Annu. Rev. Sociol. 1998. 24:395–421
Copyright 1998 by Annual Reviews. All rights reserved
INTERMARRIAGE AND
HOMOGAMY: Causes, Patterns,
Trends
Matthijs Kalmijn
Department of Sociology, Utrecht University, Utrecht, The Netherlands;
e-mail: m.kalmijn@fsw.ruu.nl
KEY WORDS: marriage markets, marriage, mate selection, assortative mating, endogamy
ABSTRACT
People have a tendency to marry within their social group or to marry a person
who is close to them in status. Although many characteristics play a role
in the choice of a spouse, sociologists have most often examined endogamy
and homogamy with respect to race/ethnicity, religion, and socioeconomic
status. I first give an overview of hypotheses on the causes of endogamy and
homogamy. The various hypotheses that have been suggested in the literature
can be distinguished as arguments about three more general factors: (a)
the preferences of marriage candidates for certain characteristics in a spouse,
(b) the interference of “third parties” in the selection process, and (c) the constraints
of the marriage market in which candidates are searching for a
spouse. Second, I summarize empirical research by answering four questions:
(a) To what extent are groups endogamous and how do groups differ in
this respect? (b) How has endogamy changed over time? (c) Which factors
are related to endogamy? (d) How do various dimensions of partner choice
coincide? Third, I discuss strengths and weaknesses of past research. Strengths
include the mass of descriptive work that has been done and the development
of a multifaceted theoretical perspective which gives sociological theorizing
an edge over psychological and economic theories of partner choice. Weaknesses
include the lack of standardization of methods in describing patterns
and trends and the relatively weak integration of empirical and theoretical
work.
0360-0572/98/0815-0395$08.00
395
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
INTRODUCTION
Since the beginning of this century, sociologists have described patterns of partner
choice and have tried to explain why people marry within their group (endogamy)
and why people marry persons close in status (homogamy). The research
literature can be divided into three traditions, depending on which type
of characteristic is considered. Research on ethnic and racial intermarriage
originated in immigrant countries such as the United States and is motivated
by the question of whether the various nationality groups would integrate with
one another and with the original population (Drachsler 1920; Wirth & Goldhamer
1944). Research on religious intermarriage has been done both in and
outside the United States and has been concerned with the extent to which
churches control the life choices of their members and the degree to which religious
involvement translates into the membership of “communal groups”
(Kennedy 1944). Research on socioeconomic homogamy was developed by
stratification researchers who used marriage patterns in conjunction with mobility
patterns to describe how open stratification systems are (Glass 1954).
Although the underlying issues are diverse, one common theme is that all
traditions characterize social differentiation by describing patterns of social
interaction. Building on the Weberian notion of status group closure, students
have argued that interaction between social groups provides a fundamental
way to describe the group boundaries that make up the social structure. Because
marriage is an intimate and often long-term relationship, intermarriage
or heterogamy not only reveals the existence of interaction across group
boundaries, it also shows that members of different groups accept each other as
social equals. Intermarriage can thus be regarded as an intimate link between
social groups; conversely, endogamy or homogamy can be regarded as a form
of group closure.
Another common theme lies in the consequences of intermarriage. First, intermarriage
decreases the salience of cultural distinctions in future generations
because the children of mixed marriages are less likely to identify themselves
with a single group. Although mixed couples may socialize their children into
the culture of a single group, these children are less likely to identify with that
group when intermarriage in society is common. Second, by intermarrying, individuals
may lose the negative attitudes they have toward other groups. Although
personal interaction between groups sometimes fosters conflicts by
making economic and cultural differences more apparent, if the relationship is
intimate, interaction gives people an opportunity to realize the individual variety
among the members of another group and, in doing so, may ultimately
weaken their prejudices and stereotypes. Because intermarriage often connects
the social networks of the two spouses, this applies to a range of outgroup
members and not just to the immediate partners.
396 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
In short, what makes intermarriage sociologically relevant lies in its inherent
dynamic: It is not just a reflection of the boundaries that currently separate
groups in society, it also bears the potential of cultural and socioeconomic
change. While marriage patterns are in this sense telling social indicators, they
do not tell us everything. First, if members of two groups do not marry one another,
it does not necessarily mean that both groups are closed. It takes two to
marry, and if one group is closed while the other is open, endogamy may still
prevail. Research on marriage is less informative in this respect than, for instance,
research on individual racial prejudice. In a similar vein, homogamy
tells a somewhat ambiguous story about the preferences and prejudices of
status groups. Homogamy will occur if people prefer to marry into high-status
groups, but it will also occur when people prefer to marry status-equals. In
high-status groups, preferences for high-status spouses and preferences for
status-equals are similar, but in lower-status groups, these are different.
Second, marriage patterns result from both preference and opportunity. Opportunity
to marry within the group depends on many factors, such as residential
segregation, the composition of local marriage markets, group size, and so
on. As a result, endogamy does not necessarily point to a personally felt social
distance toward a certain outgroup. Such preferences play a role, but to what
extent they determine the actual choices people make is an empirical question.
Marriage patterns simply tell us which groups interact with whom, and while
this is an important piece of information, they do not tell us why.
A third and final limitation of marriage patterns lies in demographic trends.
Declining marriage rates, the rise of cohabitation, and the increase in divorce
suggest that it is not always valid to treat marriage patterns as indicators of differentiation
in society as a whole. Some of these problems can be solved more
easily than others. The rise of cohabitation poses no real problem because one
can often include cohabiting couples in the analysis. Declining marriage rates
are also less of a problem because they are largely the result of marriage delays;
the vast majority of a given birth cohort eventually marries. The rise of
divorce is more problematic, because intermarriage and divorce are often positively
related. A high rate of ethnic intermarriage may point to open social
groups, but if mixed marriages are more likely to break up, such a conclusion
would need further study.
In the past decades, researchers have described patterns of intermarriage,
examined individual variations in intermarriage, and assessed changes in intermarriage
over time. In addition, both theoretical and empirical studies have
developed hypotheses about why people marry within their group and why
some do while others do not. Because such hypotheses are often not tested directly,
I divide my review into a theoretical and an empirical section. The goal
of the theoretical section is to review micro- and macro-level hypotheses about
the causes of intermarriage and homogamy and to put these into a general theoINTERMARRIAGE
AND HOMOGAMY 397
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
retical framework. The goal of the empirical section is to summarize patterns,
variations, and trends in intermarriage. I focus on the three main sociological
group characteristics (i.e. race and ethnicity, religion, and socioeconomic
status), I limit myself to Western societies, and I discuss studies conducted in
the last decade.
THEORETICAL WORK ON INTERMARRIAGE AND
HOMOGAMY
Marriage patterns arise from the interplay between three social forces: the
preferences of individuals for certain characteristics in a spouse, the influence
of the social group of which they are members, and the constraints of the marriage
market in which they are searching for a spouse (Kalmijn 1991b). Although
these factors represent analytically distinct hypotheses, they have most
often been regarded as complementary elements of a single theory, and that is
what distinguishes the sociological perspective from economic or psychological
theories on partner choice (e.g. Winch 1958).
Preferences of Marriage Candidates
To understand aggregate patterns of marriage selection, researchers use the
concept of a marriage market. Unmarried men and women operate within a
marriage market where each individual considers a set of potential spouses.
Potential spouses are evaluated on the basis of the resources they have to offer
and individuals compete with each other for the spouse they want most by offering
their own resources in return. Several kinds of resources obviously play
a role in the choice of a spouse, but sociologists have mostly focused on socioeconomic
and cultural resources. When married, spouses pool these resources
to produce family goods, such as economic well-being, status, social confirmation,
and affection.
SOCIOECONOMIC RESOURCES Socioeconomic resources are defined as resources
that produce economic well-being and status. Economic well-being is
shared by the family members and status is granted to the family as a unit
rather than to its individual members. As a result, the income and status of one
spouse contribute to the income and status of the other by raising the income
and status of the family. People maximize their income and status by searching
for a spouse with attractive socioeconomic resources. The outcome of this
competition is that the most attractive candidates select among themselves
while the least attractive candidates have to rely on one another. Competition
for socioeconomic resources on the marriage market thus leads to an aggregate
pattern of homogamy.
The nature of this competition varies with the role women play in society.
When marriage is based on the benefits that stem from the division of paid and
398 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
domestic labor in the household, prevailing gender differences in earnings
give men a comparative advantage in productive labor so that the wife’s time is
used more productively when it is spent on household labor. As a result, men
and women exchange paid and domestic labor resources. Similar arguments
have been made with regard to status and prestige. When the status of the family
depends primarily on the occupation of the husband, there will be an exchange
of male prestige and female qualities in other respects, such as class
background, physical attractiveness, and cultural participation (Jacobs & Furstenberg
1986; Stevens et al 1990; Uunk 1996).
Both types of exchange suggest that men, unlike women, do not compete
among themselves for female socioeconomic resources in the marriage market.
There are good reasons to believe that this has changed. An increasing
number of married women participate in the labor market and married
women’s work is now less often motivated by temporary economic needs of
the family. Several authors believe that these changes have made women’s socioeconomic
resources increasingly attractive to men. The wife’s human capital
may facilitate the husband’s access to networks that are helpful in his career,
her earnings may subsidize his human capital investments, and the economic
security she provides may lessen his need to settle for short-term career
benefits, thus increasing his opportunity to choose more attractive, long-term
career objectives. Because female labor is now often the reflection of women’s
desire to work outside the home, rather than a reflection of the economic needs
of the family, the wife’s socioeconomic resources may also become increasingly
important for the status of the family (Davis 1984).
CULTURAL RESOURCES While the importance of socioeconomic resources is
based on a preference to marry a resourceful spouse, independent of one’s own
resources, the role of cultural resources is based on a preference to marry
someone who is similar. Preferences for cultural similarity have been addressed
most extensively in the social psychological literature on personal attraction
(Byrne 1971). Similarity of values and opinions leads to mutual confirmation
of each other’s behavior and worldviews, similarity of taste is attractive
because it enlarges opportunities to participate in joint activities, and similarity
of knowledge creates a common basis for conversation, which enhances
mutual understanding.
Although originally developed to explain attraction between strangers in
day-to-day interaction, these notions have also been applied to marriage (DiMaggio
& Mohr 1985; Kalmijn 1994). Because cultural similarity leads to personal
attraction, it is a prerequisite for getting involved with someone. Because
of its instrumental effects, cultural similarity also encourages people to establish
a long-term relationship. Since many activities in marriage are joint, such
as the raising of children, the purchase of a house and other consumer durables,
INTERMARRIAGE AND HOMOGAMY 399
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
and the spending of leisure time, dissimilarity in taste would complicate these
shared activities. More generally, people prefer to marry someone who has
similar cultural resources because this enables them to develop a common lifestyle
in marriage that produces social confirmation and affection.
PREFERENCES AND HOMOGAMY Preferences for socioeconomic and cultural
resources do not by themselves translate into homogamy and endogamy with
respect to social characteristics. Some authors argue that social characteristics
are correlated with such resources, and that homogamy or endogamy is the unintended
by-product of individual preferences for resources in a partner. This
argument has often been made for educational homogamy, because education
is not only strongly related to income and status, but also to taste, values, and
lifestyles (Kalmijn 1991a). Similar arguments can be made for horizontally
differentiated groups, such as ethnic groups, although in this case, endogamy
is probably more the result of preferences for cultural similarity and not so
much the result of competition for economically attractive spouses.
Other authors argue that social characteristics are more than simply correlates
of the resources partners bring to the marriage market. Characteristics
such as education, occupation, race, and ethnicity are also seen as badges that
individuals wear to show others what kind of person they are. In this perspective,
spouse selection is regarded as a filter process. In the first step, people develop
a network of friends, acquaintances, and possibly marriage candidates
with whom they share some objective social characteristic. In the second step,
people find their spouse by interacting within these homogeneous networks.
The second step is also the phase in which psychological characteristics come
into play, but at that time, homogamy with respect to objective social characteristics
is already insured (Murstein 1976).
Third Parties
A second hypothesis about why people marry within their group focuses on
people who are not directly involved in the marriage. Because mixed marriages
may threaten the internal cohesion and homogeneity of the group, “third
parties” have an incentive to keep new generations from marrying exogamously.
There are two ways in which third parties prevent exogamy: by group
identification and by group sanctions.
GROUP IDENTIFICATION Children are typically brought up with a sense of
group identification. Identification either takes the form of an awareness of a
common social history, what is sometimes called a “sense of peoplehood”
(Gordon 1964), or it can take the form of a more psychological sense of being
different from others. The stronger such feelings of group identification, the
more people have internalized norms of endogamy, and the more likely it is
that they marry homogamously or endogamously. The notion of group identi-
400 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
fication has been especially important for racial and ethnic groups, where
norms of endogamy are believed to be firmly internalized (Merton 1941). Such
norms, however, may also apply to other kinds of groups such as social classes
and educational groups.
How strongly younger generations identify themselves with the group depends
to a great extent on the homogeneity of the networks in which they are embedded.
When adolescents live in neighborhoods that are homogeneous with
respect to the social and cultural characteristics of their parents, they are more
likely to develop a sense of belonging to that group. While residential segregation
in urban areas hampers opportunities to intermarry directly, as is discussed
later, it also reduced exogamy by intensifying feelings of group solidarity.
Identification with the origin group is believed to be weakened by higher
education. Owing to the emphasis on individual achievement and universalistic
principles in higher education, the college-educated may be less likely to
identify themselves with their social and cultural roots (Hwang et al 1995).
GROUP SANCTIONS Even if people have not internalized norms of endogamy,
they may still refrain from marrying exogamously because of the sanctions
third parties apply. The three most important examples of parties that
sanction intermarriage are the family, the church, and the state. Although in
Western societies parental control over children’s marriage decisions is limited,
there are still ways in which parents can interfere. They set up meetings
with potential spouses, they play the role of matchmaker, they give advice and
opinions about the candidates, and they may withdraw support in the early
years of the child’s marriage. Nevertheless, they do not have strong sanctions
when children decide against their will.
Somewhat stronger sanctions are provided by the church. Both the Catholic
church and various Protestant denominations have denounced interfaith marriages
for centuries, although the nature and strength of their disapproval have
changed over time. Religious institutions attempt to control intermarriage in
part because they are competing for members. Religious intermarriage entails
the risk of losing members and may weaken church attachment in future generations.
If interfaith marriages occur anyway, it is not always in the interest of
the church to apply sanctions because the competing church may accept the
marriage and hence gain members. This helps explain why the Roman Catholic
church, for example, has often accepted interfaith marriages on the condition
that the children be raised as Catholics. Because of competing pressures,
however, spouses in interfaith marriages often decide not to raise their children
in a religious fashion. If this occurs, both religious institutions stand to lose
strength in society.
The strongest sanctions against intermarriage have been provided by the
state. Laws on racial intermarriage in the United States—abolished in
INTERMARRIAGE AND HOMOGAMY 401
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
1967—are a well-known example (Davis 1991). When slavery was abolished,
the gradual decline in formal inequality of blacks and whites went hand-inhand
with a growing anxiety about the social boundary between the races, and
this anxiety was stronger when contacts were more intimate. Interracial dating
and marriage were condemned with great vigor, and strong social norms
emerged against interracial contacts with possible sexual undertones, such as
interracial dancing and swimming. The emerging doctrine of no social equality
was formalized in legislation that segregated the races in public facilities
(Jim Crow laws) and legislation that controlled their sexual and marital contacts
(antimiscegenation laws).
Marriage Markets
Endogamy and homogamy are not only governed by individual- and grouplevel
factors, but also by structural arrangements. The chances to marry endogamously
are higher the more often one meets people within the group and
the more often one interacts with group members on a day-to-day basis. Contact
opportunities are shaped by several structural arrangements. Some studies
focus on the demographic composition of the population as a whole, other
studies examine regional distributions of groups, and yet other studies analyze
smaller, functional settings, such as the school and the workplace.
THE LOGIC OF NUMBERS When interaction occurs randomly, the chance that a
woman in a certain group marries someone in her own group equals the proportion
of men who are in that group. As a result, members of a small group
will have lower chances of marrying endogamously than members of a larger
group. The effect of group size implies that endogamy is negatively related to
the degree of heterogeneity of a population (Blau & Schwartz 1984). To explain
this, one can think of two populations, each consisting of two groups.
One population is heterogeneous and has 50% in each group (e.g. 100 in group
A, 100 in group B), while the other is homogeneous and has 90% in one group
and 10% in the other (e.g. 180 in group A, 20 in group B). Both populations
have equal numbers of males and females in each group. In the heterogeneous
population, the number of women expected to marry within the group will be
0.5 x 50 = 25 for A and 0.5 x 50 = 25 for B, which boils down to 50% marrying
within the group. In the homogeneous population, the number of women expected
to marry within the group will be 0.9 x 90 = 81 for A and 0.1 x 10 = 1 for
B, which boils down to 82%. This shows that in a heterogeneous population,
endogamy is lower than in a homogeneous population, provided that marriage
is random.
THE GEOGRAPHY OF GROUPS The chance to encounter a member of one’s
own group does not depend on group size alone but also on the way a group is
402 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
dispersed geographically (Blau & Schwartz 1984). Groups that are concentrated
in specific regions of the country generally have more opportunity to
marry endogamously than groups that are not (Lieberson & Waters 1988). Examples
are common in the literature on ethnic groups, e.g. Asian-Americans in
California, Jewish-Americans in New York City, or Catholics and Protestants
separated in the southern and northern parts of the Netherlands. An additional
reason why it is important to consider the geography of groups is that isolation
may be correlated with group size. Smaller groups are often more isolated.
Jewish-Americans, for example, may have partly overcome the constraints of
their small group size through geographic concentration. They are a small
group in a large country, but a large group in a small region.
Although relaxing the assumption of an even geographic distribution is
more realistic, it also leads to new problems. If one controls for geographic
segregation—by calculating endogamy rates for specific regions, for instance—one
implicitly assumes that people base their decision to live in a
given area on factors that are independent of ingroup preferences. This is not
always realistic. For instance, there is much regional concentration of ItalianAmericans
in the United States, but even though this can in part be attributed to
their particular immigration history and occupational opportunities, the preferences
of Italian-Americans play a role as well (Lieberson 1980). While it is difficult
to make a precise distinction between preferences and constraints, it is
generally true that the smaller the marriage market one studies, the more the
structure of the market is affected by preferences and the less by constraints.
LOCAL MARRIAGE MARKETS Unmarried people do not just wander around a
region looking for a spouse; they spend most of their life in small and functional
places, such as neighborhoods, schools, workplaces, bars, and clubs.
Such “local marriage markets” are often socially segregated, and that is why
they are important for explaining marriage patterns. In the sociological literature,
three local markets have been considered most frequently: the school, the
neighborhood, and the workplace. Of these three, schools are considered the
most efficient markets because they are homogeneous with respect to age and
heterogeneous with respect to sex. Workplaces are considered less efficient,
but increased participation of women in the labor market and declining occupational
sex segregation suggest that this may have changed (Davis 1984). Although
it has not often been studied where couples meet, a French study shows
that the settings sociologists analyze are not the most common meeting places.
Among young French couples, fewer than 5% met in the neighborhood, fewer
than 10% met at school, and just over 10% met at work (Bozon & Heran 1989).
To clarify how local marriage markets affect homogamy, authors have
looked at the composition of these markets with respect to social characteristics.
What distinguishes the neighborhood from the school and the workplace
INTERMARRIAGE AND HOMOGAMY 403
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
is that it is homogeneous with respect to factors such as ethnicity, race, religion,
and family background, i.e. characteristics transmitted by parents (Lieberson
1980). Schools are less homogeneous in ascribed characteristics, although
there are exceptions, e.g. Catholic colleges and black colleges. At the
same time, schools are not necessarily homogeneous with respect to educational
attainment. Differences in ultimate educational attainment are larger in
high schools, for example, than in universities, simply because the educational
system works like a funnel, particularly in the United States (Mare 1991). In
general, however, it is expected that colleges promote educational homogamy
more than neighborhoods do, while neighborhoods promote ethnic endogamy
and homogamy of family background more than schools. Whether workplaces
encourage homogamy highly depends on the type of work, but on average,
they probably do not encourage socioeconomic homogamy as much as schools.
EMPIRICAL WORK ON INTERMARRIAGE AND
HOMOGAMY
Empirical work has addressed four questions: (a) To what extent are groups
endogamous or homogamous, and how do groups differ in these respects? (b)
How have endogamy and homogamy changed over time? (c) Which factors
are related to endogamy and homogamy, and in particular, what is the role of
gender, education, and geographic regions and local marriage markets? (d)
How do various dimensions of partner choice coincide? Before I summarize
the main findings, I discuss how researchers have tackled these issues method-
ologically.
Measures and Models
Intermarriage can be calculated for the stock of marriages at a given point in
time (prevalence measures) or for people who marry in a given period of time
(incidence measures). Incidence measures are generally preferable, in particular
if one analyzes trends. If the stock of marriages is used, one can analyze
characteristics at the time of survey or characteristics at the time of marriage.
The latter measures are more suitable than the former because some characteristics
change after marriage. Because partners may become more alike during
marriage—they may switch faith, for example, or influence each other’s occupational
career—current measures of homogamy tend to be biased upwardly.
To describe intermarriage, various measures have been used. To explain these,
it is helpful to consider the following marriage table.
MEASURES The most general measure is the percentage of couples intermarrying:
(CBA+CAB)/N. When calculating group-specific measures, it makes a
difference if one considers couples or individuals. The percentage of A-type
404 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
couples intermarrying is (CBA+CAB)/(CBA+CAB+CAA), while the percentage
of A-type married persons intermarrying is CAB/MA for males and CBA/FA for
females. While percentages are simple and informative measures to describe
intermarriage, they provide little information about the strength of endogamy
because they lack a reference point. If 40% of a group marries endogamously,
is this evidence for a preference to marry within rather than outside the group?
Percentages are also less useful for comparing groups because when selection
is random, small groups are less likely to marry within their group than large
groups.
These problems are overcome by a more recent measure, the odds ratio. The
odds ratio is defined as the odds that an A-type male marries an A-type female
(rather than a B-type female), divided by the odds that a B-type male marries
an A-type female, i.e. (CAA/CAB)/(CBA/CBB). The odds ratio for women is
equivalent, i.e. (CAA/CBA)/(CAB/CBB). If there are more than two groups in
the marriage table, one can calculate odds ratios for each group separately. If
CAX and CXA are marriages of A-type males and females with all other groups,
and CXX are marriages that do not involve A-type males or females, the odds
ratio can be defined as (CAA/CAX)/(CXA/CXX). Odds ratios have two important
advantages. First, they provide a reference point: Odds ratios greater than
one indicate that there is more endogamy than one would expect, and the larger
the ratio, the greater the degree of endogamy. Second, odds ratios are useful for
comparing endogamy across groups because they are independent of the relative
sizes of the groups in the marriage table.
A disadvantage of the measures discussed above is that they are based on
the married or marrying population. A measure of intermarriage that takes into
account that not everyone marries is the intermarriage index Z, which is based
on so-called harmonic mean models developed by Schoen (1988). If MP and FP
refer to the total number of males and females in the respective groups (married
and unmarried), Z is defined as (CAB/MPA+CBA/FPA+CBA/MPB+CAB/
FPB)/(MA/MPA+MB/MPB+FA/FPA+FB/FPB). This intermarriage index ranges
from zero for minimum intermarriage to one for maximum intermarriage.
When selection is random, the index takes the value of 0.5 (Schoen 1988).
Percentages, odds ratios, and the intermarriage index can be applied to both
ordered and nonordered characteristics. For ordered characteristics, another
common measure is the Pearsonian correlation between spouses’ traits. A
positive correlation means that high-status men marry higher-status women
INTERMARRIAGE AND HOMOGAMY 405
FEMALES
Group A Group B Total married
Group A CAA CAB MA
MALES Group B CBA CBB MB
Total married FA FB N
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
than low-status men; it does not necessarily mean that people marry within
their group. The correlation between the ages of husband and wife, for example,
is strongly positive, even though most men marry somewhat younger
women.
MODELS Second to measures of intermarriage, loglinear models have been
used to describe patterns of marriage selection. These models assume that the
expected counts in the marriage table are a multiplicative function of sample
size, the number of males in a group, the number of females in a group, and an
interaction parameter, which measures marriage selection independent of the
marginal row and column distributions. Many ways to model the interaction
parameter exist, but most authors present parameters for the tendency to marry
within the group (endogamy) and parameters for the tendency to avoid intermarrying
when controlling for the tendency to marry within the group (intermarriage).
The latter parameters are often equivalent to odds ratios and have
been described by the metaphors of distances or boundaries between groups
(Mare 1991; Kalmijn 1991b). When characteristics are ordered, loglinear
models also provide single measures of association that are comparable to correlations
but independent of the marginal distributions, i.e. uniform association
models (Hout 1982). When characteristics are not ordered, special types
of loglinear models exist that provide measures of the distances between
groups as revealed by the marriage frequencies in the table, i.e. logmultiplicative
models (Johnson 1980; Kalmijn 1993a).
Patterns of Intermarriage and Homogamy
In describing patterns of intermarriage and homogamy, researchers have addressed
three questions: (a) To what extent do subgroups marry endogamously?
(b) If subgroups marry out, with what groups are they most likely to
intermarry? and (c) How do subgroups compare in their degree of endogamy?
RACE/ETHNICITY Most American studies of ethnic intermarriage analyze
data from the perspective of the minority group and focus on specific types of
subgroups. Several decades ago, interest largely focused on European immigrant
subgroups and their children. More recently, new immigrant subgroups
such as Asian- and Hispanic-Americans are being studied, though there is a resurgent
interest in the descendants of the older immigrants, sometimes referred
to as white ethnic groups. Intermarriage of blacks has always been studied frequently.
Research on intermarriage of American Indians, in contrast, is scarce.
Recent national estimates of the percentage of persons who are married endogamously
vary around 95% for blacks (Sweet & Bumpass 1987), 75% for
Asian subgroups (Lee & Yamanaka 1990), 65% for Hispanic subgroups (U.S.
Bureau of the Census 1985), 45% for American Indians (Snipp 1989), and
406 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
25% for (unmixed) European subgroups (Alba & Golden 1986). Although
these percentages are high, they do not indicate whether groups are endogamous.
Loglinear models and harmonic mean analyses are more informative in
this respect and show that virtually all ethnic subgroups marry within their
group more often than can be expected under random mating (Jiobu 1988;
Schoen & Thomas 1989; Sandefur & McKinnell 1986; Alba & Golden 1986).
A low degree of endogamy does not necessarily imply integration; this also
depends on patterns of outmarriage. When Hispanic subgroups marry exogamously,
for example, they often marry with other Hispanic subgroups and
hence keep a distance from the non-Hispanic white majority (Gurak & Fitzpatrick
1982). Asian-Americans who marry out, in contrast, rarely marry other
Asian subgroups and instead marry with whites (Lee & Yamanaka 1990). Intermarriage
between European subgroups also reveals meaningful patterns. In
the first half of this century, marriage selection was characterized by a large
distance between Western and Northern Europeans on the one hand and
Southern, Central, and Eastern Europeans on the other (Pagnini & Morgan
1990; Kalmijn 1993a). This finding has been interpreted as evidence of a
boundary between the “old” and “new” European immigrants to the United
States.
How do subgroups compare in their degree of endogamy? Unfortunately,
most studies focus on one type of subgroup at a time without analyzing other
types. A broader focus is provided by Lieberson and Water, who present a list
of odds ratios for more than 20 groups (1988). European subgroups and
American Indians appear to have the lowest rates of endogamy, Hispanic and
Asian subgroups have intermediate levels of endogamy, and blacks have the
highest rates. There are also differences among European subgroups—for instance,
endogamy is lower for “old” than for “new” European groups—but
these are small when considering the range in the list. The main conclusion of
Lieberson and Water’s analysis is that groups who are more recent to the host
society have higher degrees of closure, a regularity that fits well into assimilation
theories. Blacks are the prime exception to this pattern. A similar relationship
between the newness of a group and its level of endogamy is found in
Australia (Jones & Luijkx 1996).
RELIGION Religious intermarriage has primarily been studied in religiously
heterogeneous societies. Some authors use current religious affiliation to
measure intermarriage, while others use the religion in which spouses were
raised. Endogamy is higher when current affiliation is used, because spouses
often switch faith or lose their religion after entering a mixed marriage (Glenn
1982). Estimates for the United States that use parental religion show that in
the late 1970s, 62% of Catholics were married within their group, 84% of Protestants
were married endogamously, and 80% of Jews were married endogaINTERMARRIAGE
AND HOMOGAMY 407
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
mously (Glenn 1982). Loglinear analyses further show that both Catholics and
Protestants have a tendency to marry within rather than outside their group;
this is found in the United States (Kalmijn 1991b), the Netherlands (Hendrickx
et al 1991), Germany (Hendrickx et al 1994), Australia (Hayes 1991), and
Switzerland (Schoen & Thomas 1990). Comparisons between countries using
odds ratios show that endogamy is strongest in Ireland and Northern Ireland, as
one would expect, and that Catholic endogamy is stronger in the United States
than in most European countries (Klein & Wunder 1996).
Which religious groups are most endogamous? Loglinear analyses in the
United States indicate that Catholics are somewhat more closed than Protestants
(Johnson 1980). Loglinear analyses have not included Jews, but it is safe
to say that American Jews are more endogamous than Catholics because their
intermarriage percentage is comparable to that of Catholics, while they are a
much smaller group. Detailed loglinear analyses of Protestant denominations
show that more-conservative Protestant denominations, such as the Rereformed
in the Netherlands (Hendrickx et al 1994) and Baptists in the United
States and Australia (Johnson 1980; Hayes 1991) are more endogamous than
liberal denominations. These findings are consistent with the notion of thirdparty
control: Denominations and religions that are more traditional in religious
doctrine and have higher degrees of church involvement among their
members also have the highest degree of endogamy.
When analyzing marriages between religious groups, authors have used the
concept of social distance and have developed loglinear models providing
such measures. In the United States, such analyses point to the following order
of groups: Baptists, Methodists, liberal Protestants, Lutherans, and Catholics
(Johnson 1980). These distances have been interpreted in terms of ritual and
regional dimensions. Groups on the left side of the continuum are more democratic
in their organization, have less detailed prescriptions of ritual, and emphasize
more spontaneous forms of worshiping than groups on the right side of
the continuum. Regional patterns play a role here as well. Baptists and Methodists
are concentrated in the South, for example, while Catholics are concentrated
in the Northeast.
SOCIOECONOMIC STATUS The literature on socioeconomic homogamy can be
distinguished into studies of ascribed status and studies of achieved status. Ascribed
status positions are measured by the occupational class of the father and
the father-in-law. Achieved status positions are measured by education and occupation.
Education is used more often because it is a convenient status indicator
of women and changes little after marriage. In most countries, educational
homogamy is quite strong (about 0.55), occupational homogamy is somewhat
weaker (about 0.40), while the correlation between husbands’ and wives’ class
origins is the weakest, about 0.30 (Kalmijn 1991a; Uunk 1996).
408 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
Loglinear analyses provide additional insights in the association between
partners’ status positions. Such analyses first show that people marry within
rather than outside socioeconomic groups, although some groups are more
closed than others. Groups at the top and the bottom of the educational hierarchy
are more closed than groups in the middle (Uunk et al 1996; Hendrickx
1994). These tendencies may be due to the role of opportunity: If people at the
bottom prefer to marry out, they can only choose higher groups and if people
at the top prefer to marry out, they can only choose lower groups. People
from farm background have an exceptionally high rate of endogamy, a finding
that can probably be explained in terms of the social and geographic isolation
of the rural population (Kalmijn 1991a; Uunk et al 1996; Jones & Davis
1988).
Next to a tendency to marry within the group, there is a tendency for marriage
to become less common the farther away the two status positions are.
Some status boundaries are harder to cross than others, however. For education,
the strongest boundary is that between college graduates and lessereducated
persons (Mare 1991; Kalmijn 1991a). A common interpretation of
this finding is that colleges function as local marriage markets that are physically
separated from settings in which lesser-educated persons are involved.
Patterns of occupational homogamy, like patterns of intergenerational occupational
mobility, are dominated by the line that divides blue-collar and whitecollar
occupations (Hout 1982; Hayes 1993). More detailed analyses of occupational
homogamy have shown that there is more homogamy with respect to
the cultural status of occupations than with respect to the economic status of
occupations (Kalmijn 1994). This suggests that preferences for cultural similarity
are stronger than preferences for economically attractive spouses.
Trends in Intermarriage and Homogamy
Trends have been analyzed in three ways. (a) Some researchers compare marriage
or birth cohorts at a single point in time. Such synthetic cohort studies are
potentially biased because older cohorts have been married longer than
younger cohorts. Cohorts differ in their rate of attrition, and this attrition may
be selective because the likelihood of divorce is inversely related to homogamy.
(b) Other studies compare the stock of intact marriages at different points
in time. This design has the disadvantage of containing much overlap of marriages
in the points of comparison, which leads to an underestimation of linear
trends. (c) Others, finally, compare recently formed marriages in different periods,
either through annual marriage licenses or through comparisons of newlyweds
in multiple surveys or censuses. This method is most suitable for analyzing
trends because it gives a picture of the changing incidence of intermar-
riage.
INTERMARRIAGE AND HOMOGAMY 409
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
RACE/ETHNICITY Analyses of marriage records generally reveal an increase
in intermarriage of new ethnic groups in the last decades; this applies to Hispanic
intermarriage in New York City (Gilbertson et al 1996), to Asian intermarriage
in Hawaii (Schoen & Thomas 1989), and to Asian intermarriage of
males—not females—in New York City (Sung 1990). The trend in blackwhite
intermarriage has been documented for a longer time period and in more
states. Annual marriage records in 33 states reveal that black-white intermarriage
has increased significantly in both northern and southern states since the
legal ban on intermarriage was lifted, although it remains exceptionally low
(Kalmijn 1993b). Comparisons of the 1980 and 1990 American censuses confirm
this conclusion (Qian 1997). Ethnic characteristics that are not included in
marriage licenses have primarily been analyzed through synthetic cohort
analyses. Such studies reveal growing outmarriage across birth cohorts for
European-American groups (Alba & Golden 1986; Lieberson & Waters 1988),
for American Indians (Eschbach 1995), and for ethnic groups in Australia
(Jones & Luijkx 1996).
The decline in ethnic endogamy has typically been interpreted from an assimilation
perspective: Through generational replacement, national origin
groups gradually integrate in the host society. Consistent with this perspective,
most analyses find that the children of immigrants marry out more often than
the immigrants themselves (Gilbertson et al 1996; Lee & Yamanaka 1990;
Alba 1976). Because trends also occur within generations of immigrants, individual
assimilation to the host society is not a sufficient explanation (Gilbertson
et al 1996; Sung 1990). An additional interpretation is that assimilation is a
process at the macro level: When more and more members of an ethnic group
are of the second or third generation—when an ethnic group becomes
“older”—all generations find it easier to adapt to the host society. Another interpretation
is more general in nature and points to the weakening influence of
third parties in marriage choice and the declining importance of ascription as a
basis of evaluating other people.
RELIGION In the United States, trends in religious homogamy have primarily
been assessed through surveys, largely because few places report religion on
their marriage licenses and because the census is not allowed to ask questions
on religious affiliation. By analyzing national surveys conducted between
1955 and 1989 and using a design that separates the effects of period and duration
of marriage, Kalmijn (1991b) shows that intermarriage between Catholics
and Protestants has increased in a linear fashion between 1920 and 1980.
Trends in intermarriage between Jews and non-Jews have been documented by
comparing subsequent surveys (Lazerwitz 1995) and by comparing marriage
cohorts within a single survey (Kosmin et al 1991). Both types of analyses
show that Jewish-Gentile intermarriage has increased considerably over the
410 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
last decades. An exception to these trends are conservative Christian groups,
who appeared to have stable endogamy rates over time (McCutcheon 1988).
In many other Western societies, questions on religion are included in marriage
records so that long-term trends can be documented there more easily. A
loglinear trend analysis of annual Dutch marriage records since the 1930s
shows that religious endogamy of Catholics and the conservative ReReformed
Protestants has declined (Hendrickx et al 1991). The more liberal
Dutch Reformed Protestants experienced no decline, but they had low levels of
endogamy to begin with. Marriage records in Switzerland (Schoen & Thomas
1990) and Germany (Hendrickx et al 1994) also reveal a decline in the level of
endogamy of Catholics and Protestants. That the boundaries between religious
groups in Europe and the United States have weakened during the twentieth
century is consistent with the notion of declining third-party control and
matches long-term processes such as secularization and depillarization.
SOCIOECONOMIC STATUS Trends in socioeconomic homogamy are most frequently
studied by analyzing class background and education. In most industrialized
countries, there has been a decline in the importance of social background
for marriage choice. This has been found for the United States (Kalmijn
1991a), the Netherlands (Uunk 1996), Hungary (Uunk et al 1996), and
France (Forsé & Chauvel 1995). The most common interpretation of this trend
lies in the role of third parties and opportunity. Young adults have become increasingly
independent of parents so that parents have less direct or indirect
control over the choices their children make. People also spend more time in
school settings, which are more heterogeneous with respect to social class
background than the parental neighborhood.
Trends in educational homogamy do not point in one direction. A loglinear
analysis of 18 postwar industrial nations by Ultee & Luijkx (1990) reveals that
five countries experienced a decline in educational homogamy, three countries
experienced an increase, while the remaining ten revealed no meaningful
trend. Country-specific loglinear analyses also reveal a mixed pattern, although
they do not reveal a decline in homogamy: (a) a strong increase in the
United States (Mare 1991; Kalmijn 1991a), Hungary (Uunk et al 1996), and
Germany (Blossfeld & Timm 1997); (b) a slight increase in the Netherlands
(Hendrickx 1994); and (c) stability in Australia (Jones 1987) and France
(Forsé & Chauvel 1995).
Several hypotheses have been suggested to explain these trends. Some
authors argue that opportunities for making a match on education have increased.
People marry later and spend more time in school, but the time interval
between leaving school and marriage has narrowed. As a result, it is now
more likely that unmarried people, especially the college educated, meet their
spouse in school (Mare 1991). Others point to the role of preferences. EducaINTERMARRIAGE
AND HOMOGAMY 411
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
tion has become an increasingly important proxy for both cultural taste and socioeconomic
success, and competition among men for socioeconomic resources
in women may have increased (Kalmijn 1991a; Schoen & Wooldredge
1989). There are also hypotheses predicting a decrease in educational homogamy.
Some authors argue that marriage choice has become increasingly based
on emotional or affective considerations. Because romantic considerations often
overrule status concerns, one would expect a decline in all forms of status
homogamy (Ultee & Luijkx 1990).
An attempt to reconcile these hypotheses is made by Smits et al (1998), who
argue that educational homogamy will initially increase with levels of industrialization
because in this phase, education becomes the dominant criterion
for socioeconomic success and cultural norms and values. Romantic considerations
and individualism gain importance in later stages of the industrialization
process when high standards of living are guaranteed for everyone. As a
result, educational homogamy will first increase with levels of industrialization,
but will eventually decrease. A comparison of 64 countries provides indirect
support for this claim: The relationship between educational homogamy
and the level of industrialization follows an inverted “U.”
Variations in Intermarriage and Homogamy
Next to describing patterns and trends, authors have analyzed variations in intermarriage
and have examined what factors contribute to outmarriage. Recurring
themes in the literature are differences by sex, by education, and by region.
Although these factors are generally studied in an exploratory fashion,
they also give us clues about the causes of endogamy.
SEX DIFFERENCES Sex differences have most often been studied in the literature
on racial and ethnic intermarriage. Studies on black-white intermarriage
in the United States consistently show that black men marry whites more often
than black women (Kalmijn 1993b; Schoen & Wooldredge 1989). A traditional
interpretation of this finding is that minority men are able to compensate
for their lower “ethnic prestige” by offering white women a high occupational
status or income. Although in principle one could reverse the exchange—highstatus
minority women could marry white men of lower status—under conditions
of traditional sex-roles, this type of marriage is believed to be uncommon
because the status of the family is largely dependent on the status of the hus-
band.
Although the interpretation is plausible, findings for other ethnic groups
provide a counterpoint. Asian-American women, for example, and in particular
Japanese-American women, marry whites more often than their male counterparts
(Sung 1990). A speculative interpretation of this exception is that
Asian-American women are attractive marriage candidates for white men be-
412 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
cause of their physical appearance and presumed acceptance of more traditional
power relationships in marriage. A more-plausible interpretation lies in
the role of opportunity: the presence of American soldiers in Japan and Korea.
A recent analysis shows that excluding such war brides leads to a substantial
reduction in the sex differential in Asian-American intermarriage (JJ Jacobs &
T Labov, unpublished manuscript).
Sex differences have also been studied in the analysis of socioeconomic homogamy.
A common finding is that highly educated men and men in professional
and technical occupations marry down more often than up (Mare 1991;
Kalmijn 1994). Laymen generally interpret downmarrying as evidence of a reluctance
on the part of men to marry high-status women, but most of the asymmetry
is due to differences in the composition of men’s and women’s characteristics.
On average, women have traditionally been less educated and less often
have had high-status occupations than men. Once such differences are
taken into account through loglinear analyses, researchers generally find little
evidence of asymmetry (Mare 1991). Similar conclusions apply to trends:
Educational downmarrying among men has become less common, but this is
largely due to the increased educational attainment of women (Mare 1991).
EDUCATIONAL EFFECTS Another frequently examined factor in intermarriage
is education. Many studies have found that more highly-educated members of
ethnic or racial minority groups marry exogamously more often than their
lesser-educated peers. This applies to white ethnic groups (Lieberson & Waters
1988), blacks (Kalmijn 1993b; Schoen & Wooldredge 1989), and American
Indians (Sandefur & McKinnell 1986). Less consistent evidence is found
for outmarriage of Asian-Americans (Hwang et al 1995; Wong 1989; Schoen
& Thomas 1989).
Educational effects have been interpreted in terms of both opportunity and
preference. The former interpretation states that better educated minority
members are more often exposed to settings such as colleges and high-status
occupations where they form a relatively smaller group than in the population
at large. Another interpretation states that more highly educated persons—of
both majority and minority groups—have a more individualistic attitude, are
less attached to their family and community of origin, and have a more universalistic
view on life than lesser-educated persons. As a result, they would find
ascribed characteristics less relevant in deciding whom to marry.
DIFFERENCES BETWEEN REGIONS AND SETTINGS Virtually all studies find
large regional differences in intermarriage. Asian-Americans marry out less
often in California, where they are concentrated, than in the rest of the United
States (Wong 1989), Indian-Americans are more endogamous in so-called Indian
States (Sandefur & McKinnell 1986), and blacks marry more endogaINTERMARRIAGE
AND HOMOGAMY 413
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
mously in states where the percentage of blacks in the population is larger, a relationship
that is observed both in and outside the South (Kalmijn 1993b).
Such patterns undoubtedly point to the role of opportunity: The smaller the
group, the more difficult it is to marry within the group. Hypotheses about
group size and its corollary, heterogeneity, have also been examined by analyzing
(Standard) Metropolitan Statistical Areas [(S)MSAs) or states through
correlational analyses. These studies find that relative group size is negatively
correlated with black outmarriage and ethnic outmarriage; similarly, racial,
ethnic, and occupational heterogeneity have positive effects on the respective
types of intermarriage (Blau et al 1982; Hwang et al 1994).
The role of opportunity has also been analyzed by considering local marriage
markets. In a classic study, Ramsøy (1966) analyzed marriage licenses in
a Norwegian city and found that husbands and wives lived close to each other
before marriage, and in fact closer than one would expect under conditions of
random mating in a city. While this confirms that neighborhoods are marriage
markets, Ramsøy also showed that people who lived close to one another before
marriage did not marry more homogamously with respect to occupation
than people who lived far apart. Hence, it appeared that the neighborhood did
not by itself promote occupational homogamy. Ramsøy’s analysis was criticized
on methodological grounds by Peach (1974), but later, more elaborate
analyses of newlyweds in a New Zealand city by Morgan (1981) found no
clear link between spatial and status proximity either. Analyses of ethnic endogamy
yield a more promising conclusion. Anderson & Saenz (1994), for example,
find that MSAs in which Mexican-Americans are residentially segregated
from non-Hispanic whites have lower degrees of Mexican-American
outmarriage, even when other group characteristics such as mean educational
level are controlled for.
The school is another local marriage market, but its role in promoting educational
homogamy has only been studied indirectly. Mare (1991) found that
people who marry closer to finishing school, or while in school, marry people
who are more similar in education than people who marry long after finishing
school. This relationship was found to be present only for higher levels of education.
A similar relationship is found in Germany, although there, a decline in
educational homogamy is only observed when comparing couples who married
a few years after leaving school with couples who married much later
(Blossfeld & Timm 1997). These findings provide indirect evidence that
schools function as marriage markets that favor educational homogamy.
Schools, and in particular colleges, are educationally homogeneous, while the
settings people face when they search for a partner at a later stage, such as work
settings and public places, tend to be less homogeneous. Further evidence on
the role of schools is provided by Uunk & Kalmijn (1996), who show that the
college-educated in the Netherlands have a tendency to marry someone who
414 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
has the same college major. In the Netherlands, fields of study can be regarded
as local marriage markets within the university because students choose a major
when they first enroll and do not follow courses in other fields.
Multiple Dimensions
Most early studies analyzed a single sociological characteristic at a time or
analyzed several characteristics one-by-one. Since partners choose each other
on the basis of multiple characteristics, it is important to analyze more than one
factor in marriage choice. In the last decades, several such multidimensional
analyses have been done, although most are limited to two dimensions. Research
on multiple dimensions has been guided by two hypotheses: the byproduct
hypothesis and the exchange hypothesis.
BY-PRODUCT HYPOTHESES There is a considerable overlap between social
groups in society. Ethnic groups, for example, differ in educational level, religion
and ethnicity often coincide, and education and social background are
correlated. Because the various social dimensions on which individuals select
one another are correlated, and because people are believed to take all these dimensions
into account when choosing a spouse, the question arises if and to
what extent homogamy in one group dimension is the by-product of selection
in another group dimension.
An early attempt to examine this issue empirically was done by Warren
(1966; see also Blau and Duncan 1967:354–59), who showed that the correlation
between the spouses’ fathers’ occupations is reduced substantially when
controlling for spouses’ education. This result led Warren to conclude that social
class homogamy is largely a by-product of educational homogamy. More
recent analyses confirm this and show in addition that educational homogamy
is in part a by-product of matching of social origins (Kalmijn 1991a; Uunk
1996). Hence, both forms of homogamy appear to be weaker when a multidimensional
analysis is used, although even then, educational homogamy remains
stronger than homogamy of social origins.
The by-product hypothesis has also been a theme in the study of ethnic intermarriage.
A classic study of New Haven in the first half of this century by
Kennedy (1944) showed that intermarriage is more common between groups
who have the same faith, such as between Italians and Poles on the one hand
(both largely Catholic) and between Hungarians and Russians on the other
(both largely Jewish). Kennedy used the now classic term “triple melting pot”
to describe this pattern. Kennedy’s triple melting pot confirms the by-product
hypothesis because it reveals that marriage boundaries between certain ethnic
groups are in part the result of differences with respect to religion. Recent,
more sophisticated loglinear analyses confirm that there are strong marriage
boundaries between ethnic groups who have a dissimilar faith (Alba & Golden
INTERMARRIAGE AND HOMOGAMY 415
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
1986), but no studies have simultaneously analyzed individual ethnic and religious
characteristics of husbands and wives.
EXCHANGE HYPOTHESES A second theme in multidimensional analyses is the
question of whether people trade characteristics when choosing a spouse. Several
examples of exchange have been considered, but the most debated case
was introduced by Davis (1941) and Merton (1941), who argued that members
of ethnic groups whose prestige in society is low would have better chances of
marrying outside their group if they offered a high socioeconomic status in re-
turn.
The Davis-Merton hypothesis is most frequently examined in research on
ethnic and racial intermarriage. Loglinear and harmonic mean analyses of
black-white intermarriage by Kalmijn (1993a) and Schoen & Wooldredge
(1989) show that with respect to education, white women marry up more often
when marrying a black man than when marrying a white man; similarly, black
men marry down more often when marrying a white woman than when marrying
a black woman. Similar conclusions apply when examining the marriage
choices of white men and black women. White men marry down less often
when marrying exogenously and black women marry up less often in mixed
marriages. These asymmetries in spouses’ educational characteristics are assessed
after controlling for the marginal educational distributions of race-sex
groups and thereby support the hypothesis that majority men and women
marry a minority spouse in part under the condition of socioeconomic status
gains.
While the pattern of black-white marriage provides support for the DavisMerton
hypothesis, studies of other types of ethnic homogamy are less consistent.
In a harmonic mean analysis of Asian intermarriage in Hawaii, Schoen &
Thomas (1989) show that after controlling for differences in educational distributions,
white women marry up more often when they marry Filipino and
Japanese males, consistent with the notion of exchange. The reverse is true,
however, when white women marry Hawaiian or Chinese males.
The exchange hypothesis has also been applied to other dimensions of partner
choice, such as physical attractiveness and cultural participation. To examine
exchanges, studies generally rely on correlational analyses in which socioeconomic
characteristics of the husband are regressed on socioeconomic and
noneconomic characteristics of the wife. Effects of the wife’s noneconomic
characteristics on the husband’s socioeconomic characteristics are usually
called crossing effects and are considered evidence for exchange.
In an early analysis of physical attractiveness of women and occupational
prestige of men, Taylor & Glenn (1976) show that female attractiveness has a
positive effect on the occupational prestige of the man she marries, even when
controlling for her own socioeconomic characteristics. A drawback of this
416 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
analysis is that husband’s attractiveness was not included in the model. If occupational
prestige and attractiveness are correlated within individuals, and if
people match in attractiveness, part of the effect of female attractiveness on
male prestige may be spurious. A more recent analysis, which also controls for
the physical attractiveness of the husband, confirms this. Stevens et al (1990)
find no effect of female attractiveness on husbands’ education, suggesting that
no exchanges are being made.
Another example of exchange is that between socioeconomic status and
participation in high culture. DiMaggio & Mohr (1985) find that participation
of the wife in high culture has a positive effect on the educational level of the
husband, net of the educational level of the wife. Because no measures of the
husband’s cultural participation were included in this model, the exchange effect
might again be due to homogamy with respect to cultural participation. An
analysis for the Netherlands, however, shows that this is not the case. Uunk
(1996) analyzes the correlation between the wife’s cultural participation and
the husband’s educational level while controlling not only for the wife’s educational
level but also for the husband’s cultural participation. Uunk finds a
significant partial association between female high culture and male education,
providing support for the exchange hypothesis.
CONCLUSION
Sociological research on marriage choices has generated many insights in how
modern society is differentiated. In general, social groups in society appear
closed, in the sense that men and women more often choose partners within
their group than one would expect under random mating. Although some
groups are more closed than others, examples of social groups who marry exogamously
have not (yet) been found. Research on intermarriage also reveals
how societies change. Overall, ascribed bases of group membership have become
less important, while achieved bases of group membership, and especially
those governed by education, have not lost salience. This is not to say
that ascribed groups are mixing freely now. Ethnic, religious, and particularly
racial boundaries still exist, but they are weaker than they used to be.
The focus of the literature has largely been descriptive. Researchers have
scrutinized a mass of data—coming from censuses, surveys, and marriage licenses—and
have studied many groups, several countries, and long periods of
time. Because marriage patterns are telling indicators of how closed groups in
a society are, the descriptive focus of the literature has much to say for it.
Monitoring such a social indicator also requires a certain degree of standardization,
and in this respect, the literature has its shortcomings. Some studies are
limited to percentages, which are heavily affected by relative group size, making
it difficult to compare endogamy across groups. Other studies use loglinear
INTERMARRIAGE AND HOMOGAMY 417
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
models, and while these are an advance from a methodological point of view,
they have probably made the literature less accessible to a general audience
than it deserves to be. In carrying on its descriptive mission, the literature
would gain by using simple odds ratios next to percentages or loglinear models
when describing the degree of endogamy of groups. A good example of such
an approach is provided in Lieberson & Waters (1988). In describing trends, it
would also be useful to focus on recently formed marriages rather than on the
stock of marriages at a given point in time, largely because the aim is to present
social indicators for a clearly defined period in time.
In a theoretical sense, there has also been progress in the field. There are
many theories about partner choice, and such notions provide important clues
about the causes of intermarriage and homogamy. In general, marriage patterns
arise from three social forces: the preferences of individuals for resources
in a partner, the influence of the social group, and the constraints of the marriage
market. The multifaceted perspective that has been developed over the
years gives sociological theorizing an edge over competing theories of marriage
choice such as those developed by psychologists and economists.
Considerable empirical evidence exists for these theoretical notions, but the
integration of empirical and theoretical work is less than perfect. There are two
basic problems in empirical work. First, many hypotheses are tested in an indirect
fashion. The role of third-party control, for example, is documented by
comparing ethnic groups or religious denominations, but little information is
available on what these parties in fact are doing. Similarly, the trend towards
increasing educational homogamy may point to heightened competition for
economic resources on the marriage market, but this is an interpretation, not a
test. A second and related problem in empirical work is that many of the observed
regularities and relationships can be attributed to all three types of
causes, while little is yet known about the relative strengths of these factors.
That more highly educated members of minority groups are less endogamous,
for example, may be attributed to a universalistic attitude brought about by
higher education, but can also be explained in terms of greater opportunities to
meet outgroup members. In a similar fashion, residential segregation may foster
endogamy by lowering opportunities, but its effect may also be explained
by different socialization practices in segregated areas.
While it is clear that progress can be made in integrating theoretical and empirical
work, this is not a straightforward task either. One possible solution is
to shift the focus from the aggregate to the individual level. In the past, most
studies have compared countries, groups, or time periods, but few studies have
analyzed individual differences in intermarriage. The main advantage of an individual
approach is that it facilitates the inclusion of a range of covariates for
each of the three elements of the theory. For example, one can focus on socialization
practices and characteristics of the parental home to test hypotheses
418 KALMIJN
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
about third-party control. One can also include sex-role attitudes or expectations
regarding paid and domestic labor later in life to examine the role of preferences.
Finally, one can include contextual variables, such as characteristics
of the settings in which young adults are embedded (schools, workplaces,
neighborhoods) to assess the effects of local marriage markets on intermar-
riage.
While an individual design would facilitate the multivariate analyses that
are needed to test hypotheses more directly and to compare the strength of alternative
explanations, such a design has its problems as well. Because multivariate
analyses of marriage choices use individuals as the unit of analysis,
they provide a one-sided view of marriage. It takes two to marry, and for that
reason most authors have used loglinear or harmonic mean models. Such models
correctly use marriages as the unit of analysis, rather than individuals, but
make it difficult to include multiple covariates in the model. From a methodological
point of view, such models are preferable, but if the prime concern is to
test theories, their advantage is not so obvious.
Visit the Annual Reviews home page at
http://www.AnnualReviews.org.
INTERMARRIAGE AND HOMOGAMY 419
Literature Cited
Alba RD. 1976. Social assimilation among
American Catholic national-origin groups.
Am. Sociol. Rev. 41:1030–46
Alba RD, Golden RM. 1986. Patterns of ethnic
marriage in the United States. Soc. Forces
65:202–23
Anderson RN, Saenz R. 1994. Structural determinants
of Mexican American intermarriage,
1975–1980. Soc. Sci. Q. 75:
414–30
Blau PM, Blum TC, Schwartz JE. 1982. Heterogeneity
and intermarriage. Am. Sociol.
Rev. 47:45–62
Blau PM, Duncan OD. 1967. The American
Occupational Structure. New York: Wiley
Blau PM, Schwartz JE. 1984. Crosscutting Social
Circles. London: Academic
Blossfeld H-P, Timm A. 1997. Der Einfluß des
Bildungssystems auf den Heiratsmarkt:
ein längsschnittanalyse der Wahl von Heiratspartnerern
im Lebenslaub. Kölner Z.
Soziol. Sozialpsychol. 49:440–76
Bozon M, Heran F. 1989. Finding a spouse: a
survey of how French couples meet. Population
44:91–121
Byrne D. 1971. The Attraction Paradigm.
New York: Academic
Davis FJ. 1991. Who Is Black? One Nation’s
Definition. University Park: Penn. State
Univ. Press
Davis K. 1941. Intermarriage in caste societies.
Am. Anthropol. 43:388–95
Davis K. 1984. Wives and work: consequences
of the sex-role revolution. Popul.
Dev. Rev. 10:397–417
DiMaggio P, Mohr J. 1985. Cultural capital,
educational attainment, and marital selection.
Am. J. Sociol. 90:1231–61
Drachsler J. 1920. Democracy and Assimilation:
The Blending of Immigrant Heritages
in America. New York: Macmillan
Eschbach K. 1995. The enduring and vanishing
American Indian: American Indian
population growth and intermarriage in
1990. Ethn. Racial Stud. 18:89–108
Forsé M, Chauvel L. 1995. L´évolution de
l´homogamie en France. Rev. Fr. Sociol.
36:123–42
Gilbertson GA, Fitzpatrick JP, Lijun Y. 1996.
Hispanic intermarriage in New York City:
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
420 KALMIJN
new evidence from 1991. Int. Migr. Rev.
30:445–59
Glass D. 1954. Social Mobility in Britain.
London: Routledge & Kegan Paul
Glenn ND. 1982. Interreligious marriage in
the United States: patterns and recent
trends. J. Marriage Fam. 44:555–66
Gordon MM. 1964. Assimilation in American
Life. New York: Oxford Univ. Press
Gurak DT, Fitzpatrick JP. 1982. Intermarriage
among Hispanic groups in New York City.
Am. J. Sociol. 87:921–34
Hayes BC. 1991. Religious identification and
marriage patterns in Australia. J. Sci. Study
Relig. 30:469–78
Hayes BC. 1993. Occupational homogamy
within Northern Ireland and the Republic
of Ireland: a loglinear analysis. Int. J. Sociol.
Soc. Policy 13:99–117
Hendrickx J. 1994. The analysis of religious
assortative mating: an application of design
techniques for categorical models.
PhD thesis. Nijmegen Univ., Netherlands.
312 pp.
Hendrickx J, Lammers J, Ultee WC. 1991. Religious
assortative marriage in the Netherlands,
1938–1983. Rev. Relig. Res. 33:
123–45
Hendrickx J, Schreuder O, Ultee WC. 1994.
Die konfessionelle Mischehe in Deutschland
(1901–1986) und den Niederlanden
(1914–1986). Kölner Z. Soziol. Sozialpsychol.
46:619–45
Hout M. 1982. The association between husbands'
and wives' occupations in twoearner
families. Am. J. Sociol. 88:397–409
Hwang SS, Saenz R, Aguirre BE. 1994. Structural
and individual determinants of outmarriage
among Chinese, Filipino, and
Japanese Americans in California. Sociol.
Inq. 64:396–414
Hwang SS, Saenz R, Aguirre BE. 1995. The
SES-selectivity of interracially married
Asians. Int. Migr. Rev. 29:469–91
Jacobs JJ, Furstenberg FF. 1986. Changing
places, conjugal careers, and women’s
marital mobility. Soc. Forces 64:714–32
Jiobu RM. 1988. Ethnicity and Assimilation.
Albany: State Univ. NY Press
Johnson RA. 1980. Religious Assortative
Marriage in the United States. New York:
Academic
Jones FL. 1987. Marriage patterns and the
stratification system: trends in educational
homogamy since the 1930s. Aust. NZ J. Sociol.
23:185–98
Jones FL, Davis P. 1988. Class structuration
and patterns of social closure in Australia
and New Zealand. Sociology 22:271–91
Jones FL, Luijkx R. 1996. Post-war patterns of
intermarriage in Australia: the Mediterranean
experience. Eur. Sociol. Rev. 12:
67–86
Kalmijn M. 1991a. Status homogamy in the
United States. Am. J. Sociol. 97:496–523
Kalmijn M. 1991b. Shifting boundaries:
trends in religious and educational homogamy.
Am. Sociol. Rev. 56:786–800
Kalmijn M. 1993a. Spouse selection among
the children of European immigrants: a
comparison of marriage cohorts in the
1960 census. Int. Migr. Rev. 27:51–78
Kalmijn M. 1993b. Trends in black/white intermarriage.
Soc. Forces 72:119–46
Kalmijn M. 1994. Assortative mating by cultural
and economic occupational status.
Am. J. Sociol. 100:422–52
Kennedy RJR. 1944. Single or triple melting
pot? Intermarriage trends in New Haven,
1870–1940. Am. J. Sociol. 49:331–39
Klein T, Wunder E. 1996. Regionale Disparitaten
und Konfessionswechsel als Ursache
konfessioneller Homogamie. Kölner Z.
Soziol. Sozialpsychol. 48:96–125
Kosmin BA, Goldstein S, Waksberg J, Lerer
N, Keysar A, Scheckner J. 1991. Highlights
of the 1990 National Jewish Population
Survey. New York: Counc. Jewish
Fed.
Lazerwitz B. 1995. Jewish-Christian marriages
and conversions, 1971 and 1990.
Sociol. Relig. 56:433–43
Lee SM, Yamanaka K. 1990. Patterns of Asian
American intermarriage and marital assimilation.
J. Comp. Fam. Stud. 21:
287–305
Lieberson S. 1980. A Piece of the Pie: Blacks
and White Immigrants Since 1880. Berkeley:
Univ. Calif. Press
Lieberson S, Waters MC. 1988. From Many
Strands: Ethnic and Racial Groups in
Contemporary America. New York: Russell
Sage Found.
Mare RD. 1991. Five decades of educational
assortative mating. Am. Sociol. Rev. 56:
15–32
McCutcheon AL. 1988. Denominations and
religious intermarriage: trends among
white Americans in the twentieth century.
Rev. Relig. Res. 29:213–27
Merton RK. 1941. Intermarriage and the social
structure. Psychiatry 4:361–74
Morgan BS. 1981. A contribution to the debate
on homogamy, propinquity, and segregation.
J. Marriage Fam. 43:909–21
Murstein BI. 1976. Who Will Marry Whom:
Theories and Research in Marital Choice.
New York: Springer
Pagnini DL, Morgan PS. 1990. Intermarriage
and social distance among U.S. immigrants
at the turn of the century. Am. J. Sociol.
96:405–32
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.
INTERMARRIAGE AND HOMOGAMY 421
Peach C. 1974. Homogamy, propinquity, and
segregation: a re-evaluation. Am. Sociol.
Rev. 39:636–41
Qian Z. 1997. Breaking the racial barriers:
variations in interracial marriage between
1980 and 1990. Demography 34:263–76
Ramsøy NR. 1966. Assortative mating and the
structure of cities. Am. Sociol. Rev. 31:
773–86
Sandefur GD, McKinnell T. 1986. American
Indian intermarriage. Soc. Sci. Res. 15:
347–71
Schoen R. 1988. Modeling Multigroup Populations.
New York: Plenum
Schoen R, Thomas B. 1989. Intergroup marriage
in Hawaii, 1969–1971 and 1979–
1981. Sociol. Perspect. 32:365–82
Schoen R, Thomas B. 1990. Religious intermarriage
in Switzerland, 1969–72 and
1979–82. Eur. J. Popul. 6:359–76
Schoen R, Wooldredge J. 1989. Marriage
choices in North Carolina and Virginia,
1969–71 and 1979–81. J. Marriage Fam.
51:465–81
Smits J, Ultee W, Lammers J. 1998. Educational
homogamy in 65 countries: the explanation
of differences in openness with
country-level explanatory variables. Am.
Sociol. Rev. In press
Snipp CM. 1989. American Judians: the first
of this land. New York: Russell Sage
Stevens G, Owens D, Schaefer EC. 1990. Education
and attractiveness in marriage
choices. Soc. Psychol. Q. 53:62–70
Sung BL. 1990. Chinese American intermarriage.
J. Comp. Fam. Stud. 21:337–52
Sweet JA, Bumpass LL. 1987. American
Families and Households. New York:
Russell Sage
Taylor PA, Glenn ND. 1976. The utility of
education and attractiveness for females'
status attainment through marriage. Am.
Sociol. Rev. 41:484–98
Ultee WC, Luijkx R. 1990. Educational heterogamy
and father to son occupational mobility
in 23 industrial nations: general societal
openness or compensatory strategies
of reproduction? Eur. Sociol. Rev. 6:
125–49
US Bureau of the Census. 1985. Marital characteristics.
1980 Census of Population,
Subject Report, PC80-2-4C. Washington,
DC: US Gov. Printing Off.
Uunk WJG. 1996. Who marries whom? The
role of social origins, education and high
culture in mate selection of industrial societies
during the twentieth century. PhD
thesis. Nijmegen Univ., Netherlands. 181
pp.
Uunk WJG, Ganzeboom HBG, Róbert P.
1996. Bivariate and multivariate scaled association
models: an application to homogamy
of social origin and education in
Hungary between 1930 and 1979. Qual.
Quant. 30:323–43
Uunk WJG, Kalmijn M. 1996. Wie trouwt met
wie binnen de opleidingselite: de invloed
van studierichting en onderwijsniveau. Sociol.
Gids 43:183–200
Warren BL. 1966. A multiple variable approach
to the assortative mating phenomenon.
Eugen. Q. 13:285–90
Winch RF. 1958. Mate Selection: A Study of
Complementary Needs. New York: Harper
Wirth L, Goldhamer H. 1944. The hybrid and
the problem of miscegenation. In Characteristics
of the American Negro, ed. O
Klineberg, pp. 251–369. New York:
Harper
Wong MG. 1989. A look at intermarriage
among the Chinese in the United States in
1980. Sociol. Perspect. 32:87–107
Annu.Rev.Sociol.1998.24:395-421.Downloadedfromwww.annualreviews.org
AccessprovidedbyMasarykUniversityon09/13/21.Forpersonaluseonly.