The Role of Religion in Adolescence for Family Formation in Young Adulthood

Abstract

This paper examines the role of religion in adolescence for shaping subsequent family formation. Data were drawn from the National Longitudinal Study of Adolescent Health (n=13,895). We explored the role of three dimensions of religious life—affiliation, attendance, and religious fervor, both singly and in combination for the transition to either marriage or cohabitation. Although each dimension predicted subsequent union formation, it was the particular combination of these dimensions that was important for understanding the likelihood of cohabiting. We also found evidence that patterns of religious identity, attendance, and fervor in adolescence were associated with the length of cohabitation, the likelihood of the cohabitation ending in marriage, and beliefs about the purpose of cohabitation.

Young adulthood is in the midst of a demographic revolution. Delays in marriage, rising rates of cohabitation, high rates of non-marital childbearing, delays in marital births, combined with changing patterns of schooling and work has meant that the movement into adulthood has become even less a predictable sequence of events that it once was (Arnett, 2004; Settersten, Furstenberg, & Rumbaut, 2005). The purpose of this paper is to examine a potentially important factor in the changing landscape of young adulthood: the role of earlier religious beliefs and behavior in shaping subsequent family formation choices.

The connections between religion and family life are receiving renewed attention from scholars. Recent work has examined the role religious practices and beliefs play in parenting behavior (Ellison, Bartkowski, & Segal, 1996; Wilcox, 1998; 2004), marriage (Xu, Hudspeth, & Bartkowski, 2005), intergenerational ties (King & Elder, 1999; Pearce & Axinn, 1998), and demographic behavior (Lehrer 2004; McQuillan, 2004). Scholars are finally turning their attention to the role of religion in the lives of adolescents and young adults. Work by Christian Smith (2003; 2005) has clearly demonstrated the centrality, and the complexity, of religious beliefs for American adolescents. Religious practices and beliefs also have been shown to powerfully shape adolescent sexual values and practices (Regnerus, 2007). Finally, some recent work has begun to address the role of religion in the post-teenage years, when youth are laying the foundations for their subsequent adult lives by making decisions about romantic relationships, cohabitation, and marriage (Arnett & Jensen, 2002; Thornton, Axinn, & Xie, 2007).

There are limitations to a number of these recent studies of the religious influences on family formation. Most of these studies use simple indicators of religion, focusing on one or two dimensions of religious experiences, or combine several dimensions into a single scale. Where union formation patterns are the focus, data have mainly been drawn from older cohorts, reflecting a time period when religious influences may have been stronger (Lehrer, 2000). Finally, with the exception of the work of Thornton and colleagues (Thornton et al., 2007), none of these studies have gone beyond the initial transition to examine the effect of religion on subsequent patterns of union stability or dissolution.

This paper addresses these gaps in the literature. Making use of the first and third Waves of the National Study of Adolescent Heath (Add-Health), we examine how three dimensions of religious beliefs and behavior in adolescence affect the likelihood of cohabitation and marriage in young adulthood. It is our contention that understanding the effects of religion on family formation behavior must consider the multi-dimensional nature of religiosity. In order to fully understand how religious identity, the extent or importance of religious beliefs, and how often the person attends places of worship influence family formation choices, it is important to take into account not only the effect of each of these dimensions, but also the interplay between them. Specifically, we expect that adolescents who regularly attend churches that emphasize the sanctity of marriage and whose beliefs are held to be very important are the most likely to make a transition to marriage, and the least likely to make a transition to cohabitation over the next five years, relative to other combinations of religious beliefs and experiences.

We also examine how religious behavior and beliefs in adolescence, as measured in these three dimensions, might be related to specific patterns of cohabitation experiences. Previous work has demonstrated that religious identity and religiosity reduce the odds of the first union being cohabitation (Lehrer, 2004). Nevertheless, a significant proportion of religiously inclined youth still choose to cohabit (Thornton et al., 2007). There has been little exploration of whether early religious experiences shape subsequent patterns of cohabitation. We anticipate that the particular combinations of religious identity, attendance, and fervor in adolescence are more likely be associated with briefer spells of cohabitation, that are more likely to end in a marriage, and be associated with the view that cohabitation without marriage in mind is unacceptable. Put another way, cohabiting young adults who were regularly attending churches as teenagers with distinctive teachings about marriage and cohabitation and whose religious beliefs as teens were very important are substantially more likely to use cohabitation as part of the marriage formation process.

Religion and the Formation of Intimate Relationships

How might the religious life of teenagers affect their subsequent union formation? At the most general level, Christian Smith (2003) identifies nine possible sources of religious influences on adolescents, which combine into three domains: moral order, learned competencies, and social and organizational ties. Of particular interest for us in this paper is the moral order. Smith notes that religion can promote “…specific cultural moral directives of self-control and personal virtue grounded in the authority of long historical traditions and narratives, into which members are inducted, such that youths may internalize these moral orders and use them to guide their life choices and moral commitments” (2003, 20). These can be accomplished, according to Smith, through moral directives, spiritual experiences and through role modeling.

In this paper we identify three dimensions of religious experience as reported by the adolescent that approximate several dimensions of the cultural moral order: the denominational identity—and thus, after a fashion, the particular teachings about marriage; the pattern of exposure to these teachings as measured by church attendance; and the relative importance of religion to the youth as indicated by their self-report of the importance of their beliefs and as well how often they engage in private religious practices like prayer.

Religious Identity

Particular religions, religious bodies, or religious leaders can have a direct effect on decisions about marriage or cohabitation by their doctrines or their teaching, and there are distinct differences among denominations in their teachings and/or emphasis on marriage. Mormons and Evangelical Protestants have strongly advocated marriage as God’s design for intimate relationships, and that marriage, as a sacred institution, is the corner stone of family life (Dobson, 2004; Southern Baptist Resolutions, 2004; The Church of Jesus Christ of Latter-Day Saints, 2004). As a result, Evangelical Christians and Mormons have consistently discouraged premarital sex and cohabitation and encouraged marriage (Heaton, 1992). Of course conservative denominations may not uniformly “practice what they preach”. Chaves (1997) notes, for example, that public stances of particular religious bodies are often only loosely coupled to concrete practices. Support for this observation come from an analysis of family programming practices of various religious traditions where the most conservative religious traditions (Conservative Protestant and Catholic) were no more likely than more liberal denominations to be offering traditional family programming (Wilcox Chaves, & Franz, 2004). Nevertheless, there is empirical evidence that Conservative Protestants and Mormons are significantly more likely to marry and less likely to cohabit than other types denominational affiliations (Lehrer, 2000; 2004; Xu et al., 2005). Identifying with a Conservative Protestant denomination may also influence union formation indirectly by affecting choices in other domains that influence the likelihood and timing of marriage. For example, some research shows that Mainline Protestants have higher levels of educational attainment than Conservative Protestants (Lehrer, 2000). The relatively low priority Conservative Protestants give formal education has been linked with their comparatively early age at marriage (Darnell & Sherkat, 1997). Similarly, the strong emphasis on high fertility among Mormons predisposes individuals to consider early marriage (Heaton, 1986). We hypothesize, then, that adolescents whose religious identity or affiliation is Conservative Protestant are more likely to marry and less likely to cohabit in young adulthood.

In a similar fashion, the Roman Catholic Church is “pro marriage” in its theology and teachings. It has long held that marriage is a sacrament, and that sexual behavior is only appropriate within marriage (Ellingson, Van Haitsma, Lauman, & Tebbe, 2004; U.S. Catholic Conference, 1977). However, some research suggests that Catholics behavior is less consistent or predictable than one would expect given its emphasis on marriage. While Catholic adolescents are less likely than Evangelical Protestant adolescents to be sexually active, a sizeable minority end their teenage years as nonvirgins (Regnarus, 2007). Several studies have also found that Catholics are less likely than Conservative Protestants or Mormons to enter marriage early (Lehrer, 2004; Xu et. al, 2005). Why do Catholics seemingly avoid marriage despite the strong pro-marriage emphasis of their faith? Evelyn Lehrer (2004) observes that compared to other faiths, the costs of divorce for Catholics are particularly high. These higher costs she suggests, may lead to a greater hesitancy to enter into marriage quickly. Consistent with this idea, she finds evidence that Catholics tend to enter marriage at an intermediate age, avoiding both early and late marriage. Given this, we expect that Catholic youth will be less likely than conservative Protestants to marry, but equally likely to avoid cohabitation.

In contrast to Evangelicals, Mormons, and Catholics, Protestants from the mainline denominations have become increasingly distinctive in their teachings. While continuing to emphasize the importance of strong and vital family relationships, Mainline Protestants have put less emphasis on restrictive teachings concerning sex outside of marriage, been more tolerant of divorce, and have been more accepting of alternatives to the traditional family than Evangelical Protestants (Edgell, 2005; Hargrove, 1983). The higher levels of educational attainment among Mainline Protestants might also be associated with delays in marriage (Lehrer, 2000). Given this, we might expect to find that youth from mainline protestant backgrounds to be less likely than Conservative Protestants to marry, but more likely to cohabit.

In contrast to youth who identify with a religious tradition, those with no religious affiliation are not as likely to be exposed to a compelling vision of how to behave when it comes to marriage and cohabitation. Consistent with what others have found, we expect these youths to be the least likely to marry and the most likely to cohabit (Thornton et al., 2007).

Attendance

Religious identity is far from the whole story, however. For the particular beliefs and teachings of a religious group to exert an influence, individuals must be exposed to these messages. Most typically this means they must attend religious services. We should expect then that the more teenagers attend services and/or participate in church related ministries like young groups or youth services, more likely, all things being equal, they are likely to hear and absorb teachings about what is appropriate sexual behavior, what are appropriate ways to conduct intimate relationships, and the meaning and purpose of marriage. Of course, the relationship between church attendance and behavior is more complicated for adolescents than adults. As some scholars have noted, adolescent attendance at religious services may reflect the expectations or pressure from parents or the momentum of long standing habit, rather than an authentic desire to worship God or understand and follow church teachings (Regnerus, 2007; Smith, 2005). Nevertheless, a number of studies find that church attendance remains a significant predictor of youth behavior (Regnerus, 2007; Regnerus & Elder, 2003). We hypothesize that church attendance will be positively related to the likelihood of marriage and negatively related to the likelihood of cohabiting.

The Importance of Religious Beliefs

A third dimension of religion that has consequences for subsequent marriage and cohabitation among youth is that of the importance they place on their beliefs. The messages, beliefs, and doctrines they are exposed to must be meaningful to them; they must take their religion seriously. Wimberley (1989) argues that religious identity includes both norm adherence and identity salience. Religious norm adherence includes such publicly visible aspects of religious experience as church attendance, and knowledge of church tenets or doctrine. Religious salience, on the other hand, is the internal aspects of religious experience—private religion--and “…concerns the extent to which the religious identity is dominant among other identities that make up the self”(1989, 130). In part, salience is measured by how important individuals say their religion is to them. However salience is also anchored in the extent to which individuals practice their beliefs apart from extrinsic motivations. This could mean the extent to which they engage in private religious practices like prayer, meditation, or scripture reading. It could also mean the extent to which their identity as a religious adherent “wins out” over competing identities (e.g., macho football player) for influencing sexual behavior, for example (Regnerus, 2007). Norm adherence does not necessarily lead to salience. An adolescent might be knowledgeable about Catholic teaching on marriage and sexuality, and may regularly attend mass, but these activities may say little about the extent to which the adolescent “takes his Catholicism to heart” when it comes to making decisions about sexual behavior. On the other hand, Regnerus (2007) argues, high salience can act as a “…stimulant to religious beliefs or cognitive structures, a light switch of sorts that turns on the force of religious belief” (2007:48). We hypothesize that the greater the importance of religion to the adolescent, the greater the likelihood of marriage and lower the likelihood of cohabitation.

The Interplay between Identity, Attendance, and Salience

Thus far we have argued that religiosity is multi-dimensional, and that each of these components independently effects family formation behavior. However, it is how these three dimensions interact together that may tell us more about how religion influences subsequent behavior. Put another way, the “punch” of religion on future marriage or cohabitation is much more powerful when an adolescent identifies themself as a member of particular church with distinctive teachings about marriage, and they are regularly attending church services or youth groups, and their religious beliefs are important to them.

Most of the research that has attempted to link religion and family behavior has focused on affiliation, attendance, or religiosity, but rarely more than one of these dimensions. One exception is the work of Lehrer (2004b) who examined the interaction between two of these dimensions: religious affiliation (Mainline Protestants, Conservative Protestants, and Catholics) and religious attendance at age 14. She finds that patterns of entry into marriage do not vary significantly by frequency of attendance at services for any of the three religious groups. However, she did find evidence of an interaction between affiliation and attendance when it came to cohabitation, with high attending Conservative Protestants the least likely to cohabit. Recent work by Thornton, et. al, (2007) identifies three dimensions of religiosity in a similar fashion to what is proposed in this paper; distinguish between affiliation (Catholic, Non-fundamentalist Protestant, Fundamentalist Protestant or Baptist, other, and none), and two dimensions of religiosity--participation and importance. Interestingly, contrary to what they had expected, they found no evidence that the effects of participation and importance on likelihood of cohabitation or marriage varied significantly between these religious groups. However, these findings were based on sample of heavily Catholic respondents from the Detroit metropolitan area who entered young adulthood around 1980, making generalizations difficult (Pearce & Thornton, 2007; Thornton, et. al, 2007).

Given the above considerations, we anticipate several patterns in the interplay between affiliation, attendance, and salience. We expect to find that variations in attendance and fervor will predict differences within denominations in the likelihood of marriage or cohabitation. We also expect that the differences between denominations in the likelihood of marriage and cohabitation will be considerably moderated by differences in attendance and fervor. Specifically, we expect that adolescents who identify themselves as Conservative Protestants who regularly attend worship services or youth groups, and who report their beliefs are very important, will be significantly more likely than other Conservative Protestant adolescents to marry, and will be significantly less likely to cohabit. Furthermore, we anticipate that these adolescents (high attending, high salience, Conservative Protestants) will the most likely of all the groups to marry and the least likely to cohabit. We also expect Mainline Protestant and Catholic youths to be less likely to marry than Conservative Protestant youth (for the different reasons outlined above) regardless of their level of attendance or fervor. However, we expect that the likelihood of cohabitation among Mainline Protestant and Catholic youth to be lowest among those who frequently attend religious services and for whom religion is important—although less than that of high attending, high fervor Conservative Protestant youths. In sum, we anticipate that religious identity, attendance at religious services, and fervor, separately and in concert, should affect union formation.

Religion and the Nature of Cohabitation

One of the earliest questions in the research literature on cohabitation in the United States revolved around whether the explosive growth of cohabitation in the 1970’s and 1980’s represented an emerging stage in the courtship process or a substitute for marriage (Smock, 2000). More recent work suggests that both types of cohabitations are practiced today in the United States (Casper & Bianchi, 2002; Smock & Manning, 2004). A significant fraction of cohabiting couples end up marrying, but there are also a number of cohabiting unions where the individuals indicated they have no desire to marry, that are of long duration, or that are settings for bearing and rearing children (Brown & Booth, 1996; Bumpass & Lu, 2000). What factors might differentiate between these two types of cohabitations has not received much attention beyond basic demographic characteristics such as race or SES (Blackwell & Lichter, 2000; Smock, 2000). What role might religion play in the expectations or meaning of a cohabiting experience?

One possibility is that adolescent religion has little or no effect on the course of cohabitation in young adulthood. Given the strong prohibition against non marital cohabitation among Conservative Protestants and Catholics, the effect of this particular religious background on current behavior may be small, once an individual has made the choice to cohabit. That is, once an individual has “fallen away”, the movement away from a religiously based life-style may be quite far. We do know, for example, that once individuals cohabit, their religious behavior declines (Thornton et al., 1992). If this dynamic is common, we would expect few differences by adolescent religious experiences on the likelihood of the cohabitation ending in marriage, views on the meaning of cohabitation, or the length of first cohabiting spell.

An alternative possibility builds on the idea that religion in general, apart from any particular doctrines or teachings tend to be correlated with conventional beliefs and lifestyles (Mahoney, Pargament, Jewell, Swank, Scott, Emery, & Rye, 1999). Although this conventionalization process is likely to select many individuals into marriage rather than cohabitation as a first union, it may also have an effect on those who subsequently cohabit as well. Thus cohabiting young adults with some religious roots (regardless of specific identity, attendance, or fervor) are likely to view cohabitation as most appropriate if it points to marriage, and treat their own cohabiting experiences accordingly. If religious conventionalization is the predominant pattern, we would expect to find no differences among the various combinations of identities, fervor and attendance. The major difference in cohabitation patterns, then, is between those who indicate no religious preference in adolescence, relative to all the others.

A third possibility is that particular religious beliefs and behavior evident in childhood and adolescence continue to have an effect into young adulthood. Specifically, youth who faithfully attended Conservative Protestant or Catholic churches, and who indicated their beliefs are important—a pattern that we hypothesize should exert the most influence on subsequent family formation behavior, are more likely than other young adults with differing identities, attendance, and levels of fervor to conduct their subsequent cohabitation as if it were part of the courtship process. If this pattern predominates, we would expect to find that cohabiting young adults, who were devout, high attending Conservative Protestants or Catholics as adolescents, to have shorter spells of cohabitation, have higher odds of exiting those cohabitations into marriage, and are more likely to hold to the value that cohabitation is most appropriately done with marriage in mind.

In sum, we propose three competing hypotheses about the link between adolescent religious experiences and subsequent cohabitation patterns: (a) Religious experiences have no effect on patterns and beliefs about cohabitation; (b) All religious experiences in adolescence will be associated with shorter cohabitation spells that most likely end in marriage as well as the belief that cohabitation should end in marriage; or (c) That only those adolescents who were Catholic or Conservative Protestant youth who attended regularly and whose beliefs were important will have shorter cohabiting spells that typically end in marriage, and believe that cohabitation should end in marriage.

Method

Data

The data for this study were drawn from the first and third Waves of the National Study of Adolescent Health. The Add Health survey is a longitudinal nationally representative sample of 20,745 middle and high school students first interviewed in 1995–1996. A second wave of interviews was conducted one year later, and a third round of 15, 170 persons was interviewed in 2001. Response rates for Wave 1 and 3 were 78.9% and 77.4% respectively. Attrition analyses suggest that the effect of non-response on representativeness was minimal (Chantala, Kalsbeek, & Andraca, 2004). We restricted the sample to respondents who participated in both Waves 1 and 3, who had not been in a cohabiting or marriage union prior to the first survey, and who gave information on their religious affiliation at W1, for a final sample of 13,895. The analyses in this study were all weighted using the Wave 3 weights, adjusting the sample to be nationally representative. A more detailed description of the data can be found in Harris et al. (2003).

Measures

Religion

We measured three dimensions of religious beliefs and behaviors: religious affiliation or identity, fervor or devoutness, and religious attendance. In Wave 1, respondents were asked to identify their particular denominational affiliation (e.g., Adventist, Assemblies of God, etc.). We recoded these 29 choices into five categories: No religious affiliation, Conservative Protestant, Mainline Protestant, Catholic, and Other, using a classification scheme derived from two sources: The RELTRAD method of classifying denominations described in Steensland et. al. (2000), and a scheme used by Christian Smith (2005). These schemes allowed us to categorize specific denominations into groups that share a common underlying theological or doctrinal similarity when it comes to perspectives on marriage and sexual behavior. Catholics are distinguished because of their size (they represent 22% of the sample). All other religious or denominational affiliations that could not be classified into the above three groups are lumped together in the “other” category (4 percent). Regrettably, we were not able to distinguish Muslims (n=33) or Jews (n=108) in these analyses, given their small sample sizes. Finally, about 14 percent of the respondents indicated they had no religious identify or affiliation, and were classified as “none”.

Two Wave 1 questions were used to construct our measure of religious fervor: “How important is religion to you?” and “How often do you pray?”. Responses were combined to form a scale ranging from 2 to 9 (mean 6.54), with higher scores indicating higher fervor or religiosity. Chronbach’s alpha for this scale was .75, indicating that combining measures of belief and practice makes reasonable empirical as well as theoretical sense.

The third dimension of religion we employed was worship service attendance. Attendance was drawn from the W1 question: “In the past twelve months, how often did you attend religious services?” Answers ranged from 1 (Once a week or more) to 4 (Never). This variable was reverse coded, such that higher values indicated greater attendance.

We also created a composite variable with categories that represent combinations of affiliation, high and low fervor, and high and low attendance. Utilizing a mean split on fervor, individuals with a score of 7 or above were coded as being in the high group for this variable, while those below 7 had low fervor. Similarly, individuals who attended worship services more than once per month (a score of 3 or 4) were put in the high attendance group and individuals who attended once per month or less were in the low attendance group. Thus, we divided Conservative Protestants into four groups: those with high attendance and high fervor, those with low attendance and high fervor, those with high attendance and low fervor, and those with both low attendance and low fervor. We divided Catholics and mainline Protestants in the same fashion. We do not divide respondents who are in the “other” or “none” affiliation categories. There were too few respondents in the “other” affiliation category to divide into four groups, and those who had no religious affiliation have virtually no distribution on the attendance and fervor variables. Religious identity, attendance and fervor are combined into a composite variable with sixteen categories.

Other Independent Variables

Other Wave 1 variables included in the multivariate models were age (measured in years), race (European American, African American, Hispanic American, Other), total family income, mother’s education (Less than high school, high school graduate, some college, and college graduate), region of resident at Wave 1 (South, non-south), respondent’s education as of Wave 3 (Less than high school, high school graduate, some college, and college graduate), family structure (two-parent biological, two-parent step-family, single mother, single father, and other), and gender. Descriptive statistics of these and the key independent variables are displayed in Table 1.

Table 1.

Descriptive Statistics

Variable	Mean or Proportion	STD	Range
Age	22.18	1.84	18 – 28
Gender (1 = Male, 2 = Female)	.49
Mothers Education:
Less Than High Schoola	.15
High School Graduate	.33
Some College	.29
College Graduate	.23
Family Incomeb	45.78	39.00	0 – 999
Family Structure:
Two-parent, Biologicala	.58
Two-parent, Non-Biological	.14
Mother Only	.19
Father Only	.03
Other Structure	.05
Race/Ethnicity:
European American	.68
African-American	.16
Hispanic	.07
Other	.10
Region of Residence (1=South)	.39
Respondents education:
Less than High School	.15
High School Graduate	.33
Some College	.40
College Graduate	.12
Religious Affiliation:
None	.13
Conservative Protestanta	.33
Mainline Protestant	.24
Catholic	.24
Other	.05
Religious Attendance:
Worship Service Attendance	2.71	1.17	1 – 4
Religious Fervor	6.55	2.38	2 – 9
Acceptability of Cohabitation without	3.77	1.14	1 – 5
Marriage Plansc

^aComparison Category in the Analyses

^bMeasured in thousands of dollars

^cCohabiting Participants Only N=13, 895

Dependent Variables

The key dependent variable for the analyses in Tables 2 through 4 is the transition to first union, defined here is as either marriage or cohabitation. The Wave 3 questionnaire collected cohabitation and marriage histories, from which we were able to determine the timing and type of union transition for the respondents. By Wave 3, 9.6% (1,436) of respondents had marriage be their first union transition, 39.0% (5,223) had cohabitation as a first union, and 51.5% (7,236) had no union transition.

Table 2.

Proportional Hazards Ratios of the Relationship between Religious Variables and First Union Formation.

Variables	Model 1		Model 2		Model 3		Percent in Union
	Marriage	Cohab.	Marriage	Cohab.	Marriage	Cohab.	Marriage	Cohab
No Affiliation	.41***	1.20***					6.79	51.08
Mainline Protestant	.49***	.95					7.46	37.96
Catholic	.36***	.77***					7.46	33.58
Other Religion	.56***	.87					9.18	33.17
Attendance			1.21***	.87***
Fervor					1.14***	.94***
Age	1.43***	1.22***	1.45***	1.21***	1.45***	1.22***
Family Income	.99***	.99*	.99***	.99**	.99***	.99**
Gender	2.08***	1.63***	2.08***	1.65***	2.01***	1.65***
Black	.33***	.65***	.35***	.68***	.35***	.68***
Hispanic	1.34***	.86*	1.05	.82***	1.03	.82***
Other	.96	.68***	.81*	.66***	.79*	.66***
High School	.99	1.17***	.99	1.18***	.98	1.19***
Some College	1.02	1.13*	1.02	1.14**	1.01	1.14**
College Graduate	.77*	.95	.74*	.99	.75*	.96
Stepparent	1.03	1.46***	1.11	1.43***	1.06	1.45***
Single-Mother	.66***	1.22***	.68***	1.17***	.68***	1.20***
Single-Father	1.05	1.42***	1.06	1.36***	1.06	1.38***
Other	1.11	1.43***	1.13	1.40***	1.12	1.42***
High School	.76***	.53***	.72***	.53***	.70***	.53***
Some College	.53***	.32***	.48***	.32***	.46***	.32***
College Graduate	.34***	.22***	.30***	.23***	.29***	.23***
Region	1.86***	.90***	2.12***	.93*	2.01***	.94*
N	13,895	13,895	13,895	13,895	13,877	13,877

^aComparison group is conservative protestant (13.91% married, 39.55% cohabited)

Note. Control variables in every model are age, race, gender, total family income, parent’s (usually mother’s) education, respondent’s education, region of residence, and family structure.

Participants’ attitudes on the acceptability of cohabitation without plans to marry (Table 5) were measured at Wave 3 with the following question: ‘How much do you agree or disagree with the statement “It is all right for an unmarried couple to live together even if they aren’t interested in considering marriage?”‘. The scale ranges from 1 (Strongly Agree) to 5 (Strongly Disagree). The responses were reverse coded so that higher responses indicated higher acceptability of cohabitation without plans to marry.

Table 5.

OLS Regression Analysis of the Relationship between Composite Religious Variables and Acceptability of Cohabitation without Plans to Marry

Affiliation	Attendance	Fervor	OLS Coefficient	S.E.
Nonea	N/A	N/A	.59***	.06
Other Religiona	N/A	N/A	.46***	.10
Conservative Protestanta	High	Low	.06	.11
Conservative Protestanta	Low	High	.30***	.08
Conservative Protestanta	Low	Low	.34***	.09
Mainline Protestanta	High	High	.21***	.06
Mainline Protestanta	High	Low	.35***	.11
Mainline Protestanta	Low	High	.36***	.09
Mainline Protestanta	Low	Low	.51***	.08
Catholica	High	High	.37***	.08
Catholica	High	Low	.48***	.11
Catholica	Low	High	.40***	.09
Catholica	Low	Low	.60***	.09
N = 5,298

^aComparison Group is Conservative Protestant with High Attendance and High Fervor

Note. Control variables are age, race, gender, total family income, parent’s (usually mother’s) education, respondent’s education, region of residence and family structure.

^*p < .05,

^**p < .01,

^***p < .001

Analytic Approach

Following previous work by Lehrer (2000; 2004), Cox proportional hazards models were estimated that treat marriage and cohabitation as competing risks. That is, we model the hazard of first union transition for both marriage and cohabitation, while statistically accounting for the fact that couples may choose to enter either type of union first. The analyses used the Efron method for handling ties.

We began by examining the effect of each dimension of religion separately (Table 2). Each model in Table 2 also controlled age, race, gender, family income, parent’s education, and family structure. Although our intent in this paper is to examine overall relationships between religious factors and family formation outcomes, we did explore whether there were significant gender differences in our models to warrant separate analyses. Although a few scattered gender-by-religion interactions existed, the basic relationships appeared to be similar. Following the models estimating each dimension of religion separately, we explore the effect of various combinations of the three dimensions of religious beliefs and behavior for time to first union. Table 3 contains a series of dummy variables representing the 14 combinations of the affiliation, attendance, and fervor, as well as the set of control variables. Using 3-way interaction variables produced results similar to those using the 14 combination variables. The combination variables were ultimately more interpretable and so were favored over presenting the 3-way interactions. Finally, among those whose first transition was to cohabitation, we explored religious differences in the likelihood of the cohabitation ending in marriage and at what point in the cohabitation the marriage occurred (Table 4), and in attitudes about the acceptability of cohabitation without plans to marry (Table 5).

Table 3.

Proportional Hazards Ratios of the Relationship between Composite Religious Variables and First Union Formation.

Affiliation	Attendance	Fervor	Hazard Ratio of Union Formation		Cell Size	Percent in Union
			Marriage	Cohab.		Married	Cohabited
Nonea	N/A	N/A	.43***	1.17**	1787	6.79	51.08
Other Religiona	N/A	N/A	.68**	1.01	636	9.18	33.17
Conservative Protestanta	High	Low	.69*	1.23*	301	11.19	47.62
Conservative Protestanta	Low	High	.73*	1.14	559	14.07	49.89
Conservative Protestanta	Low	Low	.60***	1.25**	462	9.76	52.34
Mainline Protestanta	High	High	.70***	.93	1883	9.39	31.25
Mainline Protestanta	High	Low	.48***	1.16	314	6.30	40.39
Mainline Protestanta	Low	High	.65**	1.28***	554	8.02	42.85
Mainline Protestanta	Low	Low	.21***	1.26***	679	2.98	49.11
Catholica	High	High	.45***	.76***	1947	7.74	27.71
Catholica	High	Low	.26***	1.32**	352	4.27	46.96
Catholica	Low	High	.57***	1.00	611	10.69	38.26
Catholica	Low	Low	.35***	.97	695	5.90	39.72
N			13,895	13,895

^aComparison Group is Conservative Protestant with High Attendance and High Fervor (cell size = 3115, Percents = 14.92% Married, 34.17% Cohabited).

Note. Control variables are age, race, gender, total family income, parent’s (usually mother’s) education, respondent’s education, region of residence and family structure.

^*p< .05,

^**p < .01,

^***p < .001

Table 4.

Multinomial Logistic Odds of a First Cohabitation Ending in Marriage at Different Cohabitation Lengths by Composite Religious Variables

Affiliation	Attendance	Fervor	Marrying within 6 months of cohabiting b	Marrying after 7 – 18 months of cohabitationb	Marrying after 19 + months of cohabitationb	Continuing Cohabitationb
Other Religiona	N/A	N/A	1.43	.93	3.39***	1.05
Conservative Protestanta	High	High	1.45	1.90**	1.63	.89
Conservative Protestanta	High	Low	1.51	2.97***	2.20*	1.30
Conservative Protestanta	Low	High	1.75*	1.39	2.10*	1.25
Conservative Protestanta	Low	Low	1.00	1.39	1.34	.98
Mainline Protestanta	High	High	.96	2.51***	1.48	1.14
Mainline Protestanta	High	Low	.46	Not Calculated	2.34*	.73
Mainline Protestanta	Low	High	.63	1.02	3.00***	.76
Mainline Protestanta	Low	Low	.76	1.51	1.52*	1.15
Catholica	High	High	1.04	1.22	1.41	1.48***
Catholica	High	Low	1.19	1.90	1.97	1.35
Catholica	Low	High	1.32	.97	.94	.98
Catholica	Low	Low	1.04	.65	.94	.95
N = 5,199

^aComparison Group is No Affiliation.

^bRelative to Odds of Simply Breaking Up.

Note. Control variables are age, race, gender, total family income, parent’s (usually mother’s) education, respondent’s education, region of residence, and family structure.

^*p < .05,

^**p < .01,

^***p < .001

Results

Turning first to Table 2, we find, consistent with other work, that affiliation, attendance, and religious fervor all strongly related to making a transition to either marriage or cohabitation. As expected, the greater the attendance at worship services at Wave 1, the higher the hazard of marrying at every person-month interval following Wave 1. Attendance at worship services also reduced the hazard of cohabitation. We also found that the stronger the religious fervor, the higher the hazard ratio of marriage, and the lower the hazard ratio of cohabiting.

The effects of affiliation were somewhat unexpected, however. Not surprisingly, Conservative Protestants were significantly more likely to marry at every person-month interval that followed W1. However, contrary to what was expected, Catholics were significantly less likely to enter a marriage or a cohabiting union than Conservative Protestants. This difference is clear in the last column, which displays the percent in marriage or cohabiting for each affiliation group. Only 7.5% of Catholics had entered marriage and only 33.6% were in a cohabiting union by third interview Wave. In contrast, nearly 14% of Conservative Protestants had entered marriage, and over 40 %, surprisingly, had transitioned to a cohabiting union as a first union—a proportion that is the highest among all the groups except those of no affiliation.

The analyses in Table 3 help clarify the picture. This table displays the hazard ratio of transition to marriage or to cohabitation for 13 combinations of the three religion dimensions, relative to Conservative Protestants who are high attendees and have high fervor. Focusing first on the three groups of Conservative Protestants, we see that high attendance, high fervor or both had to be in play for identifying as a Conservative Protestant to matter for the transition to marriage. We also observed distinct differences in the hazard of cohabitation among Conservative Protestants, depending on their attendance and fervor. Relative to those who frequently attended services and evidenced high fervor, Conservative Protestants who were less faithful were significantly more likely to cohabit. We observe similar patterns for Mainline Protestants and Catholics. Members of these groups who were high attendees and who evidenced high fervor depart significantly from their less faithful brethren when it comes to cohabitation. Interestingly, these devout were significantly less likely than devout Conservative Protestants to be in any kind of union.

In sum, we found that affiliation, attendance, and fervor and their interplay were all predictive of first union transitions. Adolescents who identified themselves as Conservative Protestants who reported high attendance and high fervor were significantly more likely than other Conservative Protestant adolescents to transition to marriage and less likely to transition to cohabitation. This pattern of within group variability was somewhat evident for Mainline Protestant and Catholic youth. As predicted, when comparing across religious identities, Conservative Protestant youth, especially those who were devout, were by far the most likely to move into marriage as a first union transition in early adulthood. Similar across group patterns were also found for cohabitation.

One of the more remarkable findings was that significant fractions (approximately a third) of religiously inclined youth entered cohabiting unions. Among those who have cohabitated, does the effect of adolescent religious affiliation on cohabitation experiences vary by level of religiosity? Tables 4 and 5 address this question. Table 4 displays the relationship between the combinations of the three dimensions of religion and likelihood that that a first cohabitation ended in marriage at different points in the cohabitation. In order to fully evaluate whether religion influences cohabitation, we switched the reference group in this analysis to cohabiting individuals who had no religious affiliation in adolescence.

There was some evidence that religious beliefs and behavior in adolescence effects the likelihood of that the first cohabitation will end in marriage. Contrary to our first hypothesis, religious respondents were significantly more likely to end their first cohabitation in marriage relative to individuals with no religious affiliation. The first three columns of coefficients in Table 4, for example, show that a fair proportion of the individuals in the different combinations of Conservative and Mainline Protestants were more likely to end their first cohabitation in marriage at some point than were individuals with no affiliation. Cohabiting Catholic youth, however, showed little inclination (relative to the non-religious) to turn their cohabitations into marriages. Consistent with the second hypothesis, there are some differences among the religious groups. Both Conservative Protestant and Mainline Protestant cohabiting youth show a tendency to turn their cohabitations into marriages.

Table 5 provides additional evidence that adolescent religious experiences moderated subsequent patterns of union formation. Opinions about the meaning of cohabitation among those whose first transition was into cohabitation were strongly influenced by adolescent religiosity—especially if they were Conservative Protestants who were high attendees and evidenced high fervor. These individuals appear most weakly attached to the position that cohabitation without plans to marry is acceptable. That religion had been measured a number of years prior to assessing attitudes about cohabitation and that these individuals had already been in cohabiting unions before being asked about cohabitation shows the strength of these relationships.

Taken together, the patterns evident across these two tables suggest that religious experience in adolescence matters in subsequent cohabiting experiences. Furthermore, the fact that those respondents who were more devout as adolescents, especially those from Conservative Protestant backgrounds, were much more likely to be understanding that their cohabitation was meant to lead to marriage, strongly suggests that religious experiences in adolescence play an important role in the union formation process beyond the nature and timing of the first union.

Discussion

Young adults are engaging in union formation practices that are markedly different from those of their parents. The fundamental change has been the delay in age at marriage, and its replacement with cohabitation in the early 20’s. Much of the demographic literature has focused on socioeconomic predictors of this change. The purpose of this paper was to explore the extent to which religious factors might be a part of the explanation. We found evidence that religious identity, attendance at worship services, and the importance of religion all mattered for subsequent union formation. However, the interaction between these three domains provides additional insight. Consistent with our main hypothesis, we found that adolescents who were Conservative Protestants were distinctive from adolescents of other religious identities in the likelihood of subsequently marrying only to the extent that they also were high attendees and evidenced high religious fervor. The interplay between these three dimensions was also found to be important for understanding the likelihood of transitioning to cohabitation with a few exceptions. This work clearly shows, as some scholars have recently argued, the multidimensional nature of religion (Regnerus & Burdette, 2006). Individuals vary not only in religious identity or affiliation, but in their practice or attendance, and in the importance they attach to the religious part of their lives. Treating any one of these dimensions as a proxy for “religion”, as is commonly done, would appear to miss-specify the nature of the relationship between religion and behavioral outcomes. More importantly, our findings suggest that the effect of religion is best understood as the interaction between dimensions of faith, practice and belief.

Contrary to our expectations, we found that devout Conservative Protestants were not the least likely to cohabit. Catholic youth who were high on attendance and evidenced high fervor were found to be less likely than devout Conservative Protestant youth to subsequently enter a cohabiting union. In addition, Mainline Protestant youth who were devout were not significantly different that Conservative Protestants in the likelihood of cohabiting. In the light of the importance Conservative Protestant denominations have placed on sexuality and marriage, this finding comes as a surprise. Other evidence, however, suggests that this lack of distinctiveness on the part of Conservative Protestants probably should not be unexpected. Recent work by Regnerus (2007), for example, shows that Conservative Protestant youth are not avoiding sexual behavior. He finds that Evangelical Protestant adolescents were far from distinctive when it came to rates of nonvirginity, with rates that were higher than both Catholic and Mainline Protestant youth. Perhaps it is a short step from being sexually active as a teenager to cohabiting as a young adult. Further analyses of these data revealed that devout Conservative Protestant adolescents are not especially good at holding onto their faith in their young adult years. Only about 26% of Conservative Protestant adolescents who were high attendees and reported high fervor remained this devout five years later. This fraction is roughly similar to the declines among devout Mainline Protestants (29%), but significantly higher than that of devout Catholic adolescents (14%) (analyses not shown). Perhaps, as Regnerus (2007) has argued, we are overestimating the force of denominational identity as a cause of behavior. While religious identity is not irrelevant when it comes to family formation behavior--witness the substantial differences by identity when it comes to marriage-- the linkage between choices about forming intimate unions and religious identity, behavior and belief is more complex, subtle, and probably dynamic, than as often portrayed. Clearly, future work should move beyond using religious identity as a sole indicator of religiosity. However, understanding the role of religion in the lives of young adults will require a careful examination of the form and structure of changing religious beliefs and behavior over this developmental period.

The fact that religiously inclined youth are cohabiting raises questions about the nature of these unions. Our admittedly exploratory work focused on three competing hypotheses: (1) Adolescent religious experiences have no effect on subsequent cohabitation experiences; (2) Any religious identity in adolescence is associated with shorter cohabitation spells; or (3) Devout Conservative Protestants and Catholics are more likely to have shorter cohabitation spells that end in marriage. We found that cohabiting young adults who were devout Conservative Protestants or Mainline Protestants were more likely to turn their cohabitation into marriage than other youth. Furthermore, cohabiting youth who were devout Conservative Protestant adolescents were significantly more likely to subscribe to a “traditional” view of cohabitation—that it should be linked to plans to marry. The evidence here is not overwhelming, but it is consistent with our predictions about the “legacy” of adolescent religious beliefs and experiences. Previous works by Thornton and colleagues (1992; 2007) have highlighted the dynamic interplay between religious beliefs and cohabiting experiences. To wit, religiously inclined youth are less likely to cohabit, but once they cohabit, their religiosity declines. Our work shows, however, that the power of religion to influence behavior does not disappear completely.

There are some limitations to this study. The denominational detail in the first Wave of the Add-Health data is quite paltry, making subtle, but important distinctions among some denominations difficult. As a result, we have probably misclassified some Conservative Protestants as Mainline Protestant and visa versa. Others have pointed out that adolescents who answered “no religion” to the religious affiliation question were not asked the subsequent religious questions on service attendance or importance (Smith, Denton, Faris, & Regnerus, 2002). The implication is we have underestimated to some degree the religious experiences of adolescents. We also have not taken into account the dynamic nature of religious behavior and beliefs. There is good reason to believe that the transition to adulthood is marked not only by considerable experimentation with relationships, but also with religious beliefs and behavior. As youth move out from under the eyes of their parents, they are likely to “fall away” from the religious practices and beliefs of their parents. For some, religion may recede in importance in their lives, and have few implications for their union formation behavior. For others, young adulthood may be time of religious experimentation, of trying different religious identities or belief systems. There is some evidence of a growing interest in being “spiritual” among young adults, which is not necessarily tied to regular church attendance or identification with a particular denomination (Marler & Hardaway, 2002). The implications of religious experimentation on union formation deserve attention.

In conclusion, an often overlooked or unappreciated aspect of American adolescents is that religion has a significance presence in their lives. A sizeable fraction of teenagers attend religious services, identify themselves as members of communities of faith, and profess that their religious faith is important in their lives (Smith, 2005). Furthermore, there is an emerging body of evidence that adolescent religious beliefs and behavior are associated with a variety of important life outcomes for these teens. Religious beliefs and behavior have been implicated in the physical and emotional health of children and adolescents, educational aspirations and attainment, delinquency, criminal behavior, and sexual behavior (see Regnerus, Smith, & Fritsch (2003) for a review of this literature). Our findings build on this work. Religion not only has implications for the immediate lives of adolescents, but, as our work here clearly indicates, continues to echo through the course of young adulthood, shaping initial and subsequent family formation choices.