Economic Factors and Relationship Quality Among Young Couples: Comparing Cohabitation and Marriage

Abstract

Are economic resources related to relationship quality among young couples, and to what extent does this vary by relationship type? To answer these questions, we estimated regression models predicting respondent reports of conflict and affection in cohabiting and married partner relationships using the National Longitudinal Study of Youth, 1997 (NLSY97, N = 2,841) and the National Longitudinal Study of Adolescent Health (Add Health, N = 1,702). We found that economic factors are an important predictor of conflict for both married and cohabiting couples. Affection was particularly responsive to human capital rather than short-term economic indicators. Economic hardship was associated with more conflict among married and cohabiting couples.

The path to a stable family life has become longer in recent decades. Many young adults cohabit during their late teens and early 20s, and an increasing proportion will cohabit multiple times in the transition to adulthood (Lichter & Qian, 2008). Furthermore, young married couples are more likely to experience separation or divorce than their older counterparts (Teachman, 2002). Yet despite young adult relationships’ sometimes-fleeting nature, the quality of those relationships has important consequences. Young people learn about relationships through these early experimentations, and those lessons are likely to hold throughout their lifetime. Thus, it is important to study young people’s relationships to discover factors that contribute to happy and healthy cohabitations and marriages.

Economics seem likely to be a key factor affecting young adults’ relationship quality. Several studies have linked financial instability to relationship dissolution and divorce (Burstein, 2007; Hoffman & Duncan, 1995; Kalmijn, Loeve, & Manting, 2007; Lewin, 2005; South, 2001). Explanations for this relationship suggest that economic hardship may place stress on couples, thereby increasing conflict and leading eventually to divorce (Ono, 1998; White & Rogers, 2000). Individuals might fight over limited resources and struggle with disappointment when financial means are meager. Economic hardship is often coupled with additional stressors, such as bill collectors. Thus, economic circumstances may diminish relationship quality by increasing conflict and reducing intimacy. Understanding how and under what circumstances economic factors affect perceived relationship quality will contribute greatly to an understanding of the sources of stability and stress for young couples.

Previous research has suggested that financial strain increases couple-level violence for cohabiting and married couples (Benson, Fox, DeMaris, & Van Wyk, 2003; Fox, Benson, DeMaris, & Van Wyk, 2002). Yet differences between the relationship types have not been tested. Furthermore, little is known about how economic factors predict positive measures of relationship quality for cohabiting couples, and there is little consensus on the relationship between economic circumstances and love or affection among married couples (White & Rogers, 2000). This article seeks to add to the literature on relationship quality by examining the role of economic resources for both marital and cohabiting relationships. We analyze two dimensions of relationship quality: affection and conflict. We focus on young couples because their relationships are more likely to endure financial stress, and the quality of those relationships can have important consequences for future relationships. Finally, we compare cohabitating and married relationships to identify differences in the association between economic factors and reported affection and conflict. Knowing if and when cohabitation differs from marriage will help clarify how cohabitation functions as a family form and will contribute to the body of knowledge on cohabitating relationships.

Economic Factors and Relationship Quality

The relationships among financial resources, family formation and dissolution, and relationship quality are a persistent focus of study in family research. According to the Conger Family Stress Model, economic strain lowers overall marital quality through its influence on marital interactions (Conger, Elder et al., 1990; Conger, Rueter, & Elder, 1999). Previous research has offered mixed support for this assertion. Papp, Cummings, and Goeke-Morey (2009) found that married couples’ arguments over money were more intense and recurrent than other sources of disagreement. Some studies demonstrate that financial distress is a salient predictor of positive aspects of relationship quality (e.g., affection, love, satisfaction), although subjective economic measures were more consistently linked to relationship assessments than objective measures (Conger, Lorenz, Elder, Simons, & Ge, 1993; Matthews, Conger, & Wickrama, 1996; Robila & Krishnakumar, 2005; White & Rogers, 2000). Other studies have failed to find evidence of this relationship (Amato & Rogers, 1997, Bulanda & Brown, 2007). Conger, Elder et al. (1990) found economic strain influences wives’ evaluations of marital quality indirectly, through its positive effect on husbands’ hostility and its negative effect on husbands’ warmth.

Although Conger, Elder et al. (1990) were concerned with the implications of financial hardship for marriage, it seems reasonable to suggest that economic hardship may affect cohabiting relationships as well. One study using a sample of both cohabiting and married couples found that family harmony was marginally related to income and perceived economic well-being (Fox & Chancey, 1998). There is also some evidence that educational attainment promotes positive interactions within cohabiting couples (Brown, 2003). Finally, several studies have found that economic well-being is positively related to the odds that a cohabiting couple will marry (Edin & Reed, 2005; Lichter, Qian, & Mellott, 2006; Manning & Smock, 1995; Sanchez, Smock, & Manning, 1998). Studies of transitions out of cohabiting relationships have offered only one measure of relationship quality, however. They do not account for all dimensions of relationship quality.

Most research on economic factors and violent conflict has combined married and cohabiting couples in one sample (e.g., Benson et al., 2003; Cunradi, Caetano, & Schafer, 2002; De Maris, Benson, Fox, Hill, & Van Wyk, 2003; Fox et al., 2002; Van Wyk, Benson, Fox, & DeMaris, 2003). This research has concluded that economic strain is positively related to violence in relationships, although there is disagreement regarding appropriate measures of economic distress. De Maris et al. (2003) found that partners’ unemployment and neighborhood disadvantage were related to physical violence, whereas Cunradi et al. (2002) showed that income, but not unemployment, was negatively associated with violent conflict. Other studies have found a relationship between objective and subjective measures of financial strain and violent aggression toward female partners (Benson et al., 2003, Fox et al., 2002, Van Wyk et al. 2003).

The existing literature on economic resources and relationship quality suggests a direct relationship between the two factors. Yet there are gaps to fill. First, most of these studies have employed out-of-date samples. Nearly all use the National Survey of Families and Households (NSFH), whose primary sample was drawn in 1988. Second, many studies of violence and conflict have used measures only of male-on-female violence. This type of measure is inappropriate for nationally representative samples because nearly all intracouple violence that such studies capture will be what Johnson and Ferraro (2000) refer to as “common couple violence.” (p. 949). For studies using nationally representative samples, a better measure of conflict would take both partners’ aggressive behaviors into account. Last, much of the theoretical and empirical work on economic factors and positive measures of relationship quality has focused on marriage. We do not know how economic factors influence affection between partners in cohabitating relationships. Research on violent conflict has been more inclusive of cohabiting couples, but it has not differentiated between married and cohabiting partnerships. Differences between the institutional and economic features of marriage and cohabitation may have implications for the processes by which economic factors affect relationship quality.

Differences Between Married and Cohabiting Couples

Cohabiting and married relationships differ in many respects, and those differences have implications for the role of economic factors in relationship quality. Cohabitation is selective of the less educated, younger adults, divorcees, non-Whites, and those who are more supportive of egalitarian gender roles (Bumpass and Lu, 2000; Clarkberg, Stolzenberg, & Waite, 1995; Edin & Reed, 2005; Smock, 2000). Many cohabitors enter into joint living arrangements out of financial necessity, which may result in more fragile partnerships (Sassler, 2004). Entry into marriage, however, is related to increases in economic well-being (Sweeney, 2002). Furthermore, married couples typically manage their resources jointly, allowing them to adjust to changing economic circumstances (Brines & Joyner, 1999), whereas cohabiting partners are less likely to pool their income (Oropesa, Landale, & Kenkre, 2003; Treas & De Ruijter, 2008). This may provide greater flexibility to engage in a wider range of economic arrangements, but it leaves cohabitors more exposed than married partners to fluctuations in income.

Previous research also has found that cohabitors report lower relationship quality (Brown & Booth, 1996) and higher levels of victimization and perpetration than married individuals (Brown & Bulanda, 2008; Johnson & Ferraro, 2000; Stets & Strauss, 1989). This may be due to demographic differences, couple-level processes, and selection effects. In this article, we focus on the role of economic resources and hardship. We argue that the quality of cohabiting relationships may be more vulnerable to economic factors because of their weaker economic and institutional foundations. Because of those fundamental differences, we examine the relationship between economic factors and relationship quality separately by relationship type and test for differences between them. Our argument suggests a divergence in the ways married and cohabiting couples experience, respond to, and are affected by financial distress. Alternatively, no differences across relationship type suggest that married and cohabiting relationships vary not in relationship processes but in means.

Young Cohabiting and Married Partners

Young cohabiting and married couples face unique challenges in building strong relationships. Young adults are particularly likely to experience poverty. One study found that, in the 1990s, more than one third of young people in their 20s spent at least one year living below the official poverty line (Sandoval, Rank, & Hirschl, 2009). Family background also differs among young cohabitors and married partners, whereby higher maternal education is associated with a lower risk of entering young cohabiting and married relationships (Schoen, Landale, Daniels, & Cheng, 2009). Such couples face uncertain economic futures and limited safety nets.

Young cohabiting and marital couples also differ in important ways from older couples. Young cohabitors are more likely to see their relationships as a precursor to marriage than are older couples, although they also report lower levels of relationship quality and stability (King & Scott, 2005). Among married couples, age is significantly and negatively associated with risk of divorce (Teachman, 2002). Prior research has demonstrated that relationship instability—including dissatisfaction and thoughts of divorce—increases when partners marry young (Booth & Edwards, 1985). Although early marriage has decreased in the United States, it is not wholly absent. Nearly one quarter of young adults marry before age 23 (Uecker & Stokes, 2008). Cohabitation is also quite common among young adults (Lichter & Qian, 2008). Thus, it is important to understand the sources and dynamics of relationship quality for such couples.

Dynamics of Relationship Quality

Relationship quality is an ambiguous term, potentially encompassing all objective and subjective measures of couple-level well-being. Measures of relationship quality include relationship satisfaction, individual-level happiness and content, frequency of arguing, conflict, violence, and relationship dissolution. Yet focusing on only one indicator of relationship quality ignores the multidimensional nature of the concept (Willetts, 2006). In this article, we seek to explore the relationship between economic factors and two dimensions of relationship quality: affection and conflict. We expect to find a direct relationship between economics and relationship quality; however, we anticipate that this relationship will differ for cohabiting and married individuals. We hypothesize that economic circumstances have a greater association with relationship quality for cohabitors than for married individuals.

Method

Sample and Data

This study used data from two nationally representative studies: the National Longitudinal Study of Youth (NLSY97) and the National Longitudinal Study of Adolescent Health (Add Health).

The NLSY97 data set was designed to represent individuals in the United States in 1997 born between the years 1980 and 1984 to document their transition from adolescence to adulthood and from school to work. The majority of respondents were still in school at the start of data collection. The original sample comprised 8,984 respondents, including a nationally representative sample of 6,748 youths and an oversample of 2,236 Latino and Black youths. Respondents were interviewed yearly from 1997 through 2005. The NLSY97 Web site provides a more detailed description of the NLSY97 study (http://www.bls.gov/nls/nlsy97.htm).

The NLSY97 has collected data on all cohabitating and marital relationships for respondents from age 16 onward. Several key relationship measures were not asked about in the early years of the study, so we limited our analysis to the years 2000 through 2005. We included anyone who was cohabiting or married during that period in our sample. We chose the most recent partnership for which we had data as the relationship of focus. Our sample consisted of 1,625 cohabiting partners and 1,216 married partners.

Add Health is a nationally representative study of teenagers in the 7th through 12th grade in the United States in 1995. This data set used a school-based and multistage cluster sample design. The study began with an in-school questionnaire and then used school rosters to randomly select 200 students from each school to participate in in-home interviews. Wave 1’s total sample size for in-home interviews was 20,745 adolescents. Respondents were reinterviewed six years later for the Wave 3 in-home interviews, which took place from August 2001 to April 2002. From that wave, 500 respondents who were then dating, cohabiting, or married were randomly drawn and asked to recruit their partners for participation in the study. Only partners who were opposite-sex partners, older than 18 years of age, and in a relationship with the respondent for at least 3 months were eligible. We drew our sample from the cohabiting and married partner samples, for a total of 838 cohabiting partners (419 cohabiting couples) and 864 married partners (432 married couples). Harris et al. (2003) have provided a more detailed description of the Add Health study.

Outcome Variable: Relationship Quality

In the NLSY97 data set, we measured affection with a scale comprising the following two items: (a) “How close do you feel towards your partner?” and (b) “How much do you feel that your partner cares about you?” Respondents were asked to rate their responses on a 0–10 scale. We averaged the two values to create an overall indicator of affection. This measure was highly skewed, so we created a dichotomous indicator of high levels of affection, where 0 indicates a score of less than 9, and 1 indicates a score of 9 to 10.

From Add Health, we measure affection with a scale that comprised the following two items: (a) “How much do you love your partner?” and (b) “How much does your partner love you?” The two measures had a 4-point scale, where 0 = a lot and 3 = not at all. We again averaged the responses and recoded them into a dichotomous measure of high levels of affection, where 1 indicates a response of “a lot” to both questions.

From NLSY97, our measure of conflict in the relationship was “Overall what is your relationship like with your partner? On a scale of 0 to 10, where 0 is no conflict and 10 is a lot of conflict, how would you rate your relationship with your partner?” This measure is skewed toward the lower end of the scale. We tested various recoding schemes, including standardizing the scale and recoding it into a smaller number of values. Results were commensurate with the original measure, so we present the unaltered version here.

In the Add Health data set, we assessed conflict by creating a scale based on each respondent’s report of whether his or her partner had, in the past year, threatened violence, pushed or shoved; slapped, hit, or kicked; forced sexual relations; or caused the respondent an injury. The scale ranged from 0 = none of these happened in the past year to 4 = all had occurred. Respondents were also asked to report how often they had done those things to their partner in the past year, which created a second scale ranging from 0 to 4. We then averaged each individual’s responses for measure of relationship conflict ranging from 0 to 4.

Explanatory Variable of Interest: Economics

We used five measures of financial security and insecurity: poverty-line adjusted family earnings, family support, government assistance, hardship, and college attendance. For the NLSY97 data set, all information on the partner was obtained from the respondent’s answers to the survey, whereas both partners answered the Add Health survey.

We computed poverty-line adjusted family earnings in three steps. First, we summed the respondent’s and partner’s earnings for the prior calendar year. Next, we converted that figure to 2005 dollars. Finally, we divided that amount by the Census Bureau’s household-size specific poverty threshold in 2005, based on the number of people living in the respondent’s household.

Next, we consider whether receiving help from family members might affect relationship quality. Parents and other family members are most likely to give money to young people who are in financial need (Fingerman, Miller, Birditt, & Zaritt 2009). Therefore, we consider receipt of family support an indicator of hardship, much like governmental program assistance. Receiving financial assistance from family members, however, may have a different effect depending on the relationship type. For married couples who share their income, receiving support is beneficial for both individuals and may strengthen the relationship after controlling for need. For cohabiting couples who do not share financial resources, receiving income from family members is a potential source of conflict when only one partner is receiving extra support. From NLSY97, we created an ordinal measure of receiving money from family members, indicating whether the respondent and his or her partner had reported receiving no money from family, $1–$500, or more than $500. The Add Health survey asked respondents only whether they had received help from family members, so we created a dichotomous variable indicating this support.

For both data sets, we included a dichotomous indicator of whether the couple received government assistance in the prior year.

We created a hardship measure using Add Health data only. This measure is an index of the six items that indicate financial hardship: whether there was a time in the past year when the individual or the household was without a telephone, was evicted for being unable to pay the full rent or mortgage, was unable to pay a full gas or electricity bill, had gas or electricity services shut off, needed to see a doctor but was unable to afford the bills, and needed to see a dentist but was unable to afford to do so, ranging from 0 = none to 6 = all.

Finally, we measured educational attainment using a dummy variable for whether the individual had attended college. Higher levels of educational attainment signify investment in human capital, which may lead to economic security in the future. We limited our measure of educational attainment because of the youth of our sample. Measures of college completion and beyond would be misleading, as many members might attain more education at an older age.

Control Variables

We controlled for several factors likely to affect relationship quality. First, we included an indicator of whether the respondent was currently enrolled in school. We included dummy variables indicating whether the respondent had previously married and whether the respondent had previously cohabited. We also controlled for whether there were any children under the age of 18 living in the home with the couple. In addition, we controlled for a variable measuring the length of the present union, in months. We top-coded this figure so that none indicated a relationship starting before age 16. We also controlled for the gender of respondents, with 1 = male and 0 = female. We constructed dummy variables for Black, Hispanic, and other race respondents, with White as the reference category. Finally, we included a variable controlling for age.

Last, in the Add Health analyses, we included a control for whether the respondent was the original Add Health respondent (1) or the sampled partner (0). Add Health respondents were asked to recruit their partners for the study. Agreeing to participate may have indicated that they were particularly happy in their relationships, and they might differ significantly along that domain from their partners. In addition, they were interviewed for a third time, whereas this was their partner’s first experience with Add Health interviewers. Respondents who had been interviewed more often may have had a greater desire to appear happy in their relationships.

Use of Two Data Sets

We chose to use data from both NLSY97 and Add Health because we believe that both data sets have unique strengths. Add Health allowed us to use reports from both members of a couple, which means that we do not need to rely on one partner’s assessment of relationship dynamics. The Add Health data set also gives us the ability to assess the impact of economic hardship on relationship quality. The Add Health sample relied on recruitment of partners by the original respondent, however, and this may have garnered a sample of particularly satisfied couples. The NLSY97 asked relationship quality questions of all individuals in a relationship. Therefore, this sample is likely to be more representative of cohabiting and married individuals. In addition, the NLSY97 contains better and more detailed measures of earnings and family support.

Furthermore, NLSY97 asked for an overall assessment of the level of conflict in a relationship. We believe that the NLSY97 measure captures instances of conflict such as arguing or tension that the Add Health measures did not assess.

Because both of our outcomes, affection and conflict, are measured differently in each data set, we anticipate that there may be slight differences in findings. We believe, however, that each data set complements and provides greater insight into the results found from the other source. In particular, we believe that the use of two different measures of conflict, a general measure (NLSY97) and a specific measure assessing violence (Add Health), can contribute greatly to the understanding of how economics affect conflict more broadly.

Analytic Strategy

We used logistic regression to assess the relationship between economic factors and affection using the NLSY97 data set. We used ordinary least squares regression for models predicting conflict.

For our analysis of the Add Health couples’ data, we employ random-effects logistic regression to estimate the relationship between economic insecurity and affection. We used random-effects ordinary least squares (OLS) regression to estimate the relationship between economic insecurity and conflict. We chose to use random-effects regression because our analyses take advantage of both individual- and couple-level data. Random-effects regression adjusts for the correlated error that is induced by the nonindependence of observations on the same couple, which then allowed us to estimate each partner’s reported levels of affection and conflict. To run a random-effects logistic regression, we used the command “xtlogit” in Stata10 and set the group variable as the couple identification number. To run a random-effects regression, we used the command “xtreg” in Stata10 and again set the group variable as the couple identification number.

We separated all models by relationship type because cohabiting and married relationships are fundamentally different relationships and are traditionally treated as such in family literature. However, whether differences in the association between economic factors and relationship quality exist for married and cohabiting couples is an open question. We test for differences by running Chow tests and interacting relationship status and economic factors. A Chow test is a statistical test that evaluates whether coefficients from a regression model are significantly different from one another across subsamples (in this case, across relationship type). Including interaction terms within an aggregate model (i.e., interacting relationship type with economic factors in a model predicting relationship quality) provides a similar test of difference. We used both tests to confirm our findings of difference or similarity across relationship type. We report our findings below, noting when our tests of differences confirm apparent differences across relationship type.

RESULTS

Table 1 summarizes the mean, standard deviation, and range of all NLSY97 and Add Health variables. In both data sets, cohabiting partners reported lower levels of affection and higher levels of conflict than did married partners, on average.

Table 1.

Cohabiting and Married Partners’ Reports of Relationship Quality, Economic Factors, and Demographic Variables: Descriptive Statistics

Variables	NLSY97						Add Health
	Cohabiting Partners			Married Partners			Cohabiting Partners			Married Partners
	M	SD	Range	M	SD	Range	M	SD	Range	M	SD	Range
Reported affectiona	0.78	0.41	0–1	0.85**	0.36	0–1	0.82	0.38	0–1	0.91**	0.29	0–1
Reported conflict	3.49	2.63	0–10	3.21**	2.58	0–10	0.95	1.43	0–4	0.81*	1.31	0–4
Poverty-adjusted household earnings	1.82	1.86	0–29	2.31**	1.79	0–15	1.87	1.75	0–21	2.21**	1.87	0–22.1
Money given by familyb							0.57	0.50	0–1	0.38**	0.49	0–1
No family money received	0.78	0.42	0–1	0.78	0.41	0–1
$1–$500	0.12	0.32	0–1	0.10†	0.29	0–1
More than $500	0.11	0.31	0–1	0.12	0.33	0–1
Received govt. program assistancec	0.36	0.48	0–1	0.33†	0.47	0–1	0.27	0.44	0–1	0.22*	0.42	0–1
Hardship							0.99	1.18	0–6	0.77**	1.13	0–6
Respondent attended colleged	0.33	0.47	0–1	0.39**	0.49	0–1	0.42	0.49	0–1	0.40	0.49	0–1
Enrolled in schoole	0.15	0.36	0–1	0.14	0.34	0–1	0.24	0.43	0–1	0.15**	0.35	0–1
Previously marriedf	0.03	0.18	0–1	0.03	0.16	0–1	0.26	0.44	0–1	0.11**	0.32	0–1
Previously cohabitedg	0.25	0.43	0–1	0.58**	0.49	0–1	0.08	0.26	0–1	0.05*	0.22	0–1
Child living in householdh	0.45	0.50	0–1	0.58**	0.49	0–1	0.29	0.46	0–1	0.53**	0.50	0–1
Length of relationship
Less than 3 months	0.11	0.31	0–1	0.02**	0.15	0–1	0.10	0.30	0–1	0.08†	0.28	0–1
3–12 months	0.27	0.44	0–1	0.13**	0.33	0–1	0.43	0.50	0–1	0.23**	0.42	0–1
Over 1 year	0.62	0.48	0–1	0.85**	0.36	0–1	0.47	0.50	0–1	0.68**	0.47	0–1
Malei	0.45	0.50	0–1	0.39**	0.49	0–1	0.50	0.50	0–1	0.50	0.50	0–1
Respondent’s race/ethnicity
White	0.49	0.50	0–1	0.60**	0.49	0–1	0.63	0.48	0–1	0.71**	0.45	0–1
Black	0.26	0.44	0–1	0.11**	0.33	0–1	0.16	0.36	0–1	0.09**	0.29	0–1
Hispanic	0.23	0.42	0–1	0.26†	0.44	0–1	0.13	0.33	0–1	0.16*	0.36	0–1
Other race/ethnicity	0.03	0.17	0–1	0.03	0.16	0–1	0.08	0.28	0–1	0.04**	0.19	0–1
Age	22.39	1.74	17–26	23.1**	1.51	18–26	22.6	2.96	18–39	23.5**	2.98	18–43
Add Health main respondentj							0.50	0.50	0–1	0.50	0.50	0–1
	n =1,625			n = 1,216			n = 838			n = 864

^aReported affection: 0 = low, 1 = high.

^bMoney given by family: 0 = no, 1 = yes.

^cReceived govt. program assistance: 0 = no, 1 = yes.

^dRespondent attended college: 0 = no, 1 = yes.

^eEnrolled in school: 0 = no, 1 = yes.

^fPreviously married: 0 = no, 1 = yes.

^gPreviously cohabited: 0 = no, 1 = yes.

^hChild living in household: 0 = no, 1 = yes.

ⁱMale: 0 = no, 1 = yes.

^jAdd Health main respondent: 0 = no, 1 = yes.

^†p < .10.

^*p < .05.

^**p < .01 (for t-tests comparing means for married versus cohabiting couples within the same data set).

Cohabiting partners also reported lower adjusted family earnings than married partners in both data sets. In the NLSY97, cohabiting partners were more likely to report receiving between $1 and $500 from family members, whereas in Add Health, cohabiting partners were more likely to report receiving family support. Cohabiting partners were more likely to report receiving government support in both data sets. Cohabiting partners in Add Health reported more economic hardship than did married partners. Finally, there were no significant differences in college attendance between cohabiting and married partners in Add Health. In the NLSY97, cohabiting partners were less likely than married partners to have attended school.

Table 2 presents results from our models of affection regressed on economic measures and other background factors for cohabiting and married partners in the NLSY97. Receipt of governmental assistance increased cohabitors’ reported level of affection for their partners. This is surprising because receiving governmental assistance is an indicator of financial need. Holding other measures of economic well-being constant, however, governmental assistance was related to higher levels of relationship quality. Income was also positively related to relationship quality among married couples at the .10 significance level. Finally, we found a positive relationship between education and affection for both cohabiting and married couples. Supplemental analyses (not shown) indicated that apparent differences across relationship type were not statistically significant for any variables except age. Advanced age was negatively associated with reported affection among married couples but not cohabiting couples.

Table 2.

Ordinary Logistic Regression of Affection on Economic Factors, Relationship Factors, and Demographics for NLSY97 Cohabiting (n = 1,625) and Married (n = 1,216) Partners

Predictor	Cohabiting Partners			Married Partners
	B	SE B	eB	B	SE B	eB
Adjusted family earnings	0.03	0.04	1.03	0.10†	0.06	1.10
Money given by familya
$1–$500	−0.04	0.19	0.96	0.31	0.41	1.36
More than $500	−0.23	0.20	0.79	0.19	1.21	0.35
Received govt. assistanceb	0.27†	0.15	1.31	−0.21	0.16	0.81
Respondent attended collegec	0.47**	0.16	1.61	0.35†	0.29	1.42
Length of relationshipd
3–12 months	−0.02	0.22	0.98	−0.25	0.62	0.78
Over 1 year	−0.05	0.21	0.95	−0.59	0.42	0.56
Child living in householde	−0.44**	0.16	0.64	−0.33	0.15	0.72
Previously cohabitedf	−0.25†	0.14	0.78	−0.11	0.16	0.90
Previously marriedg	0.66	0.40	1.93	0.80	1.39	2.23
Maleh	−0.02	0.13	0.99	−0.18	0.14	0.84
Respondent’s race/ethnicityi
Black	−0.06	0.15	0.94	−0.09	0.23	0.92
Hispanic	−0.00	0.16	1.00	−0.11	0.17	0.90
Other race/ethnicity	−0.04	0.35	0.96	0.30	0.75	1.35
Age	0.02	0.04	1.02	−0.14*	0.05	0.87
Enrolled in schoolj	−0.16	0.19	0.85	−0.50*	0.15	0.61
Constant	0.86			5.64***
Log-likelihood	−841.84			−503.22

Note: e^B = exponentiated B.

^aReference category is “No family money received.”

^bReceived govt. program assistance: 0 = no, 1 = yes.

^cRespondent attended college: 0 = no, 1 = yes.

^dReference category is “Less than 3 months.”

^eChild living in household: 0 = no, 1 = yes.

^fPreviously cohabited: 0 = no, 1 = yes.

^gPreviously married: 0 = no, 1 = yes.

^hMale: 0 = no, 1 = yes.

ⁱReference category is “White.”

^jEnrolled in school: 0 = no, 1 = yes.

^†p < .10.

^*p < .05.

^**p < .01.

Table 3 presents results for our analysis of affection using the Add Health data set. Once again, we found that educational attainment was positively related to affection for both relationship types. We did not find a significant association between other measures of economic well-being and affection. There were no statistically significant differences in the effect of educational attainment or our control variables by relationship type.

Table 3.

Random-Effects Logistic Regression of Affection on Economic Factors, Relationship Factors, and Demographics for NLSY97 Cohabiting (n = 838) and Married (n = 864) Partners

Predictor	Cohabiting Partners			Married Partners
	B	SE B	eB	B	SE B	eB
Adjusted family earnings	0.03	0.08	1.03	−0.04	0.08	0.96
Received support from familya	0.02	0.27	1.02	0.25	0.35	1.28
Received govt. assistanceb	0.01	0.31	1.01	0.07	0.39	1.07
Hardship	−0.15	0.11	0.86	−0.13	0.15	0.88
Respondent attended collegec	0.74**	0.29	2.10	0.99*	0.38	2.69
Length of relationshipd
3–12 months	0.44	0.43	1.55	−0.23	0.75	0.79
Over 1 year	0.41	0.43	1.51	−0.60	0.72	0.55
Child living in householde	0.41	0.30	1.51	−0.36	0.36	0.70
Previously cohabitedf	−0.49†	0.27	0.61	0.33	0.49	1.39
Previously marriedg	0.11	0.48	1.12	−1.09†	0.64	0.34
Maleh	0.38	0.24	0.68	−0.24	0.29	0.79
Respondent’s race/ethnicityi
Black	−1.06**	0.36	0.35	−1.81*	0.48	0.16
Hispanic	0.03	0.39	1.03	−0.50	0.43	0.61
Other race/ethnicity	−0.01	0.50	0.99	−0.68	0.74	0.51
Age	−0.05	0.05	0.95	0.07	0.06	1.07
Enrolled in schoolj	0.32	0.35	1.38	−0.47	0.47	0.63
Add Health main respondentk	0.77***	0.23	2.16	0.63*	0.29	1.88
Constant	2.69*			2.18
Log-likelihood	−354.62			−242.99

Note: e^B = exponentiated B.

^aReference category is “No family money received.”

^bReceived govt. program assistance: 0 = no, 1 = yes.

^cRespondent attended college: 0 = no, 1 = yes.

^dReference category is “Less than 3 months.”

^eChild living in household: 0 = no, 1 = yes.

^fPreviously cohabited: 0 = no, 1 = yes.

^gPreviously married: 0 = no, 1 = yes.

^hMale: 0 = no, 1 = yes.

ⁱReference category is “White.”

^jEnrolled in school: 0 = no, 1 = yes.

^kAdd Health main respondent: 0 = no, 1 = yes.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Table 4 presents results from our models of conflict regressed on economic measures and control variables for cohabiting and married couples in the NLSY97 data set. For cohabiting partners, adjusted earnings were significantly and negatively related to reported conflict. This relationship did not hold among married partners, although the difference between relationship types was not statistically significant. Among married partners, a small amount of family support was significantly related to lower levels of relationship conflict. This suggests that low levels of support from one’s family help stabilize married relationships. This was not true for cohabiting couples, however, and tests of the models show that this difference between models was significant at the .05 level. Finally, college attendance decreased the likelihood that cohabiting or married respondents reported high levels of conflict in their relationships. Tests of the models revealed significant differences in the relationship between being the male partner and reports of conflict in the relationship. Being male was associated with higher levels of reported conflict among married couples but not cohabiting couples.

Table 4.

Ordinary Least Squares Regression of Conflict on Economic Factors, Relationship Factors, and Demographics for NLSY97 Cohabiting (n = 1,625) and Married (n = 1,216) Partners

Predictor	Cohabiting Partners			Married Partners
	B	SE B	β	B	SE B	β
Adjusted family earnings	−0.10	0.04	−2.48*	−0.05	0.05	−0.99
Money given by familya
$1–$500	0.04	0.20	0.19	−0.61	0.25	−2.44*
More than $500	0.31	0.21	1.47	−0.16	0.23	−0.68
Received govt. assistanceb	−0.07	0.17	−0.44	0.23	0.19	1.22
Respondent attended collegec	−0.41	0.16	−2.49*	−0.54	0.17	−3.09**
Length of Relationshipd
3–12 months	0.40	0.23	1.70†	−0.37	0.52	−0.73
Over 1 year	0.58	0.22	2.63**	0.03	0.49	0.06
Child living in householde	0.31	0.17	1.86†	0.16	0.18	0.85
Previously cohabitedf	−0.08	0.16	−0.50	0.01	0.16	0.05
Previously marriedg	−0.21	0.38	−0.55	−0.62	0.46	−1.35
Maleh	−0.08	0.14	−0.57	0.38	0.15	2.47*
Respondent’s race/ethnicityi
Black	0.41	0.16	2.52*	0.52	0.23	2.22*
Hispanic	0.45	0.17	2.68**	0.20	0.18	1.12
Other race/ethnicity	1.06	0.38	2.83**	0.02	0.46	0.03
Age	−0.00	0.04	−0.05	−0.10	0.05	−1.94†
Enrolled in schoolj	−0.19	0.20	−0.96	0.26	0.23	1.13
Constant	3.07	0.89	3.43***	5.52	1.25	4.40***
Adjusted R2		0.04			0.04
F	4.73***			3.81***

^aReference category is “No family money received.”

^bReceived govt. program assistance: 0 = no, 1 = yes.

^cRespondent attended college: 0 = no, 1 = yes.

^dReference category is “Less than 3 months.”

^eChild living in household: 0 = no, 1 = yes.

^fPreviously cohabited: 0 = no, 1 = yes.

^gPreviously married: 0 = no, 1 = yes.

^hMale: 0 = no, 1 = yes.

ⁱReference category is “White.”

^jEnrolled in school: 0 = no, 1 = yes.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Table 5 presents the coefficients from our random-effects models of conflict using the Add Health data set. We found that financial support from family members and hardship were positively and significantly related to reports of conflict among cohabitors. Supplemental analyses revealed that receiving governmental assistance positively predicted conflict when other measures of economic well-being were not included. Among married couples, hardship was consistently and significantly related to reports of conflict. Tests of the models indicated that there were no significant differences between married and cohabiting couples in the effect of any explanatory or control variables on conflict.

Table 5.

Random-Effects Regression of Conflict on Economic Factors, Relationship Factors, and Demographics for Add Health Cohabiting (n = 838) and Married (n = 864) Partners

Predictor	Cohabiting Partners			Married Partners
	B	SE B	β	B	SE B	β
Adjusted family earnings	0.03	0.03	0.81	−0.01	0.03	−0.20
Received support from familya	0.26	0.12	2.22*	−0.01	0.11	−0.07
Received govt. assistanceb	0.21	0.14	1.52	0.09	0.13	0.70
Hardship	0.10	0.05	1.95†	0.14	0.05	2.89**
Respondent attended collegec	0.05	0.12	0.44	0.12	0.10	1.17
Length of relationshipd
3–12 months	0.34	0.19	1.78†	0.32	0.21	1.51
Over 1 year	0.40	0.20	2.05*	0.19	0.20	0.96
Child living in householde	−0.08	0.13	−0.62	0.31	0.11	2.75**
Previously cohabitedf	0.04	0.12	0.36	−0.08	0.13	−0.63
Previously marriedg	−0.28	0.20	−1.39	0.19	0.21	0.93
Maleh	−0.26	0.09	−2.86**	−0.18	0.08	−2.30*
Respondent’s race/ethnicityi
Black	0.45	0.15	2.91**	0.38	0.17	2.19*
Hispanic	0.23	0.16	1.47	0.17	0.14	1.26
Other race/ethnicity	0.37	0.19	1.90†	0.56	0.25	2.28*
Age	−0.03	0.02	−1.31	−0.00	0.02	−0.26
Enrolled in schoolj	−0.22	0.13	−1.68†	0.01	0.13	0.06
Add Health main respondentk	−0.15	0.08	−1.73†	−0.08	0.08	−1.05
Constant	1.00	0.47	2.13**	0.43	0.44	0.98
Adjusted R2		0.06			0.06
Wald χ2		49.05***			47.71***

^aReference category is “No family money received.”

^bReceived govt. program assistance: 0 = no, 1 = yes.

^cRespondent attended college: 0 = no, 1 = yes.

^dReference category is “Less than 3 months.”

^eChild living in household: 0 = no, 1 = yes.

^fPreviously cohabited: 0 = no, 1 = yes.

^gPreviously married: 0 = no, 1 = yes.

^hMale: 0 = no, 1 = yes.

ⁱReference category is “White.”

^jEnrolled in school: 0 = no, 1 = yes.

^kAdd Health main respondent: 0 = no, 1 = yes.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

DISCUSSION

This article makes several contributions to the literature on economic well-being and relationship quality for married and cohabiting partners. First, we find support for Conger, Elder et al.’s (1990) model of family stress by demonstrating that economic factors play an important role in perceived relationship quality among young cohabiting and married couples. Such young men and women are particularly susceptible to experiencing periods of poverty, which—though most likely temporary—may prove disruptive to their burgeoning romantic relationships (Sandoval et al., 2009). It is particularly important to understand which indicators of economic well-being and hardship matter for these young relationships and how they matter.

We expand on Conger, Elder et al.’s (1990) Family Stress Model in several key ways. We separated and tested for differences in the relationship between economics and relationship quality for married and cohabiting couples. Studies that have examined predictors of conflict in married and cohabiting couples have typically lumped the two relationship types together. Yet the defensibility of doing so has not been tested. The economic foundations of marriage and cohabitation differ, and prior research has suggested that dissolution rates among married and cohabiting relationships respond differently to economic factors (Brines & Joyner, 1999). This suggests a difference in the process by which economic factors affect relationship quality. Yet we found the opposite. The association between economic factors and relationship quality does not differ between married and cohabiting couples. Instead, cohabitors have fewer economic resources available than married couples, and the difference in means drives some of the overall disparity in reported relationship quality. The findings suggest that the processes by which relationships strengthen and weaken are similar across the two family forms.

We did find differences between married and cohabiting partners in the relationship between family support and conflict in the NLSY97. The receipt of family support is predictive of lower levels of reported conflict among married partners. Family support is not significantly related to conflict for cohabiting partners, however. We believe that the meaning that family support holds for married and cohabiting partners may explain this. Married couples usually enter into joint financial relationships. Receipt of family support, then, can be viewed as ameliorating economic difficulty for both members of the couple. Cohabitation, however, does not have the same norm of sharing finances, so family support may be viewed as help for an individual partner, not for both members of the couple. Receiving family support may even exacerbate conflict between cohabiting partners, if it becomes a point of contention.

We found that the role of economic factors depends on the dimension of relationship quality that is examined. Affection appears to be particularly responsive to educational attainment. Economic factors play a larger role in the level of reported conflict for both cohabiting and married partners, and this relationship emerged regardless of the type of conflict: generalized (NLSY97) or violent (Add Health). This is surprising, given the expectation that multiple dimensions of relationship quality would be interrelated. Increases in conflict (violent or otherwise) could be expected to decrease positive measures of relationship well-being in equal measure. Future work should build on this work by examining other components of relationship quality, such as commitment and satisfaction.

Finally, our findings suggest the importance of considering how a wide range of economic factors affect relationship quality and well-being. Measures beyond earnings, such as family support and hardship, play an important role in predicting relationship quality. We show that economic well-being can improve positive measures of relationship quality and that economic hardship can play a role in instigating couple conflict. Future work should build on this finding by examining marital and cohabiting relationships longitudinally to understand how changes in economic factors predict changes in relationship quality. To do this, data will need to be collected in short intervals to capture the ups and downs of economic well-being and relationship quality before a cohabiting relationship transitions to marriage or dissolves.

This article has focused on a young sample, who may be at greater risk for financial insecurity. We believe that doing so is warranted, given the susceptibility of young people both to economic hardship (Sandoval et al., 2009) and unstable relationships (Booth & Edwards, 1985; Teachman, 2002). This leaves open the question of how financial stability and instability affect older partners. Economic insecurity may have a greater influence on the quality of older couples if individuals expect to experience financial insecurity at young ages but envision reaching levels of economic security as they age. Future work should examine other aspects of relationship quality and older couples to tease apart the role economics plays in both cohabiting and marital relationships. Our article also has used cross-sectional data in its measures of economic factors and relationship assessments. This was necessary, because cohabitations are typically short in duration. Future work should endeavor to collect and use longitudinal data with closely spaced data collection periods to more fully disentangle causal ordering. Finally, we believe our measures of relationship quality can be expanded to consider a wider array of perspectives on relationship well-being and partner happiness.