectively combined dating and LAT relationships as all respondents who reported being in a sexual or romantic relationship were categorized as LATs. She found that LATs (and daters) reported less strain in their relationships than married or cohabiting older adults. It seems that the weaker obligations in LAT (and dating) relationships result in lower relationship strain, which is contrary to the incomplete institutionalization perspective. Still, Lewin (2017) also showed that LATs (and daters) were less happy in their relationships than were either cohabitors or marrieds. The ambiguous boundaries of LAT relationships may diminish positive dimensions of relationship quality while minimizing negative dimensions. The confounding of dating and LAT relationships clouds our understanding of these two partnership types. Our study teases this apart by explicitly comparing the quality of dating and LAT relationships across both positive and negative dimensions.

It is unclear how the relationship quality of daters and LATs compare. Daters are a heterogeneous group that are presumably at an earlier stage of relationship progression. Some daters will eventually break up, signaling relatively poor relationship quality, but others will segue into LAT, cohabitation, or even marriage, implying high relationship quality.

The Present Study

The current investigation is designed to examine the relationship quality of midlife adults across the full range of partnership types, including dating, LAT relationships, cohabitation, and marriage. As fewer midlife adults are married and levels of cohabitation surge among midlife adults, it is important to establish how various married and unmarried partnerships compare in terms of relationship quality, which in turn is integral for health and well-being (Carr & Springer, 2010; Umberson, et al., 2006). From an incomplete institutionalization perspective, which emphasizes the relationship challenges inherent in navigating novel union forms, we expect those in unmarried relationships to report poorer quality than those who are married. Presumably, LATs face even greater incomplete institutionalization than cohabitors and thus we anticipate the relationship quality of LATs is lower, on average, than that of cohabitors.

Then again, there is mounting evidence that cohabitation, at least, is largely comparable to marriage during the second half of life (Brown & Wright, 2017; King & Scott, 2005). Prior work reveals no marked differences in the relationship quality of older cohabitors and marrieds, which is at odds with the incomplete institutionalization perspective. It signals that cohabitation may have a unique meaning for those in midlife, functioning as an alternative to marriage. Likewise, LAT relationships, in which couples are highly committed but disinterested in cohabitation or marriage, may be akin to cohabitation (and, by extension, marriage) in terms of relationship quality. If LAT relationships are not so different from cohabitation and marriage in terms of their function and purpose then their quality should be comparable, too. However, this is partially at odds with Lewin’s (2017) study, which showed that LATs report lower levels of positive relationship quality than either cohabitors or marrieds. LATs did report comparable levels of the negative dimensions of relationship quality in her study (Lewin, 2017). This unique pattern for positive versus negative dimensions of quality may be an artifact of the conflation of LATs and daters. Or, it may be that LATs simply fare worse on some (i.e., positive) dimensions of relationship quality but not others (i.e., negative) in comparison with cohabitors and marrieds. Regardless, it underscores the value of distinguishing between positive and negative dimensions of relationship quality, as researchers have long understood that one is not the inverse of the other (i.e., values on both positive and negative dimensions can be similar) (Johnson, White, Edwards, & Booth, 1986).

In addition to testing competing hypotheses about whether LAT relationships are of comparable or worse quality than cohabitations and marriages, we also assess whether the relationship quality of LATs is equivalent to or exceeds that of daters. Dating relationships are likely to vary in terms of quality and commitment. New insights on whether and how LAT and dating relationships differ in terms of relationship quality can inform our understanding of the extent to which these two relationship types are distinctive.

Our study establishes associations between relationship type and relationship quality but does not permit us to draw causal conclusions. In fact, relationship quality may be endogeneous to relationship type. For example, individuals in especially high-quality cohabiting unions may be more likely to marry (Brown, 2000), leaving those still cohabiting with lower overall relationship quality. Similarly, some LATs may have ambivalence about their relationships, which deters them from entering cohabitation or marriage (Benson, 2016). More broadly, an individual’s willingness to progress from one relationship type to the next likely depends in part on the quality of the relationship (Ross, 1995). Unfortunately, we cannot disentangle this endogeneity problem because we only have measures of relationship type and relationship quality at a single time point.

To minimize the potential role of endogeneity, our models include controls for sociodemographic factors related to partnership type and relationship quality. Women tend to report poorer quality than men (Bulanda, 2011). We also consider whether gender interacts with partnership type since women are much less likely than men to form partnerships in middle and later life. The gender gap in relationship quality could be larger in marriage than cohabitation and LAT relationships since the former implies greater obligation and caregiving responsibilities, which could diminish women’s quality more so than men’s. Alternatively, marriage offers some distinct benefits, such as shared economic resources (e.g., health insurance), that may reduce any gender gap in relationship quality. Nonwhite respondents are especially likely to be cohabiting or dating and unlikely to be married relative to white respondents. Further, white men and women tend to report higher levels of relationship quality than their nonwhite counterparts. Cohabitors are younger than marrieds, on average, but how they compare with LATs and daters is unclear (Brown & Wright, 2017). Age is negatively associated with relationship quality. Older married individuals have higher education levels than older cohabitors (Brown & Wright, 2017). Older adult daters tend to be well-educated compared to those without any partner (Brown & Shinohara, 2013) but how educated they are relative to those in other partnership types is unknown. Education is not closely associated with relationship quality among older adults (Brown & Kawamura, 2010; Lewin, 2017). Marrieds enjoy higher household incomes than cohabitors. Presumably LATs and daters have lower household income since they do not co-reside with their partners. Home ownership is higher among marrieds than cohabitors (Brown et al., 2016). We control for living in a metro area because residence type may be an indicator of acceptance of diverse living arrangements. One reason why many older adults form a LAT relationship rather than cohabit or marry is because they have a resident child (de Jong Gierveld & Merz, 2013). Likewise, daters also might be especially likely to have a child living with them. Children are associated with poorer relationship quality. Relationship duration is likely to be longest in marriage, followed by cohabitation, LAT, and dating relationships. Relationship duration is inversely associated with relationship quality (Wright, 2019). Many midlife adults have prior marital experience, which may shape both their current relationship type and their quality of that relationship. By testing interactions between relationship type and prior marital experience, we can assess whether having been previously married modifies how individuals appraise their relationships. It also allows us to evaluate whether the quality of first marriages differs from remarriages, although we do not expect these two groups to appreciably differ based on prior work (Skinner, Bahr, Crame, & Call, 2002).

Method

Data came from the Families and Relationships Survey (FRS), a nationally representative survey of 7,517 U.S. adults ages 18–65 that was fielded in 2013. The FRS was modeled largely on the 1987–88 National Survey of Families and Households (NSFH) to permit analyses of family change over the past 25 years but it also includes new items that were not part of the NSFH, such as a novel LAT measure. The FRS survey was designed by the National Center for Family and Marriage Research (NCFMR) at Bowling Green State University, which sponsored the data collection performed by GfK Group (formerly Knowledge Networks) using their nationally representative online panel sample. GfK established the online research panel (called KnowledgePanel) based on probability sampling covering both the online and offline U.S. population. KnowledgePanel used a dual sampling frame composed of random digit dialing sampling and addressed-based sampling to randomly recruit a probability-based sample. This sampling frame included listed and unlisted phone numbers, telephone and non-telephone households, and cell-phone-only households. GfK provided hardware and internet access to panel members if needed. Other social science studies have relied on the GfK (formerly KN) panel, which has been used in federally funded data collections on couples and families (Lichter & Carmalt, 2009; Rosenfeld & Thomas, 2012; Sassler, Addo, & Lichter, 2012). The data quality of the GfK (KN) panel is comparable to or even exceeds that derived from RDD surveys (Chang & Krosnick, 2009).

The FRS is well-suited for our study because it includes several indicators of positive (happiness, support, and commitment) and negative (disagreement and instability) relationship quality. Researchers have demonstrated the utility of assessing both positive and negative dimensions as they are not simply the inverse of one another but rather are often positively correlated (Johnson, et al., 1986). Additionally, the FRS data include an innovative, original item to measure LAT relationships as distinct from dating relationships. Here, we focus on midlife adults ages 50–65 (n=2,942) to compare the relationship quality of daters, LATs, cohabitors, and married respondents. Of these midlife respondents, 776 were excluded because either they were not in a romantic relationship (711) or they were missing valid information about partnership status (65), yielding an analytic sample of 2,166 respondents. There were 70 in dating relationships, 92 in LAT relationships, 265 in cohabiting unions, and 1,739 in marriages (with 627 in remarriages and 1,112 in first marriages).

Measures

Dependent Variables

We gauged both positive and negative relationship quality since they are distinct sets of domains (Johnson et al., 1986). Positive relationship quality was captured using three indicators. Happiness was a continuous variable measured by the respondent’s response to the item “Taking thing all together, how would you describe your current relationship?” Responses ranged from 1 = Very Unhappy to 7 = Very Happy. Support was tapped by the sum of responses to three items: “My [partner] encourages or helps me to do things that are important to me,” “My [partner] shows love and affection toward me,” and “My [partner] listens when I need someone to talk to.” Each item ranged from 1 = Strongly Agree to 5 = Strongly Disagree. All items were reversed coded such that higher values signaled greater support. By summing these three items, relationship support ranged from 3 to 15. The Cronbach’s alpha was quite high at 0.86. Commitment was a single item measure that gauged how committed the respondent was to the current relationship with his/her partner. Values ranged from 1 = Not at all Committed to 5 = Completely Committed.

Negative relationship quality included two distinct dimensions. Disagreement was a three-item measure that tapped the respondent’s level of agreement with these statements: “My [partner] is fair and willing to compromise when we have a disagreement,” “My [partner] and I are good at working out our differences,” and “My [partner] tends to insult or criticize me or my ideas.” For each item, response values ranged from 1 = Strongly Agree to 5 = Strongly Disagree. The third item was reverse coded. The three items were summed to create the disagreement measure. Higher values corresponded to greater disagreement. The Cronbach’s alpha for disagreement was 0.81. Instability was a single item measure based on the question “During the past year, have you ever thought that your relationship might be in trouble?” It was coded as 0 = No and 1 = Yes.

Focal Independent Variable

Relationship type distinguished among those who were in dating, LAT (reference category), cohabiting, and married relationships. Daters and LATs were captured using a combination of two questions. All daters and LATs reported being in a relationship but not living with their boyfriend, girlfriend, or partner. Then, individuals who either strongly agreed or agreed with the statement that “Nowadays, many couples are in a committed, long-term relationship and choose to live apart (maintaining separate residences) rather than cohabit or marry. This describes my current relationship with my partner” were classified as being in a LAT relationship whereas those who neither agreed nor disagreed, disagreed, or strongly disagreed with the statement were coded as dating. Respondents were cohabiting if they reported currently living with a boyfriend, girlfriend, or partner in the household. Respondents who indicated they were married were coded as such.

Control Variables

Age was a continuous variable measured in years. Gender was coded 0 = Man and 1 = Woman. Education was a categorical variable: high school or less (reference category), some college, and a college degree or more. Race-ethnicity was coded as: non-Hispanic White (reference category), non-Hispanic Black, and Hispanic. Home ownership was a dummy variable coded 0 = Not a Homeowner and 1 = Homeowner. The presence of children was a dummy variable that gauged the presence of least one minor child in the household (0 = No and 1 = Yes). Metropolitan status indicated whether the respondent lived in a metropolitan area (0 = No and 1 = Yes). Household income included five categories: less than $25,000 (reference category), $25,000–39,999, $40,000–74,999, $75,000–99,999, and $100,000 or more. Relationship duration was a continuous variable measured in years. Previously married tapped whether the respondent had experienced at least one dissolved marriage (0 = No and 1 = Yes). Missing values on the control variables were rare (15 cases on homeownership and 116 cases on relationship duration) and were handled by recoding to the modal (homeownership) or mean (relationship duration) value according to relationship type. Missing data on at least one control variable was positively associated with identifying as Black (versus White) and poorer relationship quality.

Analytic Strategy

Our first step was to examine the distributions of the study variables by relationship type to gain some initial insights into how the characteristics of individuals in the various partnership groups compared. Next, we estimated two models for each of the five dimensions of relationship quality. The initial model included only the relationship types. The second model introduced the control variables to assess whether factors associated with relationship type and quality accounted for the bivariate differences established in the first model. The models we present used LATs as the reference group. Additional analyses were conducted using cohabitors as the reference group with significant differences indicated in Tables 2 and 3 by a superscript a. These multivariable models should not be construed as causal but rather as indicative only of correlational associations. Happiness, support, and disagreement were continuous variables and thus OLS multiple regression was used to estimate these models. For commitment, a five-category measure, the modeling strategy was ordinal logistic regression. The models for instability, a dummy variable, were estimated using logistic regression. Finally, we tested for interactions between gender and relationship type to determine whether the association between relationship type and quality differed for women and men. Additional interactions were tested between prior marital experience and relationship type to assess whether relationship quality differed according to whether individuals had been previously married as well as for those in remarriages versus first marriages. All analyses were weighted to ensure the sample represented the population.

Table 2.

Models Predicting Three Dimensions of Positive Relationship Quality

	Happiness		Support		Commitment
	OLS (b) Model 1	OLS (b) Model 2	OLS (b) Model 1	OLS (b) Model 2	OR Model 1	OR Model 2
Relationship Status
Daters	−0.43	−0.40	0.47	0.39	0.41*a	0.45a
Cohabitors	−0.06	−0.09	1.02*	0.92*	1.21	1.16
Marrieds	0.50***a	0.40** a	0.86*	0.77*	4.49***a	3.66***a
(ref: LATs)
Age		−0.01		−0.03		0.97
Woman (ref: Man)		−0.08		−0.15		1.16
Education
Some College		−0.07		−0.15		0.83
Bachelors+		−0.09		0.07		0.92
(ref: H.S. or less)
Race/Ethnicity
Black, Non-Hispanic		0.07		0.16		0.59*
Hispanic		0.003		0.22		1.40
(ref: White, Non-Hispanic)
Owns home (ref: No)		0.06		−0.06		1.03
One Child or More <18 in HH (ref: None)		−0.24*		−0.53**		0.76
Metro (ref: non-Metro)		0.08		−0.04		1.03
Household Income
$25,000--$39,999		0.02		0.47		0.94
$40,000--$$74,999		0.18		0.50		1.003
$75,000--$$99,999		0.14		0.83*		1.03
$100,000 or more		0.16		0.64*		1.09
(ref: Less than $25,000)
Duration of Relationship (years)		0.005		−0.001		1.01
Previously Married (ref: Never Married)		0.10		0.28		1.04
_cons	5.67***	5.86***	11.52***	13.00***
cut1_cons					0.04***	0.01***
cut2_cons					0.10***	0.02***
cut3_cons					0.30***	0.06**
cut4_cons					1.24	0.27
N	2146	2146	2113	2113	2151	2151
R2 / Pseudo R2	0.04	0.05	0.01	0.03	0.06	0.07

^*p<.05

^**p<.01

^***p<.001

Note: Superscript "a" indicates significant differences (p < .05) when cohabitation is the reference group.

Table 3.

Models Predicting Two Dimensions of Negative Relationship Quality

	Disagreement		Instability
	OLS(b) Model 1	OLS(b) Model 2	Odds Ratios Model 1	Odd Ratios Model 2
Relationship Status
Daters	−0.64	−0.58	0.72	0.78
Cohabitors	−0.64	−0.58	0.60	0.66
Marrieds	−1.04**a	−0.97**	0.32***a	0.0.50*
(ref: LATs)
Age		0.02		0.97
Woman (ref: Man)		−0.13		1.08
Education
Some College		0.11		1.18
Bachelors+		−0.19		1.45
(ref: H.S. or less)
Race/Ethnicity
Black, Non-Hispanic		−0.42*		0.72
Hispanic		−0.12		0.85
(ref: White, Non-Hispanic)
Owns home (ref: No)		−0.05		0.99
One Child or More <18 in HH (ref: None)		0.42*		1.19
Metro (ref: non-Metro)		0.01		1.18
Household Income
$25,000--$39,999		0.07		0.78
$40,000--$$74,999		−0.16		0.57*
$75,000--$$99,999		−0.40		0.55
$100,000 or more		−0.29		0.55*
(ref: Less than $25,000)
Duration of Relationship (years)		−0.00		0.99
Previously Married (ref: Never Married)		−0.14		1.23
_cons	7.00***	6.26***	0.64	3.22
N	2079	2079	2064	2064
R2 / Pseudo R2	0.01	0.02	0.02	0.04

^*p<.05

^**p<.01

^***p<.001

Note: Superscript "a" indicates significant differences (p < .05) when cohabitation is the reference group.

Results

Table 1 shows the distributions or means of the study variables by relationship type. The weighted distribution of the sample by relationship type reveals the prevalence of unmarried partnerships during midlife. Although most midlife adults in a partnership were married (82.9%), unmarried partnerships accounted for about 17% of all partnerships. We estimated that roughly 5.4% of partnered midlife adults were in a LAT relationship compared with 3.6% who were dating. Another 8.2% of partnered midlife adults were cohabiting. In other words, similar shares of midlife adults were in cohabiting relationships as non-coresidential relationships (9.0% total were LATS or daters). Nearly one-third (31%) of unmarried partnerships in midlife were LAT relationships. Close to one-half (48%) were cohabiting unions. The remaining 21% were dating relationships. This pattern delineated the variation among unmarried partnerships, underscoring the importance of differentiating among dating, LAT, and cohabiting relationship types.

Table 1.

Weighted Means or Percentages of Study Variables by Relationship Type

	Daters	LATs	Cohabitors	Marrieds
Relationship Quality
Happiness	5.2	5.7	5.6	6.2
Support	12.0	11.5	12.5	12.4
Commitment	3.5	4.1	4.1	4.7
Disagreement	6.4	7.0	6.4	6.0
Instability (%)	31.6	39.0	27.6	16.9
Demographic Characteristics
Age	56.5	56.3	55.6	57.0
Gender
Man	51.7	51.3	56.4	49.7
Woman	48.3	48.7	43.6	50.3
Education
H.S. or less	38.4	32.9	50.8	47.6
Some College	31.7	45.9	33.7	25.5
Bachelors+	30.0	21.2	15.5	26.9
Race/Ethnicity
White, Non-Hispanic	62.4	73.2	70.8	79.7
Black, Non-Hispanic	28.2	13.9	21.1	10.6
Hispanic	9.4	12.8	8.1	9.8
Owns Home
Yes	79.0	92.0	69.3	89.8
No	21.1	8.0	30.7	10.2
Children Living in Household
No child(ren) 18 or younger in HH	92.6	88.7	87.0	80.8
At least one child <18 in HH	7.4	11.3	13.0	19.2
MSA Status
Metro	71.9	81.9	82.0	81.8
Non-Metro	28.1	18.1	18.1	18.2
Household Income
Less than $25,000	24.9	34.3	31.5	7.2
$25,000--$39,999	33.4	25.1	17.4	12.4
$40,000--$74,999	21.6	17.4	20.6	25.5
$75,000--$99,999	7.6	16.3	15.8	20.0
$100,000 or more	12.5	6.9	14.8	34.9
Duration of Relationship (years)	3.6	7.9	11.3	26.3
Previously Married
Yes	48.2	53.6	61.3	36.5
No	51.8	46.4	38.7	63.5
Unweighted N	70	92	265	1,739
Weighted %	3.6	5.4	8.2	82.9

Across all three positive dimensions of relationship quality, it appeared that married individuals enjoyed higher average levels of quality than those in unmarried relationships. The gap between LATs and cohabitors was negligible for two of the three positive dimensions; the two groups reported nearly identical levels of happiness (5.7 and 5.6, respectively) and commitment (4.1 and 4.1, respectively). Support levels were a full point lower for those in LAT (11.5) than cohabiting (12.5) relationships. In fact, LATs also reported less support than those who were dating (12.0).

Disagreement and instability levels were highest among LATs and lowest among married individuals. The average score on the disagreement scale for LATs was 7.0 versus 6.4 for both cohabitors and daters. For marrieds it was much lower at 6.0. Nearly two in five (39%) LATs thought their relationship could be in trouble during the past year compared with 27.6% of cohabitors and 31.6% of daters. Here again, instability was considerably less common for the married at 16.9%.

The average age of midlife partnered adults in our sample hovered around 55–57 years. The gender composition across partnership type was largely comparable although cohabitors were disproportionately men at 56.4%. Education levels were lowest among cohabitors with just 15.5% holding a bachelor’s degree, followed by LATs at about 21%, marrieds at nearly 27% and finally daters at 30%. Relationship types differed in terms of their racial-ethnic composition with marrieds (79.7%) overwhelmingly identifying as White and somewhat lower percentages of LATs (73.2%) and cohabitors (70.8%) identifying as White. Black respondents were disproportionally dating (28.2%) rather than married (10.6%), LAT (13.9%) or cohabiting (21.1%). A greater share of LATs (12.8%) were Hispanic respondents than any other relationship type. Homeownership was lowest among cohabitors at 69.3% and stood at 79% among daters. Homeownership among LATs was quite high at 92% and even exceeded that of marrieds at 89.8%. Relatively few partnered midlife adults had a minor child in the household. This scenario was most common for marrieds (19.2%) and least common for daters (7.4%); LATs (11.3%) and cohabitors (13.0%) fell in between. The vast majority of midlife partnered adults resided in metro areas with about 80–83% of all relationship types, except daters. Daters were less often living in metro areas at just 71.9%. Household income levels appeared to be lowest among LATs. Just 6.9% reported incomes over $100,000 per year. Rather, the modal income category for LATs was less than $25,000 which characterizes one-third (34.3%) of all LATs. Although nearly one-third (31.5%) of cohabitors were in this lowest income bracket, there also were 14.8% in the highest bracket (over $100,000). Despite not having a resident partner, midlife daters were not far behind cohabitors with 12.5% reporting incomes over $100,000. About one-quarter (24.9%) of daters were in the lowest income bracket. Married individuals were concentrated in the upper income bracket (34.9% reported household incomes over $100,000) and underrepresented in the lowest bracket (7.2% reported less than $25,000 in household income). Relationship duration varied by relationship type. Midlife daters had been together for about 3.6 years, on average. The mean relationship duration for LATs was about twice as long at 7.9 years. For cohabitors, average duration was 11.3 years. Relationship duration was much higher, on average, for marrieds (26.3 years). Prior marital experience differed by relationship type. Most marrieds were in first marriages; just 36.5% had been married previously (and thus are currently in a remarriage). About half of daters (48.2%) and LATs (53.6%) were previously married. A majority of cohabitors (61.3%) were formerly married.

Table 2 shows the multivariable results for the three dimensions of positive relationship quality. On balance, the incomplete institutionalization perspective received the most support across these dimensions. LATs were less happy in their relationships than married respondents, although they were comparable to other unmarried respondents (both cohabitors and daters). Cohabitors reported lower levels of happiness than marrieds (indicated by superscript a). These patterns persisted in Model 2 with the inclusion of the control variables. Only one control variable was significantly associated with relationship happiness. Individuals with a child at home were less happy, on average, than their counterparts not living with a minor child. The results for relationship happiness were largely consistent with the incomplete institutionalization perspective in that married individuals enjoyed higher quality than their counterparts in unmarried relationships.

There were notable differences in relationship support by relationship type. Relative to LATs, cohabiting and married individuals reported higher levels of support, on average. Respondents in coresidential relationships reported greater levels of support than those who were not living with their partners, providing some evidence for the incomplete institutionalization perspective (but not for cohabitation). The support reported by LATs and daters was comparable. Cohabitors and marrieds enjoyed similar levels of relationship support (indicated by the absence of a superscript a), which aligned with the alternative to marriage perspective. The introduction of controls in Model 2 did not alter the pattern of associations between relationship type and support shown in Model 1. Having a child in the household was negatively associated with relationship support. Higher household income was linked to greater levels of support.

A unique pattern emerged for commitment. LATs were more committed than were daters, which is what we would have expected based on our measure of LAT membership. Likewise, cohabitors reported higher levels of commitment than daters. LATs and cohabitors were comparable in terms of commitment. In turn, both LATs and cohabitors were less committed than married individuals. These findings were largely unchanged with the inclusion of the control variables; only the differential between daters and LATs just barely (p=0.05) failed to achieve statistical significance. The only control that was related to commitment was race-ethnicity. Black respondents tended to report lower levels of commitment than White respondents. These findings aligned with the incomplete institutionalization perspective.

Turning now to the two negative dimensions of relationship quality, we uncovered partial support for the incomplete institutionalization perspective. The models shown in Table 3 reveal that the levels of disagreement experienced by LATs were similar to those of daters and cohabitors. Disagreement was higher, on average, for LATs than married respondents, which was contrary to our expectations. Cohabitors experienced greater disagreement than marrieds although this reduced to nonsignificance in the full model. The other differentials remained robust to the inclusion of controls. Blacks reported less disagreement than Whites and the presence of a child in the home was positively related to relationship disagreement. These findings only partially supported the incomplete institutionalization perspective because although LATs had more disagreement than marrieds, the levels of disagreement did not appreciably differ between cohabitors and marrieds, supporting the alternative to marriage perspective.

Similar to the disagreement findings, instability levels were comparable across unmarried partnership types (LATs were just as likely to report their relationship had been in trouble as daters and cohabitors). LATs were more likely to report being in an unstable relationship than married respondents. Although in the initial model (Model 1), cohabitors were more likely than marrieds to report relationship instability, the two groups shared similar odds of reporting their relationship had been in trouble during the past year in the full model with controls (Model 2). Higher household incomes were associated with lower odds of instability. Again, we saw partial support for the incomplete institutionalization perspective for relationship instability in that LATs reported greater relationship instability than marrieds. But we also uncovered evidence that was consonant with the alternative to marriage perspective in that cohabitors and marrieds experienced comparable levels of relationship instability.

Additional models were estimated to test whether the associations between relationship type and relationship quality differed for women and men. None of the gender by relationship type terms was significant (results not shown). Similarly, we uncovered no evidence that the effects of relationship quality differed by prior marital experience. None of the interaction terms between relationship type and previously married achieved significance (results not shown). Thus, the pattern of findings obtained for marrieds applies for both those in first marriages and remarriages (replacing the previously married indicator with two separate dummies for remarrieds and first marrieds in Tables 2 and 3 yielded no significance differences between the two groups of marrieds [results not shown]). In supplemental analyses, we examined whether relationship type operated differently according to relationship duration or the presence of a child in the household. Again, no significant interaction terms emerged (results not shown).

Discussion

Our goal was to consider a broad spectrum of relationships during midlife, including daters, LATs, cohabitors, and marrieds. Using one of the only recent data collections that includes a measure of LAT relationship status, the Families and Relationships Survey, we compared individuals across relationship types on multiple indicators of positive and negative relationship quality. The incomplete institutionalization perspective led us to anticipate that particularly LATs but also cohabitors would report poorer relationship quality than marrieds, reflecting the lack of consensus on the norms and expectations governing unmarried partnerships (Nock, 1995). Our enthusiasm for this perspective though was tempered by mounting evidence that cohabitation functions as a substitute for marriage in later life (Brown & Wright, 2017; King & Scott, 2005) and, more specifically, evidence from two prior studies (Brown & Kawamura, 2010; Wright, 2019) showing the quality of these two types of unions are largely similar.

We find that the relationship quality of midlife cohabiting and married individuals is largely indistinguishable in terms of support, disagreement, and instability, which is in line with the growing literature indicating that cohabitation during the second half of life may serve as a long-term alternative to marriage (Brown & Wright, 2017; King & Scott, 2005). At the same time our findings echo work by Brown and Kawamura (2010) as we show that midlife cohabitors are less happy in their relationships, on average, than are married individuals. We also find that cohabitors are less committed, on average, than are married individuals. Thus, there is partial support for both the incomplete institutionalization perspective and the alternative to marriage perspective for those who are cohabiting.

Consistent with the incomplete institutionalization perspective, LATs often report worse relationship quality than do married individuals. Although the flexibility of LAT relationships could be advantageous for midlife adults, minimizing disharmony by allowing individuals to preserve their independence while maintaining a long-term, intimate relationship (Lewin, 2017), we do not find any appreciable support for this argument. There is a consistent relationship quality advantage for marital relationships in midlife, especially in contrast to LAT relationships.

Across three of the five dimensions of relationship quality, the three types of unmarried relationships are essentially indistinguishable. Individuals in LAT, dating, and cohabiting relationships share similar levels of happiness, = disagreement, and instability. Cohabitors do report higher levels of support than LATs, which aligns with Lewin’s (2017) assertion that LATs fare worse that cohabitors on positive dimensions of quality. But LATs and daters do not differ in their reports of relationship support, nor do cohabitors and daters differ on this dimension. LATs report commitment levels comparable to both cohabiting and dating relationships, although daters do report less commitment than cohabitors. Ultimately though, for the most part co-residence (i.e., cohabitation) does not translate into higher levels of relationship quality among those in unmarried relationships and there are few ways to distinguish LATs and daters in terms of relationship quality.

The pattern of findings uncovered in this study reveals the gap in relationship quality between unmarried and married partnerships. The three types of unmarried relationships—dating, LAT relationships, and cohabitation—are similar to one another, and tend to be of somewhat poorer quality than married relationships. This conclusion is more characteristic of LAT relationships than cohabiting unions, as we anticipated. Still, it suggests that cohabitation and marriage are not interchangeable in midlife; at least in terms of happiness and commitment, cohabitation appears to be more akin to other unmarried partnerships such as LAT and dating than to marriage. On other dimensions, cohabitation is more comparable with marriage.

We reach somewhat different conclusions than does Lewin (2017) in her study of older adult relationship quality. Although our findings mirror hers for the positive dimensions of quality, including happiness and support, they are not in line with her results for negative dimensions of quality. Lewin (2017) found that LATs report less strain than cohabitors, whereas our results reveal LATs and cohabitors report comparable levels of disagreement and instability. Admittedly, these indicators are not precisely the same, but we would have expected them to exhibit associations in the same direction given they are all dimensions of negative relationship quality. There are a couple of reasons why our results might differ from hers. First, her measure of LATs is arguably confounded with dating. Second, her sample is older (ages 57–84) than ours (ages 50–65) and the data she analyzed were collected nearly a decade prior to the collection of the FRS. Lewin’s (2017) sample size of cohabitors was also quite modest, which could have weakened the statistical power in her models. Nevertheless, the varied patterns of findings uncovered in our study and Lewin’s work points to the need for additional research on this topic. It also foregrounds the value of examining multiple dimensions of positive and negative relationship quality (Johnson et al., 1986).

A key contribution of our work is the introduction of a novel measure of LAT relationship status that allows us to differentiate LATs from daters using a recent, nationally representative sample. Our conceptualization of midlife LAT relationships drew on theoretical (Connidis et al. 2017) and empirical research (Benson, 2016; Wu & Brown, 2021) in the North American context, but we recognize that other researchers, especially those studying LAT relationships in the European context, have specified varied working definitions of LAT relationships that differ by factors such as marital intentions or age (see Pasteels. Lyssens-Danneboom, & Mortelmans, 2017 for a detailed summary). LAT relationships likely serve unique functions across the life course (Benson & Coleman, 2016; Connidis et al., 2017), offering a combination of autonomy and commitment to midlife and older adults. Future researchers should explicitly incorporate life course stage in their conceptualization of LAT relationship status.

We believe our study is the first to examine such a broad spectrum of partnership types in midlife. Still, we also acknowledge that our work has some limitations. First, our study is cross-sectional and thus causal conclusions are not warranted. We can only identify associations. It is possible that relationship quality is endogenous to relationship type. For example, perhaps some LATs would transition into a cohabiting or marital union if they had better relationship quality (Benson, 2016). Our study provides a snapshot of the associations between relationship type and relationship quality. Second, the absence of differences between daters and LATs may have been an artifact of inadequate statistical power, however power analyses (not shown) indicated that we had sufficient sample sizes in both groups to detect mean differences. Low statistical power could have precluded us from identifying significant interaction effects between relationship type and either gender or prior marital experience. Larger samples of daters and LATs would have been preferable but our sample sizes are similar to or exceed those in prior research. Third, our measure of household income presumes sharing of economic resources between co-resident partners (a reasonable assumption for most couples, according to Eickmeyer, Manning, & Brown, 2019) but we acknowledge some couples may keep their finances separate. Likewise, household income excludes any economic resources that LATs and daters may have from their non-co-resident partners. As a robustness check, we re-estimated our multivariable models using a per capita measure of income (i.e., household income divided by two) for cohabitors and marrieds and the substantive conclusions remained the same (results not shown). Fourth, although we captured the presence of children in the household, we could not gauge (for those in non-co-residential partnerships) whether those children were also the offspring of the partner. Some couples decide to live apart together rather than cohabit or remarry precisely because they have children (from a prior partner) in the household (de Jong Gierveld & Merz, 2013). Finally, our covariates are somewhat limited in scope. Although we controlled for some of the push-pull factors (e.g., education, economic resources, prior union history, parenting) identified in the literature (Liefbroer et al., 2015; Sassler & Miller, 2017; Wagner et al., 2019) as being associated with encouraging or discouraging relationship progression (and thus shaping one’s relationship type), the data did not include other potentially relevant factors (e.g., attitudes, physical distance separating non-co-resident partners).

Our study not only informs research on relationship types and relationship quality, but it also has implications for practitioners and policymakers. We demonstrate that midlife adults are often in non-co-residential partnerships, including dating and LAT relationships, illustrating the need to recognize the diversity of partnership types that characterize adult relationships. These two partnership types appear largely comparable to cohabiting unions in terms of relationship quality. The picture for cohabitors and marrieds is mixed, with the two groups faring similarly on some dimensions of relationship quality, but not others. Because we rely on cross-sectional data, we cannot conclude that being married causes better relationship quality. Stated differently, our study does not indicate that were cohabitors to get married they would experience improvements in their relationships. Other unmeasured factors are likely at play here that shape both relationship type and relationship quality. Furthermore, we document differences in how cohabitors versus marrieds fare across various relationship quality domains, which reinforces work showing that relationship quality is multifaceted (Johnson et al., 1986). High levels of positive quality do not imply low levels of negative quality. Rather, couples can report high (or low) levels on both types. The unique pattern obtained for cohabitors, who appear largely similar to daters and LATs but also partially like marrieds, makes it a challenge to assess how cohabitation operates in midlife. For some couples, it may be a substitute for marriage, but for others, it may function in a distinct fashion. This is an important topic for future research and one that should guide practitioners as they work with cohabiting couples. It also may inform policy decisions with respect to how to treat cohabiting couples from a legal standpoint. The recent growth in cohabitation among adults aged 50 and older (Stepler, 2017) underlines the urgency of addressing how this union type functions and its role in the family life course.

Unmarried partnerships are on the rise among adults in the second half of life. A sizeable share of midlife adults in these unmarried partnerships are either in LAT or dating relationships, which have received surprisingly little attention in the literature given that our work shows non-co-residential partnerships are as common as cohabitation in midlife. The incomplete institutionalization perspective does not neatly characterize unmarried partnerships in midlife because cohabitation appears somewhat similar to marriage, supporting the alternative to marriage perspective (except for relationship happiness and commitment, on which cohabitors tend to fare worse than marrieds). The relationship quality of LAT and dating relationships aligns more closely with the incomplete institutionalization perspective in that these partnerships are marked by poorer quality than are marriages. Unmarried partnerships—whether dating, LAT, or cohabitation—are characterized by similar relationship quality, at least for most of the indicators we investigated. As these unmarried partnership types continue to gain popularity, future research should not only explore the similarities and differences among them, but also address how they are linked to health and well-being.