RELIGIOUS PATHWAYS DURING THE TRANSITION TO ADULTHOOD: A LIFE COURSE APPROACH A Dissertation Submitted to the Graduate School of the University of Notre Dame in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Jonathan P. Hill, M.A., B.A. David Sikkink, Director Graduate Program in Sociology Notre Dame, Indiana December 2007
181
Embed
Religious Pathways During the Transition to Adulthood: A Life Course Approach
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
RELIGIOUS PATHWAYS DURING THE TRANSITION TO ADULTHOOD:
A LIFE COURSE APPROACH
A Dissertation
Submitted to the Graduate School
of the University of Notre Dame
in Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
by
Jonathan P. Hill, M.A., B.A.
David Sikkink, Director
Graduate Program in Sociology
Notre Dame, Indiana
December 2007
RELIGIOUS PATHWAYS DURING THE TRANSITION TO ADULTHOOD:
A LIFE COURSE APPROACH
Abstract
by
Jonathan P. Hill
Empirical studies have identified late adolescence and early adulthood as a period
of the life course marked by relatively high levels of change in religious belief, practice,
and identity. Young people are most likely to decline in religious service attendance, but
they are also likely to disaffiliate, convert, or change particular religious beliefs during
this phase of the life course. Despite this, researchers have paid little attention to the
social sources of these changes with the exception of the study of family formation and
religious participation. This work in this dissertation begins to address this important
arena of religious change by establishing a general life course framework which
emphasizes the exogenous social forces that constrain and enable actors in their religious
worlds. Primary focus is given to two substantive areas: (1) the influence from religious
socialization and context in early adolescence on later pathways of religious participation,
and (2) the influence from higher education on religious participation, beliefs, and
affiliation.
Jonathan P. Hill
These research questions are primarily analyzed through panel data in the
National Longitudinal Study of Adolescent Health and the National Longitudinal Survey
of Youth 1997 surveys. Several key findings emerge. Religious participation in the form
of religious service attendance follows multiple pathways in late adolescence and early
adulthood. Decline in attendance is most common during adolescence, while the
proportion declining in their early twenties is approximately equal to the proportion
increasing in attendance. Further analysis reveals that religious traditions, household
religious socialization, and peer church attendance in early adolescence influence the
relative risk of decline versus stability during the transition to adulthood. Conversely,
demographic characteristics such as gender and race, along with residing in the South,
during early adolescence are key predictors of who increases religious attendance during
late adolescence and early adulthood.
Analysis of the influence of education attainment on religious practice, belief and
affiliation finds no overall decline in belief and affiliation as a result of higher education.
Further analysis reveals that college educated Catholics do not follow this general trend
and are more likely to have lower salience of faith and disaffiliate. Educated African
Americans, conversely show an increase in salience of faith and a lower likelihood of
disaffiliation. College type also matters with students attending Catholic and mainline
Protestant affiliated colleges declining in attendance more than students at other public
and private colleges and universities. A comparison with the birth cohort that attended
college during the late 1960s and early 1970s reveals that college had a stronger
secularizing effect in the past.
ii
For my wife, Cinthia.
iii
CONTENTS
Figures................................................................................................................................ vi
Acknowledgements.......................................................................................................... viii
Chapter 1: Introduction ....................................................................................................... 1 1.1 Overview........................................................................................................... 1 1.2 The Transition to Adulthood............................................................................. 3 1.3 Conceptualizing Religious Change in Social Context ...................................... 5 1.4 Methods of Inquiry ........................................................................................... 9 1.5 Organization of the Dissertation ..................................................................... 11
Chapter 2: Early Adolescent Influences on Religious Participation During the Transition to Adulthood........................................................................................ 15 2.1 Overview......................................................................................................... 15 2.2 Mapping Religious Service Attendance Trajectories ..................................... 16 2.3 The Growth Mixture Model............................................................................ 19 2.4 Data ................................................................................................................. 21 2.5 Analysis........................................................................................................... 21 2.6 Life Course Influences on Attendance Trajectories........................................ 26
2.6.1 Religious Traditions............................................................................... 27 2.6.2 Religious Socialization .......................................................................... 28 2.6.3 Ecological Contexts ............................................................................... 29 2.6.4 Social Resources and Environmental Stability ...................................... 30 2.6.5 The Mediating Influence of Life Course Events in Early Adulthood.... 31
2.7 Data ................................................................................................................. 32 2.8 Measures ......................................................................................................... 33
2.8.1 Measures of Religious Socialization...................................................... 34 2.8.2 Measures of Social Resources and Stability .......................................... 34 2.8.3 Measures of Ecological Context ............................................................ 36 2.8.4 Measures of Life Course Events in Young Adulthood.......................... 36
2.9 Plan of Analysis .............................................................................................. 39 2.10 Results........................................................................................................... 40 2.11 Discussion..................................................................................................... 50
2.11.1 Religious Tradition .............................................................................. 50 2.11.2 Religious Socialization ........................................................................52 2.11.3 Resources, Stability, and Context ........................................................ 54 2.11.4 Direct Versus Indirect Effects.............................................................. 56
Chapter 3: Higher Education and Change in Religious Identity and Practice.................. 59 3.1 Overview......................................................................................................... 59 3.2 Religion and College in Sociological Perspective .......................................... 61 3.3 Past Research .................................................................................................. 62 3.4 Differential Effects of College on Religion .................................................... 65
3.5 Data and Methods ........................................................................................... 71 3.5.1 Data ........................................................................................................ 71
3.9.1 Religious Service Attendance ................................................................ 77 3.9.2 Importance of Faith................................................................................ 80 3.9.3 Frequency of Prayer............................................................................... 83 3.9.4 Religious Disaffiliation.......................................................................... 85
Chapter 4: Religious Participation During College: Do Religious School Make a Difference?............................................................................................................ 93 4.1 Overview......................................................................................................... 93 4.2 Church Attendance and Higher Education ..................................................... 94 4.3 The Relationship Between Institutional Identity and Student Religious
Practice .......................................................................................................... 97 4.4 Differential Effects by Religious Tradition, Gender, and Race/Ethnicity ...... 99
4.7 Results........................................................................................................... 110 4.7.1 Effects from College on Service Attendance....................................... 110 4.7.2 Institutional Differences in Service Attendance .................................. 111 4.7.3 Conditional Effects .............................................................................. 120
4.8 Discussion..................................................................................................... 125 4.8.1 The Big Picture .................................................................................... 126 4.8.2 Religious service attendance at conservative Protestant schools......... 127
v
4.8.3 Religious Service Attendance at Catholic and Mainline Schools........ 129 4.8.4 Evangelical and Catholic Students....................................................... 130 4.8.5 Differences by Gender ......................................................................... 132 4.8.6 Differences by Race/Ethnicity ............................................................. 132
Chapter 5: Conclusion..................................................................................................... 135 5.1 Emerging Themes ......................................................................................... 135 5.2 A Quantitative Agenda for the Study of Religion Over the Life Course...... 138
Appendix A: College and Religiosity by Birth Cohort...................................................142 A.1 Overview...................................................................................................... 142 A.2 Religious Climate on Campus...................................................................... 142 A.3 Data .............................................................................................................. 145 A.4 Analysis I: Higher Education and Religious Outcomes in the GSS
1972-2004.................................................................................................... 145 A.5 Analysis #2: Higher Education and Religious Change in the YPSPS
1965-1973.................................................................................................... 149 A.6 A Comparison with Recent College Graduates ........................................... 151
Figure 2.1: The Unconditional Growth Mixture Model ................................................... 20
Figure 2.2: Ten Class Unconditional Growth Mixture Model Estimating Religious Service Attendance Between Ages 20 and 24 ...................................................... 23
Figure 2.3: Four Class Growth Mixture Model Estimating Religious Service Attendance Between the Ages of 16 and 20 ............................................................................ 24
Figure 2.4: Four Class Growth Mixture Model Estimating Religious Service Attendance Between the Ages of 20 and 24 ............................................................................ 25
Figure 4.1: Annual Church Attendance By College Type (Random Effects) ................ 112
Figure A.1: Weekly Religious Service Attendance Across Birth Cohort by Degree Completion.......................................................................................................... 147
Figure A.2: No Religious Affiliation Across Birth Cohort by Degree Completion ....... 147
Figure A.3: Bible “Book of Fables” Across Birth Cohort by Degree Completion......... 148
Figure A.4: Strong Religious Affiliation Across Birth Cohort by Degree Completion . 148
Figure A.5: Pray Daily Across Birth Cohort by Degree Completion............................. 149
Table 2.2 Frequency of Class Membership By Religious Tradition .............................. 41
Table 2.3 Likelihood of Declining Attendance Versus High Attending Stability Age 16-22........................................................................................................................... 43
Table 2.4 Likelihood of Increasing Attendance Versus Low Attending Stability Age 16-22........................................................................................................................... 48
Table 3.1 OLS Regression Predicting Frequency of Church Attendance Age 22-25 ...... 78
Table 3.2 OLS Regression Predicting Importance of Faith Age 22-25............................ 81
Table 3.3 OLS Regression Predicting Frequency of Prayer Age 22-25 ........................... 84
Table 3.4 Logistic Regression Predicting No Religious Affiliation age 22-25 ................ 86
Table 4.2 Random and Fixed Effects Predicting Annual Religious Service Attendance111
Table 4.3 Fixed Effects Regression Predicting Annual Religious Service Attendance . 114
Table 4.4 Fixed Effects Regression Predicting Truncated Annual Religious Service Attendance .......................................................................................................... 118
Table 4.5 Fixed Effects Regression Predicting Truncated Attendance By Religious Tradition, Gender, and Race/Ethnicity ............................................................... 121
Table A.1 The Effect of Degree Completion on Religious Attendance, Affiliation, and Belief about the Bible 1965-1973 ............................................................... 150
viii
ACKNOWLEDGEMENTS
This research uses data from Add Health, a program project designed by J.
Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by a grant P01-
HD31921 from the National Institute of Child Health and Human Development, with
cooperative funding from 17 other agencies. Special acknowledgment is due Ronald R.
Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in
obtaining data files from Add Health should contact Add Health, Carolina Population
Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524
(www.cpc.unc.edu/addhealth/contract.html).
1
CHAPTER 1:
INTRODUCTION
1.1 Overview
Commentators of American religion have frequently focused on the distinction
between religion as it exists in doctrine, liturgy, and historical traditions, and religion as it
lived out in everyday experiences (Roof and McKinney 1987; Wolfe 2005). While the
religious lives of ordinary citizens draw from doctrine and tradition, they do by being
filtered through the lens of broader cultural motifs, individual social roles, and social
structural position. This dialectic between ideas about ultimate reality and the concrete
social realities that constitute individual experience provides a nexus of rich exploration
for the social scientist. This circumscribed area of study is one of the key distinguishing
marks between the sociologist of religion and the theologian. While the theologian is
concerned with the actual attributes of ultimate reality, the sociologist describes how this
reality is perceived among the population and the social consequences of these
perceptions, saying nothing of their truth or falsehood.
Because of this focus on lived religion, sociologists find that religious knowledge,
beliefs, and practices are constantly in flux as society changes. Individuals traverse the
many relationships, institutions, and cultural systems that situate ideas about religion and
its place within the public and private spheres of American society. Religious ideas and
rituals are not determined by these social characteristics, but neither are they entirely free
to develop without them. Describing how religion is lived and understood within the
2
social cleavages of race/ethnicity (e.g. Emerson and Smith 2001; Hammond and Warner
1993; Lincoln and Mamiya 1990; Patillo-McCoy 1998), gender (e.g. De Vaus and
McAllister 1987; Miller and Stark 2002; Sullins 2006), and social class (e.g. Niebuhr
1929; Roof 1979; Park and Reimer 2002) has occupied much of the sociology of religion.
Sociologists have also focused on how religion interacts with institutions such as
education (e.g. Peshkin 1988; Sikkink 1999), the family (e.g. Edgell 2005; Wilcox 2004),
and the state/politics (e.g. McVeigh and Sikkink 2001; Sherkat 1998), as well as cultural
influences from mass media (e.g. Hoover 2006), therapeutic individualism (e.g. Bellah et
al 1985), and sexual liberation (e.g. Regnerus 2007).
There has been far less attention given to intra-individual changes in religious
practices, beliefs, and identity within the sociology of religion. Certain phases of the life
course such as adolescence have received renewed attention in recent years (Smith and
Denton 2005). There is also a small literature on aging (Bahr 1970; Krause 2006) and a
slightly larger literature on religion and the family lifecycle (Myers 1996; Stolzenberg,
Blair-Loy, and Waite 1995). Nevertheless, this leaves a large swath of intra-individual
religious change unexplored. Using a similar framework of interaction between religious
ideas and social life, one can generate hypotheses concerning changes in religious
meaning, and how beliefs and practices are altered, as individuals progress through the
social arrangements that constitute the life course. Key socializing institutions such as
the family and education take on different meaning and arrangements as individuals age.
In adolescence, peer networks become highly salient. Disposable income and the legal
ability to drive an automobile rapidly increase the slate of choices available to the young
person. Having a successful careers and the establishing a family of procreation remain
3
key goals in early adulthood. Death of parents, relationship with grown children and
continued economic gain are important social aspects of later adulthood. And finally,
impacts of physical health, death of a spouse, and institutions of assisted living and
nursing homes are principle features of late life. As the social world of the individual
changes with age, we have little understanding of the impact from these changes on
religious understandings and practices.
The most rapid changes in religious practice and belief accompany the period of
late adolescence and early adulthood. During this phase young people are more likely to
decline in attendance (Smith et al 2002; Uecker, Regnerus, and Vaaler 2007; Willits and
Crider 1989), disaffiliate (Caplovitz 1977; Sherkat and Wilson 1995; Wilson and Sherkat
1994), convert (Smith 2006), and alter other religious beliefs and practices (Uecker,
Regnerus, and Vaaler 2007). Despite popular interest in the subject (particularly from
religious advocates) and scholarly interest from other fields (such as developmental
theories in psychology), the subject has received little attention in the field of sociology
outside of a literature on religion and family formation. The documented religious
changes that occur for many young people, coupled with the rapidly changing social
world of the emerging adult, make this a ripe area for sociological analysis.
1.2 The Transition to Adulthood
Although the concept of transitioning from youth to adulthood might appear
relatively uncomplicated, the task of identifying the core indicators of this transition in
the social scientific literature remains somewhat contentious. Within the context of the
United States, researchers have traditionally used markers such as leaving the family of
4
origin, completing education, starting a full-time job, getting married, and having
children as key moments that move adolescents into adult roles (Shanahan 2000). The
ordering and spacing of these events have changed over time with recent birth cohorts
exhibiting less normative ordering and more spacing between these traditional
demographic markers (Buchman 1989; Rindfuss, Swicegood, and Rosenfeld 1987).
Although some expectation concerning the ordering and timing still exists, the normative
restraints have been loosened and off-timed transitions hold far less stigma than in the
past (Settersten 2003).
Because of these changes, some scholars have argued that demographic
transitions matter less than subjective understandings of the self as mature and
independent (Arnett 1998; Barry and Nelson 2005). Arnett (2004) has coined the term
“emerging adulthood” as the period of the life course following adolescence, but prior to
full adulthood, which is characterized by few normative expectations and relative
independence from social responsibilities. The appearance of this new phase in the life
course is made possible by the movement toward later family formation as well as
economic independence (Arnett 1998; Rindfuss 1991). During this period between
adolescence and adulthood, much of the identity work involved in becoming an adult is
undertaken. This perspective holds that the criteria for adulthood are found in subjective
understandings of independence from parents and accepting responsibility for one’s
actions as opposed to demographic transitions.
I adopt a perspective that affirms the importance of both the subjective
understanding of the self as well as the importance of the social arrangements and role
transitions that characterize the transition to adulthood. There is recent empirical
5
research that suggests these two perspectives may both be important and mutually
reinforcing (Shanahan et al 2005). While I do not regularly use the term emerging
adulthood in this dissertation, I do not object to the concept entirely. Rather, I choose to
focus on how the self is situated within social structures and institutions that both
constrain and enable certain identities and actions. Although much of this dissertation
uses crude survey measures of religious belonging and practice to assess intra-individual
change, it is still important to have an understanding of the underlying social processes
that guide these changes. I outline this perspective below.
1.3 Conceptualizing Religious Change in Social Context
Within the human sciences, theories of intra-individual religious change have
been most dominant in psychology (Pascarella and Terenzini 1991). This is not
necessarily due to a large interest in religion and spirituality among psychologists as
opposed to sociologists. Rather, psychology takes as its primary unit of analysis the
individual. Along with the frame of reference grounded in the individual, psychology has
had a long history of the study of individual development rooted in the biological
metaphor or growth and maturation (Baltes, Staudinger, and Lindenberger 1999). It is
through this lens of the structural developmental paradigm that theories of intra-
individual religious change have garnered the most attention.
Drawing heavily on both Piaget (1954) and Kohlberg (1969; 1981), psychologist
James Fowler (1981) has applied stage theory specifically to the development of faith.
Fowler hypothesizes that individuals progress through increasingly complex and
differentiated understandings of “ultimate reality” as they age. The final destination of an
6
individual varies, with some only progressing through the most elementary stages of
development. Fowler’s theory is meant to be universal in its applicability and valid
regardless of the specific content of one’s faith tradition. This universalistic claim, both
its cross-cultural and cross-faith dimensions, has come under serious criticism in recent
years (Day 2001; Streib 2001). Moreover, empirical work has questioned the validity of
claims to universal dimensions of faith that can be separated from the particularities of
any faith tradition (Leak et al. 1999).
Developmental stage theories have come under attack both within psychology
(Baltes et al. 1980; Baltes et al. 1997) and from sociology (Dannefer 1984). While the
critiques are varied, some key problems seem worth noting. First, the biological
metaphor of growth and maturation may be inappropriate, especially when the object of
interest (such as religion or faith) is distinct from physiological development. While
individuals develop biologically based changes in cognitive and sexual function as they
age, it is unwise to couple this tightly with cultural systems pertaining to super-empirical
reality. As humans develop complex, symbol-laden, socially-constructed systems of
ideas and beliefs, these ideas and belief contain emergent properties that cannot be simply
reduced to biological functioning. If our primary goal is to understand how humans
interact with these systems as they age, shift life roles, and transgress institutional
boundaries, theories that take their cue from biological change will necessarily be limited
in their usefulness.
It is also arguably beyond the purview of empirical science to distinguish
between qualitatively higher and lower understandings of the individual’s relationship to
this reality. By adopting these lower and higher stages, the researcher confines an
7
individual’s belief and value trajectory toward a pre-defined end. Moreover, when the
primary attention is given to changes within the individual, contextual influences are
usually an afterthought. Dannefer (1984) has argued that ontogenetic formulations, such
as those found in developmental stage theories, are inherently antithetical to the notion
that the phenomenological reality for the individual is socially constructed.
To integrate social reality back into the study of intra-individual religious change,
I adopt the language and methods of the life course. Even though no truly unified
schematic of the life course exists (George 1993), the work of Glen Elder (1994, 1998)
provides the clearest conceptualization of the life course paradigm from the North
American perspective (Marshall and Mueller 2003). In this perspective, the life course
consists of multiple age-graded trajectories (e.g. economic, familial, religious, and
educational) that are acted upon by exogenous social forces, and in turn act upon one
another. Within this framework the importance of historical contingency, interconnected
lives, and human agency are all emphasized. This method, which probably best defines
life course research, begins with social forces that act upon the individual and traces the
mechanisms by which differential life outcomes are constrained and enabled by these
forces (Elder 1996).
This focus on exogenous social forces provides a corrective to many of the
problems associated with developmental stage theories. Taking the social reality as a
starting point and hypothesizing how individuals alter their religious identities, beliefs,
and practices within this social reality effectively combats the reduction of the social to
the individual in developmental theories. The social world is not something to be
8
explained away or ignored from this perspective – it is fundamental to the distinctive
forms that intraindividual change takes.
At the same time, basing the method of inquiry in social structures runs the risk of
appearing fragmented, as one must choose to make salient certain aspects of the
individuals’ social worlds while minimizing or ignoring other aspects. Empirical
relationships are established, patterns emerge, but the bigger picture often remains
elusive. Much of this is the result of the complex nature of the social world that
developmental theories often ignore or oversimplify. While individuals are highly
complex, they still remain integrated into a single self which acts and is acted upon. If
the self is the frame of reference, as it is in stage theories, hypothesized changes can be
neatly packaged as matters of individual agency. This does not mean that the processes
hypothesized in developmental stage theories are somehow oversimplified. They may
actually be quite complex. However, it does mean that by virtue of the individual as
referent, theories of intra-individual change will necessarily appear more holistic than
theories that focus on the self contained within the social.
Social reality, conversely, is diffuse and has no immediate referent. In order to
gain theoretical leverage on the social world, sociologists use multiple conceptual
categories (e.g. symbols, culture, structure, institution, capital, and fields). These
categories can be thought of as useful heuristics; ways of obtaining analytic leverage
from social life. Yet, these concepts are not simply dimensions of a single concept (i.e.
“the social”). Rather, these concepts are often qualitatively distinctive from one another
with the underlying logic of each drawing from different metrics. This essential non-
9
confluence makes a singular understanding of the self as contained in social context
difficult to pin down.
Nevertheless, to the extent that the social is handled non-reductively, I believe
that our understanding of the individual will be more realistic. In this regard, the life
course method, with its loosely connected tools and concepts, is flexible enough to gain
insight from the complex social webs that individuals are embedded in.
1.4 Methods of Inquiry
This dissertation uses nationally-representative panel data to establish the broad
contours of religious life during the transition to adulthood. Data from the National
Longitudinal Study of Youth 1997, the National Longitudinal Study of Adolescent
Health, and the Youth-Parent Socialization Panel Study are used to map intra-individual
patterns of change from adolescence into young adulthood.
There are several advantages and disadvantages to using survey data to
understand changes in religion over time. If we establish that any picture of reality the
social sciences are able to provide is necessarily limited in scope, it is useful to locate
what the inherent boundaries are in a given research method. At the same time, a given
method is likely to have a particular analytic traction that distinguishes it from other
methods. By identifying the strengths and limitations of statistical analysis with panel
data, I hope to establish what sort of claims the work in this dissertation can validly
establish.
Carefully collected survey data is an excellent tool to map the general landscape
of social phenomena. The proper collection of data insures that we are able to use
10
statistical tools to make inference about the population with some degree of certainty.
Moreover, survey questions are worded and administered in a consistent way that allows
for a high degree of reliability in the constructs being measured. These properties allow
us to confidently construct a picture of the major contours of social reality. However,
many of the characteristics that allow for this confidence in measurement and inference
are the same ones that limit what kind of constructs can be measured.
Underneath these broad contours are complex webs of human identity, social
structure, and cultural systems that are difficult to measure with the blunt instrument of
survey data. Human beings generally understand themselves and their motivations in
terms of narrative (MacIntyre 1981; Smith 2003; Taylor 1989). Survey data, however,
pulls apart narrative coherence into usable chunks of data that can be manipulated via
statistical techniques. The loss of narrative coherence in the data means that the
quantitative research must carefully place a plausible narrative back into the explanation
of the data after it is analyzed. This is a delicate process and often results in necessary
simplifications. Nevertheless, charting the broad contours is an important first step in
understanding social phenomena because it supplies the broad context under which more
in depth study of motivations and meaning can occur.
The methods of this dissertation, then, focus on uncovering some of the major
contours of religious life during late adolescence and early adulthood. Much of what is
described, analyzed, and tested in the chapters to come should be very generative for
further research that more narrowly focuses on the specific motivations, structures, and
cultural tools that enable and constrain social actors and their religious worlds during
these decisive years of the life course. This does not mean that the chapters to come are
11
silent concerning these matters. These processes which underlie the empirical findings
are addressed in the theoretical foundations of each chapter as well as the post hoc
interpretation of the analyses. It is my hope, however, that the research presented in this
dissertation provides part of the larger framework for continued work on the religious
lives of young people.
1.5 Organization of the Dissertation
I have written this dissertation in the form of three scholarly articles with the
present introduction and a concluding chapter. Due to its organization as scholarly
articles, there remains some discontinuity as well as overlap between chapters, as each of
the three substantive chapters stand on their own and do not assume the reader has read
the other chapters. Furthermore, the chapters are not designed to provide a complete
overview of the many markers and transitions during the movement from adolescence to
adulthood. Rather, the chapters focus on narrowly defined areas such as religious
influences from the family of origin and the effects of higher education. The influence
from family formation on religion is not examined as there is already a robust literature
available on the subject. Part of the function of the concluding chapter, then, is to draw
together emerging findings and themes within the chapters, as well as to situate the work
done in this dissertation within the larger project of using panel data to map religious
practice and belief from a life course perspective.
Chapter two examines trajectories of religious service attendance between ages
sixteen and twenty-four using growth mixture modeling with the National Longitudinal
Study of Youth 1997. Using five waves of data from 2000 to 2004, I identify broad
12
patterns of stability, growth and decline during these ages. Once the patterns are
established, I identify key influences from early adolescence that constrain the later
patterns of religious worship attendance. More specifically, I measure the influences
from early adolescent parental, peer, and school religious socialization on the probability
of moving from a stable pattern of attendance to a pattern of growth or decline during the
transition to adulthood. I also test hypotheses that predict religious stability during the
transition to adulthood to depend upon local contexts defined by religious markets, class
structures, and racial distributions. The influence of home and neighborhood stability on
later religious trajectories is also explored in this chapter.
Chapters three and four focus on higher education and religious identity and
practice. As higher education expands, attending college has become more normative in
the transition to adulthood. In 2005, approximately sixty-nine percent of high school
graduates enrolled in a two or four year college in the fall semester immediately
following their graduation, compared to less than fifty percent in 1980 (U.S. Department
of Education, National Center for Education Statistics 2007). Higher education has also
been widely assumed to undermine traditional sources of moral authority and contribute
to individual level secularization (Caplovitz and Sherrow 1979; Hunter 1983; Wuthnow
1988). Despite all of this, there has been very little recent research focusing that focuses
on the influences from college on religion.
Chapter three compares students who attend and graduate from four year colleges
and universities to young people who do not continue their education after high school.
In order to examine if change in religious identity and practice change depending on
education attainment, I study change in indicators of religious service attendance,
13
importance of faith, frequency of prayer, and religious affiliation over time in the
National Study of Adolescent Health. This chapter uses indicators that were collected
during the students junior and senior year of high school (in 1994 and 1995) and the same
or similar measures six or seven years later (in 2001 and 2002). Examining differences
by educational trajectories allows inference concerning the effects of higher education on
religious change. This chapter also examines differences by subpopulations defined by
religious tradition, race, gender, and college major. Appendix A uses data from the
cumulative General Social Survey (1972-2004) and the Youth-Parent Socialization Panel
Study (1965-1973) to compare the findings from Chapter 3 to earlier birth cohorts.
Chapter four focuses on institutional effects on religious change while students
are enrolled in college. While the previous chapter attempts to quantify the differences in
religious change between college graduates and non-college graduates, this chapter
focuses on the differences between currently enrolled students’ religious change based on
college characteristics (e.g. religious tradition, elite status, and student body
characteristics). The key research question addressed in this chapter concerns how well
students fare religiously at religious schools versus secular schools. Data come from the
National Longitudinal Survey of Youth 1997. Similar to the analysis undertaken in
Chapter 3, I examine differential effects by subpopulations characterized by religious
tradition, race/ethnicity, and gender.
Chapter five concludes the dissertation by drawing together several key themes
across the three substantive chapters and integrates these findings with extant research on
family formation to provide a broader picture of religious practice and identity during the
transition to adulthood. I then turn to an agenda for future research on religion during
14
adolescence and early adulthood, suggesting some fruitful areas of inquiry. Lastly, this
chapter outlines further areas of life course studies and religion outside of the transition to
adulthood that would benefit from broad quantitative analysis using panel data.
15
CHAPTER 2:
EARLY ADOLESCENT INFLUENCES ON RELIGIOUS PARTICIPATION DURING
THE TRANSITION TO ADULTHOOD
2.1 Overview
Religious belief and practice in late adolescence and early adulthood arguably
constitute the most critical stage in the religious life course. During this phase young
people are more likely to decline in attendance (Smith et al 2002; Uecker, Regnerus, and
Vaaler 2007; Willits and Crider 1989), disaffiliate (Caplovitz 1977; Sherkat and Wilson
1995; Wilson and Sherkat 1994), convert (Smith 2006), and alter other religious beliefs
and practices (Uecker, Regnerus, and Vaaler 2007). Moreover, religious beliefs and
behaviors during this juncture continue to predict later levels of religiosity in adulthood
(Myers 1996; Stolzenberg, Blair-Loy, Waite 1995). Nevertheless, most research has
focused on just one or two years during this period of the life course, or focused
exclusively on events such as declining attendance, disaffiliation, or conversion. An
overall map of religious trajectories during this phase of the life course has not been
adequately specified. Additionally, most studies have focused exclusively on adolescent
religion (Smith and Denton 2005) or the religious life of young adults (Wuthnow 2007).
Because of this, an understanding the continuities and discontinuities in religious life
during the transition to adulthood is not well understood.
16
This chapter begins to address these shortcomings by focusing on the patterns of
religious participation during the transition to adulthood. Drawing on a life course
perspective, this paper argues that (1) prior research analyzing religious practice has not
adequately mapped the overall trajectories of religious service attendance during this
period of the life course. And, (2) religious traditions, religious socialization, and key
dimensions of social context during early adolescence increase the likelihood of traveling
particular trajectories of religious participation during the transition to adulthood.
This format of this chapter is divided into two sections. In the first section, I use a
growth mixture modeling method with panel data to estimate trajectories of service
attendance using five consecutive years of attendance data for respondents from ages
sixteen to twenty-four. In the second section, I model early adolescent influences on
which trajectory is traveled during late adolescence and early adulthood. Specifically, I
employ multinomial logit models to test whether key characteristics of families, peers,
and social context during early adolescence increase the relative risk of declining and
increasing religious participation during the late teenage years and early twenties.
2.2 Mapping Religious Service Attendance Trajectories
Emerging research on religiosity during adolescence and young adulthood finds
that participation in religious services declines at a substantially steeper rate than other
forms of private religious practice and belief (Uecker and Regnerus 2007; Smith et al
2002, Wallace et al 2003; Willits and Crider 1989; King, Elder, and Whitbeck 1997).
The Monitoring the Future survey estimates that the proportion of students who attend
weekly drops from 43% to 33% between the 8th and 12th grade. Uecker, Regnerus, and
17
Vaaler estimate that as much as 69% of adolescents who attend religious services at least
once a month attend less frequently in young adulthood. Likewise, Willits and Crider
calculate that 60% of youth report attending less frequently in early adulthood compared
to their adolescent years. Overall, it is clear that a decline in religious participation is a
common occurrence as individuals enter late adolescence and early adulthood.
Nevertheless, several key problems plague current estimates of religious service
attendance for young people. First, the most common multivariate techniques are geared
toward the analysis of mean levels of change only. However, analyzing mean level
change often masks patterns of decline and growth. Ozorak (1989) has suggested that
religiosity during adolescence may become more polarized. Likewise, Johnson (1997)
has proposed that religious belief and practice may become polarized during college.
Neither hypothesize is testable using methods that only assess mean levels of change over
time. Moreover, mean trends may or may not be “typical”; there is simply no way of
knowing without using different methodology. For example, although we know from the
Monitoring the Future survey that the percentage of students who attend religious
services weekly drops ten percentage points between 8th and 12th grade, we do not know
if this is because approximately ten percent of the student body moved from a state of
weekly attendance to one of less than weekly attendance. This finding could also be due
to forty percent of the student population declining from weekly attendance and thirty
percent beginning to attend weekly. These two hypothetical scenarios paint very
different pictures of religious participation during adolescence.
Studies that use panel data have the advantage of being able to estimate individual
level change over time. Using panel data allows the researcher to distinguish between the
18
two hypothetical explanations for the ten percent decline in religious participation in
adolescence. Even so, typical estimates using panel data carry some distinct
shortcomings. Most research has used two waves of data to characterize change religious
participation over time. Such methods, however, are sensitive to the length of time
between waves. For example, Uecker, Regnerus, and Vaaler (2007) find that 69% of
adolescents who attend monthly have lower levels of religious participation
approximately six or seven years later (between Waves I and III). However, Regnerus
and Uecker (2006) find that only 6% of adolescents report a substantial decrease in
religious service attendance (defined as decreasing at least two categories in the measure)
over a period of approximately one year (between Waves I and II). One would suspect
that they would have found considerably less occurrences of substantial religious decline
if the period between interviews was one month as opposed to one year. Both findings
are accurate pictures of religious change over the age span measured, but both only
represent a fraction of the total picture.
In addition to the problem of length of time, estimating the degree of change is
also a complicated matter when using only two waves of panel data. While we may
know that a certain percent decline in religious attendance between certain ages, this does
not tell us anything about the degree of decline. If a substantial proportion of young
people stop attending services altogether, this is a very different matter than if young
people simply drop slightly in their frequency of attendance with age. Of course,
examining mean change is one way to get a sense of the degree of change – but the
shortcomings of such a strategy have already been established. Alternatively, the
researcher could simply create a table of frequencies for every possible combination of
19
change or stability between waves. Of course, with a variable measured with eight
categories (such as the measure of church attendance used in this chapter), this would
require having thirty-six possible trajectories. Such a method probably needlessly
complicates the picture and makes further multivariate analysis using these trajectories
difficult.
2.3 The Growth Mixture Model
When more than two waves of data are available, more powerful methods can be
used to model and group individual trajectories. The present research uses growth
mixture modeling (GMM) to uncover the dominant trajectories of attendance for
adolescents and young adults. The GMM method is able to estimate the dominant
trajectories of religious practice during late adolescence and early adulthood. GMM
combines the power of latent growth curve analysis and latent class analysis. Unlike
standard growth curve analysis which makes the assumption that individual trajectories
deviate from one grand growth curve, growth mixture modeling assumes that individuals
deviate from multiple normative trajectories (or alternatively, that we are sampling from
distinct subpopulations that have qualitatively different growth patterns). These different
classes of trajectories are a latent function of the data, so no a priori assumptions
concerning how individuals change religiously are necessary.
Figure 2.1 presents the basic structure of an unconditional GMM model. The
general model is similar to a latent growth curve model, where the observable variables
measured over time are assumed to be a function of unobservable intercept and slope
terms. The addition in a GMM is a latent class variable, “c”. The latent class variable is
20
a nominal level variable that identifies the probabilistic membership in a set of
subpopulations. The intercept and slope terms are fixed across categories of “c”.
Random, individual-level, deviations from the intercept and slope terms are still allowed
in a growth mixture model. For additional information concerning GMM, please consult
Muthén (2004)
Figure 2.1 The Unconditional Growth Mixture Model
The GMM method fits the life course framework better than a standard latent
growth curve model. From the life course perspective, individuals follow trajectories that
are socially patterned by time, place, and social networks. This patterning generates
recognizable normative pathways of change at the aggregate level (Elder 1994). From
the standpoint of the individual, trajectories of change may take on innumerable forms.
However, from the life course perspective we recognize the social structuring of these
trajectories and the reality that normative pathways develop as social worlds constrain
Y1 Y2 Y3 Y4 Y5
i s q
c
21
individual action. By estimating multiple, distinctive trajectories of change at the
aggregate level, the GMM framework better captures these paradigmatic features of the
life course framework.
2.4 Data
Data come from the National Longitudinal Survey of Youth 1997 (NLSY97).
The NLSY97 is a nationally representative panel survey of youth, sponsored and directed
by the U.S. Bureau of Labor Statistics and conducted by the National Opinion Research
Center at the University of Chicago, with assistance from the Center for Human Resource
Research at The Ohio State University (Bureau of Labor Statistics, U.S. Department of
Labor 2006). The survey was first administered to 12 to 16 year old youth in 1997 and
has followed up annually with the most recent available wave in 2004. An additional
supplemental sample of Black and Hispanic youths was also conducted. The full number
of respondents from the representative sample and the over sample is 8,984.
Approximately 84% of the original respondents have been retained in the 2004 survey.
In the present research, data are primarily used from the base 1997 survey as well as
surveys conducted between 2000 and 2004. Supplementary geocode data have been used
to identify the respondents’ county of residence.
2.5 Analysis
Annual measures of religious service attendance are measured on an eight point
scale ranging from zero (never attend) to seven (attend everyday). In order to estimate
how trajectory membership varies with age, I analyze two subsets of the full NLSY97
22
sample. The first subset is comprised of respondents who age from 16 to 20 between
2000 and 2004 (N = 1,730). The second subset is comprised of respondents who age
from 20 to 24 between 2000 and 2004 (N = 1,605).
Class membership is estimated for each sample subset. There is not a set rule for
determining the correct number of latent classes in GMM. The Bayesian information
criterion (BIC) or adjusted BIC are commonly used. Other special statistics that rely on
bootstrapping methods are often used as well - the most reliable being the Bootstrap
Likelihood Ratio Test (see Nyland, Asparouhov, and Muthén 2006). However, it is also
important to use prior knowledge and side information to best determine the number of
latent trajectories in the data.
Using the BIC statistic and BLRT, as many as ten latent classes can be estimated
from the data. Figure 2.2 displays a GMM with ten latent trajectories estimated between
the ages of 20 and 24. However, this large number of latent trajectories proves unwieldy
for further analysis and is unnecessarily refined. The majority of estimated trajectories
can be classified in broader terms as moving from high attending to low attending or vice
versa. Moreover, class membership is very low for several of the trajectories. In order to
simplify the model, but still retain critical information, I estimate a four class GMM for
each subset of the sample.
Figure 2.3 and Figure 2.4 present the four class model for the two age groups.
Several important findings emerge. First, a majority of individuals have fairly stable
levels of church attendance in both age groups, although the percent of individuals who
are in a low attending state significantly out number those in the high attending
23
Figure 2.2 Ten Class Unconditional Growth Mixture Model Estimating Religious Service Attendance Between Ages 20 and 24
trajectory. Only 13.9% of the population age 20 to 24 can be classified as attending
approximately weekly over the entire five years measured. Secondly, approximately
twice the proportion of respondents can be classified as declining in attendance between
the ages of 16 and 20 compared to 20 and 24 (21.4% versus 12.1%). Focusing on just the
first age group, the mean level of decline is also slightly steeper between ages 16 and 18
than it is between ages 18 and 20. Lastly, growth in attendance is more evident between
the ages of 20 and 24 compared to ages 16 to 20 (12.7% versus 8.6%).
Some key implications of these findings emerge. For one, it appears that for the
recent birth cohort measured in the NLSY97, leaving a religious community has little to
24
Figure 2.3 Four Class Growth Mixture Model Estimating Religious Service Attendance Between the Ages of 16 and 20
do with leaving home and attending college. If this were the case, we would probably see
a substantial decline after age 18. However, this data suggests that the steepest decline is
actually between age 16 and 18. And, by the early twenties, patterns of decline are as
likely as patterns of increasing attendance. It is important to note, however, that the
reasons for this decline, while adolescents are presumable still residing with their parents,
are not clear. It still may be, as Hoge, Johnson, Luidens (1994) have suggested, that the
encounter with liberal values that are different from that of their household of origin are
detrimental to religious participation. Adolescents who are capable of driving and
working may be already be sufficiently cut off from the adult world that leaving home
and going to college are not drastic shifts in experience (see Smith and Denton 2005:184
Weekly
Once or Twice
18.7%
21.4%
8.6%
51.3%
25
Figure 2.4 Four Class Growth Mixture Model Estimating Religious Service Attendance Between the Ages of 20 and 24
for discussion of structural disconnect from the adult world). On the other hand, it may
also be that the “cultural broadening” associated with leaving home is non-existence for
most youth. Rather, the mundane tasks of negotiating work, school, and personal
relationships may what dominant the lifeworld of adolescents and young adults
(Clydesdale 2007).
It is also important to note the increasing attendance trajectory is not a common
experience, but neither is it rare. As the data suggest, increasing attendance is more
common during the early twenties than it is during late adolescence. While family
formation may be partially responsible for this increase in the early twenties, there is
cursory evidence to suggest that graduating from college may also contribute to this rise
in religious participation (see Table 4.2 for evidence of this influence). Regardless of the
Weekly
Once or Twice
13.9%
12.1%
12.7%
61.3%
26
social sources of this increase, the overall focus on religious decline during young
adulthood conceals the fact that approximately one in eight young adults are becoming
more active in religious communities. Further research on this group of emerging adults
seems warranted.
2.6 Life Course Influences on Attendance Trajectories
Although a majority of young people maintain stable levels of church attendance
during the transition to adulthood, substantial minorities display movement into and out
of religious communities. One of the fundamental emphases of the life course
perspective is individual continuity over time. Life experiences are cumulative and past
states shape future trajectories (McLeod and Almazan 2004). This is arguably the
primary contribution of the life course paradigm, which attempts to combat the standard
cross-sectional snapshots of population so common in quantitative research.
From this perspective, we would expect that the social characteristics of the
young person’s environment in early adolescence constrain future action. Moreover,
from their social location individuals would be expected to develop orientations and
strategies of actions that make future pathways in the life course more or less probable.
Drawing from the research on adolescent religiosity and the intergenerational
transmission of religiosity, I outline several key expectations concerning the influence
from religious traditions, religious socialization, social stability, and neighborhood
context in early adolescence on later pathways of religious participation. I also explore
whether these influences have a direct effect on later pathways, or whether they primarily
27
function through altering the probability of key life course events in early adulthood such
as educational attainment and family formation.
2.6.1 Religious Traditions
Religious traditions vary in their ability to attract and retain members.
Conservative Protestants and other “strict” churches (Iannaccone 1994) are able to retain
their members into adulthood at a greater rate than either Catholicism or mainline
Protestantism (Sherkat 2001; Smith 1998). Individuals are also less likely to switch out
of conservative Protestantism than other Christian religious traditions, and this likelihood
has decreased over time. In fact, most of the proportional increase in conservative
Protestantism over the last several decades is due to increasingly high retention rates and
relatively greater levels of fertility (Hout, Greeley, and Wilde 2001).
There are many possible reasons why conservative Protestants are better able to
retain their members over time. Using economic theory, Iannaccone (1994) suggests that
the strict rules that often times accompany congregational life in conservative Protestant
churches effectively screen out free riders and generate more efficient levels of collective
religious production. Smith (1998) argues that evangelicals have historically maintained
a tension between being engaged with the culture and distinct from it at the same time.
This combination of engagement and distinction generates subcultural solidarity that
makes evangelicalism thrive. Ammerman (1987) illustrates how a distinctive
congregational culture and theological beliefs among fundamentalist Christians is capable
of generating a meaningful world view that encompasses the lives of congregants.
For these reasons, I expect that evangelicals will be less likely to decline in
attendance during the transition to early adulthood compared to other religious traditions.
28
At the same time, growing up affiliated with a conservative Protestant denomination is
expected to decrease the stability of those infrequent attendees, resulting in a greater
relative probability of increasing attendance.
2.6.2 Religious Socialization
Apart from the church, there are three primary sources of religious socialization:
parents, peers, and schools. Research has typically identified parents as the primary
agents of religious socialization during youth (Erickson 1992; Ozorak 1989; Smith and
Denton 2005). Parental knowledge, belief, and behavior concerning religion provide the
framework from which the young person must navigate through his or her own religious
world. The literature on the intergenerational transmission of religious belief has fairly
consistently concluded that parental religious practice (and maternal practice in
particular) is one of the most robust predictors of offspring religiosity (Myers 1996). In
addition to parental church attendance, religious activities in the home are also an
indicator of household socialization processes and impact measures of youth religiosity
(Erickson 1992).
The other agents of religious socialization have received less attention in the
literature. The influence from peers on all spheres of life grows considerably in
adolescence (Youniss and Smollar 1985). The voluntary time spent with peers is greater
than the equivalent time spent with parents during the adolescence years (Furman 1989).
Consistent with this, research on the religion of peers does indicate that peer religiosity
has influence over religious practice during adolescence (Erickson 1992; Regnerus,
Smith and Smith 2004). Nevertheless, it is not clear whether peers exert any long term
influence over religious participation, or if, alternatively, peer influence is primarily
29
related to social comparison processes that are largely contemporaneous and fleeting in
nature.
The stated goal of many religious schools is to provide an explicitly moral and
religious education that counters the secular education available in public schooling.
And, many parents who oppose public schooling hope that religious alternatives function
in just this way (Sikkink 1999). Nevertheless, there is mixed evidence concerning the
estimated effectiveness of religious institutions in the religious socialization process.
Erickson (1992) finds that religious education does influence adolescent religiosity.1
Regnerus, Smith, and Smith (2004), on the other hand, find no discernable influence from
religious schooling on measures of adolescent church attendance. As with peers, it is an
open question as to whether any long-term religious effects exist from religious
schooling.
It seems likely that the most enduring effects on religious pathways in young
adulthood will come from the religious socialization of parents. The sustained influence
from peers and religious schools is expected to be milder, if any influence exists at all.
2.6.3 Ecological Contexts
Adolescents are embedded in local structures that shape their social world. The
social ecology defined by religious composition, class structures, racial distributions, and
urbanicity might be expected to alter the available stock of religious knowledge, customs,
and beliefs. Arguments noting the regional nature of religious commitment in the United
1 Although this may be partially due to the inclusion of congregational religious education (e.g. Sunday school) in his measure of religious education.
30
States point to these demographic characteristics as evidence for why the South is
particularly religious while the Northeastern and Western United States is not (Hunter
1983). While local contexts are certainly not expected to dictate the content of religious
belief and practice, young people are likely to develop religious ideas and practices that
are congruent with the environment they occupy. Because of this, the push to regularly
attend religious services would be expected to be prominent in areas with large
evangelical populations, high proportions of ethnic and racial minorities, and rural
communities. These environments are expected to provide religious examples and shared
cultural repertoires that nurture religious life in the transition to adulthood as well as the
social capital to more efficiently monitor behavior and punish nonconformity.
2.6.4 Social Resources and Environmental Stability
Research concerning the intergenerational transmission of religious belief and
practice has documented measures of household social capital as influencing the
successful parent-offspring transfer of religiosity. In particular both family structure
(Biblarz and Raftery 1993; Myers 1996) and the quality of parent-child relationships
(Bao et al. 1999; King, Elder, and Whitbeck 1997) influence adolescent levels of
religiosity. Measures of human capital in the form of parental educational attainment,
however, have shown mixed effects in the literature. Hoge, Petrillo, and Smith (1982)
find no relationship between parental education and a measure of theological belief.
Myers (1996) finds that father’s educational attainment is positively related to a
composite index of offspring religiosity; while mother’s educational attainment is
negatively related to the same measure. Smith and Sikkink (2003) find that father’s
education is positively related to offspring retention in the same religious tradition, but
31
only for mainline Protestants. In addition to the conflicting results, none of these studies
directly measure religious participation – the outcome of interest in the present study.
No known studies have documented the influence from neighborhood stability on
religious outcomes for youth or young adults. However, several possible expectations
can be developed from related literature. Studies have found that as individuals become
mobile, they break social ties resulting in lower levels of religious participation
(Wuthnow and Christiano 1979; Welch and Baltzell 1984). Residential turnover in a
geographic area has the same effect. As the proportion of individuals who move into and
out of a geographic area increase, social networks at the congregational level are broken
and it becomes more difficult to maintain a vibrant religious culture (Welch 1983; Finke
1989). We might suspect that young people would be more likely to decrease religious
participation in areas marked by high residential turnover.
At the same time, we might suspect that other forms of neighborhood instability
increase the likelihood of joining and participating in religious communities. Freeman
(1985) and Cook (2000) have both suggested that young people may join churches in
unstable neighborhoods as a deliberate way to avoid deviance and gang life. By
becoming involved in a church, young people can generate alternative social networks
that provide resources and direction toward positive life outcomes. Nevertheless, it is not
clear whether living in an unstable neighborhood during early adolescence would affect
pathways of religious participation that occur several years later in the life course.
2.6.5 The Mediating Influence of Life Course Events in Early Adulthood
Extant research on the life course and religious change has identified education
(Cohen 1983; Cornwall 1989; Gaede 1977; Mueller and Johnson 1975; Roof 1978) and
32
family formation based influence (Bahr 1970; Mueller and Cooper 1986; Roozen et al.
1990; Stolzenberg, Blair-Loy, and Waite 1995) on religious practice in early adulthood.
Parents, peers, and social environments all play a part in shaping educational and family
related outcomes in young adulthood. Many of the influences hypothesized to influence
differing pathways of religious participation would also be expected to influence these
other life events in young adulthood. It is quite possible, then, that any enduring
influence from social location in early adolescence on religious participation is mediated
through these key life course events in young adulthood. Alternatively, the measures of
social location in early adolescence may have a distinct impact on later religious
pathways apart from education and family related life course events.
2.7 Data
In order to measure the influence of early adolescence on religious pathways in
early adulthood, the independent variables are primarily drawn from the 1997 wave of the
NLSY97. Due to skip patterns in the data, several key variables measuring family life
were only administered to respondents between the ages of twelve and fourteen in 1997.
Thus, the final models are limited to these respondents (N=5,254). The NSLY97 also
contains a survey given to a parent or guardian who resided with the youth in 1997 from
which several variables are utilized.2 A small number of variables measuring life course
events in early adulthood are also drawn from the 2004 wave of the NLSY97.
2 A biological parent was given preference as the parental informant. If a biological parent was unavailable, preference was given to adoptive parents, stepparents, legal guardians, foster parents, and parental figures in this order. Preference was given to the mother before the father. The majority of parent questionnaires were completed by the respondent’s biological mother.
33
The 2000 Religious Congregations and Membership Study (RCMS 2000) has
been merged with the NLSY97 data to obtain county level measures of religious
adherence and denominational market share. The 2000 RCMS contains information on
religious members and religious adherents for 149 major religious bodies in the United
States.
2.8 Measures
The dependent variable is the categorical latent variable that measures the
underlying class membership in one of four distinct religious trajectories: (1) High-
attending stability, (2) Low-attending Stability, (3) Declining attendance, and (4)
Increasing attendance. In order to simplify the presentation of the models, only results
from certain categorical comparisons in the multinominal logit portion of the growth
mixture model are presented in the models.
Religious tradition is measured by grouping the self-reported denominational
affiliation of respondents into categories that approximate those established by
Steensland et al (2000). Religious tradition is coded into the following categories: (1)
Catholic, (2) Mainline Protestant, (3) Evangelical, (4) Jewish, (5) Other Religion, and (6)
No Religion.3 Respondents, age twelve to fourteen in 1997, closely identified their
religious affiliation with that of the household parental informant. The only substantial
difference in affiliation between parents and children was the greater proportion of youth
who claimed to have no religious affiliation (11.5% of youth versus 6.9% of parents).
3 The Protestant denominations reported in the NLSY97 are relatively unrefined when compared to the categorization scheme utilized in the General Social Survey. Because of this, creating a category for historically Black Protestant denominations was not possible.
34
2.8.1 Measures of Religious Socialization
Several variables are included in the models that measure the religious influence
of parents, peers, and schools in early adolescence. The annual level of parental religious
service attendance is measured on an eight point scale that ranges from “never” to
“everyday”. In addition to this measure, I also include a measure of the frequency of
family religious activity. Respondents were asked to estimate the number of days in a
typical week that the family did something religious together. The influence of peer
religious activity is measured by the estimated proportion of students in the respondent’s
grade at school who regularly attend church services. This item contains five response
categories ranging from “almost none” to “almost all”. Although this item does not
measure the influence from the respondent’s primary network of friends, it does tap into
the degree to which the adolescent perceives church attendance to be normative among
his or her peers. Lastly, I include two dichotomous variables that measure whether the
respondent’s school they presently attended (or most recently attended) in 1997 could be
classified as (1) a private Catholic school, or (2) a non-Catholic religious school.
2.8.2 Measures of Social Resources and Stability
Several variables are included which measure the degree of family resources and
neighborhood stability. At the household level, family structure is measured by a
dichotomous variable that indicates whether or not the respondent lived with both
biological parents during early adolescence. Parent’s educational attainment is measured
by the sum of the highest grade completed by the respondent’s residential mother and
father. If either the mother or father was absent, the remaining parent’s attainment was
weighted to equal that of two adults. Household income was measured in thousands of
35
dollars. The natural log of household income is used in the models to account for the
strong positive skew in the variable. Lastly, two composite scales, one measuring the
parent-youth relationship (primarily the emotional supportiveness of the parent toward
the youth),4 and one measuring the degree to which parents monitor the behavior of the
youth5, are included in the models. Both scales are created from items in the youth
questionnaire in 1997. Separate scales for residential mothers and fathers, as well as non-
residential mothers and fathers, are included in the NLSY97 data. In the present analysis,
I use the mean of the residential parents for both the parent-youth relationship scale and
the parental monitoring scale. For more information on the validity and reliability of
these scales, please consult Appendix 9 of the NLSY97 codebook supplement.
Several indicators of neighborhood stability are also included as covariates. The
degree of residential turnover is measured by the percent of the county population age 5
years and older in 1990 who were residing in the same house in 1985. This measure is
included in the geocode supplemental data to the NLSY97 and is derived from the 1994
County and City Data Book (CCDB) prepared by the U.S. Census Bureau. I also include
two indicators of crime. One is a measure of “serious crimes” per 100,000 individuals 4 The parent-youth relationship scale consists of the sum of scores on eight items. Three of the items involve likert categories indicating the youth’s agreement with the following statements: (1) I think highly of him/her (2) S/he is a person I want to be like (3) I really enjoy spending time with him/her. The answers ranged from “strongly disagree” with a score of 0, to “strongly agree” with a score of 4. Five additional questions measure the perceived frequency of supportive parental activity. The youth was asked to respond to the following statements: (4) How often does s/he praise you? (5) How often does s/he criticize you or your ideas? (reverse coded) (6) How often does s/he help you do things that are important to you? (7) How often does s/he blame you for her/his problems? (reverse coded) (8) How often does s/he make plans with you and cancel for no good reason? (reverse coded) Responses to these items were measured on a scale of “Never” with a score of 0, to “Always” with a score of 4.
5 The parental monitoring scale was created by summing the scores to four survey items. Respondents were asked to respond to the following four questions: (1) How much does he/she know about your close friends, that is, who they are? (2) How much does he/she know about your close friends’ parents, that is, who they are? (3) How much does he/she know about who you are with when you are not at home? (4) How much does he/she know about who your teachers are and what you are doing in school? The responses ranged from “knows nothing” coded as 0, to “knows everything” coded as 4.
36
(measured at the county level). This measure is also included in the geocode supplement,
derived from the 1994 CCDB. From the youth questionnaire, I also include the number
of days the respondent reports hearing gunshots in his or her neighborhood during a
typical week.
2.8.3 Measures of Ecological Context
I include several indicators of the social context of the adolescent between the
ages of 12 and 14. I use the standard census categories of region which categorizes the
state of residence as Northeast, Midwest, South, and West. I also include several county-
level indicators of race (percent of the population that is Black and percent of the
population that is Hispanic), the median level family income, and the percent of the
county that can be classified as urban as opposed to rural. All of these variables are
included in the geocode supplement and were obtained from the 1994 CCDB. Lastly, I
include two measures of the religious affiliation of the county population. Using the
2000 RCMS data, I include measures of the rate of evangelicals per 1,000 individuals and
the rate of Catholics per 1,000 individuals. Evangelical affiliation is determined by
adherence to particular denominations associated with evangelicalism. For more
information, please consult Jones et al (2002).
2.8.4 Measures of Life Course Events in Young Adulthood
In order to measure the direct versus indirect effect of social location in early
adolescence (age 12-14) on later religious pathways of participation, I include several
indicators of important life course events that have been singled-out in the literature to
influence religious participation. Educational attainment is measured by two
37
dichotomous variables. The first indicates whether the respondent had graduated with a
bachelor’s degree or a higher degree in 2004. I also include a variable that measures
whether the respondent has attended at least some college or graduated with an
associate’s degree or its equivalent by 2004. Research has demonstrated that college
graduates with a bachelor’s degree or more are more likely to attend religious services
regularly compared to those with lesser degrees or no academic degrees (Uecker,
Regnerus, and Vaaler 2007). Nevertheless, while students are enrolled in higher
education, their religious service attendance temporarily declines. Due to the age of the
sample in 2004 (age 20-22), only a small percentage of the had graduated with a BA or a
higher degree. See Chapters three and four for a more detailed analysis of higher
education and religious participation.
Prior studies have also demonstrated the importance of marriage and family for
religious participation. Variables measuring whether the respondent has a biological
child living in the household, as well as a variable indicating whether or not the
respondent is currently married or has even been married, are included in the final
models. Research has also demonstrated that engaging in cohabiting relationships prior
to marriage is associated with declining levels of religious service participation
(Thornton, Axinn, and Hill 1992). A variable indicating whether the respondent has ever
been in a cohabiting relationship is included in the analyses. Missing data are handled
via multiple imputation methods.6 Descriptive statistics for all variables are listed in
Table 2.1.
6 The ICE software program available in STATA was used to impute missing values (Royston 2005). ICE uses a maximum likelihood procedure to fill in the most likely values for missing data. In regression analysis, the standard errors of the regression coefficients are correctly computed by statistically including
38
TABLE 2.1
DESCRIPTIVE STATISTICS
Mean Std.Dev. Range Church Attendance 2000 2.526 2.161 0 to 7 Church Attendance 2001 2.353 2.077 0 to 7 Church Attendance 2002 2.059 1.985 0 to 7 Church Attendance 2003 2.023 1.975 0 to 7 Church Attendance 2004 1.855 1.910 0 to 7 Latent Class 1 to 4 Catholic 0.302 0 to 1 Evangelical 0.359 0 to 1 Mainline Protestant 0.215 0 to 1 Jewish 0.011 0 to 1 Other Religion 0.020 0 to 1 No Religion 0.115 0 to 1 Age 13.008 0.816 12 to 14 Female 0.483 0 to 1 Black 0.270 0 to 1 Hispanic 0.208 0 to 1 White Non-Hispanic 0.495 0 to 1 Other Race 0.035 0 to 1 Frequency of Parental Church 3.166 2.034 0 to 7 Frequency of Family Religious Activity 1.648 2.027 0 to 7 Frequency of Peer Church Attendance 1.838 1.138 0 to 4 Attends Catholic School 0.038 0 to 1 Attends Non-Catholic Religious School 0.021 0 to 1 Lives with Both Biological Parents 0.523 0 to 1 Parents’ Educational Attainment 24.751 5.694 2 to 40 Household Income (natural log) 3.479 0.987 0 to 5.511 Parental Emotional Support 24.717 4.710 2 to 32 Parental Monitoring 9.393 3.343 0 to16 Residential Turnover 47.391 8.664 19.6 to 73.400 Frequency of Gun Shots 0.548 1.346 0 to7 Frequency of Serious Crime 5.784 2.878 0 to 16.031 Midwest 0.223 0 to 1 West 0.371 0 to 1 South 0.230 0 to 1 Northeast 0.176 0 to 1 Percent Black (County) 14.558 15.513 .037 to 75.796 Percent Hispanic (County) 10.673 15.231 .3 to 85.2 Median Family Income (County) 35.330 8.795 12.136 to Percent Urban (County) 75.409 43.067 0 to 100 Evangelical Rate (Per 1,000 county 143.92 118.982 0 to 809.585 Catholic Rate (Per 1,000 county 221.22 152.609 0 to 695.877
a degree of uncertainty for the values that have been imputed. To avoid losing these and other respondents, most missing values have been imputed.
39
TABLE 2.1 (continued)
BA or More 0.028 0 to 1 Some College 0.474 0 to 1 Biological Child in Household 0.189 0 to 1 Ever Cohabit 0.300 0 to 1 Ever Married 0.101 0 to 1 N 5254
SOURCE: National Longitudinal Survey of Youth 1997
2.9 Plan of Analysis
I address the influence of early adolescent influences on religious participation
pathways in two ways. First, in order to understand how different religious traditions fare
among the four class trajectories presented in the first part of this chapter, I examine the
frequency of class membership by religious tradition. This is presented in basic
frequency tables. Then, I present multivariate models that contrast specific trajectory
categories.7 Specifically, I present multinomial logit models that examine predictors of
being in the declining attendance class versus the high stable attending class. I also
present a set of multinomial logit models that examine the predictors of being in
increasing attendance class versus the low stable attending class. These models are then
compared to the expected influences from religious tradition, religious socialization,
7 Although covariates can be included within the growth mixture modeling framework, the number of covariates in the present analysis makes the computation of such a model problematic. In order to test for consistency between a correctly specified GMM with covariates and the multinomial logit models presented in the tables, I compared a GMM with covariates for religious tradition to the first models in Table 2.3 and Table 2.4. Effect size and statistical significance were nearly identical, with the exception of Jewish youth not being significantly different from the omitted category of evangelical youth in either model.
40
family and neighborhood resources and stability, contextual influences, and life course
events in early adulthood.
2.10 Results
Table 2.2 presents the frequencies of class membership by religious identification.
Several contrasts are particularly noteworthy. Among the three largest religious
traditions of Catholicism, mainline Protestantism, and evangelical Protestantism, there is
only a slight difference in the proportion that fall into the declining attendance category
(with evangelicals being slightly less likely to be in this class). Somewhere between one
out of four and one out of five young people who claimed an affiliation with one of the
major Christian traditions will decline substantially in their religious participation
between the ages of sixteen and twenty-two. On the other hand, Jewish adolescents,
individuals with no religious affiliation, and individuals who adhere to other religions are
relatively unlikely to be in the declining category in young adulthood compared to those
with a Christian affiliation. For those in other religions, and especially for those without
a religious affiliation, this is likely due to low levels of overall religious participation
(evidenced in the low proportions who participate regularly between the ages of 16 and
22). However, Jewish youth are actually more likely than Catholics to regularly attend
worship services, suggesting that it is not simply overall low participation that keeps
Jewish young adults out of the declining attendance class. Still, some caution should be
exercised interpreting this finding, as the overall number of Jewish youth in the sample is
only 47.
41
TABLE 2.2
FREQUENCY OF CLASS MEMBERSHIP
BY RELIGIOUS TRADITION
Latent Class Catholic Evangelical Mainline Jewish Other No Religion
Total
Decreasing 375 24.48%
406 22.22%
267 24.88%
2 4.26%
10 11.49%
50 21.52%
1110
Stable-Low 825
53.85% 749 41.00%
514 47.90%
35 74.47%
59 67.82%
478 80.88%
2660
Increasing 110
7.18% 183 10.02%
71 6.62%
1 2.13%
8 9.20%
41 6.94%
414
Stable-High 222
14.49% 489 26.77%
221 20.60%
9 19.15%
10 11.49%
22 3.72%
973
Total 1532 1,827 1073 47 87 591 5157
SOURCE: National Longitudinal Survey of Youth 1997
Table 2.2 also presents the proportion of respondents from each religious tradition
that are best classified as increasing their religious participation in young adulthood.
Here we find less overall differences between religious traditions when compared to the
increasing attendance class. The exception may be among Jewish youth who seem
particularly unlikely to increase attendance during the late adolescent and young adult
years (although once again this is based on a small sample). Overall, evangelicals are the
most likely to increase in attendance, with those from other faiths following close behind.
Those with no religious affiliation are nearly as likely to substantially increase their
religious service attendance as those who adhere to the mainline Protestant and Catholic
42
traditions. However, once we consider the number of low, stable attending non-affiliates
(approximately four out of five non-affiliates fit this description), the relative proportion
in the increasing category is still low compared to other groups. Analyzing the table in
this manner would suggest that only around one in eleven 16 to 22 year non-affiliates
move from a state of low attendance to a substantial increase in attendance. The
equivalent comparison for evangelicals would be almost one in four. Overall, this table
reveals that those adhering to a Christian religious tradition between the ages of 12 and
14 have far less stable religious trajectories in late adolescent and early adulthood
compared to other religious groups and non-religious youth.
Table 2.3 presents a series of multivariate models that assess the relative risk of
an individual having his or her attendance trajectory being classified as declining versus
being classified as high attending and stable. It is important to remember conceptually
that these models do not predict whether a single individual switches from one trajectory
to another. Rather, each individual can be probabilistically categorized as belonging to
one trajectory or another. Nevertheless, those in the declining category are necessarily
declining from a state of high attendance. Whether their prior state was that of stable,
high attendance is not known. Nevertheless, they provide a useful comparison category
to those who do not decline, but maintain high levels of religious participation during the
transition to young adulthood. Table 2.3 and Table 2.4 are both estimated using
multinomial logit regression with coefficients displayed as odds ratios.
In the first model of Table 2.3, I include only covariates that identify the youth’s
religious tradition between the ages of twelve and fourteen, omitting evangelicals as the
reference category. The results, as would be expected, mirror the relative distributions
43
TABLE 2.3
LIKELIHOOD OF DECLINING ATTENDANCE
VERUS HIGH ATTENDING STABILITY
AGE 16-22
(1) (2) (3) (6) Religious Traditiona Catholic 2.035** 2.030** 1.784** 1.721** Mainline Protestant 1.455** 1.407** 1.397** 1.415** Jewish 0.268+ 0.253+ 0.219+ 0.215+ Other Religion 1.204 1.030 1.081 1.159 No Religion 2.737** 2.679** 1.708* 1.549 Individual Characteristics Age 0.832** 0.827** 0.871* Female 0.807* 0.791** 0.867 Blackb 0.893 0.936 0.881 Hispanicb 0.832 0.871 0.848 Other Raceb 1.197 1.175 1.167 Religious Socialization (Age 12-14) Frequency of Parental Church Attendance 0.876** 0.895** Frequency of Family Religious Activity 0.881** 0.885** Frequency of Peer Church Attendance 0.883** 0.911* Attends Catholic Schoolc 0.944 0.952 Attends Non-Catholic Religious Schoolc 0.745 0.674 Social Resources and Stability (Age 12-14) Lives with Both Biological Parents 0.791* Parents’ Educational Attainment 0.998 Household Income (natural log) 1.087 Parental Emotional Support 0.982 Parental Monitoring 1.001 Residential Turnover 0.995 Frequency of Gun Shots 0.976 Frequency of Serious Crime 1.006 Contextual Influence (Age 12-14) Midwestd 0.863 Westd 0.917 Southd 0.980 Percent Black (County) 0.993 Percent Hispanic (County) 0.995
44
SOURCE: National Longitudinal Survey of Youth 1997 NOTE: Multinomial Logistic Regression; Coefficients are Odds Ratios; aReference group is Evangelical; bReference group is White; cReference Group is all Non-Religious Schools; dReference group is West; eReference Group is High School Degree or Less +p<.10 *p<.05 **p<.01 (two-tailed test) seen in Table 2.2. Those who claimed to have no religious affiliation during early
adolescence were the most likely to experience relative decline versus high attending
stability. Among religious groups, Catholics were approximately twice as likely as
evangelicals to experience relative decline while Jewish adolescents and young adults
were the least likely. Those from other religious traditions were not statistically different
from evangelicals.
Including individual level control variables in Model 2 does little to alter these
initial differences by religious tradition. Older respondents (measured by whether they
were twelve, thirteen, or fourteen in 1997) were less likely to decline in attendance
compared to being in the high attending category. This corroborates Figures 2.3 and 2.4
which display a substantial decrease in membership in the declining attendance category
for the older group of respondents. Consistent with other research documenting the
religious difference between males and females, females are less likely to be in the
TABLE 2.3 (continued)
Median Family Income (County) 1.007 Percent Urban (County) 1.001 Evangelical Rate (Per 1,000 county residents) 1.000 Catholic Rate (Per 1,000 county residents) 1.001 Life Course Events (Age 20-22) BA or Moree 0.555* Some Collegee 0.632** Biological Child in Household 0.713* Ever Cohabit 1.725** Ever Married 0.573** N 5254 5254 5254 5254
45
declining category versus the high stable attending category when compared to males.
No statistically significant differences by race were found.
Model 3 includes the bulk of the predictors that measure the social location of the
individual during early adolescence. An examination of the primary agents of religious
socialization reveals that families as well as peers influence the relative risk of decline in
late adolescence and early adulthood. Individuals who had attended a religiously
affiliated school were not significantly different from other adolescents who attended
other public and private schools. The variables designed to measure social resources and
stability were mostly ineffective predictors of the relative risk of decline during the
transition to adulthood. The one exception to this is the variable measuring family
structure. Respondents who resided with both biological parents between the ages of
twelve and fourteen were at less risk of later declining attendance. Emotional support
from parents also lowered the likelihood of decline, but this was only significant at the
.10 level. Measures of neighborhood instability were non-significant. Lastly, the broader
social context, primarily measured at the county level (with the exception of geographic
region), were mostly non-significant predictors of relative risk of declining attendance.
Respondents residing in counties with large African American populations appear to be
slightly less likely to be in the declining category versus stable high attendance. Region,
median family income, urbanicity, and the distribution of religious adherents at the
county level do not predict the relative risk of decline.
The inclusion of these variables designed to account for the social location of the
youth in 1997 alter the effect size for Catholics and those without a religious affiliation.
In both cases, the difference between these groups and evangelicals becomes smaller.
46
Separate analyses confirm that this is largely due to parental religious attendance for
those without a religious affiliation and primarily due to the frequency of family religious
activities for Catholics.
Model 4 in Table 2.3 includes variables measured in 2004 (when respondents
were between the ages of twenty and twenty-two) that tap into key life course events in
early adulthood. As expected, higher educational attainment, having children, and
getting married are all associated with a lower relative risk of decline, while cohabiting is
associated with an elevated risk of declining religious practice. The inclusion of these
indicators of life course activities does little to influence the effect of religious tradition.
The only exception to this is the group without a religious affiliation drops from being
slightly significantly different from evangelicals, to no longer being significantly
different (although the effect size still suggests they are at 55% higher risk of decline, we
can not be confident that this difference is true in the population). This suggests that the
effect of religious tradition on later risk of decline is largely a direct effect, unmediated
by later key life course events. One of the most striking findings concerns the effect of
gender on relative risk of decline. A substantial portion of the difference between males
and females, according to Model 4, can be attributed to differences in life course events
related to education, marriage, and children. Separate analyses find that having a child
and attending at least some college largely account for the differences in the relative risk
of decline between males and females.
Table 2.4 examines predictors of the relative risk of increasing attendance versus
low attending stability between the ages of sixteen and twenty-two. Several differences
exist by religious tradition according to Model 1. Specifically, compared to evangelicals,
47
all groups are at less risk of a relative increase in attendance (although those from non-
Christian and non-Jewish faiths were not statistically different from evangelicals). Those
at the least risk were Jewish youth, followed by those with no religious affiliation.
Catholic and mainline Protestant affiliates were nearly half as likely to belong to an
increasing attendance class versus low attending stability.
Model 2 includes individual level covariates. Unlike in Table 2.3, age group is
not a significant predictor of relative increase. Females are more likely to be in the
increasing attendance class relative to low attendance, while non-Hispanic whites (the
reference category) are relatively less likely than African Americans and Latinos to be in
the increasing religious participation category.
Model 3 includes variables measuring religious socialization, social resources and
stability, and contextual influences. Neither peer church attendance nor the religious
affiliation of the school attended during early adolescence predicts the relative risk of
belonging to the increasing attendance trajectory. However, both family measures of
religious practice (parent’s attendance and frequency of family religious activities)
increase the likelihood of increasing religious practice versus remaining at a stable low
level of attendance. Family structure does not appear to alter the relative risk of increase,
while having educated parents increases the likelihood of remaining at a low stable level
of attendance. No other family or neighborhood measures of resources and stability
predict the relative likelihood of increasing attendance. The particular variables chosen
to measure contextual influence on the relative risk of increasing religious participation
were all non-significant, with the exception of geographic region of residence. Those
who resided in the South between the ages of twelve and fourteen were 87.7% more
48
TABLE 2.4
LIKELIHOOD OF INCREASING ATTENDANCE
VERSUS LOW ATTENDING STABILITY
AGE 16-22
(1) (2) (3) (4) Religious Traditiona Catholic 0.546** 0.616** 0.624** 0.738+ Mainline Protestant 0.565** 0.691* 0.696* 0.783 Jewish 0.117* 0.178+ 0.200 0.325 Other Religion 0.555 0.909 0.897 1.070 No Religion 0.351** 0.408** 0.519** 0.587** Individual Characteristics Age 1.046 1.045 1.014 Female 1.467** 1.501** 1.513** Blackb 2.033** 1.824** 1.720** Hispanicb 1.669** 1.583** 1.486+ Other Raceb 0.498 0.485 0.512 Religious Socialization (Age 12-14) Frequency of Parental Church Attendance 1.090** 1.085* Frequency of Family Religious Activity 1.101** 1.087** Frequency of Peer Church Attendance 1.029 1.010 Attends Catholic Schoolc 0.955 1.286 Attends Non-Catholic Religious Schoolc 0.596 0.662 Social Resources and Stability (Age 12-14) Lives with Both Biological Parents 1.035 Parents’ Educational Attainment 0.972* Household Income (natural log) 1.056 Parental Emotional Support 0.998 Parental Monitoring 0.996 Residential Turnover 1.003 Frequency of Gun Shots 1.024 Frequency of Serious Crime 0.988 Contextual Influence (Age 12-14) Midwestd 1.431+ Westd 1.066 Southd 1.835* Percent Black (County) 1.000 Percent Hispanic (County) 1.000
49
SOURCE: National Longitudinal Survey of Youth 1997 NOTE: Multinomial Logistic Regression; Coefficients are Odds Ratios; aReference group is Evangelical; bReference group is White; cReference Group is all Non-Religious Schools; dReference group is West; eReference Group is High School Degree or Less +p<.10 *p<.05 **p<.01 (two-tailed test)
likely to increase their attendance versus those whose residence was in the Northeastern
United States. Those who lived in the Midwest were 46.1% more likely to increase
attendance compared to those who lived in the Northeast.
The addition of variables measuring life course events related to education and
family formation in Model 4 do not substantively alter the relationship between any of
the other independent variables and the relative risk of increasing religious attendance.
Those who graduate from college with at least a bachelor’s degree are approximately
twice as likely to experience a relative increase in attendance versus those who did not
attend college at all. Having a biological child in the household does not influence the
likelihood of following an increasing attendance pathway in early adulthood. It is highly
likely that the majority of parents in this data (age twenty to twenty-two) had children
that were quite young at the time of their interview. Past research has identified
TABLE 2.4 (continued)
Median Family Income (County) 0.991 Percent Urban (County) 1.001 Evangelical Rate (Per 1,000 county residents) 0.999 Catholic Rate (Per 1,000 county residents) 0.999 Life Course Events (Age 20-22) BA or Moree 2.071* Some Collegee 0.887 Biological Child in Household 1.141 Ever Cohabit 0.707** Ever Married 1.876** N 5254 5254 5254 5254
50
increasing patterns of religious service attendance when children reach school age, but
not prior to this (Stolzenberg, Blair-Loy, and Waite 1995). Lastly, cohabiting reduces the
likelihood of being in the increasing attendance class, while being married increases the
likelihood of being in the increasing class.
In Model 4, with the inclusion of the full set of covariates, the effect from
religious tradition is considerably lessened from the effect size estimated in Model 1.
Catholics are the only religious tradition to remain slightly significantly different from
Evangelicals (at the .10 level). Those without a religious affiliation are still the least
likely statistically significant category of individuals to be classified in the increasing
attendance class versus the low attending class. The inclusion of individual demographic
characteristics, measures of family religious activity, and region of residence explained a
substantial portion of the difference between religious traditions.
2.11 Discussion
2.11.1 Religious Tradition
The religious traditions measured in the NLSY97 play different roles in religious
retention versus religious uptake in late adolescence and early adulthood. Between the
Christian traditions of Catholicism, mainline Protestantism, and evangelical Protestantism
there exists statistically significant differences in the relatively likelihood that individuals
will decline in attendance versus maintaining high levels of religious participation.
Jewish youth seem to be at far less risk than even evangelicals for this relative risk of
religious decline, although caution should be taken in this finding, given the low number
51
of Jewish youth available in the sample. The overall difference between the religiously
affiliated and the non-affiliated is not statistically different when we examine this relative
risk of decline.
This finding suggests that there is something about evangelicalism, beyond
demographic characteristics and family religious practice, which keep youth involved in
church to a greater extent than other Christian traditions during the critical years of late
adolescence and early adulthood. Lytch (2003) argues that the congregational level
processes of religious socialization and religious experience work in tandem to shape
youth loyalty to religious traditions. The particular combination of strong socialization
into the particular rituals, beliefs, and religious language of a tradition is cemented as
one’s own during personal religious experience. Those congregations and traditions that
are better able to give young people the religious tools to interpret religious experience as
salient to their sense of self and identity would be expected to retain individuals into
young adulthood at higher rates. Denominations that are classified as evangelical (the
majority being Southern Baptist in this analysis), are arguably more effective in this task.
Evangelicals have developed a vibrant subculture in the United States, one which
includes a distinct terminology, sanctions against nonconformity, and an emphasis on
individualist commitment and experience. Socialization into the particularities of this
subculture commonly includes the expected cues and signals of religious experience
along with language with which to interpret this experience. In this way, the
congregational experience of a young adolescent in an evangelical congregation would
likely be distinct from that of a mainline Protestant or Catholic congregation.
52
Religious tradition, however, matters less in accounting for the difference
between those who do not regularly attend and those that increase their attendance
substantially during the transition to adulthood. With the exception of a slightly
significant difference between Catholics and evangelicals, the primary cleavage is
between those who maintain a religious affiliation and those who do not. Religiously
affiliated youth, including those who are non-practicing, likely have a greater degree of
exposure to specific religious cultures and their accompanying stories, beliefs, and rituals.
This knowledge, then, would be expected to reduce the traversable cultural boundaries
necessary in gaining entrance into a religious community. The particular effectiveness of
this process between religious traditions, however, can largely be accounted for by
gender, race, family religious practices, and regional culture.
2.11.2 Religious Socialization
The primacy of parental religious socialization and household religious practices
was confirmed in the analyses presented here. Parental religious service attendance
during early adolescence is a strong predictor of the religious pathways a young person
takes as they develop a more autonomous self. However, beyond the frequency of church
attendance, youth who report frequent religious activities in the home are also
substantially influenced by this in their decisions to participate in a religious community
during the transition to adulthood. This suggests that there is something additional
beyond simple modeling behavior that influences religious participation. Young people
appear to internalize faith practices partly due to religious life being reinforced in the
household. These influences from parents, as expected, act to reinforce the likelihood of
retention in a religious community for the youth. However, these parental practices also
53
act to decrease the likelihood of maintaining a low, stable attendance pattern in late
adolescence and early adulthood. This suggests that it is difficult for a young person to
maintain a pattern of religious participation that conflicts with that of their household of
origin, even as they enter the relatively autonomous stage of early adulthood.
Early adolescent peer involvement in religious communities appears to exert some
degree of lasting influence over religious retention. To the extent that young adolescents
perceive their peers (defined as youth in the same grade who with whom they attend
school) participating frequently in religious services, they themselves are more likely to
maintain their own religious participation stably into young adulthood.8 Although
parents are arguably the most important lasting influence from early adolescence, the
enduring influence from peers appears to extend beyond the sphere of popular culture to
the sphere of religious practice. This particular finding suggests that future research
would benefit from specifying more precisely the impact from peers on the religious life
of young people. Nevertheless, no lasting influence from peers on those who
substantially increase their religious service in early adulthood is found in this analysis.
It is likely that immediate social networks and contemporaneous peer culture exert a
larger influence on the decision to join a religious community, as opposed to the decision
to stay in a religious community.
No long-term effect on religious participation patterns from attending a religious
school between the ages of twelve and fourteen could be found. Although the difference
8 The variable measuring peer attendance does not measure the attendance of the respondent’s group of primary friends. Rather, the variable taps into the perception of the frequency of religious service attendance for all similarly aged youth attending his or her school. Because of this, the measure may be capturing what the respondent thinks the norm is regarding church attendance for young people in his or her community. Undoubtedly this is influenced by the respondent’s own friendship network, yet it is not narrowly focused on this dimension of peer attendance.
54
between those attending a non-Catholic religious school (with a likely conservative
Protestant affiliation) and a non-religious school were moderate in effect size, no
statistically significant difference was found (perhaps due to the low numbers of
respondents attending such schools). Although many schools have as a stated goal to
provide religious education, in a society that places great import on education for its
utility in the market, even religious schools are likely to be dominated by educational
programs primarily designed to enroll the majority of their students in higher education
and, ultimately, achieve occupational and economic security. The time a typical student
spends learning geometry and studying British literature likely dwarfs any time spent
learning theology or religious rituals. Although a more directed study of conservative
Protestant schools may be able to trace the influences of religious education into early
adulthood, this analysis suggests that schools are less significant for later religious
pathways when compared to the household, congregation, or peer group.
2.11.3 Resources, Stability, and Context
Overall, most of the variables measuring social resources and environmental
stability were poor predictors of religious pathways in early adulthood. Even so, the two
variables that did seem to have an influence were both measures of household
characteristics, as opposed to neighborhood characteristics. Family structure, specifically
living with both biological parents in early adolescent, reduces the likelihood of dropping
in attendance versus staying in church. This finding, consistent with the research on the
intergenerational transmission of religious practice (Myers 1996), is likely due to the
difficulty of successful religious socialization with only one parent. Having highly
educated parents, however, reduces the likelihood of becoming active in a religious
55
congregation during early adulthood. Past research on the transmission of parent-
offspring religious values has obtained mixed results when examining the influence of
parent’s education on the likelihood of successfully transmitting religiosity (Hoge,
Johnson, and Luidens 1994; Myer 1996; Smith and Sikkink 2003). The inconsistent
results may be partly due to the different mechanisms at work. Having highly educated
parents reduces the likelihood of beginning to regularly attend religious services, but it
has no influence on the likelihood of retaining those who already attend. At the same
time, if the respondent graduates from college, he or she is more likely to increase their
attendance. These counteracting forces may be partially accounted for by cohort
differences in the influence of higher education on religious practice (see Appendix A).
The measures of geographic context also had little influence over later pathways
of religious participation with the exception of geographic region. Living in the
Midwestern and Southern United States increases the likelihood of becoming active in a
religious congregation during the transition to early adulthood. Extant research identifies
high rates of religious participation in these regions for both adults (Finke 1989) and
adolescents (Smith and Denton 2005). Moreover, regional differences in religious
participation persist even despite continued geographic mobility across regions
(Iannaccone and Makowsky 2007; Smith, Sikkink, and Bailey 1998). However, the
analysis presented in this chapter indicates that this regional influence is not due to the
proportion of the population that is evangelical, the geographic distribution of race, the
wealth of residents, or the lower levels of urban population in the South and Midwest.
Southern culture and institutions in particular (and perhaps to a less extent the culture and
institutions of the Midwest) do not appear to provide the support for stable, low religious
56
participation as is the case in the Northeastern and Western United States. However, the
mechanisms that keep individuals in church once they attend regularly do not appear to
be related to regionalism.
2.11.4 Direct Versus Indirect Effects
The analysis presented in this chapter largely corroborates the evidence
regarding the religious influence of education and family formation (see Chapters 3 and 4
for more on higher education and religious practice and belief). Most of the significant
predictors from early adolescence have a direct effect, unmediated by these life course
indicators in early adulthood.
The differences in religious retention between men and women are one exception
to this finding. Once life course events in early adulthood are controlled for, the
differences between men and women were no longer significant. This result is important
because it suggests that women are not less likely to decline in religious particpation
during early adulthood because of firmly rooted gender differences cemented in early
childhood socialization (Thompson 1991), or because they are less prone toward risky
behavior (Miller and Stark 2002), but rather because a larger proportion of women now
attend higher education compared to men along with the added religious responsibilities
of raising a young child for some women.
2.12 Conclusion
This chapter makes several key contributions to our understanding of religious
participation during the transition to early adulthood. Past research has typically
57
analyzed mean changes in religious service participation, or, alternatively, identified the
proportion of young people who report increasing or decreasing attendance over some
predefined period of time. Unfortunately these methods are plagued by certain pitfalls.
Analyzing mean change often times masks important patterns of growth and decline
while measuring the proportion who are stable, increase, or decrease is not able to convey
the degree of growth or decline, and it is sensitive to the time frame between
measurements. To provide a better map of change in religious participation during this
period of the life course, I use growth mixture modeling along with five waves of data on
church attendance to uncover latent patterns of religious participation trajectories
between the ages of sixteen and twenty-four.
This research also has the advantage of using concurrently collected panel data
as opposed to retrospective life histories. Although the latter have distinct advantages,
they are often times plagued by an individualist interpretation of life events that masks
the social influences on individual action. Moreover, studies that have used panel data
have not typically focused their analysis on the key ages of late adolescence and early
adulthood, when changes in religious practice are most likely to occur (for an exception
see Uecker, Regnerus, and Vaaler 2007).
Lastly, this study expands the focus outward from family or origin and family of
procreation influences. Although families are a critical part of the religious lifecycle, the
adoption of the life course framework draws our attention to the myriad social influences
of networks and ecological contexts on patterned trajectories of change. One of the
shortcomings of the present study is the broad county-level measures of context
employed, possibly underestimating the effects of the proximate context. Greater
58
precision in measurement is necessary to better specify the influence from ecological
context on religious outcomes. Future research would also benefit from the introduction
of more direct measures of social networks and congregational life. Understanding the
social influences outside of the family will greatly expand our understanding of religion
during this key juncture in the life course.
59
CHAPTER 3:
HIGHER EDUCATION AND CHANGE IN RELIGIOUS IDENTITY AND PRACTICE
3.1 Overview
Sociologists of religion have long recognized that religious beliefs and practice
vary over time with the changes in social arrangements that constitute the individual life
course (Bahr 1970). Recent research on life course transitions and religion has focused
primarily on the Family Life Cycle model; namely that marriage and childbearing bring
about increases in religious participation (Bahr 1970; Mueller and Cooper 1986; Roozen
et al. 1990; Stolzenberg, Blair-Loy, and Waite 1995). Nevertheless, it is recognized that
the increases in religious service attendance typically brought about by family formation
are commonly preceded by a period in late adolescence and early adulthood marked by
low levels of religious practice (Bellah et al. 1985; Roozen 1980; Wilson and Sherkat
1994). While this period of decline in religious practice is widely acknowledged in the
popular culture, and more formally documented in sociological research, little is known
about the processes that produce this decline. The purpose of this chapter is to focus on
the religious effect of one highly formative institution in the transition to adulthood – the
university. Higher education has been targeted as a primary secularizing agent in the
sociological literature (see Caplovitz and Sherrow 1977; Hammond and Hunter 1984;
Hunter 1987; Roof and McKinney 1987; Wuthnow 1988) as well as among concerned
religious advocates (see Budziszewski 1999; Campolo and Willimon 2002; Nye 2005;
60
Wheaton 2005). Despite this, the impact of college on religious identity and practice has
not been specified adequately, looming larger in anecdotal accounts and a priori
assumptions.
The intent of this research is to re-examine the relationship between higher
education and religious belief and practice in light of both methodological advances in
the social sciences and the changing nature of higher education. As higher education
continues to expand, the importance of understanding how college shapes students’
attitudes and beliefs becomes ever more crucial. Past research has documented that
college students become more politically tolerant, display an increase in altruistic values,
and support civil rights and values to a greater extent (see Pascarella and Terenzini
1991:276-280). Despite this, recent research on religious beliefs has been relatively
scarce (Johnson 1997; Pascarella and Terenzini 2005).
This chapter corrects for this oversight by examining nationally-representative,
longitudinal data for changes in religious salience, practice, and identity over the college
years. Specifically, changes in religious service attendance, frequency of prayer, self-
rated importance of faith, and religious affiliation are assessed over a six year period
ranging from high school to young adulthood. Moreover, the design used in this study
incorporates a control group of non-college graduates that is essential for separating the
effects of educational attainment from more general maturational and life course effects
associated with early adulthood. Further discussion of the methodological advantages
will be addressed shortly, but first we direct our attention to some broader theoretical and
empirical issues that shape our understanding of the effect of college on religious identity
and practice.
61
3.2 Religion and College in Sociological Perspective
Sociologists of religion have primarily discussed the effects of higher education in
the context of secularization (Hammond and Hunter 1984; Hunter 1987; Roof and
McKinney 1987; Wuthnow 1988). In most of these studies, the actual college
environment is not explicitly examined, but rather implicitly assumed to undermine
traditional religious belief. With the adoption of the Bergerian language of “pluralism”
and “plausibility structures”, these sociological accounts contend that the encounter with
alternative ideas and lifestyles on campus (i.e. pluralism) undermines the foundation of
dogmatic religious beliefs (Berger 1967). The expansion of higher education, therefore,
contributes to an overall secularizing influence in society.
Of course, with the questioning and refinement of secularization theory among
sociologists (see Stark and Finke 2000; Warner 1993), it is appropriate to reconsider how
powerful an effect the college environment has on the plausibility of religious belief.
Berger (1999) has himself acknowledged that religious belief may be more resilient than
he once theorized. Moreover, Smith (1998) has argued that certain aspects of modern
pluralism may actually serve as fodder for social psychological cohesion within religious
communities. From this perspective, college campuses could be rife breeding grounds
for subcultural solidarity. Similarly, Hammond and Hunter (1984) find that evangelicals
on secular campuses appear to be maintaining their faith more than evangelicals on
religious campuses. They posit a similar explanation as Smith, hypothesizing that
minority identities require more active maintenance than dominant identities, resulting in
62
a stronger overall identity. These alternative conceptualizations of the interplay between
pluralism and religion warrant a fresh examination of the relationship between higher
education and personal religious identity.
3.3 Past Research
Quantitative research on the effects of higher education on religious belief and
practice has been relatively sparse over the past few decades. Prior to the decade of the
1970s, numerous studies were accumulated by social scientists exploring this particular
relationship (for a review see Feldman and Newcomb 1969:23-28), but the number of
studies since then has dropped off to a small handful (Johnson 1997). While the early
studies were frequent, what they were able to conclude was tentative at best and often
plagued by inadequate methodologies and limited samples that placed a damper upon
claims to causality and inference to the larger population. An examination of the studies
reviewed by Feldman and Newcomb (1969) reveals small samples selected from single
colleges and universities with only a few having longitudinal data. Most of these studies
use a college population to study cross-sectional differences between freshman and
seniors, finding seniors less orthodox in their belief, more skeptical about God, and less
trustful in the institution of the church as compared to freshman. Despite the lack of
generalizability, the studies were numerous and consistent enough to establish what was
considered a moderate secularizing effect of college on students.
With the growth of large-scale survey research in the 1970s, sociologists began
using general population surveys to examine the relationship between years of education
and religiosity among the general population. Though the studies are fewer in number
63
than the prior research, the evidence from these general population surveys concur with
the conclusions of the college population studies: orthodox religious belief is lower
among the college educated (Cohen 1983; Gaede 1977; Hunsberger 1978; Roof 1978).9
More recently, two key studies have used panel data to examine educational
attainment and changes in religious belief and practice. Using data from the Youth-
Parent Socialization Panel Study (YPSPS), Sherkat (1998) concludes that college
graduates are less likely to have traditionally orthodox views of the Bible, less likely to
pray, and no different from those with lower educational attainment in their frequency of
religious service attendance. Using the same data, Sherkat and Wilson (1995) and
Wilson and Sherkat (1994) also find that college graduates are substantially more likely
to disaffiliate from organized religion entirely. Likewise, Funk and Willits (1987)
examine changes in belief about God and religion over an eleven year period and find
that those respondents who graduate from college display more secular change in belief
than those who did not.
While these panel studies are probably the most complete of their kind, the data is
also dated. The first wave of the YPSPS data is a nationally representative sample of
high school seniors in 1965. The data that Funk and Willits analyze is originally drawn
from high school sophomores residing in Pennsylvania in 1970. Research has
documented the changing demographic backgrounds of college students (Dougherty
1997; Hout 1988), the changing attitudinal and religious beliefs of undergraduates (Astin
9 The effect of higher education on religious service attendance remains one notable exception to this trend (Cohen 1983; Gaede 1977; Mueller and Johnson 1975; Petersen 1994; Roof 1978). One post hoc explanation for this anomaly focuses on the social status rewards associated with attendance. According to this account, higher education brings with it status expectations that are manifest in religious service attendance as well as other voluntaristic behaviors (Petersen 1994). An alternative account suggests the opposite causal influence; positive associations with adults and peers among frequent church goers may actually encourage continuing with schooling versus other life course paths (Loury 2004).
64
1997; Hoge et al. 1987; Moberg and Hoge 1986), and the movement of religiously
sympathetic faculty into prestigious positions (Schmalzbauer 2003; Schultze 1993).
Because college campuses are likely different in a myriad of ways from the late 1960s
and early 1970s, caution should be taken in inferring the results of these studies to recent
college attendees.
The ethnographic work of Cherry, Deberg, and Porterfield (2001) underscores the
point that college campuses may be religiously and spiritually different from the near
past. The authors conducted studies of four major universities in the United States to
assess the vitality of religion on American campuses. Contrary to what past sociologists
hypothesized, they found religion to be alive and well on college campuses. A variety of
religious traditions were practiced at the schools and interest in religion from an
academic perspective was high. However, in accordance with what Robert Wuthnow
(1998) has observed about religion in the United States more generally, most of the
undergraduates could be classified as religious seekers, not religious dwellers. In other
words, these students more freely constructed their religious identities from available
religious symbols without regard for more traditional denominational separations. The
students also preferred to refer to themselves as “spiritual” as opposed to “religious”.
Nevertheless, the authors were surprised at the popularity of evangelical parachurch
groups on two of the campuses. These ethnographic observations paint a picture that is
far more complex than past sociological theories associated with secularization would
allow.
65
Although some recent literature casts doubt on declining religious practice and
identity due to educational attainment, I will nonetheless draw the first hypothesis from
the best quantitative studies available. Therefore:
H1: Declines in religious practice, salience, and identity among recent
college graduates will be significantly larger compared to non-college
graduates of similar age.
3.4 Differential Effects of College on Religion
3.4.1 Gender
An alternative avenue of study has examined how the college experience may be
predicated on the specific characteristics of the students, and how this, in turn, may affect
religious behaviors and beliefs. Several studies have examined the differences between
college-age males in females. Fairly consistently, this research finds that religiosity
declines more among males as compared to females (Astin 1978; Frantz 1971; Funk and
Willits 1987; Trent and Medsker 1968). Extant research has continually reinforced the
near “universal” difference in religiosity between males and females (Beit-Hallahmi
1997; Bensen et al. 1989; Miller and Stark 2002). Females are expected to retain
religious belief and practice because of a greater investment in such beliefs and practices.
From a young age, females are more likely than males to be socialized into the role of
caretaker of the family. Therefore characteristics such as nurturing and submissiveness
develop to a greater degree in females as opposed to males (Mol 1985). These particular
66
characteristics are more conducive to religious belief and adherence to religious
institutions (Thompson 1991).
Nevertheless, one might suspect that the progressive institution of higher
education may quell this tendency toward sex differentiation regarding religion. Gender
convergence has emerged among 12 of the 18 “values” questions adminstered by CIRP in
its annual freshman survey (Astin 1997). Women have also moved into the traditionally
male-dominated fields of science and engineering at a rapid rate worldwide (Ramirez and
Wotipka 2001). The increasingly similar experiences of males and females in college
may serve to negate the “universal” difference in religiosity. In fact, only two studies
were identified by Pascarella and Terenzini (1991) that tested for statistical interactions
between gender and college education, and neither of these studies found significant
differences in sex-based college effects (see Frantz 1971; Funk and Willits 1987). Thus,
my second hypothesis:
H2: College graduates will display significantly more gender similarity in
religiosity than non-college graduates.
3.4.2 Academic Field
Past research on faculty has documented differential rates of church attendance
and religious beliefs among the various disciplines – with irreligiousness highest in the
social sciences and humanities (Hoge and Keeter 1976; Lehman and Shriver 1968; Stark
and Finke 2000; Wuthnow 1989). While some of this research has suggested that there is
something inherent about the disciplines themselves that cause these differences, there is
also conflicting evidence that suggests religious belief or lack of belief occurs prior to
entrance into the particular discipline (Thalheimer 1965). For students, no published
67
study could be found that explicitly examined changes in religious belief by college
major. Nevertheless, college major has been examined as an influence on political
attitudes and values. Research has documented that students in the social sciences and
humanities exhibit gains in liberalism, even after controlling for initial political status
(Abravanel and Busch 1975; Bachman 1972; Solmon and Ochsner 1978; Spaeth and
Greeley 1970). It seems likely, then, that the social sciences and humanities may exhibit
a secularizing effect on students as their values become more liberal. Thus, I frame my
third hypothesis with the expectation that college graduates will differ by college major:
H3: College graduates who majored in the social sciences or
humanities will secularize more than college graduates with other
majors.
3.4.3 Catholicism
There may also be reason to suspect that a disparity between Catholics and
Protestants exists in the effects of higher education on religiosity. Caplovitz and Sherrow
(1977) documented much lower rates of apostasy among Catholics than Protestants
among a sample of college graduates in 1961. However, by 1969 Catholics and
Protestants had virtually identical rates of apostasy. More recent research on religious
switching has suggested that Catholics are more likely to drop out of religion altogether
than switch to some form of Protestantism (Sherkat and Wilson 1995). If higher
education engenders demand for a more autonomous, anti-institutional style of religion as
the ethnographic reports from Cherry et al. (2001) suggest, this may pose more of a
problem for Catholics than Protestants. Catholics, having their religious beliefs more
intimately entwined in the institution of the Church itself, may abandon some of these
68
beliefs as anti-institutional attitudes are embraced during the college years. The arguably
more rigid nature of the Catholic Church as opposed to Protestant denominations may not
allow Catholics to reject the Church and hold onto religious belief as easily as some
Protestants would be able to.
On the other hand, Greeley (1977) has made the argument that many American
Catholics are able to reject certain aspects of the Church’s teachings, yet remain loyal to
Catholicism and its sacramental tradition as a whole. Dillon (1996) demonstrates that
Catholic college students, even among the most elite universities, are essentially no
different than American Catholics as a whole when it comes to their rejection of certain
teachings but their continued identification with the Church and its traditions.
Again, despite counter evidence, I structure the hypothesis with the expectation
that educational attainment affects Catholics differently:
H4: College education will have a greater secularizing effect for
Catholics than non-Catholics.
3.4.4 Conservative Protestantism
Conservative Protestant groups have been documented to have lower educational
attainment when compared to other religious groups (Beyerlein 2004; Darnell and
Sherkat 1997; Sherkat and Darnell 1999).10 The conservative Protestant emphasis on a
literalist or semi-literalist interpretation of scripture, disagreement with modern scientific
accounts of human origins, and emphasis on traditional sources of moral authority
10 Beyerlein (2004) makes the case that this is solely due to the influence of Pentecostals and fundamentalists. Those belonging to evangelical denominations had reported levels of educational attainment similar to other religious groups.
69
seemingly clash with the scientific rationalism that has become dominant in the realm of
higher education (Hunter 1987; Marsden 1996; Smith 2000). It is little wonder that
conservative Protestants may view avoiding higher education as an important strategy in
the maintenance of their faith.
Nevertheless, there is evidence that secular higher education may not be as
detrimental to conservative Protestant faith as once thought. Wuthnow (1988:187)
documents that between 1960 and 1972 the proportion of evangelicals who had attended
college tripled. Moreover, McConkey (2001) tracks a weakening of the negative
relationship between evangelical identification and educational attainment between the
years 1988 and 1998. Responding to Hunter’s (1983) findings that evangelicals are less
educated, Smith (1998) finds that self-identified evangelicals have higher educational
attainment than fundamentalists, liberal Protestants, Catholics, and those with no
religious affiliation (Smith 1998:76). Additionally, the research by Darnell and Sherkat
(1997) and Sherkat and Darnell (1999) use data in which the majority of their
respondents would have attended college in the late 1960s. It is difficult to know whether
or not their conclusions are still relevant for a younger generation of conservative
Protestants. All of this suggests that conservative Protestants may not be shirking higher
education as in the past.
Still, my fourth hypothesis is structured with the prevailing assumption that higher
education is detrimental to conservative Protestant faith:
H5: College education will have a greater secularizing effect on
Conservative Protestants in comparison to other religious groups.
70
3.4.5 African-Americans
The overwhelming majority of black religious adherents belong to one of seven
major Black Protestant denominations (Roof and McKinney 1987; Steensland et al 2000).
Roof and McKinney (1987) estimate that 85% of black Christians belong to a historically
black denomination while Lincoln and Mamiya (1990) estimate that more than 80%
belong to these denominations. While the Black Church has always occupied a unique
niche in the history of American Protestantism, their theological emphasis on a literalist
interpretation of scripture and traditional sources of moral authority most closely
resembles white evangelicals (Lincoln and Mamiya 1990; Steensland et al 2000). In this
regard, we might suspect that Black religious adherents will have similar difficulties as
other conservative Protestants with the secular milieu of higher education.
Nevertheless, minority status may function to strengthen black identity, as racial
categories provide salient markers for group identity on heterogeneous campuses. Those
students that have been raised in the Black Church may draw upon their religious identity
and accompanying theology and symbols to help them negotiate their identity as a
minority student. In this manner, the socializing processes associated with the college
experience may actually provide an opportunity for deeper religious engagement.
Despite the possibility that minority status on campus could solidify black
religious identity, the following hypothesis reflects the traditional assumption that the
college campus will be detrimental to the more conservative religious beliefs of many
blacks:
71
H6: Educational attainment will more negatively affect the religious
practices and identity of African-Americans in comparison to other
races.
3.5 Data and Methods
3.5.1 Data
The data in this analysis come from the National Longitudinal Study of
Adolescent Health (Add Health) (Udry 2003). Add Health was mandated by congress
and funded by the National Institute of Child Health and Development (NICHD) and 17
other federal agencies. The National Opinion Research Center (NORC) at the University
of Chicago conducted the fieldwork, consisting of interviews with adolescents, their
parents, friends, siblings, school administrators, and romantic partners. The data are
nationally representative and longitudinal. Wave I of the survey contains adolescent
respondents in grades 7-12, interviewed between April and December of 1995. The
adolescents were surveyed again approximately one year later between April and August
of 1996. The most recent wave of interviews (Wave III) was completed between August
2001 and April 2002 when respondents were 18-26 years of age. Additionally, only
those students in grades 11 and 12 in Wave I, who were also surveyed in Wave III are
used. Restricting the data in such a way allows for an adequate number of respondents to
graduate from college prior to the Wave III collection of data. The final number of
respondents used in this analysis is 4,259.11
11 It is important to point out that the Add Health dataset draws its sample from students currently enrolled in secondary schools. Due to high school drop-outs, the data is not nationally representative of all
72
3.6 Methodological Issues
There are two central aspects of the design of this study that substantially improve
our confidence in the results. First, the longitudinal design corrects for past research
done with large-scale, cross-sectional secondary data sources, such as the General Social
Survey, in which the issue of self-selectivity on the outcome of study is always a
problem. For example, it is difficult to know whether higher education really causes
decreased orthodoxy and increased church attendance, or whether these outcomes are
simply artifacts of changes that were in the works before entrance into college. While
researchers attempt to control for this possibility in cross-sectional data, longitudinal data
containing information on the outcome at two separate time points is ideal (In this
particular case, religious data collected prior to college attendance and then immediately
after college graduation is best.).
Secondly, the data were collected from high school juniors and seniors, and this
provides us with a control group that does not graduate from college. Both longitudinal
and cross-sectional data collected from college freshman make it difficult to assess
whether changes in religious belief and practice are actually the result of maturation or
life-course processes associated with early adulthood as opposed to causal effects
associated with college attendance. Perhaps young adults go to church less because they
have left the family home, not because of the college environment. This is difficult to
parse out without a non-college control group. In addition, historical events (e.g.
September 11, 2001) could leave their mark on all students and falsely be attributed to the
college environment without a control group.
secondary school-aged youth. Additionally, it should be noted that attrition of approximately 27% of the total sample occurred between Wave I and III of the data collection.
73
3.7 Dependent Measures
The dependent variables in this analysis are four religious measures gathered in
Wave III. The first measure is an indicator of the respondent’s public religious practice:
religious service attendance within the past 12 months. There were seven response
categories ranging from “never” to “more than once a week”. The second measure of
religiosity is a measure of religious saliency: self-rated importance of faith. The
respondent was asked how important his or her religious faith was. Four response
categories ranging from “not important” to “more important than anything else” were
available as answers. As an indicator of private religious practice, frequency of prayer is
included as a dependent variable (the survey question specifically asks the respondent not
to include prayer in a church service). There were eight possible response categories
ranging from “never” to “more than once a day”. The fourth measure is a dichotomous
variable indicating whether or not the respondent had a religious affiliation at Wave III.12
3.8 Independent Measures
The primary independent variables in this analysis are measures of college
attendance and college graduation. The variable “some college” indicates that the
12 While the dependent variables are selective indicators of the key concepts of practice, salience, and identity, they enjoy a long precedent in sociological research using population surveys. It is not the intention of this research to establish the psychometric validity and reliability of these indicators, although their shortcomings in this regard are acknowledged.
74
respondent has attended at least some college and either has a degree less than a
bachelor’s degree, or does not have a degree at all. The second main independent
variables is a dichotomous variable that codes those who have obtained a bachelor’s
degree or higher as “1” and all other respondents as “0”. The time elapsed between the
Waves I and III is approximately six years, enough time for both juniors and seniors in
high school to have completed the requirements for a bachelor’s degree. Approximately
twenty-five percent of the respondents in this analysis had completed a bachelor’s degree
by Wave III. Separate models will also look at the interaction of this variable with the
following: gender, college major (social sciences or humanities), Catholic identity,
conservative Protestant identity, and race (African American).
The primary controls in this analysis will be religious beliefs and practice taken
from the Wave I interview. To control for prior religious practice and belief, the
following measures are included: (1) attendance at religious services, (2) frequency of
private prayer, (3) self-rated importance of faith, (4) a dummy variable indicating
Catholic affiliation, (5) a dummy variable indicating conservative Protestant affiliation13,
and (6) a dummy variable indicating that the respondent considers the scriptures to be the
“word of God, without any mistakes”.
In addition to the primary controls, additional control variables will be included
from both Wave I and Wave III. From Wave I, academic skill is controlled for using the
percentile ranking from a shortened version of the Peabody Picture Vocabulary test
administered to all respondents. Controlling for the respondent’s prior achievement level
13 Conservative Protestant identity is measured as adherence to Adventist, Assemblies of God, Baptist, Pentecostal, or Holiness categories for non-Black respondents at Wave I of the survey.
75
separates the effects of attending college from personal academic skill. Also,
dichotomous variables for gender and race (African-American/Other) are included as
controls.
Familial measures are also included from the Wave I as controls. Consistent with
past research on intergenerational religious transmission, it is hypothesized that
adolescents whose parents are religious, and adolescents who feel close to their parents,
will be more likely to retain their religious beliefs into young adulthood (Bader and
Desmond 2006; Hunsberger 1983; Ozorak 1989; Sherkat and Wilson 1995). The
religious beliefs of the parent or guardian answering the survey are therefore included as
controls. The specific measures are: religious service attendance, frequency of private
prayer, and self-rated importance of faith. In addition to these, variables indicating how
close the adolescent feels to his/her parents and if the adolescent has attended religious
services with his/her parents within the past four weeks are included as controls. Lastly,
family socio-economic status is included as a control. The measure of family SES
includes parent’s income and parent’s educational attainment.
Wave III controls consist of life course events that occur between the Wave I and
Wave III interviews that could potentially confound the relationship between college
attendance and religious outcomes. A variable indicating whether or not the respondent
is currently living at home is included in the analyses. Living with parents is
hypothesized to result in higher levels of religious practice as it reduces the opportunities
to deviate from the religious identity of the family. Research on the life course has
identified both marriage and having children as events that increase levels of religiosity
(Hoge and Hoge 1984; Sandomirsky and Wilson 1990; Wilson and Sherkat 1994). Both
76
will be included as dichotomous variables in the analysis. The literature also documents
cohabiting with an unmarried partner resulting in lower levels of religious participation
(Stolzenberg, Blair-Loy, Waite 1995). A variable indicating whether the respondent is
currently cohabiting will also be included.
In addition, high social status has been associated with adherence to mainline
Protestant denominations (Niebuhr 1929; Park and Reimer 2002; Roof and McKinney
1987), and hence more liberal religious beliefs (although this relationship may be
weakening). Personal income at Wave III has been included in this analysis as an
indicator of social status. Lastly, a variable indicating whether the respondent has
participated in any unpaid volunteer work within the last 12 months is included.
Volunteer work is used as proxy for community involvement and should decrease the
positive relationship between higher education and church attendance based on the theory
that increased church attendance is the result of civic activism and not religious piety.
Missing data are handled via multiple imputation methods.14 A table of descriptive
statistics is available upon request.
14 The ICE software program available in STATA was used to impute missing values (Royston 2005). ICE uses a maximum likelihood procedure to fill in the most likely values for missing data. In regression analysis, the standard errors of the regression coefficients are correctly computed by statistically including a degree of uncertainty for the values that have been imputed. To avoid losing these and other respondents, most missing values have been imputed. Unfortunately, missing data can only be validly imputed if the type of missingness is missing completely at random (MCAR) or missing at random (MAR). Data that does not fit the MCAR or MAR criteria are referred to at non-ignorable (NI). Non-ignorable missingness occurs when missing data depends on some unobservable variables. Due to a programmed skip pattern in Wave I of Add Health, religiously unaffiliated respondents, and religious unaffiliated parents of respondents, were not asked any of the explicitly religious questions. We believe this data is best classified as NI, and have applied list-wise deletion for these cases. Approximately 15% of cases have been deleted. For more information on multiple imputation please consult King et al (2001).
77
3.9 Results
Tables 3.1 through 3.3 present OLS regressions predicting religious service
attendance, self-rated importance of faith, and frequency of prayer when most of the
respondents were age 22 through 25 (Wave III).15 16 Table 3.4 presents odds ratios
predicting no religious affiliation at age 22 to 25. Each table follows a similar modeling
strategy. Model 1 displays the group mean differences on the religious outcome by
education level. Model 2 provides the adjusted group differences by education net of
demographic and life course controls. The lagged dependent variable at age 16-19 (Wave
I) is added to model 3.17 Model 4 adds in additional individual and familial religious
variables. Finally, model 5 provides a series of interaction effects by education level.
3.9.1 Religious Service Attendance
Consistent with past research, Table 3.1 reveals that those respondents with a
college degree also tend to attend religious services more regularly. As expected, Model
2 suggests African Americans, females, those who felt close to their parent(s) at age 16-
19, those currently in their first marriage, those who have children, and those who
volunteer are more frequent attendees. On the other hand, those who scored high on the
Peabody Picture Vocabulary test, and those who are currently cohabiting attend less
15 Ordinal regression techniques were also tested with no substantial differences evident. For ease of interpretation, OLS regression has been used in these tables.
16 Approximately 98% of the sample used was between the ages of 22 and 25 at Wave III of the survey. More than 80% were either 23 or 24 years of age.
17 Approximately 99% of the sample used was between the ages of 16 and 19 at Wave I of the survey. More than 80% were either 17 or 18 years of age.
78
TABLE 3.1
OLS REGRESSION
PREDICTING FREQUENCY OF CHURCH ATTENDANCE
AGE 22-25
(1) (2) (3) (4) (5) Education Some College .277* .247* .200 .185 .029 Bachelor’s Degree or More .805*** .643*** .518*** .495*** .408* Wave 1 General Controls Vocabulary Score -.004* -.004* -.004* -.004* Female .372*** .322*** .284*** .390** Black .739*** .491*** .199 .123 Feels Close to Parents .125* .102 .063 .060 Parent’s Income (logged) .046 .023 .051 .056 Parent’s Educational Attainment .030 .013 .010 .009 Wave 3 General Controls Currently Living in Parent’s Home .180 .180* .187* .184* Currently in First Marriage .782*** .643*** .550*** .543*** Have Children .176* .247** .260*** .247** Currently Cohabiting -.699*** -.536*** -.438*** -.435*** Personal Income (logged) -.014 -.002 -.000 -.001 Volunteered During Past Year .634*** .520*** .479*** .492*** Lagged Dependent Rel Service Attendance Age 16-19 .666*** .261*** .260*** Individual Religion (Age 16-19) Importance of Faith .199** .209*** Frequency of Prayer .127*** .126*** Bible is the “Word of God” .071 .056 Catholic -.056 -.245 Conservative Protestant .226* .056 Family Religion Parent’s Church Attendance .215*** .213*** Parent’s Importance of Faith .139 .142 Parent’s Frequency of Prayer -.004 -.007 Attends Church With Parents .427*** .429***
79
SOURCE: National Longitudinal Study of Adolescent Health *p<.05 **p<.01 ***p<.001
frequently. The various controls present in Model 2 slightly diminish the relationship
between education and church attendance.
Model 3 adds a measure of church attendance from Wave I of the survey. As
expected this is highly correlated with the similar Wave III measure. The addition of the
earlier measure of church attendance diminishes the effect of having a BA further, while
rendering those with some college non-significant from those with no college. However,
most of the other significant findings from Model 2 maintain their significance (with the
exception of feeling close to parents). This suggests that blacks and females do not suffer
as large of a change in church attendance between late high school and early adulthood
compared to non-blacks and males. Currently living with parents teeters from non-
significance to significance at the .05 level with the addition of the lagged dependent
variable.
TABLE 3.1 (Continued)
(1) (2) (3) (4) (5) Differential Effects Female * Some College -.143 Female * BA or More -.169 Soc Sci/Human Maj * Some College .142 Soc Sci/Human Maj * BA or More -.111 Catholic * Some College .358 Catholic * BA or More .171 Black * Some College -.092 Black * BA or More .451 Conserv Protestant * Some College .278 Conserv Protestant * BA or More .268 Constant 1.909 .991 -.666 -1.921 -1.851 N 4259 4259 4259 4259 4259 R2 ( significance of R2 change) .026 .150*** .274*** .325*** .329
80
Model 4 adds both individual and family measures of religion. Several of these
variables are significant predictors of church attendance at age 22-25, even after
controlling for prior church attendance. The addition of these other dimensions of
religiosity reduces the effect of church attendance at age 16-19 by more than 60%. These
variables also reduce the effect from African Americans to non-significance.
Nevertheless, the relationship between those who have a BA and the frequency of
religious service attendance is nearly unaltered by these controls. Model 5 includes
several variables that interact education with gender, college major, religious
identification, and race. No significant differences are found.
3.9.2 Importance of Faith
Unlike church attendance, Table 3.2 suggests that there is no overall difference in
importance of faith by level of education. Despite this, the other variables behave
similarly in Table 3.2 and Table 3.1. Model 2 finds importance of faith higher for blacks,
females, those who felt close to their parents, those currently in their first marriage, and
those who volunteer. Individuals with higher vocabulary scores, as well as those who
cohabit, claim faith is not as important to them. The introduction of the lagged dependent
variable in Model 3 reduces the effect of the vocabulary test and feeling close to parents
to non-significance. The effect from gender and African American are also reduced, yet
still significant at the .001 level. As in Table 3.1, the additional religious variables
included in Model 4 reduce the partial correlation between importance of faith at Wave I
and III. Interestingly enough, neither individual church going in high school or parental
church going affect how important faith is early adulthood. However, attending church
81
TABLE 3.2
OLS REGRESSION
PREDICTING IMPORTANCE OF FAITH
AGE 22-25
(1) (2) (3) (4) (5) Education Some College -.025 -.000 -.019 -.016 -.081 Bachelor’s Degree or More .046 .063 .046 .045 .070 Wave 1 General Controls Vocabulary Score -.002** -.001 -.001* -.001 Female .166*** .099*** .108*** .102* Black .347*** .195*** .096 .024 Feels Close to Parents .074** .021 .024 .026 Parent’s Income (logged) -.050 -.029 -.028 -.025 Parent’s Educational Attainment .003 .002 -.002 -.002 Wave 3 General Controls Currently Living in Parent’s Home -.009 -.011 -.007 -.009 Currently in First Marriage .268*** .197*** .163*** .156*** Have Children .009 .021 .031 .027 Currently Cohabiting -.205*** -.127*** -.101** -.101** Personal Income (logged) -.010 -.007 -.005 -.005 Volunteered During Past Year .169*** .116** .106** .109** Lagged Dependent Importance of Faith at Age 16-19 .420*** .269*** .270*** Individual Religion (Age 16-19) Religious Service Attendance .023 .025 Frequency of Prayer .054** .055** Bible is the “Word of God” .075* .066* Catholic -.035 .029 Conservative Protestant .112** .093 Family Religion Parent’s Church Attendance .013 .012 Parent’s Importance of Faith .122*** .120** Parent’s Frequency of Prayer .013 .014 Attends Church With Parents .107* .118**
82
SOURCE: National Longitudinal Study of Adolescent Health *p<.05 **p<.01 ***p<.001
regularly with parents during high school seems to increase importance of faith, net of
other factors. In all models, education level is unrelated to self-rated importance of faith.
Model 5 uses interaction effects to test whether higher education differentially
effects key subpopulations. Unlike the similar test in the regression predicting church
attendance, this model finds some differences between groups. Catholics are less likely
to say their faith is important to them if they graduated from college. Catholics who do
not go to college are significantly different from their college counterparts, and the
similar degree of decline in saliency of faith is not evident. Conversely, Blacks who
graduate from college are significantly more likely to claim their faith is important to
them compared to those who do not go to college. Model 5 also reveals that females are
more religious than males, regardless of whether or not they go to college. Neither
conservative Protestant affiliation, nor majoring in the social sciences or humanities, have
TABLE 3.2 (continued)
(1) (2) (3) (4) (5) Differential Effects Female * Some College .007 Female * BA or More .021 Soc Sci/Human Maj * Some College .005 Soc Sci/Human Maj * BA or More .011 Catholic * Some College .045 Catholic * BA or More -.274** Black * Some College .082 Black * BA or More .205* Conserv Protestant * Some College .071 Conserv Protestant * BA or More .009 Constant 1.517 1.366 .145 -.297 -.303 N 4259 4259 4259 4259 4259 R2 ( significance of R2 change) .006 .105*** .255*** .291*** .299**
83
significant interactions with educational variables. It is also important to note that while
group differences in the form of interaction effects are indeed significant, the contribution
to the overall variance explained is minimal.
3.9.3 Frequency of Prayer
As with Table 3.2, Table 3.3 finds no difference in frequency of prayer by
educational attainment. This lack of difference holds up across the various models.
Model 2, once again, finds differences by vocabulary score, race, gender, and closeness
toward parents in high school. The life course effects of marriage and cohabiting are also
evident in this table as in prior analyses. Living with parents and having children are
non-significant. Volunteer activity has a positive and significant effect, while personal
income measured in Wave III has a negative effect.
Most of these findings hold up (albeit with diminished effect size) when the
lagged dependent variable is added in Model 3. Cohabitation and income are no longer
significant in this model. Model 4 adds the rest of the religion variables. As in the
analysis predicting importance of faith, prior church attendance does not influence later
frequency of prayer. It is also important to note that unlike the previous two tables,
measures of parents’ religiosity have no influence on later frequency of prayer (including
how frequently parents pray). Model 5 finds no significant differences in any of the
interaction effects measured.
84
TABLE 3.3
OLS REGRESSION
PREDICTING FREQUENCY OF PRAYER
AGE 22-25
(1) (2) (3) (4) (5) Education Some College .171 .216 .135 .166 -.150 Bachelor’s Degree or More .347 .303 .172 .249 .160 Wave 1 General Controls Vocabulary Score -.005* -.004 -.003 -.003 Female .816*** .585*** .601*** .417* Black 1.220*** .904*** .553*** .494 Feels Close to Parents .285*** .195** .153* .154* Parent’s Income (logged) -.097 -.092 -.052 -.050 Parent’s Educational Attainment .028 .019 .015 .015 Wave 3 General Controls Currently Living in Parent’s Home
-.022 .017 .005 -.010
Currently in First Marriage .729*** .572*** .453*** .428*** Have Children .108 .151 .159 .166 Currently Cohabiting -.346** -.137 -.063 -.065 Personal Income (logged) -.043** -.027 -.025 -.025 Volunteered During Past Year .684*** .481*** .464*** .479*** Lagged Dependent Frequency of Prayer at Age 16-19 .761*** .555*** .557*** Individual Religion (Age 16-19) Religious Service Attendance .038 .039 Importance of Faith .351*** .359*** Bible is the “Word of God” .145 .136 Catholic -.087 -.187 Conservative Protestant .383** .346 Family Religion Parent’s Church Attendance .002 -.003 Parent’s Importance of Faith .154 .140 Parent’s Frequency of Prayer .124 .126 Attends Church With Parents .118 .136
85
SOURCE: National Longitudinal Study of Adolescent Health *p<.05 **p<.01 ***p<.001
3.9.4 Religious Disaffiliation
Table 3.4 presents the odds ratios of reporting no religious affiliation on Wave III
of the survey.18 Once again, there are no differences across any of the models by
educational attainment. Model 2 suggests that females, those who feel close to their
parents in high school, those still living with their parents in early adulthood, and those in
their first marriage all are less likely to disaffiliate from a religious tradition. Unlike all
of the other analyses carried out, cohabitation and volunteer activity have no effect on
disaffiliation. The effect from the vocabulary test becomes non-significant once religious
variables are added in Model 3. Similar to the analysis predicting frequency of prayer,
18 Approximately 99% of the sample used was between the ages of 16 and 19 at Wave I of the survey. More than 80% were either 17 or 18 years of age.
TABLE 3.3 (continued)
(1) (2) (3) (4) (5) Differential Effects Female * Some College .303 Female * BA or More .304 Soc Sci/Human Maj * Some College -.191 Soc Sci/Human Maj * BA or More -.106 Catholic * Some College .504 Catholic * BA or More -.297 Black * Some College -.049 Black * BA or More .341 Conserv Protestant * Some College .098 Conserv Protestant * BA or More .058 Constant 3.712 2.450 -.062 -1.836 -1.690 N 4259 4259 4259 4259 4259 R2 ( significance of R2 change) .003 .119*** .262*** .291*** .296
86
TABLE 3.4
LOGISTIC REGRESSION
PREDICTING NO RELIGIOUS AFFILIATION
AGE 22-25
(1) (2) (3) (4) Education Some College .882 .798 .845 .797 Bachelor’s Degree or More .798 .719 .731 .532 Wave 1 General Controls Vocabulary Score 1.007** 1.005 1.005 Female .632*** .669** .700 Black .722 1.078 1.336 Feels Close to Parents .740** .796* .783* Parent’s Income (logged) .895 .878 .870 Parent’s Educational Attainment 1.003 .982 .982 Wave 3 General Controls Currently Living in Parent’s Home .643** .652** .658** Currently in First Marriage .536** .610* .631* Have Children .957 .935 .942 Currently Cohabiting 1.115 .905 .917 Personal Income (logged) .995 .982 .980 Volunteered During Past Year .821 .914 .898 Individual Religion (Age 16-19) Religious Service Attendance 1.003 .999 Importance of Faith .654*** .647*** Frequency of Prayer .883* .878* Bible is the “Word of God” .628** .639** Catholic .608* .357** Conservative Protestant .817 .770 Family Religion Parent’s Church Attendance .839 .847 Parent’s Importance of Faith .830 .840 Parent’s Frequency of Prayer .963 .956 Attends Church With Parents 1.116 1.097
87
TABLE 3.4 (continued)
(1) (2) (3) (4) Differential Effects Female * Some College 1.005 Female * BA or More .791 Social Science/Humanities Major * Some College 1.903 Social Science/Humanities Major * BA or More 1.634 Catholic * Some College 1.593 Catholic * BA or More 3.253* Black * Some College .818 Black * BA or More .308* Conservative Protestant * Some College .959 Conservative Protestant * BA or More 1.411 N 4259 4259 4259 4259 Pseudo R2 (asterisks indicate sig. of R2 change) .001 .048*** .097*** .124* *p<.05 **p<.01 ***p<.001
SOURCE: National Longitudinal Study of Adolescent Health NOTE: Coefficients are odds ratios *p<.05 **p<.01 ***p<.001 only individual religious variables prove significant. Additionally, church attendance in
high school does not affect the likelihood of disaffiliation in early adulthood (nor did it
affect frequency of prayer or importance of faith in previous tables).
Lastly, in the interaction effects of Model 4, we see similar patterns to the analysis
predicting importance of faith. Catholics, who are less likely overall to disaffiliate, seem
to be less likely only if they do not graduate from college. Catholics who graduate from
college are significantly more likely to disaffiliate. The opposite trend is apparent for
African Americans. Graduating from college results in blacks being significantly less
likely to disaffiliate compared with blacks who do not attend college at all. No
significant interaction by gender, college major, or conservative Protestant affiliation is
evident.
88
3.10 Discussion
Several important findings arise from this research. First, and perhaps most
importantly, recent college graduates do not seem to be secularizing at a faster rate than
their non-college attending counterparts. In other words, the declines on various
measures of religious faith and affiliation associated with young adulthood do not seem
to be stemming from the college environment. Gauged in terms of religious attendance,
salience of faith, frequency of prayer, and religious affiliation, college graduates never
measure lower in early adulthood than those did not attend college at all. For the
outcomes examined, Hypothesis 1 cannot be accepted. Nonetheless, it is important to
point out that these measures are certainly not exhaustive in terms of capturing the many
dimensions of religion and spirituality. Other measures may very well behave differently
than the outcomes examined.
Moreover, college does not affect the religious practice and identity of the entire
population uniformly. Hypothesis 2 posits that educational attainment may produce
gender convergence on religious outcomes. No such evidence is garnered from the
outcomes studied. While college educated individuals may adopt more progressive
political and moral stances, this does not seem to result in college educated men and
women with similar religious outcomes. Rather, women appear to be more religious than
men regardless of whether or not they attend college. What the tables do suggest,
however, is that being female is still a significant predictor of religiosity at age 22-25,
even after controlling for prior religiosity. This indicates that women do not evidence the
same degree of religious falling out that men do during early adulthood. Moreover,
college attendance does not seem to shake this female advantage in religious retention.
89
Hypothesis 3 should also be rejected. Majoring in the social sciences or
humanities has no significant effect on religious outcomes at age 22-25. Faculty in these
fields have been documented as the least religious in the university. Nevertheless, once
initial religious status in high school is controlled for, these students look no different
from the rest of the college population in their religious practice and identity. In many
ways, this may be the most surprising finding. It could be hypothesized that college no
longer has a secularizing effect because of the growth of business departments and other
religiously innocuous developments at the university. Nevertheless, in this formulation it
still seems reasonable to expect certain arenas of the university to undermine religious
authority. Perhaps faculty are careful not to convey their religious ideas, or equally
likely, students are less influenced by these ideas than many have suspected. More
detailed study is certainly warranted to understand why secular faculty have little
influence over their students.
Partial support is found for Hypothesis 4. Respondents who identified themselves
as Catholic in Wave I traveled a different religious trajectory compared to both their non-
Catholic college-educated counterparts as well as their Catholic peers who did not attend
college. For college-educated Catholics, faith becomes less important and their identity
as Catholics is not as firm. I have suggested that this may be due to the anti-institutional
and anti-authoritarian attitudes accumulated during the college experience. These
particular attitudes may be more problematic for young Catholics compared to other
religious groups who have more options for switching and greater flexibility in their ideas
about organized religion. Nevertheless, it important to point out that declines in mass
attendance are actually less severe for Catholics who graduate from college. However,
90
this is not unique to Catholics; all groups who graduate college attend religious services
at higher rates than those who did not attend college at all.
Hypothesis 5 expected that conservative Protestants would secularize more than
other religious groups as a result of college. Contrary to this expectation, no significant
interaction effects between educational attainment and conservative Protestant identity
were found. While this study was not able to measure aspects of the college experience,
it is possible that the institutional support for conservative Protestantism in the form of
explicitly religious schools (such as those belonging to the Coalition of Christian
Colleges and Universities) or strong parachurch organizations (such as InterVarsity
Christian Fellowship and Campus Crusade for Christ) provide the resources needed to
maintain conservative Protestant belief and identity during college. Once again, more
research in this area would be better able to establish the processes involved in
conservative Protestant identity maintenance during college.
In addition, some significant interactions for educational attainment and African
Americans are evidenced in the analyses. However, these interactions were in the
opposite direction than what is posited in Hypothesis 6. College educated blacks are
more likely to claim their faith was important and less likely to disaffiliate from a
religious denomination compared to other college graduates and blacks who did not go to
college at all. This finding calls attention to the manner in which black identity and
religious identity may reinforce one another during college. Future research is necessary
to understand the processes that trigger educational attainment to open up religious
trajectories in early adulthood for African Americans.
91
Specific limitations of this study provide important departing points for future
research. While the virtues of panel data over cross-sectional data are considerable,
caution should still be practiced in concluding that the college environment directly
causes any differences demonstrated in these analyses. Future research should attempt to
unpack all of the meanings young adults attach to educational attainment and how these
meanings interact with religious identity. Separating indicators of status associated with
educational attainment from the actual content and values learned from the college
experience is an important first step in this process.
Lastly, the lack of data on the institution of higher education is a shortcoming that
has theoretical implication that could be addressed in future work. Hammond and Hunter
(1984) found religious students fared differently at religious and non-religious schools.
Unfortunately, the Add Health data currently contains no information on the type of
school attended by the respondent. Examining the type of school, as well as
characteristics of the school environment would also serve to shed some light on the
relationship between higher education and religious identity and practice.
3.11 Conclusion
This chapter systematically analyzed the role higher education plays in the
religious decline associated with the transition to adulthood. As higher education
expands and more students graduating from high school delay their entrance into the
occupational field and opt to continue their education, it is important to attempt to
quantify the secularizing effect college may have on young adults. Using panel data with
92
multiple religious measures, the conclusion reached in this analysis suggests there is no
overall secularizing effect. This finding contradicts research conducted with older data
(see Funk and Willits 1987; Sherkat 1998). Among recent birth cohorts, those who
graduate from college look quite similar to those who do not on outcomes such as prayer,
self-reported importance of faith, and rates of religious disaffiliation. Rates of religious
service attendance are higher across the board for college graduates compared to the rest
of the population. Catholics, on the contrary, do seem to disaffiliate more and report that
their faith is less important to them if they are college graduates. However, African
Americans actually appear to do increase salience of faith and retain their religious
affiliation at a higher rate if they graduate from college.
93
CHAPTER 4:
RELIGIOUS PARTICIPATION DURING COLLEGE:
DO RELIGIOUS SCHOOLS MAKE A DIFFERENCE?
4.1 Overview
A growing body of research has been addressing the religious lives of college students in
the United States. Scholars have documented high levels of interest in spirituality among
undergraduates (Cherry, Deberg, and Porterfield 2001; Higher Education Research
Institute 2004; Lee, Matzkin, and Arthur 2006), growing campus ministry and parachurch
organizations (Bramadat 2000; Cawthon and Jones 2004; Hunt and Hunt 1991;
Schmalzbauer 2007), and the resurgence of the academic study of religion (Leege and
Kellstedt 1993; Sherkat and Ellison 1999; Stark and Finke 2000). In one national survey,
nearly three times as many students report becoming more religious during college as
opposed to less religious (Lee 2002). Moreover, work with nationally representative
panel data has confirmed that recent college graduates are not secularizing at faster rates
than their non-college counterparts (Hill 2007; Uecker, Regnerus and Vaaler 2007).19
19 The research documenting the religious climate on campuses should not be equated with some sort of return to traditional religious belief and practice. Scholars have documented that most students do not emphasize public religious identities (Clydesdale 2007), and that a respect for religious pluralism is the overarching norm (Higher Education Research Institute 2004). The role of religion in the academy has also shifted. Whereas religious studies have proliferated in recent years, the specifics of religious content do not often inform the substantive domains of study (Marsden 1994; Wuthnow 2007). Rather, religion is often studied through the lens of the humanities and social sciences.
94
While this work has advanced our understanding of religious life during college,
institutional level effects have largely gone unexamined. Extant research has found that
fundamental institutional types (e.g. public, private liberal arts, religiously affiliated)
shape student change in spheres as diverse as cognitive growth (Bohr and Pascarella
1994; Bohr et al 1995), moral development (Good and Cartwright 1998; McNeel 1994),
and economic earnings (Monks 2000). Nevertheless, these between-college effects have
received little attention in the realm of religious outcomes. This research begins to
address this gap in the literature by using nationally-representative panel data to examine
changes in religious service attendance during the college years. More specifically, three
key questions are addressed: (1) Are students who are currently enrolled in college more
or less likely to attend religious services than those who are not attending college?; (2)
Among those currently enrolled, do religious institutions develop student religious
practice, or do religious students fare better at non-religious schools?; And, (3) do key
subpopulations (as measured by religious tradition, gender, and race/ethnicity) differ in
how their religious practice is shaped by the institutional characteristics of the college?
4.2 Church Attendance and Higher Education20
Research has consistently found that educational attainment is positively related
to religious service attendance (Cohen 1983; Cornwall 1989; Gaede 1977; Mueller and
Johnson 1975; Roof 1978). However, the causal direction between education attainment
20 Although more diverse measures of religious belief and practice would be ideal, the data limits this analysis to religious service attendance. Nevertheless, religious service attendance is a key sociological measure that taps into commitment to a religious community. For the college student, largely free from parental religious demands, measures of voluntary participation in a religious community provide a glimpse of the priority given to religious life over other spheres of influence in early adulthood.
95
and religious service attendance is not clear. Several studies demonstrate that youth
involved in religious communities positively interact with adults and peers in such a
manner as to encourage pro-social behavior such as avoiding deviance and pursuing
academic trajectories (Loury 2004; Smith 2003; Muller and Ellison 2001). Moreover,
there is evidence that students who regularly attend religious worship services do better
on a variety of academic outcomes while enrolled in college (Mooney 2005). These
findings suggest that at least some of the relationship between educational attainment and
religious involvement is due to religiously involved students being more likely to attend
and graduate from college.21
An alternative explanation has been posited that suggests educational attainment
is coupled with civic engagement expectations that result in educated individuals joining
religious organizations (Hoge and Roozen 1979; Roof 1978). Recent studies that control
for initial religious attendance prior to college confirm that there is an independent effect
from educational attainment on religious service attendance in early adulthood (Hill
2007; Uecker et al 2007). From this I develop my first hypothesis:
H1: Net of self-selection mechanisms, graduating from college will, on
average, result in increased levels of religious service attendance.
Despite the positive relationship between educational attainment and church
attendance, precipitous drops in religious attendance have been documented during the
college years. Bryant, Choi, and Yasuno (2003) report that 46.4% of entering freshman
21 Although regularly attending youth are more likely to attend and graduate from college, youth with fundamentalist and Pentecostal orientations are less likely to attain a college degree (Beyerlein 2004; Darnell and Sherkat 1997; Glass and Jacobs 2005; Sherkat and Darnell 1999).
96
report attending religious services frequently. By the end of the first year, the percent
regularly attending drops to 27.1%. Likewise, the number of students not attending at all
climbs from 16.4% to 43.4% during the first year. Although somewhat dated, Astin
(1977) finds a freshman to senior decline in church attendance between 31 and 35
percent. Freshman to senior declines are also found in Clark (1972) and McAllister
(1985).
While declines in religious service attendance during the college years could be
due to disenchantment with institutional religion, several factors lead us to believe that
the sources may be rooted in the social arrangements and institutional constraints
associated with student life, and not in the abandonment of religious beliefs. For those
who leave home and reside on a college campus, the geographic move itself may mean
cutting ties with one religious community. Residing on campus also commonly means
that family influences on religious involvement are lessened (Astin 1972; Astin 1977).
Moreover, many college campuses provide social and religious opportunities that
functionally replace the role of a religious community by providing opportunities for
religious practice, friendships, social support, and potential mates (Cherry et al 2001).
The social forces that push young adults to be active in a church or other place of worship
are diminished on college campuses. It is likely that students who attend worship
services during college are primarily comprised of individuals who entered college with
strong normative beliefs about the importance of regular corporate worship in their
religious tradition. Thus, my second hypothesis:
H2: Young adults who attend college will temporarily exhibit lower levels of
church attendance compared to young adults not in college.
97
4.3 The Relationship Between Institutional Identity and Student Religious Practice
More than 35% of all four year colleges and universities in the United States
claim some sort of religious affiliation. Moreover, most of these religiously affiliated
institutions make some sort of explicit reference to moral or spiritual development in
their mission statements. I expect that strongly religious schools would do a better job
than non-religious schools nurturing the religious and spiritual life of students as religious
socialization is central to the identity of many religiously affiliated colleges. The
religious background of the student and the religious mission of the school positively
interact to create networks of religious support and encouragement. From this
perspective, I expect that students attending such schools display less precipitous drops in
overall religious service attendance despite being away from family and home church.
Consistent with this perspective, Railsback (2006) finds that seniors attending schools
associated with the evangelical Council for Christian Colleges and Universities were
nearly twice as likely as the average student to report that their religious beliefs and
convictions were stronger since enrolling (44% versus 82%). From this I develop a third
hypothesis:
H3: Students attending strongly religious institutions will be more likely to
maintain pre-college levels of religious practice compared to students
attending less-explicitly religious or non-religious schools.
However, a homogenous religious environment may have unintended
consequences. If the institution provides opportunities for religious activities outside of
regular worship, students may choose to substitute these activities for church attendance.
98
While such an outcome does not necessarily indicate some underlying level of religious
decline, it still represents a move away from the normative expectation of corporate
worship that most religious traditions espouse.
The insular religious environments on some campuses may also breed spiritual
apathy. From such a perspective, WE might suspect that religious heterogeneity on
campus actually encourages students to examine their own religious identities and either
strengthen, revise, or abandon them. Lack of such heterogeneity leaves religious
identities unexamined. Smith (1998) makes the case that evangelical Christians actually
thrive in a religiously pluralistic environment. Consistent with Smith, Hammond and
Hunter (1984) find that evangelicals on secular campuses actually strengthen their faith
over time (the authors find a “slow erosion” of faith on insulated evangelical campuses).
In a case study of an evangelical group at a large Canadian university, Bramadat (2000)
finds that students use the group both as way to protect themselves from secular
influences as well as a tool to connect themselves with the non-evangelical intellectual
and social life on campus. This perspective proposes the counterintuitive notion that
religiously committed young adults may actually develop their religious selves more
attending an explicitly non-religious campus as they must use their religious identities to
engage and cope with secular college life. From this, we develop an alternative
hypothesis:
H3alt: Students who choose to attend non-religious schools will experience, on
average, less decline in religious practice than students attending religiously
affiliated institutions.
99
4.4 Differential Effects by Religious Tradition, Gender, and Race/Ethnicity
Just as the college experience varies across institutional boundaries, it also is not a
uniform experience for students that bring minority identities to campus. Several strands
of developmental theory have emphasized the essential discontinuity between the patterns
of identity development in college for African-Americans and women when compared to
white males (Helms 1995; Josselson 1996). Students that carry self-conscious identities
to college must filter their student experience through these identities. For highly
religious students as well as minority students that claim a religious identity, religious
ideas and practices become part of the tools to negotiate social and intellectual demands
during college. Because of this, religion often becomes more integrated into public
selves for these students (Clydesdale 2007). Nevertheless, very little is known about the
effect from institutional context on the religious practice of these key subpopulations
during college. Below I present some hypotheses outlining how institutional
characteristics may facilitate and constrain religious practice during college for these
students.
4.4.1 Religious Tradition
Although research suggests that religious change is quite common, especially
during adolescence and young adulthood (Smith 2006), it is best understood in the
context of religious trajectories shaped during childhood and early adolescence (Regnerus
and Uecker 2006; Wuthnow 1999). The tools to further religious socialization during
college will be comparably greater when the institution’s religious identity corresponds
most closely with the stock of religious knowledge, rituals and beliefs that students
develop in early life. In the case of conservative Protestants and Catholics, I expect that
100
attending a college or university associated with their religious tradition will provide
them with more opportunities to practice their faith, provide relatively religiously
homogenous friendship networks, and offer more opportunities for formal religious
education. Stated more formally:
H4: Conservative Protestants and Catholic students will display the highest
levels of religious practice while enrolled in institutions that identity with
their own religious tradition.
4.4.2 Gender
Past research indicates that men and women do not differ in the net effect from
educational attainment on religiosity (Funk and Willits 1987; Hill 2007). Nevertheless,
the effects from the religious identity of the college may be masked in these studies, as
only one in ten students are enrolled in a religiously affiliated college (Mahoney,
Schmalzbauer and Younis 2001). For institutions that have strong programs of religious
socialization, males and females may respond differently. Research on religion and
gender consistently finds that men are less amenable to religious socialization than
women (Stark 2002; Walter and David 1998). Men tend to take more risks and engage
more in deviant behavior (Miller and Stark 2002), while women tend to develop
characteristics of nurturance and submissiveness (Mol 1985; Thompson 1991).
Moreover, conservative Protestants belief systems commonly reinforce these traditional
gender characteristics (Peek, Lowe and Williams 1991). While most institutions of
higher education no longer have strong religious socialization programs, the few that do
may find that men are less receptive than women. This leads us to the following
hypothesis:
101
H5: Men attending evangelically affiliated colleges will have lower levels of
religious practice than women attending these institutions.
4.4.3 Race/Ethnicity
There is very little research concerning the effects of college on religiosity for
African Americans and Latinos. Hill (2007) reported that African Americans who
attended and graduated from a four year college place higher value on their faith and are
less likely to disaffiliate from their religious tradition than those with lower educational
attainment. Although not statistically significant, the same study also found that Blacks
with a college degree attended religious services more frequently. Despite these findings,
little is known about why education has a positive effect on African American religiosity.
One possible source of this increase in religiosity could be due to the effect of
Historically Black Colleges and Universities (HBCUs). There is a tight link between
African American congregations and HBCUs with an estimated 86% of black
congregations provide financial support to HBCUs (Lincoln and Mamiya 1990). Despite
lower levels of enrollment in HBCUs among recent black college attendees, HBCUs still
grant 21.5% of all Bachelor degrees earned by Blacks in recent counts (Provasnik and
Shafer 2004). This tight coupling between the Black Church and HBCUs may allow
maintaining traditional faith and practice easier for Blacks who attend HBCUs compared
to attending other institutions. Therefore:
H6: African Americans who attend black Protestant HBCUs will be exhibit
higher levels of religious practice compared to African Americans attending
other institutions.
102
No published research could be found concerning the effect of higher education
on religiosity for Latinos. Nevertheless, we can draw on some existing knowledge to
hypothesize the effect of different college types on Latino religious practice. Most
Latinos identify as Catholic (68%), with a sizable minority identifying as Protestant
(15%). Although Latinos are, on average, more religious than non-Hispanic Whites,
religious practice is often oriented around ethnic particularities with influences from the
charismatic and Pentecostal traditions. A recent study by the Pew Hispanic Center
(2007) finds that approximately two-thirds of Latino worshipers attend place of worship
that have Latino religious leaders, offer services in Spanish, and are comprised of a
Latino majority.
For Hispanic students who identify with ethnic religious beliefs and practices,
finding support during college may be easier in some institutional settings than others.
Among non-Hispanic white Catholics enrolled in four year institutions, nearly 14% are in
Catholic schools. The equivalent number for Hispanic Catholics is 5%.22 Among
evangelical schools associated with the Council for Christian Colleges and Universities
(CCCU), the total Hispanic enrollment is less than 5% of their full time undergraduates.
Seen in terms of the number of Hispanics on campus, the mean number at public four
year colleges and universities is 618 full-time undergraduates. For Catholic schools this
is only 177 students. At CCCU institutions, the average is only 68 Latinos.23 Although
economic restraints may partially explain the low levels of enrollment (Darnell and
Sherkat 1997), the reality remains that college-bound Latinos are not attending religious
22 These estimates are derived from the National Longitudinal Survey of Youth 1997.
23 The Latino enrollment at CCCU institutions and the median number of Latinos at public, Catholic, and CCCU institutions are derived from the Integrated Post-Secondary Education Data System.
103
schools. Because of this, it is unlikely that the religious programs at many Catholic and
evangelical colleges provide institutional support to the ethnic religious practices of
Latino students. I expect the following:
H7: Hispanic students attending Catholic and evangelical colleges will exhibit
greater declines in religious practice, on average, than non-Hispanics
attending the same schools.
4.5 Data and Measures
Data are derived from two national datasets: The National Longitudinal Survey of
Youth 1997 (NLSY97), and the Integrated Post Secondary Data Service (IPEDS). The
NLSY97 is a nationally representative panel survey of youth sponsored and directed by
the U.S. Bureau of Labor Statistics and conducted by the National Opinion Research
Center at the University of Chicago, with assistance from the Center for Human Resource
Research at The Ohio State University (Bureau of Labor Statistics, U.S. Department of
Labor 2006). The survey was first administered to 12 to 16 year old youth in 1997 and
has followed up annually with the most recent available wave in 2004. An additional
supplemental sample of Black and Hispanic youths was also conducted. The full number
of respondents from the representative sample and the over sample is 8,984.
Approximately 84% of the original respondents have been retained in the 2004 survey.
In the present research, data are primarily used from the base 1997 survey as well as
surveys conducted between 2000 and 2004. The geocode supplementary data is used to
identify the respondents’ postsecondary institution.
104
The IPEDS data is a census of postsecondary educational institutions in the
United States. Data on institutional characteristics are collected annually for the IPEDS
and housed at the National Center for Educational Statistics. Data on religious affiliation,
sector, institutional size, racial and gender distribution are used in the present analysis.
4.5.1 Analytic Strategy
While my primary aim is to analyze how students at the differing religious and
secular institutions fare in terms of religious service attendance for students while
enrolled, I begin by establishing the general level of religious service attendance during
college versus the level when not enrolled in college. Once established, I then
graphically examine overall trends over time for students who reported being enrolled at
both religious and non-religious schools. I then use multivariate statistical techniques to
test the robustness of these trends, accounting for both college level characteristics and
individual aging and life course events. Finally, I examine the full multivariate model by
the following subpopulations: (1) White Evangelicals, (2) Catholics, (3) Women, (4)
Men, (5) African Americans, and (6) Latinos.
4.5.2 Measures
The dependent measure used in this analysis consists of an eight category measure
of religious service attendance ranging from “Never” to “Everyday”. The survey
question was first administered in 2000 and has been included in all subsequent surveys.
In the present analysis, each respondent contain religious service attendance data for five
years from 2000 to 2004. Because the eight category measure does not approximate a
105
linear measure of annual church attendance, the church attendance data have been
converted to annual attendance to better suit analyses with linear regression methods.24
The main independent variables mark the religious affiliation or sector of the
college the respondent is enrolled in. Binary variables indicate whether the college is
identified with a non-CCCU evangelical, CCCU, mainline Protestant, Catholic, or black
Protestant denominational group if the college is religiously affiliated.25 If the college is
not religiously affiliated, a set of variables indicating whether they are public, private (not
for profit), or private (for profit). A small number of colleges labeled “Other” have been
removed from analysis (approximately 2% of students enrolled in college). For each
wave of data between 2000 and 2004, these variables are coded to indicate the total
cumulative number of years the respondent has reported attending a college or university
in each group. The affiliations and sectors are derived from the religious affiliation
available in the IPEDS data. The specific denominational affiliations have been grouped
according to the guidelines set forth in Steensland et al (2000).
Additional control variables include institutional characteristics such as a
dichotomous variable indicating required chapel attendance, student body size, status as a
four-year institution, a dichotomous variable indicating whether the university is
classified as academically elite, two measures of racial composition, and a measure of
gender composition. Chapel attendance policies were collected through email and phone
calls to campus ministry or student affairs offices for all 300 religiously affiliated
24 The following conversion was used: “Never” = 0; “Once or Twice” = 1.5; “Less than Once a Month/ 3-12 Times” = 7.5; “About Once a Month / 12 Times” = 12; “About Twice a Month / 24 Times” = 24; “About Once a Week” = 52; “Several Times a Week” = 104; “Everyday” = 365.
25 No respondents attended Jewish institutions. Institutions affiliated with the Church of Jesus Christ, Latter Day Saints (Mormon) were included in the non-CCCU evangelical category. Institutions were classified as black Protestant if they were religiously affiliated and identified as an HBCU.
106
institutions in the NLSY97 sample. If the institution had a policy between 2000 and 2004
requiring students to attend a fixed number of chapel services other than convocation, a
value of “1” is assigned to the institution. Otherwise the institution is assigned a “0”.
Student body size is derived from the available IPEDS data from 2005. The top 51
national universities according to 2007 US News & World Report college rankings have
been classified as elite institutions. Using a similar measure of elite status, Gross and
Simmons (2007) found faculty at these institutions to be considerably more secular in
orientation when compared to other colleges and universities. Racial composition is
measured by the complement of a Herfindahl index which measures the degree of racial
pluralism on campus as well as a measure of the percent Caucasian on campus.26 The
gender composition is measured by the percent of male undergraduates on campus. The
racial and gender compositional measures are derived from the IPEDS data for 2005.
Individual-level variables include religious tradition, graduating with a bachelors
degree, graduating with an associates degree, residing with parents (or parental figures),
living in a dorm or dorm-like setting, marital status, cohabiting status, and whether the
respondent has biological or adopted children. Religious tradition is derived from the
NLSY 1997 wave of data. Two dichotomous variables indicating belonging to a Catholic
or evangelical Protestant denomination were created.27 Two dummy variables were
created indicating whether the respondent had graduated with a bachelor’s degree or
graduated with an associate’s degree. A dichotomous variable was created indicating
whether the respondent was living with at least one parent or parental figure at the time of 26 Using the proportion of Caucasians, African-Americans, Hispanics, and Asians/Pacific Islanders on campus, the racial diversity index takes the following form: 1 - ∑pi
2.
27 The denominational categories were approximated using Steensland et al (2000). Only white respondents were included in the evangelical Protestant category.
107
the survey. Another variable was constructed indicating whether the respondent was
living in a dorm or other group quarters. Past research has also shown that marrying and
having children commonly increase frequency of church attendance (Roozen, McKinney,
and Thompson 1990; Stolzenberg, Blair-Loy and Waite 1995). Likewise, cohabiting
with an unmarried partner has been associated with declines in church attendance
(Thornton, Axinn and Hill 1992). Dichotomous variables indicating each of these states
are included in the models.
Missing data is minimal and handled via list-wise deletion. Slightly over 3% of
the sample is missing in the final models. Table 4.1 provides descriptive statistics.
4.6 Methods
4.6.1 Random Effects Modeling
Standard regression techniques are not valid in panel data with repeated measures.
In OLS regression, the stochastic error term is assumed to random. However, when a
variable is repeatedly measured, the error terms will be correlated within individuals
across time. This violates a fundamental assumption of OLS regression. The random
effects model is one way to deal with this problem. In a random effects model, the
individual level intercepts are allowed to randomly vary between individuals. The model
takes the following form:
yit = β0 + βxit + u0i + εij
where i = 1,…,N individuals; t = 1,…,N waves of data; β0 is the population-level
intercept; βxit is a vector of predictors; u0i is a random intercept; and εij is a random
108
TABLE 4.1
DESCRIPTIVE STATISTICS
Mean Std. Dev. Range Service Attendance (Annually) 20.2162 33.6019 0 to 365 Service Attendance (Truncated) 16.3094 20.4549 0 to 52 Years at Evangelical College (Non-CCCU) .0448 .3478 0 to 5 Years at CCCU College .0342 .3038 0 to 5 Years at Mainline Protestant College (Non-CCCU) .0585 .3879 0 to 5 Years at Black Protestant College .0281 .2728 0 to 5 Years at Catholic College .0795 .4509 0 to 5 Years at Public College 1.7966 1.3648 0 to 5 Years at Private Non-Religious College .1737 .6701 0 to 5 Years at Private for Profit College .0637 .3561 0 to 5 Age 18 (or less) .1064 0 to 1 Age 19 .2105 0 to 1 Age 20 .2587 0 to 1 Age 21 .1936 0 to 1 Age 22 .1351 0 to 1 Age 23 .0684 0 to 1 Age 24 (or more) .0272 0 to 1 Chapel Required .0296 0 to 1 Four Year College or University .664 0 to 1 Elite University .0512 0 to 1 Racial Pluralism Index (Standardized) 0 1 -1.8862 to 2.0412 Percent White on Campus 67.7322 25.9983 0 to 100 Percent Male on Campus 42.7294 9.1394 0 to 100 Educational Attainment: Bachelors Degree or More .051 0 to 1 Educational Attainment: Associates Degree .0325 0 to 1 Living With Parents or Parental Figures .7031 0 to 1 Currently Married .0455 0 to 1 Living with own Kids .039 0 to 1 Currently Cohabiting .063 0 to 1 Evangelical Protestant (reported in 1997) .2958 0 to 1 Catholic (reported in 1997) .3061 0 to 1 Female .5612 0 to 1 Black .2157 0 to 1 Latino .1669 0 to 1 N 13400
SOURCE: National Longitudinal Study of Youth 1997
109
disturbance term. This effectively corrects for the autocorrelation in error at the
individual level.
4.6.2 Fixed Effects Modeling
The majority of models are estimated using individual-level fixed effects. Fixed-
effects, like random effects models, are an effective way to deal with the problem of
autocorrelation. Fixed effects models take the following form in panel data:
yit = βxit + αi + εij
where i = 1,…,N individuals; t = 1,…,N waves of data; βxit is a vector of time-varying
predictors; αi is a vector of time-invariant predictors; and εij is a random disturbance term.
In a fixed-effects model, the αi is treated as fixed for every individual as opposed to
random. Practically speaking, this is equivalent to entering a dummy variable for every
respondent in the survey. Treating these values as fixed eliminates all between-subject
variance and only leaves within-subject variance to be predicted by the time-variant
predictors. There are two disadvantages to the fixed effects method over the random
effects method: (1) Because the model only uses part of the overall population variance
of the dependent variable, standard errors become larger and statistical power is reduced.
(2) Fixed effect models are also at a disadvantage if time invariant characteristics of the
individual are substantively important to what is being tested.
The most obvious advantage to a fixed-effects method over a random effects
method lies in the ability to control for stable, unobservable characteristics of individuals
that commonly result in model misspecification (Allison 2005; Halaby 2004). In my
particular application, this resolves much of the difficulties arising from selection effects.
110
Because respondents have not been randomly assigned colleges, their decision to attend
college at all, or to attend a religious or non-religious school, likely results from
particular individual and familial religious commitments. While some of these religious
commitments and preferences can be measured with survey data, other important
constructs present difficulty. For example, the religiously committed may have a strong
set of intrinsic religious motivations that both affect their likelihood of attending a
religious school as well as a instilling an intention to remain religiously active in the
transition to adulthood. In the absence of reliable and valid indicators of such constructs,
the fixed-effects approach allows us to achieve unbiased estimates.
4.7 Results
4.7.1 Effects from College on Service Attendance
Table 4.2 uses the full dataset to estimate the level of annual religious service
attendance for students currently enrolled and individuals with college degrees
(respondents not enrolled and without a post-secondary degree provide the base for
comparison). Model 1 presents an uncontrolled random effects model. Respondents who
are currently enrolled along with respondents with those who attain an associate’s degree
attend less than those without a degree and not presently enrolled. Those with a BA are
not statistically distinguishable from the reference group.
Model 2 adjusts for aging effects. Net of the population level declines in church
attendance associated with age, students currently enrolled in college have the lowest
levels of annual attendance while those with a BA or more have the highest. The
111
TABLE 4.2
RANDOM AND FIXED EFFECTS
PREDICTING ANNUAL RELIGIOUS SERVICE ATTENDANCE
(1) Rand Effects
(2) Rand Effects + Age
(3) Fixed Effects + Age
Enrolled in College -2.712** -1.058** -3.292** Associates Degree -2.886* 0.866 0.627 Bachelor’s Degree -1.102 3.278** 2.210* # of Observations 38767 38767 38767 # of Individuals 8623 8623 8623 SOURCE: National Longitudinal Study of Youth 1997 +p<.10 *p<.05 **p<.01 (two-tailed) reference group and those with an associate’s degree are in the middle and
indistinguishable from one another. Lastly, Model 3 uses a fixed effects approach to
estimate the group differences in service attendance. Unlike the previous two models,
Model 3 uses only within person variance. The difference between Models 2 and 3
suggests that some of the differences between groups are due to high attending
individuals being more likely to enroll in college and graduate, which is consistent with
past research (Loury 2004; Muller and Ellison 2001). The adjusted effects reveal that
enrolling in college results in something closer to a drop in 3 services annually (rather
than 1) and that graduating with a bachelor’s degree results in an average increase of 2
services (compared to when the individual was not enrolled).
4.7.2 Institutional Differences in Service Attendance
Figure 4.1 displays the estimated number of religious services attended annually
for each year enrolled in both religiously affiliated and non-religiously affiliated
112
institutions of higher education. It is important to note that the strong first year declines
in service attendance reported by Bryant, Choi, and Yasuno (2003) have probably already
occurred for students represented in Figure 4.1. The results are obtained from a random
effects model with dichotomous variables entered in for each year enrolled in each type
of school. A random effects model takes into account the autocorrelation within subjects
between subsequent waves of measurement in panel data and provides more accurate
estimates of church attendance when compared to the alternative of group means. The
random effects regression results are available upon request.
Figure 4.1 Annual Church Attendance By College Type (Random Effects)
Two distinct groups emerge in Figure 4.1: (1) The high attending CCCU,
Evangelical, and black Protestant colleges and (2) the moderate to low attending
113
Mainline, Catholic, Public, and Private colleges. Among the first group, the CCCU
attendees begin their college careers attending worship considerably more often than their
fellow evangelical and black Protestant college attendees. The average worship
attendance for this group of college students is more than once a week. Between years 2
and 3 a sharp drop places them much closer to the other respondents in the high attending
group. By year 4, the mean level of all three groups in the high attending categories
merge at approximately 40 religious services in the past year.
The low attending groups, by contrast, look remarkably similar in religious
service attendance trends. Those who attended Catholic, Mainline, or religiously
unaffiliated public or private higher education began their schooling at an average of
approximately 20 religious worship services a year and display a modest, steady drop in
attendance to 15 services a year by the fourth year of schooling.
Model 1 in Table 4.3 displays the fixed effects regression coefficients for each
cumulative year of education by college type. Consistent with the results displayed in
Figure 4.1, most college students display downward trends in religious service attendance
over subsequent years of schooling. The exceptions to this are students enrolled in black
Protestant colleges and student enrolled at private for profit schools. Neither slope is
significantly different from zero. Also, the sharp drop displayed by students attending
CCCU schools is consistent with the trend lines displayed in Figure 4.1. Even though the
level of decline is substantial for these students, this group still has a high mean level of
attendance in their later years of schooling.
114
TABLE 4.3
FIXED EFFECTS REGRESSION
PREDICTING ANNUAL RELIGIOUS SERVICE ATTENDANCE
(1) (2) College Type (Years Enrolled) Evangelical -3.004* -2.467+ CCCU -8.290** -7.719** Mainline Protestant -2.248* -1.828 Black Protestant 0.436 0.877 Catholic -2.555** -2.009+ Public -1.382** -0.915 Private (non-profit) -2.368** -1.926* Private (profit) -1.032 -0.623 Age 18 (or under) 3.171* 19 1.243 20 0 21 -0.657 22 0.145 23 0.218 24 (or over) 2.078 # of Observations 13400 13400 # of Individuals 4547 4547 R-squared (within) 0.01 0.02
SOURCE: National Longitudinal Study of Youth 1997 +p<.10 *p<.05 **p<.01 (two-tailed test)
The literature has already established that traditional college-aged individuals
experience overall declining levels of religious practice regardless of whether they enroll
in higher education (Willits and Crider 1989). To statistically control for this generalized
decline, I include dummy variables for the age of the respondent, centered around the age
of 20. The decline in attendance is most evident from age 18 to 20. There also appears
to be a slight rebound in church attendance at the age of 24, although this is not
115
statistically different from attendance at age 20. The declines by college type become
slightly more modest in their effect. The declines for those attending mainline and public
colleges or universities become insignificantly different from no decline. The slope
associated with CCCU students is still considerably steeper than any other group, even
after accounting for the age of the respondent.
Before we reach any conclusions about the secularizing influence from this
explicitly religious group of colleges, we need to carefully consider who is declining in
attendance at CCCU schools. Mean level declines in worship attendance could occur
primarily from high attending individuals, moderate to low attending individuals, or both.
As is clear from Figure 4.1, many of the CCCU students must attend worship services
more than once a week. If the decline is primarily due to students attending less often,
but still at least once a week, then these students would still be regularly involved in a
religious community during the college years. It seems plausible to posit that
maintaining an extremely high frequency of religious service attendance may be
logistically difficult when in residence on nearly any college campus. Moreover, those
campuses that house disproportionately high numbers of greater than weekly service
attendees will thus suffer the greatest overall decline. However, if the mean level decline
in service attendance is due to students who attend worship weekly or less than weekly
declining in attendance, than we could rightly assign a secularizing effect evidenced by
these students attending a purposefully religious college.
In order to test this hypothesis, I recode the attendance variable and truncate all
attendance greater than once a week to equal once a week. This forces all variance in the
attendance variable to occur in the weekly or less categories. In addition to testing the
116
hypothesis concerning the level of decline in CCCU schools, I believe that this truncated
variable better captures the essence of the question I have proposed – namely, is decline
or disaffection from church more or less likely at religiously affiliated schools. In this
regard, decline from more than once a week to once a week should be considered a
categorically different sort of decline, as weekly attendance is still the desired norm in
most congregations. If we want to capture declining religious practice that best
approximates what scholars generally mean when referring to individual-level
secularization, then this truncated variable better approximates this type of decline.
Table 4.4 replicates Table 4.3 but uses the truncated attendance variable as its
outcome. Model 1 reveals a substantially smaller effect size (still significant at the .10
level) for students attending campuses affiliated with non-CCCU evangelical
denominations. Students at CCCU schools display no statistically significant overall
drop in attendance, suggesting that the substantial negative effect in Table 4.1 is due
almost entirely to declining levels of attendance in the weekly or greater categories. The
remaining religious schools display similar effect sizes using the truncated attendance
outcome and the outcome that varies across all of the attendance categories. Students
attending mainline Protestant and Catholic schools show statistically significant levels of
decline while students as black Protestant colleges and universities are not significantly
different from a null change in attendance. Students at the non-religious schools all have
significant levels of decline, although they are smaller in effect than the declines of
students at Mainline and Catholic schools. The addition of age categories in Model 2
does little to change the effect size of the primary variables, however non-CCCU
evangelical and private for profit schools both drop from slightly significant to non-
117
significant. The age variables themselves reveal the same “U” shaped movement of
attendance with the lowest point of attendance occurring at age 20.
Model 3 in Table 4.2 includes additional college characteristics. Overall, these
college characteristics do little to explain the differences in average student attendance
associated with the main college categories. Despite this, these variables still provide
illumination into some interesting contextual effects of the college campus on religious
service attendance. As expected, attending a campus requiring chapel attendance
increases the reported frequency of attendance. Of course, it is impossible to know from
this whether students on these campuses are substituting required chapel for off-campus
worship, or whether attending chapel acclimates students to religious community and
practice, encouraging further participation off campus (or neither of these). Students
attending four year schools are more likely to decline in attendance when compared to
students at less traditional two-year or specialty schools. The effect size is even larger
when we narrow down these four year schools to only elite institutions. However, due to
the smaller number of students attending elite schools, we cannot be statistically
confident in this result. Neither the level of racial diversity nor the percent Caucasian on
campus alter the attendance rates of students. Lastly, students attending schools which
enroll a high ratio of men to women are more likely to show some level of decline in
attendance. This finding suggests that net of one’s own gender, the gender composition
of one’s environment has an independent role on religious practice.
Model 4 includes additional variables that measure the effect from life course
events on the frequency of religious service attendance. Similar to the effects associated
with specific college characteristics in Model 3, these have little effect on size and
118
TABLE 4.4
FIXED EFFECTS REGRESSION
PREDICTING TRUNCATED ANNUAL RELIGIOUS SERVICE ATTENDANCE
(1) (2) (3) (4) College Type (Years Enrolled) Evangelical -1.222+ -1.105 -1.143 -1.329+ CCCU -0.394 -0.218 -0.410 -0.722 Mainline Protestant -2.120** -2.117** -2.048** -2.356** Black Protestant -0.850 -0.809 -0.740 -0.891 Catholic -2.585** -2.457** -2.390** -2.570** Public -1.133** -1.085** -1.048** -1.116** Private (non-profit) -1.808** -1.772** -1.746** -1.930** Private (profit) -1.215+ -1.242 -1.077 -1.071 Age 18 (or under) 2.163* 2.216* 1.939* 19 0.913+ 0.946+ 0.851+ 20 0 0 0 21 0.028 0.050 0.065 22 0.907 0.936 0.720 23 1.226 1.251 0.783 24 (or over) 3.599* 3.683* 3.140+ College Characteristics Chapel Required 4.020* 4.219* Total Enrollment (In Thousands) 0.016 0.016 Four Year -1.416* -1.238* Elite University -2.509 -2.465 Racial Diversity -0.148 -0.180 Percent White 0.018 0.017 Percent Male -0.063+ -0.058+ Life Course Events Bachelors Degree 1.976** Associates Degree -0.256 Living with parents 1.204** Living in Dorm 0.456 Married 1.790* Children 1.902 Cohabiting -0.196
119
SOURCE: National Longitudinal Study of Youth 1997 +p<.10 *p<.05 **p<.01 (two-tailed test) significance of the institutional variables. Consistent with past research, obtaining a
bachelors degree results in an increase in the frequency of church attendance. However,
obtaining an associates degree has no significant affect on the frequency of worship
attendance. As expected, living with parents or parental figures results in slightly higher
levels of service attendance. However, living in a dorm does not affect the level of
attendance. Getting married also seems to result in a slight increase in church attendance.
Unexpectedly, neither having children nor cohabiting significantly alters the frequency of
worship attendance. It is likely that the children born to respondents are still too young to
attend a congregational religious education program. Past research has demonstrated an
increase in service attendance when young children reach school age, but not before
(Argue, Johnson and White 1999; Stolzenberg et al 1995). The unexpected result
regarding the non-effect from cohabiting may be partly due to cohort differences in
sexual mores. Among younger cohorts, living with an unmarried partner may no longer
represent a moral barrier to entrance into congregational life as it once did.
(1) (2) (3) (4) (5) (6) Evangelical Catholic Female Male Black Latino # of Observations 2035 4102 7520 5880 2891 2237 # of Individuals 710 1373 2489 2058 1020 821 R2 (within) 0.04 0.05 0.03 0.04 0.03 0.04
SOURCE: National Longitudinal Study of Youth 1997 +p<.10 *p<.05 **p<.01 (two-tailed test) Life course events also differ by religious tradition. Catholics are more likely to attend
after obtaining a bachelors degree, but degree attainment does not significantly increase
evangelical attendance. Evangelicals attend more while living at home, but Catholics
attend the same whether living at home or living elsewhere. This may be partially due to
the fact that evangelicals come from disproportionately high attending homes. If the
family of origin attends frequently, the norm of attending would be expected to be greater
in these households. White evangelical students attend more frequently when they marry,
although Catholics do not. However, there is a significant increase in attendance after
having children for Catholics but not for evangelicals. This is likely due the emphasis on
the sacrament of infant baptism in Catholicism. Catholics who want to raise their
children in the faith begin much earlier than most evangelicals who do not require infant
baptism. Lastly, just as in the general population, cohabiting does not negatively affect
the level of attendance for evangelicals or Catholics. Both of these traditions still
maintain prohibitions against living with an unmarried partner, although this does not
seem to affect the frequency of attendance for evangelical and Catholic college students.
Models 3 and 4 examine difference by gender. Overall, decline by college type
are more substantial for men than for women. Women have significant declining
123
attendance at Catholic and non-religious private (non-profit) schools. Men, in contrast,
decline in attendance at CCCU schools, Mainline, Catholic, public, and private (non-
profit) schools. Neither sex declines significantly in attendance at non-CCCU
evangelical, black Protestant, or private (for profit) schools.
Several contextual campus characteristics have a greater impact on women than
men. Perhaps the most notable finding in this regard is the difference in reports of
attendance for women who attend a college that requires chapel versus men who attend
such colleges. It seems that men do not increase their reporting of service attendance at
such schools while women do. There are also slight differences in the effect from
attending a four year college, with women being significantly likely to decline in
attendance but not men. Neither women nor men are affected in their frequency of
service attendance by attending elite universities or the racial composition of campus.
Lastly, women seem to decline slightly in attendance when they attend colleges with a
large male to female ratio of studios. This suggests that the overall lower attendance of
males influences females on male dominant campuses.
Lastly, in examining life course measures, only women seem to increase
attendance after graduating with a bachelor’s degree. On every other measure life course
measure – obtaining an associates degree, living at home, getting married, having kids,
and cohabiting – neither men nor women are significantly affected in their frequency of
attendance. This is partially due to the decreased power associated with the smaller
samples. In the full sample (Model 4 in Table 4.4) there are significant positive effects
associated with getting married and living at home.
124
Models 5 and 6 examine differential effects by African-Americans and Latinos.
No significant effects for either minority group could be found by college type. Even
though both groups are oversampled in the data, the number of minority students in the
sample attending private religious schools is small. Nevertheless, the effect size indicates
substantial drops in services for African Americans at Catholic schools and Latinos at
CCCU schools. Without larger samples of these minority groups, little should be made
of these statistically insignificant findings.
Blacks also show no significant differences in attendance level by institutional
characteristics. As with males, Blacks do not seem to increase their reporting of church
attendance when they attend an institution that requires regular chapel attendance.
Perhaps more surprisingly, neither the degree of racial diversity nor the percent of the
student body who are White affect the level of religious service attendance for Blacks.
Latinos, in contrast, are considerably more likely to report an increase in attendance
while they attend colleges that require chapel. Additionally, Latinos seem to increase in
attendance when the percent of non-Hispanic Whites on campus is high. For every 10%
increase in non-Hispanic Whites on campus, Latinos add another church service to their
annual level on average.
When we examine the effects of life course events for Blacks and Latinos, we
also discover some significant differences between the two groups. Blacks, on average,
increase their attendance after graduating with bachelor’s degree. Latinos do not (at least
not at a level of statistical significance). Conversely, Latinos actually decline in
attendance after they graduate with an associates degree. Apart from this, Blacks are not
significantly affected by the life course events measured in this model. Latinos, however,
125
seem to be affected by parent-child relationships. Both living at home and having kids
seem to boost service attendance. About two-thirds of Latinos in the United States are
Catholic, so this increase in attendance after having kids may be similar to the effect that
was evident for Catholics as well. However, there is also extant research that emphasizes
the link between religiosity and marriage and fertility for many Latinos (Mosher, Johnson
and Horn 1986; Ritchey and Dietz 1990). This may also contribute to the effects evident
in Model 6.
4.8 Discussion
The relationship between the religious status of a college or university and the
religious service attendance of students is complex. Nevertheless, several key findings
emerge that deserve special attention:
(1) Most students experience some decline in religious service attendance while
enrolled in college. This remains true even after holding constant all time
invariant individual characteristics and controlling for variant characteristics
such as age, college characteristics, and life course events.
(2) Students attending institutions that belong to the Council for Christian
Colleges and Universities exhibit the greatest overall declines in annual counts
of church attendance. However, this is not a result of students dropping out of
church. Rather, students who did attend multiple times per week were simply
less likely to do so while attending a CCCU school (although these students
still attended at least weekly).
126
(3) Students are the most likely to decline away from the norm of weekly
attendance while attending a Catholic or mainline affiliated college or
university. On average, these students declined at a rate between 2 and 2.5
religious services for each year enrolled, considerably more than the
approximate 1.1 service decline annually for students enrolled in public
colleges and universities.
(4) Evangelicals and Catholics students look remarkably similar to the rest of the
student population, even while attending institutions that are affiliated with
their own religious tradition.
(5) Differences by the gender and race of the student are evident. Net of other
variables, males are more likely to drop in attendance while enrolled in most
types of colleges (with the exception of non-CCCU evangelical, black
Protestant, and Private for profit institutions). African-Americans have the
sharpest decline at Catholic institutions (~5.8 services annually), while
Latinos decline the most at CCCU institutions (~4.7 services annually).
While the effects by race/ethnicity have comparably large effect sizes, they do
not reach statistical significance.
4.8.1 The Big Picture
The first two hypotheses predicted a temporary decline in religious service
attendance during college, followed by an increase after graduation. The models
presented here confirmed these hypotheses. On average, students decline in attendance
while enrolled in colleges and universities in the United States. Moreover, there is
evidence from these models that graduating with a bachelor’s degree boosts religious
127
service attendance. Table 4.5 suggests that Catholics, women, and African-Americans
are the most likely groups to increase attendance after graduating. Nevertheless, on
average, this increase in service attendance does not fully counter the effect from being
enrolled in a college or university.
It is likely, however, that the effect from graduating with a BA is underestimated
in the full multivariate models presented here (see Model 4 in Table 4.4). Table 4.2
found that, net of aging effects, those with a BA in the young adult population had the
highest level of religious service attendance (compared to those enrolled in college and
not enrolled without a degree). This corroborates extant research that examines church
involvement and educational attainment using survey data (Cornwall 1989; Hoge,
Johnson, and Luidens 1994). These studies are better able to capture the long term
effects from higher education that result from new social networks, transitioning into the
workforce, and civic expectations. The data used in the full models of Table 4.4 reflect
only respondents who report being enrolled in a college or university within the past year.
The full social script that educational attainment brings to religious life has probably not
played itself out for the respondents who have just graduated. If the aforementioned
contention concerning college graduation and religious service attendance is true, then a
college education does no enduring harm to basic religious practice. Rather, educational
attainment may actually provide the structural means and cultural tools to integrate into
religious congregations during young adult life.
4.8.2 Religious service attendance at conservative Protestant schools
Hypothesis 3 had two opposing formulations. In the first formulation, strongly
religious schools were posited to have a positive religious impact; in the second
128
hypothesis they were posited to have negative religious influence. Overall, there seems
to be more evidence for the first formulation for conservative Protestant schools, but with
some important caveats. While students at non-CCCU evangelical schools did not decline
as much in service attendance as students at Catholic and mainline colleges, they were
not different from public college and universities (Model 4 in Table 4.4). Moreover, a
statistically significant decline at these schools was still evident.
CCCU schools are arguably some of the most explicitly religious colleges in the
United States. When the eight category attendance variable is converted into annual
religious service attendance, students as CCCU experience the most dramatic decline
over the course of four years. In the uncontrolled random effects model (Figure 4.1),
students at CCCU schools drop nearly 25 services, primarily between their second and
third year of attendance at these institutions. While CCCU students start significantly
higher in religious service attendance, they look roughly similar to non-CCCU
evangelical and black Protestant schools by senior year. This sharp drop is confirmed in
the fixed effects models in Table 4.2. However, once I truncate the annual attendance at
52 times a year, I find that the decline associated with attendance at CCCU schools nearly
disappears (Table 4.3). This suggests that the drop in attendance is mostly due to regular
attendees (which CCCU schools have a disproportionate number of) dropping modestly,
but not to levels lower than once a week. This is probably due to the constrained
opportunities for attending more than weekly worship services while enrolled at nearly
any institution of higher education. Perhaps students attending CCCU schools substitute
their multiple weekly services with other on campus religious activities.
129
Nevertheless, even using the truncated variable, men are still likely to display a
statistically significant annual decline of more than two services at CCCU schools. Part
of this is due to the fact that men decline more at nearly all types of schools when
compared to women. However, this decline is statistically robust in the face of
counteracting forces. Men, as a whole, decline less in attendance on small campuses.
CCCU schools are disproportionately small, with most enrolling less than 2,000 students.
CCCU schools also have a disproportionately high number of schools that require chapel
attendance. Nevertheless, either men substitute these required chapels for an even greater
number of off campus religious services, or they do not consider these chapel services in
their definition of “worship service” (while women do). The latter definitional gulf
between men and women seems far less plausible then the former behavioral gulf. The
picture painted here seems to suggest that women do not substitute chapel for outside
church services, while men do. Net of this difference in chapel, men still decline in their
overall attendance at CCCU schools while women do not.
4.8.3 Religious Service Attendance at Catholic and Mainline Schools
Unlike the conservative Protestant institutions, the evidence clearly points to
students at mainline and Catholic institutions declining in religious service attendance –
more so than at any other type of institution. Unlike CCCU students, this steeper rate of
decline does not appear to be from students as these schools attending at higher overall
rates (see Figure 4.1). So why is the rate of decline at these schools more than double the
rate of public colleges and universities where more than 70% of all undergraduates in the
United States enroll? Unfortunately, the data available are not able to directly address the
manner in which the religious self-identification of the institution is manifest in the
130
different pedagogical philosophies, structural organization, and general campus milieu.
Nor are we able to determine from this data how student religious life interacts with these
campus characteristics. Nevertheless, some plausible explanations are available.
Students at CCCU schools (and to a certain extent non-CCCU evangelical
schools) may be relatively well served by the in locus parentis policy established on these
campuses. They are given plenty of opportunity to worship in addition to strict religious
instruction. However, this does not necessarily imply that a lack of religious direction on
other campuses necessarily results in a spiritually vacuous environment. At large public
colleges and universities, religious students must actively participate in student led
groups if they choose to be religiously active on campus. Moreover, the hands-off
approach most faculty take to religious guidance (Higher Education Research Institute
2005) may actually result in interested students finding alternative outlets for knowledge
and belief from their religious peers. Where does this place mainline and Catholic
schools? While there is substantial variation in the religious mission of these schools,
students are probably provided more opportunity for worship (with regularly available
chapel and mass on campus), and spiritual direction, but with lower religious
expectations compared to conservative Protestant institutions. Thus, the institutional
structure allows religious identities to remain private and unexamined on these campuses.
Further research should aid in developing and testing these hypotheses regarding
religious identity on campus.
4.8.4 Evangelical and Catholic Students
Hypothesis 4 predicted that evangelical and Catholic students would have the
highest levels of religious practice on campuses affiliated with their own tradition. The
131
evidence does not support this hypothesis. Both Catholics and evangelicals fare about the
same as other students at these institutions (see Table 4.5: Models 1 and 2). For
evangelicals, this finding suggests that it is not simply the pairing of evangelical campus
culture with evangelicals that produce the lower rates of decline at these schools. While
we should be careful not to extrapolate too far from this finding, there is cursory evidence
here that stricter religious schools are slightly more effective in inculcating religious
practice for both evangelicals and students from non-evangelical denominations.
While attending evangelical colleges may curb certain tendencies to decline in
religious practice, evangelicals themselves are more likely than other groups to decline in
attendance when they leave their family of origin. Likewise, evangelicals are more likely
to increase their worship service attendance when they marry. These findings coincide
with the literature that emphasizes the tight ties between traditional family ideals and
religion for white evangelicals (Bendroth 2007; Edgell 2003). In this way, evangelical
higher education may play an important role in the period of time that exists between
residing with family of origin and family of procreation.
Catholics, like evangelicals, fare approximately the same as non-Catholics at
Catholic affiliated institutions. Indeed, Catholics appear to be very similar to the general
trends in the student population at large, exhibiting the highest rates of declining
attendance at Mainline and Catholic schools. This finding suggests that there is nothing
significant about the interaction with Catholics at Catholic institutions that explain the
overall higher levels of decline at these schools. Rather, Catholic institutions seem to
exert an independent effect on students regardless of their religious background.
132
4.8.5 Differences by Gender
Hypothesis 4 expected that men and women would have different rates of decline
in service attendance at evangelically affiliated colleges. The evidence in the present
analyses suggests that men declined more than women at most types of institutions,
whether they were religious or not. Furthermore, women have substantially higher
increases in service attendance after graduation when compared to men. These findings
are seemingly inconsistent with past research that finds no interaction between
educational attainment and gender on religious outcomes (See Funk and Willits 1987;
Hill 2007). Nevertheless the analyses presented here only apply to students during their
college experience and not the years following college graduation. One possible
explanation for these gender differences may reside in the different reaction to an
environment largely free of family constraints. Research suggests that male religious
development is closely tied with family life (Browning 2003; Wilcox 2004). In one
study, marital dissolution resulted in a significantly larger drop in religious service
attendance for men than women (Stolzenberg et al 1995). It is quite probable that the
temporary shift away from family influence results in lower attendance for men, but is
more minimal of an influence on women. This suggests that the gender differences in
religious practice during college may be short term, and that the long-term effect from
educational attainment may be similar for men and women as other research has
indicated.
4.8.6 Differences by Race/Ethnicity
Hypothesis 6 and 7 expected differences in institutional influence for African
Americans and Latinos. While none of the coefficients of college types are significant,
133
the effect sizes are larger than most in the full-population model. Part of the reason for
this lies in the smaller N of Models 5 and 6 of Table 4.5. What is more, the fixed effects
method produces larger standard errors when compared to random or mixed modeling.
African-American students attending Catholic colleges and Latino students attending
CCCU colleges have relatively large rates of annual decline (-5.8 and -4.7 services per
year enrolled respectively). While we cannot have statistical confidence in these results,
future studies may want to consider testing this finding to see if it is substantiated with
different data and methods.
African American also seem to be unaffected by the racial ecology of the campus.
This suggests that the possible Catholic school effect is not due to Catholic affiliated
schools having low levels of racial diversity and a high percentage of non-Hispanic
Whites (characteristics that, on average, the Catholic affiliated institutions in the
NLSY97 do have). Latino students, while unaffected by the overall diversity of the
campus population, are affected by the percent of Caucasians on campus. Net of other
effects, they increase their religious service attendance by one service annually for every
10% increase in Caucasian population. This would seem to suggest that being a minority
on campus, may actually result in higher levels of religious practice for the Latino
population. Future research should also address this finding in more detail to understand
this process for Latinos.
4.9 Conclusion
The findings of this research suggest that students who enroll in higher education
experience temporary declines in religious service attendance compared with other young
134
adults of the same age. However, these declines are not the same at all types of
institutions. Students who attend a religious college associated with the Council for
Christian Colleges and Universities drop the most in absolute terms, but not below the
norm of weekly service attendance. When service attendance is truncated to once a week
or less, students attending Catholic and Mainline schools decline the most - more than
twice as much as students attending public colleges and universities. The religious
tradition of the student does not seem to influence these institutional differences.
However, men are more likely to decline than women at most types of institutions.
Lastly, although minorities may have unique religious experiences at some of these
institutions, the analyses carried out here were unable to distinguish statistically between
institutional influences on religious attendance for African Americans and Latinos.
135
CHAPTER 5:
CONCLUSION
5.1 Emerging Themes
Although the three substantive chapters of this dissertation do not form a
sustained argument, several emerging themes arise throughout the chapters that may
serve to guide a future agenda for study in the area of religion during the transition to
adulthood. Perhaps one of the most surprising findings is the relatively early age that
religious disengagement begins. The possibility of leaving the religious community of
one’s youth was typically considered an activity that began once one left the household.
The findings in Chapter 2 suggest that substantial decline is occurring at the very latest
by age 16. Not only that, but by age 20 the downward trend has stopped and mean levels
of religious service attendance are steady.
This early disengagement may partially explain why college attendance no longer
has a clear secularizing effect on young adults as it did in the past. If the bulk of religious
decline has occurred while residing at home, then for many young people there is nothing
more to disengage from when they move to a college campus. Clydesdale (2007) argues
that the first year after high school is mostly a continuation of lifestyle for most young
people. The everyday existence of the average sixteen to twenty year old involves
preoccupation with navigating social relationships, maintaining academic standing in
school, and spending time in recreational activities. The supposed “big” questions of
136
religion and meaning are likely of secondary concern for the majority during the college
years.
The story just described is appears to be different than what the baby-boomers
experienced at a similar age. For those who came of age during the 1960s, college did
have a secularizing influence on religious practice and belief. This can be seen in
Appendix A, using the panel data for young people who were high school seniors in
1965. Approximately 1.5% of the entire nationally-representative sample claimed no
religious affiliation when they were high school seniors. Yet, by 1973, 12.5% had no
religious affiliation. Moreover, if the baby-boomers graduated from college, they were
more than twice as likely to disaffiliate as those who had not attended college at all. For
this rather unique cohort, college campuses served as a mobilizing arena to oppose
traditional sources of authority, including religion.
Yet, it may be that this very purposeful religious disengagement for those that
came of age in the 1960s, has now become institutionalized in the children of the baby-
boomers. Expectations of religious engagement (or at the very least religious
identification) were high for this previous generation all the way through high school.
Yet, as the analysis in Chapter 2 revealed, as of 1997, 11.5% of twelve to fourteen year
olds had no religious affiliation. This picture is certainly congruent with the description
given by Smith and Denton (2005) of adolescent religiosity as inarticulate and
disconnected from historical religious traditions.
At the same time as this move toward early adolescent religious disengagement
has occurred, young adults are delaying marriage and having children, the events
typically associated with increasing religious practice (Wilson and Sherkat 1994).
137
Wuthnow (2007) finds that a decrease in religious participation has occurred among
young adults when compared to the generation of young adults in the early 1970s.
However, the difference in religious practice between these generations can largely be
accounted for by the differential rates of marriage and childbearing.
These two movements, one toward younger disaffiliation, the other toward older
reaffiliation, open up a new period in the life course where most young adults have no
ties to religious institutions. I will outline several promising avenues of study in this
period of emerging adulthood.
One neglected population in the study of young adults and religiosity, are the
young people who do not attend college. Sociologists had assumed for some time that
higher education was detrimental to religious practice and belief. However, the research
in this dissertation (Chapters 3 and 4) suggests that college campuses may actually be a
relatively safe place for religion, and in the case of religious participation, may actually
provide the tools to integrate into religious communities upon graduation. However, we
know very little about the transition to adulthood and the accompanying religious
practices of those who enter the workforce as opposed to entering higher education. The
rough sketches provided by the General Social Survey in Appendix A suggest that the
changes in the non-college population by birth cohort have been more severe than the
changes within the population that attended college. Understanding the structural
disconnect between this population and religious institutions will be an important key to
understanding the religious lives of young adults.
Another key area of study will involve examining the role that the timing of key
life course events in emerging adulthood plays. As the normative ordering of transitional
138
events such as leaving home, completing education, getting a job, getting married, and
having children have become less defined (Shanahan 2000), the ordering and timing of
these events become part of the conscious decision-making processes of young adults as
they navigate their way through emerging adulthood. This process brings with it new
questions about religiosity in young adulthood. For example, marriage and having
children bring new roles that require responsibility for others. If we suppose that one of
the primary reasons that college has little effect on the religious lives of young people is
due to the difficulty getting young persons engaged with the world outside of ego-centric
concerns, then we might expect that college would have a greater religious impact on
non-traditionally aged students, many who have already formed families.
Another important issue may be the length of time spent away from religious
institutions in young adulthood. There may be a nonlinear relationship between time
spent in apostasy and the probability of returning. If young adults spend a long enough
time outside the confines of the church developing meaningful networks of relationships
and establishing career trajectories, then they may be less likely to return even if they do
marry and have children. Understanding the full implications of this extended period of
disengagement appears to be the next step in understanding religion in early adulthood
for the current generation.
5.2 A Quantitative Agenda for the Study of Religion Over the Life Course
Studies of intra-individual religious change and its social correlates are still
relatively sparse within the sociological literature. Research that cover large periods of
the life course are quite rare, principally due to the difficulty in attaining adequate data
139
(for exceptions, see Dillon and Wink 2007 and McCullough et al 2005). Collecting
retrospective data is commonly a remedy for this problem. Research that has primarily
relied upon retrospective life histories or open-ended interviews has been relatively
successful in providing broad pictures of religious change and the aging process (see
Ingersoll-Dayton, Krause, and Morgan 2002). Nevertheless, these methods are often
hampered by an individualist narrative interpretation wherein the actor is overemphasized
and many of the social forces that sociologists theorize as important - such as institutions,
networks, and cultural contexts - remain hidden. The combination of retrospective data,
and its ability to capture large swaths of time, is probably best used in tandem with
concurrent panel data that is better able to capture the hidden realities that the analyst
wishes to uncover.
The use of panel data has proliferated in sociological research over the past two
decades (Halaby 2004). And although panel data on religion is still in its infancy, several
promising developments will likely result in future research on American religion relying
more heavily on panel data. The National Study of Youth and Religion is entering its
third wave of data collection and will span a period of five years for the study participants
from adolescent into early adulthood. This promises to be a rich source of material for
analyzing multiple dimensions of religious and social change during this period of the life
course. The Panel Study on American Religion and Ethnicity, completing its first wave
of data, also promises to be a key resource on the changing beliefs, rituals, and identities
of American adults as more waves are collected. The widely used National
Congregations Study is collecting a second wave that includes a subsample of
congregations from the first wave of study, allowing for the investigation of institutional
140
dynamics of religious congregations over time. The primary data used in this
dissertation, the National Longitudinal Study of Adolescent Health and the National
Longitudinal Study of Youth 1997 are also continuing to collect further waves of data,
likely containing survey indicators of religious practice and belief. In addition, there is a
concerted effort to add high quality questions on religion to already existing panel data. I
expect that these developments will encourage sociological researchers to increasingly
study the religious dynamics of the life course as the tools to study such change become
more widely available.
Although this dissertation has focused exclusively on religion during the
transition to adulthood, studying the social correlates of religious change in mid-life and
late-life is a relatively new territory. Panel data, such as Marital Instability Over the Life
Course, and Americans’ Changing Lives, contain multiple waves of data that include
some basic religious indicators repeated over time. While there have been some studies
tracking aging and indicators such as church attendance and belief in life after death,
many basic indicators such as frequency of prayer and importance of faith have gone
unstudied. In addition, the social mechanisms associated with aging have not yet been
specified across the adult lifespan. For example, little is known concerning the influence
on religiosity from such important markers of adult life as marital quality, career
pathways, social mobility, and community attachment.
Religion in late life has been studied to a greater extent by gerontologists.
Nevertheless, a more socially orientated perspective would likely open up new questions
and debates within the field. Grown children commonly become important in this phase
of the life course. Just as the parent-child relationship is key to religious socialization in
141
early life, the reversal of the care-taker role may once again increase the importance of
this relationship in late life. An examination of this relationship could be carried out
using data from the Longitudinal Study of Generations. Investigating the role of other
defining features of late life such as failing health, the death of a spouse, and entry into
professional institutions of assisted living may also be fruitful avenues for studying what
social changes predicate religious changes in late life.
Understanding the social embeddedness of religious belief is one of the
fundamental tasks of the sociologist of religion. Yet, we still only have a cursory
understanding of how religious beliefs are patterned by the many intra-individual changes
in social and institutional arrangements that give the life course its specific texture. It is
my hope that future research will move beyond theories of the family lifecycle and begin
to theorize the ways that specific religious beliefs and practices are shaped by the diverse
domains that constitute the entire life course.
142
APPENDIX A:
COLLEGE AND RELIGIOSITY BY BIRTH COHORT
A.1 Overview
Chapter 3 documented the relationship between educational attainment and religious
practice and affiliation. As was noted, the hypothesized secularizing effect from college
is not evident among recent college graduates. This finding, however, applies only to a
particular birth cohort and may not be applicable to past generations. Colleges and
universities have undergone substantial shifts in the United States during the second half
of the twentieth century. Higher education rapidly expanded during this time as more
young people transitioned directly into post-secondary education following high school
graduation (Knapp, Kelly-Reid, and Whitmore 2007). Moreover, recent incoming
freshman are more likely to desire financial security from college compared to
developing a meaningful philosophy of life. This is substantially different from students
who entered college in the late 1960s (Astin 1998). In this appendix I briefly explore
how religion fares alongside these other changes, and examine how the effect of a college
education on religiosity has altered over time.
A.2 Religious Climate on Campus
A recent New York Times Op-Ed has made the claim that current students are
more traditionally religious than in the recent past (Taylor 2006). This view is not,
143
however, embraced by all social scientists who study religion on campus. There certainly
does appear to be an increase in evangelical presence on college campuses, as
documented by students adopting the “born-again” label (Bryant 2005), and the growth
of evangelical parachurch organizations on campus (Schmaulzbauer 2007). Others,
however, note that a majority of students are unreflective in their faith, with mundane
activities such as navigating classes and maintaining an active social life occupying the
majority of student life (Clydesdale 2007). While the majority of students do identify
themselves as spiritual (Higher Education Research Institute 2004), this does not
necessarily entail that they are engaged with historic strands of religious and spiritual
traditions (Bender 2007). Still, it is difficult to make claims that these students are
different than in the past without adequate measures of these dimensions of religiosity
over time.
There is some survey evidence that religious practice and belief have changed
over time for college students, although the studies are now somewhat dated. Moberg
and Hoge (1986) studied Catholic students at Marquette University in 1961, 1971, and
1982. Between 1961 and 1971, the authors report a substantial increase in individualistic
interpretations surrounding morals and religion, increased religious doubt, and decreased
mass attendance. However, between 1971 and 1982 there was a move toward more
traditional Catholic positions, along with a decreased demand for intellectual autonomy
from the Church. The authors point to the turbulence of Vatican II, Humanae Vitae, and
the Kennedy presidency between 1961 and 1971 as a possible cause for such a change.
The 1970s, on the other hand, were more stable and changes in religious beliefs among
Catholic students were similar to changes occurring among other students.
144
Hoge et al. (1987) documented these more general trends among all Dartmouth
college students with the administration of five identical surveys to undergraduates in
1952, 1968-69, 1974, 1979, and 1984. The major declines in values and orthodox
religious beliefs occurring from the 1950s and into the 1970s had reversed themselves in
many instances by the 1984 survey. For example, those with no religious preference
were highest in 1974 and subsequently drop sharply. In addition, belief in a “Divine
God, Creator of the universe” began rising in 1979 and 1984. These studies, although not
updated for recent college students, establish that considerable historical variation has
occurred in religious practice and belief over the past few generations.
There is also limited evidence that faculty dispositions toward religion may be
changing. Schmalzbauer (2003) and Lindsay (2007) have documented the increased
number of Catholic and evangelical faculty moving into the major research universities
across the United States. Along with this, several explicitly conservative, evangelical
colleges are employing faculty that are contributing more seriously to their particular
fields of study (Schultze 1993). As religious conservatives re-emerge from their enclaves
and separatist institutions of the past, it is difficult to say what effect, in any, this will
have on the religious beliefs of students attending institutions of higher education.
However, both the trends among students and faculty may signal that past studies were
picking up a historically conditioned cohort effect in addition to any enduring effect of
college upon students.
In order to examine the possibility that the relationship between religiosity and
college attendance has changed over time, two brief analyses will be carried out. The
first involves examining the bivariate relationship between attaining a college degree and
145
several key religious outcomes separated by birth cohort in the cumulative General Social
Survey. The second involves using panel data from a national sample of high school
seniors in 1965 to examine the relationship between attaining a college degree and
changes in church attendance, religious affiliation, and beliefs about the Bible between
1965 and 1973.
A.3 Data
Data come from the cumulative General Social Survey (GSS) 1972 to 2004 as
well as the Youth-Parent Panel Socialization Study. The GSS is a nationally
representative survey administered annually or biennially by the National Opinion
Research Center since 1972. A question on religious service attendance and religious
affiliation has been asked on every GSS questionnaire administered. Strength of
religious affiliation has been measured since 1974, frequency of prayer since 1983, and
beliefs about the Bible since 1984. The Youth-Parent Panel Socialization Study is a
nationally representative, four wave panel study of youth and their parents beginning
with high school seniors in 1965. Follow-up waves were conducted in 1973, 1982, and
1997. For the purposes of this analysis, only the 1965 and 1973 youth panels will be
used. Both waves contain questionnaire items on religious service attendance, affiliation,
and belief about the Bible.
A.4 Analysis I: Higher Education and Religious Outcomes in the GSS 1972-2004
Figure A.1 displays the percent of individuals age twenty-five or older who attend
church “nearly every week” or more by birth cohort for those with a bachelor’s degree or
146
more and for those with less than a bachelor’s degree. Looking within birth cohorts, this
figure displays a consistent, slight boost in weekly church attendance for those with
college degrees. While there is some fluctuation over time, there is always at least a
slight difference between college graduates and non-college graduates.28 Figure A.2
displays the percent age twenty-five or older who claim no religious affiliation across
birth cohorts. We can see that the college-educated born before 1960 were more likely to
claim no religious affiliation. However, cohorts born in the 1960s and later appear to be
roughly equal on measures of non-affiliation. Figure A.3 presents data on the percent of
respondents believing the Bible to be a “book of fables”. This measure taps into “non-
orthodox” Christian beliefs. Here we see that there is substantial difference between
college educated respondents and the rest of the population in all birth cohorts. The
difference is slightly less in recent cohorts but still hovers around ten percentage points.
Figure A.4 displays differences between the two populations in strength of affiliation. Of
those who do affiliate, the GSS asks how strongly they affiliate with their tradition. We
can see from this graph that over the majority of birth cohorts (and certainly over recent
cohorts), the college educated who identify with a religious tradition, so more strongly
than religious affiliates with less than a bachelor’s degree. Lastly, Figure A.5 examines
the differences in the percent who pray at least daily by cohort. Like the graph for non-
affiliation, we see a near convergence between the college educated and the rest of the
population among recent birth cohorts.
28 It is important to not interpret too much from the decline across birth cohorts, as this could also be due to aging effects. The important difference is within cohorts between those who graduated from college and those who did not.
Figure A.5 Pray Daily Across Birth Cohort by Degree Completion
A.5 Analysis #2: Higher Education and Religious Change in the YPSPS 1965-1973
In order to assess the possibility that recent birth cohorts react differently to the
college environment, a similar analysis to the main analysis in Chapter 3 has been
conducted with panel data from an earlier cohort. The youth respondents in the Youth-
Parent Socialization Panel Study were high school seniors in 1965. They were given a
questionnaire that asked them to report their religious affiliation, religious service
attendance, and beliefs about the Bible among other items. These same questions were
then replicated in the 1973 follow-up questionnaire. The dependent variables in Table
A.1 examine three different outcomes: (1) how often they attend church on a five-point
scale ranging from “Never” to “Every Week”, (2) a dichotomous variable reporting no
religious affiliation, and (3) a dichotomous item with the responses the Bible “is a good
book” but “God had nothing to do with it”, or the Bible “was written by men who lived
150
so long ago that it is worth very little today” versus two responses that acknowledge
divine influence in the Bible.
TABLE A. 1
THE EFFECT OF DEGREE COMPLETION
ON RELIGIOUS ATTENDANCE, AFFILIATION, AND BELIEF ABOUT THE BIBLE
1965-1973
SOURCE: Youth-Parent Socialization Panel Study +p<.10 *p<.05 **p<.01 (two-tailed test) The independent variables are similar to several used in the analysis with the Add
Health data. The main independent variable indicates whether or not the respondent has
attained a bachelor’s degree or more in the eight years between Wave I and Wave II.
Religious service attendance and Bible beliefs at Wave I are included as controls. Lastly,
basic demographic controls from Wave I have been included here as well.
Table A.1 presents the results. The first set of models examines whether religious
service attendance depends on graduating from college. Unlike the more recent cohort in
the Add Health panel data, this earlier birth cohort saw no differences between those that
received college degrees and those that did not in the frequency of religious service
attendance. This holds true even after statistically controlling for prior service attendance
and other potentially confounding factors. The second set of regression equations in
Table A.1 present the odds ratios of disaffiliating from a religious denomination or group
between wave I and wave II. Model 2 reveals that even after statistical controls, college
graduates were nearly twice as likely to drop out of religion compared to those that did
not graduate from college. Once again, this is in rather stark contrast to the more recent
cohort represented in the Add Health data. Lastly, the set of equations on the right hand
side of Table A.5 display the odds ratios of believing the Bible is not divinely inspired
versus the belief that the Bible is divinely inspired. The coefficients from the equation
furthest to the right demonstrate that college graduates were more than 2.5 times as likely
to believe that the Bible was not divinely inspired compared to non-college graduates.
A.6 A Comparison with Recent College Graduates
Overall, the effect of educational attainment on religious outcomes appears
sufficiently different for recent college graduates compared to college graduates of the
past – particularly those that attended college in the late 1960s. Both the data analyzed
from the GSS as well as the models from the Youth-Parent Socialization Panel Survey
suggest that secularization as a result of college attendance may be waning, although not
uniformly along all measures.
Religious disaffiliation, no different for college graduates and the rest of the
population in recent data, is about twice as likely to occur for college graduates who
152
came of age during the late 1960s and early 1970s according to the findings presented
here. This pattern is confirmed in the GSS, with younger birth cohorts equally likely to
disaffiliate, while college graduates from earlier birth cohorts were more likely to
disaffiliate. There are several plausible explanations for this change. Nearly all of the
high school seniors in the YPSPS in 1965 were affiliated with a religious tradition or
denomination (less than 1.5 % had no affiliation). Among recent high school seniors the
rate without a religious affiliation is nearly 13%. Yet, by the follow-up surveys, each
approximately seven years later, the rates of disaffiliation are much closer for both birth
cohorts (12.5% versus 18.9%). The earlier cohort experienced much higher rates of
disaffiliation once they left home, while the more recent birth cohort underwent far less
disaffiliation after leaving home. Moreover, the rate of religious affiliation seems to be
approximately the same from 8th grade to 12th grade in the Add Health data. While this
may mean that adolescents are disaffiliating prior to the 8th grade, the more likely
explanation is that they are reflecting the religious identity of their parents. Although
religious disaffiliation was much more common for the entire birth cohort represented in
the YPSPS, graduating from college still resulted in nearly twice the rate of disaffiliation.
In this way, college campuses likely provided the arena for attacking many of the cultural
institutions that defined their parents’ generation, including religious institutions. This
generational conflict over values does not seem to define more recent birth cohorts. In
this way, the birth cohort that came of age during the late 1960s may be the exception
rather than the norm.
The positive relationship between graduating from college and religious service
attendance evidenced in the Add Health data is non-existent in the YPSPS birth cohort.
153
College, which I have suggested provides the social and cultural tools for integrating into
religious congregations in young adulthood (see Chapter 3 and Chapter 4), may only
function in this manner for recent birth cohorts. Alternatively, this function may have
been present in the past, but resulted in no mean level change due to the high level of
disaffiliation. The GSS data, however, suggest that college educated individuals have
been attending religious services slightly more frequently since the birth cohort of the
1930s (which would have most likely attended college during the 1950s). It is not
entirely clear why there is a discrepancy between these datasets, although it is quite
possible that selection effects outlined in Chapter 4 influence the cross-sectional GSS
data. The YPSPS data are a better comparison with the Add Health data because they are
both longitudinal and cover similar periods in the life course.
Bible beliefs were altered by college for this earlier generation. According to
YPSPS, college graduates were nearly two and one half times as likely to believe that
God had nothing to do with Bible compared to non-college graduates (controlling for
their prior beliefs about the Bible). However, data from the GSS in Figure A.3 suggests
that differences in Bible beliefs by educational attainment are still substantial in younger
birth cohorts. Part of this is certainly due to the lower likelihood of high educational
attainment for people who hold to literalist interpretation of the Bible (Darnell and
Sherkat 1997). Nevertheless, there may be effects from college on beliefs about
scripture among recent graduates, but information pertaining to the content of religious
belief at multiple points in time is not available in the Add Health data. It is possible that
religious practice is no longer affected by educational attainment, but that the actual
content of religious belief is still altered. As this data becomes available, researchers
154
would be wise to address the relationship between education and a wider variety of
religious indicators – particularly measures of religious belief and orthodoxy.
The GSS has measured the frequency of prayer since 1984. Consistent with the
data from the Add Health analysis in Chapter 3, college graduates seem virtually the
same as non-college graduates when measuring the percent that pray daily. However,
college graduates in the birth cohorts prior to 1950 have a relatively lower percent who
pray daily compared to those who did not graduate from college. There is, unfortunately,
no equivalent variable in the YPSPS to compare to the GSS or Add Health data.
Although we should be careful not to interpret the data over birth cohorts as decline
(much of these effects could also be accounted for by age), it is important to note that the
convergence between college graduates and non-college graduates seems to be due to the
percent praying daily more rapidly declining overall in the non-college category. Steeper
slopes for the non-college category over time are evident in several of the Tables. This
indicates the importance of future research attempting to understand how the transitions
in early adulthood affect religiosity for those that do not continue their education as well
as those that do.
For those who are affiliated with a religious tradition, Table A.4 suggests that a
higher proportion of individuals strongly identify with their religious tradition if they
have graduated from college. This seems to have consistently been the case at least since
the birth cohort born in the 1930s. This finding adds some nuance to these other
analyses. Although college graduates in the past seem more likely to disaffiliate and less
likely to believe that the Bible was divinely influenced than those that did not graduate, a
higher percentage of the affiliated strongly identified with their religious tradition. If
155
college is a time for the development of a meaningful understanding of the self, then it
seems plausible that individuals would engage their religious traditions. This may result
in disaffiliating (as it did with those that attended college in the late 1960s), or it may
result in more strongly identifying with one’s religious tradition. On the other hand, it
may simply be that those who have a stronger religious identity are more likely to attend
college (just as those that attend church more frequently are more likely to attend
college). Religious youth are likely to be more “planfully competent” (Clausen 1993),
which is correlated with educational attainment.
Regardless of the source of this difference, it is important to note that the
changing effect of college on religion is not a singular story of declining or increasing
influence on religion. The overarching trend seems to be that educational attainment may
have been related to some forms of religious decline in the past, however this is less the
case for recent college graduates. Nevertheless, it is possible that traditional and orthodox
religious beliefs are still in decline as a result of college. There are, unfortunately, no
adequate measures of this in a form which would allow us to have methodological
confidence in the results. Future research on the effect of college on religious life would
do well to reflect the complexity of forms that religious beliefs, rituals, and identity take
on during the college years. Such an analysis would shed light on how education and
religion interact and help us to generate plausible theories about how this relationship
may have changed over time.
156
BIBLIOGRAPHY
Abravanel, Martin D. and Ronald J. Busch. 1975. “Political Competence, Political Trust, and the Action Orientations of University Students.” Journal of Politics 37:57-83.
Allison, Paul D. 2005. Fixed Effects Regression Methods for Longitudinal Data Using SAS. Cary, NC: SAS Institute.
Ammerman, Nancy T. 1987. Bible Believers: Fundamentalists in the Modern World. Rutgers University Press.
Argue, Amy, David R. Johnson, and Lynn K. White. 1999. "Age and Religiosity: Evidence from a Three-Wave Panel Analysis." Journal for the Scientific Study of Religion 38:423-435.
Arnett, Jeffrey J. 1998. “Learning to Stand Alone: The Contemporary American Transition to Adulthood in Cultural and Historical Context.” Human Development 41:295-315.
——. 2004. Emerging adulthood. Ebrary.
Astin, Alexander W. 1972. "Measured Effects of Higher Education." Annals of the American Academy of Political and Social Science 404:1-20.
——. 1977. Four Critical Years: Effects of College on Beliefs, Attitudes, and Knowledge. San Francisco: Jossey-Bass.
——. 1978. Four Critical Years. Jossey-Bass.
——. 1997. “The Changing American College Student: Thirty Year Trends, 1966-1996.” Review of Higher Education 21:115-136.
Bachman, Jerald. 1972. Young Men in High School and Beyond: A Summary of Findings from the Youth in Transition Project. Office of Education, Bureau of Research.
Bader, Christopher D. and Scott A. Desmond. 2006. “Do as I Say and as I Do: The Effects of Consistent Parental Beliefs and Behaviors upon Religious Transmission.” Sociology of Religion 67:313-29.
Bahr, Howard M. 1970. “Aging and Religious Disaffiliation.” Social Forces 49:59-71.
157
Baltes, Paul B. 1997. "On the Incomplete Architecture of Human Ontogeny - Selection, Optimization, and Compensation as Foundation of Developmental Theory." American Psychologist 52:366-380.
Baltes, Paul B., H. W. Reese, and L. P. Lipsitt. 1980. "Life-Span Developmental-Psychology." Annual Review of Psychology 31:65-110.
Baltes, Paul B., Ursula M. Staudinger, and Ulman Lindenberger. 1999. "Lifespan Psychology: Theory and Application to Intellectual Functioning." Annual Review of Psychology 50:471-507.
Bao, Wan- Ning, Les B. Whitbeck, and Dan R. Hoyt. 1999. “Perceived Parental Acceptance as a Moderator of Religious Transmission among Adolescent Boys and Girls.” Journal of Marriage and the Family 61:362-374.
Barry, Carolyn McNamara and Larry J. Nelson. 2005. “The Role of Religion in the Transition to Adulthood for Young Emerging Adults.” Journal of Youth and Adolescence 34:245-255.
Beit-Hallahmi, Benjamin. 1997. The Psychology of Religious Behavior, Belief and Experience. Routledge.
Bellah, Robert N., Richard Madsen, William M. Sullivan, Ann Swidler, and Steven M. Tipton. 1985. Habits of the Heart: Individualism and Commitment in American Life. University of California Press.
Bender, Courtney. 2007. "Religion and Spirituality: History, Discourse, Measurement." SSRC Web Forum on the Religious Engagements of American Undergraduates. Available online at http://religion.ssrc.org/reforum/Bender.pdf
Bendroth, Margaret. 2007. “Evangelicals, Family, and Modernity.” Pp. 56-69 in American Religions and the Family: How Faith Traditions Cope with Modernization and Democracy, edited by D.S. Browning and D.A. Clairmont. New York: Columbia University Press.
Bensen, Peter L., Michael J. Donahue, and Joseph A. Erickson. 1989. “Adolescence and Religion: A Review of the Literature from 1970–1986.” Research in the Social Scientific Study of Religion 1:153–81.
Berger, Peter L. 1967. The Sacred Canopy. Anchor.
——. 1999. “The Desecularization of the World: A Global Overview.” Pp. 1-18 in The Desecularization of the World: Resurgent Religion and World Politics, edited by Peter L. Berger. Eerdmanns.
Beyerlein, Kraig. 2004. “Specifying the Impact of Conservative Protestantism on Educational Attainment.” Journal for the Scientific Study of Religion 43:505-18.
158
Biblarz, Timothy J. and Adrian E. Raftery. 1993. “The Effects of Family Disruption on Social Mobility.” American Sociological Review 58:97-109.
Bohr, Louise, and Ernest Pascarella. 1994. "Cognitive Effects of Two-Year and Four-Year Institutions: A Preliminary Study." Community College Review 22:4-11.
Bohr, Louise, Ernest T. Pascarella, Amaury Nora, and Patrick T. Terenzini. 1995. "Do Black Students Learn More at Historically Black or Predominantly White Colleges?" Journal of College Student Development 36:75-85.
Bramadat, Paul A. 2000. The Church on the World’s Turf: An Evangelical Christian Group at a Secular University. New York: Oxford University Press.
Browning, Don S. 2003. Marriage and Modernization: How Globalization Threatens Marriage and What to Do About It. Grand Rapids, MI.: William B. Eerdmans.
Bryant, Alyssa N. 2005. "Evangelicals on Campus: An Exploration of Culture, Faith, and College Life." Religion and Education 32:1-30.
Bryant, Alyssa N., Jeung Yun Choi, and Maiko Yasuno. 2003. “Understanding the Religious and Spiritual Dimensions of Students’ Lives in the First Year of College.” Journal of College Student Development 44:723-745.
Buchman, Marlis. 1989. The Script of Life in Modern Societies: Entry into Adulthood in a Changing World. University of Chicago Press.
Budziszewski, J. 1999. How to Stay Christian in College: An Interactive Guide to Keeping the Faith. NavPress.
Bureau of Labor Statistics, U.S. Department of Labor. 2006. National Longitudinal Survey of Youth 1997 cohort, 1997-2004 (rounds 1-8) [computer file]. Columbus, OH: Produced by the National Opinion Research Center, the University of Chicago and distributed by the Center for Human Resource Research, The Ohio State University.
Campolo, Tony, and William Willimon. 2002. Survival Guide for Christians: How to be Students and Disciples at the Same Time. Howard Books.
Caplovitz, David and Fred Sherrow. 1977. The Religious Drop-Outs: Apostasy Among College Graduates. Sage Publications.
Cawthon, Tony W. and C. Jones. 2004. “A Description of Traditional and Contemporary Campus Ministries.” College Student Affairs Journal 23:158-172.
Cherry, Conrad, Betty A. DeBerg, and Amanda Porterfield. 2001. Religion on Campus. Chapel Hill: University of North Carolina Press.
159
Clark, Burton R. 1972. Students and Colleges: Interaction and Change. Berkeley: Center for Research and Development in Higher Education, University of California.
Clausen, John A. 1993. American Lives: Looking Back at the Children of the Great Depression. Free Press.
Clydesdale, Timothy T. 2007. The First Year Out: Understanding American Teens after High School. Chicago: University of Chicago Press.
Cohen, Steven Martin. 1983. American modernity and Jewish identity. New York: Tavistock Publications.
Cook, Kaye V. 2000. “You Have To Have Somebody Watching Your Back, And If That.” Adolescence 35:717-730.
Cornwall, Marie. 1989. "The Determinants of Religious Behavior – A Theoretical Model and Empirical Test." Social Forces 68:572-592.
Dannefer, Dale. 1984. "Adult Development and Social Theory: A Paradigmatic Reappraisal." American Sociological Review 49:100-116.
Darnell, Alfred, and Darren E. Sherkat. 1997. “The Impact of Protestant Fundamentalism on Educational Attainment.” American Sociological Review 62:306-15.
Day, James M. 2001. “From Structuralism to Eternity? Re-Imagining the Psychology of Religious Development After the Cognitive-Developmental Paradigm.” International Journal for the Psychology of Religion 11(3):173-183.
De Vaus, David and Ian McAllister. 1987. “Gender Differences in Religion: A Test of the Structural Location Theory.” American Sociological Review 52:472-481.
DeSousa, Jason D., and Patricia M. King. 1992. “Are White Students Really More Involved in Collegiate Experiences than Black Students?” Journal of College Student Development 33:363-69.
Dillon, Michele. 1996. “The Persistence of Religious Identity Among College Catholics.” Journal for the Scientific Study of Religion 35:165-7.
Dillon, Michelle and Paul Wink. 2007. In the Course of a Lifetime: Tracing Religious Belief, Practice, and Change. University of California Press.
Dougherty, Kevin J. 1997. “Mass Higher Education: What is its Impetus? What is its Impact?” Teachers College Record 99:66-72.
Edgell, Penny. 2003. “In Rhetoric and Practice: Defining ‘The Good Family’ in Local Congregations.” Pp. 164-178 in Handbook of the Sociology of Religion, edited by M. Dillon. New York: Cambridge University Press.
160
——. 2005. Religion and Family in a Changing Society. Princeton University Press.
Elder, Glen H., Jr. 1994. “Time, Human Agency, and Social Change: Perspectives on the Life Course.” Social Psychology Quarterly 57:4-15.
Elder, Glen H., Jr. 1998. “The Life Course as Developmental Theory.” Child Development 69:1-12.
Elder, Glen H., Jr.. 1996. “Human Lives in Changing Societies: Life Course and Developmental Insights.” Pp. 31-62 in Developmental Science, edited by Robert B. Cairns, Glen H. Elder, Jr., and E. Jane Costello. Cambridge Studies in Social and Emotional Development. New York: Cambridge University Press.
Emerson, Michael O. and Christian Smith. 2000. Divided by Faith: Evangelical Religion and the Problem of Race in America. Oxford University Press, USA.
Erickson, Joseph A. 1992. “Adolescent Religious Development and Commitment: A Structural Equation Model of the Role of Family, Peer Group, and Educational Influences.” Journal for the Scientific Study of Religion 31:131-152.
Feldman, Kenneth A. and Theodore M. Newcomb. 1969. The Impact of College on Students. Jossey-Bass.
Finke, Roger. 1989. “Demographics of Religious Participation: An Ecological Approach, 1850-1980.” Journal for the Scientific Study of Religion 28:45-58.
Fowler, James W. 1981. Stages of Faith: The Psychology of Human Development and the Quest For Meaning. San Francisco: Harper and Row.
Frantz, T. 1971. “Student and Non-Student Change.” Journal of College Student Personnel 12:49-53.
Freeman, Richard B. 1985. “Who Escapes? The Relation of Church-Going and Other Background Factors to the Socio-Economic Performance of Black Male Youths from Inner-City Poverty Tracts.” NBER Working Paper.
Funk, Richard B. and Fern K. Willits. 1987. “College Attendance and Attitude Change: A Panel Study, 1970-81. Sociology of Education 60:224-231.
Furman, Wyndol. 1989. “The Development of Children's Social Networks.” Pp. 151-172 in Children's Social Networks and Social Supports, edited by D. Belle. New York: Wiley.
Gaede, Stan. 1977. “Religious Participation, Socioeconomic Status, and Belief-orthodoxy.” Journal for the Scientific Study of Religion 16:245-253.
George, Linda K. 1993. "Sociological Perspectives on Life Transitions." Annual Review of Sociology 19:353-373.
161
Glass, Jennifer and Jerry Jacobs. 2005. “Childhood Religious Conservatism and Adult Attainment among Black and White Women.” Social Forces 84:555-579.
Good, J. L. and C. Cartwright. 1998. “Development of Moral Judgment among Undergraduate University Students.” College Student Journal 32:270-276.
Greeley, Andrew M. 1977. The American Catholic: A Social Portrait. Basic Books.
Gross, Neil and Solon Simmons. 2007. “How Religious are America’s College and University Professors?” SSRC Web Forum on the Religious Engagements of American Undergraduates. Available online at http://religion.ssrc.org/reforum/Gross_Simmons.pdf
Halaby, Charles N. 2004. "Panel Models in Sociological Research: Theory Into Practice." Annual Review of Sociology:507-544.
Hammond, Phillip E. and James Davison Hunter. 1984. “On Maintaining Plausibility: The Worldview of Evangelical College Students.” Journal for the Scientific Study of Religion 23:221-238.
Hammond, Phillip E. and Kee Warner. 1993. “Religion and Ethnicity in Late-Twentieth-Century America.” Annals of the American Academy of Political and Social Science 527:55-66.
Helms, Janet E. 1995. “An Update of Helm’s White and People of Color Racial Identity Models.” Pp. 181-198 in Handbook of Multicultural Counseling, edited by J. Ponterotto, J. Casas, L. Suzuki, and C. Alexander. Thousand Oaks, CA: Sage.
Higher Education Research Institute. 2004. The Spiritual Life of College Students: A National Study of College Students’ Search for Meaning and Purpose (Full Report). Los Angeles: Higher Education Research Institute at University of California, Los Angeles.
——. 2006. Spirituality and the Professoriate: A National Study of Faculty Beliefs, Attitudes, and Behaviors. Los Angeles: Higher Education Research Institute at University of California, Los Angeles.
Hill, Jonathan. 2007. “Higher Education and Change in Religious Identity and Practice.” University of Notre Dame, Notre Dame, IN. Unpublished manuscript.
Hoge, Dean R., Gregory H. Petrillo, and Ella I. Smith. 1982. “Transmission of Religious and Social Values from Parents to Teenage Children.” Journal of Marriage and the Family 44:569-580.
Hoge, Dean R. and Larry G. Keeter. 1976. “Determinants of College Teachers’ Religious Beliefs and Participation.” Journal for the Scientific Study of Religion 15:221-35.
162
Hoge, Dean R., and David A. Roozen. 1979. Understanding Church Growth and Decline, 1950-1978. New York: Pilgrim Press.
Hoge, Dean R., and Jann L. Hoge. 1984. “Period Effects and Specific Age Effects Influencing Values of Alumni in the Decade After College.” Social Forces 62: 941-62.
Hoge, Dean R., Benton Johnson, and Donald A. Luidens. 1994. Vanishing Boundaries: The Religion of Mainline Protestant Baby Boomers. Louisville: Westminster/John Knox Press.
Hoge, Dean R., Jann L. Hoge, and Janet Wittenberg. 1987. “The Return of the Fifties: Trends in College Students’ Values Between 1952 and 1984.” Sociological Forum 2:500-519.
Hoover, Stewart M. 2006. Religion in the Media Age. Routledge.
Hout, Michael. 1988. “More Universalism and Less Structural Mobility: The American Occupational Structure in the 1980s.” Americal Journal of Sociology 93:1358-140.
Hunsberger, Bruce. 1978. “The Religiosity of College Students: Stability and Change over Years at University.” Journal for the Scientific Study of Religion 17:159-64.
——. 1983. “Apostasy: A Social Learning Perspective.” Review of Religious Research 25:21-38
Hunt, Keith, and Gladys M. Hunt. 1991. For Christ and the University: The Story of InterVarsity Christian Fellowship of the U.S.A. 1940-1990. Downer’s Grove, Ill.: InterVarsity.
Hunter, James Davison. 1983. American Evangelicalism: Conservative Religion and the Quandary of Modernity. Rutgers University Press.
——. 1987. Evangelicalism: The Coming Generation. University of Chicago Press.
Iannaccone, Larry R. 1994. "Why Strict Churches Are Strong." American Journal of Sociology 99:1180-1211.
Iannaccone, Laurence R. and Michael D. Makowsky. 2007. “Accidental Atheists? Agent-Based Explanations for the Persistence of Religious Regionalism.” Journal for The Scientific Study of Religion 46:1-16.
Ingersoll-Dayton, Berit, Neal Krause, and David Morgan. 2002. "Religious Trajectories and Transitions Over the Life Course." International Journal of Aging and Human Development 55:51-70.
163
Johnson, Daniel Carson. 1997. “Formal Education vs. Religious Belief: Soliciting New Evidence with Multinomial Logit Modeling.” Journal for the Scientific Study of Religion 36:231-46.
Jones, Dale E., Sherri Doty, Clifford Grammich, James E. Horsch, Richard Houseal, and Mac Lynn. 2002. Religious Congregations & Membership in the United States 2000: An Enumeration by Region, State and County Based on Data Reported for 149 Religious Bodies. Glenmary Research Center.
Josselson, Ruthellen. 1996. Revising Herself: The Story of Women's Identity from College to Midlife. New York: Oxford University Press.
King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. “Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation.” American Political Science Review 95:49-69.
King, Valerie, Glen H. Elder Jr., and Les B. Whitbeck. 1997. “Religious Involvement Among Rural Youth: An Ecological and Life-Course Perspective.” Journal of Research on Adolescence 7:431-456.
Knapp, Laura G., Janice E. Kelly-Reid, and Roy W. Whitmore. 2007. "Enrollment in Postsecondary Institutions, Fall 2005; Graduation Rates, 1999 and 2002 Cohorts; and Financial Statistics, Fiscal Year 2005." (NCES 2007-154). U.S. Department of Education. Washington, D.C.: National Center for Education Statistics.
Kohlberg, Lawrence. 1969. Stage and Sequence: The Cognitive-developmental Approach to Socialization. Rand McNally.
——. 1981. The Meaning and Measurement of Moral Development. Clark University Press.
Krause, Neal. 2006. “Church-Based Social Support and Mortality.” Journals of Gerontology Series B: Psychological Sciences and Social Sciences 61:140-146.
Leak, Gary K., Anne A. Loucks, and Patricia Bowlin. 1999. "Development and Initial Validation of an Objective Measure of Faith Development." International Journal for the Psychology of Religion 9:105-124.
Lee, Jenny J. 2002. “Religion and College Attendance: Change among Students.” The Review of Higher Education 25:369-384.
Lee, Jenny, Aurora Matzken, and Sara Arthur. 2006. “Understanding Students’ Religious and Spiritual Pursuits: A Case Study at New York University.” Journal of College and Character 2.
Leege, David C., and Lyman A. Kellstedt. 1993. Rediscovering the Religious Factor in American Politics. Armonk, N.Y.: M.E. Sharpe.
164
Lehman, Edward C. and Donald W. Shriver, Jr. 1968. “Academic Discipline As Predictive of Faculty Religiosity.” Social Forces 47:171-82.
Lincoln, C. Eric, Lawrence H. Mamiya. 1990. The Black Church in the African American Experience. Durham: Duke University Press.
Lindsay, D. Michael. 2007. Faith in the Halls of Power: How Evangelicals Joined the American Elite. Oxford University Press.
Loury, Linda D. 2004. “Does Church Attendance Really Increase Schooling?” Journal for the Scientific Study of Religion 43:119-127.
Lytch, Carol E. 2004. Choosing Church: What Makes a Difference for Teens. Westminster John Knox Press.
MacIntyre, Alasdair C. 1981. After Virtue: A Study in Moral Theory. Gerald Duckworth & Co Ltd.
Mahoney, Kathleen A, John Schmalzbauer, and James Younis. 2001. “Religion: A Comeback on Campus.” Liberal Education 87:36-41.
Marsden, George M. 1994. The Soul of the American University: From Protestant Establishment to Established Nonbelief. New York: Oxford University Press.
Marshall, Victor W. and Margaret M. Mueller. 2003. “Theoretical Roots of the Life-Course Perspective.” In Social Dynamics of the Life Course: Transitions, Institutions, and Interrelations, edited by Walter R. Heinz and Victor W. Marshall. New York: Aldine de Gruyter.
McAllister, Edward W. C. 1985. "Religious Attitudes of Women College Students: A Follow-Up Study." Adolescence 20:797-804.
McConkey, Dale. 2001. “Whither Hunter’s Culture Wars? Shifts in Evangelical Morality, 1988-1998.” Sociology of Religion 62:149-74.
McCullough, Michael E., Craig K. Enders, Sharon L. Brion, and Andrea R. Jain. 2005. "The Varieties of Religious Development in Adulthood: A Longitudinal Investigation of Religion and Rational Choice." Journal of Personality and Social Psychology 89:78-89.
McLeod, Jane D. and Elbert P. Almazan. 2003. “Connections between Childhood and Adulthood.” Pp. 391-412 in Handbook of the Life Course, edited by J.T. Mortimer and M. J. Shanahan. Springer.
McNeel, Stephen P. 1994. “College Teaching and Student Moral Development.” Pp. 26-47 in Moral Development in the Professions: Psychology and Applied Ethics, edited by J. Rest and D. Narvaez. Hillsdale, NJ: Erlbaum.
165
McVeigh, Rory and David Sikkink. 2001. “God, Politics, and Protest: Religious Beliefs and the Legitimation of Contentious Tactics.” Social Forces 79:1425-1458.
Miller, Alan S. and Rodney Stark. 2002. “Gender and Religiousness: Can Socialization Explanations Be Saved?” American Journal of Sociology 107:1399-1424.
Moberg, David O. and Dean R. Hoge. 1986. “Catholic College Students’ Religious and Moral Attitudes, 1961 to 1982: Effects of the Sixties and Seventies.” Review of Religious Research 28:104-117.
Mol, Hans. 1985. The Faith of Australians. Boston: Allen & Unwin.
Monks, James. 2000. "The Returns to Individual and College Characteristics: Evidence from the National Longitudinal Survey of Youth." Economics of Education Review 19:279-289.
Mooney, Margarita. 2005. Religion at America’s Most Selective Colleges: Some Findings from the National Longitudinal Survey of Freshmen (NLSF). Paper presented at the annual meeting of the Association for the Sociology of Religion, Philadelphia, PA, August 2005.
Mosher, William D., David P. Johnson, and Marjorie C. Horn. 1986. "Religion and fertility in the United States: the importance of marriage patterns and Hispanic origin." Demography 23:367-379.
Mueller, Charles W., and Weldon T. Johnson. 1975. "Socioeconomic Status and Religious Participation" American Sociological Review 40:785-800.
Mueller, Daniel, and Phillip W. Cooper. 1986. "Religious Interest and Involvement of Young-Adults - A Research Note." Review of Religious Research 27:245-254.
Muller, Chandra, and Christopher G. Ellison. 2001. “Religious Involvement, Social Capital, and Adolescents’ Academic Progress: Evidence from the National Educational Longitudinal Study of 1988.” Sociological Focus 34:155-183.
Muthén, Bengt. 2004. "Latent Variable Analysis: Growth Mixture Modeling and Related Techniques for Longitudinal Data." Pp. 345-368 in Handbook of Quantitative Methodology for the Social Sciences, edited by D. Kaplan. Newbury Park, CA: Sage Publications.
Myers, Scott M. 1996. “An Interactive Model of Religiosity Inheritance: The Importance of Family Context.” American Sociological Review 61:858-866.
Niebuhr, H. Reinhold. 1929. The Social Sources of Denominationalism. Henry Holt and Co.
Nye, Abby. 2005. Fish Out of Water. New Leaf Press.
166
Nylund, Karen L., Asparouhov, Tihomir, and Muthen, Bengt. 2006. "Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study." Structural Equation Modeling 14:535-569.
Ozarak, Elizabeth Weiss. 1989. “Social and Cognitive Influences on the Development of Religious Beliefs and Commitment in Adolescence.” Journal for the Scientific Study of Religion 28:448-63.
Park, Jerry Z. and Samuel H. Reimer. 2002. “Revisiting the Social Sources of American Christianity 1972-1998.” Journal for the Scientific Study of Religion 41:733-747.
Pascarella, Ernest T. and Patrick T. Terenzini. 1991. How College Affects Students: Findings and Insights from Twenty Years of Research. Jossey-Bass.
——. 2005. How College Affects Students: A Third Decade of Research. Jossey-Bass.
Patillo-McCoy, Mary. “Church Culture as a Strategy of Action in the Black Community.” American Sociological Review.
Peek, Charles W., George D. Lowe, and L. Susan Williams. 1991. "Gender and God's Word: Another Look at Religious Fundamentalism and Sexism." Social Forces 69:1205-1222.
Peshkin, Alan. 1986. God's Choice: The Total World of a Fundamentalist Christian School. University of Chicago Press.
Petersen, Larry R. 1994. “Education, Homogamy, and Religion Commitment.” Journal for the Scientific Study of Religion 33:122-135.
Pew Hispanic Center. 2007. “Changing Faiths: Latinos and the Transformation of American Religion.” Washington, D.C.: Pew Research Center.
Piaget, Jean. 1954. The Construction of Reality in the Child. New York: Basic Books.
Provasnik, Stephen and Linda L. Shafer. 2004. “Historically Black Colleges and Universities, 1976 to 2001 (NCES 2004-062).” U.S. Department of Education, National Center for Education Statistics. Washington, D.C.: Government Printing Office.
Railsback, Gary. 2006. "Faith Commitment of Born-Again Students at Secular and Evangelical Colleges." Journal of Research on Christian Education 15:39-60.
Ramirez, Francisco O. and Christine Min Wotipka. 2001. “Slowly But Surely? The Global Expansion of Women’s Participation in Science and Engineering Fields of Study, 1972-92.” Sociology of Education 74:230-251.
167
Regnerus, Mark D., and Jeremy E. Uecker. 2006. "Finding Faith, Losing Faith: The Prevalence and Context of Religious Transformations during Adolescence." Review of Religious Research 47:217-237.
Regnerus, Mark D., Christian Smith, and Brad Smith. 2004. “Social Context in the Development of Adolescent Religiosity.” Applied Developmental Science 8:27-38.
Regnerus, Mark. 2007. Forbidden Fruit: Sex and Religion in the Lives of American Teenagers. Oxford University Press.
Rindfuss, Ronald R. 1991. “The Young Adult Years: Diversity, Structural Change, and Fertility.” Demography 28:493-512.
Rindfuss, Ronald R., C. Gray Swicegood, and Rachel A. Rosenfeld. 1987. “Disorder in the Life Course: How Common and Does It Matter.” American Sociological Review 52:785-801.
Ritchey, P. N., and B. Dietz. 1990. "Catholic/Protestant Differences in Marital Status" Review of Religious Research 32:65-72.
Roof, Wade Clark and William McKinney. 1987. American Mainline Religion. Rutgers University Press.
Roof, Wade Clark. 1978. Community and Commitment: Religious Plausibility in a Liberal Protestant Church. New York: Elsevier.
Roof, Wade Clark. 1979. “Socioeconomic Differentials among White Socioreligious Groups in the United States.” Social Forces 58:280-289.
Roozen, David A. 1980. "Church Dropouts - Changing Patterns of Disengagement and Reentry." Review of Religious Research 21:427-45.
Roozen, David A., William McKinney, and Wayne Thompson. 1990. "The Big Chill Generation Warms to Worship - a Research Note." Review of Religious Research 31:314-322.
Royston, Patrick. 2005. “Multiple Imputation of Missing Values: Update.” The Stata Journal 5:1-14.
Sandomirsky, Sharon and John Wilson. 1990. “Processes of Disaffiliation: Religious Mobility Among Men and Women.” Social Forces 68: 1211-3.
Schmalzbauer, John. 2003. People of Faith: Religious Conviction in American Journalism and Higher Education. Cornell University Press.
168
——. 2007. “Campus Ministry: A Statistical Portrait.” Web Forum on the Religious Engagements of American Undergraduates. Available online at http://religion.ssrc.org/reforum/Schmalzbauer.pdf.
Schultze, Quentin. 1993. “The Two Faces of Fundamentalist Higher Education.” In Fundamentalisms and Society, edited by Martin Marty and R. Scott Appleby. University of Chicago Press.
Settersten Jr, Richard A. 2003. “Age Structuring and the Rhythm of the Life Course.” Pp. 81-98 in Handbook of the Life Course edited by J.T. Mortimer and M. J. Shanahan. Springer.
Shanahan, Michael J. 2000. "Pathways to Adulthood in Changing Societies: Variability and Mechanisms in Life Course Perspective." Annual Review of Sociology 26:667-692.
Shanahan, Michael J., Erik J. Porfeli, Jeylan T. Mortimer, and Lance D. Erickson. 2005. “Subjective Age Identity and the Transition to Adulthood: When Do Adolescents Become Adults.” Pp. 225-255 in On the Frontier of Adulthood: Theory, Research, and Public Policy, edited by R.A. Settersten, Jr., F.F. Furstenberg, Jr., and Ruben G. Rumbaut. University of Chicago Press.
Sherkat, Darren E. 1998. “Counterculture or Continuity? Competing Influences on Baby Boomers’ Religious Orientations and Participation.” Social Forces 76:1087-114.
——. 2001. “Tracking the Restructuring of American Religion: Religious Affiliation and Patterns of Religious Mobility, 1973-1998.” Social Forces 79:1459-1493.
Sherkat, Darren E., and Alfred Darnell. 1999. “The Effect of Parents' Fundamentalism on Children's Educational Attainment: Examining Differences by Gender and Children's Fundamentalism.” Journal for the Scientific Study of Religion 38:23-35.
Sherkat, Darren E., and Christopher G. Ellison. 1999. "Recent Developments and Current Controversies in the Sociology of Religion." Annual Review of Sociology 25:363-394.
Sherkat, Darren E., and John Wilson. 1995. “Preferences, Constraints, and Choices in Religious Markets: An Examination of Religious Switching and Apostasy.” Social Forces 73:993-1026.
Sikkink, David. 1999. “The Social Sources of Alienation from Public Schools.” Social Forces 78:51-86.
Smith, Christian and David Sikkink. 2003. “Social Predictors Of Retention in and Switching From The Religious Faith of Family of Origin: Another Look Using Religious Tradition Self-Identification.” Review of Religious Research 45:188-206.
169
Smith, Christian, David Sikkink, and Jason Bailey. 1998. “Devotion in dixie and beyond: A test of the "Shibley thesis" sim the effects of regional origin and migration on individual religiosity.” Journal for The Scientific Study of Religion 37:494-506.
Smith, Christian, Melinda Lundquist Denton, Robert Faris, and Mark Regnerus. 2002. “Mapping American Adolescent Religious Participation.” Journal for the Scientific Study of Religion 41:597-612.
Smith, Christian. 1998. American Evangelicalism: Embattled and Thriving. Chicago: University of Chicago Press.
——. 2000. The Secular Revolution: Power, Interests, and Conflict in the Secularization of American Public Life. University of California Press.
——. 2003a. Moral, Believing Animals: Human Personhood and Culture. Oxford University Press.
——. 2003b. "Theorizing Religious Effects Among American Adolescents." Journal for the Scientific Study of Religion 42:17-30.
Smith, Christian. and Melinda L. Denton. 2005. Soul Searching: The Religious and Spiritual Lives of American Teenagers. Oxford University Press, USA.
Smith, Tom W. 2006. "The National Spiritual Transformation Study." Journal for the Scientific Study of Religion 45:283-296.
Solmon and Ochsner 1978. “Attitude Changes of College Graduates.” In Increasing Student Development Options in College, edited by D. Drew. Jossey-Bass.
Spaeth, Joe L. and Andrew M. Greeley. 1970. Recent Alumni and Higher Education: A Survey of College Graduates. McGraw-Hill.
Stark, Rodney and Roger Finke. 2000. Acts of Faith: Explaining the Human Side of Religion. Los Angeles: University of California Press.
Stark, Rodney. 2002. "Physiology and Faith: Addressing the ‘Universal’ Gender Difference in Religious Commitment." Journal for the Scientific Study of Religion 41:495-507.
Steensland, Brian, Jerry Z. Park, Mark D. Regnerus, Lynn D. Robinson, W. Bradford Wilcox, and Robert D. Woodberry. 2000. “The Measure of American Religion: Toward Improving the State of the Art.” Social Forces 79:291-318.
Stolzenberg, Ross M., Mary Blair-Loy, and Linda J. Waite. 1995. "Religious Participation in Early Adulthood - Age and Family-Life Cycle Effects on Church Membership." American Sociological Review 60:84-103.
170
Streib, Heinz. 2001. “Faith Development Theory Revisited: The Religious Styles Perspective.” International Journal for the Psychology of Religion 11(3):143-158.
Sullins, D. Paul. 2006. “Gender and Religion: Deconstructing Universality, Constructing Complexity.” American Journal of Sociology 112:838-80.
Taylor, Charles. 1989. Sources o The Self : The Making of the Modern Identity. Cambridge: Harvard University Press.
Taylor, Mark C. 2006. "The Devoted Student." New York Times, December 21, 2006.
Thalheimer, Fred. 1965. “Continuity and Change in Religiosity: A Study of Academicians.” Pacific Sociological Review 8:101-108.
Thompson, Edward H. 1991. “Beneath the Status Characteristic: Gender Variations in Religiousness.” Journal for the Scientific Study of Religion 30:381–394.
Thornton, Arland, William G. Axinn, and Daniel H. Hill. 1992. "Reciprocal Effects of Religiosity, Cohabitation, and Marriage." American Journal of Sociology 98:628-651.
Trent, James W and Leland L. Medsker. 1968. Beyond High School: A Psychological Study of 10,000 High School Graduates. Jossey-Bass.
Udry, J. Richard. 2003. The National Longitudinal Study of Adolescent Health (Add Health), Waves I & II, 1994–1996; Wave III, 2001–2002 [machine-readable data file and documentation]. Carolina Population Center, University of North Carolina at Chapel Hill.
Uecker, Jeremy E., Mark D. Regnerus, and Margaret L. Vaaler. 2007. "Losing My Religion: The Social Sources of Religious Decline in Early Adulthood." Social Forces 85:1667-1692.
Walter, Tony, and Grace Davie. 1998. "The Religiosity of Women in the Modern West." The British Journal of Sociology. 49:640-660.
Warner, R. Stephen. 1993. “Work in Progress Toward a New Paradigm for the Sociological Study of Religion in the United States.” American Journal of Sociology 98:1044-93.
Welch, Kevin. 1983. “Community Development and Metropolitan Religious Commitment: A Test of Two Competing Models.” Journal for the Scientific Study of Religion 22:167-181.
Welch, Michael R. and John Baltzell. 1984. "Geographic Mobility, Social Integration, and Church Attendance." Journal for the Scientific Study of Religion 23:75-91.
171
Wheaton, David. 2005. University of Destruction: Your Game Plan for Spiritual Victory on Campus. Bethany House.
Wilcox, W. Bradford. 2004. Soft Patriarchs, New Men: How Christianity Shapes Fathers and Husbands. Chicago: University of Chicago Press.
Willits, Fern K. and Donald M. Crider. 1989. “Church Attendance and Traditional Religious Beliefs in Adolescence and Young Adulthood: A Panel Study.” Review of Religious Research 31:68.
Wilson, John and Darren Sherkat. 1994. “Returning to the Fold.” Journal for the Scientific Study of Religion 33:148-162.
Wolfe, Alan. 2005. The Transformation of American Religion: How We Actually Live Our Faith. University Of Chicago Press.
Wuthnow, Robert and Kevin Christiano. 1979. "The Effects of Residential Migration on Church Attendance in the United States" Pp. 257-276 in The Relgious Dimension: New Directions in Quantitative Research, edited by Robert Wuthnow. New York: Academic Press.
Wuthnow, Robert. 1988. The Restructuring of American Religion. Princeton University Press.
——. 1989. The Struggle for America’s Soul: Evangelicals, Liberals, and Secularism. W.B. Eerdmans.
——. 1998. After Heaven: Spirituality in America Since the 1950s. University of California Press.
——. 1999. Growing up Religious: Christians and Jews and their Journeys of Faith. Boston: Beacon Press.
——. 2007. “Can Faith Be More Than a Side Show in the Contemporary Academy?” Web Forum on the Religious Engagements of American Undergraduates. Available online at http://religion.ssrc.org/reforum/Wuthnow.pdf.
Youniss, James and Jacqueline Smollar. 1985. Adolescent Relations with Mothers, Fathers, and Friends. University Of Chicago Press.