A Cohort Comparison of Predictors of Young Adult Union Formation and Dissolution in the US

Abstract

The theory of the second demographic transition argues that as educated Americans began valuing self-actualization and individual autonomy, delays in union formation spread through the US. The accelerated adulthood theory suggests that socioeconomic disadvantage distinguishes young adulthood such that those with fewer resources have shorter, more informal (i.e. cohabitation) unions, and those with more resources delay but achieve marriage and have greater union stability. We use two large, nationally representative samples of young adults collected about twenty years apart, the National Longitudinal Surveys of Youth 1979 and 1997 cohorts to examine cohort differences in union formation and dissolution and test interaction effects in demographic and socioeconomic correlates. We found that the NLSY97 cohort 1) entered into unions earlier than the NLSY79 cohort, 2) entered direct marriage (marriage without premarital cohabitation) later than the NLSY79 cohort, and 3) entered cohabiting unions earlier than the NLSY79 cohort. A greater proportion of young adults in the NLSY97 cohort dissolved their first union between ages 16 and 30. We found that socioeconomically disadvantaged young adults had earlier unions by some indicators (e.g. lower maternal education) and later unions by other indicators (e.g. unemployment) in both cohorts. We also found that in both cohorts, socioeconomic disadvantage undermined union stability. We also found evidence for interaction effects; some indicators of socioeconomic disadvantage (e.g. income, employment, and maternal education) had exacerbated effects on union formation and stability in the NLSY97 as compared to the NLSY79 cohorts perhaps because inequality grew over the twenty years between cohorts.

The age at marriage is at a worldwide high and some people will never marry (Ortega, 2014). Scholars have argued that the benefits of marriage have decreased (Musick & Bumpass, 2012), and instead of planning weddings, young adults are hooking up (Monto & Carey, 2014), and moving from relationship to relationship (Cohen & Manning, 2010). Yet other scholars argue that marriage is still part of the American dream (Lauer & Yodanis, 2010). They cite that adolescents continue to strive to marry (Anderson, 2016), most young adults only cohabit with their spouse (Teachman, 2003), and the divorce rate has declined for young adults (Kennedy & Ruggles, 2014). Although trends in the prevalence of cohabitation and marriage and the median age of cohabitation and marriage have been a focus of family demographers (Manning, Brown, & Payne, 2014; Ortega, 2014), trends in the correlates of union formation and dissolution have received less attention.

Cohort comparisons can include two types of effects (Williams, 2015). The first is a compositional effect. Compositional effects suggest that levels of the independent variable may differ between groups, but that the association between the independent variable and the outcome of interest do not differ. For example, just 7% of the NLSY79 cohort identified as Hispanic, whereas 13% of the NLSY97 cohort identified as Hispanic due to the growth in Hispanic migration and immigration to the US over this twenty year period (Flores, 2017). But, has the association between identifying as Hispanic and age at union entry changed over time? If the answer is no, then there is a compositional effect only. If the answer is yes, for example the association between Hispanic identity and age at first union is stronger in the NLSY79 as compared to the NLSY97, there would be an interaction effect. That is, the effect of the independent variable would differ between groups, or cohorts (Williams, 2015). When cohort comparisons have been conducted, the focus has been on these compositional differences between cohorts by race and education (Kuo & Raley, 2016; Manning et al., 2014). We focus on interaction effects in the demographic and socioeconomic predictors of first union formation and dissolution in two US cohorts. Specifically, using two cohorts of young adults in the US collected about twenty years apart from the National Longitudinal Surveys of Youth 1979 and 1997, we ask: have the predictors of union formation and dissolution in the US changed as trends in union formation and dissolution have changed? Guided by the second demographic transition and accelerated adulthood theories (Amato, 2010; Sassler, 2010), predictors examined include gender, race/ethnicity, education, school enrolment, employment, income, pregnancy, children, family structure, maternal education, and religion. Shedding light on persistent, and diverging, trends in the predictors of union formation and dissolution in the US context has the potential to inform conversations about the implications of family formation change around the world.

Cohort Trends and Comparisons

Overall, the median age at first union in the US changed little from 1984 to 2010; the median age at first union was about 22 for women and 23 to 24 for men between 1984 and 2010 (Manning et al., 2014). What changed was the type of first union. The proportion of women who ever cohabited almost doubled from one to two-thirds between 1987 and 2013 (Hemez & Manning, 2017a), and 70% of women who married between 2010 and 2014 cohabited before marriage, as compared to 40% of women marrying between 1980 and 1984 (Hemez & Manning, 2017b). The median age at first cohabitation, about 22 for women and 24 for men, changed little between the 1980s and 2000s. The median age at marriage rose during that period from 23 to 26 for women and 25 to 28 for men (Manning et al., 2014). Turning to union dissolution, Guzzo (2014) found that the dissolution risk of cohabitation increased between 1980 and 1995, and increased even more rapidly after 2000. Cohabiting unions were more likely than marriages to dissolve in the 2000s as compared to the 1980s (Guzzo, 2014). In contrast to cohabitation dissolution, divorce rates fell for those in their twenties from 1980–1990 to 2008–2010; the divorce rate was around 42% for 25 to 29 year olds in 1980 and 1990, and around 36% for 25 to 29 year olds in 2008–2010 (Kennedy & Ruggles, 2014).

The research that has examined interaction effects in the predictors of union formation and dissolution have found mixed results. For union formation, Manning et al. (2014) established a delay of first marriage and the earlier onset of cohabitation occurred similarly for all race/ethnic groups from 1988 to 2010 in the National Survey of Family Growth. Similarly, Manning et al. found that marriage was delayed for all education levels, but the delay was greater for those with less education, an interaction effect. Across education levels, the median age at first cohabitation remained relatively stable over time, but there was a single exception. Female high school graduates experienced a greater decline in their age at first cohabitation over the 20 years (Manning et al., 2014). Teachman (2002) examined whether the predictors of divorce have changed as divorce increased. Also using the National Survey of Family Growth, Teachman (2002) found little cohort variation in the predictors of divorce. The one interaction effect found a convergence in the effect of race on divorce; young adults who were White experienced a more rapid growth in their risk of divorce than did young adults who were Black. Thus, the risk of divorce for young adults who were Black was greater in magnitude compared to Whites in earlier periods compared to later periods.

This paper focuses on additional sociodemographic predictors of union formation and dissolution that have not been tested for interaction effects. Established predictors of union formation and dissolution include the following. Socioeconomic advantage and education is associated with later marriage, decreased cohabitation, and decreased union dissolution (Addo, 2014; Martin, 2006; Sweeney, 2002). Religiosity is associated with earlier marriage, decreased cohabitation, and decreased union dissolution (Eggebeen & Dew, 2009; Uecker & Stokes, 2008; Vaaler, Ellison, & Powers, 2009). Black women are more likely to cohabit and less likely to marry compared to their White and Hispanic counterparts (Manning et al., 2014). Other demographic characteristics associated with socioeconomic problems that are also associated with delayed marriage or cohabiting unions that end in dissolution including having a nonmarital birth and a less educated mother, and growing up in a non-intact family (Guzzo, 2014; South, 2001). Pregnancy and the presence of children are associated with earlier unions, but increased cohabitation and later marriage, and decreased union dissolution (Graefe & Lichter, 2002; Lichter, 2012). Because predictors of union formation, type, and dissolution have been established and reviewed elsewhere (Amato, 2010; Sassler, 2010), we focus our theoretical discussion on hypothesized interaction effects.

Theoretical Perspectives

As union formation and dissolution trends have changed around the world (Lesthaeghe, 2010; Ortega, 2014), theories have emerged to explain them. We draw on literature from several disciplines to paint a broad picture of two such theories: 1) the second demographic transition, and 2) accelerated adulthood. Each theory led to a set of hypotheses, sometimes competing, about interaction effects in the predictors of union formation and dissolution.

Second demographic transition theory.

The changes to union formation and dissolution since the 1950s, which include delayed marriage and childbearing, increased nonmarital cohabitation and births, and union instability have been characterized as the second demographic transition (Lesthaeghe, 2010; Lesthaeghe & Neidert, 2006). Undergirded by the widespread availability of oral contraceptives, which led to the sexual revolution and the gender revolution, Lesthaeghe (2010) suggested that ideational change occurred. Specifically, ideational theory suggests that as countries became increasingly educated, economically developed and more secularized, societal focus shifted from the fulfillment of basic material needs to higher order needs including self-actualization and individual autonomy. In the US, economists have argued that gains to marriage have shifted over the past 50 years from maximizing efficiency to maximizing personal satisfaction (Stevenson & Wolfers, 2007). Sociologist Cherlin (2004) has suggested that marriage is now an “individualized, choice based achievement” (pp. 858). Psychologist Eli Finkel and colleagues contended that spouses now ask marriage to help them “achieve goals relevant to esteem and self-actualization” whereas before marriage was meant to fulfill basic physiological and safety needs (pp. 2, Finkel, Hui, Carswell, & Larson, 2014).

The theory of the second demographic transition further argues that as countries prosper and their educated citizens find it easier to have their basic physiological and safety needs met, the entire society shifts to valuing self-actualization and individual autonomy (Lesthaeghe & Surkyn, 1988). As a result, religiosity decreases, gender and sexual behavior norms become less rigid and family formation behavior change follows, including delayed marriage, increased cohabitation, and increased union dissolution. Importantly, ideational theory argues that the most highly educated in society will be the vanguard of these changes because they will be the first to be exposed to self-actualization and individual autonomy as values (Lesthaeghe & Surkyn, 1988; McLanahan & Jacobsen, 2015). In particular, cohabitation offers flexibility for couples to explore a long-term union with the possibility of an easy exit (Perelli-Harris & Bernardi, 2015). As ideational change and relaxed gender and sexual behavior norms spread, nonmarital sex, cohabitation, and single motherhood become increasingly acceptable (Cherlin, 2014).

The second demographic transition had already begun in the United States during the young adulthoods of the NLSY79, and twenty years later during the young adulthoods of the NLSY97, theoretically ideational change should be spreading to the less educated across the US (Lesthaeghe & Neidert, 2006). We hypothesized that the association between education and union formation would diminish as ideational change spread. Further, due to the relaxing of gender, sexual, and divorce norms as ideational change spreads, we also expected that the association between pregnancy, the presence of children, and growing up in an intact family and union formation would also diminish. We made no specific hypotheses regarding interaction effects for union dissolution because the second demographic transition theory supports competing hypotheses. The vanguards of the second demographic transition, those with more education (Lesthaeghe & Surkyn, 1988), may expect their unions to fulfill their higher order needs and choose a partner who maximizes their personal satisfaction (Finkel et al., 2014; Stevenson & Wolfers, 2007). If the vanguards were successful, they should have more stable unions. Indeed, Lesthaeghe and Neidert (2006) found that conservative US states, which they argue have been less affected by the second demographic transition, have higher divorce rates. However, the expectations of those with more education may be too high for any one person to fulfill (Finkel et al., 2014) and they also hold more favorable attitudes toward divorce (Toth & Kemmelmeier, 2009). These forces would undermine union stability.

Hypothesis 1a.

Higher education, higher maternal education, school enrollment, lack of pregnancy, and no children in the household will be more strongly associated with a delay in union formation in the NLSY79 cohort as compared to the NLSY97 cohort.

Hypothesis 1b.

Higher education, higher maternal education, school enrollment, lack of pregnancy, and no children in the household will be more strongly associated with an earlier entrance into cohabitation and a delay in marriage in the NLSY79 cohort as compared to the NLSY97 cohort.

Accelerated adulthood.

Given the shrinking middle class and rising income inequality in the US and growing stratification (Saez & Zucman, 2016), the failure of theories such as the second demographic transition to account for the role of socioeconomic diversity is significant (Zaidi & Morgan, 2017). Lee (2014) and Roy, Messina, Smith, and Waters (2014) proposed that disadvantaged young adults experience an “accelerated adulthood” due to economic and educational disadvantage. Many disadvantaged young adults enter adult roles early, as children or adolescents, and do so not out of choice, but because of constraints (Lee, 2014). Roy et al. (2014) found that among young men in youth development programs, many had taken on adult roles such as caring for others including cooking and dressing younger children, providing financially, and dealing with community violence as boys and adolescents. Low-income youth are more likely to leave home early and become parents early, but delay marriage (Berzin & De Marco, 2010), strategies that may even be developmentally beneficial as they reduce uncertainty and put disadvantaged youth on a known pathway (Friedman, Hechter, & Kanazawa, 1994).

Disadvantaged youths lack resources and their early transition to adult roles may prevent them from focusing on identity development and self-actualization, and hence delay modern marriage with its focus on maximizing personal satisfaction. The accelerated adulthood theory suggests that socioeconomically disadvantaged young adults may enter unions to save money through economies of scale by sharing rent, groceries, and other living expenses. Indeed, McLanahan and Jacobsen (2015) argued that the destinies of more and less socioeconomically stable young adults in developed countries have diverged such that those with fewer resources enter cohabitation and unmarried parenthood earlier and fail to marry, and those with more resources delay but achieve marriage, most often after cohabitation, and have children after marriage. Similarly, Perelli-Harris and Gerber (2011) contended that cohabitation was the result of a pattern of disadvantage, rather than a culture choice as suggested by the theory of the second demographic transition. In support of the theory of accelerated adulthood, practical considerations such as convenience and economic need are often cited as motivations for cohabitation in the US, particularly among working-class young adults (Sassler & Miller, 2011). In addition, attitudinal studies find rising and perhaps unrealistic expectations for marriage among lower income young adults (Gibson-Davis, Edin, & McLanahan, 2005; Silva, 2013) which may make it more difficulty for a relationship to lead to marriage as those expectations are hard to fulfill. When less socioeconomically stable young adults do form unions, those unions may be less stable. Trail and Karney (2012) found that low-income couples had similar attitudes toward marriage and similar relationships problems as higher-income couples. However, the economic and social problems lower-income couples faced undermined these relationships. The gap between the advantaged and disadvantaged grew over the 20 year period of this study as income inequality grew (Saez & Zucman, 2016). This rising income inequality could lead to two possibilities. First, there may be compositional changes such that the NLSY97 has earlier union formation, earlier cohabitation, delayed marriage, and an increased risk of union dissolution as compared to the NLSY79. But, there may be no significant interaction effects because the association between socioeconomic risk factors and union formation and dissolution may be the same for the two cohorts. It is also possible that the rising income inequality has led to an exaggeration of the association between socioeconomic risk factors such that socioeconomic risk factors are more strongly associated with early union formation and union dissolution in the NLSY97 as compared to the NLSY79. Due to these competing hypotheses, we do not assert specific hypotheses related to the theory of accelerated adulthood.

Data

Data came from the National Longitudinal Survey of Youth 1979 and 1997 (NLSY79 and NLSY97), two comparable, US nationally representative studies. The NLSY79 includes 12,686 individuals born in 1957 to 1964; respondents were interviewed annually from 1979 to 1994 and biennially since (data through 2012 are publicly available). The NLSY97 includes 8,984 individuals who were born between 1980 and 1984; respondents have been interviewed annually since 1997 and data through 2013 were publicly available. The two datasets are well suited for cohort comparisons as both provide a wide range of life course experiences for these two cohorts and have similar survey designs (Bureau of Labor Statistics, 2014, 2015). We restricted our analytic samples to ages 16 to 30 for each cohort.

Measures

First union formation and dissolution.

The NLSY79 provides information on the current spouse or unmarried partner (i.e., living with an opposite sex partner) at each interview, and the exact dates of marriages and divorces (up to three). The exact dates of cohabitation entrances and exits were not available until 2002, but starting in 2002, retrospective data were collected on nonmarital cohabiting relationships during any unmarried spells. We first used the current partner information at each wave. If these data were missing, the exact dates of marriage and cohabitation were used. Monthly marital and cohabiting data are available in the NLSY97, where cohabitation was defined as “a sexual relationship in which partners establish one household and live together.” In both cohorts, annual union formation and dissolution histories were created; first union formation was measured as either direct marriage (marriage without cohabitation), or cohabitation (living with a non-married partner), and union dissolution was measured as direct marriage dissolution (divorce of first direct marriage) and cohabitation dissolution (the end of cohabiting relationship in break-up). Note that cohabiting unions that became marriages were treated as continuing cohabiting unions.

Pregnancy and children in the household.

Pregnancy was measured as 9 months before the date of the birth of a child for both NLSY79 and NLSY97; a time-varying dichotomous variable was created each year (0=no, 1=pregnant). While NLSY97 provides detailed information of current pregnancy (e.g., exact weeks of pregnancy), no symmetric variable is available in NLSY79, especially for men (Center for Human Resource Research, 2017). A limitation is that the NLSY only captures pregnancies resulting in a live birth. Respondents in both cohorts were asked about the number of biological, step, and adopted children in the household each year. A time-varying dichotomous variable was created to indicate 0=no, 1=children in the household each year.

Individual characteristics.

Gender was coded as 0=male and 1=female. Race/ethnicity was included as non-Hispanic Black and Hispanic; non-Hispanic non-Black (White) served as a reference group. Education included a time-varying measure of the highest degree completed (4-year college, 2-year college, high school, and less than high school [reference group]), and a time-varying measure of the current school enrollment status (0=no, 1=yes). Employment status was included as a time-varying measure of full-time employment (worked more than 35 hours per week for at least 50 weeks) and part-time employment (worked other than full-time) each year; those not working were in a reference group. Personal income was included as a yearly time-varying measure of continuous personal incomes from annual wages, salary and tips. Religion was included as a categorical variable of the present religious affiliation of respondents in the first interview (Catholic, Protestant, other religion, and no religion as a reference category).

Family background.

Maternal education was included as the highest grade the respondent’s mother completed. Intact family was coded as a dichotomous indicator of whether respondents had lived with both biological parents until age 18 for NLSY79 and at the first interview in 1997 for NLSY97.

Analytic Plan

Descriptive statistics included the comparison of individual and family background characteristics by cohorts. Detailed union formation and dissolution patterns between ages 16 and 30 were also reported by cohort along with Kaplan-Meier survival estimates. The difference in the Kaplan-Meier survival estimates by cohorts was tested using the log-rank test of equality.

Cox proportional hazards models were used to examine how first union formation changed across cohorts in relation to individual and family background characteristics. The onset of risk was at age 16, and the failure, union entrance, was the year respondents first married or cohabited. We also conducted a competing risks model in which the hazard of direct marriage and cohabitation were presented as a competing risk to being single. These models included three sets of analyses in which cohort was treated as a covariate, as well as models estimated separately by cohort. To test for interaction effects, a model was also run in which each covariate was interacted with a dummy variable for the NLSY97 cohort. A similar strategy was followed for union dissolution. The onset of the union dissolution was set to the year of union formation and failure was set to the year the respondent ended their marriage or cohabitation through breakup. Separate Cox models were used for divorce of first direct marriage and first cohabitation dissolution.

Missing values on predictor variables were imputed via multiple imputation in Stata. The chained-equations method (i.e. truncated regression and multinomial logit) was used; the MI prefix allowed for pooled estimates of missing values across 25 imputed datasets (Johnson & Young, 2011). The survey setting command in Stata were also implemented to account for the complicated sampling strategy of the NLSY data sets and thus provide robust statistical estimates (Cleves, Gutierrez, Gould, & Marchenko, 2010).

Results

Table 1 reports weighted descriptive statistics of individual and family background characteristics, and union formation and dissolution patterns by cohort. About half of respondents in both cohorts were female. A majority of the sample was White although the NLSY97 was more racially diverse than the NLSY79. A similar proportion of young adults in both cohorts completed 2-year or 4-year college degrees but a greater proportion of young adults in NLSY97 were enrolled in school. A greater proportion of those in NLSY79 were employed. The NLSY79 reported more pregnancies and were more likely to have children in the household. Young adults in NLSY79 reported greater incomes. A greater proportion of NLSY97 respondents reported no religion. The NLSY79 was more likely to grow up in an intact family. Mothers of NLSY79 young adults had on average less than a high school education, whereas mothers of NLSY97 young adults had slightly more than a high school education on average.

Table 1.

Weighted Descriptive Statistics

Variable		NLSY79 (n = 12686)			NLSY97 (n = 8984)
		M	SD	Range	M	SD	Range
Individual characteristics
Age		23.00a	4.32	16–30	22.54	4.01	16–30
Female		0.49	-	0–1	0.49	-	0–1
Race/ethnicity
White (ref)		0.79a	-	0–1	0.72	-	0–1
Black		0.14	-	0–1	0.15	-	0–1
Hispanic		0.07a	-	0–1	0.13	-	0–1
Education
Less than High School (ref)		0.12a	-	0–1	0.16	-	0–1
High school or equivalent		0.58a	-	0–1	0.50	-	0–1
2 year College		0.07	-	0–1	0.07	-	0–1
4 year College		0.23	-	0–1	0.26	-	0–1
Enrollment status		0.24a	-	0–1	0.35	-	0–1
Employment
Not working (ref)		0.12a	-	0–1	0.21	-	0–1
Part-time employment		0.50a	-	0–1	0.55	-	0–1
Full-time employment		0.38a	-	0–1	0.24	-	0–1
Ever pregnant		0.64a	-	0–1	0.45	-	0–1
Children in the household		0.60	-	0–1	0.47	-	0–1
Incomes ($ in 2013)		19277.25a	19622.3	0–204751.7	18481.64	19410.2	0–180331.0
Religion
No religion (ref)		0.04a	-	0–1	0.19	-	0–1
Catholic		0.32a	-	0–1	0.21	-	0–1
Protestant		0.52a	-	0–1	0.45	-	0–1
Other religion		0.12a	-	0–1	0.14	-	0–1
R lived in an intact family		0.63a	-	0–1	0.53	-	0–1
Maternal education (years)		10.87a	3.17	0–20	12.44	2.91	1–20
Union formation
Any union	Proportion	0.81a	-	0–1	0.73	-	0–1
	Mean age	23.24a	0.07	16–30	22.48	0.08	16–30
	Median age	23.00a	-	16–30	22.00	-	16–30
Direct Marriage	Proportion	0.57a	-	0–1	0.19	-	0–1
	Mean age	25.06a	0.11	17–30	25.89	0.17	18–30
	Median age	25.00a	-	17–30	26.00	-	18–30
All Marriage	Proportion	0.69a	-	0–1	0.35	-	0–1
	Mean age	23.31a	0.08	16–30	23.86	0.10	16–30
	Median age	23.00a	-	16–30	24.00	-	16–30
Cohabitation	Proportion	0.24a	-	0–1	0.54	-	0–1
	Mean age	23.98a	0.10	16–30	22.23	0.08	16–30
	Median age	24.00a	-	16–30	22.00	-	16–30
Union dissolution
Any union		(n=9823)			(n=6456)
	Proportion	0.34a	-	0–1	0.49	-	0–1
	Mean duration	3.31a	0.08	1–15	2.54	0.05	1–13
	Median duration	2.00a	-	1–15	2.00	-	1–13
Direct marriage		(n=6846)			(n=1667)
	Proportion	0.32a	-	0–1	0.29	-	0–1
	Mean duration	3.99a	0.10	1–15	4.47	0.14	1–12
	Median duration	3.00a	-	1–15	4.00	-	1–12
Cohabitation		(n=2977)			(n=4789)
	Proportion	0.38a	-	0–1	0.56	-	0–1
	Mean duration	1.92a	0.06	1–14	2.19	0.05	1–13
	Median duration	1.00	-	1–14	1.00	-	1–13

Note:

^aindicates significant difference between cohorts.

Union Formation

The middle panel of Table 1 presents union formation descriptive statistics by cohort. Overall, about 81% and 73% of young adults in the NLSY79 and NLSY97 entered their first union by age 30; mean age at first union was between 22 and 23. A majority of the NLSY79 directly married as a first union. The mean age at first direct marriage was around 25 for both cohorts. The average age at first cohabitation was younger for the NLSY97. Kaplan Meier estimates (see Figure 1) of union formation showed that those in NLSY97 entered unions earlier than their counterparts in NLSY79, a compositional effect.

Figure 1.

Kaplan-Meier Survival Estimates for union formation

Note: χ² = 0.02 (p = .90) significant differences were tested using log-rank test for equality of survivor fucntions.

Any union.

The Cox Proportional Hazard models predicting union formation are presented in Table 2. Overall, the NLSY97 had earlier first unions than the NLSY79. Males, young adults who identified as Black, and those enrolled in school entered unions later. Those employed either part or full-time and those pregnant entered unions earlier, but those with children in the household delayed union formation. Those who were Catholic and other religion entered unions later than those with no religion. The children of intact families entered unions later than the children of divorce. Income was positively related to the hazard of union formation but maternal education was negatively related to the hazard of union formation. Turning to the interaction effects, young adults in the NLSY97 who had 4 years or more college entered a union later than those in the NLSY79 who had 4 years or more college. The positive association between income and earlier union formation was diminished for the NLS97 cohort as compared to the NLSY79.

Table 2.

Cox Proportional Hazards Models Predicting First Union Formation

Variable	NLSY79+97		NLSY79		NLSY97		NLSY79+97 with interaction effects
	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.
Cohort1997	0.13 (0.04) ***	1.13					0.61 (0.20) **	1.85
Gender
Female	0.45 (0.02) ***	1.56	0.36 (0.02) ***	1.44	0.33 (0.03) ***	1.39	0.36 (0.03) ***	1.43
Race/ethnicity (ref: Whites)
Black	−0.46 (0.03) ***	0.63	−0.53 (0.03) ***	0.59	−0.47 (0.04) ***	0.63	−0.52 (0.03) ***	0.59
Hispanic	−0.01 (0.04)	0.99	−0.04 (0.04)	0.96	−0.04 (0.04)	0.96	−0.04 (0.04)	0.96
Education (ref: less than HS)
High school or equivalent	0.00 (0.04)	1.00	−0.06 (0.04)	0.94	0.05 (0.04)	1.05	0.02 (0.04)	1.02
2-year college	0.04 (0.08)	1.04	−0.08 (0.08)	0.93	0.11 (0.10)	1.11	0.03 (0.08)	1.03
4-year college or more	0.07 (0.06)	1.07	0.06 (0.06)	1.06	0.05 (0.07)	1.05	0.16 (0.06) **	1.18
Enroll in school	−0.18 (0.04) ***	0.83	−0.18 (0.03) ***	0.83	−0.22 (0.04) ***	0.80	−0.22 (0.03) ***	0.80
Employment (ref: not working)
Employed part-time	0.16 (0.06) **	1.18	0.28 (0.05) ***	1.32	0.22 (0.05) ***	1.25	0.24 (0.05) ***	1.27
Employed full-time	0.33 (0.07) ***	1.39	0.43 (0.06) ***	1.53	0.35 (0.06) ***	1.42	0.39 (0.06) ***	1.48
Pregnant	0.50 (0.06) ***	1.65	0.53 (0.06) ***	1.71	0.53 (0.07) ***	1.70	0.81 (0.08) ***	1.72
Children in the household	−0.21 (0.07) **	0.81	−0.16 (0.06) **	0.85	−0.21 (0.07) **	0.81	−0.16 (0.06) **	0.85
Religion (ref: no religion)
Catholic	−0.20 (0.04) ***	0.82	−0.12 (0.06)	0.89	−0.12 (0.06) *	0.89	−0.12 (0.07)	0.88
Protestant	−0.04 (0.04)	0.96	0.00 (0.06)	1.00	0.04 (0.04)	1.04	−0.00 (0.06)	1.00
Other religion	−0.09 (0.04) *	0.91	0.04 (0.07)	1.04	−0.12 (0.05) *	0.89	0.03 (0.07)	1.03
Logged income	0.04 (0.01) **	1.04	0.06 (0.01) ***	1.06	0.06 (0.02) ***	1.06	0.07 (0.01) ***	1.07
R lived in an intact family	−0.25 (0.02) ***	0.78	−0.16 (0.02) ***	0.85	−0.16 (0.03) ***	0.85	−0.15 (0.02) ***	0.86
Maternal education (years)	−0.04 (0.00) ***	0.96	−0.02 (0.00) ***	0.98	−0.02 (0.01) ***	0.98	−0.02 (0.00) ***	0.98
Cohort interaction
Female*1997							−0.05 (0.04)	0.95
Black* 1997							0.02 (0.05)	1.02
Hispanic* 1997							−0.02 (0.06)	0.98
High school/equivalent*1997							−0.03 (0.05)	0.97
2 yr college * 1997							−0.00 (0.13)	1.00
4 yr college or more * 1997							−0.17 (0.09) *	0.84
Enroll in school * 1997							0.09 (0.05)	1.09
Employed part-time* 1997							−0.02 (0.07)	0.98
Employed full-time* 1997							−0.07 (0.09)	0.94
Pregnant * 1997							0.00 (0.10)	1.00
Children * 1997							−0.06 (0.09)	0.94
Catholic * 1997							−0.01 (0.08)	0.99
Protestant * 1997							0.04 (0.08)	1.04
Other religion * 1997							−0.17 (0.09)	0.84
Logged income* 1997							−0.04 (0.02) *	0.96
R in an intact family*1997							−0.05 (0.04)	0.95
Maternal education* 1997							−0.01 (0.01)	0.99
n	21188		12478		8710		21188

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Direct marriage, cohabitation, and single as competing risks.

Findings from competing risks models for first union formation type are presented in Table 3. Compared to young adults in the NLSY79, those in the NLSY97 directly married later and cohabited earlier. Although cohabitation grew over this time (Guzzo, 2014), initial cohabitation dates were not recorded for the NLSY79 until the 1990s, hence cohabitation was under-reported and direct marriage may be overestimated in the NLSY79. Females were married and cohabited earlier than males. Young adults with 4 year college or more education directly married earlier but cohabited later. Young adults employed full-time directly married and cohabited earlier than those not working although only those employed part-time cohabited earlier. Pregnancy was significantly related to greater hazards of both direct marriage and cohabitation but having children in the household was associated with lower hazards of cohabitation. Those who were Catholic, Protestant, and other religion cohabited later than those without any religion. Income was positively associated, and maternal education negatively associated, with the hazards of direct marriage.

Table 3.

Competing Risks Models Predicting First Union Formation by Type

Variables	Ref: no union formation
	Direct Marriage				Cohabitation
	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.
Cohort1997	−0.84 (0.06) ***	0.43	−1.96 (0.45) ***	0.14	1.05 (0.06) ***	2.86	2.12 (0.44) ***	17.46
Gender
Female	0.47 (0.04) ***	1.60	0.48 (0.04) ***	1.62	0.42 (0.03) ***	1.52	0.44 (0.07) ***
Race/ethnicity (ref: Whites)
Black	−0.67 (0.05) ***	0.51	−0.62 (0.06) ***	0.54	−0.24 (0.05) ***	0.79	−0.23 (0.08) **	0.96
Hispanic	0.06 (0.06)	1.07	0.02 (0.07)	1.02	−0.09 (0.05)	0.92	−0.18 (0.11)	1.00
Education (ref: less than HS)
High school or equivalent	0.02 (0.07)	1.02	0.03 (0.09)	1.03	−0.03 (0.04)	0.97	0.02 (0.10)	1.14
2 yr college	0.12 (0.14)	1.13	0.10 (0.17)	1.10	−0.06 (0.11)	0.94	0.14 (0.19)	1.27
4 yr college or more	0.24 (0.09) *	1.27	0.29 (0.11) *	1.34	−0.18 (0.09) *	0.83	−0.09 (0.15)	0.92
Enroll in school	−0.10 (0.06) f	0.90	−0.07 (0.07)	0.93	−0.29 (0.05) ***	0.75	−0.51 (0.09) ***	0.52
Employment (ref: not working)
Employed part-time	0.08 (0.11)	1.09	0.10 (0.13)	1.10	0.26 (0.07) ***	1.30	0.20 (0.17)	1.48
Employed full-time	0.31 (0.11) **	1.37	0.34 (0.14) *	1.40	0.33 (0.08) ***	1.39	0.22 (0.19)	1.41
Pregnant	0.45 (0.10) ***	1.57	0.55 (0.12) ***	1.74	0.57 (0.08) ***	1.76	0.52 (0.14) ***	2.00
Children in the household	−0.20 (0.12) f	0.82	−0.21 (0.15)	0.81	−0.23 (0.08) **	0.80	−0.15 (0.16)	0.98
Religion (ref: no religion)
Catholic	0.08 (0.10)	1.08	−0.21 (0.13)	0.81	−0.26 (0.05) ***	0.77	−0.27 (0.14) *	0.68
Protestant	0.36 (0.09) ***	1.44	0.02 (0.13)	1.02	−0.29 (0.05) ***	0.75	−0.41 (0.13) **	0.66
Other religion	0.38 (0.10) ***	1.46	0.05 (0.14)	1.05	−0.44 (0.07) ***	0.65	−0.40 (0.15) **	0.60
Logged income	0.07 (0.02) **	1.08	0.06 (0.03) *	1.06	0.00 (0.02)	1.00	0.07 (0.04) f	1.14
R lived in an intact family	−0.10 (0.04) **	0.91	−0.14 (0.05) **	0.87	−0.43 (0.04) ***	0.65	−0.59 (0.07) ***	0.59
Maternal education	−0.06 (0.01) ***	0.94	−0.06 (0.01) ***	0.94	−0.02 (0.01) **	0.98	0.00 (0.01)	0.99
Cohort interaction
Female*1997			−0.09 (0.08)	0.91			−0.08 (0.08)
Black* 1997			−0.25 (0.12) *	0.78			−0.04 (0.09)	0.90
Hispanic* 1997			0.09 (0.11)	1.09			0.10 (0.13)	0.88
High school/equivalent*1997			0.09 (0.12)	1.09			−0.10 (0.11)	0.83
2 yr college * 1997			V0.10 (0.24)	1.11			−0.33 (0.23)	0.73
4 yr college or more * 1997			−0.24 (0.16)	0.79			−0.13 (0.17)	0.99
Enroll in school * 1997			−0.10 (0.11)	0.91			0.40 (0.10) ***	1.74
Employed part-time* 1997			−0.15 (0.18)	0.86			0.06 (0.18)	0.78
Employed full-time* 1997			−0.19 (0.19)	0.82			0.09 (0.20)	0.89
Pregnant * 1997			−0.29 (0.25)	0.75			0.05 (0.17)	1.10
Children * 1997			0.08 (0.23)	1.08			−0.07 (0.18)	1.14
Catholic * 1997			0.65 (0.19) **	1.92			−0.05 (0.15)	1.06
Protestant * 1997			0.96 (0.18) ***	2.61			0.13 (0.14)	1.16
Other religion * 1997			0.95 (0.20) ***	2.59			−0.16 (0.17)	0.93
Logged income* 1997			0.05 (0.04)	1.05			−0.10 (0.04) *	0.86
R in an intact family*1997			0.31 (0.09) ***	1.36			0.22 (0.08) **	1.22
Maternal education* 1997			0.00 (0.01)	1.00			−0.04 (0.02) *	0.96
n	21188

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Turning to interaction effects, the negative association between being Black and the hazard of direct marriage grew over time such that Black young adults in the NLSY97 directly married later, or forewent marriage, more than Black young adults in the NLSY79. School enrollment was more strongly positively associated with the hazards of cohabitation in the NLSY97 as compared to the NLSY79. Young adults in NLSY97 who were Catholic, Protestant, or with other religion directly married earlier than their counterparts in NLSY79. Greater income and higher maternal education were more strongly associated with delayed cohabitation in NLSY97 as compared to the NLSY79. Growing in an intact family was associated with greater hazards of both direct marriage and cohabitation in the NLSY97 as compared to the NLSY79

Union Dissolution

Descriptive statistics.

The bottom panel of Table 1 includes descriptive statistics of first union dissolution. A greater proportion of young adults in NLSY97 dissolved their first union between ages 16 and 30. Specifically, among those who directly married, 32% and 29% of NLSY79 and NLSY97 respectively were divorced, and among those cohabiting, 38% and 56% of NLSY79 and NLSY97, respectively, dissolved their union. The average duration of first cohabitations was significantly longer for those in the NLSY97. Figure 2 presents Kaplan-Meier estimates of first union dissolution within 5 years; a greater proportion of those in NLSY97 ended their first union and dissolved it earlier than those in NLSY79.

Figure 2.

Kaplan-Meier Survival Estimates for union dissolution

Note: χ² = 315.96 (p < .001) significant differences were tested using log-rank test for equality of survivor fucntions.

Any union dissolution.

Findings from Cox proportional hazards models for union dissolution are reported in Table 4. Young adults in NLSY97 ended their first union more quickly than those in the NLSY79. In both cohorts, respondents who were Black were more likely to end their first union. Those who were pregnant, had any religion, had a greater income, and were from an intact family ended their first union later in both cohorts. Turning to the interaction effects, being Hispanic was associated with lower hazards of union dissolution in the NLSY97 as compared to the NLSY79. Having more than high school education (i.e., high school diploma, 4 year college or more education) and not working were more strongly positively associated with the hazard of union dissolution in the NLSY97 as compared to the NLSY79. The NLSY97 youth who had a mother with more education had lower hazards of union dissolution compared to NLSY79 young adults.

Table 4.

Cox Proportional Hazards Models Predicting First Union Dissolution

Variable	NLSY79+97			NLSY79		NLSY97		NLSY79+97 with interaction effects
	Coef. (s.e.)		H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)	H.R.
Cohort1997	0.13	0.05 *	1.13					1.30 (0.45) **	3.31
Gender
Female	0.01	0.05	1.01	0.06 (0.07)	1.06	0.02 (0.05)	1.02	0.07 (0.07)	1.07
Race/ethnicity (ref: Whites)
Black	0.42	0.05 ***	1.53	0.49 (0.07) ***	1.63	0.32 (0.06) ***	1.38	0.46 (0.07) ***	1.59
Hispanic	0.18	0.06 **	1.20	0.34 (0.09) ***	1.41	−0.01 (0.07)	0.99	0.34 (0.10) ***	1.40
Education (ref: less than HS)
High school or equivalent	−0.20	0.05 ***	0.82	−0.37 (0.08) ***	0.69	−0.08 (0.07)	0.92	−0.33 (0.08) ***	0.72
2-year college	−0.58	0.14 ***	0.56	−0.83 (0.20) ***	0.44	−0.32 (0.16) *	0.73	−0.83 (0.21) ***	0.43
4-year college or more	−0.83	0.10 ***	0.44	−1.13 (0.14) ***	0.32	−0.55 (0.12) ***	0.58	−1.10 (0.14) ***	0.33
Enroll in school	−0.01	0.07	0.99	0.05 (0.12)	1.05	−0.03 (0.08)	0.97	0.07 (0.11)	1.07
Employment (ref: not working)
Employed part-time	0.00	0.07	1.00	0.19 (0.11)	1.20	−0.25 (0.08) ***	0.78	0.16 (0.12)	1.18
Employed full-time	−0.20	0.08 *	0.82	−0.01 (0.13)	0.99	−0.45 (0.09) ***	0.64	−0.04 (0.14)	0.96
Pregnant	−0.25	0.07 ***	0.78	−0.22 (0.09) *	0.80	−0.32 (0.09) ***	0.72	−0.25 (0.09) **	0.78
Children in the household	−0.04	0.06	0.96	−0.05 (0.08)	0.95	−0.05 (0.06)	0.95	0.00 (0.08)	1.00
Religion (ref: no religion)
Catholic	−0.32	0.07 ***	0.73	−0.36 (0.13) *	0.69	S-0.10 (0.08)	0.90	−0.39 (0.14) **	0.68
Protestant	−0.20	0.07 **	0.82	−0.09 (0.14)	0.91	−0.33 (0.08) ***	0.72	−0.12 (0.14)	0.88
Other religion	−0.32	0.09 ***	0.73	−0.20 (0.15)	0.82	−0.42 (0.09) ***	0.66	−0.22 (0.16)	0.80
Logged income	−0.09	0.02 ***	0.92	−0.08 (0.03) *	0.93	−0.08 (0.02) ***	0.92	−0.07 (0.03) *	0.94
R lived in an intact family	−0.17	0.05 ***	0.84	−0.16 (0.07) *	0.86	−0.01 (0.01)	0.99	−0.15 (0.07) *	0.86
Maternal education (years)	0.01	0.01	1.01	0.03 (0.01) *	1.03	−0.19 (0.06) **	0.83	0.03 (0.01) **	1.03
Cohort interaction
Female*1997								−0.05 (0.09)
Black* 1997								−0.09 (0.10)	0.92
Hispanic* 1997								−0.34 (0.12) **	0.91
High school/equivalent*1997								0.24 (0.11) *	1.26
2 yr college * 1997								0.50 (0.27) f	1.16
4 yr college or more * 1997								0.52 (0.19) **	1.37
Enroll in school * 1997								−0.15 (0.14)	1.04
Employed part-time* 1997								−0.33 (0.15) *	0.92
Employed full-time* 1997								−0.37 (0.18) *	0.77
Pregnant * 1997								−0.07 (0.14)	0.91
Children * 1997								−0.15 (0.10) f	0.82
Catholic * 1997								0.30 (0.16) f	0.73
Protestant * 1997								−0.26 (0.16)	0.48
Other religion * 1997								−0.29 (0.18)	0.41
Logged income* 1997								−0.03 (0.04)	0.97
R in an intact family*1997								−0.05 (0.09)	0.92
Maternal education* 1997								−0.04 (0.02) *	0.99
n	15545		9438		6107			15545

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

First direct marriage divorce and first cohabitation dissolution.

Findings from Cox proportional hazard models of direct marriage divorce and cohabitation dissolution are reported in Table 5. The NLSY97 ended their direct marriages later and their cohabiting unions earlier than the NLSY79. Respondents who were Black ended their direct marriage earlier than Whites. Those with less than a high school education dissolved their direct marriage earlier than those with a high school degree. Respondents who were Catholic or other religion delayed direct marriage dissolution compared to those with no religion. Respondents who had more income and grew up in an intact family also dissolved their direct marriage’s later. Respondents who had more than a high school diploma had lower hazards of dissolving their cohabiting union than those respondents with less than a high school diploma. Employment, pregnancy, and higher income was associated with delayed cohabitation dissolution. Protestants also delayed cohabitation dissolution compared to those with no religion.

Table 5.

Cox Proportional Hazards Models Predicting Union Dissolution by Type

Variables	First Direct Marriage Dissolution					Cohabitation Dissolution
	Ref: stay in first direct marriage					Ref: stay in cohabitation
	Coef. (s.e.)		H.R.	Coef. (s.e.)	H.R.	Coef. (s.e.)		H.R.	Coef. (s.e.)	H.R.
Cohort1997	−0.24	0.08 ***	0.79	1.74 (0.76) *	5.67	0.25	0.08 **	1.29	0.33 (0.68)	1.39
Gender
Female	0.11	0.07	1.12	0.15 (0.09) f	1.16	−0.07	0.06	0.93	−0.14 (0.13)	0.87
Race/ethnicity (ref: Whites)
Black	0.32	0.07 ***	1.38	0.37 (0.08) ***	1.45	0.53	0.06 ***	1.69	0.78 (0.13) ***	2.19
Hispanic	0.21	0.09 *	1.23	0.32 (0.11) **	1.37	0.19	0.08 *	1.21	0.47 (0.19) *	1.60
Education (ref: less than HS)
High school or equivalent	−0.24	0.09 *	0.79	−0.32 (0.13) *	0.73	−0.13	0.06 *	0.88	−0.40 (0.13) **	0.67
2 yr college	−0.65	0.21 **	0.52	−0.85 (0.27) **	0.43	−0.42	0.17 *	0.66	−0.76 (0.35) *	0.47
4 yr college or more	−1.04	0.16 ***	0.35	−1.22 (0.20) ***	0.30	−0.52	0.13 ***	0.60	−0.87 (0.28) **	0.42
Enroll in school	−0.25	0.13 f	0.78	−0.20 (0.17)	0.82	0.14	0.08†	1.15	0.46 (0.14) **	1.59
Employment (ref: not working)
Employed part-time	0.15	0.13	1.16	0.29 (0.17) f	1.33	−0.19	0.09 *	0.83	−0.21 (0.20) ***	0.81
Employed full-time	0.03	0.13	1.03	0.19 (0.17)	1.21	−0.46	0.09 ***	0.63	−0.64 (0.20) **	0.53
Pregnant	−0.18	0.10 f	0.84	−0.13 (0.11)	0.88	−0.29	0.08 ***	0.75	−0.56 (0.19) **	0.57
Children in the household	0.00	0.08	1.00	0.00 (0.10)	1.00	−0.11	0.07†	0.89	−0.34 (0.14) *	0.71
Religion (ref: no religion)
Catholic	−0.43	0.12 ***	0.65	−0.46 (0.16)**	0.63	−0.14	0.10	0.87	−0.23 (0.26)	0.80
Protestant	−0.13	0.12	0.88	−0.10 (0.16)	0.91	−0.20	0.08 *	0.82	−0.20 (0.27)	0.82
Other religion	−0.33	0.12 **	0.72	−0.27 (0.17)	0.76	−0.19	0.11†	0.82	−0.08 (0.32)	0.92
Logged income	−0.08	0.03 *	0.93	−0.06 (0.04)	0.94	−0.09	0.03 ***	0.92	−0.09 (0.05)†	0.91
R lived in an intact family	−0.23	0.07 ***	0.79	−0.17 (0.08)*	0.84	−0.06	0.06	0.94	−0.05 (0.13)	0.96
Maternal education	0.01	0.01	1.01	0.03 (0.02)†	1.03	0.01	0.01	1.01	0.05 (0.03)†	1.05
Cohort interaction
Female*1997				−0.14 (0.18)	0.87				0.18 (0.15)	1.20
Black* 1997				−0.27 (0.19)	0.76				−0.41 (0.14) **	0.66
Hispanic* 1997				−0.39 (0.19)*	0.68				−0.43 (0.21) *	0.65
High school/equivalent*1997				0.24 (0.19)	1.27				0.34 (0.15) *	1.40
2 yr college * 1997				0.76 (0.38)*	2.14				0.45 (0.40)	1.57
4 yr college or more * 1997				0.67 (0.31)*	1.95				0.37 (0.31)	1.45
Enroll in school * 1997				−0.10 (0.24)	0.90				−0.48 (0.17) **	0.62
Employed part-time* 1997				−0.47 (0.25)†	0.63				0.04 (0.22)	1.04
Employed full-time* 1997				−0.62 (0.30) *	0.54				0.23 (0.23)	1.26
Pregnant * 1997				−0.53 (0.26) *	0.59				0.38 (0.21)	1.47
Children * 1997				0.03 (0.14)	1.03				0.25 (0.16)	1.29
Catholic * 1997				0.15 (0.25)	1.17				0.21 (0.28)	1.24
Protestant * 1997				−0.36 (0.24)	0.70				−0.07 (0.28)	0.93
Other religion * 1997				−0.48 (0.27)†	0.62				−0.21 (0.34)	0.81
Logged income* 1997				−0.07 (0.07)	0.94				0.02 (0.06)	1.02
R in an intact family*1997				−0.24 (0.15)	0.78				−0.01 (0.14)	0.99
Maternal education* 1997				−0.05 (0.03)†	0.95				−0.05 (0.03)†	0.95
n	8179					7366

^†p ≤ .10.

^*p ≤ .05.

^**p ≤ .01.

^***p ≤ .001.

Turning to interaction effects, young adults who identified as Hispanic, were employed, or who had a pregnancy in the NLSY97 had lower hazards of direct marriage divorce compared to the NLSY79. In contrast, young adults with 2-year college or more education had greater hazards of divorce in the NLSY97 cohort as compared to the NLSY79 cohort. Young adults who identified as Black or Hispanic or who were enrolled in school had lower hazards of cohabitation dissolution in the NLSY97 cohort as compared to the NLSY79 cohort.

Results by Gender

We also conducted all analyses separately by gender; results reported in Appendix 1. The overall pattern of results was very similar for men and women. We found that the significant interaction effects such that those who were enrolled in school, had less income, and who had less educated mothers entered unions earlier in the NLSY97 than in the NLSY79 was greater in magnitude and significant for men only. Employment hastened direct marriage was only significant for men. Three interactions predicting cohabitation formation were only significant for men. Men who reported a pregnancy, with less income, and with less educated mothers entered cohabiting unions more quickly in the NLSY97 as compared to the NLSY79. We found that employment and pregnancy delayed union dissolution in both cohorts for men. However, we found interaction effects for women such that employed and pregnant women delayed union formation more in the NLSY97 than the NLSY79. In contrast, we found that higher education hastened union dissolution for women in the NLSY97 in comparison to the NLSY79. This pattern of results for union dissolution held true for direct marriage dissolution but not for cohabitation dissolution.

Women with more highly educated mothers in both cohorts also dissolved their unions more quickly in both cohorts. Hispanic men in the NLSY97 cohort as compared to the NLSY79 cohort had more stable direct marriages. In both cohorts, Men with less income and who identified as Hispanic had shorter cohabiting unions.

Discussion

The second demographic transition theory suggested that as countries across the world have become increasingly industrialized, citizens focus on self-actualization and individual autonomy rather than basic material needs because these lower-level Maslowian needs are already fulfilled (Lesthaeghe & Neidert, 2006; Zaidi & Morgan, 2017). According to the theory, ideational changes are led by those with the most resources, most often those with socioeconomic advantages, because their safety and material needs are met. Following these ideational shifts, the importance of marriage as an institution should decline and cohabitation should rise, as should union dissolution. Lesthaeghe and Neidert (2006) highlight several trends in the US that suggest the second demographic transition happened over the twenty year span between the cohorts in the NLSY79 and NLSY97, thus we hypothesized change in union formation and dissolution behavior between the cohorts. Specifically, if as the theory states the purpose of marriage has shifted to self-actualization and esteem goals in the US, it follows that cohabitation would become increasingly popular as a trial run for marriage that could ultimately lead to more stable marriages as the least satisfying cohabitations would dissolve before marriage (Manning & Cohen, 2012). In support of the theory, the NLSY79 was more likely to directly marry and less likely to cohabit compared to the NLSY97. Further, the NLSY97 experienced more rapid union dissolution overall, and cohabitation dissolution specifically, compared to the NLSY79 perhaps because it is much more difficult to meet someone’s self- actualization needs as compared to meeting their basic safety and material needs (Finkel et al., 2014). However, the NLSY97 had more stable direct marriages. Perhaps the NLSY97 made better choices of whom to marry due to their increased expectations (Finkel et al., 2014).

Yet in critiquing second demographic transition theory, Zaidi and Morgan (2017) point out that the theory’s “silence on inequality and its emphasis on ideology suggest that all young adults have the agency and power to exercise individual freedom, achieve self-actualization, and shape their life course” (p. 486). The theory of accelerated adulthood focuses squarely on this inequality in that those with fewer resources may enter unions not to fulfill their self-actualization needs, but to fulfill basic needs for shelter and food, which may be easier to meet under economies of scale. Individuals who are more socioeconomically disadvantaged may enter cohabitation earlier, fail to marry, and have less stable unions due to economic stress and the structural challenges of inequality (Lee, 2014; Zaidi & Morgan, 2017) thus this theory also supports the pattern of results we found for union formation and dissolution. Importantly, inequality in the US increased over the twenty years between the two cohorts (Saez & Zucman, 2016). This growth in inequality, and the educational divide in marriage (Manning et al., 2014), could also explain the more stable direct marriages of the NLSY97 cohort; those most at risk for divorce have less access to marriage (McLanahan & Jacobsen, 2015).

Union formation.

We found some support for our hypotheses based on the theory of the second demographic transition. In support of Hypothesis 1A, we found that education was more strongly positively associated with a delay in union formation in the NLSY79 cohort as compared to the NLSY97. Young adults in the NLSY79 cohort with college degrees entered unions more slowly than those who dropped out of high school, and this education effect was diminished in the NLSY97 cohort. Also in support of Hypothesis 1A, we found that school enrollment was more strongly associated with earlier cohabitation in the NLSY79 cohort as compared to the NLSY97 cohort. We also found that religion was a stronger predictor of early direct marriage in the NLSY79 as compared to the NLSY97. Although not hypothesized, the theory of the second demographic transition could help explain this increased effect of religion on direct marriage formation. It may be that as ideational change spread between the NLSY79 and NLSY97 cohorts, those young adults not following the dominant societal norms may be more unique, and in the case of direct marriage, more uniquely religious given that over 70% of young adults cohabited before marriage between 2010 and 2014 (Eggebeen & Dew, 2009; Hemez & Manning, 2017b).

We also found support for the theory of accelerated adulthood. The theory of accelerated adulthood could support two patterns of results. First, although inequality grew over the twenty years between the NLSY79 and NLSY97 cohorts, the compositional effects of socioeconomic risk factors may have changed little, suggesting few interaction effects. In contrast, if the growth in inequality exacerbated the negative effects of socioeconomic risk factors on family formation, there may have been stronger associations in the NLSY97 cohort as compared to the NLSY79 cohort. We found evidence of both patterns of results. Beginning with union formation, we found the following socioeconomic risk factors predicted earlier unions: non-school enrollment, pregnancy, growing up in a nonintact family, and having a less educated mother. Early union formation is a risk factor for later union problems because age at marriage and cohabitation are both associated with union dissolution (Guzzo, 2014; Teachman, 2002), thus the positive association between these socioeconomic risk factors and earlier union formation was expected. However, we also found that young adults who were White (as compared to Black), employed, and who had no children in the household entered unions earlier. Considering the well-documented delays in marriage for young adults who are Black, unemployed, and who have children from previous relationships (Lichter, 2012; Manning et al., 2014; Sweeney, 2002), this finding may also support the theory of accelerated adulthood. Indeed, we found that in both cohorts, young adults who were White (as compared to Black), college graduates, employed full-time, and with more income entered direct marriage earlier, or entered direct marriage at all. Cohabitation was accelerated for young adults who were White (as compared to Black), had less than high school education, were employed part-time or full-time, were pregnant, and who had no children in the household. Similar to the mixed pattern of results for union formation, these results for cohabitation entrance also suggest that earlier cohabitation can serve as a pathway to marriage for young adults who are White, employed, and have no children in the household (Lichter, 2012; Manning et al., 2014; Sweeney, 2002), but can also serve as a way to enjoy economies of scale without marriage for others, including those who have less education.

Turning to the interaction effects that supported the theory of accelerated adulthood, we found that young adults who were Black and who grew up in non-intact families directly married more slowly, or forewent marriage, in the NLSY97 as compared to the NLSY79. We also found that young adults with less income and who had mothers with less education entered cohabitation more quickly in the NLSY97 as compared to the NLSY79. The growth in inequality in the US may have exacerbated the effects of these socioeconomic risk factors on direct marriage and cohabitation formation.

Union dissolution.

The theory of the second demographic transition could posit two patterns of results for union dissolution. As ideational change spread and the expectations for unions shifted from basic to higher order self-actualization needs (Finkel et al., 2014), young adults may make better choices about who to partner with, fostering more stable unions, but may also have higher expectations of relationships that are difficult to meet, undermining union stability and potentially leading to serial cohabitation, a phenomenon that has become increasingly common (Guzzo, 2014). In support of higher expectations begetting dissolution, we found that education was more strongly associated with union stability in the NLSY79 cohort; the protective effect of education diminished in the NLSY97. Kamp Dush and Arocho (2018) found a similar pattern of results in a cohort analysis of the National Survey of Family Growth such that education was less protective of union dissolution in later cohorts. Another mechanism that could underlie this finding is that as inequality grew in the US in the twenty years between the two cohorts (Saez & Zucman, 2016), the less educated maintained their unions to enjoy the benefits of economies of scale (McLanahan & Jacobsen, 2015). Further, as a result of increasing homogamy in union formation (Greenwood, Guner, Kocharkov, & Santos, 2014), those with less education may have fewer options for repartnering as compared to those with more education who may have a higher quality pool of potential partners to choose from. In examining the results by type of union, we found the same pattern of results as above for direct marriage dissolution, but not for cohabitation dissolution. We found that higher education was associated with union stability in both cohorts. This pattern of results lends some additional support to both theories. That education hastened dissolution for direct marriage only supports the second demographic transition argument that more educated, and individuated, Americans hold particularly high expectations for marriage that may not apply to cohabitation (Finkel, 2017). Further, that less education is associated with higher union dissolution supports the theory of accelerated adulthood’s argument that due to inequality, cohabitation functions as a stepping stone to marriage for those with more resources as evidenced by longer cohabiting unions of the more educated, but is an end game for those with less resources as evidenced by the shorter unions of the less educated (Guzzo, 2014).

Additionally, young adults who were Black dissolved their unions more rapidly than Whites. This result held for direct marriage, but there was an interaction effect for cohabitation dissolution. Young adults who were Black dissolved their cohabiting union less rapidly in the NLSY97 as compared to the NLSY79. These findings support the theory of accelerated adulthood such that racial inequality for young adults who were Black continued to undermine union stability, but rising income inequality may have influenced young adults who were Black in the NLSY97 cohort to stay in cohabiting unions longer because of economies of scale. Further, Black Americans are more likely to experience cohabitation dissolution rather than marriage as the outcome of their cohabiting unions (Guzzo, 2014), meaning that Black young adults may perceive fewer acceptable alternatives in the union market compared to White young adults. Interestingly, young adults who were Hispanic were at an increased risk of union dissolution in the NLSY79 cohort, but no longer had an increased risk of union dissolution in the NLSY97 cohort. This result held true for both direct marriage and cohabitation dissolution. Hispanics’ share of the US population grew over this time period (Flores, 2017). Given the rise of assortative mating over this time (Greenwood et al., 2014), perhaps the growth in the Hispanic population led to a more diverse marriage market for Hispanics in later years, leading to the formation of better matched first unions.

Interaction effects were found for employment and maternal education. The positive association between employment and union stability was greater in the NLSY97 cohort as compared to the NLSY79 cohort, but these results held true only for direct marriage dissolution and not cohabitation dissolution. This finding supports the theory of accelerated adulthood such that the negative effect of employment grew across cohorts, particularly for direct marriage, as inequality, and inequality in access to marriage specifically, also grew (McLanahan & Jacobsen, 2015). Further, higher maternal education had a negative association with union stability in the NLSY79 cohort, but a positive association with union stability in the NLSY79 cohort. In support of the theory of the second demographic transition, having a more educated mother may have led to more instability in the NLSY79 as individuation was spreading and more educated mothers taught their children to expect more from their unions. In support of the theory of accelerated adulthood, having a more educated mother in the NLSY97 cohort may have been particularly beneficial as assortative mating by education grew (Greenwood et al., 2014), leading those with less educated mothers to have fewer partnering options.

Having a pregnancy was associated with more union and cohabitation stability, but an interaction effect was found for direct marriage dissolution. Pregnancy in a direct marriage more strongly delayed marital dissolution in the NLSY97 as compared to the NLSY79. Perhaps waiting until marriage to have a first birth stabilized marriage more in the NLSY97 than in the NLSY79 because it was more normative to have a marital first birth for the NLSY79 cohort (Lamidi, 2016). We found a mixed pattern of results for religion. Young adults who were not religious dissolved their unions more rapidly than those who were Catholic, Protestant, or other religion. For direct marriage dissolution, only young adults who were Catholic and other religion had more stable marriages than those who were not religious. There are greater prohibitions to divorce for Catholics, and some other religion groups, such as Mormons and Muslims, as compared to Protestants (Cherlin, 2009). In contrast, only Protestants had more stable cohabiting unions than those who were not religious. These complicated associations suggest additional research on religion and union dissolution among younger cohorts is warrented.

In support of the theory of accelerated adulthood, we found that growing up in an intact family and having more income were positively associated with union stability and direct marriage stability, but only income had a stabilizing effect on cohabitation. Finally, we found a unique effect for school enrollment. Being enrolled in school was associated with more stable cohabiting unions in the NLSY97 cohort as compared to the NLSY79 cohort. In support of the second demographic transition, those enrolled in school may make better choices about who to partner with as they learn to become more individuated in higher education (Lesthaeghe, 2010), stabilizing their cohabiting unions.

Limitations.

Although data from the NLSY79 and 97 allowed us to compare union formation and dissolution across cohorts, several limitations of our study should be noted. First, about 20% of respondents in the NLSY97 had not reached age 30 by the last interview of 2013, and thus were right-censored in the analyses. Second, the NLSY79 began to ask a detailed cohabitation history including short-term and non-premarital cohabiting relationships in 2002. Although we drew on available data about cohabitation from the NLSY79 collected prior to 2002 (i.e., partner/spouse history from 1979 to 2012, premarital cohabitation history collected from 1990) as well as the 2002 data, cohabitation was likely underestimated for the NLSY79. Thus, there is additional error in our union formation and dissolution results for the NLSY79, and we could be overestimating direct marriage due to missing data on premarital cohabitation. Further, due to a lack of monthly data, we used yearly intervals to model union formation and stability, yielding a less precise estimate given that some unions may have been very short (Guzzo, 2014).

Third, period effects were not addressed in our study. It is possible that the meaning of cohabitation was not compatible across cohorts due to increases in cohabitation over the past few decades (Manning et al., 2014). That is, in the NLSY79, cohabitation could have been viewed more as a trial run for marriage, and in the NLSY97, as an alternative to marriage, although there is evidence that this change in the meaning of cohabitation has not occurred in the US as Americans continue to aspire to marry (Cherlin, 2009). Fourth, the exact timing of the second demographic transition in the US is up for debate. Lesthaeghe (2010) has argued that the North Atlantic, Mountain, and Pacific states in the US have more fully transitioned, whereas the South and Great Plains states have not yet fully embraced individuation. Thus, if the theory of the second demographic transition holds true, in another 20 years we may find a different pattern of results as more states, and Americans, experience the rise of individualism.

Future Research.

The theory of the second demographic transition and other ideational theories suggest that as ideational change spreads throughout a society, self-actualization and individual autonomy are increasingly prioritized (Finkel et al., 2014; Lesthaeghe & Neidert, 2006). This prioritization of individual needs is hypothesized to undermine intimate relationships because individual’s expectations of their romantic partners become so high that modern partners cannot meet them (Finkel et al., 2014). Expectations are unmet because partners do not have the time, energy, and relationship skills necessary to meet their partner’s self-actualization and esteem needs in an era where romantic partners are competing with texts, Snapchats, and news alerts for an individual’s time and attention, and most of social media promotes an idealized image of intimate relationships (Ansari & Klinenberg, 2016). We did not have symmetrical data across cohorts on relationship conflict or satisfaction, nor did we have data on time spent together or other indicators of relationship functioning. The major assumptions of ideational theories should be tested with age-period-cohort data that would allow for period and cohort effects of ideational change on intimate relationship functioning, formation, and dissolution to be disentangled.

Finally, our results supported the accelerated adulthood theory, but in complicated patterns. We found evidence that socioeconomically disadvantaged young adults had earlier unions by some indicators (e.g. lower maternal education) and later unions by other indicators (e.g. unemployment). But, socioeconomic disadvantage undermined union stability overall. We did find evidence that some indicators of socioeconomic disadvantage (e.g. income, employment, and maternal education) had exacerbated effects on union formation and stability in the NLSY97 as compared to the NLSY79 as inequality grew over that time period (Saez & Zucman, 2016). Ideational change may be spreading across the US, leading to better intimate relationships if partners are successful at meeting the higher order needs begot by higher expectations. But, if partners are coping with the stress of poverty and racism while also being expected to meet their partner’s self-actualization and esteem needs, forming and sustaining a successful long-term union may become increasingly difficult for large segments of the US population.