National Context, Parental Socialization, and Religious Belief: Results From 15 Nations

Abstract

How much does a nation's religious environment affect the religious beliefs of its citizens? Do religious nations differ from secular nations in how beliefs are passed on from generation to generation? To find out, we use data from the 1991 International Social Survey Programme collected in 15 nations from 19,815 respondents. We use diagonal reference models estimated by nonlinear regression to control for a nation's level of economic development and exposure to Communism, and for the individual's denomination, age, gender, and education. We find that (1) people living in religious nations will, in proportion to the religiosity of their fellow-citizens, acquire more orthodox beliefs than otherwise similar people living in secular nations; (2) in relatively secular nations, family religiosity strongly shapes children's religious beliefs while the influence of national religious context is small; (3) in relatively religious nations family religiosity, although important, has less effect on children's beliefs than does national context. These three patterns hold in rich nations and in poor nations, in formerly Communist nations and in established democracies, and among old and young, men and women, the well-educated and the poorly educated, and for Catholics and Protestants. Findings on the link between belief and church attendance are inconsistent with the influential "supply-side" analysis of differences between nations.

Full text

National Context, Parental Socialization, and Religious Belief: Results from 15 Nations
Author(s): Jonathan Kelley and Nan Dirk De Graaf
Source:
American Sociological Review,
Vol. 62, No. 4 (Aug., 1997), pp. 639-659
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2657431
Accessed: 14/09/2008 14:56
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=asa.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to
American Sociological Review.
http://www.jstor.org

NATIONAL CONTEXT,
PARENTAL SOCIALIZATION, AND RELIGIOUS BELIEF:
RESULTS FROM 15 NATIONS
Jonathan Kelley Nan Dirk De Graaf
Institute of Advanced Studies, Nijmegen University,
The Australian National University The Netherlands
How much does a nation's religious environment affect the religious beliefs
of its citizens? Do religious nations differfrom secular nations in how beliefs
are passed on from generation to generation? To
find out, we use data from
the 1991 International Social Survey Programme collected in 15 nations from
19,815 respondents. We use diagonal reference models estimated by nonlin-
ear regression to control for a nation's level of economic development and
exposure to Communism, andfor the individual's denomination, age, gender,
and education. We find that (1) people living in religious nations will, in
proportion to the religiosity of their fellow-citizens, acquire more orthodox
beliefs than otherwise similar people living in secular nations; (2) in rela-
tively secular nations, family religiosity strongly shapes children's religious
beliefs, while the influence of national religious context is small; (3) in rela-
tively religious nations family religiosity, although important, has less effect
on children's beliefs than does national context. These three patterns hold in
rich nations and in poor nations, in formerly Communist nations and in es-
tablished democracies, and among old and young, men and women, the well-
educated and the poorly educated, andfor Catholics and Protestants. Find-
ings on the link between belief and church attendance are inconsistent with
the influential "supply-side" analysis of differences between nations.
R
eligion remains a central element of
modern life, shaping people's world-
views, moral standards, family lives, and in
many nations, their politics. But in many
Western nations, modernization and secular-
ization may be eroding Christian beliefs,
with profound consequences that have in-
trigued sociologists since Durkheim. Yet this
much touted secularization may be over-
stated-certainly it varies widely among na-
tions and is absent in the United States
*
Direct all correspondence to Jonathan
Kelley,
ISP RSSS ANU, Canberra 0200, Australia
(jonathan.kelley@anu.edu.au),
or to Nan Dirk De
Graaf, Department of Sociology, Nijmegen Uni-
versity, P.O. Box 9104, Nijmegen 6500 HE, The
Netherlands (n.degraaf@mailbox.kun.nl).
Results
"available from the authors"
can be obtained via
the internet at international-survey.org/publica-
tions. Also available on the internet are the raw
data, SPSS specifications for all data transforma-
tions, the transformed
data in a form convenient
for reanalysis, command files, and complete re-
(Benson, Donahue, and Erickson 1989:154-
57; Felling, Peters, and Schreuder 1991;
Firebaugh and Harley 1991; Stark and Ian-
naccone 1994). We explore the degree to
which religious beliefs are passed on from
generation to generation in different nations.
Devout parents socialize their children, in-
culcating religious beliefs in most of them.
But inevitably some offspring break with
their parents' beliefs-especially between
ages 10 and 30 when children come in con-
tact with the wider world, with teachers and
peer groups, when they mature and take jobs,
acquire new friends, and eventually marry
suits for all analyses. An earlier version of this
paper was presented to the International Socio-
logical Association's Research Committee on So-
cial Stratification
and Mobility (RC28) in Zurich,
Switzerland, May 9-11, 1995. We thank James
A. Davis, Rob Eisinga, M. D. R. Evans, Andrew
Greeley, Janne Jonsson, Jan Peters, Frans Van der
Slik, Tom Smith, and Krzysztof Zagorski for
helpful comments.
American
Sociological Review, 1997, Vol. 62 (August:639-659) 639

640 AMERICAN SOCIOLOGICAL REVIEW
and form their own families (Need and De
Graaf 1996). If there were no influences
other than the family to inculcate belief, even
a small loss in each generation would even-
tually produce a secular society.
We propose that one source of the durabil-
ity of religious belief is the religious context
of the nation as a whole: In proportion
to the
orthodoxy of their fellow citizens, people
born into religious nations will acquire more
orthodox beliefs than otherwise similar
people born into secular nations. Following
the literatures on religious socialization,
secularization, and geographic differences in
religious beliefs, we argue that this contex-
tual effect comes about in part through
people's exposure to religious culture (and
perhaps to pro-religious government poli-
cies), and in part because the pools of poten-
tial friends, teachers, colleagues, and mar-
riage partners are predominantly devout.
Conversely, in secular societies, individuals
are likely to acquire secular friends, teach-
ers, colleagues, and marriage
partners
and so
become secular themselves. We test this ar-
gument, which has not previously been rig-
orously tested.' Moreover, we go beyond
these traditional claims, arguing that these
processes interact with the family's religious
background in ways that make family back-
ground more important in secular nations,
but make national context more important
in
religious nations.
THEORY
Devout parents inculcate religious beliefs in
their children directly by explicit teaching
and by example, and indirectly by shaping
their children's views of life (Benson et al.
I An important cross-national literature as-
sesses church attendance, mainly using aggregate
data. In a review, Stark and lannaccone (1994:
239-41) argue that church attendance reflects the
interplay of "potential demand"
for religious ser-
vices (general beliefs in the supernatural
that pro-
vide a motive for attending religious services)
and the "supply" of religious services offered in
each nation, with the "supply-side" characteris-
tics of nations strongly influencing church atten-
dance. Our argument
and data speak to the poten-
tial demand for services, and thus complement
their approach. Unlike Stark and Tannaccone's
analysis, however, ours relies on individual-level
data rather than aggregate-level data.
1989:162-66; Francis and
Brown 1991; Hoge
and Petrillo 1978; Myers 1996). This process
is not unique to religion-parents also incul-
cate moral values, political preferences, and
a wide variety of other attitudes, values, and
preferences (Acock and Bengtson 1978;
Jennings, Allerbeck, and Rosenmayr 1979;
Jennings and Markus 1984).
But parents are not the only force affecting
religious beliefs. Children acquire diverse
friends outside the family, forming peer
groups that by adolescence exert a strong in-
dependent influence on their religious beliefs
(Hoge and Petrillo 1978; Spika, Hood, and
Gorsuch 1985). Children also come into con-
tact with schools and teachers who may shape
their values (Benson et al. 1989:166-67;
Greeley and Gockel 1971). Children are ex-
posed to religious values (or their absence) in
school curricula, the mass media, and the
nation's culture. Some are exposed to gov-
ernment sponsored propaganda that can
shape their views-either pro-religious (as in
Ireland and many Islamic nations) or antireli-
gious (as in Eastern Europe in Communist
times) (Stark and lannaccone 1994:236-39).
In time, children leave home, reducing their
parents'
impact on their
beliefs, behavior, and
values (Need and De Graaf 1996). They ac-
quire new friends, new colleagues at work,
and new peer groups, all of which may shape
their religious views (Wuthnow 1994). Even-
tually, they marry,
founding new families that
become central influences on their lives,
strongly shaping (and being shaped by) their
religious and other values (Hoge and Petrillo
1978; Need and De Graaf 1996). Religious
beliefs thus depend not only on parents' reli-
gious beliefs, but also on the religious con-
tent of school curricula and the mass media,
on the religious policies of the government
and churches, on the general religious con-
tent of the nation's culture and dominant val-
ues, and especially on the religious "environ-
ment"
that
people live in-their friends, peer
groups, schools, teachers, and marriage
partners.
To some extent, people shape their reli-
gious environments by choosing religiously
compatible friends, colleagues, and marriage
partners.
They may even convert their friends
and (especially) their marriage partners to
their own religious beliefs. Thus to some ex-
tent, religious environment
is a consequence

NATIONS, PARENTS, AND RELIGIOUS BELIEF 641
of prior religious convictions, and not an in-
fluence on them.
But a person's religious environment is
also shaped by factors other than their own
and their parents' religious beliefs, and hence
is a potential cause of those beliefs, not a
consequence of them. Friends and spouses
are chosen on the basis of propinquity,
per-
sonality, interests, education, occupation, po-
litical views, sheer chance, and many other
traits besides religion. Schools are chosen on
the basis of quality, cost, or location, not just
religious compatibility. Jobs and employers
are chosen more on the basis of opportunity,
pay, interest, and propinquity
rather
than for
religious reasons.
We argue that prominent among these
"unchosen" aspects of one's religious envi-
ronment is birthplace. People have no con-
trol over where they are born. Hence they
have no choice about whether they grow up
in a predominantly religious or predomi-
nantly secular culture, and no choice about
whether they are exposed to pro- or antireli-
gious policies and propaganda
by church and
state. If they are born into a predominantly
devout nation, the pool of potential friends,
teachers, colleagues, and marriage partners
will differ from the pool available in a pre-
dominantly secular nation. As a conse-
quence, and other things being equal, people
born into a religious nation are likely to ac-
quire religious friends, teachers, colleagues,
and marriage partners and therefore are
likely to become (or remain) religious. Con-
versely, people born into secular nations are
likely to acquire secular friends, teachers,
colleagues, and spouses, and therefore are
likely to become secular themselves. The
same logic applies to people born
during par-
ticularly religious or irreligious periods of a
nation's history, or in particularly religious
or irreligious geographic regions. Hence:
Hypothesis 1: People born into religious na-
tions will, in proportion
to the orthodoxy
of their fellow-citizens, acquire more or-
thodox beliefs than otherwise similar
people born into secular nations.
This type of contextual hypothesis, using
the dependent variable to define the context,
has a long history in sociology and political
science (e.g., Blalock 1984:353-59; Prze-
worski 1974). In our case, the dependent
variable is the individual's religious belief,
and nations are classified according to the
average intensity of their citizens' religious
beliefs. At first glance, such reasoning may
seem circular, but such relationships are far
from tautological (Blalock 1984:363-69;
Boyd and Iversen 1979:22-59). For instance,
the familiar and seemingly obvious ecologi-
cal hypothesis that local political context in-
fluences voting behavior (Butler and Stokes
1974:130-37) turns out on closer analysis to
be false (Kelley and McAllister 1985).
Further, we argue that national religious
context has different effects on the strategies
of devout and secular families. In a predomi-
nantly secular society, children are likely to
acquire secular friends, teachers, work col-
leagues, and marriage partners. This poses a
serious problem for devout parents and their
churches: To ensure that their children ac-
quire and retain orthodox religious beliefs,
they need to control the children's social en-
vironments and restrict their choices of
friends to those with compatible religious
beliefs. They do this by screening potential
friends, teachers, and marriage partners; by
enrolling their children in church groups or
sending them to religious schools to ensure
an appropriate pool of potential friends and
marriage partners;
by socializing their chil-
dren to reject the irreligious; and in many
other ways (Benson et al. 1989:164-67;
Stark and McCann 1993). Insofar as parents
succeed in controlling their children's social
environments, they effectively shut out most
of the irreligious national environment's
secularizing pressures. Hence, the effect of
the family's religious background will be
large and the effect of national environment
will be small.
Conversely, in a predominantly religious
society, devout parents need not worry about
the possibility of their children acquiring
secular beliefs from friends, teachers, col-
leagues, or spouses, because almost everyone
is devout. Therefore, devout parents need not
invest time, effort, or money in controlling
their children's social environments; they
need not endure the emotional strain and po-
tential conflict this imposes on parent-child
relations; and they need not accept the po-
tential loss of desirable friends and marriage
partners
that such restrictions would impose
on their children.

642 AMERICAN SOCIOLOGICAL REVIEW
Irreligious parents face the opposite con-
straints. In a secular society, they need not
worry about their children getting into a de-
vout social environment, because most po-
tential friends, colleagues, and marriage part-
ners are secular. In a religious society, how-
ever, their children are at risk of being drawn
into a devout social environment.
Thus, if the
prospect of their children acquiring
religious
beliefs is distasteful, parents have a strong
incentive to control their children's environ-
ment.
We suggest, however, that most secular
parents will not strongly object to their chil-
dren accepting some religious tenets. Few
secular parents
are committed atheists-most
are agnostic or believe in some vague higher
power if not in a personal, anthropocentric
god (Greeley 1992:66-68). Even if they are
convinced there is no God, parents may see
little harm in their children becoming reli-
gious-the duties imposed by religion are
rarely onerous, the emotional support and
sense of meaning and purpose religion pro-
vides are valuable (Poloma and Pendelton
1990), and there is usually no "antichurch"
institution encouraging secularism (save
those countries under Communist rule).
Moreover, being secular in a devout nation
can have practical disadvantages because of
prejudice on the part of the religious, and the
restrictions that prejudice can impose on
choices of friends, schools, jobs, and mar-
riage partners. Hence, we suggest, most
secular parents will make little effort to in-
sulate their children from the religious pres-
sures of a devout society. Nor will secular
parents prevent their children from acquiring
devout friends, colleagues, and marriage
partners.
As a consequence, many children of
secular parents will become religious. Thus:
Hypothesis 2a: In relatively secular societies,
devout families usually insulate their
children from secular pressures, hence
family background strongly shapes reli-
gious beliefs while national influences
are small.
Hypothesis 2b: In relatively devout societies,
secular families do not usually insulate
their children from religious pressures,
hence family background
has little effect
on religious beliefs and national influ-
ences are large.
DATA
Data are from the 1991 "Religion" module
of the International Social Survey Pro-
gramme (ISSP), an international
consortium
composed primarily of academic survey or-
ganizations (Zentralarchive 1993). Each
year, the ISSP creates a module containing
exactly the same questions, answer catego-
ries, and sequencing for all countries sur-
veyed. This module is then fielded in con-
junction with each country's regular annual
survey. Each country also collects detailed
data on background and demographic vari-
ables, using questions appropriate to local
circumstances and institutions.
The samples are all large, representative
national samples of adults. The most com-
mon procedure is to hold face-to-face inter-
views with a stratified random sample (occa-
sionally a panel on an earlier sample), fol-
lowed by a leave-behind self-completion
questionnaire containing the ISSP module
(Zentralarchive 1993). Some surveys are
conducted entirely by interview, and a few
are entirely self-completion. Completion
rates average over 60 percent (counting
losses both at the interview and the self-
completion stages). These rates compare fa-
vorably with recent experiences in many in-
dustrial
nations. For example, the completion
rates for the highly regarded 1989 "Interna-
tional Crime Victimization Survey" averaged
41 percent over 14 nations (Van Dijk, May-
hew, and Killias 1990). ISSP data appear
ac-
curately to reflect opinion in each nation and
have been widely used in international com-
parisons (e.g., Kelley and Evans 1993, 1995).
Data were processed by the Zentralarchiv
(Scheuch and Uher 1993).
In all there are 19,815 cases: 2,203 in Aus-
tralia
(Kelley, Evans, and Bean 1993); 984 in
Austria (Haller and Hoellinger 1993); 1,486
in East Germany (Mohler and Braun 1992);
1,257 in Great Britain (Jowell et al. 1992);
1000 in Hungary (Kolosi 1992); 1,005 in Ire-
land (Ward and Whelan 1992); 983 in Italy
(Calvi 1992); 1,635 in the Netherlands (So-
cial and Cultural Planning Office 1992);
1,070 in New Zealand (Gendall 1992); 838
in Northern Ireland (Jowell et al. 1992);
1,506 in Norway (Norwegian Social Science
Data Services 1992); 1,063 in Poland
(Cichomski 1992); 2,080 in Slovenia (Tos

NATIONS, PARENTS, AND RELIGIOUS BELIEF 643
Table 1. Percentage Distribution of Responses to Questions about Religious Beliefs: Pooled Data for
15 Nations, 1991
Scoring
Questionnaire
Item Percent (in Points)
Please indicate which statement below comes closest to expressing what
you believe about God. (Degree of Belief)
I don't believe in God ................................................................... 12 0
I don't know whether there
is a God and
I don't believe there
is any way to find out... 10 20
I don't believe in a personal
God, but I do believe in a higher power
of some kind
...... 15 40
I find myself believing in God some of the time, but not at others
.............. ....................
9 60
W hile I have doubts,
I feel that I do believe in God ..............................................
............ 20 80
I know God really exists and
I have no doubts
about
it
.....................................................
34 100
Number
of cases 19,528
Mean 63
Loading
on factor I a .92
Which best describes your beliefs about God? (Belief Timing)
I don't believe in God now and
I never have
................................................. 16 0
I don't believe in God now, but I used to .............. ................................... 14 33
I believe in God now, but I didn't use to ............. .................................... 6 67
I believe in God now and I always have ............. .................................... 64 100
Number
of cases 17,002
Mean 72
Loading
on factor I
a .89
How close to you feel to God most of the time? (Feel Close)
Don't believe in God
............................................... 16 0
Not close at all
.1............................................... 1 25
Not very close........................................................................................................................
21 50
Somewhat
close ............................................... 38 75
Extremely
close ............................................... 16 100
Number
of cases 17,710
Mean 57
Loading
on factor la .92
There is a God who concerns Himself with every human being, personally. (God Cares)
Strongly
agree........................................................................................................................
19 100
Agree......................................................................................................................................
.24 75
Neither agree nor disagree
.................. 18 50
Disagree
.................. 19 25
Strongly
disagree...................................................................................................................
190
Number
of cases 17,614
Mean 51
Loading
on factor la .81
a Factor loadings are from a principal axis factor analysis with commonalities estimated iteratively.
Country-specific factor analytic results and correlations are available from the authors at
www
.international-survey.org.

644 AMERICAN SOCIOLOGICAL REVIEW
and Stebe 1992); 1,346 in West Germany
(Mohler and Braun 1992); and 1,359 in the
United States (Davis and Smith 1991). In to-
tal, there are 17,635 cases with complete data
on all variables in our model.2
MEASUREMENT
Religious Belief
Our religious belief scale, which conceptu-
ally and empirically resembles standard reli-
gious orthodoxy scales used in previous re-
search (Felling et al. 1991; Van der Slik
1994), measures belief in a supernatural
be-
ing who is concerned with each individual
human (Table 1). Most of the items have
been used previously in the NORC General
Social Survey in the United States, the
SOCON survey in the Netherlands, or the
World Values Survey in many nations.
In our pooled sample, 12 percent do not
believe in God, 10 percent are agnostic, 15
percent believe not in a personal God but in
a higher power of some kind, 9 percent be-
lieve in God some of the time but not at
other times, 20 percent believe but have
some doubts, and 34 percent believe in God
and have no doubts (Table 1, item 1). Only
1 percent did not answer. We score these an-
swers conventionally in equal intervals,
from a low of 0 ("do not believe in God") to
a high of 100 ("believe and have no
doubts"). This scoring gives a clear and con-
venient metric (Evans, Kelley, and Kolosi
1992:468-69), but any other equal-interval
scoring would lead to mathematically iden-
tical standardized results and metric results
differing only by a linear transformation.
Using our scoring, belief in God averages
63 points out of 100, with a standard
devia-
tion of 36.
Using this 0 to 100 scoring, answers to a
question which focuses on stability of belief
over time average 72 points (question 2),
feelings of personal closeness to God aver-
age 57 (question 3), and degrees of belief in
a God who concerns himself with every per-
son (question 4) average 51.
2 We do not use the Israeli
dataset,
the Russian
dataset (which
omits some key variables),
or the
Philippines
dataset
(which
has a serious
process-
ing error).
These four items are highly correlated in
all 15 nations, suggesting that they all mea-
sure a single underlying factor (Table 2). In
the pooled sample, inter-item correlations
average .79 with a scale reliability (alpha) of
.93; reliabilities are between .79 and .95 in
each nation separately. Factor loadings aver-
age .88 in the pooled analysis, with equally
high figures in separate analyses for each
country. The four items also have very simi-
lar correlations with other variables in the
model both in the pooled sample and within
each nation (results available from the au-
thors).
Our religious orthodoxy scale is the aver-
age of answers to the four questions in Table
1. Respondents who answered some but not
all questions are assigned the average of the
questions they did answer. Those who failed
to answer any of the questions (1 percent) are
omitted from the analysis.
Parents' Church Attendance
We measure the religious orientation of the
family in which each respondent was raised
by their mother's and father's church atten-
dance when the respondent was 14 or 15
years old. Previous research indicates that
church attendance-a clear-cut behavior-is
reliably reported and is generally the key
family influence, with strong direct and indi-
rect effects on respondent's religion (Benson
et al. 1989:163-66; Hoge and Petrillo 1978).
Because we are interested in the overall ef-
fect of the family, we average mother's and
father's church attendance. If data were
available for only one parent, we used that.
Only 7 percent failed to answer for at least
one parent.3
Mother's and father's church at-
tendance are highly correlated (r = .77 in the
pooled sample), and have similar correlations
with other variables (results available on re-
quest). Averaging them gives a reliable mea-
sure (alpha = .93 in the pooled sample);
separating
them would unnecessarily compli-
cate the analysis and would be difficult be-
cause of their high correlation. Previous re-
search on their relative importance is incon-
clusive (Benson et al. 1989:165-66; Acock
3
An analysis
excluding
cases with
missing
data
on either
parent's
church
attendance leads to the
same conclusions as our
preferred
model.

NATIONS, PARENTS, AND RELIGIOUS BELIEF 645
Table 2. Belief in a Personal God: Inter-Item Correlations, 15 Nations 1991
Degree of Belief a Belief Timingb Feel Closec
Belief
Nation Timingb Feel Closec God Caresd Feel Closec God Caresd God Caresd
All .83 .84 .76 .83 .71 .76
Australia .84 .84 .81 .80 .71 .79
Austria .62 .67 .57 .71 .57 .64
East Germany .84 .87 .74 .87 .75 .77
Great Britain .83 .84 .78 .81 .72 .80
Hungary .81 .81 .65 .84 .62 .66
Ireland .63 .55 .46 .50 .41 .59
Italy .74 .72 .61 .71 .50 .58
Netherlands .84 .87 .81 .88 .75 .78
New Zealand .82 .83 .79 .80 .71 .78
Northern Ireland .68 .63 .58 .58 .54 .60
Norway .84 .88 .84 .85 .76 .83
Poland .62 .56 .44 .54 .41 .48
Slovenia .83 .88 .69 .87 .66 .70
United States .55 .58 .56 .61 .48 .63
West Germany .63 .76 .65 .74 .61 .68
Note: Complete question wording for all variables is in Table 1. Reliability (alpha) is .93 in the pooled
sample and ranges from .83 to .95 for each nation separately.
a Degree of belief in God, from "I don't believe in God" to "I know God really exists and I have no
doubts about it."
b Belief timing from "I don't believe in God and never have" to "I believe in God now and always have."
c "How close do you feel to God most of the time?"
d "There is a God who concerns Himself with every human being, personally."
and Bengtson 1978), so there should be little
loss in ignoring these differences.
We group parents' church attendance into
five groups (scored 1 to 5 in the analysis):
Secular(l) Parents never attend church, or
have no religion;
(2) Parents attend about once a year;
(3) Parents attend several times a
year up to once a month;
(4) Parents attend several times a
month up to almost every week;
and
Devout (5) Parents attend every week or
more often.
One complication is that church attendance
might have different meanings in different
nations. Regular church attendance could in-
dicate that parents are unusually devout in
nations in which church attendance is atypi-
cal or politically unpopular (as in the former
East Germany during the Communist era), or
where "monopolistic" churches make little
effort to provide attractive and diverse ser-
vices (Stark and Iannaccone 1994). If so, this
could introduce some error in our analysis,
which assumes that the meaning of parents'
church attendance is comparable in all 15
nations. Lacking data on parents' beliefs, we
can not test for this possibility directly. But
we can test it indirectly by looking at the re-
lation between respondent's church atten-
dance and the intensity of respondent's reli-
gious belief, on the reasonable assumption
that this reflects the usual pattern for the
respondent's nation (Table 3).
Regular church attendees have much the
same religious beliefs in all the nations for
which we have data (Table 3). Those who at-
tend weekly average 89 points (out of 100)
in religious belief, and this figure varies only

646 AMERICAN SOCIOLOGICAL REVIEW
Table 3. Respondent's Mean Religiosity Score by Respondent's Church Attendance: 13 Nations, 1991
Frequency of Church Attendance
Almost A Few Number
Country Weekly Weekly Monthly Times a Year Never of Cases
Total 89 83 78 61 34 16,466
Australia 92 86 78 61 38 2,147
East Germany 87 78 a 42 12 1,475
Great Britain 88 82 77 65 45 1,203
Hungary 89 84 76 56 28 984
Ireland 87 80 76 67 49 1,003
Italy 88 80 78 67 38 981
Netherlands 88 77 71 58 33 1,620
New Zealand 91 82 78 58 41 1,045
Northern Ireland 91 84 81 74 65 820
Norway 96 95 85 61 32 1,462
Poland 89 85 78 63 33 1,057
United States 93 89 82 76 63 1,327
West Germany 85 79 a 58 34 1,342
Note: Respondent's church attendance was not asked in Austria or Slovenia.
a Category not available in survey
slightly from a low of 85 in West Germany
to a high of 96 in Norway. Those who attend
church almost every week are a little less de-
vout on average, those who attend monthly
are a little less devout again, and those who
attend only a few times a year are noticeably
less devout; but for all groups the pattern is
similar in all 13 nations. The largest varia-
tion occurs among those who never attend
church: Their religious beliefs average 34
points, indicating that for the most part they
do not believe in a God who is personally
concerned with each human being but they
are uncertain about the matter, or they be-
lieve in some kind of higher power. However
East Germans who never attend are notice-
ably less devout than the average, scoring
only 12 points, while Americans (63 points)
and the Northern
Irish (65 points) are notice-
ably more devout. Thus, there is some uncer-
tainty about
parents
who never attend
church.
To test whether these difficulties affect our
results, we excluded the deviant cases from
the analysis (East German, American, and
Northern
Irish respondents raised by parents
who never went to church) and reestimated
our preferred
model. The results are virtually
unchanged (results available from the au-
thors), and we conclude that these uncertain-
ties are not consequential for our analysis.
Note that these findings are not consistent
with the influential "supply-side" argu-
ment-that nations with religious monopo-
lies have large unmet religious needs, while
churches in religiously competitive nations
like the United States more successfully meet
the population's diverse religious needs. This
argument implies that those who do not at-
tend church are more devout in monopolistic
societies. But our findings show exactly the
opposite: People who don't attend church are
actually more devout in religiously competi-
tive societies like the United States and
Northern
Ireland.
Secular and Religious Nations
Measuring how secular or religious a
respondent's
national environment
was when
he or she was growing up, requires a mea-
sure of the religiosity of other people in the
nation-the people among whom the respon-
dents would find their friends, teachers, and
spouses. We measure national religiosity by

NATIONS, PARENTS, AND RELIGIOUS BELIEF 647
an unweighted average of parental church at-
tendance in the nation as a whole and reli-
gious belief in the nation as a whole. Char-
acterizing contexts by the average level of
the dependent variable, or some similar vari-
able, is usual and appropriate (Blalock 1984:
363-68; Boyd and Iversen 1979:22-59). It
would not be wise to characterize each na-
tion separately using dummy variables
(Blalock 1984:357; Farkas 1974). We di-
vided the nations into five groups ranging
from most secular to most religious (scored
1 to 5 in the analysis):
Secular (1) East Germany
and Norway;
(2) Great Britain, New Zealand,
Australia, Hungary, Slovenia,
the Netherlands;
(3) West Germany
and Austria;
(4) United States and Italy;4
and
Religious (5) Northern Ireland, Poland, and
Ireland.
Other plausible measures give much the
same ranking of nations and lead to virtually
identical results (results available from the
authors).
Other Contextual Characteristics
of Nations
Modernization theory suggests that reli-
gious belief will decline as nations become
more prosperous, educated, and modern
(e.g., Becker and Vink 1994; Peters 1993).
To take these factors into account, we mea-
sure a nation's level of development by its
gross national product per capita in 1987, at
parity purchasing power (World Bank
1996). For clarity, we express these figures
as percentages of the U.S. GNP, ranging
from a low of 21 (Poland) to a high of 100
(United States).
For some 40 years preceding our surveys,
many Communist nations followed explicit
antireligious policies. We take this into ac-
count by including a dummy variable coded
4Although the United States and Italy have
very different
religious
institutions
("supply")-
an established
monopolistic
church
in Italy, but
religious diversity
in the United States-the key
issue for our analysis
is the level of belief in the
general public ("demand"),
and
that is similar in
the two countries.
1 for formerly Communist nations and 0
otherwise.
A large literature argues for "American
exceptionalism" in religion. In theory,
America's competitive, entrepreneurial,
open
"religious market" contrasts so sharply with
the typical European situation-in which
there is an established church with a reli-
gious monopoly-that two different para-
digms seem appropriate
(Warner 1993). Em-
pirically, the traditional paradigm (Berger
1969) based on European
experience sees the
high levels of religiosity in the United States
as atypical of modern industrial
societies. We
therefore include a dummy variable coded 1
for the United States and 0 otherwise.
Individual Characteristics
Because Catholics and Protestants differ in a
variety of ways (Peters and Schreuder 1987;
Weber [1921] 1972), we include two dummy
variables for parents' denomination: one for
Catholics and one for non-Christians. The
reference category is Protestant (together
with a few atheists and nondenominational
Christians). We also test for interactions in-
volving Catholicism, family background,
and
national context because Catholics differ
from Protestants in the balance of the indi-
vidual and institutional orientations in their
religions (Weber [1921] 1972).
Because religious belief typically in-
creases with age (Benson et al. 1989:157-
59), and nations differ in their age structures,
we include respondent's age (in years).
Men typically are less devout than women
(Benson et al. 1989:159-60), and nations dif-
fer somewhat in their sex composition-so
we include sex (coded 1 for men and 0 for
women).
Well-educated individuals are generally
less devout than are the poorly educated
(Benson et al. 1989:166-67), and education
is correlated
both with parents' church atten-
dance and GNP. We therefore include educa-
tion measured in years of schooling follow-
ing ISSP definitions (Zentralarchiv
1993).
METHODS AND MODELS
In its simplest form, the relation between
parents' churchgoing and the strength of
respondent's religious belief in different na-

648 AMERICAN SOCIOLOGICAL REVIEW
PARENTS'CHURCH
ATTENDANCE
NATIONAL. Secular Devout
RELIGIOSITY (1) (2) (3) (4) (5)
Religious () wa5+-w) a, wa5+(1 -w) a2 wa5+(1-w) a3 W
a5+ 1
0-w)a4
(3) wa3+(1-w)a1 wa3+(1-w)a2 wa3+(1-w)a4. wa3+(1-w)a5
...:s,",,,,,,',<,.,.,.s.,,:,,
~
~ ~ ~ ~ ~ ~ ~ ~ ~
~
.... ......
(2) wa(
3+(1 + (1-w)wa3 w
wa2 + (I-w) a4 wa2 + (1-w) a5
L
Secular (1) 1'AEESE'#0 wa1 + (1E wa2 I wa1 + (l-w) a3 w
wa2 + (l-w) a4 wa2 + (1-w) a5
Figure 1. Design of the Diagonal Reference Model: Combinations of Parents' Church Attendance and
National Religiosity
Note: Shaded cells represent the average religiosity of each consistent ideal-typical cell. The order of the
rows is reversed (5 to 1 rather than the usual 1 to 5) to correspond with Figures 2 through 4.
tions can be represented in a three-dimen-
sional table, with nations in the rows (ranked
from secular to religious), family back-
ground in the columns (ranked
from secular
to devout), and each cell displaying the aver-
age religious belief score of respondents liv-
ing in that religious a nation and coming
from that devout a family (for the moment
we ignore control variables like age and
GNP). For clarity, we make the table square,
with five categories of national religiosity
and five categories of parents' church atten-
dance. The ideal-typical religious settings are
on the main diagonal: The religious extreme
is represented by people living in devout na-
tions and coming from devout families; the
secular extreme is represented
by people liv-
ing in secular nations and coming from secu-
lar families. Between these extremes are
other ideal-typical settings-people living in
nations of intermediate levels of religiosity
and coming from families with intermediate
levels of devoutness. In these diagonal cells,
the religiosity of the nation is congruent with
the devoutness of the family. We model the
religiosity of cells with an inconsistent con-
text-cells in which the devoutness of the
nation is not consistent with the devoutness
of the family-as a function of cells with
consistent contexts.
Diagonal reference models (Hendrickx et
al. 1993; Sobel 1981) are designed to model
situations of this sort in a clear and parsimo-
nious manner. These models have been ap-
plied to a variety of similar inconsistency and
mobility problems, such as social mobility
and fertility (Sobel 1985), educational mobil-
ity and apostasy (Van der Slik, De Graaf, and
Peters 1995), social mobility and politics (De
Graaf, Nieuwbeerta, and Heath 1995; De
Graaf
and Ultee 1990), and the effects of het-
erogamy (De Graaf and Heath 1992; Van
Berkel and De Graaf 1995). Although diago-
nal reference models originally were de-
signed to model the impact of individual
variables, they can be used to model the im-
pact of macro variables like national context,
giving unbiased point estimates of contextual
effects of national characteristics.5
The design can be visualized as shown in
Figure 1, where w is the weight for the na-
tional devoutness and 1-w is the weight for
5 The disadvantage is that the standard errors
will be slightly
low. Multi-level methods
(Gold-
stein 1987)
would allow us to estimate
similar but
much less parsimonious linear models while giv-
ing asymptotically correct
standard
errors. The
diagonal reference model
applied
to this problem
produces
a single standard error that applies to
both the individual's
family background
and the
macro context
of the nation,
and the two levels
cannot
be separated
statistically.
We believe that
the conceptual advantages
and statistical
parsi-
mony
of the diagonal
reference models
outweigh
this small
statistical
disadvantage.

NATIONS, PARENTS, AND RELIGIOUS BELIEF 649
parents' church
attendance: Here al to a5 (the
shaded cells) represent the average religios-
ity of each consistent ideal-typical cell. The
implication for the religiosity of a person in
an inconsistent cell, for example someone in
a religious nation who has secular parents
(upper left-hand corner), is a weighted aver-
age of the religiosity of two consistent
cells-those living in a religious nation and
having devout parents (a5), and those living
in a secular nation and having secular par-
ents (a,).
Formally, the diagonal reference model
can be written:
= wai + (1-w)aj + Ejk;
where i =1,2,3,4,5; j = 1,2,3,4,5;
k =1,. ..
., N;
and 0<w<1. (1)
In equation 1, Yijk
represents the dependent
variable, the intensity of religious belief of
respondent k; i represents the religiosity of
the nation; and j represents parents' church
attendance.
There is one parameter
a for each
diagonal cell, which represents the expected
mean degree of religious belief of respon-
dents within consistent combinations of pa-
rental church attendance and national religi-
osity. Parameters w and (1-w) are weights
that indicate the relative importance of pa-
rental church attendance and national reli-
gious context respectively on respondent's
religious beliefs. Eijk
is a stochastic error term
with an expected value of 0. One advantage
of the model is that it makes no assumption
about the linearity of the association.
A diagonal reference model with co-
variates in addition to the main variables can
be written as:
Yijk
= wai+ (1-w)a + ,
3mX +E (2)
For each covariate Xm
we estimate a beta pa-
rameter Pm. There are, therefore, m co-
variates and m beta parameters. This equa-
tion describes Model A (with covariates for
age, gender, education, Catholic, non-Chris-
tian, previous Communist society, GNP, and
a dummy for the United States), which ex-
plains 35 percent of the variance. Table 4
presents the goodness-of-fit statistics for
Model A and the other models.
Model B relaxes the restriction that the
weight parameters be the same for each
level of national religiosity and gives a more
flexible model:
w = p+ dl(row2)
+d2(row3)+ d3(row4)+ d5(row5). (3)
Substituting
equation 3 into equation 2 gives
the equation for Model B with five weight
coefficients, one for each religiosity level: p
is the reference parameter for a secular na-
tion (row 1); p + d, is the weight for row 2; p
+ d2 is the weight for row 3; and so forth.
Weights for parental
church attendance, 1-w,
decline conversely. This model, which ex-
plains 36 percent of the variance, leads to a
substantial improvement in fit compared to
baseline Model A (x2 =193.3, d.f. = 4). The
BIC criterion
(Raftery 1986) also suggests an
improvement
in fit.
Finally, because the weight coefficients for
rows 2 and 3, and for rows 4 and 5 are al-
most equal, we put an extra restriction on the
weights to create Model C:
w = p + dI
(row2,3) + d2(row4,5)* (4)
In Model C, the nation weight is p for people
living in a secular nation (row 1). The weight
increases by d1 for people living in a nation
of intermediate religiosity (rows 2 and 3),
while the parental
weight (1-w) decreases by
dj. The same logic holds for d2 for religious
nations (rows 4 and 5). Model C fits almost
as well as Model B, but it is more parsimoni-
ous. The BIC criterion also suggests that it is
preferable to Model B.
Alternative Models
The relative weights of national religiosity
and of parental church attendance may also
vary between Catholics and Protestants (We-
ber [1921] 1972), between old and young,
between men and women, between the well-
educated and the poorly educated, between
formerly Communist nations and other na-
tions, or between rich nations and poor na-
tions. To test these possibilities, in Model D
we allow a dichotomous variable (e.g.,
Catholic background) to interact with the
weight parameters:
W = p + dl (roW2,3) + d2 (row4,5)
+d3(parent Catholic). (5)

650 AMERICAN SOCIOLOGICAL REVIEW
Table 4. Goodness-of-Fit Statistics for Diagonal Reference Models Predicting Respondent's Religios-
ity: 15 Nations, 1991
Degrees Difference in Difference in
of Chi-Square BIC from
Model Description Freedom from Model A Model A
A Baseline diagonal
reference model 14
with covariates (Equation 2)
B Model A + separate weights for each 18 193.3 -154.2
level of national religiosity (Equation
3) (4)
C a Model A + separate weights for inter- 16 193.8 -174.3
mediate
nations
(rows 2 and 3) and
devout (2)
nations (rows 4 and 5) (Equation 4)
D Model C + separate weight for those 17 196.6 -167.3
with Catholic
parents (Equation 5) (3)
Model C + separate weight for age 17 192.8 -163.5
(3)
Model C + separate weight for males 17 194.0 -164.7
(3)
Model C + separate weight for education 17 196.3 -167.0
(3)
Model C + separate weight for formerly 17 194.5 -165.2
Communist
societies (3)
Model C + separate weight for GNP 17 205.1 -175.8
(3)
Note: Numbers
in parentheses
are degrees of freedom for the chi-square
calculation. Number of cases is
17,635. Model C explains 36 percent
of the variance.
For Model A, BIC = -7572.1.
a Model C is the preferred
model.
For each of our main independent variables
in turn, we substitute equation 5 into equa-
tion 2.
Some of these interactions show small but
statistically significant improvements in fit.
The BIC criterion, however, suggests that
only one of the improvements might justify
the increased complexity of these models.
We therefore prefer the more parsimonious
Model C.6
RESULTS
Description
Table 5 and Figure 2 describe the joint effect
of the nation's religious environment
and
par-
6 The version of Model D that includes a sepa-
rate weight for GNP has a slightly lower BIC.
However with only 15 nations there is necessarily
some uncertainty, especially as GNP is highly cor-
related with the dummy variables for the United
States (r = .52) and communism (r = -.79), and its
effects vary depending on whether or not the
United States is included in the analysis.
ents' church attendance on respondent's reli-
gious belief, without adjusting for other indi-
vidual or contextual variables. (1) Clearly,
parental religiosity matters greatly: People
with devout parents are far more likely to ac-
quire orthodox beliefs than are those with
secular parents. (2) People born into religious
societies are much more likely to hold ortho-
dox religious beliefs than are those born into
secular societies. For example, people with
secular parents living in secular nations
(lower left-hand corner
of Figure 2 and Table
5) score only 16 points on religious ortho-
doxy, while those with equally secular par-
ents residing in religious nations are far more
orthodox, averaging
60 points (upper left cor-
ner). Similarly, respondents from devout
families living in secular nations (lower right
corner) score 73 on religious belief while
those born to equally devout families in reli-
gious nations score 85 points (upper right
corner). (3) Even in secular nations, devout
families are largely successful in inculcating
orthodox
beliefs in their children (lower right
corner). Hence, the difference between those

NATIONS, PARENTS, AND RELIGIOUS BELIEF 651
Table 5. Observed Mean Religiosity Score by Religiosity of Nation and Parents' Church Attendance:
15 Nations, 1991
Parents' Church Attendance
National Secular Devout
Religiosity (1) (2) (3) (4) (5) Total
Religious (5) 60 67 72 81 $ 81
.... .. ... ..
{ A X KA AQ 7 < .. . .............:Q g .. 5 :: 5 QX 7Q~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. . . . .
.... . . . . . . .
(4) 64 68 75 80 ....... 85 78
(3) 36 46 69 76 62
(2) 314 56 6
4 72 55
Secular (1) 16 40 53 68 73 43
Total 32 47 59 70 79 61
Number
of cases 2,495 2,926 4,059 4,065 4,770 18,315
from devout families and those from secular
families is large (lower right corner versus
lower left corner: 73 - 16 = 57 points). The
pattern
is similar in nations with intermediate
levels of religiosity. (4) In religious nations,
even individuals born into secular families
are likely to acquire relatively orthodox be-
liefs (upper left corner). Hence in religious
nations, the difference between those from
devout and those from secular families is not
so large (upper
right corner versus upper
left
corner: 85 - 60 = 25 points).
Hypothesis 1: National Context
A clear test of our hypotheses is provided by
the multivariate results which adjust for dif-
ferences among nations in modernization,
the experience of Communist antireligious
policies, American exceptionalism; and for
differences among individuals in denomina-
tion, age, sex, and education (see Tables 4
and 6, and Figures 3 and 4).
After adjusting for all these differences,
parents still strongly influence their
offspring's religious beliefs. It is also clear
that those living in religious societies are
much more likely to acquire religious beliefs
than are those living in secular societies-
even if they are from equally devout fami-
lies, live in nations at the same level of mod-
ernization, are the same age, sex, and de-
nomination, and have the same level of edu-
cation. In all, the religious environment of
100- /
z 80c/ _
o
/ r
~~~~~~~~~~~~~~~~~~~~................g 1
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~._..
. ........., ..........
@ Dn c / ~~~~~~~~~~~~~~. . . . . . . . . . . . . . . . . . . . . . . . . . ... . :|l
z vv / ~~~~~~~~~~~~~~~.-.-...........
- . .-..E i .........
0 ~ ~ / g g | ,'..,.,.,...'.,.,'...'
~..:...
.... ...... ......
@
40 L t g W
~~~~~~~~~~~~. .. ........ ..
....
. ............... ... .....
X g , ~~~~~~~~~~~~~~~. ...... ..... .... ,.....
..
O
/ 0 ~~~~~~~~~~~~~~~. . . . . . . . . . . . . . . . . . . .
z ,~~~~~~~~~~~~~~~~.
.. , , . . .. , .. . .. -..
O 6 z i ,~~~~~~~~~. .. . .. . ... . ...... ecla
Secular~ ~ ~~~~~....... ... vout
.
Parents'~~~~~~~~~~.......
Churc .............
Figure~~~~~~~~~~6 ........ ....ve ReigostySor.b.Ntonl .elgostyan.
Prnt'.huc.Atedac

AMERICAN SOCTOTOGTCAT REVIEW
100/-
~~~~~~~~~~~~~~~~~~~~~~.m/Z .-. ...'..,':
....
F0~~ ' '(..-i .'.'..'.-""'"'''" ":";-:""""""""""".... ...''';;'""";;
as , ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~. . ... . . . . . . . .
.
. ...... -...
X 4 . ::.i.--l :: -- . ......~~~~~~~~~~~.............
i - : : :
:i.i.g.-:::.
o : : | -: -: -I0 En;- ~~~~~~~~~~~~~~~...........
Reliiou
z i...'.-'.
' ....
"""".:
:."'- --
'I "''i";""""
"-"
E': '' ' "'
00
.... .
......
"'' . ...
0 a R t S Secular~~~~~~~~~~~~........
.
~~~~~~~~~~~~~~. .
.eu a ....u
Parents' Church Attendance~~~..........
..
Figure 3 Prediced Religosity Sore by Ntional eligiosty and Prents' hurch Atendance
Model.
Note: For women from Protestant families, of average age and education, living in a non-Communist
...
nation of average GNP, not the United States.~~~~~~~~~~~~~~~~~.........
..
the nation matters
about half as much as par-
ents' church attendance.
Hypothesis 2: Interaction Between Nation
and Family
Model A, which is portrayed in Figure 3,
captures two of the most striking features of
the data, but it clearly does not do full jus-
tice to the data (compare Figures 2 and 3,
particularly the left-hand corners). Model B
reveals a strong, statistically significant in-
teraction
between parents' church attendance
and national religiosity (2 = 193.3, d.f. = 4,
p < .001). Model C, portrayed in Figure 4,
has a simpler description of the interaction
with an equally good fit, and so, on the
grounds of parsimony, it is our preferred
model (Table 6 presents the parameter esti-
mates for Model C; Table 7 presents the re-
sulting weight parameters).
The religiosity of the nation matters more
100 ..........
..............
..........
..
..............
.......... ..
..............
.............
..............
.............
Z
/ ..............
.......... ...
8 0 ...... ... ...
..............
0 ..... .....
... ...
...............
......
. .... .............
............
... ............
...
.
.. ..............
.............
.... .............
...
... ............
....
.... ....... ...
.-
..........
. . . . . . . . . . . . . . . . .
.
. .
. .
. . . . . . . . . . . . . . . . .
... ....
........ ....
...
.. ...
.........
..
XX . ............
...
.............
.... .............
... ............
....... .....
.... .... ....
...............
............
6 0 . .. ...... ..... ..
............. ...
............
. ........ ....
... .............
...
.
. .
. . . . . . . . . . . . . .
.
. .
..... ..... ...
........
. ... ......... .. ....
. . . . . . . . . . . . .
. . . . . . . . . . . . . . - . . . . . .
.
. . . . . . .
. .
.. .........
... .....
....... ...
.............
....
... ...
............ .... ..........
. .. .............
...
........... ...
...... . ...
.. ... .. .....
... . . . . . . .......
. . . . . . . . . .
. . . . . . .
. . . . .
. . . . . . .
. .
. . .
. . . . . . . . . . . .
.
.
. .
. .
. . . . . . . . . . . . . .
. . . . .
.............
.... ........ .........
.........
.... ........
:X ... .............
... .............
...
....... .. . . .... .......
..... ... ............ ....
. . . . ...... . . . . . . . . . .
. .
.
. . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . . .
.......... ... ............ ....
. . .
.
. .
. . . . . . . . . .
. . . . . . , , .
. . . . .
. . . . . . . . . . .
....
...... .
. .... ... ..................
....
........
.... ..........
4 0 .......
..... ..
............ . ... ..
..............
.... ...................
... .............
.
............. . ............
.............
.... ............
. .
. .
.
.
. . . . . . . .
. . . .
. . . . . . . . . . . . . . .
.
.
............
....
.............
. .
.
. . . . . . . . .
. .
. . . . . . . . . . .
........................
............
.. ..........
.. ...
. . . . .
. . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
.
. . .
............
... .... ........
... .... ............
....
..............
. .
. . . . . . . . . . . . . . . . . .
. . .
.
. . . . . . . . . .
. .
.. .. .............................
.......
....... .........
... ... ............ Religious
C L .............
.... ..
. . .... ... ..
.............
.... ... .. ..
............. ..... ..............
.
.
. . . . . . . . .
. . . . . . . . . . . .
.
.
.
. . .
X
. .
.
. . . .
. . . . .
. .. ....
.......:.....
... ........ ... .............
... .... .......
. . . . .
.
.
.
.
. . . .
. . .
. .
.
. . . . . . . . . . . . . .
. . . . . . . .
. . . .
'A 2 0 .... .... ... . ........
.... ..... ......
. . . . . .
. . . . . . .
. . .
. . .
.
. . . . . . . . . . .
. .
.
. . . . . . .
CD ............ . .....
... ... ...... N
a
tio n a l
. ........... .
......
............
...
.............
... ...
.. ..... . . ..............
.............
....
............. ...
. ........ .. ........... . .... ..............
. . . . .
.
.
. . . .
.... . . . . . . . . . . . . . . . . . . . . . .
.
.
. . . . . .
.
.
.
.
. .
. . . . . . .
. . . . . .
. . . .
.
.
. . . . . . . . . . .
.
.
. . . . . . . . .
.
. . . . . . . . . . . . . . . . . . . . .
. . . .
. . . . . . .
. . . . . .
. . . . . .
.
.
.
. .
............
. . . . .
.
.
. . . . . .
.
.
. . . . . . . . . . .
. . . . . . .
. . .
. . . . . . . . . . .
.
.....
............
... .............
.... .............
. ........
... . R
e lig io s ity
. . .
. . . . . . . . . . . . . . .
.
.
. . . . . . . . . .
.
.
. . . .
. . .
.
.
. . . . . .
. . .
. . . . .
.
.
. . . . .
. . .
.
.
. . . . . . . . . . . . . . . . .
.
.
. . . . . . . . . . . . . .
. . . . . . .
. .
. . .
. .
. . . .
. . . . . . . . . . .
. . . .
.
. . . . . .
. . . . . . . . . . .
. . . . .
. .
. . . .
. . . .
.
. . .
. . . . . .
. . .
.
. . . . . . .
. . . . . .
............
.... ...... .... S e c
u la r
0 Secular Devout
Parents' Church Attendance
Figure 4. Predicted Religiosity Score by National Religiosity and Parents' Church Attendance: Model C
Note: For women from Protestant families, of average age and education, living in a non-Communist
nation of average GNP, not the United States.

NATIONS, PARENTS, AND RELIGIOUS BELIEF 653
in some circumstances than in others (see
Table 7). (1) In relatively secular nations, na-
tional context matters little or not at all (w =
.06), while family background matters a lot
(1-w = .94, p < .001). (2) In societies of in-
termediate devoutness, national context mat-
ters more (w = .26, p < .001), and family
background
matters correspondingly
less (1-
w = .74, p < .001). (3) And in devout societ-
ies, national context matters more (w = .58,
p < .001) than family background
(1-w = .42,
p < .001), although family background re-
mains important.
These differences are sub-
stantively large and statistically significant,
and they offer strong support
for Hypotheses
2a and 2b.
The same pattern
holds for both Catholics
and non-Catholics, old and young, men and
women, the well educated and the poorly
educated, formerly communist nations and
other nations, and rich nations and poorer
nations (results available from the authors).7
OLS analyses for each nation separately
show the same pattern (Appendix Table A).
Other Aspects of Nation
Modernization, as measured by the nation's
GNP, slightly reduces religious orthodoxy, in
conformity with modernization
theory (Table
6). Compared to people living in a poor
country like Poland (GNP = 21 percent of
the U.S. GNP), those in a prosperous Euro-
pean nation like West Germany (GNP = 76
7 Our comparison
of the effect of parents'
church
attendance
on religious
orthodoxy
in secu-
lar and religious
societies could be biased
by ceil-
ing effects. Religiosity
is close to the maximum
score in devout countries,
biasing
it downward,
while there
is little or no downward bias in secu-
lar countries. However comparisons
in which
ceiling effects are
unlikely-say the three
or four
most secular rows and columns of Table 5-
shows the same pattern
as the whole table. For
example,
when
comparing
the most secular fami-
lies with those one category
up, religiosity in-
creases
from 16 to 40 (24 points)
in secular na-
tions but only from 64 to 68 (4 points)
in rela-
tively devout
nations.
Applying
Model C to the 4
x 4 table
omitting
the most
religious
nations
and
the most
devout
parents
gives a more
rigorous
test
which
produces
the same
pattern
as was obtained
for the full sample (details
available
on request).
We conclude that
ceiling
effects do not apprecia-
bly bias our
results.
Table 6. Nonlinear Regression Estimates Pre-
dicting Religious Belief: Model C
Parameter
Parameter Estimate
Weights
w: Religiosity of nation .06
(.03)
1-w: Parents' church
attendance .94**
(.03)
Interaction Terms
for Weights
d2,3: Intermediate
religious nation .20**
(rows 2 and 3) (.03)
d4,5: Religious nation
(rows 4 and 5) .52**
(.04)
Diagonal Reference Cells
a,: Secular parents,
secular
nation 37.4**
(1.00)
a2 54.8**
(.63)
a3 67.4**
(.70)
a4 81.7**
(.79)
a5: Devout parents, religious nation 89.2**
(.76)
Individual Context Variables
b
1: Age (years)
a .17**
(.01)
b2: Male (1 = yes) -7.6**
(.39)
b3:
Education
(years)a -1.2**
(.07)
b4: Catholic
parents (1 = yes) -1.3**
(.53)
b5:
Non-Christian
parents (1 = yes) 5.1**
(1.34)
National Context Variables
b6: Communist
nation
(1 = yes) -19.7**
(1.29)
b7:
GNP per capita -.26**
(.03)
b8: United
States (1 = yes) 18.8**
(1.55)
Number
of cases 17,635
Note: Numbers
in parentheses
are standard errors.
a Scored as deviations from their means.
** < .01 (two-tailed tests)

654 AMERICAN SOCIOLOGICAL REVIEW
Table 7. Weight Parameters Implied by Model C
National Parental
National Religiosity Religiosity
Religiosity (w) (1 - w)
Secular .06 .94
Intermediate .26 .74
Devout .58 .42
percent of the U.S. GNP) would, all else
equal, hold beliefs 14 points less orthodox
(-.26 x (76-21) = -14). However, this effect
(unlike others in the model) is not entirely
robust: A plausible alternative model yields
different results (available from the authors).
Specifically, if the United States (which is
both rich and devout) is not exceptional
(contrary to the assumption of Model C),
then an increase in GNP increases religious
orthodoxy slightly. So no strong claim is
warranted.
Communism's persistent campaign against
religion seems to have shaped people's be-
liefs. People living in formerly Communist
nations acquired religious beliefs 20 points
less orthodox than otherwise similar people,
a large and statistically significant difference
(p < .001).
As many have argued, the United States
seems to be exceptional. Americans hold be-
liefs that are, on average, 19 points more or-
thodox than otherwise similar people in other
countries, a large and statistically significant
difference (p < .001).
Individual Effects
Denominational differences matter slightly.
People from Catholic families acquire
slightly less orthodox beliefs, other things
being equal (1 point out of 100, p < .05;
Table 6).
Other individual effects are small but sta-
tistically significant and resemble those
found in previous studies (Table 6). (1) Or-
thodox beliefs increase slightly with age.
Other things equal, a 70-year-old is likely to
be 9 points more orthodox than a 20-year-
old. If this is a cohort effect rather than a life-
cycle effect, it implies a gradual
decline over
time in national religiosity. (2) Men are 8
points less orthodox than women, other
things being equal. (3) The well-educated are
a little less orthodox than the poorly edu-
cated: The difference in beliefs between a
university graduate and someone with only a
secondary school education is about 5 points,
other things being equal.
CONCLUSION
The religious environment of a nation has a
major impact on the beliefs of its citizens:
People living in religious nations acquire, in
proportion to the orthodoxy of their fellow
citizens, more orthodox beliefs than those
living in secular nations. This is not because
they come from more devout families (al-
though most do), nor because religious na-
tions differ from secular nations in modern-
ization or exposure to Communism
(although
they do), nor because of differences in an
individual's denomination,
education, age, or
sex (although such differences exist). Rather,
the religious character of the nation itself
matters. In some circumstances, national
context is more important even than family
background in shaping people's beliefs. A
nation's culture and the policies of its
churches and government are part of the ex-
planation. But especially important, we ar-
gue, is the pool of potential friends, teach-
ers, work colleagues, and marriage-partners:
In a religious nation this pool is mostly de-
vout, and in a secular nation this pool is
mostly secular-and these differences mat-
ter greatly. Our findings about the impact of
a nation's religious environment are consis-
tent with previous research about differences
between urban and rural areas, and between
secular regions and devout regions within a
single nation, but such differences have not
previously been systematically tested on data
from many nations. Our empirical results
from a powerful multivariate model using
comparable data from 15 nations and over
17,000 individuals offer strong evidence for
such contextual effects.8
8The logic of our
argument
extends
beyond
re-
ligion, suggesting
that national context
shapes
in-
dividuals' views on any attitude,
value,
or belief
on which nations
differ, and on which friends,
colleagues or marriage partners
influence each
other.
Thus,
we predict
that national context will
influence attitudes
to, among
other
things, poli-
tics (e.g., views on unions, government
owner-
ship of industry,
welfare
provision,
and
inequal-

NATIONS, PARENTS, AND RELIGIOUS BELIEF 655
Our results also clearly demonstrate
that a
nation's religious environment shapes the
way in which religious beliefs are passed on
from parent to child. In relatively secular na-
tions, the religious views of secular families
are reinforced. However, devout families in
these societies are usually able to insulate
their children from secular pressures. Hence
in relatively secular nations, the effect of
family religiosity on children's religiosity is
strong, and the effect of national religious
context is small. By contrast,
in relatively re-
ligious societies, devout families' views are
reinforced. But secular families in such soci-
eties generally do not insulate their children
from religious pressures, so many acquire the
beliefs of their fellow citizens. Hence, in
relatively religious nations, family religios-
ity is less important for children's religious
views than it is in secular societies.
This strong interaction between a nation's
religious environment and the way in which
family background influences religious be-
lief has important consequences for future
research. Studies on religious socialization
(e.g., Hoge, Petrillo, and Smith 1982) usu-
ally do not take national context into ac-
count. Our results clearly show that studies
that neglect the religious context of the na-
tion can be misleading with respect to the
importance of parental socialization. In de-
vout societies like the United States, Poland,
or Ireland, family religiosity effects will not
be as strong. But in secular societies like
much of Western Europe and most of East-
ern Europe, the effect of family religiosity
can be expected to be strong. If our argu-
ments are correct, these apparent
differences
do not reflect differences in how families
function in these societies, but instead reflect
differences in the religious environments of
the nations themselves.
Our results also speak to the long-running
debate about U.S. religious exceptionalism
(Warner 1993): They support the view that
the United States is unusually religious. Both
fundamentalist beliefs in a personal god (the
potential "demand" for religion) and church
attendance are markedly higher in the United
States.
ity), social issues (e.g., feminism,
divorce,
sexu-
ality, or abortion),
and social groups
(e.g., preju-
dice against
racial or ethnic
groups).
Our results do not support Stark and
Iannaccone's (1994) "supply-side" analysis
of differences between nations which argues
that nations with religious monopolies have
substantial unmet religious needs, while
churches in religiously competitive nations
like the United States do a better job of meet-
ing diverse religious needs. This argument
implies that those who do not attend church
are more devout in monopolistic societies.
But our results suggest the opposite: Those
who do not attend church are actually more
devout in religiously competitive societies
like the United States and Northern
Ireland.
Modernization theory and related argu-
ments by many (mainly European) sociolo-
gists of religion predict that religious belief
declines as nations become more modern.
They contend that this has happened
in many
European nations in recent decades. But
many other (mainly American) scholars dis-
agree, citing the high levels of religious be-
lief in the United States. Our results suggest
that the answer to this argument
turns on the
issue of U.S. exceptionalism. If the United
States is taken to be exceptional, as in our
preferred
model, our data suggest that mod-
ernization (as measured by GNP per capita)
leads to a modest but statistically significant
decline in religious belief in the 14 non-U.S.
nations in our analysis. But if the United
States is not treated as exceptional, this one
case-very high in both GNP and religious
belief-tips the scales in the opposite direc-
tion. In this case, modernization appears to
produce a slight increase in religious belief.
But the effect of modernization is not strong
in either model.
Our results suggest that religious beliefs
endure in large part because the religious en-
vironment of a nation shapes the beliefs of
its citizens. Most residents of European na-
tions in the eighteenth and nineteenth centu-
ries-and the European immigrants who
populated the New World-believed in a su-
pernatural,
at least vaguely anthropomorphic
God who was personally concerned with in-
dividual humans. They agreed on these fun-
damental
points, while often differing on de-
nominational loyalty and specific points of
doctrine. Devout parents raised mostly de-
vout children, helped by the prevailing reli-
gious atmosphere of the nation. Crucially,
even when parental socialization failed, the

656 AMERICAN SOCIOLOGICAL REVIEW
religious atmosphere of the nation and the
devout beliefs of the overwhelming majority
of potential friends, teachers, colleagues, and
marriage partners inculcated belief.
Not until external forces like moderniza-
tion, the Enlightenment, the growth of edu-
cation, or the rise of science bring the aver-
age levels of belief in the nation down ap-
preciably do things begin to change. Our re-
sults suggest that religiosity then changes
rapidly-not declining slowly and gradually
but dropping precipitously. The offspring of
devout families mostly remain devout, but
the offspring of more secular families now
strongly tend to be secular. A self-reinforc-
ing spiral of secularization then sets in, shift-
ing the nation's average religiosity ever fur-
ther away from orthodoxy. So after genera-
tions of stability, religious belief declines
abruptly in the course of a few generations
to the modest levels seen in many Western
European nations today.
Jonathan Kelley
is Director
of the International
Survey Program at the Institute of Advanced
Studies,
The Australian National
University.
He
is also Professorial Associate at the Melbourne
Institute of Applied Economic and Social Re-
search, University of Melbourne, and is principal
investigator of the International Survey of Eco-
nomic Attitudes (ISEA) and of the International
Social Science Survey. With M. D. R. Evans he is
studying the ideology of inequality in 30 nations
and continuing a participant-observation study of
twins. With A. 0. Haller he is studying inequality
in Brazil, and with Krzysztof Zagorski he is in-
vestigating attitudes toward inequality and the
economy during the transition from communism
in Eastern Europe.
Nan Dirk De Graaf is a member of the Depart-
ment of Sociology, Nijmegen University, The
Netherlands. He is also an associate member of
Nuffield College, Oxford, England. His primary
research interests are social stratification and the
political and cultural consequences of social mo-
bility and mixed marriages. He has been super-
vising various projects supported by the Nether-
lands Organization for Scientific Research. He is
currently studying the decline of religious- and
class-based voting, the decline of religiosity,
post-industrialism and class, new consumer styles
and post-materialism, nonvoting, gender inequal-
ity, and the impact of the field of education on
inequality.
Appendix Table A. Ordinary Least Squares Regression Models Predicting Respondent's Religiosity: 15 Na-
tions, 1991
Standardized Metric
Coefficient Coefficienta
Parents' Parents'
Church Church Educa- Number
Nation Attendance Attendance Age Male tion Catholic Constant R2 of Cases
Total
Standardized .49 .09 -.13 -.04 -.03 .27 19,712
Metric 11 .18 -8 -.5 -2 27
East Germany .53 13 .19 -8 0 -1 0 .35 1,485
Slovenia .52 13 -.15 -7 -3 -2 7 .36 2,042
Netherlands .58 13 .22 -1 -1 -12 16 .33 1,633
Hungary .45 11 .39 -10 -2 1 18 .39 985
Norway .40 13 .33 -12 -1 18 30 .29 1,497
Great Britain .31 7 .28 -10 -2 4 41 .21 1,244
West Germany .50 12 .08 -7 0 -5 25 .25 1,346
New Zealand .30 7 .21 -10 -1 8 44 .17 1,067
Australia .32 7 .12 -9 -1 7 43 .16 2,200
Austria .39 9 .13 -8 0 -3 37 .18 980
Italy .28 6 .09 -9 -1 23 32 .18 983
(Appendix Table A continued on next page)

NATIONS, PARENTS, AND RELIGIOUS BELIEF 657
(Appendix Table A continued
from previous page)
Standardized Metric
Coefficient Coefficienta
Parents' Parents'
Church Church Educa- Number
Nation Attendance Attendance Age Male tion Catholic Constant R2 of Cases
Poland .33 8 -.01 -5 -1 19 29 .21 1,062
Ireland .16 5 .22 -7 -1 4 57 .14 1,005
United States .28 5 .09 -8 -1 -3 70 .16 1,350
Northern Ireland .25 4 .23 -7 -1 2 67 .17 833
Note: Nations are listed in order of respondent's religiosity, from most secular (East Germany) to most reli-
gious (Northern Ireland).
In a pooled analysis controlling for characteristics of nations, parents' church attendance has a standardized
effect of .40, and the effect for religiosity of the nation is .24; the effect for Communist nation = -.1 1, GNP = .06,
age = .09, male = -.12, education = -.08, Catholic = -.04, with R2 = .35. Including a multiplicative interaction
between religiosity of nation and parents' church attendance as in Model C gives b = -1.48 (t = -11.6, p < .001).
a Except for row 1, which presents standardized
coefficients.
f Estimate not significant at p < .05 (two-tailed tests). All other estimates are significant.
REFERENCES
Acock, Alan, C. and Vern L. Bengtson. 1978.
"On the Relative Influence of Mothers and Fa-
thers: A Covariance Analysis of Political and
Religious Socialization." Journal of Marriage
and the Family 519-30.
Becker, Jos W. and R. Vink. 1994. Secularisatie
in Nederland, 1966-1991 (Secularization
in the
Netherlands). Rijswijk, the Netherlands:
Sociaal en Cultureel Planbureau.
Benson, Peter, Michael J. Donahue, and Joseph
A. Erickson. 1989. "Adolescence and Reli-
gion." Research in the Social Scientific of Re-
ligion 1:153-81.
Berger, Peter L. 1969. The Sacred Canopy: Ele-
ments of a Sociological Theory of Religion.
Garden City, NJ: Doubleday.
Blalock, Hubert M. 1984. "Contextual-Effects
Models." Annual Review of Sociology 10:353-
72.
Boyd, Lawrence H. and Gudmund R. Iversen.
1979. Contextual Analysis. Belmont, CA:
Wadsworth.
Butler, David and Donald Stokes. 1974. Political
Change in Britain. London, England:
MacMillan.
Calvi, Gabriele. 1992. Indagine Sociale Inter-
nationale 1990/1991 (International
Social Sur-
vey). Milan, Italy: Eurisko.
Cichomski, Bogdan. 1992. Poglady Polakow
1991 (ISSP 1991: Religion). Warsaw, Poland:
Institute for Social Studies, University of War-
saw.
Davis, James A. and Tom W. Smith. 1991. Gen-
eral Social Survey 1991. Chicago, IL: National
Opinion Research Center.
De Graaf, Nan Dirk and Anthony F. Heath. 1992.
"Husbands' and Wives' Voting Behaviour in
Britain: Class-Dependent Mutual Influence of
Spouses" Acta Sociologica 35:311-22.
De Graaf, Nan Dirk, Paul Nieuwbeerta, and An-
thony F. Heath. 1995. "Class Mobility and Po-
litical Preferences: Individual and Contextual
Effects." American Journal of Sociology 100:
997-1027.
De Graaf, Nan Dirk and Wout Ultee. 1990. "Indi-
vidual Preferences, Social Mobility and Elec-
toral Outcomes." Electoral Studies 9:109-32.
Evans, M. D. R., Jonathan Kelley, and Tamas
Kolosi. 1992. "Images of Class: Public Percep-
tions in Hungary and Australia." American So-
ciological Review 57:461-82.
Farkas,
George. 1974. "Specifications, Residuals,
and Contextual Effects." Sociological Methods
and Research 2:333-63.
Felling, Albert, Jan Peters, Osmund Schreuder.
1991. Dutch Religion. Nijmegen, the Nether-
lands: Instituut voor Toegepaste Sociale
Wetenschappen.
Firebaugh, Glenn and Brian Harley. 1991.
"Trends in U.S. Church Attendance" Journal
for the Scientific Study of Religion 30:487-500.
Francis Leslie J. and Laurence B. Brown. 1991.
"The Influence of Home, Church and School
on Prayer among Sixteen-Year-Old Adoles-
cents in England." Review of Religious Re-
search 33:112-22.
Gendall, Philip. 1992. New Zealanders' Attitudes
to Religion-ISSP: 1991 Survey. Palmerston
North, New Zealand: Massey University.
Goldstein, Harvey. 1987. Multilevel Models in

658 AMERICAN SOCIOLOGICAL REVIEW
Educational and Social Research. London, En-
gland: Griffin.
Greeley, Andrew M. 1992. "Religion in Britain,
Ireland and the USA." Pp. 51-70 in British So-
cial Attitudes: The 9th Report, edited by R.
Jowell, L. Brook, G. Prior, and B. Taylor. Lon-
don, England: Dartmouth.
Greeley, Andrew. M. and Galen L. Gockel. 1971.
"The Religious Effects of Parochial Educa-
tion." Pp. 264-301 in Research on Religious
Development, edited by M. P. Strommen. New
York: Hawthorn
Books.
Haller, Max and Franz Hoellinger. 1993. Austrian
Social Survey 1993 Survey. Graz, Austria:
Institut fuer Soziologie der Universitaet Graz.
Hendrickx, Jan, Nan Dirk De Graaf, J. Lammers,
and Wout Ultee. 1993. "Models for Status In-
consistency and Mobility." Quality and Quan-
tity 27:335-52.
Hoge, Dean R. and Gregory H. Petrillo. 1978.
"Determinants
of Church Participation
and At-
titudes among High School Youth."
Journal
for
the Scientific Study of Religion 17:359-79.
Hoge, Dean R., Gregory H. Petrillo, and Ella I.
Smith. 1982. "Transmission of Religious and
Social Values from Parents to Teenage Chil-
dren." Journal of Marriage and the Family
44:569-80.
Jennings, M. Kent, Klaus Allerbeck, and Leopold
Rosenmayr. 1979. "Generations
and Families."
Pp. 449-86 in Political Action, edited by S. N.
Barnes and M. Kasse. Beverly Hills, CA: Sage.
Jennings, M. Kent and G. B. Markus. 1984. "Par-
tisan Orientations over the Long Haul: Results
from the Three-Wave Political Socialization
Panel Study." American Political Science Re-
view 78:1000-18.
Jowell, Roger, Lindsay Brook, Gillian Prior, and
Bridget Taylor. 1992. British Social Attitudes
1991 Survey. London, England: Social and
Community Planning Research.
Kelley, Jonathan
and M. D. R. Evans. 1993. "The
Legitimation of Inequality: Norms on Occupa-
tional Earnings in Nine Nations." American
Journal of Sociology 99:75-125.
. 1995. "Class and Class Conflict in Six
Western Nations." American Sociological Re-
view 60:157-78.
Kelley, Jonathan, M. D. R. Evans, and Clive
Bean. 1993. National Social Science Survey,
1992. Canberra,
Autralia: International
Survey
Program, Research School of Social Sciences,
the Australian National University.
Kelley, Jonathan and Ian McAllister. 1985. "So-
cial Context and Electoral Behaviour in Brit-
ain." American Journal of Political Science
29:564-86.
Kolosi, Tamas. 1992. ReligiQn-The Hungarian
Survey. Budapest, Hungary:
TARKI.
Mohler, Peter Ph. and Michael Braun. 1992.
AllBus: ISSP 1991. Mannheim, Germany:
Zentrum fuer Umfragen, Methoden, und
Analysen.
Myers, Scott M. 1996. "An Interactive Model of
Religiosity Inheritance: The Importance of
Family Context." American Sociological Re-
view 61:858-66.
Need, Ariana and Nan Dirk De Graaf. 1996.
"'Losing My Religion': A Dynamic Approach
of Leaving the Church
in the Netherlands."
Eu-
ropean Sociological Review 12:87-99.
Norwegian Social Science Data Services (NSD).
1992. Attitudes to Religion-ISSP: 1991 Sur-
vey. Bergen, Norway: Norwegian Social Sci-
ence Data Services.
Peters, Jan 1993. Individualisering en Secular-
isering in Nederland in de Jaren Tachtig (Indi-
vidualization and Secularization in the Nether-
lands in the 1980s). Nijmegen, the Netherlands:
Instituut voor Toegepaste Sociale Weten-
schappen.
Peters, Jan and Osmund Schreuder. 1987.
Katholiek en Protestant. Een Historisch en
Contemporain Onderzoek Naar Confessionele
Culturen (Catholic and Protestant: A Histori-
cal Investigation of Confessional Cultures).
Nijmegen, the Netherlands: Instituut voor
Toegepaste Sociale Wetenschappen.
Poloma, Margaret M. and Brian F. Pendelton.
1990. "Religion and Religious Well-Being."
Social Indicators Research 22:255-76.
Przeworski, Adam. 1974 "Contextual Effects of
Political Behavior." Political Methodology
1:27-61.
Raftery, Adrian. 1986. "Choosing Models for
Cross-Classifications." American Sociological
Review 51:145-46.
Scheuch, Erwin K. and Rolf Uher. 1993. ISSP
1991: Religion. Koeln, Germany: Universitat
zu Koeln, Zentralarchiv fuer Empirische
Sozialforschung.
Social and Cultural Planning Office. 1992. ISSP
1991. Rijswijk, the Netherlands: Sociaal en
Cultureel Planbureau.
Sobel, Michael. 1981. "Diagonal Mobility Mod-
els." American Sociological Review 46:893-
906. .1985. "Social Mobility and Fertility Re-
visited" American Sociological Review 50:
699-712.
Spika, Bernard,
Ralph W. Hood, Jr., and Richard
J. Gorsuch. 1985. The Psychology of Religion.
Englewood Cliffs, NJ: Prentice Hall.
Stark, Rodney and Laurence R. Iannaccone. 1994.
"A Supply-Side Reinterpretation
of the 'Secu-
larization' of Europe."
Journal of the Scientific
Study of Religion 33:230-52.
Stark, Rodney and James C. McCann. 1993.
"Market Forces and Catholic Commitment."
Journal for the Scientific Study of Religion

NATIONS, PARENTS, AND RELIGIOUS BELIEF 659
32:111-24.
Tos, Niko and Janez Stebe. 1992. Slovensko
Javno Mnenje (Sloveniann Social Survey).
Ljubljana, Slovenia: University of Ljubljana,
Public Opinion and Mass Communication Re-
search Centre.
Van Berkel, Michel and Nan Dirk De Graaf.
1995. "Husband's and Wife's Culture Partici-
pation and Their Levels of Education." Acta
Sociologica 38:131-49.
Van der Slik, Frans. 1994. "Measuring Christian
Beliefs." Journal of Empirical Theology 7: 5-
34.
Van der Slik, Frans, Nan Dirk De Graaf, and Jan
Peters. 1995. "Kerkverlating in Nederland."
Sociale Wetenschappen
38:23-50.
Van Dijk, Jan J. M., Pat Mayhew, and Martin
Killias. 1990. Experiences of Crime
Across the
World.
Deventer, the Netherlands: Kluwer Low
and Taxation.
Ward, Conor and Brendan Whelan. 1992. Reli-
gion and Role of Government Modules. Dublin,
Ireland: Economic and Social Research Insti-
tute.
Warner,
R. Stephen. 1993. "Work in Progress to-
ward a New Paradigm for the Sociological
Study of Religion in the United States." Ameri-
can Journal of Sociology 98:1044-93.
Weber, Max. [1921] 1972. Wirtschaft und
Gesellschaft (Economy and Society).
Tilbingen, Germany: Mohr.
World Bank. 1996. World Development Report
1996: From Plan to Market. Oxford, England:
Oxford University Press (for the World Bank).
Wuthnow, Robert. 1994. Sharing the Journey:
Support Groups and America's New Quest for
Community.
New York: Free Press.
Zentralarchiv. 1993. International Social Survey
Programme: ISSP 1991, Religion. Koeln, Ger-
many: Universitat zu Koeln: Zentralarchiv fuer
Empirische Sozialforschung (Bachemer Strasse
40, D-5000 Koeln 41, Germany).