Population Research and Policy Review

, Volume 33, Issue 2, pp 287–305

Cohabitation Expectations Among Young Adults in the United States: Do They Match Behavior?

Authors

    • Sociology Department and Center for Family and Demographic ResearchBowling Green State University
  • Pamela J. Smock
    • Sociology Department and Population Studies CenterThe University of Michigan
  • Cassandra Dorius
    • Department of Human Development and Family StudiesIowa State University
  • Elizabeth Cooksey
    • Department of SociologyOhio State University
Article

DOI: 10.1007/s11113-013-9316-3

Cite this article as:
Manning, W.D., Smock, P.J., Dorius, C. et al. Popul Res Policy Rev (2014) 33: 287. doi:10.1007/s11113-013-9316-3

Abstract

Cohabitation continues to rise, but there is a lack of knowledge about expectations to cohabit and the linkage between expectations and subsequent cohabitation. We capitalize on a new opportunity to study cohabitation expectations by drawing on the National Longitudinal Study of Youth (NLSY79) main youth and two waves (2008 and 2010) of the NLSY young adult (YA) surveys (n = 1,105). We find considerable variation in cohabitation expectations: 39.9 % have no expectation of cohabiting in the future and 16.6 % report high odds of cohabiting in the next 2 years. Cohabitation expectations are associated with higher odds of entering a cohabiting relationship, but are not perfectly associated. Only 38 % of YAs with certain cohabitation expectations in 2008 entered a cohabiting union by 2010. Further investigation of the mismatch between expectations and behaviors indicates that a substantial minority (30 %) who entered a cohabiting union had previously reported no or low expectations, instances of what we term “unplanned cohabitation.” Our findings underscore the importance of considering not only just behavior but also individuals’ expectations for understanding union formation, and more broadly, family change.

Keywords

FamilyCohabitationUnion formationEmerging adulthoodYoung adultsUnplanned cohabitation

Young adults (YA) are part of a transformation in union formation behaviors as demonstrated by a dramatic growth in cohabitation and delayed marriage. The role of cohabitation in union formation in the United States continues to increase such that nearly two-thirds (66 %) of YAs have cohabited, and two-thirds of recent first marriages are preceded by cohabitation (Manning 2013). At the same time, age at marriage is at a historic high (U.S. Bureau of the Census 2011). While change in family experiences has been well documented, little attention has been paid to YAs expectations to engage in family behaviors in the future; such expectations, whether or not they are acted upon, provide clues about continuing family change (Bumpass 1990; Smock and Greenland 2010).

Drawing on data from a large, longitudinal, nationally representative U.S. survey of parents, adolescents, and YAs (the National Longitudinal Survey of Youth or NLSY), this paper extends knowledge about cohabitation by investigating YAs cohabitation expectations and the linkage between expectations and subsequent union formation behavior. To date, no study based on national-level data has evaluated whether and how expectations to cohabit are associated with subsequent entry into a cohabiting union. Taking advantage of new questions added to the NLSY, this study builds on a large literature in demography that explores the connections between intentions and family-related behaviors.

Background

The study of childbearing expectations has a long history in demography (see, e.g., Freedman et al. 1965), with interest in the topic even catalyzing the 1955 and 1960 Growth of American Families surveys (Kiser 1967; Whelpton et al. 1966). The association between fertility expectations and actual fertility has been of particular interest. As Freedman et al. wrote nearly 50 years ago: “On the practical side, expectations about future births may be an important datum for population forecasting. Whelpton and his associates have shown that such expectations expressed by a national probability sample of the white married women in the childbearing years in 1955 were remarkably accurate in the aggregate in predicting for the period 1955–1960 the cohort’s actual fertility” (Freedman et al. 1965, p. 250).

Although some of the emphases have changed and there are more detailed data with which to work, demographers continue to study fertility expectations and their ties to behaviors (see, e.g., Ajzen and Koblas 2013; Hagewen and Morgan 2005; Hayford 2009; Morgan and Rackin 2010). A key finding of this work is while there is a close congruence at the aggregate level between intentions and completed fertility in the U.S., this is not the case at the individual level. Morgan and Rackin (2010), for example, show that by age 45, only 43 % of women had achieved their intended parity measured when they were 24; some under-achieved and others had more children than intended. Correspondingly, Hayford (2009) describes how fertility expectations change over the course women’s childbearing years; expectations are not fixed.

Investigation of marriage desires or expectations is an extension of this type of research. While many such studies focus on documenting expectations or how they may vary by subgroups (Bulcroft and Bulcroft 1993; Crissey 2005; South 1993), a few studies examine the connection between marital expectations or desires and actual marriage. While the general desire to marry is high overall, it does not necessarily predict marriage. Licther et al. (2004) find that while over 70 % of women in their sample desire marriage, only about one-fifth do so in the next 4 years. Among unmarried couples with a recent nonmarital birth, the linkage is even less strong (Waller and McLanahan 2005). They report that in 61 % of cases, both partners are quite optimistic that they will marry. But by a year to a year and a half after childbirth, only 12 % of the couples had gotten married (see also Brown 2000; Guzzo 2009).

There has been little attention paid to expectations to cohabit. A large share of demographic research on cohabitation focuses solely on behavioral outcomes, with numerous studies attempting to predict the timing and likelihood that an individual will enter a cohabiting union, whether a first union is a cohabiting or marital one, or the likelihood of marriage among cohabiting individuals or couples (e.g., Carlson et al. 2004; Oppenheimer 2003; Sassler and Schoen 1999; Raley et al. 2007; Smock and Manning 1997; Thornton et al. 1995; Zito 2013). Other studies focus on attitudes toward cohabitation. The main findings are well known; there has been a dramatic increase in support for cohabitation among youth over time. In 1976, 40 % of high school seniors reported that cohabitation was a testing ground for marriage compared to nearly 70 % in 2008 (Bogle and Wu 2010; see also Thornton and Young-DeMarco 2001). Drawing on nationally representative data collected in 2001 and 2002 from unmarried YAs ages 20–24, the majority agreed (70 % of cohabitors, 59 % of daters, and 60 % of singles) that cohabitation is “all right” even if there are no plans for marriage (Scott et al. 2009).

With the exception of one article focusing on adolescents in a mid-size city (Manning et al. 2007), there has been virtually no attention to expectations to cohabit in the first place and no research on the connection between cohabitation expectations and subsequent behavior in the United States (see Liefbroer et al. 1994 for a study examining cohabitation expectations in the Netherlands). This gap in the knowledge is not so surprising given cohabitation has only recently become a typical relationship experience and, until recently, national-level data did not allow the study of this issue.

However, the omission is important because a central tenet of social psychology is that the primary individual-level factor determining whether a behavior will occur is the intention to perform that particular activity (Fishbein and Ajzen 1975, 2010). According to the Theory of Planned Behavior, as it has come to be called, intentions are key proximate determinants of behavior. This theory, while it has been criticized and elaborated in various ways, still serves as an important framework for fertility research (Ajzen and Koblas 2013; Barber 2011; Morgan and Bachrach 2011; Philipov 2011). While intentions do not perfectly predict behavior, as we have seen, there is a correlation between intentions and behaviors (Brown 2000; Guzzo 2009; Licther et al. 2004; Liefbroer 2011; Miller and Pasta 1995; Schoen et al. 1999; Waller and McLanahan 2005). It is important to note that whether one calls them “expectations,” as we do here, or “intentions,” there is evidence that people respond to these and similar terms in much the same way (Philipov 2011).

Beyond the large gap in knowledge about cohabitation expectations and links to behavior, there are at least three other reasons to study cohabitation expectations and their connections with entry into a cohabiting union. The first pertains to illuminating some of the mechanisms underlying and sustaining the increased prominence of this union type. For example, as YAs expand the number and type of relationships they experience, we may observe greater growth in expectations to cohabit. Similarly, expectations to cohabit may be responsive to economic conditions as financial constraints may impair young couple’s abilities for independent living. A second rationale for studying expectations, especially among young people, is that they can be interpreted as an early harbinger of broader changes in social norms. This is so, in part, because barriers to a behavior (e.g., a poor economy) may prevent achieving behavioral goals but not necessarily the desire to realize that goal (Gibson-Davis et al. 2005; Halpern-Meekin 2012). That is, a behavioral measure cannot as completely tap the perceived desirability of a behavior as an intention indicator, especially in the demographically dense years of young adulthood (Rindfuss 1991). This rationale motivated past studies on race differences in the desire to marry and the actual occurrence of marriage (e.g., Bulcroft and Bulcroft 1993; South 1993). A third reason is to more rigorously examine a finding from qualitative research that cohabitation is often not a purposeful decision, but more a “slide” into cohabitation (Manning and Smock 2005). If this is the case, we would expect fairly high levels of what we term “unplanned cohabitation.”

Current Investigation

Our study thus has two primary aims. The first is to identify key correlates of YAs’ expectations to cohabit. The second is to use our expectations measure, along with a range of other covariates, to predict whether a YA subsequently enters a cohabiting union. Because of the recent addition of new questions about cohabitation expectations to a large, nationally representative survey (the NLSY), we have a unique opportunity to address these aims. The new 2008 NLSY questions ask whether respondents expect to cohabit and then query about the chance they will cohabit in the next 2 years based on a scale from 0 to 100. We extend the second aim to determine the extent of a mismatch between expectations and actual cohabitation. We specifically draw on earlier qualitative work (Manning and Smock 2005) to consider unplanned cohabitations. Unplanned cohabitations are those that occurred between interview waves but were not anticipated at the time of the 2008 interview.

We consider key factors found to influence cohabitation including relationship history, sociodemographic characteristics, activity status, and religious attitudes. Adolescents who were in dating relationships more often subsequently cohabited than those who were not (Raley et al. 2007). Parents have higher odds of cohabiting than childless YAs (Guzzo 2006a; Schoen et al. 2007; Stewart et al. 2003). Respondents with prior cohabitation or marriage experience more often cohabit than remain single (Guzzo 2006b). Childhood family experiences, specifically family structure, are tied to adult union formation (e.g., Axinn and Thornton 1993; McLanahan and Sandefur 1994; Ryan et al. 2009; Sassler et al. 2009; Zito 2013). An additional indicator of family background used in prior work is maternal education. Cohabitation is associated with lower maternal education (Landale et al. 2010; Lehrer 2004; Schoen et al. 2007). The traditional demographic indicators include gender, age, race and ethnicity, and region. The motivations for cohabitation differ for men and women (Huang et al. 2011). Younger age is positively associated with cohabitation (Guzzo 2006b). Whites and Hispanics experience higher rates of cohabitation as a first union than Blacks (Copen et al. 2013; Guzzo 2006a;), but similar proportions of Whites, Blacks, and Hispanics have ever cohabited (Manning 2013). Given regional variation in cohabitation (Lesthaeghe and Neidert 2006), we include variables distinguishing four regions. Activity status taps respondents’ education and employment experiences which are still in process during young adulthood. Based on prior work the odds of cohabitation are higher among YAs who are employed and lower among those enrolled in school (Guzzo 2006a; Landale et al. 2010). Strong religious convictions are an indicator of conservative attitudes (Pearce and Thornton 2007) and associated with lower likelihoods of cohabitation (Eggebeen and Dew 2009).

Data and Methods

We draw on nationally representative data from the 1979 National Longitudinal Study of Youth (NLSY79) main youth and 2 waves (2008 and 2010) of the YA surveys. To date, no other longitudinal data set in the United States includes cohabitation expectations data along with subsequent cohabitation behavior. Born between 1957 and 1965, main youth respondents represent the later Baby Boom birth cohorts who entered young adulthood in the late 1970s and early 1980s when American families were in flux with increasing divorce rates, nonmarital childbearing and cohabitation. These NLSY79 respondents have been interviewed every year from 1979 through 1994 and biennially thereafter. Data collected on fertility and union experiences permit the construction of living arrangement and childbearing histories.

In 1986, biennial interviewing began with all children born to the NLSY79 women, and starting in 1994 all children ages 15 and older were interviewed every other year as “Young Adults.” We focus on these YAs and link their data to information about their mothers, female NLSY respondents, on their own relationship histories. Because men’s biological children were not included in the YA sample, they were not eligible for inclusion in our analyses. The YA questionnaire also includes a wealth of information relating to education, employment, dating, fertility, and expectations for the future.

The analytic sample is restricted to firstborn children over 18 years old who had not married or cohabited prior to age 16, reported being single (not cohabiting or married) in 2008, and were categorized as White, Hispanic or Black (n = 1,534). The sample was limited to those who answered cohabitation and marital expectations variables in 2008 (n = 1,347). We then excluded respondents who did not respond (n = 207) or did not provide valid union formation data in the 2010 survey (n = 35). The final analytic sample is 1,105. Analyses of attrition from the 2008 to 2010 surveys indicate that overall those who were less likely to participate were older in 2008 and had a child in 2008. The mean cohabitation expectations are not statistically different across the 2008 and 2010 samples.

Measures

Our key independent variable is cohabitation expectations measured in 2008. The question reads: “Do you think you will [ever/ever again] live with an opposite-sex partner, to whom you were not married?” The responses are a dichotomous “yes” or “no.” Respondents who were same-sex were provided a separate response and coded appropriately. A subsequent timing question, answered by those who responded affirmatively, asks “On a scale of zero to 100, where zero means you are sure this will not happen and 100 means you are sure it will, how likely are you to live with an opposite-sex partner, to whom you were not married in the next 2 years?” Note that the second question in the sequence introduces a timing component (e.g., next 2 years); thus, a “yes” to the first does not necessarily indicate a high value on the second. Based on the distribution of responses and only for illustrative purposes we categorize the expectations into a four category variable ranging from no expectation, low expectation (1–24 %), moderate expectations (25–74 %), and high expectations (75–100 %).

Subsequent union formation behavior is based on responses to questions about the start of cohabitations and/or marriages since the 2008 interview. Due to sample size limitations, we identify respondents who entered a cohabitation (e.g., only cohabit or cohabit and later married), giving them a value of 1 on this variable and all other respondents a value of zero. Clearly, the 2 year time frame provides only a short period for union formation. While we create an indicator predicting marriage, we do not present findings because of the focus of the paper and the small sample size (n = 61). We note that the results focusing on behavior may not be broadly generalizable because analyses are limited to behavior in a 2-year time window.

The models include four indicators of the YAs sociodemographic characteristics measured as of the 2008 interview. We include age, gender, race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic), and region of residence in 2008 (Northeast, West, South, Northcentral). Mother’s education is a four category variable measured when the child was age 15: less than high school, high school, some college, and college graduate.

The NLSY provides an opportunity to include independent variables measuring the child’s exposure to particular family types from birth to age 18, drawing on mothers’ reports of union experiences. Creating family histories is complex, but we believe that the quality of the data produced is high. At each survey, respondents reported whether they were currently in a residential relationship, provided information on relationship type (marriage, cohabitation, single), up to three changes in relationship status that occurred since the prior survey (divorce, move out, marriage, move in), and start and end dates of each relationship (coded as century months). Since 1990, the NLSY79 has included a series of additional cohabitation questions about whether the participant cohabited before marriage (including a retrospective report of cohabitation prior to their current marriage). In later survey years, respondents are asked if the cohabiting relationship was continuous, if a cohabiting partner was present at the time of the survey, whether there was a gap of singlehood in the past year in which cohabitation could occur, the month cohabitation began and ended, and the number of cohabitations occurring during the past year. We also drew on the NLSY79 Fertility File and Household Roster because it provides two constructed variables allowing us to identify individual men in the household: A unique partner ID number for every residential partner, and a variable identifying cohabiting partnerships where men were identified as living in the household, but for which no cohabitation data were collected prior to 1990. Because each of the mother’s partners was given a unique ID number that was maintained for every year the man was present in the household, it is possible to identify birth fathers and calculate the total amount of time children lived with biological and step fathers. This strategy makes it possible to link children to specific residential relationships, with corollary information on the biological mother’s and father’s relationship start date, end date, duration, and type. We create a series of family experience variables, two biological parents marriage, two biological parent cohabitation, stepparent marriage, stepparent cohabitation, and single parent single, to measure whether the respondent ever lived in each specified family between birth and age 18.

Our analyses also include several controls tapping relationship history. The dating status of respondents is established with a question in the 2008 interview that asks “Are you going out with one particular person, or are you dating more than one person?” We account for the respondent’s parenthood status with a dummy variable indicating whether he or she had given birth to a child prior to the 2008 interview. We include a variable measuring whether the respondent cohabited prior to 2008 and another variable indicating whether the respondent had married prior to 2008.

An indicator of activity status was generated to measure whether the respondent was in school or employed in 2008. The variable has three mutually exclusive categories: enrolled in school (may be working or not); employed at least 10 h a week, and neither in school or working. While it is challenging to measure employment and education during young adulthood, we believe this measure is a reasonable proxy.

Religiosity is included as a continuous variable that is based on a question in 2008 asking “In past year how often have you attended religious services: more than once a week, once a week, two to three times a month, once a month, several times a year, or not at all?” and was reverse coded to go from low to high religious attendance. Similar results are obtained when relying on an indicator of religiosity at age 18 and mother’s religiosity reported in 1979.

Methods

First, we present descriptive statistics on all the variables with an emphasis on the cohabitation expectations indicators. Next, we estimate tobit models predicting cohabitation expectations. Tobit models are appropriate because it accounts for the skewed distribution of cohabitation expectations; a substantial proportion does not expect to cohabit or marry within the next 2 years. The tobit model uses all the information available, including information on censoring, and tobit regression results in less biased and more consistent estimators than ordinary least squares regression (DeMaris 2004). The results are similar when we rely on ordinary least squares regression. The initial model is a zero-order model and then we present the full model including all the covariates.

Next, we assess how expectations are related to subsequent cohabitation behavior, using logistic regression models. We present models that predict the log odds of cohabiting within a 2-year time frame. We first present a model with cohabitation expectations, a model with traditional predictors, and a full model that includes the relationship expectations and all control variables.

Our final set of results focus on how cohabitation expectations match or mismatch observed behavior. These descriptive results rely on the four category cohabitation expectation indicator. First, we present the percentage of YAs who cohabited at differing levels of certainty of future cohabitation. Second, we examine the distribution of expectations among those who cohabited and those who did not cohabit. Our specific interest is in respondents who cohabited without strong expectations to cohabit, unplanned cohabitation.

Results

Table 1 provides a descriptive profile of the sample. Three-fifths, 60 %, of respondents expected to cohabit at some point in their future. Once timing is incorporated into the measure with responses ranging from zero to 100 % chance of forming a cohabitation in the next 2 years, the mean score of cohabitation expectations was 34 %. This mean level of cohabitation expectations was heavily affected by the proportion not expecting to cohabit. We report in the table the mean score on expectations among those who expect to cohabit in the future, scoring 1–100. On average respondents reported a 63 % chance of cohabiting in the next 2 years among those who expect to cohabit in the next 2 years. For illustrative purposes, we present a four indicator variable indicating the strength of cohabitation expectations. We find that 46 % did not expect to cohabit within a 2 year time frame, 6 % reported low (1–25 % chance) expectations, 25 % indicated moderate (25–74 % chance) expectations, and 22 % stated high (75 % or higher chance) expectations to cohabit.
Table 1

Descriptive statistics for sample

 

Weighted p/mean

Expectations to cohabit

 Ever expect to cohabit (0,1)

60.1

 Expect to cohabit next 2 years (0–100)

33.6

 Expect to cohabit next 2 years (1–100)

62.6

 Expect to cohabit next two years

  None

46.3

  Low

6.3

  Moderate

25.2

  High

22.2

Union formation

 Cohabit (2008–2010)

  Yes

16.3

  No

83.7

Independent variables

 Dating 2008

66.9

 Ever cohabited 2008

23.0

 Ever married 2008

3.5

 Parent 2008

16.5

 Family structure experiences

 

  Two biological cohabiting parents

7.0

  Two biological married parents

74.7

  Step cohabiting parents

21.2

  step married parents

24.6

  Single mother

45.5

 Mother’s education

  Less than high school

5.9

  High school diploma/GED

43.6

  Some college

26.2

  Bachelors degree

24.3

Female

46.0

Age

22.8

Race/ethnicity

  White Non-Hispanic

77.1

  Black Non-Hispanic

17.0

  Hispanic

5.9

Region

 Northeast

17.1

 Northcentral

28.5

 South

40.6

 West

13.8

Activity status

 Enrolled school

50.0

 Employed (at least l0 h/week)

32.5

 None

17.5

Religiosity (1–6)

2.8

 Unweighted N

1,105

Source NLSY

Chances of cohabitation and marriage: low = 1–24, moderate = 25–74, high = 75–100

Table 1 also provides information on relationship experiences. Two-thirds of the sample was currently dating at the time of interview. Nearly one-quarter (23 %) of the sample had cohabited prior to interview and 3.5 % had been married. We find 16.5 % of respondents had a child prior to 2008. In terms of family background, we include indicators that are not mutually exclusive. We find 7 % ever resided with cohabiting two biological parents, 74.7 % ever lived in a married two biological parent family, 21.1 % ever lived in a cohabiting stepparent family, 24.6 % ever resided in a married stepparent family, and 45.5 % ever lived with a single mother. The modal category for maternal education was high school graduate. The sociodemographic variables indicated that 46 % of respondents were female and the mean age in 2008 was 22.8. In terms of race and ethnic categorizations, 77 % were White Non-Hispanic, 17 % Black Non-Hispanic, and 6 % Latino. Respondents were distributed more heavily in the south and fairly evenly across the remaining regions of the United States. Half of respondents were enrolled in school at time of interview, one-third were employed at time of interview (and not enrolled), and 17.5 % were neither working nor in school. The mean level of religiosity was 2.8 indicating attending on average once a month.

Table 2 presents estimates from tobit models predicting cohabitation expectations (Zero-order and Model 1) in the next 2 years. Dating at the time of interview was associated with greater expectations to cohabit. YAs who had experience cohabiting reported greater cohabitation expectations. Prior marital experience was not associated with cohabitation expectations. In the zero-order and multivariate model parenthood status was not tied to cohabitation expectations.
Table 2

Tobit regression coefficients cohabitation expectations in next 2 years

 

Cohabitation expectations

Zero order

Model 1

Dating 2008

15.90***

18.21***

Ever cohabited 2008

20.38***

10.95*

Ever married 2008

14.25

7.55

Parent 2008

5.32

5.42

Family structure experiences

 Two biological cohabiting Parents

14.81

11.64

 Two biological married parents

1.75

0.75

 Step cohabiting parents

7.64

−4.94

 Step married parents

8.98

7.57

 Single mother

7.38

3.44

Mother’s education (ref. high school)

 Less than high school

1.49

2.20

 Some college

−10.06

−7.00

 Bachelors degree

−15.78**

−5.77

Female

−11.97**

−10.91**

Age

0.89

−0.29

Race/ethnicity (ref. White)

 Black Non-Hispanic

−18.77***

−13.44*

 Hispanic

−10.36

−7.01

Region (ref. Northeast)

 Northcentral

3.54

6.89

 South

−10.01

2.55

 West

1.41

4.24

Activity status (ref. Employed)

 Enrolled School

−11.24*

−3.38

 None

−25.06*

−20.61*

Religiosity (1–6)

−14.32***

−12.79***

N = 1,105. Source NLSY

p  < 0.05; ** p  < 0.01; *** p < 0.001

Family background was not strongly linked to cohabitation. Table 2 presents the indicators of ever experiencing each family type and shows that they were not related to cohabitation expectations in the zero-order or multivariate model. At the bivariate level, respondents who have better educated mothers reported lower cohabitation expectations, but maternal education was no longer related to cohabitation expectations in the multivariate model. Being female was related to weaker cohabitation expectations, and age was not associated with cohabitation expectations. In the zero-order and multivariate model Black respondents reported lower cohabitation expectations than Whites and Hispanics shared similar cohabitation expectations as whites. Region was not associated with cohabitation expectations. With regard to activity status, enrollment in school was negatively associated with lower cohabitation in the bivariate, but not multivariate model. YAs who were not in school or employed had lower expectations to cohabit. Religiosity was strongly negatively associated with cohabitation expectations.

Table 3 presents the logistic regression log odds of cohabitation between 2008 and 2010. We tested a series of models: Model 1 presents the expectations indicators, Model 2 shows the traditional predictors of cohabitation, and Model 3 includes all the variables. Model 1 shows that cohabitation expectations reported in 2008 were significantly associated with higher odds of cohabiting by 2010. The magnitude of the odds ratio appears relatively modest, but this value represents a one unit change in expectations, ranging from 0 to 100. A ten unit change in expectations resulted in 15 % higher odds of forming a cohabiting union.
Table 3

Logistic regression odds ratio of cohabitation

 

Model 1

Model 2

Model 3

Cohabitation expectations 2008

1.01***

 

1.01***

 Dating 2008

 

1.95***

1.67*

 Ever cohabited 2008

 

1.60*

1.47

 Ever married 2008

 

1.39

1.30

Parent 2008

 

1.88**

1.81**

Family structure experiences

   

 Two biological cohabiting parents

 

1.56

1.48

 Two biological married parents

 

1.33

1.32

 Step cohabiting parents

 

0.94

0.99

 Step married parents

 

0.89

0.82

 Single mother

 

1.19

1.18

Mother’s education (ref. high school)

 Less than high school

 

0.96

0.94

 Some college

 

0.90

0.92

 Bachelors degree

 

0.64

0.66

Female

 

0.86

0.91

Age

 

0.99

0.99

Race/ethnicity (ref. White)

   

 Black Non-Hispanic

 

0.91

1.01

 Hispanic

 

0.82

0.84

Region (ref. Northeast)

   

 Northcentral

 

1.62

1.60

 South

 

1.31

1.33

 West

 

1.85*

1.89*

Activity status (ref. Employed)

 Enrolled school

 

0.68

0.69

 None

 

0.74

0.80

Religiosity (1–6)

 

0.84**

0.88*

N = 1,105. Source NLSY

p  < 0.05; ** p  < 0.01; *** p < 0.001

Model 2 presents the traditional predictors of cohabitation. As expected respondents who were dating at time of interview more often cohabited than those were not in a relationship. Prior cohabitation experience was positively associated with cohabitation while prior marriage was not related to the odds of cohabitation. Single parents had higher odds of cohabiting than YAs without children. In terms of sociodemographic indicators, only region was related to the odds of cohabitation within the 2008 and 2010 interval. YAs living in the West had significantly higher odds and those in the North Central region experience only marginally (p = 0.10) higher odds of cohabiting than YAs the North East region. The activity status indicators show that YAs who were enrolled in schooled had marginally (p = 0.07) lower odds of cohabiting than respondents who were working. Respondents who reported higher levels of religiosity had significantly lower odds of cohabitation.

Model 3 adds the expectations to the model and shows that cohabitation expectations remained positively associated with the odds of cohabiting. In this model prior cohabitation was only marginally (p = 0.07) associated with the odds of cohabitation. The remaining variables maintained their significance and magnitude of influence on the odds of cohabitation. The cohabitation expectation variable significantly (p = 0.002) added to the fit of model and appeared to have an independent effect on cohabitation.

While cohabitation expectations were positively associated with the odds of cohabitation, expectations were not perfectly aligned with behavior. At least half of the sample experienced some form of mismatch in expectations and behavior and the discrepancies are detailed in Table 4. The first panel of Table 4 presents the distribution of cohabitation between 2008 and 2010 according to cohabitation expectations in 2008. Only 30 % of respondents who reported high expectations (75–100) in the two-year time window in fact cohabited within 2 years. Among the subset of respondents who stated 100 % certainty of cohabitation within the timeframe, 38 % had cohabited by 2010 (results not shown). In contrast, 9 % of YAs who reported no chance of cohabitation indeed cohabited.
Table 4

Cohabitation expectations and behavior (2008–2010)

Cohabitation (2010)

Expect to cohabit in next 2 years (2008)

None (46.3 %)

Low (6.3 %)

Moderate (25.2 %)

High (22.2 %)

No

90.5

91.4

81.7

69.6

Yes

9.5

8.6

18.3

30.4

 

100.0

100.0

100.0

100.0

Cohabitation expectations (2008)

Cohabitation (2010)

Yes (16.3 %)

No (83.7 %)

None

26.9

50.1

Low

3.3

6.9

Moderate

28.3

24.5

High

41.5

18.5

 

100.0

100.0

n = l,105. Source NL5Y

Weighted %

Chances of cohabitation: low = 1–24, moderate = 25–74, high = 75–100

Shifting the focus to cohabitation behavior, the second panel of Table 4 shows that 41.5 % of respondents who cohabited in 2010 had declared a high expectation of cohabitation in 2008. About 30 % of respondents who initiated a cohabiting union between the 2008 and 2010 interview waves had not expected or had low expectations to cohabit by 2010. These findings show that a substantial minority of respondents who cohabited experienced an “unplanned” cohabitation (cohabited with no or low prior expectations). The next column shows that half of respondents who did not cohabit in the interval had no expectation of cohabitation by 2010. Nearly one-fifth of those who did not cohabit by 2010 had reported high cohabitation expectations and one-quarter had expressed moderate cohabitation expectations in 2008. Thus, a substantial minority (43 %) did not achieve their cohabitation expectations and experienced what could be termed “impaired” cohabitation (no cohabitation with moderate or high expectations).

Discussion

As age at first marriage continues to rise and a troubled economy makes it difficult for many YAs to form independent households, we argue that studying expectations is an increasingly important avenue for gauging changing social norms about families in addition to their value in signaling future union formation behavior. This study is the first to examine YAs’ cohabitation expectations and subsequent cohabitation behavior because, until recently, cohabitation expectation indicators were not available in national surveys. We leveraged new data available in the NLSY to investigate expectations and their connections to behavior.

Consistent with cohabitation having become commonplace, our results indicate that 60 % of single YAs expect to cohabit sometime in the future. We find nearly one-quarter of single YAs indicate a high or certain expectation of cohabiting within the next 2 years. We also find that cohabitation expectations are positively linked to cohabitation even with the inclusion of an array of conventional predictors of cohabitation.

What factors are associated with expecting to cohabit and with doing so? A first set of important variables concern past and current relationship experiences. Dating is positively associated with expecting to cohabit. Daters likely contemplate the potential future of an actual relationship compared to YAs who do not report being in a dating relationship. Not surprisingly, too, being in a dating relationship is positively linked with actually forming a cohabiting relationship. An implication of these findings is that “single” is a very broad term, underscoring the relevance of past calls for more attention to dating in research on union formation. Clearly, dating is quite relevant. Future research might consider not only whether respondents are dating, but also whether and how the qualities of their relationships influence both expectations and later union formation behavior. Prior experience with cohabitation is also associated with stronger expectations to cohabit another time and with subsequent cohabitation. As a consequence, we are likely to observe growth in what has come to be called “serial” cohabitation (i.e., cohabiting more than once). These findings are also consistent with the notion that YAs are experiencing considerable “relationship go rounds” as they enter and exit numerous relationships during their YA years (Arnett 2004; Cohen and Manning 2010).

A third relationship indicator that we investigated is having a child. Given that cohabiting unions increasingly include children and that growing shares of children are born into cohabiting unions (Kennedy and Bumpass 2011), it is important to examine how children shape union formation expectations and behavior. In this analysis of singles, already having a child does not appear to influence expectations about forming a cohabiting union. There may be countervailing forces operating in that having a child may be a motivation to form a coresidential union but, simultaneously, engender caution on the part of parents to form what are most likely stepfamilies (Gibson-Davis et al. 2005; Manning et al. 2010). At the same time, our results show that parenthood is positively associated with the odds of actually cohabiting. Couples who have a child may be on the pathway to family formation and may experience some social pressure to parent together rather than apart, resulting in higher odds of cohabiting. As cohabitation continues to become a more viable, and visible, family context in which to raise children, further attention to cohabitation expectations and behavior among single parents is warranted.

Activity status is also consequential for cohabitation expectations and behavior. YAs who are not enrolled in school or employed have lower expectations to cohabit, a union that does not have as high of a perceived economic bar as marriage but nonetheless requires resources to live independently as a couple. Our results regarding school enrollment are consistent with other recent studies (Guzzo 2006a; Landale et al. 2010). Interestingly, neither enrollment nor employment is associated with actual cohabitation within the time interval. Our time span, however, encompasses a recessionary period, and it is thus quite possible that the activity status measures are not operating in the expected manner. Not only is this time period characterized by economic uncertainty, but also the uncertainty may be particularly salient for YAs given the flux that characterizes employment and education in early adulthood.

In addition to relationship experiences and activity status, we find that religiosity is also associated with cohabitation expectations and behaviors. Qualitative data indicates that religious beliefs create a “frame” around union formation decisions (Manning et al. 2011). As expected, our results indicate that single YAs with stronger religious beliefs, proxied by frequency of attendance at religious services, report lower cohabitation expectations and subsequently lower odds of cohabitation (see also Eggebeen and Dew 2009). Thus, the values of YAs align with both expectations and behavior. Still, there are YAs with high levels of religiosity who also either expect or do enter a cohabiting union.

Family background measures are standard in models of union formation. Our results suggest that family type indicators are not strongly linked to cohabitation expectations. Respondents who spent time in cohabiting parent families (either biological or quasi step parent families) do not report stronger cohabiting expectations than YAs who did not live with cohabiting parents. Moreover, family structure indicators do not influence the odds of cohabitation. There may be counteracting forces at work here; YAs may have observed successful cohabiting relationships, resulting in positive views of cohabitation while others may have had negative experiences resulting in objections to cohabitation. To date, there are limited opportunities via available data sources to test why parental cohabitation may, or may not, matter.

Women report lower expectations to cohabit than men but share similar odds of cohabiting. While we cannot interpret the source of the gender differences in YAs’ cohabitation expectations, the finding is consistent with prior work that shows men and women differ in some very specific ways regarding motivations for and the meaning of cohabitation (Huang et al. 2011). Further work on the perceived role of cohabitation in the union formation process may explain men’s and women’s differing cohabiting expectations. While many of the respondents were dating at the time of the interview, our evidence rests on expectations from just one member of the couple. Thus, we cannot establish how men and women in the same relationship view and experience cohabitation. Further work on how couples manage disparate expectations and the possible ramifications for relationship stability is warranted.

One of the main purposes of this paper was to explore the link between cohabitation expectations and behavior, similar in spirit to research on fertility expectations or fertility outcomes. YAs’ expectations are indeed tied to union formation behavior, albeit imperfectly. That is, expectations to cohabit in the next 2 years are not typically met. Nearly one-third (30 %) of YAs who had strong expectations to cohabit in a two-year time horizon did so, but over two-thirds did not. Further analyses of these optimistic singles and perceived barriers to cohabitation, and to marriage, may be helpful for understanding the types of societal changes or life circumstances that would facilitate YAs’ achieving their relationship aspirations.

While this paper illuminates the role of expectations to cohabit as part of the story of family change, we acknowledge several limitations. First, while we focus on expectations at a single point in time, expectations may be somewhat fluid over time (Hayford 2009) and vary according to the qualities of dating relationships. Also the expectations indicators do not reference expectations to cohabit with their current boyfriend or girlfriend, instead just expectations within a specific time frame. Second, our sample is limited to respondents who have met specific criteria in terms of participating in several surveys. Our sample may not be as representative of the most disadvantaged who experience higher sample attrition. Third, our measures of parental union experiences only include those occurring during the respondent’s childhood and reference only the mother’s experience. Ideally, we would include maternal and paternal cohabitation, but the detailed relationship histories necessary to code cohabitation experience in the NLSY are based on mother’s reports and not father’s reports. Another limitation of this study is the focus on mothers without attention to the potentially important influences of peers or the community. These are likely increasingly important for YAs’ expectations and behavior, with parents’ influence waning as YAs establish greater independence and engage in broader communities (Manning et al. 2011). Finally, our focus on cohabitation expectations means that we are limited to assessments of cohabitation from the 2008 interview forward; our analysis of cohabitation does not follow the traditional approach of timing to first union and may explain why few covariates are tied to cohabitation. Longitudinal data that start asking questions about cohabitation expectations in adolescence would be ideal.

Our results are broadly consistent with prior research on expectations and behaviors. Cohabiting expectations are predictive of behavior, but for a high proportion of respondents, expectations and behavior do not match. Given the informal nature of cohabitation and evidence that entering a cohabiting union is often more a slide than a distinct event (Manning and Smock 2005), we anticipated “unplanned cohabitations” (cohabitations that were not expected). And we found them: About 10 % of YAs with no expectation to cohabit and 9 % with very low expectations wound up cohabiting. Further investigation into the planning status of cohabitations may provide clues about the stability and implications of cohabitation. As cohabitation continues to rise, the composition of cohabitors will become more heterogenous (Smock 2000) providing new opportunities and challenges to understand how cohabitation not only fits in, but also has changed, the American family system.

Copyright information

© Springer Science+Business Media Dordrecht 2014