4-6 page

 Subject: Essay #1 

In 1986, J. Glass argued that a parent’s political orientation is the strongest determining factor in predicting a child’s future political preferences (i.e., most (or many) Republicans grew up in Republican-leaning households and most (or many) Democrats grew up in Democrat-leaning households).  See Attitude similarity in three generational families: Socialization, status inheritance, or reciprocal influence? American Sociological Review , Vol 51, 685-698 (1986).  You can access the article using the hyperlink below.  Please read it as part of your response to the essay questions:

1.   In your opinion (and experience) did Glass correctly identify the primary source of our collective “political socialization?” 

2.   What other sources or factors influence our political preferences?

3.   Glass’s article was published in 1986.  In your opinion, is political socialization today largely similar to political socialization in 1986?  If so, how is it similar?  If not, how is it different?  What implications do your observations have for political candidates, organizations, and/or institutions?

4.   What aspects of your reading in our course text help you to understand political socialization differently than you did prior to beginning our class?

Your essay must be between 4-6 pages in length, be double-spaced with standard (1-inch) margins and 12-point Times New Roman font.  Your essay will be graded on a scale of 0-100.  

Glass.ASR.1986

Attitude Similarity in Three-Generation Families: Socialization, Status Inheritance, or
Reciprocal Influence?
Author(s): Jennifer Glass, Vern L. Bengtson and Charlotte Chorn Dunham
Source: American Sociological Review, Vol. 51, No. 5 (Oct., 1986), pp. 685-698
Published by: American Sociological Association
Stable URL: http://www.jstor.org/stable/2095493

.

Accessed: 23/08/2013 17:46

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. 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

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=asa

http://www.jstor.org/stable/2095493?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN THREE-GENERATION FAMILIES:
SOCIALIZATION, STATUS INHERITANCE,

OR RECIPROCAL INFLUENCE?*

JENNIFER GLASS VERN L. BENGTSON
University of Notre Dame CHARLOTTE CHORN DUNHAM

University of Southern California

This study examines hypotheses of attitude transmission across three ideological domains
(gender roles, politics, religion) to access the adequacy of direct socialization, status
inheritance, and reciprocal influence models in a developmental aging perspective. Data
are from 2,044 individuals, members of three generation families, grouped to form
parent-youth (G2-G3) and grandparent-parent (GJ-G2) dyads. Results suggest, first, that
there is little convergence of parent-child attitudes with age when viewed cross-sectionally.
Second, status inheritance processes do account for a substantial amount of observed
parent-child similarity, but parental attitudes continue to significantly predict childrens’
orientations after childhood. Third, child influences on parental attitudes are relatively
strong and stable across age groups, while parental influence decreases with age, although
the exact pattern of influence varies by attitude domain.

For many years, social theorists have considered
the role of the family in maintaining continuity in
social ideologies over time (Engels [1884], 1967;
Adorno et al., 1950; Parsons and Bales, 1955;
Thomas and Znaniecki, 1958; Chodorow, 1978).
The resulting view of the family as conservative
(for example, slowing the pace of social change)
and monolithic (influencing individual beliefs in a
forceful and consistent manner) perhaps reached its
ultimate expression in the attempts of some
revolutionary movements (such as in Cambodia or
China) to break up generational ties in order to
foster rapid social change. And, in fact, contempo-
rary research on the intergenerational transmission
of attitudes has shown that parents’ attitudes,
especially mothers’ attitudes, are significant posi-
tive predictors of children’s attitudes in adulthood
(Acock and Bengtson, 1978; Bengtson, 1975;
Dalton, 1980; Jennings and Niemi, 1982; Smith,
1983).

However, this typification of the family as
conservative and monolithic in its influence on
ideological orientations has come under increasing
scrutiny among family scholars, as they point out
the diversity of influences on children and the
complexity of family relationships. In this study,
we focus on two basic empirical questions: 1) How

much actual similarity in social ideologies is found
between American parents and children at different
points across the life cycle? 2) What are the forces
generating that similarity over the life course?

To answer these questions, we first describe the
traditional view of attitude transmission derived
from childhood socialization theory. Then we
explore conceptual criticisms of the socialization
paradigm from alternative theoretical perspectives.
Finally, we empirically examine the dynamics of
attitude transmission using responses to specific
attitudinal scales from a sample of three-generation
families.

SOCIALIZATION THEORY AND
DEVELOPMENTAL AGING

Traditional conceptions of socialization have viewed
the family, speficially parents, as the principal
agent of socialization in childhood (Freud, 1933;
Erickson, 1950; Heilbrun, 1965). One of the
functions of the family is seen as the provision of
stability and continuity to individual members.
Families are thought to provide systematic social-
ization through which children are taught the
norms of the social order. Attitude similarity
between generations, from this view, is the
consequence of successful parental socialization of
beliefs and values. Children learn their parents’
values, beliefs, and attitudes through both direct
teaching and indirect observation, as part of the
information and guidance that children either
actively seek out (in the Piagetian sense) or
passively accept (through social conditioning) in
maneuvering their way through life.

While childhood socialization theories do not
directly address the issue of parent-child similarity

*Address all correspondence to Jennifer Glass, Depart-
ment of Sociology, University of Notre Dame, Notre
Dame, IN 46565.

This study, based on data from the USC Longitudinal
Study of Three-Generation Families, was supported by
grants from the National Institute of Aging (#AG-04092)
and the National Institute of Mental Health (#MH-38244).
Special thanks go to Rich Williams, Richard Miller and
Donna Polisar for their assistance in developing this
paper, as well as to three anonymous reviewers for their
helpful comments and criticisms.

American Sociological Review, 1986, Vol. 51 (October:685-698) 685

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

686 AMERICAN SOCIOLOGICAL REVIEW

in adulthood,’ the implicit assumption of tradi-
tional conceptualizations has been that childhood
socialization is so intense, prolonged, and
psychodynamically important that the attitudes and
values formed in the family context persist well
into adulthood (Chodorow, 1978; Campbell, 1969;
Adorno et al., 1950). Thus, it might be expected
that parents and children would continue to exhibit
attitude similarity across the life course, and into
later adulthood-though perhaps diminishing with
time as the intensity of parent-child contact
diminishes.

This traditional approach to socialization has
been challenged by scholars in recent years for
failing to address two important issues. First, from
a macro-structural point of view, parent-child
attitude similarity may be viewed more as the
result of social forces that generate the inheritance
of social status than as the product of individual
psycho-social influence. One of the central issues
in the interpretation of findings of parent-child
attitude similarity is whether such similarity can be
attributed to successful parental socialization, per
se, or whether it has more to do with successful
intergeneration transmission of class, race, reli-
gious affiliation, marital status, and other promi-
nent social statuses that structure life experience
and mold social attitudes (Acock, 1984). What
parents transmit may be social statuses, more than
attitudes and values. In this way, similarities in
social structural position may create attitudinal
similarities between parents and adult children
through a common-cause association (i.e., parents
and children have undergone similar attitude-
shaping experiences).

The second conceptual challenge to an uncriti-
cally traditional perspective of socialization is the
possibility that similarity in attitudes between
parents and children could equally be due to the
influence of children’s attitudes on those of their
parents, especially as children age. The traditional
perspective on socialization focuses on young
children and adults and ignores the possibility of
variability across the life course by the age and
developmental stages of the parents and children at
each point in life (Hagestad, 1981; Featherman,
1983). Proponents of an interactionist perspective
(Bell and Harper, 1977; Lerner and Spanier, 1978;
Bengtson and Troll, 1978; Hagestad, 1984) argue
that children increasingly influence their parents
with age, and that attempts to model intergenerational
influence as a one-way process-flowing from
parents to children-may be fundamentally errone-
ous, since reciprocal effects occur.

Only a few studies, however, have emprically
tested the reverse influence process with respect to

social attitudes. Hagestad (1984) noted that about
two-thirds of parents, and one-third of children,
reported “successful” influence by children in her
three-generation sample. Angress (1975) found
that mothers of radical college students changed
their attitudes about cohabitation based on their
children’s behavior. Chaffee et al. (1971) reported
that adolescents influenced their parents’ television
behavior. Curiously, studies of reciprocal influ-
ence began in the literature on infant development
(Lewis and Rosenblum, 1974), despite the fact that
older children are presumably much more capable
of altering parents’ stated beliefs or behavior.

Very little is known about intergenerational
attitude similarity across the lifespan or the forces
generating similarity past childhood (Bengtson, et
al., 1985). Some theorists emphasize the impor-
tance of parent-child bonds across all stages of the
life cycle (Shanas, 1979; Troll et al., 1979),
implying that substantial intergenerational similar-
ity exists across the life course. Others emphasize
flexibility and change in parent-child relations at
different stages of the life course.

Theories of developmental aging (Bengtson and
Kuypers, 1971; Hess and Waring, 1978; Moss and
Abramowitz, 1982; Baltes, 1979; Rossi, 1980)
suggest that parents and children have different
investments in family relationships and different
sources of power in family interaction as they
move through the life course. For example,
children in late adolescence may share few of the
adult statuses that their parents hold and may be
facing the developmental tasks of independence
and differentiation from parents2 (Erickson, 1950).
Such processes would suggest relatively larger
absolute discrepancies between parents’ and adult
children’s attitudes. This position implies that
social status similarity3 should account for rela-

1 This closely parallels Manheim’s (1952) notion of
youths’ “fresh contact” with the social order (see
discussion in Bengtson et al., 1985).

2 We refer throughout this paper to “social statuses”
occupied by parents and children. Our use of this term is
broad and inclusive, intending to cover family and
community statuses as well as general socio-economic
status.

3 Although the developmental aging paradigm has
proved to be a useful tool in the investigation of family
relationships over time, it needs to be amended with a
theory of social change. Elder (1974) has demonstrated
the importance of looking at the impact of social and
historical changes on family functioning. Rapid social
changes (revolutions, economic recessions, technological
advances) may encourage adult children to increasingly
look to non-parental sources of information and support.
Improvements in health status and economic well-being
may decrease the dependence of elderly parents on their
children in the future. These historical changes may limit
the impact of internal family dynamics in attitude
formation. Unfortunately for analytic purposes, the data
on three-generation families utilized in this research
report are cross-sectional rather than longitudinal. As
such, we have confounded historical (or cohort) effects
and developmental/life stage effects. Therefore, our

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 687

tively less of the relationship between parents’ and
childrens’ attitudes at this point in the life cycle,
since young adults have not yet attained many of
the social statuses that inform their parents’ beliefs
(few have married or become parents, many have
not finished their education and have had minimal
opportunities for occupational achievement, etc.).
With respect to the reciprocal influence between
young adults and their parents, there is reason to
believe that parental influence may still be quite
strong at this stage, relative to child influence.
From a social exchange perspective, the ability of
parents and children to influence each other should
be determined by the relative resources and
rewards that each bring to the interaction. At this
stage, parents have only recently relinquished their
authority over their children, and young adult
children may still rely on their parents for material
support and guidance. Moreover, young adults
have little of the experience or resources that
would enable them to influence their parents’
attitudes.

Turning now to middle-aged children and their
elderly parents, a developmental aging perspective
suggests that a different set of dynamics may
characterize their relationship. As youth age, they
are more likely to attain social structural position
similar to their parents with respect to marital
status, income property ownership, etc. By
mid-life, children have achieved a variety of adult
social statuses. The life experiences generated by
these adult roles are likely to replace direct parental
influence in the modification of social attitudes.
However, this similarity of adult social roles
between the generations may lead to smaller mean
differences in attitudes between them. In other
words, youth may gradually come to hold views
more similar to their parents’ as they have children
of their own, buy property, and obtain full-time
employment, although their parents’ influence
does not directly cause them to alter their beliefs.
The respective developmental stages of middle-
aged children and elderly parents suggest further
that the pattern of influence between parents and
children may change over time. Middle-aged
children are in many ways at the height of their
social power in industrialized western societies
(Riley, et al., 1982). Aged parents in later life, on
the other hand, may become more dependent on
their middle-aged children for advice and informa-
tion than before, reflecting both physical decline
and a loss in social power. In this context,
middle-aged parents may not view their elderly
parents as appropriate social referents. These

factors suggest that adult children may increasingly
influence elderly parents over time, while elderly
parents’ influence on their adult childrens’ attitudes
may have declined since mid-life.4

In summary, it can be said that traditional views
on socialization have focused upon the process of
influence from parent to child without adequately
considering the impact of inherited social status,
the possibility of mutual influence, and variability
across the life course due to developmental aging.
More recent literature suggests that attitude
transmission may indeed be mutual and grounded
in social and historical milieu (Elder, 1984). The
degree of similarity and difference between parents
and children will be affected by the dynamics of
mutual influence and developmental change. Re-
ciprocal influence will also be played out within a
broader structural and historical context.

Keeping in mind the importance of both
developmental change and mutual influence, we
have selected three attitudinal domains for study-
religious ideology, political ideology, and gender
ideology. Three ideological areas, rather than one,
were selected both to test the generalizability of the
developmental aging perspective outlined here to a
variety of social attitudes, and to avoid heavy
reliance on one particular content area in address-
ing broad conceptual issues in attitude transmis-
sion. Although variability of results across atti-
tudes scales is discussed, it is not the primary focus
of this paper. Empirical research on religious,
political, and gender ideology has tended to show
both moderate intergenerational transmission and
cohort effects of varying sizes (Bengtson and
Troll, 1978). Prior research on religious ideology
has demonstrated moderate to high parent-child
agreement (Acock, 1984; Weiting, 1975; Hoge et
al., 1982) and small but significant age cohort
difference in religious ideology (Hyman, 1959).
Although the literature on political socialization is
too extensive to review here, studies generally
indicate that transmission is moderate for political
ideology (Jennings and Niemi, 1968, 1982).
Finally, recent studies of gender role attitudes
(Mason et al., 1976; Thornton et al., 1983) show
that while attitudes across age cohorts have
liberlized in the past 20 years, younger age cohorts
have changed their opinions to a much greater
extent than older age cohorts. Smith and Self
(1980) report only minor transmission of gender

ability to directly assess the impact of rapid social change
on attitude similarity over the life course is limited.
However, follow-up data on these families 12 years later
are currently being collected and will be used to confirm
or amend the findings discussed here.

4 Few prominent theorists have made firm statements
about the strength or endurance of dispositions socialized
in childhood (Goslin, 1969). Most socialization theorists
view socialization as a process begun in infancy and
ending in death. However, the term “resocialization” is
often used to refer to specific attempts to alter the content
of earlier socialization. The presumption seems to be that
socialized outcomes are relatively stable unless and until
specific attempts at resocialization are made in adult-
hood.

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

688 AMERICAN SOCIOLOGICAL REVIEW

ideology from mothers to daughters in a college
sample. Overall, these studies suggest significant
variability in the impact of recent social changes on
attitudes and attitude transmission in the three
areas (social change most pronounced in the area
of gender ideology, followed by political ideology
and religious ideology).

In the analyses to follow, we focus on three
major hypotheses. The first hypotheses is that
elderly parents and their middle-aged children
show smaller mean differences across all three
domains of attitudes than middle-aged parents and
their young adult children. In other words,
attitudes of parents and children converge with
age. Both developmental aging and status inherit-
ance explanations suggest smaller attitude differ-
ences in older generation dyads, although tradi-
tional socialization theory suggests larger attitude
differences with age. Developmental theories
emphasize the rebelliousness of youth as they
attempt to separate and establish independence
from their parents. As children move into adult
roles and establish their independence, their need
to differentiate themselves from their parents
decreases. Traditional socialization theories, how-
ever, emphasize parent-child contact and parental
control of resources as the forces generating
parent-child attitude similarity. To disentangle
these effects requires further regression analysis.

Our second hypotheses, therefore, is that
parents’ attitudes continue to significantly predict
children’s attitudes, after controlling for children’s
current social status. However, parental influence
(controlling for social status) should be weaker for
elderly parents than for middle-aged parents,
reflecting the diminishing intensity of parent-child
interaction. Conversely, the developmental trajec-
tory of status inheritance suggests that status
effects should increase with age, as children take
on adult roles similar to those their parents held in
adulthood.

Finally, the possibility of reciprocal influence
must be addressed, to insure that the causal
direction of parental effects are correctly specified,
The developmental aging perspective suggests that
as children become older, they are better able to
influence their parents. Therefore, our third
hypothesis is that child influences on parental
attitudes increase with age, while parental influ-
ences on children’s attitudes decline with age.

Three basic analyses are performed: 1) compar-
ison of absolute differences between generation
pairs on three attitudinal scales (political ideology,
gender ideology, religious ideology); 2) regression
analysis of parental attitude score on adult child’s
attitude score, with and without demographic
indicators of child’s social status; and 3) structural
equation modeling of both parent and child attitude
scores across successive generations. The first
analysis addresses the question of whether parent-
child similarity declines in successive generations

within families. The second analysis attempts to
disentangle the effects of adult children’s social
structural position (which may be quite similar to
their parents’) from the effects of parental attitudes
per se on adult children’s attitudes. The third
analysis addresses the question of reciprocal effects
in attitude transmission, controlling for the effects
of social structural variables on individual re-
sponses. This analysis compares the magnitude and
significance of parent-child versus child-parent
influence, and compares patterns of influence
within dyads across generational positions in the
family.

METHODS

Sample Selection

The data for this analysis are based on responses
for 2,044 individuals drawn from a broader study
of three generations conducted in Southern Califor-
nia in 1973. The sample was drawn from a
population of 840,000 members of a Los Angeles
area health care plan (described in Bengtson,
1975). To be eligible for inclusion in the original
sampling frame, members had to have been males
over 65 with at least one grandchild between 16-26
years of age. Sample construction proceeded by
sending questionnaires to all eligible grandchildren
between 16-26, their parents, and related grandpar-
ents in the original sampling frame. 77 percent of
the grandparents (Gi) had only one biological
child respond; another 20 percent had two children
respond. Among the parents (G2), 48 percent had
only one child respond, 35 percent had two, and 13
percent had three or more. Although not a random
sample, this sample does represent a wide group of
individuals from various ethnic, economic and
social backgrounds. The sample is generally
representative of white, economically stable,
middle- and working-class families. This sample
has an advantage over some other studies of
three-generation families which have drawn their
sample from among college students and their
parents. The mean age of the grandparent genera-
tion (GI) is 67.1. Mean age for the parent
generation (G2) is 43.8, while mean age for the
adult child generation (G3) is 19.7.

Attitudes and opinions were measured in a
self-administered, mailed questionnare which had
a response rate of 70 percent (N = 2044) over all
3 generations. The questionnaires were mailed in
two waves: a period of six months elapsed between
the mailing of the first and second wave. The
religious and political items were measured on the
first wave. Because the response rate on the second
wave was somewhat smaller, the number of cases
used in the construction and analysis of the gender
ideology scale is somewhat smaller (N = 1585).
Comparison of respondents and non-respondents
on age, education, sex, and income revealed no

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 689

significant attrition bias between the two panel
waves.

Dyad Selection
Attitudes concerning three substantive areas (polit-
ical, religious and gender ideologies) were com-
pared in order to determine the differences between
parents and their children at different stages of the
life-course in these areas. This was accomplished
by the construction of dyad comparisons in which
the summated scores of the children on these items
were subtracted from those of the parent. Because
we are studying three-generation lineages, two
types of dyads exist in this analysis. The first set of
dyads consisted of first and second generation
(G1-G2) parents and children (N = 478); the
second set, the second and third generation
grouping (G2-G3) (N = 1004). We are comparing
the opinions and attitudes of grandparents (Gl)
with the opinions and attitudes of parents (G2), as
well as comparing the opinions and attitudes of
parents (G2) with those of grandchildren (G3).
Because every child in the sample was compared to
each participating parent, some respondents were
entered into the analysis more than once (in the
case of families with more than one child or more
than one parent responding to the survey). A total
of 1482 dyads were constructed for the analysis.
Approximately 45 percent of the dyads contain
either a parent or child record that has appeared as
such elsewhere in the remaining dyads. 30 percent
of the dyads contain both a parent and a child
record that have already appeared in the sample of
dyads. Approximately 9 percent of the dyads
contain parent records that are replicated more than
once in other dyads in the sample. This over-
representation of some dyad members presents the
potential for attenuation of the distribution that
might not occur otherwise. However, a similar
sampling procedure was employed by Acock and
Bengtson (1980) using these same data; they tested
the degree to which such sampling-with-replication
resulted in any increase or decrease in the level of
predictivity and found it did not.

The lack of independence among the sampling
units is a serious issue. However, each parent-child
dyad is an unique unit of analysis that is not
duplicated; even though one member may appear
more than once in the total sample of dyads. Not to
include the duplicated member dyads would also
have the effect of underrepresenting large families
and two parent families, further jeopardizing the
representativeness of the sample. The nature of the
data collection process in this survey precludes
easy elimination of replicated member dyads. Such
an elimination process would effectively halve the
sample.

MEASUREMENT
From the attitude and opinion items within the
mailed questionnaire, three distinct scales, measur-

ing three substantive areas, were used. The items
in these scales have a forced choice, Likert-type
format with four response options that range from
strongly agree to strongly disagree. Factor analysis
was used in order to determine how the scales
would be constructed from these items. One
advantage of this procedure is that scale score
comparisons could be made in order to avoid the
attenuation of correlations that occurs when using
single-item comparisons (Bohrnstedt and Carter,
1971). Once appropriate items for inclusion were
ascertained for each scale, scale scores were
computed by adding to relevant items together and
dividing the result by the total number of items
used. Thus, mean scores are comparable across
scales.

Gender Ideology Scale

Out of 12 possible questions that might have been
included in the scale measuring gender ideology,
five had high factor loadings. These items and their
factor loadings for the entire sample are: “Wives
should obey their husbands” (.62); “Men cannot
respect a fiancee who has had sex” (.45);
“Husbands should have the main say in marriage”
(.60); “Womens’ lib makes sense” (.53); and
“Women should not have authority over men”
(.45). Coefficient alpha was used as a measure of
reliability for each of these scales-for the overall
sample as well as for each generation. The alpha
coefficients for the overall sample, G1, G2, and
G3, are .72, .62, .72, and .75 respectively.
Because each generation represents a unique
position within the lineage, factor coefficients by
generation were examined as well. The factor
loadings followed a similar pattern for each
generation.

Religious Ideology Scale

The religious ideology scale (reflecting conserva-
tive Christian opinions) contains four items with
high factor loadings. For the total sample, the
factor loadings for each item are: “Every child
should have religious instruction” (.66); “God
exists as in the Bible” (.90); “The United States
would be better if religion had more influence”
(.78); and “We are all decendents of Adam and
Eve” (.76). Coefficient alpha for each generation
for this scale is: .85 (Gl), .85 (G2) and .83 (G3).
The total sample reliability is .85.

Political Ideology Scale

The factor loadings for the five items in this scale
are: “The United States should be ready to answer
any challenge to its power, anywhere in the
world” (.45); “Student demonstrators deserve
strongest punishment possible” (.58); “Society’s
most important task is law and order” (.66); “It is

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

690 AMERICAN SOCIOLOGICAL REVIEW

a man’s duty to work; it is sinful to be idle” (.53);
and “Most people on welfare are lazy; they just
won’t do a good day’s work and so cannot get
hired” (.56). The alpha coefficients for GI and G3
are somewhat low (GI = .62, G2 = .72, and
G3 =.66). However, dropping one item from the
scale would have increased unreliability, so the
entire five item scale was retained with total
sample reliability of .67. Unfortunately, both this
scale and the religious ideology scale contain only
items that are worded in a conservative direction,
raising the possibility that response biases exist.

Mean scale scores by generation for each
attitude scale are shown in Appendix 1. Note that
N’s are somewhat smaller for the gender ideology
scale, due to the lower response rate on wave B.
Scores on all three attitude scales increase with
generational position, indicating greater conserva-
tism in older generational cohorts. Differences
between the generations were statistically signifi-
cant across the three scales (F = 47.56 for
political ideology, F = 10.03 for gender, F –
17.37 for religious ideology).

Social Status Variables

Factors that are assumed to influence one’s
opinions and attitudes concerning religion, gender
ideology, and political issues include certain
measures of social status that can be used to
describe an individual’s experiences and interests
within society. It is assumed that the experiences
implied by such variables as marital status and
gender, for example, have an impact upon the
attitudes and opinions that one holds regarding
these issues. In this analysis, eight social status
variables were used to predict differences in
attitude and opinion between parents and children.
These variables are: gender, age, marital status,
number of children, occupational prestige, labor
force participation, educational attainment, and
income. The variables of labor force status and
marital status were dichotomized into working or
non-working and married or non-married. Occupa-
tional prestige was measured with Duncan’s
occupational prestige scores. Educational attain-
ment and income were both measured by ordinal
scales, containing more than seven categories.

RESULTS

Trends in Generational Similarity

Hypotheses 1 predicted larger generational differ-
ences between youth and their parents than
between middle-aged adults and their parents.
Inspection of Table 1 indicated that this was not
true across the three scales. None of the dyad
contrasts were statistically significant, meaning
that there is no evidence from this table to suggest
any convergence in attitudes between adjacent
generations with age. Contrary to our original

Table 1. Mean Absoloute differences between Genera-
tional Dyads in 3-generation Families

Political ID Gender ID Religious ID

Grandfather/ .63 .59 .64
Father .06 .07 .07

(75) (57) (80)
Grandfather/ .64 .76 .69
Mother .03 .07 .05

(167) (114) (161)

Grandmother/ .65 .55 .67
Father .05 .07 .07

(84) (57) (90)
Grandmother/ .68 .78 .61
Mother .04 .05 .05

(152) (123) (144)

Gl-G2 Total .65 .70 .65
.02 .03 .03

(478) (351) (475)

Father/ .66 .72 .78
Son .03 .05 .04

(206) (146) (209)

Father/ .61 .75 .65
Daughter .03 .04 .04

(246) (196) (251)

Mother/ .57 .68 .77
Son .03 .04 .04

(255) (177) (241)

Mother/ .59 .75 .60
Daughter .02 .04 .03

(297) (228) (275)

G2-G3 Total .60 .73 .69
.01 .02 .02

(1004) (747) (976)

Standard errors are reported under each mean.

(N in parentheses)

expectation, attitude differences appear to be of
approximately the same magnitude whether one is
looking at grandparent-parent (G1-G2) dyads or
parent-adult child dyads (G2-G3).

It is important to note that this result may arise
out of an unspecified cohort effect. Older cohorts
may have entered adulthood with smaller parent-
child differences than modem cohorts, but time
since childhood has increased those parent-child
differences until they equal the current parent-child
differences of young adults. However, it is
plausible that today’s young adults will also
increase their parent-child differences as they age
in the future. Longitudinal data, which will be
available shortly from these same families, can be
used to test this hypothesis.

Table 1 reports the results of the first analysis
for all three scales by generation and sex as well.
In general, generational differences are modest in
size, with means less than 1 on a maximum
five-point scale. The largest aggregate differences
of opinion between parents and children were

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 691

found for the gender ideology scale (.70 for G1-G2
dyads; .73 for G2-G3 dyads). No consistent effects
of sex composition on dyadic agreement across
generations were found, supporting Acock and
Bengtson’s (1978) earlier contention that few
sex-specific influences exist in comparing attitude
differences between generations.

Status Inheritance

The evidence from Table 1 demonstrating substan-
tial intergenerational agreement should not neces-
sarily be taken as evidence of strong parental
influence across generational dyads. At this point,
both methodological and substantive problems
impede a straightforward interpretation of mean
attitude agreement measures. Mean attitude agree-
ment measures cannot conceptually address the
issue of whether parents’ attitudes actually influ-
ence their children’s attitudes (Acock, 1984).
Obviously, mean attitude differences across gener-
ations can obscure the variation among families in
the extent of generational agreement. Therefore,
we have refined our initial findings of intergenera-
tional agreement by constructing regression models
in which children’s attitudes are predicted as a
function of both parent’s attitude and a set of
variables describing the child’s own social sta-
tuses. It should be noted that such a regression
model allows for cohort effects between genera-
tions on these attitude scores.

All regression models have been estimated with
LISREL, to incorporate a measurement model for
parents’ and children’s scale scores into the
regression framework. In addition, regression
models for G1-G2 dyads and G2-G3 dyads were
estimated together in one LISREL model using the
option for multiple group data. This framework
insured that the factor structures for the attitude
scales could be made invariant across all genera-
tions, and that statistical tests could be performed
to detect significant differences in regression
coefficients across the two dyad types.

The constrained measurement models for politi-
cal, gender, and religious ideology fit the data
quite well.5 For the measurement model for
political ideology, the overall coefficient of
determination (R2) was .77 for Gl-G2 dyads and
.78 for G2-G3 dyads. For gender, the coefficients
were .86 for GI-G2 and .93 for G2-G3 dyads.
Finally, the measurement model for religious
ideology produced coefficients of .99 for G1-G2

dyads and .97 for G2-G3 dyads. These models
were tested statistically by constructing null
models in which factor structures were not
constrained across generations (Sobel and Bohrnstedt,
1985). The resulting chi-square difference tests
revealed statistically significant generational differ-
ences in factor structure for political ideology (X2
= 400, 12 df), religious ideology (X2 = 53, 9
df), and gender ideology (X2 = 45, 12 df). We
attach substantive importance to the findings for
the political scale only, since the increments to
chi-square for the other two scales are relatively
small. Because of the theoretical importance of
constraining factor loadings across generations
(Thomson and Williams, 1982), the constrained
models were used in all analysis. However, these
constraints are not empirically supported by the
data for the political ideology scale, in particular.

As is shown in the analyses to follow, greater
generational agreement is generally synonymous
with greater parental influence in these dyads, as
measured by the amount of variance in children’s
attitudes explained by parental attitudes. Table 2
presents the results of the regressions of adult
child’s attitudes on parents’ attitudes and adult
child’s social status variables for the two dyad
types, using LISREL to model the measurement of
social ideologies. This insures that unreliability or
measurement error does not attenuate the relation-
ships between child attitudes, child status vari-
ables, and parental attitudes. It is clear from Table
2 that adding social status variables to the null
model of parental influence alone results in a
significant improvement in model fit. The decrease
in chi-square was 266 for political ideology, 257
for gender ideology, and 289 for religious ideology
(all with 28 df).

Hypothesis two predicts that parental attitudes
should continue to significantly predict children’s
attitudes in these dyads. The significant impact of
parental attitudes on adult children’s attitudes does
persist across all three scales, even after controls
for age, marital status, labor force status, educa-
tion, number of children, and family income are
added to the equation. It is clear from these results
that parental influence is not reducible to the
transmission of social status, although the coeffi-
cients for parental attitudes drop with the addition
of social status variables to the equation. Status
transmission can account for some of the attitude
continuity displayed across generations, but there
are definitely family socialization effects that exert
an independent influence on children’s attitudes
past young adulthood.

Hypothesis two also predicts that parental
attitudes will have a stronger impact in G2-G3
dyads than in G1-G2 dyads. Constraining the
parental coefficients to be equal across dyad type
produced an insignificant increase in the overall
chi-square statistic for each model, indicating that

5 The errors of measurement were not correlated by
design in these models. LISREL modification indices
showed only scattered error correlations that, if esti-
mated, might improve model fit. These instances fit no
pattern or a priori theory of measurement error (for
instance, correlating one parent error term with a
different child error term).

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

692 AMERICAN SOCIOLOGICAL REVIEW

Table 2. Regression of Child’s Attitude on Parent’s Attitude and Child Social Status Variables

Political Gender Religious

Variables G1-G2 G2-G3 G1-G2 G2-G3

G1-G2 G2-G3

Parental .56** .50** .53** .49** .44** .27** .37** .32** .51** .44** .59** .55**
Attitude (.54) (.49) (.51) (.48) (.40) (.25) (.34) (.30) (.51) (.44) (.58) (.55)

Age -.01 .02 -.001 -.02 -.01 -.01
(-.08) (.17) (-.004) (-.09) (.04) (-.04)

Labor Force (.03) .06** -.01 .07** -.01 .03*
Status (.10) (.19) (-.02) (.14) (-.02) (.11)

Education -.10** -.12* -.11** -.15** -.01* -.04
(- .24) (- .31) (- .18) (- .24) (- .02) (-.10)

Income -.003 .00 -.01* .001 -.01* .00
(- .07) (.01) (-.16) (- .02) (- .13) (.02)

Number of .02 .09** .08** .16 .05** .10
Children (.04) (.20) (.11) (.23) (.12) (.22)
Marital

Status -.05 -.05* 1.l3* -.05 -.09* -.02
(Married = 1) (-.11) (- .11) (-.17) (-.06) (-.20) (-.04)

Sex -.14** .01 -.17 -.11 .01 .12
(- .15) (.01) (-. .12) (- .07) (.01) (.12)

R 2 .22 .30 .30 .38 .14 .25 .12 .19 .23 .28 .34 .38
Goodness of

fit Index .927 .936 .927 .953 .912 .939
X2/df 1109/332 669/332 710/243

* p c .05.

** p ‘ .01.
(standardized effects in parentheses)

the absolute size of the parental coefficient does
not differ across older and younger dyads.

Turning to Table 3, we separated the unique
contributions of parental attitudes and child social
status variables as a group to the explained
variance in the attitude equations displayed in
Table 2. Substantial differences were found in the
predictive power of parental attitudes across scales
and across generations. Controlling for social
status, grandparents predicted parents’ scores less
well than parents predicted their young adult
children’s scores across all three attitude scales.
The R2 increments for G1 parental attitudes on G2
children’s scores were .15, .04, and .14 for the
political, gender, and religious ideology scales,
respectively. The corresponding figures for G3

children were higher, .24, .08, and .26, respec-
tively. Conversely, social status variables indepen-
dently predicted slightly more of the variance in
G2 (parents’) scores than G3 (adult children’s)
scores, for the political and gender ideology scales.
While not definitive, these results suggest that the
importance of parental attitudes as determinants of
children’s attitudes decreases with age, while the
importance of social structural variables as deter-
minants of attitudes only slightly increases with
age.

Looking at between-scale differences, one can
see that parents’ scores were much more predictive
of children’s scores for the religious and political
ideology scales than for the gender scale. The
predictive power of parents’ scale scores seems to

Table 3. Decomposition of R2 into Unique Contributions of Parental Attitudes and Child’s Social Status Variables

Political Gender Religious
Ideology Ideology Ideology

GJ-G2
Parental Attitude .15 .04 .14
Social Status Variables .09 .11 .05

Total R2 .30 .25 .28

G2-G3
Parental Attitude .24 .08 .26
Social Status Variables .08 .07 .04

Total R2 .38 .19 .38

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 693

complement the degree of parent-child similarity in
scale scores. Parents’ scores predicted at least 14
percent of the variance in their children’s religious
and political attitudes, even after relevant social
status variables were controlled. Yet, parents’
scores explained well under 10 percent of the
variance in their children’s gender ideology scores.
This is not overwhelming evidence of direct family
influence on gender attitudes. In contrast, social
structural variables explain a roughly consistent
proportion of the variance across attitude scales.
For example, among G2-G3 dyads, social struc-
tural variables explain 8 percent, 7 percent, and 7
percent of the variance in G3 responses for
political, gender, and religious ideology respec-
tively. For G1-G2 dyads, structural variables
account for 9 percent, 11 percent, and 5 percent of
the explained variance in scale scores. Clearly,
social structural variables do not replace parental
influence where influence is low.

Reciprocal Influence

Our last analysis addresses the reciprocal nature of
the attitude influence process-both parent and

child influence. A path model describing the model
of the influence process estimated with LISREL is
shown in Figure 1. It is assumed that parents’
social statuses affect children’s attitudes only
indirectly through parents’ attitudes6. These mod-
els were initially estimated with and without
correlated disturbance terms. The simple uncorre-
lated model was more parsimonious in each case,

6 Examination of the normalized residual correlations
indicates that this assumption is accurate for the older
dyads; less so for the younger dyads. For instance,
among younger dyads, there is evidence that parents’
income and education independently affect political
ideology. However, it is plausible that these parent
effects are proxies for unmeasured neighborhood or
peer-group influences on young adult’s political ideol-
ogy. It appears that parent’s marital status may also
independently influence religious ideology, although the
mechanism through which this effect operates is open to
speculation. Finally, parents’ employment status may
independently affect young adults’ gender ideology,
although the effect is weak and, we suspect, confined to
mothers’ employment status.

Figure 1. Model of Reciprocal Influence Between Generations

, _

/*
p C

income income

educ marital C dcmarital p status P educ status
p LFstatus #kids p c LFstatus / kidsC

age se x age sex

PARENT’S ATTITUDE CHILD’S ATTITUDE
( I ) political (I) political
(2) gender (2) gender
(3) religious (3) religious

7K I \\ Xup Uc 7K
pl p2 p3 p,4 p5 cl c2 c3 c4 c5

E1 ?2 E3 E4 5 6l1 E2 83 84 85

STRUCTURAL EQUATIONS: MEASUREMENT MODELS:

Pattz BCatt +rpXp +Up pi APtt+E

Catt BPatt + rcXc + Uc Cj a ACatt + Ej

* all social status variables were allowed to freely correlate with all other social status variables

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

694 AMERICAN SOCIOLOGICAL REVIEW

with corresponding increases in chi-square insignif-
icant.

Table 4 displays the unstandardized estimates
obtained from LISREL models for each attitude
scale. The decrease in chi-squares between the
models in Tables 2 and 4 indicate that the
reciprocal effects models in Table 4 fit the data
better than the parental influence models of Table 2
(X2=7,12,21, respectively, with 2 df). The
pattern of results weakly supports hypothesis 3,
that parent influences decline with age while child
influences increase with age, controlling for social
status. Large differences appear in the pattern of
influence between parents and children across the
three attitude scales and across the life-span. The
results for political ideology best support hypothe-
sis 3. In the younger G2-G3 dyads, parent
influence on children is significant, while child
influence on parents is not. However, in the older
GI-G2 dyads this pattern is reversed-middle-aged
children significantly influence their parents’
political ideology but parental influence is insignif-
icant at this stage. For gender ideology, child
effects on parents’ scale scores appear significant
in both G1-G2 and G2-G3 dyads, while parent
effects are insignificant across dyad type. This
pattern of influence “upward” through the gener-
ations challenges long held notions about the
impact of family socialization on gender attitudes
in adulthood. The last attitude scale, religious
ideology, shows reciprocal influence of parents
and children on each other among younger G2-G3
dyads, while older dyads show only child-to-parent

influence. We see that, once again, parent effects
seem to become less significant with age.

Within LISREL, the chi-square statistic repre-
sents the ability of the estimated parameters to
reproduce the original variance-covariance matrix
of the input data. The degrees of freedom represent
the number of free sample moments (variances and
covariances) unused in the process of estimating
the number of parameters included in each model.
Because the samples used are relatively large, it is
difficult to estimate models which reproduce the
original data well, or provide a “good fit.” None
of our models in Tables 2 or 4 were able to
reproduce the original data with a probability
greater than .05. For each model, we report the
ratio of chi-square to degrees of freedom. For all
scales, the chi-square ratios suggest acceptable fits
to the data (ratios ranging from 1.99 to 3.33). The
LISREL-generated goodness-of-fit index is also
uniformly high for all models (1.00 indicating
perfect fit). In addition, model fit was assessed
using Hoelter’s (1983) critical N method. This
method determines the sample size needed to
reproduce the data with a given model at an
acceptable probability level. If that critical N
exceeds 200 per group (400, in this case), then a
given model fits the data reasonably well. Using
Hoelter’s criteria, all the models produce accept-
able fits to the data (CN = 431, 530, and 601
respectively).

To test the significance of differences in the
influence process between younger and older
dyads, models were estimated in which parent-

Table 4. Models of Reciprocal Influence on Attitudes, Controlling for Own Social Status Variables, by Dyad Type

Political Gender Religious

G1-G2 G2-G3 G1-G2 G2-G3 GI-G2 G2-G3

Parent Child Parent Child Parent Child Parent Child Parent Child Parent Child
Parent’s

Attitude – .21 – .43** – -.15 – .13 -.31 – .34**
Child’s Attitude .27** – .11 – 34** – .21** – .64** – .28** –

Age .01** -.00 -.00 .02 .00 .00 .01* -.01 .01 .00 -.00 -.01
Labor Force

Status – .02 .03 – .00 .06** .02 .00 – .01 .08** .02 .01 .02 .05**
Education -.07** -.11** -.11** -.13** -.08** -.17** -.12** -.17** -.02 -.06* -.04** -.05*
Income -.Ol* -.00 .00 -.00 -.02** -.02** -.01* -.00 -.O1* -.01** -.01** .00
Number of

Children -.01 .02 .03 .08* .05 .07** .11** .15 .01 .07** .05** 11**
Marital Status -.02 -.06 -.01 -.05* .12 -.19** -.12 -.07 .03 -.14** -.15** -.04*
Sex -.03 -.15** -.20** .01 -.03 -.21* -.22** -.10 .07 -.08 .03 .14**

x2 (df) 1102(330) 657(330) 689(241)
x2ratio 3.33 1.99 2.86
Goodness-of
fit index .93 .94 .93 .95 .91 .94
Critical N 431.41 530.02 601.10

*p < .05.

p < .01.

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 695

child effects, child-parent effects, or both were
constrained to be equal across dyad type. Surpris-
ingly few significant differences were uncovered
among any of the attitude scales when constrained
and unconstrained models were compared. For
political ideology, the difference in chi-square
between the fully constrained and unconstrained
models was 1.76, with two degrees of freedom
gained. Thus, no statistically significant difference
exists between the size of the parent effect for
younger and older dyads or between the size of the
child effect for younger and older dyads. In
addition, both the child-parent and parent-child
effects were significantly different from zero in the
newly estimated constrained model. The best-
fitting constrained model for political ideology is
displayed in Table 5.

For gender ideology as well, the most parsimo-
nious model was the fully constrained model in
which both parent-child and child-parent effects
were constrained to have equal effects across dyad
type. However, in this constrained model, only
child effects on parents were significant. Finally,
the best-fitting model for religious ideology was a
partially constrained model in which only child-to-
parent influences were set equal in the two dyad
types. When parent-child effects were also con-
strained, the difference in chi-square between
models rose to 6.28 with only one degree of
freedom gained. So, the difference in elderly and
middle aged parents’ influence on children’s
religious ideology was significant. Only middle
aged parents’ religious attitudes significantly shaped
their children’s religious beliefs; elderly parents
apparently did not exert a strong independent
influence on their children’s religious beliefs.

Looking across best-fitting models in Table 5, it
is hard to avoid the conclusion that significant
child-to-parent influence is more prevalent than
parent-to-child influence. Child-parent effects are
significant and equal across dyad types for all three
scales. Parent-child influences are significant in
younger dyads for political and religious ideologies
only; for older dyads, it is political ideology alone
that shows significant parent-child transmission.

DISCUSSION

This research has examined three issues concerning
the transmission of attitudes across generations.
The first involves the amount of ideological
similarity between parents and children across
life-course positions, as this may reflect increasing
status similarity or the resolution of parent-
adolescent conflict. Second, causal mechanisms
underlying apparent continuity across generations
were explored. We wished to test the possibility
that observed similarities in attitudes are due to
social status similarities and not to socialization.
The third issue concerns the possibility of
reciprocal influence. Intergenerational agreement
can not necessarily be taken as evidence of parental
influence, since observed similarity may be due to
influence of children on their parents. Data
addressing these issues suggest some important
modifications of existing socialization and devel-
opmental aging theory.

The first hypothesis suggested that parent-child
attitudes converge with advancing age; specifically
that G1-G2 dyads would show smaller attitudinal
differences than would G2-G3 dyads. This hypoth-
esis was not supported by the data. Attitude
differences were small throughout the generational
pairs, and the differences observed were the same
in the younger (G2-G3) dyads as in the older
(GI-G2) dyads. Some differences were manifest
across ideological domains, with gender showing
the greatest contrasts. These findings suggest that
an uncritical use of life-course position to predict
varying levels of parent-child difference can lead
to overgeneralization; substantial continuity is seen
across different points in life represented in this
study.

The second hypothesis attempted to disentangle
developmental aging and status inheritance as
sources of attitude similarity. The hypothesis that
parents’ attitudes predict children’s attitudes, after
controlling for children’s social status, was con-
firmed. However, the level of parental prediction
drops with the addition of social status variables,
indicating the importance of status transmission
mechanisms. It should be noted that parents’

Table 5. Final Estimates of Reciprocal Effects’

G1-G2 G2-G3

parent child parent child

Political parent’s attitude – .379** – .379**
ideology child’s attitude .168* – .168*

Gender parent’s attitude – .074 – .074
ideology child’s attitude .237** – .237**

Religious parent’s attitude – .14 -.31**
ideology child’s attitude .31** – .31**

1 Since social status effects were listed in Table 4, they are omitted from this table.

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

696 AMERICAN SOCIOLOGICAL REVIEW

scores are more predictive of children’s scores on
religious and political ideology scales than on the
gender scale. Also confirmed was the crucial
second part of the hypothesis, that the parental
effect would decrease in importance with age; that
is, parental attitudes would have a stronger
influence in G2-G3 dyads than in G1-G2 dyads.
But social status variables do not replace parental
influence in scales where influence is low, nor do
they increase in predictive importance with age, as
measured by generational position.

The principal conclusion of our second analysis,
that parental attitudes exert an influence indepen-
dent of social status inheritance, and that these
effects, though diminished, exist past early adult-
hood, can be taken with the results of the first
hypothesis to suggest the continuing importance of
influence processes across generational positions.
However, observed parent-child similarity need
not reflect parental influence, and indeed the
life-course or developmental aging perspective
points to reciprocal influence as the causal
mechanism. The third analysis estimated LISREL
models of reciprocal influence, testing the hypoth-
esis that child influences on parental attitudes
increase with age, while parental influences on
children’s attitudes decrease with age. The data did
not confirm this hypothesis. Child influences were
significant and equal in magnitude across younger
and older dyads for all three attitude domains.
However, when significant parental influence
existed among younger generational pairs, that
influence did tend to decline among older genera-
tional pairs. Gender ideology, the domain affected
most by rapid social change, showed a pattern of
only upward transmission from children to parents
across the generations.

These results, taken together, suggest that while
the extent of parent-child attitude similarity
appears relatively stable across successive genera-
tions, the forces generating these similarities
appear to change over time. Direct parental
influence declines in older generation dyads, while
social structural variables only slightly increase in
importance as predictors of attitudes. However,
child influences on parents are strong in early
adulthood, and stay strong over the life course.
This implies that social-structural similarities and
child influence produce parent-child similarities
later in the life course, while reciprocal influence
may produce more parent-child similarity in
younger generational dyads.

The findings of this study suggest three
conclusions concerning the family as an agent of
socialization over the life course. First, it is
important to recognize relational change beyond
primary socialization. Evidence of significant
influence upward through the generations suggests
that the family may act as an agent of change, not
an impediment to change as is implied by many
conceptualizations of family socialization.

Second, one must examine the causal mecha-
nisms behind observed continuity or change in
socialization outcomes, especially those structural
or status similarities between parents and children
that are often undifferentiated from “parental
influence.” Status inheritance may be, as sug-
gested by these results, an important alternative
route to inter-generational similarity.

Finally, variability in the impact of parent-child
relations across social ideologies should be acknowl-
edged. While religious and political ideologies
clearly emerge as areas of strong independent
family influence, gender ideology seems less
affected by internal family dynamics. Perhaps
fewer competing agents of socialization exist for
religious or political attitudes, or perhaps these
domains are less profoundly related to daily living
than gender ideology, giving children little reason
to question their parents’ beliefs. It may also be
true that parental influence weakens during periods
of rapid changes in social behavior. Whatever the
source, it is clear that the family is neither a
monolithic nor necessarily conservative source of
influence on attitudes or beliefs past childhood.

Appendix 1. Mean Scale Scores for each ideological
scale by generation

61 62 63 F
Political 3.0 2.7 2.3 47.56*
ideology (484) (661) (779)
Gender 2.6 2.4 2.2 10.03*
ideology (387) (518) (597)

Religious 3.3 3.3 2.7 17.37*
ideology (479) (645) (754)

(N in parentheses)

Appendix 2. LISREL factor loadings for each attitude
scale (constrained to be invariant across generations)

Item # Political Gender Religious
1 1.00 1.00 1.00
2 1.184 .692 2.094
3 .056 .956 1.473
4 1.390 .776 2.070
5 1.310 .965 –

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

ATTITUDE SIMILARITY IN FAMILIES 697

Appendix 3. LISREL generated Error and Disturbance Terms for each latent variable (full models displayed in
Table 4)

Political Gender Religious

P’s C’s P’s C’s P’s C’s
G1-G2 attitude attitude attitude attitude attitude attitude

disturbance (psi) .107 .186 .257 .409 .164 .274
El 1.126 .904 1.070 .679 .275 .407

Errors E2 .613 .696 1.202 .873 .148 .148
of E3 1.033 1.232 .836 .670 .350 .374

measurement E4 .596 .617 .942 .559 .455 .525
Es .853 .602 .784 .614 – –

G2-G3
disturbance (psi) .162 1.28 .337 .472 .132 .147
El 1.004 .828 .745 .631 .339 .569
E2 .633 .871 .989 .659 .237 .417
E3 1.032 1.040 .676 .718 .340 .605
E4 .719 .581 .542 .655 .465 .567
E5 .574 .739 .611 .550 – –

REFERENCES

Acock, Alan C. 1984. “Parents and Their Children: The
Study of Inter-Generation Influence.” Sociology and
Social Research 68:151-71.

Acock, Alan C. and Vern L. Bengtson. 1980. “Social-
ization and Attribution Processes: Actual vs. Perceived
Similarity Among Parents and Youth.” Journal of
Marriage and the Family 40:519-30.

Adorno, Theodor W., et. al. 1950. The Authoritarian
Personality. New York: Harper.

Angress, Selma. 1975. Intergeneration relations and
value congruence between young adults and their
mothers.” Unpublished Ph.D. dissertation. University
of Chicago, Committee on Human Development.

Baltes, Paul B. 1979. “Life Span Developmental
Psychology: Some Converging Observations on His-
tory and Theory. Pp. 256-74 in Life Span Develop-
ment and Behavior, Vol. 2, edited by Paul Baltes and
Orville Brim. New York: Academic Press.

Bell, Robert and L. V. Harper. 1977. Child Effects on
Adults. New York: Wiley.

Bengtson, Vern L. 1975. “Generation and Family
Effects in Value Socialization.” American Sociologi-
cal Review 40:358-71.

Bengtson, Vern L., Neal E. Cutler, David J. Mangen,
Victor W. Marshall. 1985. “Generations, Cohorts,
and Relations Between Age Groups. Pp. 304-38 in
Handbook of Aging and the Social Sciences, edited by
Robert Binstock and Ethel Shanas. New York: Van
Nostrand Reinhold Company.

Bengtson, Vern L. and Joseph A. Kuypers. 1971.
“Generational Difference and the Developmental
Stake.” Aging and Human Development 2:249-60.

Bengtson, Vern L. and Lillian Troll. 1978. “Youth and
Their Parents: Feedback and Intergenerational Influ-
ence in Socialization.” Pp. 215-40 in Child Influences
on Marital and Family Interaction: A Life-Span
Perspective, edited by Richard M. Lerner and Graham
B. Spanier. New York: Academic Press.

Bohrnstedt, George W. and T. Michael Carter. 1971.
“Robustness in Regression Analysis.” Pp. 118-46 in
Sociological Methodology, edited by Herbert L.
Costner. San Francisco: Jossey-Bass.

Campbell, Ernest Q. 1969. “Adolescent Socialization.”
Pp. 821-60 in Handbook of Socialization Theory and
Research, edited by David A. Goslin. Chicago:
Rand-McNally.

Chaffee, Stephen H., Jack M. McLeod and Charles K.
Atkin. 1971. “Parental Influences on Adolescent
Media Use.” American Behavioral Scientist 14:323-40.

Chodorow, Nancy. 1978. The Reproduction of Mother-
ing. Berkeley: University of California Press.

Dalton, Russell J. 1980. “Reassessing Parental Social-
ization: Indicator Unreliability Versus Generational
Transfer. ” American Political Science Review 74:421-31.

Elder, Glen H., Jr. 1974. Children of the Great
Depression. Chicago: University of Chicago Press.

. 1984. “Families, Kin, and the Life Course: A
Sociological Perspective.” Pp. 80-136 in Advances in
Child Development Research: Vol. 7: The Family,
edited by Ross Parke. Chicago: University of Chicago
Press.

Engels, Frederik. 1967. The Origin of the Family,
Private Property, and the State. New York: Interna-
tional Publishers.

Erikson, Erik. 1950. Childhood and Society. New York:
Norton.

Featherman, David L. 1983. “Life-Span Perspectives in
Social Science Research.” Pp. 1-59 in Life-Span
Development and Behavior, vol. 5, edited by Paul B.
Baltes and Orville G. Brim, Jr. New York: Academic
Press.

Freud, Sigmund. 1933. New Introductory Lectures in
Psychoanalysis. New York: Norton.

Goslin, David A. 1969. Handbook of Socialization
Theory and Research. Chicago: Rand McNally.

Hagestad, Gunhild 0. 1981. “Problems and Promises in
the Social Psychology of Intergenerational Relations.”
Pp. 11-46 in Aging: Stability and Change in the
Family, edited by Robert W. Fogel, Elaine Hatfield,
Sara B. Kiesler, and Ethel Shanas. New York:
Academic Press.

. 1984. “The Continuous Bond: A Dynamic
Multigenerational Perspective on Parent-child Rela-
tions between Adults.” Pp. 129-58 in Minnesota
Symposium on Child Psychology, edited by Marion
Perlmutter. New York: Erlbaum.

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

698 AMERICAN SOCIOLOGICAL REVIEW

Heilbrun, Authur B., Jr. 1965. “An Empirical Test of
the Modeling Theory of Sex-role Learning.” Child
Development 36:789-99.

Hess, Beth B. and Joan M. Waring. 1978. “Parent and
Child in Later Life: Rethinking the Relationship.” Pp.
241-74 in Child Influence on Marital and Family
Interaction, edited by Richard M. Lerner and Graham
B. Spanier. New York: Academic Press.

Hill, Reuben, Nelson Foote, Joan Aldous, Robert
Carlson and Robert MacDonald. 1970. Family Devel-
opment in Three Generations. Cambridge, MA:
Schenkman.

Hoelter, Jon W. 1983. “The Analysis of Covariance
Structures: Goodness of Fit Indices.” Sociological
Methods and Research 11:325-44.

Hoge, Dean R., Gregory Petrillo and Ella Smith. 1982.
“Transmission of Religious and Social Values from
Parents to Teenage Children.” Journal of Marriage
and the Family 44:569-80.

Hyman, Herbert. 1959. Political Socialization. New
York: Free Press.

Jennings, M. Kent and Richard Niemi. 1982. Genera-
tions and Politics: A Panel Study of Young Adults and
Their Parents. Princeton, NJ: Princeton University
Press.

. 1968. “The Transmission of Political Values
from Parent to Child.” American Political Science
Review 42:169-84.

Lerner, Richard M. and Graham B. Spanier. 1978. Child
Influence on Marital and Family Interaction. New
York: Academic Press.

Lewis, Michael and Leonard A. Rosenblum. 1974. The
Effect of the Infant on its Caregiver. New York:
Wiley.

Mannheim, Karl. 1952. “The Problem of Generations
(1928).” Pp. 276-320 in Essays on the Sociology of
Knowledge, edited by D. Kecskemeti. London:
Routledge and Kegan Paul.

Mason, Karen O., John L. Czajka and Sara Arber. 1976.
“Change in U.S. Women’s Sex-role Attitudes,
1964-1974.” American Sociological Review 41:573-96.

Moss, Nancy E. and Steven I. Abramowitz. 1982.
“Beyond Deficit-filling and Developmental Stakes:
Cross-disciplinary Perspectives on Parental Heritage.”
Journal of Marriage and the Family 44:357-66.

Parsons, Talcott and Robert Bales. 1955. Family,
Socialization, Interaction Process. Glencoe, IL: Free
Press.

Riley, Matilda W., Beth B. Hess, and Kathleen Bond.
1982. Aging in Society: Selected Reviews of Recent
Research. Hillsdale, NY: Erlbaum.

Rossi, Alice A. 1980. “Aging and Parenthood in the
Middle Years.” Pp. 138-207 in Life-Span Develop-
ment and Behavior, vol. 3, edited by Paul B. Baltes
and Orville G. Brim. New York: Academic Press.

Shanas, Ethel. 1979. “Social Myth and Hypotheses: The
Case of the Family Relations of Old People.” The
Gerontologist 19:3-9.

Smith, M. Dwayne and George D. Self. 1980. “The
Congruence Between Mother and Daughters in Sex
Role Attitude: A Research Note.” Journal of Marriage
and the Family 42:105-109

Smith, Thomas E. 1983. “Parental Influence: A Review
of the Evidence of Influence and Theoretical Model of
the Parental Influence Process. ” Pp. 13-45 in
Research in Sociology of Educational Socialization,
vol. 4, edited by Alan Kerkhoff. Greenwich, CT: JAI
Press.

Sobel, Michael and George Bohrnstedt. 1985. “Use of
Null Models in Evaluating the Fit of Covariance
Structure Models.” Pp. 152-78 in Sociological
Methodology, edited by Nancy Tuma. San Francisco:
Jossey Bass.

Thomas, William I. and Florian Znaniecki. 1958. The
Polish Peasant in Europe and America. New York:
Dover.

Thomson, Elizabeth and Richard Williams. 1982.
“Beyond Wives’ Family Sociology: A Method for
Analyzing Couple Data.” Journal of Marriage and the
Family 44:999-1008.

Thornton, Arland, Duane F. Alwin and Donald Camburn.
1983. “Sex-role Attitudes and Attitude Change.”
American Sociological Review 48:211-27.

Troll, Lillian, Sheila J. Miller, and Robert Atchley.
1979. Families in Later Life. Belmont, CA: Wadsworth.

Weiting, Stephen G. 1975. “An Examination of
Intergenerational Patterns of Religious Belief and
Practice and Practice.” Sociological Analysis 36:
137-49.

This content downloaded from 128.110.224.115 on Fri, 23 Aug 2013 17:46:05 PM
All use subject to JSTOR Terms and Conditions

http://www.jstor.org/page/info/about/policies/terms.jsp

• Article Contents

p.685
p.686
p.687
p.688
p.689
p.690
p.691
p.692
p.693
p.694
p.695
p.696
p.697
p.698

• Issue Table of Contents

American Sociological Review, Vol. 51, No. 5 (Oct., 1986), pp. i-vi+591-742+I-XXVI
Front Matter [pp.i-vi]
E-State Structuralism: A Theoretical Method [pp.591-602]
Expectations, Legitimation, and Dominance Behavior in Task Groups [pp.603-617]
Class Struggle American Style: Unions, Strikes and Wages [pp.618-633]
Worker Attachment and Workplace Authority [pp.634-649]
Labor Market Structure, Intragenerational Mobility, and Discrimination: Black Male Advancement Out of Low-Paying Occupations, 1962-1973 [pp.650-659]
Race, Instruction, and Learning [pp.660-669]
The Settlement Process Among Mexican Migrants to the United States [pp.670-684]
Attitude Similarity in Three-Generation Families: Socialization, Status Inheritance, or Reciprocal Influence? [pp.685-698]
Structure as Process: Organization and Role [pp.699-716]
Using Adjusted Crosstabulations to Interpret Log-Linear Relationships [pp.717-733]
Comments
Marital Coital Frequency: Unnoticed Outliers and Unspecified Interactions Lead to Erroneous Conclusions [pp.734-737]
Is It Outlier Deletion or Is It Sample Truncation? Notes on Science and Sexuality [pp.738-742]
Back Matter [pp.I-XXVI]