Вы находитесь на странице: 1из 36

Public Opinion Quarterly, Vol. 70, No. 4, Winter 2006, pp.

530–564

MEASURING RELIGIOUS IDENTITIES


IN SURVEYS

DUANE F. ALWIN
JACOB L. FELSON
EDWARD T. WALKER
PAULA A. TUFIS

Abstract The purpose of this article is to acquaint the broader pub-


lic opinion research audience with what has been a salient issue within
the community of scholars of religion. We address the question of how
best to conceptualize and measure religious identities in research on
contemporary American society. We consider the main approaches to
the measurement of religious identification with regard to their back-
grounds, their assumptions about the importance of understanding
religious identities in historically relevant terms, and the practical con-
siderations of survey measurement. Using data from the General Social
Survey, particularly recent innovative efforts to obtain information on
subjective association with particular religious traditions and/or move-
ments (e.g., Pentecostal, fundamentalist, evangelical, mainline, or
liberal Protestant), we compare the two main approaches: the traditional
“denominational” approach, where religious identities are assumed
to be associated with religious denominations, and the subjective
approach, where religious identities are assumed to be captured by a set
of “nondenominational” reference categories linked to particular histor-
ical religious traditions or social movements. We conclude that both
approaches have substantial predictive validity, and the most effective
strategy for future research may be one that uses a combination of
approaches, rather than one that relies entirely on a single method of
measurement.

DUANE F. ALWIN is the McCourtney Professor of Sociology and Demography at Pennsylvania State
University. JACOB L. FELSON, EDWARD T. WALKER, and PAULA A. TUFIS are graduate students in the
Department of Sociology, Pennsylvania State University. The authors acknowledge research sup-
port from the Population Research Institute, Pennsylvania State University, and the McCourtney
endowment. We thank Roger Finke, Peter Miller, and anonymous reviewers for providing helpful
comments on an earlier version of the article. Address correspondence to Duane F. Alwin; e-mail:
dfa2@psu.edu.
doi:10.1093 / poq / nfl024
© The Author 2006. Published by Oxford University Press on behalf of the American Association for Public Opinion Research.
All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.
Measuring Religious Identities 531

Introduction
It has been a commonly accepted view among twentieth-century social scien-
tists that religious identities play a powerful role in people’s lives. This view
can be traced to the origins of sociological thought concerning the importance
of religion in social life. For example, Emile Durkheim ([1912] 1995) argued
that religion—defined as a unified system of beliefs and practices oriented to
sacred things—is organized around collective identities that motivate and
maintain beliefs and behavior. Drawing attention to the social differentiation
of religious identities, Max Weber (1922–23 in Gerth and Mills 1958) and
Ernst Troeltsch ([1931] 1960) recognized the importance of the social organi-
zational distinction between “church” and “sect” for understanding religious
behavior in post-Reformation society, and this distinction continues to be an
important basis for the religious identities of individuals in the contemporary
world (e.g., Finke and Stark 1992; Stark and Finke 2000).
In the American context, the study of the history of religion, particularly the
history of Protestantism, involves a study of an array of social movements that
resulted in different religious denominations. It is apparent that such denomi-
national affiliations were critical to the identities of individuals. Building on
the Weber-Troeltsch church-sect typology, Richard Niebuhr (1929) traced the
development of Protestant denominations in American society, concluding
that religious identities were more highly dependent on socioeconomic factors
than theological differences. The pioneering work of Gerhard Lenski (1961)
employed the concept of denominational identity to capture the key elements
of religion that were linked to economic, political, and family life. Lenski
based his argument in part on Will Herberg’s (1960) tripartite distinction for
the American case of Protestant, Catholic, Jew. Herberg had argued that
despite increasing secularization, religion was paradoxically gaining strength;
and with the declining importance of ethnic group identification for indivi-
duals, the distinctions between major religious groups—Protestant, Catholic,
or Jewish—were taking a stronger position in terms of their influences on
“self-identification” (Herberg 1960, pp. 18–39).
At least from Lenski’s work (1961) onward, it is perhaps fair to say that
denominational identities have been the dominant approach to the conceptual-
ization of “religion” and its measurement (e.g., Roof 1999; Roof and McKinney
1987; T. Smith 1987, 1990; Steensland et al. 2000). In fact, when social scientists
make reference to the “religion variable” (or what Lenski [1961] called the
“religious factor”), they often mean (implicitly or explicitly) religious denom-
ination. There is, of course, more to religion than denominational identities,
and several efforts have been undertaken to define a concept of generic religi-
osity. Indeed, it has been argued by several observers (e.g., Alwin 1986,
pp. 434–36) that in some spheres of social life denominational differences
may be less important than other salient aspects of religious experience—for
example, attendance at religious services.
532 Alwin et al.

Recently, the denominational approach to the conceptualization and mea-


surement of religious identities has been seriously questioned, and contempo-
rary researchers have argued that other forms of religious identity—those that
consider the identification of individuals with specific historical religious
movements—can be employed in analyses to more favorably capture the criti-
cal variation in the religious sphere. Christian Smith and his colleagues (1998;
see also Smith and Sikkink 2003), for example, have cautioned against the
exclusive usage of the denominational approach, arguing that religious identi-
ties rooted in labels associated with twentieth-century religious movements or
“historical religious traditions” are the best way to capture important religious
categories that shape social life, at least with respect to Protestant religious
identities. The survey research literature reflects this reorientation, as in the
1990s several political scientists and sociologists introduced the idea of meas-
uring religious identity by asking respondents whether they subjectively iden-
tified as “fundamentalist,” evangelical,” “mainline,” “theologically liberal,”
“charismatic,” or “Pentecostal” (Green et al. 1996; Kellstedt and Smidt 1991;
C. Smith et al. 1998; Wilcox, Jelen, and Leege 1993). Prior to the 1990s,
movement identity questions had been used only with specialized religious
populations such as clergy (Beatty and Walter 1988, 1989). In a national multi-
method study of churchgoing Protestants, Smith et al. (1998, pp. 233–36)
found that most respondents were familiar with the names of Protestant move-
ments and had little trouble identifying themselves with at least one of them.
Following these themes, we explore the concept of “religious identity” and
how it is conceptualized and measured in contemporary survey research. We
contrast these two approaches with respect to Protestant religious identities.1
The first is what we here call the “denominational” approach, wherein reli-
gious identities are assumed to be associated with religious denominations,
and that adherence to particular denominational (and in some cases nonde-
nominational) religious groups is used to capture meaningful individual varia-
tion in religious belief and practice. From the point of view of measurement,
there are several approaches that can be used to assess religious identities,
given a set of responses with respect to denominational preferences (T.Smith
1987, 1990; Steensland et al. 2000). We compare these approaches to a reli-
ance on “movement” identification, or what has been called the “self-identifi-
cation” approach to religious identification (see C. Smith et al. 1998; Smith
and Sikkink 2003), wherein religious identities are assumed to be captured by
a set of “nondenominational” reference categories linked to particular histori-
cal religious traditions—for example, “evangelical” or “fundamentalist” (Kell-
stedt and Smidt 1991; C. Smith et al. 1998; Wilcox, Jelen, and Leege 1993).
We compare the denominational and the self-identification approaches to mea-
suring religious identities and evaluate their similarities and differences with

1. We acknowledge the importance of considering Catholic, Jewish, and other non-Protestant


religious identities as they vary within these broad groupings, but this is beyond the scope of the
present article.
Measuring Religious Identities 533

respect to measures of religious beliefs and behavior. Specifically, we use data


from the 1996, 1998, and 2000 General Social Survey (GSS), which employed
multiple measures of religious identification, tracing the implications of these
results for the study of religious belief and behavior.

The Research Problem


We conceive of contemporary religious organizations as “faith communities”
within which religious identities are lodged. Whether we think of these as
“churches,” “congregations,” “denominations,” or simply as “religious com-
munities,” the assumption is that, through such religious organizations, reli-
gions function to provide an important set of experiences that create and
maintain the identities of individuals, and they therefore, like other social
institutions, provide key socialization and social support functions that help
shape religious identities, beliefs, and practices. Such religious communities,
thus, both reflect the beliefs and orientations of their members and at the same
time shape and influence those same beliefs and orientations. There is a
widely observed correspondence between what are often referred to as “religious
subcultures”—the ecological nexus within which religious identities are
located—and a range of religious and nonreligious beliefs, attitudes, and
behaviors (e.g., Gay and Ellison 1993; Gay, Ellison, and Powers 1996; Gay,
Lynxwiler, and Peek 2001).
In order to understand the link between religious organizations and reli-
gious identities there are at least three important considerations. The first is
that religious identities function in ways similar to other aspects of social
identity. There is a body of social psychological literature that is relevant to
understanding the nature of identities, and this literature can be used to better
understand the nature and development of religious identities (see Burke 2004).
The second is that religious identities must be understood in historically rele-
vant terms. Virtually all religious denominations or other aspects of religious
identity are a complex result of social movements embedded in particular his-
torical circumstances, and it is a mistake to ignore the nature of the historical
process that has resulted in particular forms of religious identification. Finally,
a third consideration of relevance here involves the nature of measurement. In
particular, we assume that the measurement of religious identity in the context
of survey research involves recognizable features of survey measurement that
can be evaluated according to standard criteria of validity and reliability.

What Are Religious Identities?


Identities, according to Rosenberg (1981, p. 601), “are the fundamental bases
upon which society, independent of the special and unique features of each
individual, orders and arranges its members.” It is a basic tenet of interactionist
534 Alwin et al.

social psychology that individuals take on such “identities,” which then


become salient within particular social contexts (Stryker 1980). Social identi-
ties are of two types—those identities taken on voluntarily, such as social
roles, and those resulting from processes of ascription. Regarding the latter,
Rosenberg (1981, p. 601) notes that “scarcely has an infant entered the world
than he or she is immediately classified according to race, sex, religion,
nationality, and so forth.” These ascribed identities are attached to the individual
based on biological or group characteristics, but as we emphasize throughout
this article, religious identities are ultimately a matter of choice. We empha-
size the social psychological nature of religious identities here both for
conceptual reasons and because of their implications for measurement (see
Burke 2004).
One set of processes by which religious identities are constructed involves
what Hewitt (1989, pp. 191–201) refers to as the “strategy of exclusivity”
(although other processes are clearly involved), in which actions and experi-
ences can be meaningfully interpreted by identifying with a single religious
community. This may be a religious denomination, a church/synagogue/tem-
ple, a particular congregation, or a religious movement. According to Hewitt
(1989, p. 193), those who take this strategy “view their community as an
enclave, often portraying it as opposed to society at large or to other commu-
nities . . . . its members are apt to feel that their community is in sole posses-
sion of the truth, and occasionally they will claim moral ownership of
society.” Examples of a pronounced emphasis on this strategy of exclusivity
are Orthodox Jews and fundamentalist Christians, whose day-to-day existence
is organized by the demands of the religious subculture.
An important element in the understanding of social identities involves the
issue of continuity, whereby particular emergent identities are linked across
time. Most identities that develop early in childhood are highly stable over the
lifespan (see Alwin 1994; Wells and Stryker 1988). Although there is clear
evidence that sexual, religious, and political identities have their roots in early
childhood, these types of identities are thought to be formed during adoles-
cence and young adulthood. Identities that develop later in adolescence and
young adulthood may be just as deeply held and therefore may be quite stable
over people’s lives. These types of identities may be susceptible to the influ-
ences of historical factors and social change, and the new identity may have a
powerful effect on the individual’s beliefs and behavior. Students of “religious
switching” have argued that, once established, there is relatively little change
in religious identities (e.g., Sherkat 2001). In the 1988 GSS module dealing
with this topic, for example, after being asked about their current religious
affiliation and the religious denomination in which they were raised, respon-
dents were asked whether they “ever had another religious preference.”
Slightly less than one-third of GSS respondents reported any change in reli-
gious identification, and of those who had switched religious preferences at
least once, the vast majority had made only one change, most of which occurred
Measuring Religious Identities 535

at a young age (Roof 1989, p. 534). The patterns of stability and change were
related to religious denomination: 94 percent of Jews, 84 percent of Catholics,
and 68 percent of Protestants retained their childhood identity (Roof 1989,
p. 532).

The Historical Nature of Religious Identity


The history of religious movements involves competition and conflict
between a variety of religious identities and beliefs. Given the cultural dynam-
ics of the last half-century, it is not surprising that several prominent religion
scholars began to prefer movement identification to denomination as a way to
measure religion in the early 1990s. In an influential book Wuthnow (1988b)
described how the consensus across Protestant denominations in the 1950s
gave way to internecine conflict in the 1960s and 1970s, as theological liber-
als sparred with theological conservatives over the morality of liberalizing
social norms. As conflict increased both across and within denominations,
new religious boundaries emerged. Doctrinal feuds that divided denomina-
tions mattered less. Attitudes about gender roles and perspectives on sexual
morality mattered more. While more liberal churchgoers embraced many of
the societal changes that brought greater freedom for more individuals, con-
servative churchgoers found themselves at odds with a culture they viewed as
increasingly secular.
Since the mid-1980s, sociologists of religion have often classified survey
respondents into schemas organized along a continuum from liberal to conser-
vative based on denominational identities. But in the last ten years or so,
scholars have come to view the largest Protestant denominations—Baptists,
Lutherans, and Methodists—as heterogeneous with regard to basic religious
values and attitudes (Woodberry and Smith 1998). Thus, some scholars have
come to believe that denominational classification schemas are fraught with
excessive measurement error. Also in recent years, more people identify as
“nondenominational” and thus cannot be categorized using traditional meth-
ods. In the context of recent religious dynamics, it makes sense that some
scholars have begun to advocate direct measurement of religious movement
identification. While movement identification has now been used in a wide
range of research analyses (e.g., Beatty and Walter 1988; Denton 2004;
Gallagher 2004; Kellstedt and Smidt 1991; Park and Smith 2000; Schwadel
2005; C. Smith et al. 1998; C. Smith and Sikkink 2003; Woodberry and Smith
1998), it remains secondary to denominations, at least in part because most
surveys include denominations but not religious identities.
An understanding of the nature of variation in religious identities in American
society, thus, cannot be divorced from the historical connections between reli-
gious traditions, especially the historical connections among various Protestant
denominations. While it is apparent that such denominational affiliations are
536 Alwin et al.

critical to the identities of individuals, other aspects of religious movements


are equally important. A critical issue in the understanding of Protestant reli-
gious identification has focused on how to characterize conservative religion’s
response to modernity (Marsden 1980). Throughout the past century there
have been a number of theological divisions along a liberal-conservative
dimension (Hadden 1995; Wuthnow 1988a, 1988b). Prominent within Protes-
tant denominations was the emergence of fundamentalism. It is generally
agreed that fundamentalism resulted from a reaction against the emergence of
liberal and scientific thought—obtaining its name from 12 booklets published
between 1910 and 1915 called “The Fundamentals,” which articulated what
the authors considered to be the fundamental doctrines of the Christian faith
(Ahlstrom 1972, pp. 815–16). Conservative Protestant religious identities are
tied historically to this movement in American Protestantism, and while the
term has gone out of fashion (both in academic circles and among conserva-
tive Protestants themselves), the concept of fundamentalism embodies an
objective set of standards against which Protestant religions can be evaluated
(e.g., T. Smith 1990), as well as a set of subjective categories that can be
applied as an alternative, which may or may not correspond to objective defi-
nitions (C. Smith et al. 1998).
Another concept that is relevant to recent historical Protestant religious
identifications is the term “evangelical,” which in many circles has come to
embrace the general notion of theologically conservative Protestantism
(Hunter 1987, p. 3). This term was prominent in eighteenth- and nineteenth-
century Protestantism and referred generally to an emphasis on the authority
of the New Testament in opposition to the institutional authority of the church
itself, an emphasis on the belief that salvation is achieved by personal conver-
sion to faith in the atonement of Christ, and often marked by an ardent or zeal-
ous enthusiasm for the religious cause. More recently (especially since the late
1970s) the term has been used in a fashion that eschews reference to a desig-
nation of fundamentalist but still holds to a conservative interpretation of the
Bible.
Thus, while the term “conservative” is more often applied in the political
arena, an argument can be made for the use of either of these terms—
“evangelical” or “fundamentalist”—as a reflection of a disposition to preserve
or restore what is established and traditional and to limit change or adaptation
to modern scientific thinking. Analysts tend to prefer one or the other of these
terms for particular lines of religious tradition; and while there may some
value in maintaining these distinctions, we believe a solution to this issue is
probably not possible.2 Suffice it to say that both are a type of conservative
religion—and there are probably more similarities than there are differences.
In the following analysis we employ both concepts, but we do so primarily

2. It is worth noting in this regard that in some circles fundamentalism is viewed “as a faction
within Evangelicalism and not as a movement distinct from Evangelicalism” (Hunter 1987, p. 4;
also see Hunter 1983, pp. 7–9).
Measuring Religious Identities 537

because of the way in which they are used in the approaches we compare.3 In
addition, there are at least two additional elements of religious history—two
recent trends that must be addressed in any discussion of the historical founda-
tions of religious identities. The first is the fact that there are increasing num-
bers of “nondenominational” responses to standard survey questions about
religious preferences (Woodberry and Smith 1998). The second involves the
temporal growth in the “no religious preference” category in American soci-
ety (see Hout and Fischer 2002).

The Measurement of Religious Identities


From the point of view of measuring religious identities in a contextually
valid manner, it is important to understand the critical elements involved in
the standard approaches employed in survey measurement of religion. As
noted above, Lenski (1961) employed a denominational approach to measur-
ing religious identity, using a series of questions that is now typical of the
standard approach to measuring denominational identity. There are two
important features to this approach. First, in a world of religious denomina-
tions, nothing could be more fundamental than unique designations of group
differences. It is essential therefore to tap the meaning that matters to the
individuals themselves. Except for the differentiation between major denomi-
national categories—Catholic, Protestant, Jewish, or something else—the
denominational approach is entirely open-ended. Specifically, Protestants are
allowed to “self-identify” in the sense that they may mention any religious
group, whether recognized by the investigator or not, and they may reject
standard denominational labels altogether. Second, through the use of detailed
coding of the responses to these types of questions, combined with other
information, this approach can then be used to construct categories of reli-
gious denomination that reflect dimensions of interest, such as religious con-
servatism (T. Smith 1990). There are, however, serious difficulties that arise
in the use of these approaches when respondents provide a “nondenomina-
tional” affiliation or do not provide a denomination that can be classified into
standard categories (see Steensland et al. 2000).
In contrast to this approach, other investigators have urged the measure-
ment of other forms of religious identity that can be employed to represent key
variation in the religious sphere. Christian Smith and his colleagues (1998),
for example, advocate an approach that involves “self-identifications” relying
on the measurement of religious identities rooted in labels associated with
twentieth-century religious movements, such as “evangelical, “fundamentalist,”

3. Another subdivision of conservative Protestantism with origins at the beginning of the twentieth
century refers to a subset of denominations that are considered “Pentecostal,” a reference that per-
tains to any of several fundamentalist Protestant groups that emphasize the activity of the Holy
Spirit, stress holiness of living, and express their religious feelings uninhibitedly, as in speaking in
tongues.
538 Alwin et al.

“mainline Protestant,” and “liberal Protestant” (Smith et al. 1998, p. 233).


Although the authors refer to this as an approach involving “self-identifications,”
the measures do not involve open-ended questions. Unlike the denominational
approach discussed above, the subjective identification approach relies on a
fixed set of categories reflecting the religious identities thought to be relevant
by the investigators and therefore are not “self-identifications” in the same
sense as the measures of denominational identities are.

Data and Methods


Here we employ data from the General Social Survey (GSS), a periodic cross-
sectional survey of the noninstitutionalized household population of the conti-
nental United States aged 18 years and older (see Davis and Smith 2003). It
was conducted nearly annually since 1972 on approximately 1,500 respon-
dents per year (surveys were not conducted in 1979, 1981, and 1992), and
beginning in 1994 it became a biennial survey interviewing approximately
3,000 respondents at each time. During the first few years of the GSS, the
samples were either entirely or in part nonprobability samples, but since 1977
a full probability household sample has been employed. Beginning in 1973,
respondents were asked to report the religious affiliation in which they were
raised, as well as their current religious affiliation. As of the writing of this
article, both questions have been included in 25 national cross-sectional sur-
veys, although changes in the manner in which religious denominations were
classified place limitations on the usefulness of the entire GSS series. We
focus solely on data from the GSS series from 1984 through 2002 for current
religious affiliation only (see T. Smith 1990).4
For purposes of this analysis, we employ the GSS question on religious
preference as the denominational approach to measuring religious identity, as
follows: “What is your religious preference? Is it Protestant, Catholic, Jewish,
some other religion, or no religion?” (If Protestant) “What specific denomination
is that?” (If Jewish) “Do you consider yourself Orthodox, Conservative,
Reform, or none of these?” We employ the following categories of religious
identity: (1) Protestant—Conservative denomination; (2) Protestant—
Nonconservative denomination; (3) Protestant—African-American denomination;
(4) Protestant—No denomination given or nondenominational church; (5)
Catholic; (6) Jewish; (7) Other religions; and (8) No religious preference.5 In
developing these categories we rely principally on the influential work of Tom
Smith (1990) and Steensland et al. (2000), although we employ some slight
modifications for our present purposes. Specifically, for the purpose of

4. We employ GSS data from 1984 and after due to the inadequacy of the GSS religion codes
prior to 1984 (see T. Smith 1990).
5. Because of the small number of cases involved, we ignore the differences in Catholic and
Jewish identity in the present study.
Measuring Religious Identities 539

estimating differences among religious groups we employ two strategies—


one that uses Tom Smith’s (1990) distinctions between “fundamentalist” and
“nonfundamentalist” denominations, and a second that employs the Steensland
et al.’s (2000) distinctions between “evangelical” and “mainline” Protestant
denominations. Finally, we should note that the other religious adherents cate-
gory is relatively small and heterogeneous—it contains both other, specified
non-Western religions and Protestant groups that cannot be easily classified
into any of our other categories.
In addition to these questions, in 1996 and 1998 the following question was
employed to obtain a subjective assessment of religious identity: (If Protes-
tant) “When it comes to your religious identity, would you say you are a fun-
damentalist, evangelical, mainline, or liberal Protestant, or do none of these
describe you?” In the 2000 survey the following question was employed:
“When it comes to your religious identity, would you say you are a Pentecos-
tal, fundamentalist, evangelical, mainline, or liberal Protestant, or do none of
these describe you?” In the 2000 survey respondents were allowed to give up
to three different responses; however, this was followed up with a query that
asked for the one that “best describes what you are.” We rely on the latter
response in some cases in the present analysis.6 In the 1998 and 2000 surveys
an effort was made to assess the religious identities of Catholics as well, and
in these two surveys Catholics were asked the following question: “When it
comes to your religious identity, would you say you are a traditional, moder-
ate, or liberal Catholic, or do none of these describe you?”7 While we focus
primarily on Protestant identities in this article, we do include respondents
with non-Protestant identities (using the denominational approach) in the
main body of our analyses in order to gauge the location of Protestants relative
to other groups on the criterion variables.
For purposes of evaluating the utility of these various approaches to mea-
suring religious identity, we employ four dependent variables as criteria (the
text of the survey question is given in parentheses):
1. Belief in Biblical literalism (“Which of these statements comes closest
to describing your feelings about the Bible? (a) The Bible is the actual
word of God and is to be taken literally, word for word; (b) The Bible is
the inspired word of God but not everything in it should be taken liter-
ally, word for word; (c) The Bible is an ancient book of fables, legends,
history, and moral precepts recorded by men.”)
2. Belief in an afterlife (“Do you believe there is a life after death? Yes or
no?”)

6. We use the first response as the main source of information for this variable, except for cases
where the “best” description does not match the first response, in which case we use the “best”
description.
7. For present purposes we ignore the religious identity assessments for Catholics.
540 Alwin et al.

3. Frequency of prayer (“About how often do you pray? Several times a


day, once a day, several times a week, once a week, less than once a
week, never, don’t know?”)
4. Frequency of church attendance (“How often do you attend religious
services? Never, less than once a year, about once or twice a year, sev-
eral times a year, about once a month, two to three times a month,
nearly every week, every week, several times a week, don’t know?”)
The first two of these—belief in Biblical literalism and belief in an
afterlife—were treated as 1–0 outcomes, and we employed logistic regression
in their prediction.8 Belief in Biblical literalism contrasted those who believe
the Bible is “the actual word” to all other responses, and belief in an afterlife
contrasted those who responded “yes” to those saying “no.” The frequency of
prayer was measured using the following categories: several times a day; once
a day; several times a week; once a week; less than once a week; and never.
The frequency of church attendance was measured as follows: every week,
several times a week; nearly every week; two to three times a month; about
once a month; several times a year; about once or twice a year; less than once
a year; and never. Ordered logistic regression techniques were employed in
the prediction of the frequency of prayer and church attendance. We chose
these four variables—Biblical literalism, belief in the afterlife, prayer fre-
quency, and church attendance—because they are widely considered to be
valid measures of religious belief and behavior, and therefore they suit our
purposes as indicators of the influence of religious identifications.9

Results
We present the results of the two basic approaches to measuring religious
identities. We first present the univariate distributions of the subjective reli-
gious identities obtained in the 1996–98 and 2000 GSS surveys and compare
these results with similar distributions obtained using the denominational
identity approach. In the latter case we employ both the approaches due to

8. We reduced the measure of Biblical literalism to two categories (three were used in the survey
question) in order to focus on the contrast between literalism and all other responses. We also
explored three additional belief measures from the GSS—belief in “heaven,” “hell,” and “religious
miracles”—results for which we do not report here. Data for these measures exist for only one-
third of the GSS sample in 1998, and results are therefore somewhat less dependable. Moreover,
these measures are considerably less informative since there is much less variation in these vari-
ables than in the other criterion variables, results for which are presented here. We present only
those results for measures that are available in both 1996 and 1998 and that provide sufficient
variation across the sample and a sufficient number of cases on which to test our hypotheses.
9. On Biblical literalism as an indicator of religious beliefs, see, for example, Achtenmeier
(1980); Boone (1989); Greeley (1993); and Ellison and Sherkat (1993). Studies of belief in the
afterlife include Harley and Firebaugh (1993); Hynson (1975); and Nelsen (1981). Works on the
significance of prayer and church attendance in religious life are Boyd (1999); Hadaway, Marler,
and Chaves (1993); and Hout and Greeley (1987).
Measuring Religious Identities 541

Tom Smith (1990) and Steensland et al. (2000). Following the presentation of
the univariate results, we then cross-tabulate the two denominational
approaches using the 1984–2002 GSS data in order to demonstrate their key
differences, leading to a comparison of the Steensland et al. (2000) approach
to the subjective approaches employed in the 1996–98 and 2000 GSS surveys.
Finally, we predict the four criterion variables—belief in Biblical literalism,
belief in an afterlife, frequency of prayer, and frequency of church attendance—
and summarize the relative ability of these different approaches to measuring
religious identity to predict variation in these outcomes.
Table 1 presents the distributions of subjective religious identification for
respondents who responded “Protestant” to the GSS question: “What is your
religious preference? Is it Protestant, Catholic, Jewish, some other religion, or
no religion?” Two percentage distributions are given in the table for each set
of surveys: one for all respondents and one for Protestants alone.10 The results
for the two sets of surveys appear to be quite different for reasons that are not
altogether clear, although there are some similarities. A major difference in
the two sets of results is that there are substantially fewer “No answer,”
“Don’t know,” and “Other” responses recorded in the 2000 survey. We cannot
account for these differences in terms of the manner in which the question was
posed. There are also substantially more “None” responses in the 2000 survey.
The source of the discrepancy appears to be related in part to the relatively
fewer numbers of responses referred to above (“No answer,” etc.), but also in
part to a diminished frequency of response involving “mainline” Protestant.11
We speculate, although we have no firm basis for this, that the 2000 survey
involved additional probing, which reduced the number of “Don’t know,” “No
answer,” and “Other” responses, creating a substantially greater number of
“None” responses in the 2000 survey compared with what was present in the
1996–98 surveys without the additional probing.
One of the problems with forced-choice questions is that they may suggest
responses that may not have been on the respondent’s mind (Schuman and
Presser 1981). Thus, if the desired approach is to employ “self-identifications”
(e.g., C. Smith et al. 1998), then an open-ended question might have been better.

10. We should note that given the tendency for some groups that are historically derivative of the
Protestant faith and other “nondenominational” Protestants to avoid the Protestant label, there are
understandably some Protestants in the non-Protestant category.
11. The responses obtained from respondents to the General Social Survey differed somewhat
from the responses Christian Smith reports from his Evangelical Identity and Influence Survey
(EIIS), conducted from 1995 to 1997. While the GSS presented respondents with a “none” cate-
gory, Smith’s survey did not present this option. Moreover, Smith’s survey included one addi-
tional probe about religious identity. For these reasons, more respondents in Smith’s survey
identified with one of the religious identities they were offered than in the GSS. Smith et al.
(1998) reported results only for respondents who attended church more often than once a month
or indicated that religion was very important. We compared the GSS results for 1998 with the
Smith et al. (1998) results, finding that the lack of a “none” category raises the number of people
identifying as liberal and, to a lesser extent, mainline and fundamentalist. However, the percent-
age of respondents who identify as evangelical remains the same regardless of whether respon-
dents are given the option of identifying with “none” of the movements read to them.
Table 1. Distribution of Subjective Religious Self-Identifications: General Social Surveys, 1996–98 and 2000

Religious Self-Identification

GSS 1996–98 GSS 2000

Frequency Total (%) Protestant (%) Frequency Total (%) Protestant (%)

Pentecostal — — — 118 4.33 8.67


Fundamentalist 238 4.26 9.97 75 2.73 5.46
Evangelical 296 5.29 12.37 124 4.52 9.06
Mainline 399 7.14 16.70 164 5.99 11.98
Liberal 356 6.37 14.89 200 7.33 14.68
None 756 13.52 31.64 612 22.38 44.81
Other (SPECIFY) 81 1.45 3.39 33 1.21 2.42
Don’t Know 28 0.51 1.18 9 0.33 0.66
No Answer 235 4.21 9.86 31 1.13 2.26
N/A 3201 57.26 — 1368 50.05 —
Total 5591 100.00 100.00 2733 100.00 100.00
Measuring Religious Identities 543

Our results support the conventional wisdom—that what you get depends
entirely on what you ask—although there are interesting findings resulting from
the comparison of the 1996–98 and 2000 surveys. Despite the presentation of
three (Pentecostal, fundamentalist, and evangelical), rather than two (funda-
mentalist and evangelical), conservative categories, the relative proportion of
conservative responses is nearly identical in the two sets of surveys (22.3 per-
cent in 1996–98 and 23.2 percent in 2000). Pentecostals appear to choose the
identity that best suits them in a particular context—either fundamentalist or
evangelical, rather than choosing “None” and downwardly biasing the estimate
of the proportion of conservative Protestants.
In Table 2 we compare the univariate distributions of the subjective reli-
gious identification measures with the denominational religious identification
measures. Here we display the distributions for four approaches: (1) a modi-
fied version of Tom Smith’s (1990) approach to classifying Protestants
(defined by denomination), using FUND to define “conservative” and “non-
conservative” Protestants; (2) the RELTRAD approach of Steensland et al.
(2000) to classifying “evangelical” (conservative) and “mainline” (nonconser-
vative) Protestants; (3) the 1996–98 GSS “subjective” identity approach; and
(4) the 2000 GSS “subjective” identity approach.12 The table also presents the
distribution that includes Catholics, Jewish, Other, and Black Protestants
(using RELTRAD to define this category, see Steensland et al. 2000). In this
table the two “denominational” approaches are applied to the combined 1996–
2000 GSS data.
These results support several conclusions. First, given the reliance on a
common question to define the relative proportion of Protestants, all approaches
generally agree that slightly less than one-half of GSS respondents identify
themselves as Protestant, but the allocation of the respondents to categories
within Protestantism is quite different depending on the approach taken. Sec-
ond, the Smith (1990) and Steensland et al. (2000) approaches differ in their
allocation of Protestants to “conservative” and “nonconservative” categories.
Specifically, the Steensland et al. (2000) RELTRAD approach classifies sub-
stantially more Protestants as “conservative” and fewer as “nonconservative.”
Third, the denominational approach produces vastly more classifiable respon-
dents, owing we believe to the ambiguity of the categories employed in the
subjective identification approach. Some of these categories may be more rec-
ognizable than others. Christian Smith et al. (1998) argue, for example, that
the evangelical label may be one that enjoys greater recognition in contempo-
rary society. On the other hand, the term “mainline” Protestant may be one that
is not in common use in the general population. For example, a respondent
may know that s/he is a “United Methodist” or “Evangelical Lutheran,” but may

12. In this table we combine “fundamentalist,” “evangelical,” and “Pentecostal” in the conserva-
tive category and the “mainline” and “liberal” in the nonconservative category.
Table 2. Distribution of Denominational Religious Identities and Subjective Religious Identities: General Social Surveys,
1996–98 and 2000

Denominational Identity Subjective Identity

Modified Smith (1990) Steensland et al. (2000)

GSS 1996–2000 GSS 1996–2000 GSS 1996–98 GSS 2000

Frequency Total(%) Frequency Total(%) Frequency Total(%) Frequency Total(%)

Conservative Protestant 1487 19.21 1706 22.04 477 9.55 281 10.22
Nonconservative Protestant 1571 20.30 1203 15.54 656 13.13 317 11.56
Black Protestant 593 7.66 593 7.66 355 7.10 238 8.67
Protestant—answer unclassifiable 0 0.00 176 2.27 673 13.48 508 18.51
Protestant—don’t know/no answer 25 0.32 25 0.32 164 3.28 34 1.22
Total Protestant 3676 47.49 3703 47.83 2325 46.54 1378 50.18
Catholics 2145 27.71 2145 27.71 1428 28.58 717 26.11
Jews 173 2.23 173 2.23 110 2.20 63 2.28
Other Religions 576 7.44 550 7.10 363 7.27 186 6.79
No Religion 1123 14.51 1123 14.51 726 14.53 398 14.49
Religion—don’t know/no answer 48 0.62 48 0.62 44 0.88 4 0.14
Total Sample 7741 7741 4995 2746
Measuring Religious Identities 545

not know that sociologists of religion consider these “mainline” religions, and
therefore may not incorporate such a concept into his/her religious identity.
In order to compare the “denominational” and “subjective” approaches, we
employ the Steensland et al. (2000) approach and cross-classify respondents
in categories of this scheme with the categories employed in the 1996–98 GSS
(table 3) and the 2000 GSS (table 4) for Protestant denominations only. What
these tables, taken together, illustrate is that there is quite a wide variety of
subjective labels chosen for members of a given denominational category.
Nearly as many “conservative” Protestants consider themselves “nonconser-
vative” when they are asked to use the subjective categories. For example,
taking the first row of table 3, we can see that of those respondents classified
as “Evangelical Protestants” in the Steensland et al. (2000) scheme, some 11.5
percent consider themselves “fundamentalist,” 13 percent as “evangelical,”
11percent as “mainline,” 9 percent as “liberal,” and 23 percent as “none.” In
the largest Protestant denomination—the Southern Baptist Convention
(SBC)—essentially equal percentages (28.3 and 32.5, respectively) of people
identify as conservative (fundamentalist or evangelical) versus nonconserva-
tive (mainline or liberal) Protestants (data not shown). The same result does
not apply to the “Mainline Protestants”—most of them (35 percent) consider
themselves “nonconservative” (i.e., mainline or liberal), and only about 11
percent consider themselves “conservative.” These examples are borne out in
table 4 as well. Needless to say, there is not a close correspondence between
subjective identities and standard denominational categories.
There are several observations that can be made about these results. First,
there is a relationship between ambiguities in religious identities across the
two (denominational versus subjective) approaches. Higher percentages of
people who gave vague denominational responses also gave “None,” “Don’t
know,” and “Other” responses on the subjective measure. For example, about
50 percent of the Protestant respondents giving a “nondenominational” or “no
denomination” response chose “None/Don’t know/Other” categories. Second,
the detailed examination of these tables reveals that higher percentages of
people belonging to mainline denominations self-identify as “mainline” or
“liberal” than do people belonging to conservative denominations subscribe to
“fundamentalist” or “evangelical” identities. Third, as would be expected, the
smaller denominations in the evangelical group, when aggregated, are more
likely to self-identify as “fundamentalist” or “evangelical” than members of
larger evangelical denominations. Smaller denominations should be more
homogeneous, since their sometimes “sect-like” quality reinforces distinctive-
ness among group members. Even so, the percentage of people belonging to
the smaller evangelical denominations who self-identify as “fundamentalist”
or “evangelical” is no larger than the percentage of people in mainline
denominations who self-identify as “mainline” or “liberal.”
To summarize, when we looked more closely at specific denominations
(data not shown), we found even more support for the conclusion that there is
Table 3. Cross-Tabulation of Denominational Identity and Subjective Identity: General Social Surveys, 1996–98

Subjective Identity

Fundamental Evangelical Mainline Liberal None Other DK N/A Missing Total

N N N N N N N N N N

Col% Col% Col% Col% Col% Col% Col% Col% Col% Col%

Evangelical Protestant 161.00 184.00 155.00 125.00 316.00 41.00 18.00 87.00 314.00 1400.00
67.56 62.16 38.72 35.03 41.86 50.66 62.27 37.10 9.80 25.04
Row% 11.49 13.13 11.04 8.90 22.60 2.93 1.26 6.24 22.41 100.00
Mainline Protestant 39.00 71.00 186.00 153.00 230.00 15.00 6.00 34.00 222.00 956.00
16.33 23.96 46.60 43.11 30.37 18.42 22.64 14.48 6.94 17.11
Row% 4.07 7.41 19.44 16.04 24.01 1.56 0.67 3.57 23.23 100.00
Black Protestant 30.00 32.00 49.00 53.00 142.00 19.00 2.00 21.00 107.00 454.00
12.53 10.81 12.28 14.97 18.75 23.03 7.55 8.82 3.33 8.12
Row% 6.57 7.04 10.80 11.74 31.22 4.11 0.47 4.58 23.47 100.00

NOTE.—The rows of this table employ categories defined by the approach of Steensland et al. (2000).
Table 4. Cross-Tabulation of Denominational Identity and Subjective Identity: General Social Surveys, 2000

Subjective Identity

Pentecostal Fundamental Evangelical Mainline Liberal None Other DK N/A Missing Total

N N N N N N N N N N N

Col% Col% Col% Col% Col% Col% Col% Col% Col% Col% Col%

Evangelical Protestant 71.00 55.00 79.00 60.00 58.00 220.00 18.00 5.00 11.00 34.00 611.00
59.91 74.29 64.22 36.48 28.99 35.98 53.23 58.82 34.52 2.45 22.34
Row% 11.61 9.08 13.00 9.77 9.51 36.04 2.88 0.87 1.75 5.50 100.00
Mainline Protestant 10.00 11.00 37.00 77.00 97.00 185.00 7.00 1.00 13.00 2.00 441.00
8.56 15.00 30.17 46.91 48.67 30.31 20.97 11.77 41.42 0.12 16.12
Row% 2.30 2.54 8.46 17.41 22.13 42.08 1.57 0.24 2.90 0.36 100.00
Black Protestant 31.00 6.00 4.00 22.00 25.00 132.00 4.00 3.00 5.00 7.00 238.00
26.13 7.86 3.45 13.36 12.23 21.52 12.90 29.41 15.43 0.51 8.69
Row% 13.01 2.47 1.79 9.19 10.32 55.39 1.79 1.12 2.01 2.92 100.00

NOTE.—The rows of this table employ categories defined by the approach of Steensland et al. (2000).
548 Alwin et al.

considerable diversity within denominational categories in the subjective


identities respondents apply to themselves. It appears, then, that while there is
some overlap, whatever it is that these two approaches to measuring religious
identity capture, it does not necessarily appear to be the same thing. The above
results support the conclusion that the two approaches seem to be measuring
somewhat different things. The SBC example given above is one prominent
example of why it may be important to consider both denominational identity
and subjective identity in classifying Protestants. As noted by Ammerman
(1987, 1990), the SBC is rife with internal division. In the 1980s, as the SBC
showed signs of “liberal drift,” conservatives fought back, gaining control and
inciting the resentment of self-styled moderates, who were still theologically
conservative but were more flexible and tolerant of diversity in scriptural
interpretation. More recently, Ammerman (1990) argues, the fundamentalists
were in power, but the base followers were a much more diverse group. The
battles that occurred in the mid-1980s both reflected and may have further
promoted “mainline” self-identification in some congregations. Despite this
internal dissension, the denomination avoided schism in part because the SBC
denominational identity is still strong. The Southern Baptist belief that they
are God’s Chosen People—“God’s Last and Only Hope”—reinforces the
point we will make more clearly below, that a reliance on subjective identity
alone, without regard for denominational identity, may illustrate why the
failure to acknowledge relevant church history can lead one astray.
In large denominations like the SBC, individuals may both echo and react
against their denomination in forging a religious identity. This may also apply
to non-Protestant groups, such as Catholics, although we are limiting our
empirical attention in this article to the religious identities of Protestants.
Abbott (2001) has observed the tendency for persons in groups to subdivide
themselves into “self-similar” smaller groups. This could to some extent
describe the contentious factions in the SBC. Despite a strong group identity
that informs their sociopolitical beliefs and attitudes, they also divide into fun-
damentalist/evangelical and moderate/liberal factions in reaction to each
other. A similar process occurs in other denominations, even though theological
differences within denominations may at times be rather slight. Ammerman
(1990) notes, for example, that although both SBC factions are theologically
conservative, individuals perceive the differences among their peers to be
large, and they may react accordingly in considering a subjective religious
identity presented by questions posed by a survey researcher.
In order to evaluate the extent to which these measures produce predictable
relationships, we present patterns of associations of the categories of subjec-
tive religious identities with aspects of religious belief and behavior (belief in
Biblical literalism, belief in an afterlife, frequency of prayer, and frequency of
church attendance) in the 1996–98 GSS data within the three Protestant
denominational categories in the Steensland et al. (2000) scheme—conservative
Protestants, nonconservative Protestants, and black Protestants. Owing to the
Measuring Religious Identities 549

possibility of reciprocal relationships between these criterion variables and the


measures of religious identity, we think of this exercise therefore as one
involving “prediction” rather than one of determining the substantive
“effects” of one set of variables on the other. We thus approach this task from
the point of view of establishing the predictive or “criterion-related” validity
of the two approaches to measuring religious identity.13 The four measures we
have chosen to establish predictive validity are useful in discerning the com-
parative predictiveness of subjective and denominational religious identities,
given that people who adopt more conservative religious identities should
have higher levels of commitment to their faith. According to Stark and Finke
(2000, pp. 144–47), this is in part because religious groups that are in higher
tension with their surroundings, such as more conservative, sect-like groups,
will tend to require much more from their members. In exchange for receiving
higher religious rewards from their religious groups, members of sect-like
groups will tend to express a more otherworldly set of religious beliefs, dem-
onstrate a more Bible-centered religious disciple, endorse the importance of
prayer, and participate at higher levels than members of less sect-like religious
groups. Faced with relatively high demands from their faith, those who adopt
more sect-like religious identities are faced with the choice of either partici-
pating at higher levels on these dimensions of belief and behavior or disaffili-
ating altogether.
The results in figures 1–4 provide strong evidence that the subjective mea-
sures of religious identity are related in predictable ways to aspects of reli-
gious belief and behavior, independent of denominational identities. Generally
speaking—but not always—the ordering of the categories is consistent across
denominational categories. The clearest example of this involves the measure
of belief in Biblical literalism. Predictably, self-identified “fundamentalists”
within each of the three denominational categories have the highest proportion
of agreement with the statement that “the Bible is the actual word of God and
is to be taken literally”; “liberals” have the lowest level of agreement within
each group. The other groups are ordered in predictable fashion. Overall,
black Protestants have the highest endorsement of Biblical literalism, followed
by conservative Protestants and nonconservative Protestants.
Results for belief in an afterlife present a less predictable pattern. In this
case it is “evangelicals” who have the highest endorsement of belief in an
afterlife, with the exception of the nonconservative Protestant denominations
where it is the self-identified fundamentalists who are most likely to endorse
this belief. Protestants responding “none” to the subjective identification question

13. There are other criterion variables that can be used, but because we are measuring religious
identities, we believe this exercise requires specifically religious criterion variables. Although the
approach of Lenski (1961) and the contributors to the Leege and Kellstedt (1993) volume focuses
on the effects of religious identity on factors external to it, we are primarily concerned with the
means by which we may understand the primary independent variable, rather than demonstrating
effects on “nonreligious” outcome variables.
Figure 1. Predicted probabilities of belief that the Bible is the actual word of God among Protestants.
Figure 2. Predicted probabilities of belief in the afterlife among Protestants.
Figure 3. Predicted probabilities of praying more than once per day among Protestants.
Figure 4. Predicted probabilities of attending church more than once per week among Protestants.
554 Alwin et al.

have the lowest endorsement of the afterlife within all groups, slightly lower
than “liberal” and “mainline” Protestants. The overall pattern for belief in the
afterlife is replicated in the results for frequency of prayer (figure 3) and fre-
quency of church attendance (figure 4). In both these measures self-identified
“evangelicals” have the highest expression of religious behavior, followed in
rough order by “fundamentalists,” “mainline,” “liberal,” and “none.” Again, the
patterns are roughly parallel across denominational categories. The “other”
category is difficult to interpret in all of these measures, and there seems to be
little patterning to its levels across these measures.
We also employed regression techniques to ascertain the extent to which
these measures perform as expected, and we found results consistent with the
above findings. These analyses, presented in tables 5 and 6, included two sets
of models: (1) for all respondents in the combined 1996–98 GSS data, and (2)
for the subset of Protestants in these data. For each of the dependent vari-
ables—belief in Biblical literalism, belief in an afterlife, frequency of prayer,
and frequency of church attendance—we present a set of regression models
that assess (1) the importance of religious denominational identities relative to
a set of baseline control variables—education, year of birth, year of survey,
income, southern residence, gender, and race (African-American versus all
others) among all respondents, and (2) the importance of subjective religious
identities relative to a model that includes both baseline predictors and
denominational identities among Protestants. It is necessary to restrict this
second set of models to Protestants only because the measurement of the sub-
jective identities considered here was restricted to that group.
The first set of models in each of these tables presents a set of baseline con-
trol variables as predictors (model 1), followed by a model (model 2) that adds
a set of dummy variables representing the seven categories of denominational
religious identities (labeled “RELTRAD” in the tables): evangelical, mainline,
black Protestant, Catholic, Jewish, other religions, and no religious affiliation
(the last category is the omitted or “reference” category in these two regres-
sion models). These analyses employ logistic regression for the measures of
belief in Biblical literalism and belief in an afterlife and ordered logistic
regression for the measures of frequency of prayer and church attendance.14 In
addition, these tables present a set of models for Protestants that include the
control variables and a set of dummy variables representing the three Protes-
tant denominational categories (here the mainline Protestant group is the ref-
erence category) (model 3) and a model (model 4) that adds a set of dummy
variables representing the six subjective religious identity variables (labeled
“RELIGID” in the tables): fundamentalist, evangelical, mainline, liberal, none,
and other (in this model the “mainline” subjective category is the reference
category).

14. On such models, see Henry (1982); and Winship and Mare (1984).
Table 5. Logistic Regression Results for Religious Beliefs—GSS 1996, 1998

Biblical Literalism Belief in an Afterlife

All Respondents Protestants All Respondents Protestants


N = (2,782) N = (1,016) N = (2,520) N = (926)

Predictor Variables Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

RELTRAD
Evangelical 7.126*** 4.242*** 3.528*** 4.250*** 1.320 1.191
Mainline 1.752*** 2.952***
Black Protestant 3.974*** 2.922*** 2.990*** 2.742*** 1.195 1.297
Catholic 1.308* 2.324***
Jewish 1.201 0.546**
Other 2.091*** 2.279***
RELIGID
Fundamentalist 3.086*** 1.288
Evangelical 1.876*** 2.332*
Liberal 0.507*** 0.922
None 0.963 0.486**
Other 0.919 1.016
CONTROL VARIABLES
Education (centered at 12) 0.863*** 0.874*** 0.904*** 0.893*** 1.042** 1.060*** 1.092** 1.069
Year of birth (centered at 1960) 0.992*** 0.996 0.996 0.999 1.001 1.006 1.005 1.010
Table 5. (Continued)

Biblical Literalism Belief in an Afterlife

All Respondents Protestants All Respondents Protestants


N = (2,782) N = (1,016) N = (2,520) N = (926)

Predictor Variables Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

Year of survey (centered at 1996) 1.054 1.096** 1.016 1.005 1.018 1.049 0.902 0.892
Income (centered at mean) 0.991*** 0.993*** 0.995 0.995 0.998 0.999 1.008 1.008
Southern residence 1.527*** 1.235* 1.379** 1.452** 1.094 0.976 0.853 0.860
Gender (female) 1.041 0.996 0.902 0.974 1.209* 1.121 0.889 0.893
Black 2.471*** 2.120*** 1.778* 1.750* 0.684*** 0.651** 0.464** 0.432**
Log Likelihood –2940.693 –2664.251 –1153.348 –1101.761 –2203.206 –2105.745 –611.137 –589.884
Pseudo R2 0.120 0.242 0.185 0.251 0.012 0.085 0.077 0.148
Increment to R2 — 0.122*** — 0.066*** — 0.073*** — 0.071
% Outcomes predicted correctly 72.90 74.59 64.47 68.01 82.62 82.62 88.88 88.77

*p ≤ .05.
**p ≤ .01.
***p ≤ .001.
Table 6. Ordered Logit Regression Results for Religious Behaviors—GSS 1996, 1998

Frequency of Prayer Religious Service Attendance

All Respondents Protestants All Respondents Protestants


N = (1,596) N = (720) N = (1,560) N = (701)

Predictor Variables Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

RELTRAD
Evangelical 8.619*** 2.452*** 2.153*** 15.733*** 2.536*** 2.022***
Mainline 3.990*** 6.538***
Black Protestant 3.480*** 1.451 1.535 7.957*** 1.296 1.247
Catholic 4.660*** 8.465***
Jewish 0.755 4.108***
Other 7.720*** 7.030***
RELIGID
Fundamentalist 1.295 1.592*
Evangelical 1.655 2.935***
Liberal 0.488*** 0.486***
None 0.446*** 0.419***
Other 0.560 0.874
CONTROL VARIABLES
Education (centered at 12) 1.054* 1.066** 1.089* 1.071* 1.075*** 1.109*** 1.111*** 1.076**
Year of birth (centered at 1960) 0.974*** 0.977*** 0.976*** 0.980*** 0.982*** 0.986*** 0.989* 0.995
Year of survey (centered at 1996) 0.998 1.021 1.071 1.081 1.030 1.094** 1.181** 1.165**
Income (centered at mean) 0.992*** 0.994** 0.998 0.999 0.999 0.999 1.002 1.004
Southern residence 1.370** 1.238 1.260 1.330 1.092 0.935 1.067 1.087
Gender (female) 1.729*** 1.660*** 1.994*** 2.250*** 1.480*** 1.355*** 1.310* 1.376**
Black 2.431*** 3.427*** 2.381* 2.477* 1.966*** 2.229*** 2.222*** 2.453***
Table 6. (Continued)

Frequency of Prayer Religious Service Attendance

All Respondents Protestants All Respondents Protestants


N = (1,596) N = (720) N = (1,560) N = (701)

Predictor Variables Model 1 Model 2 Model 3 Model 4 Model 1 Model 2 Model 3 Model 4

THRESHOLDS
Threshold 1 –3.526*** –2.465*** –4.975*** –5.432*** –1.193*** 0.434*** –1.703*** –2.249***
Threshold 2 –0.580*** 0.750*** –0.591** –0.990*** –0.566*** 1.187*** –0.745*** –1.266***
Threshold 3 –0.125 1.269*** –0.009 –0.381 0.091*** 1.966 0.088 –0.388
Threshold 4 0.590*** 2.059*** 0.976*** 0.647** 0.698*** 2.676 0.849 0.437
Threshold 5 2.007*** 3.567*** 2.391*** 2.135*** 1.030*** 3.051*** 1.211*** 0.834***
Threshold 6 1.481*** 3.547*** 1.735*** 1.421***
Threshold 7 1.783*** 3.872*** 2.073*** 1.798**
Threshold 8 3.221*** 5.373 3.379*** 3.214
Log Likelihood –2481.547 –2365.981 –1041.418 –1014.951 –7806.550 –7397.263 –2702.810 –2614.283
Pseudo R2 0.032 0.077 0.047 0.072 0.013 0.065 0.023 0.055
Increment to R2 — 0.045*** — 0.025** — 0.052*** — 0.032***

*p ≤ .05.
**p ≤ .01.
***p ≤ .001.
Measuring Religious Identities 559

It is clear from these regressions that both the objective, denominational-


based identities and the subjective religious identity variables contribute
significant portions of explained variance. The coefficients for the
“RELTRAD” dummy variables in model 2 indicate that there are substantial
differences among denominational categories in their religious beliefs and
behavior, across all dependent variables. These denominational identities
contributed significant increments to R2—.122 in the case of the belief in
Biblical literalism, .073 for the belief in an afterlife, .045 for frequency of
prayer, and .052 for frequency of church attendance.15 In all cases these fac-
tors exceed the contribution to the explained variance attributable to the set
of controls variables taken alone (see model 1). Looking at the models esti-
mated for Protestants alone, as expected on the basis of the figures presented
above, the subjective identity variables make a significant contribution to
the prediction of religious beliefs and behaviors across all dependent vari-
ables, net of the denominational identities. The increments to R2 in this case
are .066 for the belief in Biblical literalism, .071 for belief in an afterlife,
.025 for the frequency of prayer, and .032 for frequency of church atten-
dance. The patterns suggested by the coefficients in these models fit the
graphic portrayal of these same data in figures 1–4. Of course, the regres-
sion models are additive in the sense that the effects of the subjective iden-
tity variables are constrained equal across religious denominations. In order
to evaluate this constraint we examined a fifth model for each of our crite-
rion variables, which included a set of interactions between the denomina-
tional identity and subjective identity variables (results not presented here).
These results indicate that for the most part the interaction terms were not
significant, although there were a few instances where the pattern of means
for the subjective identity categories was not parallel across denominational
categories, as the patterns in figures 1–4 suggest. These departures from
additivity were registered in the analyses of interaction effects, but taken as
a whole the interactions produced a significant contribution to explained
variance in only one case—belief in an afterlife—but the results in this case
were ambiguous due to substantial multicollinearity among the interaction
and main effect terms. In summary, there was not much to be learned by
including the interactions, and the weight of the evidence suggests that the
denominational and subjective identity measures both contribute impor-
tantly (and additively) to the prediction of variation in the criterion variables
studied here.

15. We employ “pseudo R2” coefficients and increments to “pseudo R2” in this analysis to evalu-
ate their explanatory power (see DeMaris 2002). Also, for the logistic regression models (table 5),
we present the “percent of outcomes predicted correctly” (see Long 1997, pp. 104–7) in order to
provide an additional basis for talking about the improvement in the explanatory power of the
more inclusive models. We know of no way to compute such a measure for the ordered logit
regression models.
560 Alwin et al.

Conclusions
Is there a “gold standard” for assessing religious identities? In some quarters,
the answer to this question is likely to be “yes.” The denominational approach
has dominated the terrain for nearly a half-century. Others have questioned
these traditional practices, and recent research has shown that there is nothing
sacred about this approach, especially given the increasing numbers of nonde-
nominational Christians and others who do not identify with traditional
denominational categories. Given the need to capture those dimensions that
are most salient to understanding individual religious identities, it makes a
great deal of sense to consider using the subjective assessments of religious
identity—the so-called self-identification approach. We believe such an
approach has great potential, but there are several problems with its applica-
tion, and there is much more work that needs to be done. We consider the
basic idea to be sound, but as the above results illustrate, it appears to be
highly dependent on the particular response categories employed and the extent
of effort undertaken on the part of interviewers to secure a valid response.
Moreover, there are large numbers of respondents that are unclassifiable using
the subjective approach, at least as implemented in the GSS interviews.
The most striking finding presented herein involves the amount of variabil-
ity that there is in subjective religious identities within denominational catego-
ries. Our results suggest that a combination of the two approaches may be the
most valid approach to measuring religious identity. We therefore find our-
selves in at least partial agreement with scholars who have recently argued in
favor of the subjective identities approach to the measurement of religion.
Christian Smith and colleagues (1998, p. 233) have argued, “standard approaches
to using denominations (organizational locations) and theology (belief posi-
tions) in religious research, when done well, can offer helpful ways to situate
people in religious identity-space . . . [and] that registering people’s identifica-
tion with different historical religious traditions provides another effective
means of mapping the terrain of religious identity.” We find that each type of
variable appears to measure something different, and to the extent that the cri-
terion variables we have examined here—measures of beliefs and religious
practices—are standards against which to evaluate the predictive validity of
the measures, we have to conclude that both are equally valid. The results pre-
sented here point toward the general conclusion that there is a predictable pat-
terning to subjective religious identities within denominational categories.
Indeed, this supports the conclusions of Hout and Wilde (2004), who argue
that the distinction between subjective and denominational identities should
not be drawn too heavily, as those who are strong members of a denomination
are slightly more likely to identify with one of the four religious movements
identified by Christian Smith and colleagues (1998). These patterns of overlap
between subjective and denominational identities differ somewhat according
to the particular aspect of religious belief and/or behavior considered, but
Measuring Religious Identities 561

there is substantial support for the conclusion that subjective identities are pre-
dictably related to these criteria.
Although there is evidence that the subjective assessments add something to
the explanation of aspects of religious belief and behavior, we believe they are
probably best thought of as supplementary rather than substitute measures. The
denominational and subjective approaches each contribute something unique.
There is a certain amount of exclusive variance attributable to each, in that each
contributes to the predictable variance in our criterion measures while controlling
for the other. Specifically, the addition of the subjective religious identity catego-
ries to a model that includes the denominational religious identity categories pro-
duces significantly greater explained variance. And in this same model the
addition of the subjective religious identity categories does not remove (i.e., does
not completely mediate) the effects of the objective denominational categories.
These results, consistent with those reported by others (see Kellstedt and
Smidt 1991; Wilcox, Jelen, and Leege 1993), suggest that there may be some
benefits to using both approaches. There are several advantages to the denom-
inational approach, among which is the fact that it is essentially “open-ended.”
In contrast to the GSS studies of subjective identification reviewed here, the
denominational approach does not force a set of categories on the respondent,
beyond the broad categories of Catholic, Protestant, Jewish, other, and none.
In addition, the denominational approach has a proven track record, and at
least two classification systems are available to researchers (T. Smith 1990;
Steensland et al. 2000) that reveal a great deal of overlap. There are, however,
problems with implementing such classification systems, owing to the ambi-
guity of the denominational information provided by respondents, as discussed
above, and discrepancies that arise in about 4 percent of the cases. Such mea-
surement can clearly be improved, and this may be one area in which subjec-
tive information can be quite helpful in building denominational classification
systems that have an improved degree of reliability. In addition, and more
important, the use of subjective placement may assist in achieving greater
understanding of important variation in religious identities within denomina-
tional categories that may reflect greater ecological validity in the representation
of the “faith communities” within which religious identities are embedded.
Whether these are conceptualized as “churches” or “congregations” or some
other type of “religious community,” based on the evidence presented here we
believe that, in combination with denominational referents (if relevant), the
use of subjective categories that reflect relevant historical religious move-
ments may go a long way to assessing those social networks that shape reli-
gious identities, beliefs, and practices.

References
Abbott, Andrew. 2001. Chaos of Disciplines. Chicago: University of Chicago Press.
Achtenmeier, Paul. 1980. The Inspiration of Scripture. Philadelphia: Westminster.
562 Alwin et al.

Ahlstrom, Sydney E. 1972. A Religious History of the American People. New Haven, CT: Yale
University Press.
Alwin, Duane F. 1986. “Religion and Parental Child-Rearing Orientations: Evidence of a
Catholic-Protestant Convergence.” American Journal of Sociology 92:412–40.
———. 1994. “Aging, Personality, and Social Change: The Stability of Individual Differences
across the Life Span.” In Life-Span Development and Behavior, ed. David L. Featherman,
Richard M. Lerner, and Marion Perlmutter, vol. 12, pp. 135–85. Hillsdale, NJ: Lawrence
Erlbaum.
Ammerman, Nancy T. 1987. Bible Believers: Fundamentalists in the Modern World. New
Brunswick, NJ: Rutgers University Press.
———. 1990. Baptist Battles: Social Change and Religious Conflict in the Southern Baptist
Convention. New Brunswick, NJ: Rutgers University Press.
Beatty, Kathleen, and B. Oliver Walter. 1988. “Fundamentalists, Evangelicals, and Politics.”
American Politics Quarterly 61:362–74.
———. 1989. “A Group Theory of Religion and Politics: The Clergy as Group Leaders.” Western
Political Quarterly 42:129–58.
Boone, Kathleen C. 1989. The Bible Tells Them So: The Discourse of Protestant Fundamentalism.
Albany, NY: SUNY Press.
Boyd, Heather Hartwig. 1999. “Christianity and the Environment in the American Public.” Jour-
nal for the Scientific Study of Religion 38:36–44.
Burke, Peter J. 2004. “Identities and Social Structure.” Social Psychology Quarterly 67:5–15.
Davis, James A., and Tom W. Smith. 2003. General Social Surveys, 1972–2002: Cumulative Code-
book. Chicago: National Data for the Social Sciences at the National Opinion Research Center.
DeMaris, Alfred. 2002. “Explained Variance in Logistic Regression: A Monte Carlo Study of
Proposed Measures.” Sociological Methods and Research 31:27–74.
Denton, Melinda Lundquist. 2004. “Gender and Marital Decision Making: Negotiating Religious
Ideology and Practice.” Social Forces 82:1151–80.
Durkheim, Emile. (1912) 1995. The Elementary Forms of Religious Life, trans. Karen E. Fields.
New York: Free Press.
Ellison, Christopher G., and Darren E. Sherkat. 1993. “Obedience and Autonomy: Religion and
Parental Values Reconsidered.” Journal for the Scientific Study of Religion 32:313–29.
Finke, Roger, and Rodney Stark. 1992. The Churching of America, 1776–1900. New Brunswick,
NJ: Rutgers University Press.
Gallagher, Sally K. 2004. “Where Are the Antifeminist Evangelicals? Evangelical Identity, Sub-
cultural Location, and Attitudes toward Feminism.” Gender and Society 18:451–72.
Gay, David A., and Christopher G. Ellison. 1993. “Religious Subcultures and Political Tolerance:
Do Denominations Still Matter?” Review of Religious Research 34:311–32.
Gay, David A., Christopher G. Ellison, and Daniel A. Powers. 1996. “In Search of Denomina-
tional Subcultures: Religious Affiliation and ‘Pro-Family’ Issues Revisited.” Review of Religious
Research 38:3–17.
Gay, David A., John P. Lynxwiler, and Charles W. Peek. 2001. “The Effects of Switching on
Denominational Subcultures.” Journal for the Scientific Study of Religion 40:515–25.
Gerth, Hans H., and C. Wright Mills, eds. 1958. From Max Weber: Essays in Sociology.
New York: Oxford University Press.
Greeley, Andrew. 1993. “Religion and Attitudes toward the Environment.” Journal for the Scien-
tific Study of Religion 32:19–28.
Green, John C., James L. Guth, Corwin E. Smidt, and Lyman A. Kellstedt. 1996. Religion and the
Culture Wars: Dispatches from the Front. New York: Rowman and Littlefield.
Hadaway, C. Kirk, Penny Long Marler, and Mark Chaves. 1993. “What the Polls Don’t Show: A
Closer Look at U.S. Church Attendance.” American Sociological Review 58:741–52.
Hadden, Jeffrey K. 1995. “Religion and the Quest for Meaning and Order: Old Paradigms, New
Realities.” Sociological Focus 28:83–100.
Harley, Brian, and Glenn Firebaugh. 1993. “Americans’ Belief in an Afterlife: Trends over the
Past Two Decades.” Journal for the Scientific Study of Religion 32:269–78.
Henry, Frank. 1982. “Multivariate Analysis and Ordinal Data.” American Sociological Review
47:299–307.
Herberg, Will. 1960. Protestant, Catholic, Jew: An Essay in American Religious Sociology.
Garden City, NY: Doubleday.
Measuring Religious Identities 563

Hewitt, John P. 1989. Dilemmas of the American Self. Philadelphia: Temple University Press.
Hout, Michael, and Claude Fischer. 2002. “Americans with ‘No Religion’: Why Their Numbers
Are Growing.” American Sociological Review 67:165–90.
Hout, Michael, and Andrew M. Greeley. 1987. “The Center Doesn’t Hold: Church Attendance in
the United States, 1940–1984.” American Sociological Review 52:325–45.
Hout, Michael, and Melissa J. Wilde. 2004. “The Denominational Society of the USA: A
Reappraisal.” In Patterns and Processes of Religious Change in Modern Industrial Societies:
Europe and United States, ed. Alasdair Crockett and Richard O’Leary, pp. 51–80. Lewiston,
NY: E. Mellen Press.
Hunter, James Davison. 1983. American Evangelicalism: Conservative Religion and the Quan-
dary of Modernity. New Brunswick, NJ: Rutgers University Press.
———. 1987. Evangelicalism: The Coming Generation. Chicago: University of Chicago Press.
Hynson, Lawrence M., Jr. 1975. “Religion, Attendance, and Belief in an Afterlife.” Journal for
the Scientific Study of Religion 14:285–87.
Kellstedt, Lyman, and Corwin Smidt. 1991. “Measuring Fundamentalism: An Analysis of Differ-
ent Operational Strategies.” Journal for the Scientific Study of Religion 30:259–78.
Leege, David C., and Lyman A. Kellstedt. 1993. Rediscovering the Religious Factor in American
Politics. Armonk, NY: M. E. Sharpe.
Lenski, Gerhard. 1961. The Religious Factor: A Sociological Study of Religion’s Impact on
Politics, Economics and Family Life. New York: Doubleday.
Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables.
Thousand Oaks, CA: Sage.
Marsden, George M. 1980. Fundamentalism and American Culture: The Shaping of Twentieth-
Century Evangelicalism, 1870–1925. New York: Oxford University Press.
Nelsen, Hart M. 1981. “Life without Afterlife: Toward Congruency of Belief across Genera-
tions.” Journal for the Scientific Study of Religion 20:109–18.
Niebuhr, H. Richard. 1929. The Social Sources of Denominationalism. New York: Henry Holt
and Company.
Park, Jerry Z., and Christian Smith. 2000. “‘To Whom Much Has Been Given . . .’: Religious
Capital and Community Voluntarism among Churchgoing Protestants.” Journal for the Scien-
tific Study of Religion 39:272–86.
Roof, Wade Clark. 1989. “Multiple Religious Switching: A Research Note.” Journal for the
Scientific Study of Religion 28:530–35.
———. 1999. Spiritual Marketplace: Baby Boomers and the Remaking of American Religion.
Princeton, NJ: Princeton University Press.
Roof, Wade Clark, and William McKinney. 1987. American Mainline Religion: Its Changing
Shape and Future. New Brunswick, NJ: Rutgers University Press.
Rosenberg, Morris. 1981. “The Self-Concept: Social Product and Social Force.” In Social Psy-
chology: Sociological Perspectives, ed. Morris Rosenberg and Ralph H. Turner, pp. 593–624.
New York: Basic Books.
Schuman, Howard, and Stanley Presser. 1981. Questions and Answers in Attitude Surveys.
New York: Academic Press.
Schwadel, Philip. 2005. “Education and Churchgoing Protestants’ Views of Highly Politicized
Christianity.” Review of Religious Research 47:150–61.
Sherkat, Darren E. 2001. “Tracking the Restructuring of American Religion: Religious Affiliation
and Patterns of Religious Mobility, 1973–1998.” Social Forces 79:1459–93.
Smith, Christian, with Michael Emerson, Sally Gallagher, Paul Kennedy, and David Sikkink.
1998. American Evangelicalism: Embattled and Thriving. Chicago: University of Chicago
Press.
Smith, Christian, and David Sikkink. 2003. “Social Predictors of Retention in and Switching from
the Religious Family of Origin: Another Look Using Religious Tradition Self-Identification.”
Review of Religious Research 45:188–206.
Smith, Tom. 1987. Classifying Protestant Denominations. GSS Methodological Report no. 43.
Chicago: NORC.
———. 1990. “Classifying Protestant Denominations.” Review of Religious Research 31:
225–45.
Stark, Rodney, and Roger Finke. 2000. Acts of Faith: Explaining the Human Side of Religion.
Berkeley: University of California Press.
564 Alwin et al.

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.
Stryker, Sheldon. 1980. Symbolic Interactionism: A Social Structural Version. Menlo Park, CA:
Benjamin/Cummings.
Troeltsch, Ernst. (1931) 1960. The Social Teaching of the Christian Churches, trans. Olive Wyon.
New York: Harper.
Wells, L. Edward, and Sheldon Stryker. 1988. “Stability and Change in Self over the Life
Course.” In Life-Span Development and Behavior, ed. Paul B. Baltes, David L. Featherman,
and Richard M. Lerner, pp. 191–229. Hillsdale, NJ: Lawrence Erlbaum.
Wilcox, Clyde, Ted G. Jelen, and David C. Leege. 1993. “Religious Group Identifications:
Toward a Cognitive Theory of Religious Mobilization.” In Rediscovering the Religious Factor
in American Politics, ed. David Leege, Lyman Kellstedt, et al., pp. 72–99. Armonk, NY: M. E.
Sharpe.
Winship, Christopher, and Robert D. Mare. 1984. “Regression Models with Ordinal Variables.”
American Sociological Review 49:512–25.
Woodberry, Robert D., and Christian S. Smith. 1998. “Fundamentalism et al.: Conservative
Protestants in America.” Annual Review of Sociology 24:25–56.
Wuthnow, Robert. 1988a. “Sociology of Religion.” In Handbook of Sociology, ed. Neil J.
Smelser. Beverly Hills, CA: Sage.
———. 1988b. The Restructuring of American Religion: Society and Faith Since World War II.
Princeton, NJ: Princeton University Press.

Вам также может понравиться