Changes in expectations to marry and to divorce across the transition to adulthood

Abstract

Marriage and divorce expectations predict family life and personal outcomes. Understanding how expectations are associated with varying characteristics over emerging adulthood (ages 18-28) will inform understanding of emerging adult development. Panel Study of Income Dynamics Transition into Adulthood 2005-2015 data were used in hybrid-effects ordinal logistic regression to parse interindividual and intraindividual variation associated with relationship experiences, socioeconomic and contextual characteristics, and mental and emotional wellbeing. Partnerships were associated with optimistic expectations: both dating and cohabiting predicted greater marriage expectations and lower divorce expectations within individuals. Between individuals, greater time in full-time employment predicted more positive marital expectations, greater responsibility was associated with lower marital expectations, stronger religious identity predicted higher marital expectations and lower divorce expectations, having been arrested predicted greater divorce expectations, greater wellbeing predicted greater marriage expectations, and older age predicted lower marriage expectations. Both between and within individuals, greater worry predicted lower marriage expectations.

Most American youth expect to marry someday (Anderson, 2016), but likely recognize that marriage is not always happy in the ever after. Although declining (Payne, 2018), the relatively high divorce rate in the recent past, which suggested that nearly half of marriages started in the 1980s would end in divorce (Cherlin, 2010), means that most emerging adults will have witnessed divorce in their own or another family and may be concerned about the longevity of hypothetical marriages (Boyer-Pennington, Pennington, & Spink, 2001). Marriage and divorce expectations have been studied chiefly cross-sectionally, though both likely vary over time (Arnett, 2000; Willoughby, 2014). Understanding how these expectations fluctuate with experiences and circumstances is important to understanding how expectations are formed and enacted as well as predicting behavior in emerging adulthood (Arocho & Kamp Dush, 2016, 2017; Willoughby, 2014; Willoughby & Dworkin, 2009). This study used the Panel Study of Income Dynamics Transition into Adulthood Supplement (PSID Transition into Adulthood Study 2011 User Guide, n.d.) to study how characteristics and experiences predicted interindividual differences and intraindividual variation in expectations for marriage and divorce in emerging adulthood (18 to 28 years old in this study). The analyses build on the work conducted by Arocho and Purtell (2018) by taking a longitudinal view on both interpersonal differences and intrapersonal variation in predicting characteristics and in the expectations themselves.

Expectations of Marriage and Divorce

Emerging adulthood, a period of identity formation and exploration (Arnett, 2000), often includes exploring the role of marriage in one’s life (Arnett, 2015; Willoughby & James, 2017). Although most Americans do not marry until almost the fourth decade of their lives (Manning, Brown, & Payne, 2014), individuals may think about marriage and divorce even in adolescence (Arnett, 2015; Arocho & Kamp Dush, 2017; Carroll et al., 2007). According to marital paradigms framework, marriage rhetoric is so pervasive in American culture that youth cannot help but develop attitudes towards marriage (Willoughby, Hall, & Luczak, 2015).

Different relationship expectations are not developed alone and may be associated with similar predictors, though possibly in opposite directions. Like marriage itself (Finkel, Hui, Carswell, & Larson, 2014), marital expectations are multi-faceted. Marital paradigms framework (Willoughby, Hall, et al., 2015) suggests there are six categories of marital opinions that individuals may hold. One, marital salience, a belief about getting married, details feelings towards marriage in one’s life. Marital permanence, beliefs about how well marriage can and should last, is about being married. These two beliefs are likely interrelated and may be individually and jointly associated with subsequent behaviors (Arocho, 2019).

Expectations for experiences and achievement are important both as indicators of present experiences and predictors of personal development and family life behaviors. Expectations for both marriage and divorce are cross-sectionally associated with a number of personal characteristics, including relationship experiences (Arocho & Purtell, 2018), witnessing parents’ marriage experiences (Boyer-Pennington et al., 2001; Dennison & Koerner, 2008), and socioeconomic factors (Arocho & Kamp Dush, 2018; Arocho & Purtell, 2018). Expectations for marriage, especially younger or stronger expectations, are also associated with a variety of outcomes, ranging from actually getting married to behaving more like a “married person,” such as by engaging in less drug use or fewer delinquent activities (Arocho & Kamp Dush, 2016; Carroll et al., 2007; Willoughby & Dworkin, 2009). Greater expectations of divorce are associated with delayed marriage and increased cohabitation (Miller, Sassler, & Kusi-Appouh, 2011; Arocho, 2019; Perelli-Harris, Berrington, Sánchez Gassen, Galezewska, & Holland, 2017).

Development of Expectations over Time and Experience

Although adolescent relationships are important (Collins, 2003), emerging adulthood is when relationship expectations may be more readily put into practice and new challenges emerge (Shulman & Connolly, 2013). Given the myriad changes in emerging adults’ lives, expectations for romance are likely dynamic. Indeed, studies of marital expectations over time have found that romantic experience is associated with more positive attitudes towards marriage (Barr, Simons, & Simons, 2015; Willoughby, Medaris, James, & Bartholomew, 2015). However, many studies are limited to single measurements of expectations, especially divorce (Boyer-Pennington et al., 2001; Dennison & Koerner, 2008). The aim of the current study was to use rigorous within- and between-person comparisons to elucidate associations between changing experiences and characteristics of emerging adults and their expectations for marriage and divorce. By isolating the variation in expectations associated with circumstances, this study provides evidence for the role personal context may play in any specific observation, helping researchers limited to cross-sectional data better understand the values they observe.

In the past, transitioning to adulthood often followed a pattern of education, employment, marriage, and parenthood (Settersten Jr., 2012). With increased freedom and fewer institutionalized bounds to becoming an adult, however, later generations have experienced more opportunities for choice and fewer prescribed goals (Settersten & Ray, 2010). As such, personal measures of achievement and “adulthood” may both be formed from and proceed other traditionally “adult” experiences (Settersten & Ray, 2010; Silva, 2013). For many emerging adults, achieving adulthood is needed before marriage, rather than achieved through it (Kefalas, Furstenberg, Carr, & Napolitano, 2011). As such, several personal characteristics and experiences in emerging adulthood could predict variations in expectations for marriage and marriage success. In the expectancy-value theory of achievement motivation (Wigfield & Eccles, 2000), past experiences and characteristics feed into task value and expectations of success which then together predict behavior towards a goal. Task value is the subjective value of the goal, whereas expectations for success measure confidence in accomplishment. Applying this model to marriage as a goal, with expectations for marriage as task value in the absence of a direct measure (Arocho & Purtell, 2018), expectations for divorce becomes (the inverse of) expectations for success. According to the model, past and present personal characteristics and experiences predict both of these expectations, and these expectations then in conjunction predict behavior (Arocho, 2019).

One of the most proximal predictors of expectations for success is an individual’s self-schemata, partially constructed from self-concept of abilities and perceptions of task demands (Wigfield & Eccles, 2000). In this study, several characteristics that may predict feelings of marriage readiness and maturity fit within this umbrella. These various experiences and characteristics may also be in flux; understanding their time-varying associations with expectations for romance provides insight into development. Thus, variation in expectations was predicted by changing characteristics in four groups: relationship experiences, socioeconomic characteristics, context and maturity, and mental and emotional wellbeing.

Relationship experiences.

A proximal predictors of attitudes towards marriage and divorce is experience with relationships, including those observed. Indeed, intergenerational similarities in marriage and divorce behavior exist (Amato & DeBoer, 2001; Amato & Patterson, 2017), as well as associations between parents’ marital behaviors and offspring expectations (Arocho & Kamp Dush, 2017; Dennison & Koerner, 2008). However, there is more to the story, as children of divorce are often still willing to marry (Amato, 1988) and having had a parent divorce does not predict absolute expectations for marriage failure (Arocho & Purtell, 2018). Experiencing relationships, including dating and cohabiting, may help shape attitudes towards marriage and marital success, often positively (Arocho & Purtell, 2018; Barr et al., 2015; Crissey, 2005; Manning, Longmore, & Giordano, 2007; Willoughby, Medaris, et al., 2015). Youth’s experiences as parents may also be associated with marital expectations. Having a child may make one more confident in their expectation to marry, but perhaps less sure of their success, as potential partners may not be as willing to enter a relationship with a parent (Goldscheider, Kaufman, & Sassler, 2009). Taken together, relationship experiences such as stable parental marriage, dating, cohabiting, or parenting likely predict greater expectations to marry; but witnessing divorce or having had a child may predict increased expectations for divorce.

Socioeconomic characteristics.

Socioeconomic experiences may predict attitudes towards marriage and divorce because they represent potential economic success. Indeed, economic stability and an ability to care for oneself or a future family have been cited as prerequisite for marriage (Carroll et al., 2009; Gibson-Davis, Gassman-Pines, & Lehrman, 2018; Smock, Manning, & Porter, 2005). Educational attainment may represent economic potential (Torpey, 2018) and many believe education should be completed before marriage is undertaken (Carroll et al., 2009; Kefalas et al., 2011). Economic success or potential for success, such as having achieved higher education, embarked on a career path, or achieved financial wellbeing, should predict both increased expectations to marry and decreased expectations for divorce.

Context and maturity.

Similar to economic achievement and stability, maturity, evidenced by behavior and contextual factors, may also play a role in “adulthood” (Silva, 2013) and readiness for marriage (Carroll et al., 2009). For instance, living away from parents is, for some, necessary for marriage readiness (Carroll et al., 2009). Delinquent behavior may indicate proclivity towards non-traditional attitudes, and although not predictive of marital expectations (Arocho & Kamp Dush, 2016; Manning et al., 2007) may predict expectations for divorce. Religiosity can change considerably over emerging adulthood (Chan, Tsai, & Fuligni, 2015) but is associated with relationship expectations (Manning et al., 2007) and behavior (Eggebeen & Dew, 2009). Finally, peer context, especially peers’ experiences with relationships and adult roles, may also play a role in expectations (Arnett, 2015; Manning, Cohen, & Smock, 2010). Thus, markers of adulthood, such as leaving the parental home or having friends who are married, as well as characteristics such as higher religious importance or normative behavior, should predict greater expectations of marriage and lower expectations of divorce.

Mental and emotional wellbeing.

Finally, mental and emotional wellbeing likely play a role in quantification of marriage and divorce risk. When traditional markers of achievement are hard to attain, emerging adults may turn to narratives of overcoming hardships, including mental and emotional distress, as a form of achievement (Silva, 2013). In addition, mental health interacts with relationships throughout the life course. In adolescence and emerging adulthood, poorer mental health predicts less relationship progression (Sandberg-Thoma & Kamp Dush, 2014), but entering relationships is associated with improved mental health (Mernitz & Kamp Dush, 2016). Mental and emotional health is also associated with expectations of divorce (Arocho & Purtell, 2018), but this has only been shown cross-sectionally. It is likely that greater emotional wellbeing and lower mental and emotional distress will predict more optimistic expectations of both marriage and divorce.

Hypotheses

In summary, expectations for both marriage and divorce will likely be predicted by interpersonal factors and intrapersonal variation within such characteristics, including relationship experiences, socioeconomic characteristics, contextual and maturity markers, and mental and emotional wellbeing. As these characteristics improve (i.e., greater socioeconomic status is achieved, emotional health improves), marital expectations should increase and expectations for divorce decrease within and between individuals.

Data and Method

Data

Data were from the 2005 to 2015 biennial waves of the Panel Study of Income Dynamics (PSID) Transition into Adulthood Supplement (TAS) subsample (PSID Transition into Adulthood Study 2011 User Guide, n.d.). The original data are publicly available from the PSID website, psidonline.isr.umich.edu, and the prepared analytic file is openly available in the PSID Public Data Extract Repository at http://doi.org/10.3886/E111182V1. The PSID was a U.S.-nationally-representative sample of families in 1968 and the study has continued with descendants of the original sample. Starting in 2005, the TAS sample was drawn from the respondents in the Child Development Study (CDS) portion of the PSID in the 1990s and 2000s. Respondents entered the TAS when they either finished high school (by graduation or dropping out) or reached age 18 and were interviewed in the TAS through age 28. Observations were limited to waves when the youth were unmarried for three reasons: 1) those with prior marriage experience may differ in their marital and divorce expectations after they have experience (Edin & Kefalas, 2005), 2) youth who marry in their teens or early twenties are a select and unusual group (Uecker & Stokes, 2008), and 3) the survey was designed such that those who were already married do not receive all the same expectations questions, limiting comparability. Measuring within-individual change required at least two interviews from each respondent. Between 2005 and 2015, 2893 youth were interviewed at least once. Removing waves in which individuals were married (or previously married) dropped 935 observations, plus 5 observations with missing marital data. Further limiting the sample to those with at least 2 interviews left 2195 individuals and 7215 person-waves. Imputation and analyses were weighted using each individual’s initial TAS weight as suggested by the PSID user guides (PSID Transition into Adulthood Study 2011 User Guide, n.d.). The prepared analytic dataset is available online at https://www.openicpsr.org/openicpsr/search/psid/studies.

Measures

Dependent variables

The two dependent variables, expectations to marry and to divorce, were asked as, “What do you think are the chances that you will get [married/divorced]?” and had five response categories: no chance, some chance, about 50-50, pretty likely, it will happen; lower values indicated less certainty of the event. Few responses indicated certainty of divorce (0.55%) or remaining unmarried (2.73%) and were collapsed with the nearest category, resulting in four categories for each expectation: expectations of divorce became no chance, some chance, 50-50 chance, pretty likely/it will happen; and expectations of marriage became no/some chance, 50-50 chance, pretty likely, it will happen. In the case that respondents reported no expectation to marry, they were not asked to report expectations to divorce; those responses (197 observations) were assigned a value of “no chance” for divorce expectations as well.

Relationship experiences.

Relationship status was measured as cohabiting if respondents answered yes to the question, “are you living with a partner in a marriage-like relationship?” and as dating if they answered yes to the question, “Are you currently involved in a romantic relationship?” and were not cohabiting. Respondents were coded as single if they answered no to both. Although the questions did not specify same- or different-gender relationships, the gender of the partner was not reported, so same- and different-gender relationship cannot be differentiated. Also included was a time-varying measure of the marital status of the adult identified to be the respondents’ primary caregiver in 1997, which is the first year the Child Development Study was conducted. Caregivers were coded as never married, in first marriage, ended first marriage and not remarried, and entered second or higher-order marriage. These data were drawn from the retrospective marital history collected on all adults ages 15-44 in the PSID (Institute for Social Research University of Michigan, 2017). Finally, at each wave, respondents were coded as a parent if they had at least one child (biological, step, or adopted) or were pregnant or reported their partner was pregnant.

Socioeconomic characteristics.

At each interview, respondents reported currently being enrolled in college (yes or no) and completed education (coded as less than high school, high school or equivalent, and 2-year college degree or more). Respondents also reported average hours per week at up to five jobs in the past year. The average of these hours was used to classify respondents as having had no, part time (1-34 hours), or full time (35+ hours) work in the past year. Respondents also indicated whether or not their current job was “part of your long-term career plan” (respondents without a current job were coded with those who responded “no”). Finally, respondents reported their level of personal savings (asked as “do you yourself have a checking or savings account” and then either reporting a dollar amount or indicating if the amount was worth a certain amount or more), coded as no savings, up to $1,000 saved, and $1,000 or more saved; as well as whether or not they had student loans, either through a direct question (“do you yourself currently have any student loans?”) in 2009 through 2015, or after reporting debts, a question asking “of the amount just reported, is any of that for student loans?” in 2005 and 2007.

Context and maturity.

Respondents reported their regular residence in the previous fall/winter and last summer and were coded as living with their parents all, part, or none of the year. Respondents were asked “how much responsibility do you currently take for…” earning their own living, paying their rent/mortgage, paying bills, and handling money. Responses were on a scale from 1 to 5, with 5 indicating highest responsibility (“I am completely responsible for this all of the time”); responses were averaged (α = .76-.84 each year). This scale was originally developed for the Michigan Study of Adolescent and Adult Life Transitions (Survey Research Center Institute for Social Research, 2008). Religious importance was measured by responses to the question, “how important is religion to you?” Because responses were skewed towards high religious importance (in the raw data, depending on the year, between two-third and three-fourths of respondents reported “very” or “somewhat” high importance), and additionally those who indicated no religious preference were not asked this question, respondents were collapsed into three categories: not religious, somewhat or less important, and very important. Respondents were asked each interview if they had been arrested (coded ever or never). Finally, respondents were asked “What percent of your closest friends not including your romantic partner are married or living with a romantic partner?” with responses ranging from 1 (none) to 5 (all).

Mental and emotional wellbeing.

Respondents reported on how often they worried about money or future jobs or felt discouraged about the future, on a scale from 1 (never) to 7 (daily). These three items were averaged to construct a scale of worry (α = .68-.76 each year); like the responsibility scale, this was developed for the Michigan Study of Adolescent and Adult Life Transitions (Survey Research Center Institute for Social Research, 2008). Respondents also completed a six-item measure of nervousness, hopelessness, restlessness, effort, sadness, and worthlessness (Kessler et al., 2003) scored 1-5, summed (α=.72-.79) and coded so that higher scores indicated greater non-specific psychological distress. Finally, respondents completed 14 items dealing with psychological wellbeing (such as, “In the last month, how often did you feel that you have experiences that challenged you to grow or become a better person?”), social wellbeing (“…how often did you feel that you belonged to a community like a social group, your school, or your neighborhood?”), and emotional wellbeing (“…how often did you feel happy?”), and responses (1-6) were averaged to produce the wellbeing scale (α = .87-.88) as used in the Midlife in the United States study (Survey Research Center Institute for Social Research, 2008); higher scores indicated greater wellbeing.

Control variables.

Gender, race, ethnicity, age, and socioeconomic background are associated with expectations of both marriage and divorce (Arocho & Purtell, 2018; Manning et al., 2007). Models controlled for respondent’s gender (1=male), first reported racial identity (White, Black, and Other/Multiracial) and Hispanic ethnicity as well as maternal education (less than high school, high school, some college, and 4-year college degree or more) as a proxy for socioeconomic background. Age at interview was the measure of time (Allison, 2009).

Multicollinearity.

Multicollinearity of variables is a concern with many variables in each model. The variance inflation factor (VIF) for all the variables in the models (O’brien, 2007) was computed using the -collin- command in Stata15. The average VIF was 1.30, and the highest was 1.72, produced by age. As values of 4, 10, and 20 have been suggested as cause for concern (O’brien, 2007), the relatively low VIF in these data indicated the models were safe to run.

Missing data.

In total, about 30% of observations had some missing values, but no variable was missing more than 11% of observations, and most were missing much less (Table 1, right column). Assuming the data to be Missing at Random (MAR), that is, the missingness of values was conditional on other observed variables, missing data in both covariates and the expectations outcomes was imputed using Multiple Imputation using Chained Equations in the long-form data over 10 datasets using the MI suite in Stata15 (StataCorp, 2017; White, Royston, & Wood, 2010). Simulations have suggested that fixed-effect panel regression is robust to long-form imputation with similar results to data imputed in wide-form (Young & Johnson, 2015).

Table 1.

Nonimputed, Weighted Descriptive Statistics.

Individuals = 2195
	Mean/%	Standard Error	Minimum	Maximum	Missing (%)
Male	48
Age at First Interview	19.5	0.03	17	25
Age at Last Interview	24.41	0.06	19	28
Race					0.1
White	65
Black	15
Other/Multi	20
Hispanic	16				0.05
Mother Education					2
Less HS	25
HS or GED	27
Some Col	24
College+	24
Number of Observations	3.29		2	5
Observations = 7215
Expectation of Divorce					0.08
No Chance	39
Some Chance	47
About 50-50	13
Pretty Likely/It Will Happen	2
Expectation of Marriage					0.03
No Chance/Some Chance	11
About 50-50	18
Pretty Likely	36
It Will Happen	34
Relationship Experiences
Relationship Status					0.04
Single	51
Dating	32
Cohabiting	16
Have a Child	14				0.04
Parental Marital Status					11
Never Married	7
In First Marriage	55
Ended First Marriage	18
Second+ Marriage	20
Socioeconomic Characteristics
Education Completed					3
< High School	7
High School or GED	80
College+	14
In College	44				0.2
Employment					11
Unemployed	8
Part Time	51
Full Time	41
Job in Career Plan	29				2
Savings					2
No Savings	19
Up to $1K	41
$1K+	40
Student Loans	34				0.1
Context and Maturity
Responsible	3.97	0.02	1	5	2
Residence with Parents					1
Parents All Year	41
Parents Part Year	20
Not with Parents	39
Religious Importance					0.2
Not Religious	22
Somewhat Important	47
Very Important	31
Arrested	20				0.2
Friends Married	1.99	0.01	1	5	0.3
Mental and Emotional Wellbeing
Worry	3.56	0.02	1	7	0.04
Psychological Distress	5.1	0.06	0	24	0.1
Wellbeing	13.56	0.04	3	18	1

Analysis

Hybrid panel regression (“between-within”) models (Allison, 2009) were used to estimate two components of time-varying independent variables: a personal mean effect of differences between individuals (interindividual); and a deviation at each observation from each individual’s personal mean, modeling within-person change (intraindividual). Time-invariant variables were also included. To account for interdependence between observations of the same individual, the clustered standard errors (StataCorp, 2017) option was specified. Because of the categorical distribution of the outcome variables, ordinal logistic regression was appropriate. These models are interpreted as either increasing or decreasing odds of selecting a higher category of response, and accordingly decreasing or increasing the odds of a lower response. Each model was tested for the parallel lines (proportional odds) assumption (Buis, n.d.; Williams, 2006) inherit to ordinal logistic regression. Both models (marriage and divorce) violated the assumption, so each model was then specified as a partial probability generalized ordinal logistic regression model (Williams, 2006). Out of all the variables in each model, only 12% and 10% of the variables in the marriage and divorce model, respectively, violated the assumption. Due to the increased complexity in the generalized ordinal model and the overall similarity between the results in the generalized and traditional models, only the traditional model results are reported below for parsimony. Results are reported in odds ratios (exponentiated coefficients) to aid in interpretation; values above 1 indicate increased odds of reporting a higher response category and values below 1 lower odds of reporting a higher category (thus greater odds of reporting a lower category). Due to the large number of comparisons, significance is reported starting at p < 0.01 (Williams, 2016).

Results

Descriptive Statistics

Table 1 shows the descriptive statistics of the sample both by individuals (for time-invariant variables) and pooled observations (for time-varying variables). Respondents were included for an average of 3.29 interviews, ranging from 2 to 5, and were on average 19 years old at their first interview and 24 at their last. Looking at the 7215 overall observations, expectations for marriage and divorce were most often optimistic, with the majority of marital expectations observations falling into higher categories (36% in “pretty likely” and 34% in “it will happen) and divorce expectations generally falling into the lower categories (47% in “some chance” and 39% in “no chance”).

Regression Results

Table 2 displays the results of each model, one for marriage expectations and the other for divorce expectations. For both models, one column displays the association between the independent variable and the differences in the outcome when looking between individuals, and the second column shows the associations when comparing observations within the same individual over time. The first line of the table reports the associations between the expectations themselves; all associations were negative and statistically significant. For expectations of marriage, a one-unit increase in expectations of divorce was associated with 34% lower odds of selecting higher marital expectations when comparing individuals to each other, and 19% lower odds when comparing respondents to themselves over time. For expectations of divorce between individuals, greater expectations of marriage predicted 26% lower odds of selecting higher divorce expectations, and within individuals greater expectations of marriage were associated with 11% lower odds of high expectations for divorce.

Table 2.

Results of Hybrid Ordinal Logistic Regression Model Predicting Expectations for Marriage or for Divorce, individuals = 2195, observations = 7215.

	Marriage Expectations		Divorce Expectations
	Between	Within	Between	Within
Expectation of Other Event†	0.66***	0.81**	0.74***	0.89*
Relationship Experiences
Relationship Status (ref: Single)
Dating	1.93***	1.22*	0.76	0.76**
Cohabiting	4.49***	2.21***	0.72	0.56***
Has a Child	0.93	1.46	1.03	0.65
Parental Marital Status (ref: In First Marriage)
Never Married	0.78	0.91	0.98	1.07
Ended First Marriage	0.80	1.04	1.38	0.88
In Second+ Marriage	1.21	1.50	1.24	0.90
Socioeconomic Characteristics
In College Now	1.07	1.05	1.22	1.07
Education Completed (ref: High school or GED)
< High School	0.77	1.06	0.97	1.28
College+	1.03	1.27	1.45	0.94
Employment (ref: Full-time)
Not Working	0.44**	0.84	1.28	1.07
Part-time	0.64*	0.85	1.21	0.95
Job in Career Plan	1.01	1.03	1.54	1.09
Savings (ref: between $1 and $1k)
No Savings	0.64	0.80	1.33	0.88
Savings >$1K	1.13	0.98	1.22	0.95
Student Loans	1.17	0.94	0.85	1.05
Context and Maturity
Responsible for Self	0.77***	0.97	1.06	1.00
Residence (ref: Not with Parents)
Parents All Year	1.22	0.97	1.24	1.06
Parents Part Year	1.01	0.86	0.95	1.04
Religious Importance (ref: Somewhat Important)
Not Religious	0.57***	0.92	0.97	1.18
Very Important	1.46*	1.13	0.42***	0.95
Ever Arrested	1.02	0.82	1.45*	1.13
Friends Married	0.92	1.05	1.05	0.96
Mental and Emotional Wellbeing
Worry	0.84***	0.92*	1.11	1.03
Psychological Distress	1.01	0.98	1.04	1.03
Wellbeing	1.19***	1.05	0.96	0.99
Control Variables
Race (reference: White)
Black	0.78		0.74
Other Race	0.93		0.95
Hispanic	0.86		0.90
Maternal Education (reference: High School)
< High School	1.10		1.04
Some College	1.06		0.87
College+	1.06		0.91
Male	0.96		1.24
Age	0.94**		1.00

Note: *p < 0.01, **p < 0.001, *** p < 0.0001. †“Other Event” refers to expectations for divorce when expectations for marriage is outcome and vice-versa.

Relationship experience.

Looking at relationship experience, only respondents’ own relationship status was associated with variation in either expectations for marriage or divorce. The pattern of findings supported prior research on marital expectations: having a relationship, both dating and cohabiting, was associated with greater expectations of marriage both between (93% increased odds for dating and 349% increased odds for cohabiting) and within (22% for dating and 121% for cohabiting) individuals. Expectations for divorce, however, were only predicted by variations in relationship status within individuals. In this case, being in a dating relationship was associated with 24% lower odds of greater expectations to divorce, and cohabiting was associated with 44% lower odds. Having a child was not significantly associated with either expectation, nor was parents’ marital experience.

Socioeconomic characteristics.

Overall, only employment was significantly associated with expectations for marriage in these models, and no socioeconomic characteristics were significantly associated with expectations for divorce. For expectations for marriage, lower expectations for marriage were predicted by being unemployed (56% lower odds of higher responses) or employed only part-time (36% lower odds), and only when looking between individuals. Within observations of the same individual, variations in employment status did not predict variations in expectations for marriage. Variations in education, both completed and ongoing; having started on a career trajectory; level of savings; and carrying student loans were all not associated with either type of expectation.

Context and maturity.

Turning now to characteristics of the respondents’ context, again, only a few key variables were associated with variation in expectations. For marriage expectations, taking greater responsibility for oneself predicted lower expectations to marry when looking between individuals; one step greater responsibility was associated with 23% lower odds of high expectations. For expectations to divorce, having been arrested, at least when looking between individuals, was associated with 45% higher odds of reporting greater expectations to divorce. Only religious importance predicted both types of expectations. For marriage expectations, those reporting low or no religious importance had 43% lower odds of higher marital expectations, and those reporting religion as very important had 46% higher odds of high expectations. Conversely, a very important religious identity was associated with 58% lower odds of high expectations to divorce. In both cases, these significant associations were only apparent when examining the variation between respondents, not variation of religious importance within observations of the same individual.

Mental and emotional wellbeing.

Variables from the category of mental and emotional wellbeing predicted only expectations to marry. Greater worry predicted lower expectations, and greater wellbeing predicted higher, as might be expected. Specifically, a one-unit increase in worry was associated with 16% lower odds of higher expectations for marriage when looking between individuals, and even within the same individual, one-unit greater worry was associated with 8% lower odds of reporting higher expectations for marriage. Greater wellbeing was only significantly associated with expectations when comparing between individuals, such that one-unit greater wellbeing was associated with 19% higher odds of selecting higher expectations for marriage.

The only control variable associated with expectations was age, in that each year of age was associated with 6% lower odds of higher expectations of marriage.

Discussion

Understanding nuanced predictors of emerging adults’ expectations for marriage and marital outcomes can aid theoretical understanding of expectation development and enactment as well as aid in understanding how these expectations predict behavior into adulthood. Drawing on the expectancy-value theory of achievement motivation (Wigfield & Eccles, 2000), four categories of variables were hypothesized to predict expectations, and a rigorous method was used to quantified the associations both between individuals and within observations of the same respondents. Overall, there were few significant associations uncovered within the stringent parameters of the model. Those that were evident, however, often supported the hypotheses. A close look at these predictors, especially looking at comparisons of the same person at different time points, suggests that variations in expectations are predicted by a few key variables. This may be especially helpful for researchers limited to one or few observational periods who want to understand what expectations mean for a respondents’ context and other experiences.

As expected, both dating and cohabiting predicted optimistic expectations for both marriage and divorce. For marriage, time spent in relationships was associated with between-person differences, and when looking within individual’s multiple observations both expectations were more optimistic (greater expectations of marriage and lower expectations of divorce) when respondents were in relationships than when they were single. This corroborates past work which has suggested that being in relationships is associated with stronger expectations for marriage (Crissey, 2005) and more positive attitudes towards marriage (Barr et al., 2015) and extends it to expectations of divorce (Arocho & Purtell, 2018), suggesting that even as individuals shift in relationship status their feelings towards divorce may vary accordingly. These optimistic expectations may be an extension of seeing one’s current relationship through rose-colored glasses, as other work has suggested with married individuals considering risks of divorce (Baker & Emery, 1993). Emerging adults who allow the inertia of a relationship to sweep them into marriage, often through cohabitation, may actually be at an increased risk of divorce (Stanley, Rhoades, & Markman, 2006). On the other hand, however, it could be that respondents feel more confident in their ability to navigate the stages and challenges of relationships (Shulman & Connolly, 2013), and avoid the pitfalls, once they have some experience in hand. In this case, increased feelings of relationship competence would likely help respondents form more satisfying relationships and reach their goals of family life, whatever those may be. For emerging adults struggling with feelings of low relationship competence, psychotherapy may be an option (Atzil-Slonim, Reshef, Berman, Peri, & Shulman, 2016).

Parental marital status, hypothesized to play a role in respondents’ expectations because of links between parental divorce and emerging adult behavior (e.g. Amato & Patterson, 2017), was not significantly associated with either expectation. The research on divorce expectations in adolescents and emerging adults has generally focused on parental divorce as a main indicator (Boyer-Pennington et al., 2001; Dennison & Koerner, 2008), but research including other measures has suggested that the role of parental marital experience may be small once other experiences are considered (Arocho & Purtell, 2018). These findings support this.

Although socioeconomic factors predict actual marriage quite well (Gibson-Davis et al., 2018) and prior work as found socioeconomic predictors of marital intentions (Arocho & Kamp Dush, 2018) the subtle variation in socioeconomic factors measured in this study did not predict variation in either marriage or divorce expectations, except for employment being associated with marriage. The observed association was only apparent when looking between individuals, suggesting that an overall greater time spent in full-time employment was more important for predicting higher expectations of marriage than were individual bouts of less employment. Employment, especially full-time employment and starting a career, is associated with greater feelings of readiness for marriage (Carroll et al., 2009), and characteristics of a person’s job predict differential likelihoods of actually getting married (Kuo & Raley, 2016), thus future research may do well to examine links between variation in additional employment characteristics and marriage expectations.

Interestingly, the finding on responsibility for self contradicted the hypothesis that those holding greater responsibility for their own living would expect marriage more strongly. Those who feel greater independence and who recognize their ability to care for themselves may be less likely to expect marriage because they recognize the possibility of gaining fulfillment in life without it. Indeed, many individuals can and do make meaningful, satisfying lives without romantic partners (DePaulo, 2014). Perhaps, then, respondents reporting expectations for marriage may be doing so out of both desires for companionship and a belief that they will need the partnership to build the life they want. Although some theories of the role of marriage suggest that U.S. marriages have moved towards a self-actualization model (Finkel, Hui, Carswell, & Larson, 2014), for many Americans, marriage still helps to fulfill basic needs for security, safety, and providing for a household (Pietromonaco & Perry-Jenkins, 2014). Those that feel they have less of a need for the financial and instrumental aspects of marriage may report lower expectations of marrying.

For divorce expectations, the positive association with arrest was not necessarily unexpected, but is interesting. Prior research has shown that marital expectations are not damped by being involved with law enforcement (Arocho & Kamp Dush, 2016; Manning et al., 2007), and the results of this study supported that for marital expectations. However, perhaps those who have been involved with the justice system, at least as measured through experiencing arrest, are not necessarily more pessimistic about getting married, just about staying married if they do. The greatest benefits of marriage on crime desistance may come for those least likely to marry (King, Massoglia, & Macmillan, 2007), but those with criminal justice experience may already harbor concerns about divorce before they approach marriage. As pessimism towards divorce may be associated with delaying or avoiding marriage in favor of less formal unions (Perelli-Harris et al., 2017; Arocho, 2019), those who could most benefit from marriage (in terms of criminal desistance) may be less willing to try it.

Also in the contextual measures, respondents who reported no religious identity were less likely to expect to marry and respondents with the strongest religious identity were likely to have stronger expectation for marriage and also lower expectations for divorce. Stronger religious identity is indeed associated with earlier marriage (Uecker & Stokes, 2008), so this is not surprising. What is interesting is that variation in religious identity within individuals was not associated with expectations, suggesting that, although religious identity and involvement often change over emerging adulthood (Chan et al., 2015), general religious inclination, measured in this study by the mean values of religious identity, may be more important for predicting expectations than particular measures of identity over time. Adolescent religiousness is predictive of later union behavior (Eggebeen & Dew, 2009), also suggesting that early or overarching religiousness may be key.

Finally, mental and emotional wellbeing measures were predictive of marriage expectations only. Those with greater mean worry were less optimistic in their marital expectations, and intraindividual variation in worry functioned the same way. Higher mean wellbeing was associated with more optimistic marital expectations between individuals. However, wellbeing and worry did not extend to expectations for divorce. The scale of worry in this study dealt mainly with issues of financial or socioeconomic concerns for the future. As such, this finding may be more of a reflection of feeling able to provide or care for a household and relationship, which is itself often a marker of readiness for emerging adults (Carroll et al., 2009). Having greater wellbeing, conversely, predicted greater expectations for marriage. This generalized measure of success and contentment in various aspects of life may indicate that an individual has achieved a personalized marker of adulthood (Silva, 2013). Silva has suggested that, with the challenges to accomplishing “traditional” markers of adulthood that many youth face, especially those from the working class, resolving personal challenges may be one way emerging adults can reconcile their developmental achievements and feel ready for adult roles (2013).

Overall, these results suggest that individuals’ expectations for marriage and divorce at any given time over early emerging adulthood may be driven somewhat by their context, especially their relationship status, religious importance, employment or employability, socioeconomic worry, deviant behavior, and overall wellbeing. These may be the most important things to keep in mind when using single measures of these items or trying to understand what these expectations actually represent when respondents are reporting their self-perceived risks or expectations of experiencing certain behaviors.

As discussed briefly above, perhaps expectations for marriage and divorce are extensions of feelings of competence in young adulthood. Although likely directly related with relationship experience and competence, they seem to stem from some other markers of adulthood or adult competence, such as employment or socioeconomic worry, as well. As such, psychotherapies aimed at helping emerging adults develop competence in areas important to them may be a way to help emerging adulthoods achieve their goals and live the life they want (Atzil-Slonim et al., 2016).

Limitations

Although hybrid regression highlighted differences both between and within individuals’ repeated interviews, this method did not provide a causal test of directionality. There is only limited evidence that expectations stimulate later behavior and achievement (Arocho & Kamp Dush, 2018), but it is possible that the directions of change were not as proposed. Also, expectations for success in marriage may mean different things to different people. This study focused on expectations to avoid divorce as “success,” but more nuanced measures of anticipated success would be important in future research. It should be noted that the sample is of younger emerging adults, with the average age at last interview being only 24. Expectations for marriage were associated with age, but with additional years more associations may be uncovered. Finally, although the original PSID was nationally-representative in 1968 (Institute for Social Research University of Michigan, 2017), even with some adjustment to the sample in the 1990s the TAS sample was not sampled from the current diversity of the U.S. Results may not be totally generalizable to groups that may have grown and changed since the selection of the original families, such as newer immigrant families.

Conclusion

All told, these results support that expectations may vary as young adults develop and gain experiences, especially romantic experiences (Barr et al., 2015; Willoughby, Medaris, et al., 2015), and suggest that future research should consider individuals’ current context when studying expectations, especially if expectations are measured only once. Experiences with romantic relationships consistently predicted optimistic expectations, but variables such as personal responsibility or mental and emotional health might also be helpful to consider in future research. Socioeconomic characteristics, contextual experiences, and mental and emotional wellbeing also played roles in predicting differences and variation in these expectations in emerging adulthood. The evidence provided by this study should inform future research on expectations, particularly by noting the predictable ways in which cross-sectional measures of expectations may be associated with respondents’ current situations and overall proclivities.