Using Marital Attitudes in Late Adolescence to Predict Later Union Transitions

Abstract

Using a sample of 982 late adolescents and tracking them throughout young adulthood, this study investigated if marital attitudes held during the last year of high school were predictive of union transitions to both cohabitation and marriage during young adulthood. Results using both logistic regression and discrete event history models found that marital attitudes did not have significant associations with the transition to cohabitation but did significantly predict the probability of transitioning to marriage during young adulthood. Specifically, having a younger expected age of marriage and placing more importance on marriage at the end of adolescence was associated with an increased likelihood of transitioning to marriage earlier than other young adults.

Changing demographic trends over the last 50 years, such as a later age of first marriage (Kreider, 2005) and increased attendance in secondary education (McClanahan, 2004), have changed the probable relational trajectories of individuals in contemporary Western society. Marriage is now often delayed into the late 20’s for most young adults in the United States (Kreider, 2005) and the transition to marriage has become increasingly varied as social norms surrounding marital timing have shifted. New committed relationship patterns involving cohabitation have also complicated the pathway from adolescent dating to committed marital relationships, with many couples now choosing to cohabit instead of marrying or utilizing cohabitation as a precursor to marriage (Bumpass & Lu, 2000).

Recent research focused on union formation during young adulthood has focused not on marital transitions themselves but on the marital attitudes and marital plans of young adults as they prepare for and move toward future marriages (Carroll et al., 2007; 2009). Three recent studies (Carroll et al., 2007; 2009; Willoughby & Dworkin, 2009) have suggested that sexual behavior and alcohol use during emerging adulthood are partially influenced by attitudes toward marriage, leading some scholars to suggest that marital attitudes may change trajectories through young adulthood(Carroll et al., 2007). Despite these attempts to understand how marital attitudes impact trajectories and behaviors during young adulthood, no study to date has documented how marital attitudes that form in late adolescence and young adulthood subsequently impact actual couple formation transitions. Given that the transition to marriage has been documented as a protective factor against some forms of risky behavior, such as alcohol use (Miller-Tutzaner, Leonard, & Windle, 1991), the transition to cohabitation may be associated with eventual negative couple outcomes (Dush, Cohan & Amato, 2003), and understanding union transitions continues to be of importance to family, developmental, and sociological scholars, understanding how marital attitudes may help shape trajectories into marriage and cohabitation would have important empirical and policy ramifications. In the present study, I explore how the timing of marital and cohabiting transitions during young adulthood might be predicted based on how individuals orientate toward marriage in late adolescence.

Background

Couple Formation Trends in Young Adulthood

A delay in marital timing (Schoen, Landale, & Daniels, 2007), as well as recent increases in the rates of cohabitation (Bumpass & Lu, 2000; Heuveline & Timberlake, 2004), have drawn increased scholarly attention to the union formation of young adults. Most young married adults now have cohabitated prior to marriage (Kennedy & Bumpass, 2008) and many individuals who enter cohabitation see it as precursor to marriage (Bumpass & Lu, 2000). As these trends have emerged and continue to evolve, social scientists have become interested in how behaviors and experiences during an extended period of young adulthood influence later family formation patterns and outcomes.

These changing patterns also interest researchers as some scholars have speculated that these cultural and demographic shifts in couple formation may have important ramifications for individual and family development. Developmental scholars have argued that increased variability in union formation trajectories has created unique challenges and opportunities for adolescents and young adults growing up in an increasingly eclectic world (Côté, 2000). Some research has supported this notion, with the transition to marriage during young adulthood being associated with numerous behavioral changes, particularly a decrease in health compromising risk-taking behavior (Arnett, 1998; Donovan, Jessor, & Jessor, 1983). Other union formation forms, such as cohabitation, tend to be associated with negative couple outcomes such as higher divorce rates if the couple goes on to marry and less relationship satisfaction (Dush, Cohan & Amato, 2003; Hansen, Moum & Shapiro, 2007).

Knowing that eventual transitions to committed union forms such as marriage and cohabitation are important components of individual well-being and development, one of the primary research priorities in this area is developing conceptual models through which young adults navigate the transition from adolescence to adulthood and family formation. Although many factors influence young adults’ decisions regarding union transition, attitudes toward marriage have become increasingly of interest to developmental and family scholars.

Marital Attitudes and Couple Formation Transitions

Research on the influence of marital attitudes on actual family formation transitions has been limited (Sassler & Schoen, 1999). A few older studies have found a link between positive attitudes toward marriage and couple formation behavior Clarkberg and colleagues (1995) found that the importance placed on marriage during young adulthood increased the likelihood of a transition to marriage or cohabitation and lowered the probability that the transition was to a cohabiting union. Axinn and Thornton (1992) found similar results, suggesting that more positive attitudes toward marriage among men lowered the probably of transitioning into a cohabiting relationship while more positive attitudes toward marriage among women increased the probability of marriage.

Although empirical studies linking early marital attitudes to union formation in young adulthood are limited, important recent theoretical work has suggested this link may be of particular importance. One such model, marital horizon theory (Carroll et al., 2007; 2009), proposes that the combination of the importance individuals place on marriage, their criteria for marriage readiness and their desired marital timing will have an important impact on decisions regarding other behaviors such as risk-taking and dating behavior. This direct influence on individual behavior in turn will affect eventual transitions into couple relationships, employment and education. From this viewpoint, marital attitudes during young adulthood take an important role in shaping and altering pathways through young adulthood.

Marital horizon theory is used in the current study as a theoretical framework to inform the interpretation and discussion of results. Carroll et al. (2007) suggest that each young adult has a unique marital horizon which will alter both their movement through young adulthood and their decisions to engage in risky behaviors prior to marriage. If marital horizons do help frame trajectories toward young adulthood, then marital attitudes and beliefs held earlier in life, before adolescents enter young adulthood, should influence union formation transitions that typically mark the end of young adulthood. To test this assumption, in this study I examine whether marital attitudes held during the last year of high school, when adolescents are about to enter young adulthood, influence couple transition in their 20’s and early 30’s. Previous research has found marked differences between marriage and cohabitation in both relationship quality and stability (Brown, Sanchez, Nock & Wright, 2006; Dush, Cohan, & Amato, 2003; Teachman, 2003) and finding distinctive characteristics of young adults who choose to either cohabit or marry may yield important information about which couples and individuals are at risk for negative relational and personal outcomes.

In addition to exploring this previously untapped area, I also attempt to expand the scope of measurement when considering how marital attitudes influence union transitions. Most studies that do focus on marital attitudes rely on one or two item measures that represent an eclectic variety of different constructs. Clarkberg et al. (1995), the only study to look at how attitudes predict both marriage and cohabitation behavior, utilized one item asking participants to rate how important marriage was for them on a three point scale, a limited assessment of one’s perception and beliefs about marriage. Some scholars (see Carroll et al., 2007; Hall, 2006) have suggested that marital attitudes and beliefs are multidimensional constructs that should be assessed from multiple angles in order to truly understand how individuals situate and prioritize marriage compared to other life goals. With this in mind, instead of focusing on only one type of marital attitude, this study uses marital attitudes measures which tap the constructs of both marital importance and marital timing, two core factors in marital horizons theory (Carroll et al., 2007).

Scope of Study

It is important to consider how attitudes toward marriage during young adulthood influence relational transitions and patterns later in the life course. Some previous research, with limited assessment of marital attitudes, has suggested that such a link might exist (Clarkberg et al., 1995; Sassler & Schoen, 1999) but these studies are now dated and may not reflect current cohorts of young adults. Additionally, making full use of multiple measures of marital attitudes will allow for a more complex understanding of how marital attitudes might influence couple formation transitions.

In this study, I examine the transitions to both marriage and cohabitation. I look at whether marital attitudes, in conjunction with other predictors previously found to be associated with couple formation transitions, provide any significant prediction of the probability of making one of these union transitions. Although it is expected that marital attitudes will influence the transition into marriage, previous work has suggested that the transition into cohabitation may be less planned and more likely to occur due to proximate personal and environmental factors (Sassler, 2004). For this reason I expect that marital attitudes during late adolescence will have little impact on the transition to cohabitation later in the life course.

In this study, I tested the following two hypotheses:

• Hypothesis 1:

  The likelihood of transitioning to cohabitation will not be predicted by marital attitudes held during late adolescence.

• Hypothesis 2:

  The likelihood of transitioning to marriage will be predicted by marital attitudes in adolescence, with more importance being place on marriage, a lower expected age of marriage and a higher expectation to marry increasing the likelihood of marriage.

Method

Participants

Data for this project were taken from the Youth Development Study (YDS). The YDS is a longitudinal data set comprised of an initial random sample of 1,010 ninth graders enrolled in the public school system in St. Paul, Minnesota. Census comparisons have shown that St. Paul has social and economic indicators similar to national averages (Mortimer, 2003). For the present study, waves 4 through 15 of the YDS were utilized. Wave 4 was collected when participants were seniors in high school (1992) with wave 15 being collected in 2004 when most participants were 30. In total, data from 982 participants who participated in wave four of the original study and had some relationship history data at subsequent waves were included in this study. Gender was roughly split within the sample with 47% of the sample being male while the sample was 74% white. Ninety two percent of the sample indicated they were born in the United States. Sixty eight percent of the sample indicated their parents were still married. Fifty nine percent of the sample indicated their parents had completed at least some college education. Table 1 provides an overview of sample demographics.

Table 1.

Demographic Characteristics of the Sample (N = 982)

Variable	Mean (SD)	Proportion
Whitea		.74
Male		.47
Two-Parent Biological Familyb		.68
Parent’s Incomec	5.78 (2.42)
U.S. Nationality		.92

^aproportion white versus “non-white”

^bproportion who living with two biological parents versus any other family structure

^cAssessed on a 13-point scale (1 = under $5,000; 13 = more than $100,000)

Procedure

Data for the first four waves (1988–1991) were obtained by questionnaires completed by students in school classrooms. Students who were not available during the administration of the questionnaire completed the questionnaire by mail. Extensive contact information was obtained for each student to ensure that students who dropped out of school or those who transferred to another school were able to be tracked for subsequent data collection waves. Data for subsequent waves (1992–2004) were obtained through completed questionnaires mailed to each respondent. Most years involved a short survey assessing demographic transitions regarding relationships and education. Longer follow-up data was obtained during waves 8, 12 and 15. The original participation rate for the study was 67%. The retention rate as of 2004 was 75%. Previous studies investigating if study participation was linked to differences in family structure, race, education outcomes, and socioeconomic background found no differences (Finch, Shanahan, Mortimer, & Ryu, 1991).

Measures

The YDS measured attitudes toward marriage in each of the first four waves of data collection. This study used several types of marital attitudes to get a more complete picture of any potential differences which might exist based on the type of marital attitudes assessed. Three types of marital attitudes were assessed at the initial wave used in this study (wave 4).

Marital importance

Marital importance was assessed with one item which asked participants to rate how important “Marriage, relationship with my husband or wife” would be when they were an adult. This item was assessed on a four point scale (1 = not at all important; 4 = extremely important).

Expectations to marry

Expectations to marry were assessed by one item asking participants “Do you expect you will get married someday?” Reponses ranged from 1 (No, I’m quite sure I won’t marry) to 5 (Yes, I am quite sure I will marry).

Expected age of marriage

Expected age of marriage was assessed by asking participants to indicate the oldest and youngest ages they expected to marry by. These two values were averaged to create an average expected age of marriage for each individual.

Union transitions

Each wave of data (5–15) contained life tables where participants indicated if they entered or exited marriage and/or cohabitation. Information was obtained on both marriage and cohabitation transitions separately, thus providing information on each individual regarding if they transitioned into cohabitation or marriage in a given year. These tables were used to construct variables for each year of data collection for both marriage and cohabitation. The first variable indicated that the participants either did not transition (0) or did transition (1) into a marital union during each year of data collection. A second variable indicated if the participants transitioned into a cohabiting union for each year of data collection. In both cases, once a transition occurred subsequent years were coded as missing in order to censor each individual at the year of first transition. These yearly intervals serve as discrete time intervals which were coded to determine if each participant in the risk set transitioned at each year of data collection. Couple formation transitions were measured by both first marriage and first cohabitation. Additional variables were created which aggregated yearly data into two variables which indicated if the participant had transitioned to marriage or cohabitation at any point during the study (0 = no; 1 = yes).

Controls

Several control variables were included in all analyses. Participants were asked to indicate their gender while family structure was also assessed and coded as either 0 (living with two parent biological family) or 1 (not living with two parent biological family). Race was assessed by asking each participant to indicate which racial category they identified with the most. This variable was re-coded into a two category variable indicated with either 1 (white) or 0 (non-white). Participants were also asked to indicate their nativity (coded as either born in U.S. (1) or not (0). Finally, family socio-economic status was assessed by asking parents in a separate survey at baseline to indicate their annual income which was assessed on a thirteen point scale with higher numbers indicated higher family income (1 = under $5,000; 13 = over $100,000).

Data Analysis Plan

First, hierarchical logistic regression models were used to predict any transition during data collection to both cohabitation and marriage. Two-step models were utilized with the first step including all control variables and the second step including wave four marital attitudes. This allowed for an examination of if marital attitudes at the end of high school predicted the likelihood of any transition during the study to both marriage and cohabitation above and beyond controls by examining differences in −2 log likelihood values.

Next, in order to assess the effect of marital attitudes on the probability of transitioning into either cohabitation or marriage across the waves of data collection, discrete time event history models were utilized. The risk set at each time point t for this sample constituted all participants who, up until time t, had not yet had a first union transition. Once a participant transitioned into either marriage or cohabitation, they were censored from the risk set. General models were first run to determine the population probabilities of transitioning into marriage or cohabitation at each age across the sample. After these general patterns were examined, controls and marital attitude covariates were added to the model to examine their effect on the probability of making union transitions. Participants’ attitudes were utilized as time invariant covariates. In total, three measures of marital attitudes were used as covariates to predict the transition to first marriage and cohabitation.

Missing data were assumed to be missing at random (MAR) and were handled using the full-info-max likelihood (FIML) function of Mplus. This method has been shown to be an effective way to handle incomplete data (Wothke, 2000). Overall, 133 individuals (14.5% of the original sample) had incomplete data. Preliminary analyses did not reveal any differences between participants with missing data and those with valid data on any of the variables of interest, including assessments of marital attitudes, demographics or union formation.

Results

Predicting any Transition to Cohabitation

To first predict any transition to cohabitation, control variables were entered into the first step of a logistic regression model and were significantly predictive of transitioning. Final models then included the addition of marital attitude covariates into the model. Table 2 summarizes final regression models predicting any transition to cohabitation across all waves of data collection. For models predicting any transition to cohabitation, being female (OR = 1.44, p < .05) or born in the United States (OR = 2.11, p < .05) increased the likelihood of transitioning to cohabitation. The marital attitude variables did not significantly predict any transition to cohabitation once control variable were accounted for (block χ² (3, 680) = 1.23, p = .75). None of the three marital attitudes significantly predicted if participants transitioned to cohabitation during the study.

Table 2.

Hierarchical logistic regression final model results and odds ratios predicting any transition to cohabitation or marriage

Variable	Cohabitation			Marriage
	B	SE	OR	B	SE	OR
Race	.199	.213	1.22	.424	.215	1.53*
Gender	.363	.163	1.44*	.126	.161	1.13
Family Structure	.003	.195	1.00	.121	.193	1.13
Parent’s Income	.052	.039	1.05	.039	.038	1.04
Nationality	.747	.366	2.11*	.258	.374	1.29
Importance of Marriage	.119	.140	1.13	.451	.143	1.57**
Expectation to Marry	.030	.189	1.03	−.169	.188	.844
Expected Timing	.021	.029	1.02	−.039	.029	.962

^*p < .05;

^**p < .01

Predicting any Transition to Marriage

Full model results predicting transitions to marriage showed different results. Full model results predicting any transition to marriage are also found in Table 2. The block of marital attitude variables significantly predicted any transition to marriage, even when adjusting for controls (block χ² (3, 680) = 12.69, p < .01). Specifically, a higher importance placed on marriage (OR = 1.57, p < .01) was significantly associated with an increased probability of transitioning to marriage during the study. Being white was associated with an increased likelihood of transitioning to marriage (OR = 1.53, p < .05). Expected age of marriage and expectation to marry were not significant predictors of any transition to marriage.

First Transition to Cohabitation

Baseline event history models with no covariates were next run to examine the hazard probabilities for the sample regarding first transitions to both cohabitation and marriage. Figure 1 shows the hazard probabilities for cohabitation and marriage for each year of data collection. The risk of transitioning into cohabitation made a sharp increase around 1994 when most participants would be 22–23. The proportion of the sample transitioning into cohabitation in 1993 was 1.2% compared to 4.1% in 1994 and 18.1% in 1995. The risk of cohabiting remained fairly constant over the rest of the data collection period before slowly decreasing in the final few waves of the study.

Figure 1.

Hazard probabilities for transitions to cohabitation and marriage

Covariates were next added to the discrete time event history model in order to determine how marital attitudes held as late adolescents might influence the hazard probability of transition into cohabitation at each wave of data collection. Control measures of gender, nativity, family structure, parent’s income and race were added along with all three measures of marital attitudes. Results are summarized in Table 3. None of the three measures of marital attitudes and none of the controls significantly influenced the probability of transitioning into a first cohabiting union at each wave of data collection.

Table 3.

Logistic regression results and odds ratios predicting hazard probabilities for transitioning to cohabitation or marriage

Variable	Cohabitation			Marriage
	β	SE	OR	β	SE	OR
Race	−.174	.172	.840	.169	.186	1.18
Gender	.071	.118	1.07	−.229	.123	.795
Family Structure	−.086	.144	.918	−.011	.153	.989
Parent’s Income	−.013	.026	.987	−.022	.027	.978
Nativity	.520	.308	1.68	−.046	.327	.955
Importance of Marriage	−.005	.106	.995	.320	.118	1.38**
Expectation to Marry	−.169	.127	.845	−.138	.134	.871
Expected Timing	−.002	.022	.998	−.063	.023	.939**

^*p < .05;

^**p < .01

First Transition to Marriage

As shown in Figure 1, the hazard probability of first marriage gradually increases across the study waves, peaking in 2003 where 15% of the at-risk population made a first transition into marriage. Table 3 summarizes the results predicting hazard probabilities once covariates were added. After controls were added to the model, two of the three marital attitudes had a significant impact on the event history model. Expected age of marriage had a significant and negative relationship with the probability of transitioning to marriage at each wave (OR = .939, p < .01). Each year later that late adolescents expected to get married resulted in a roughly 6% decrease in their likelihood of transitioning into marriage at each time point. The general importance of marriage item was also significant (OR = 1.38, p < .01). For each unit increase on the general importance of marriage scale for late adolescents, there was a corresponding 38% increase in the probability that they transitioned to marriage at each time point. None of the control variables had a significant impact on the model.

Discussion

Patterns of cohabitation and marriage in the study seemed to indicate similar patterns as those found in previous studies. The transition to cohabitation was likely to happen earlier than marriage, with a peak in cohabitation found in the early 20’s. As participants entered their late 20’s and early 30’s there was a general decrease in the likelihood of a transition into cohabitation for the first time. As seen from the marriage hazard plot, the probability of transitioning to marriage steadily increased as participants enter their late 20’s and early 30’s. As more participants began to transition into marriage, the likelihood of transitioning into cohabitation decreased across the sample, suggesting that as individuals approach their early 30’s, marriage becomes the transition of choice for many individuals.

When predicting actual union transition behavior, controls used as covariates had surprisingly little impact on the union behavior of young adults. Although we found weak associations between race and transitions to marriage consistent with other studies (Crowder & Tolnay, 2000) along with associations between gender, nativity and the transition to cohabitation, demographic controls were not major predictors of transitions to both marriage and cohabitation. Although possibly a byproduct of the Midwestern sample utilized with less demographic variability, it is also possible that background factors are less influential than current contextual factors in predicting actual union transitions for most young adults.

Results from this study suggest that marital attitudes during late adolescence have little impact on the transition into cohabitation later in the life course, confirming hypothesis one. None of the three marital attitudes investigated had a significant impact on the survival probability for cohabitation and or the probability of transitioning to cohabitation at any point during the study. This would suggest that the decision regarding when and if to cohabit might be made independently of marital plans and expectations during late adolescence. Because the transition to cohabitation can be sudden or unplanned for many couples (Sassler, 2004), it is not surprising that there would be little association between adolescent marital attitudes and subsequent transitions into cohabitation during young adulthood.

Marital attitudes from late adolescence did have an impact on the probability of transitioning to marriage. Specifically, late adolescents who expected to marry earlier and who placed a higher importance on marriage were more likely to transition into marriage at each time point during the study. This provided partial support for hypothesis two. The fact that the importance place on marriage impacts the probability of transitioning to marriage has been found in previous research (Clarkberg et al., 1995). Adolescents who place a high importance on marriage are likely to be more marriage-centered in their relationships and be actively preparing and expecting to marry. That adolescents’ expected age of marriage had a significant effect on the probability of transitioning into marriage is an important new finding for the study of young adulthood. When late adolescents expect to marry young, they are more likely to transition into marriage and more likely to make that transition earlier than peers who do not expect to marry young.

Both of these findings lend support to marital horizon theory. How adolescents viewed marriage prior to high school graduation had a significant impact on their transition into marriage later in young adulthood. This suggests that young adults may be on some level altering relational goals, patterns and practices based on their global views and expectations regarding marriage. Marriage has largely been left out of the discussion and research surrounding young adulthood as most young adults will not transition to marriage until later into their twenties (Carroll et al., 2007). Results from this study suggest that although marriage is not a common event during young adulthood, marital attitudes may have important ramifications for behaviors, decisions, and trajectories through young adulthood.

Limitations and Future Directions

Several limitations should be considered while interpreting the results of this study. Although significant effects were found across most marital attitudes measured; most of these effects were small. Additionally, data limitations did not allow for the tracking of marital attitude change through young adulthood as marital attitudes were only assessed during the high school years. It is unknown if marital attitudes held during young adulthood are significantly different than those held during late adolescence and if those attitudes held during young adulthood would be associated differently with union transition behavior. It should also be noted that although demographic variables such as family structure and race were not predictive of union formation, this may be due to the limited assessment of these factors available in the dataset. Future studies should seek to understand if more specific measurement might find union formation differences across family structures and racial groups. Additionally, family income in high school may not be a good predictor later union formation during young adulthood. More proximate assessments of economic stability during the twenty would likely have more influence on eventual marital and cohabiting transitions.

The present study also suggests avenues for future scholarship. While the current study found little link between marital attitudes and cohabiting behavior, attitudes toward cohabitation during late adolescence may be more predictive of cohabitation and may predict marital formation behavior. Attitudes toward cohabitation remain an understudied area of attitudinal research (blinded Author citation) and should be a focus of future research. Another area of future work lies in understanding the links between marital attitudes, education and union formation. Previous research has shown that adolescents who aspire to post-secondary education tend to place marriage farther out in the life course (Willoughby, 2010). Although it is likely that educational attainment will alter union formation trends, scholars should continue to investigate the links between marital attitudes, educational trajectories and union formation to understand how young adults prioritize relational and educational goals.

The results of this study establish a clear link between marital attitudes in adolescence and the transition to marriage later in the life course. With so many varied pathways developing through young adulthood, models which take a life course perspective and find predictive power in values or attitudes during adolescence will continue to be important in the future. Marriage and cohabitation have both been the focus of research across multiple disciplines and both are a staple of relational development in the United States. Marital attitudes help shape and determine those transitions and should continue to be an important priority for couple and family formation scholars.