ArticlePDF Available

The Changing Role of Education in the Marriage Market: Assortative Marriage in Canada and the United States Since the 1970s

Authors:

Abstract and Figures

Whether or not relative rates of assortative marriage have been rising in the affluent democracies has been subject to considerable dispute. First, we show how the conflicting empirical findings that have fueled the debate are frequently an artifact of alternative methodological strategies for answering the question. Then, drawing on comparable census data for Canada and the United States, we examine trends in educational homogamy and intermarriage with log-linear models for all marriages among young adults under 35 over three decades. Our results show that educational homogamy, the tendency of like to marry like, has unambiguously risen in both countries since the 1970s, with no sign of the U-turn in levels of intermarriage reported in some earlier comparative studies. Rising levels of marital homogamy were the result of declining intermarriage at both ends of the educational distribution. However, while trends for men and women were quite similar in Canada, they differed significantly in the United States. The overall rise in marital homogamy In the United States was partially offset by an increased tendency of women with some college education to marry `down¿ the educational hierarchy. In Canada, the only sign of abatement in the trend toward greater educational homogamy was a slight increase in intermarriage among university-educated men and women during the 1990s.
Content may be subject to copyright.
Analytical Studies Branch Research Paper Series
The Changing Role of Education in the Marriage
Market: Assortative Marriage in Canada
by Feng Hou and John Myles
Business and Labour Market Analysis
24th floor, R.H. Coats Building, 100 Tunney’s Pasture Driveway, Ottawa, K1A 0T6
Telephone: 1-800-263-1136
T
Catalogue no. 11F0019MIE — No. 299
ISSN: 1205-9153
ISBN: 978-0-662-46028-2
Research Paper
and the United States Since the 1970s
The Changing Role of Education in the Marriage Market:
Assortative Marriage in Canada and the United States Since
the 1970s
by Feng Hou and John Myles*
11F0019 No. 299
ISSN: 1205-9153
ISBN: 978-0-662-46028-2
Statistics Canada
Business and Labour Market Analysis Division
24-H, R.H. Coats Building, 100 Tunney’s Pasture Driveway, Ottawa K1A 0T6
* University of Toronto and Statistics Canada
How to obtain more information:
National inquiries line: 1-800-263-1136
E-Mail inquiries: infostats@statcan.ca
May 2007
We would like to thank Dr. Dana Hamplova for important comments on an earlier version of this paper.
Published by authority of the Minister responsible for Statistics Canada
© Minister of Industry, 2007
All rights reserved. The content of this electronic publication may be reproduced, in whole or in part, and by any
means, without further permission from Statistics Canada, subject to the following conditions: that it be done solely
for the purposes of private study, research, criticism, review or newspaper summary, and/or for non-commercial
purposes; and that Statistics Canada be fully acknowledged as follows: Source (or “Adapted from,” if appropriate):
Statistics Canada, year of publication, name of product, catalogue number, volume and issue numbers, reference
period and page(s). Otherwise, no part of this publication may be reproduced, stored in a retrieval system or
transmitted in any form, by any means—electronic, mechanical or photocopy—or for any purposes without prior
written permission of Licensing Services, Client Services Division, Statistics Canada, Ottawa, Ontario, Canada
K1A 0T6.
La version française de cette publication est disponible (no 11F0019MIF au catalogue, no 299).
Note of appreciation:
Canada owes the success of its statistical system to a long-standing partnership between Statistics Canada, the
citizens of Canada, its businesses, governments and other institutions. Accurate and timely statistical information
could not be produced without their continued cooperation and goodwill.
Table of contents
Executive summary......................................................................................................................... 5
1. Introduction............................................................................................................................ 7
2. Literature review: Has educational homogamy changed?..................................................... 8
3. Now you see it, now you don’t: Measuring changes in educational homogamy ................ 11
3.1 Education groupings ................................................................................................................. 11
3.2. First marriage versus marriage stock and real versus synthetic cohorts ...................................13
3.3 Measuring heterogeneous and overall changes in homogamy with log-linear models............. 14
4. Data sources......................................................................................................................... 18
5. Results.................................................................................................................................. 19
5.1. Changes in absolute rates: Educational homogamy among the married.................................. 19
5.2 Changes in relative rates ........................................................................................................... 24
6. Conclusion ............................................................................................................................... 28
Appendix 1 Trends in educational homogamy among newly formed first marriages................. 33
References..................................................................................................................................... 38
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 3 -
Abstract
Whether or not relative rates of assortative marriage have been rising in the affluent democracies
has been subject to considerable dispute. First, we show how the conflicting empirical findings
that have fueled the debate are frequently an artifact of alternative methodological strategies for
answering the question. Then, drawing on comparable census data for Canada and the United
States, we examine trends in educational homogamy and intermarriage with log-linear models
for all marriages among young adults under 35 over three decades. Our results show that
educational homogamy, the tendency of like to marry like, has unambiguously risen in both
countries since the 1970s, with no sign of the U-turn in levels of intermarriage reported in some
earlier comparative studies. Rising levels of marital homogamy were the result of declining
intermarriage at both ends of the educational distribution. However, while trends for men and
women were quite similar in Canada, they differed significantly in the United States. The overall
rise in marital homogamy in the United States was partially offset by an increased tendency of
women with some college education to marry ‘down’ the educational hierarchy. In Canada, the
only sign of abatement in the trend toward greater educational homogamy was a slight increase
in intermarriage among university-educated men and women during the 1990s.
Keywords: educational homogamy, intermarriage, log-linear models.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 4 -
Executive summary
Educational homogamy—the tendency of men and women with the same level of education to be
married to one another—has important implications for social stratification, family income
inequality, and intergenerational mobility. While absolute levels of educational homogamy have
unambiguously been rising, whether or not relative rates of educational homogamy, net of
changes in the educational composition of husbands and wives, have also been rising has been a
contested issue in the literature. Changes in relative rates are indicative of whether the function
of education in mate selection is strengthening or weakening.
In this paper our aim is to answer three questions. First, what is the overall trend in the relative
rate of educational marital homogamy in Canada and the United States over the three decades
from the beginning of the 1970s to the turn of the century? Second, are the highly educated more
likely now than in the past to marry within their own education level or to marry down, and are
the less educated more likely than in the past to marry within their own educational class or to
marry up? Third, do women and men experience different trends in educational homogamy given
that the rapid increase in women’s educational attainment relative to men’s implies declining
opportunities for women to marry better-educated men but rising opportunities for men to marry
more-educated women?
Our analysis of marriages among young adults shows that the overall level of both absolute and
relative rates of educational homogamy have unambiguously increased in both countries over the
three decades. In Canada, 54% of couples had the same level of education in 2001, up from 42%
in 1971. In the United States, some 55% of marriages consisted of couples with the same level of
education in 2000, up from 49% in 1970. These overall trends have been driven mainly by
changes in the association of husbands’ and wives’ education rather than by changes in the
relative supply of more- and less-educated partners. In Canada, the change in association, net of
changes in the distributions of wives’ and husbands’ education levels, accounts for almost 10
percentage points of the 12-percentage-point increase in educational homogamy from 1971 to
2001. In the United States, the change in association accounts for 4 percentage points of the 6-
percentage-point increase.
In both countries, intermarriage across education levels occurs primarily between adjacent
education levels, and intermarriage across more than one education level is relatively rare.
Therefore, changes in intermarriage between adjacent education levels dominate the overall
trends in educational homogamy.
Rising educational homogamy has been driven mainly by changes at the top and at the bottom of
the educational hierarchy.
Declining intermarriage between those with university degrees and those with less education was
a major factor in both countries. In Canada, the relative rate of intermarriage between the
university educated and those with only some post-secondary education fell by 38% and in the
United States, by 45%. Whereas Canadian trends were quite similar for men and women, the
U.S. decline was almost entirely driven by declining intermarriage among university-educated
men.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 5 -
Similarly, the odds of intermarriage between high-school graduates and those with less than
high-school completion fell by 30% in the United States and by 58% in Canada. In Canada, this
was a common pattern among both men and women while in the United States the trend was
mainly confined to men.
There were two important exceptions to the general trend towards rising homogamy. The first
was an increase in intermarriage between women with some post-secondary education and men
with high-school graduation in the United States. The second was a small increase in
intermarriage in the 1990s among male and female university graduates in Canada and among
female university graduates in the United States, suggesting that levels of educational
homogamy among the better educated may be stabilizing.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 6 -
1. Introduction
The inequalities in the distribution of market incomes among families has become markedly
more important in most affluent nations over the past quarter century (Kenworthy, 2004), and
rising rates of marital homogamy—the tendency of men and women with high earnings to be
married to one another—is increasingly cited as one of the major factors underlying this trend
(Esping-Andersen, 2007; Kenworthy, 2004). As women’s labour force participation and
education levels have risen, the correlation between husbands’ and wives’ earnings has also risen
(Hyslop, 2001; Fortin and Schirle, 2006), reflecting the fact that well-educated men and women
tend to marry one another and form families with high earnings and few risks of unemployment.
Less well-educated couples have lower wages, and both partners are far more likely to
experience periods without work.
While absolute levels of educational homogamy have unambiguously been rising, whether or not
relative rates of educational homogamy, net of changes in the educational composition of
husbands and wives, have also been rising has been a contested issue (see Table 1).1 Changes in
absolute rates of educational endogamy/exogamy are of interest, when the question concerns the
effects of educational homogamy on outcomes such as family earnings inequality (Goldthorpe,
1987). However, if changes in marital patterns are entirely exogenous to the mate selection
process (i.e., entirely a result of changes in the marginal education distributions of men and
women), the nature of the question changes entirely. If the rising share of university-educated
men married to university-educated women is mainly due to the rising share of women with
university degrees, for example, the only task that remains is to explain the rise in women’s
educational attainment. Changes in relative rates, in contrast, are indicative of whether the
function of education in mate selection is strengthening or weakening.
In this paper, our aim is to answer three questions. First, what is the overall trend in the relative
rate of educational marital homogamy in Canada and the United States over the three decades
from the beginning of the 1970s to the turn of the century? Since the average trend may conceal
offsetting changes at different levels in the educational hierarchy, however, we also examine
trends among the main subcomponents of the educational hierarchy.2 Are the highly educated
more likely now than in the past to marry within their own educational class or to marry down?
1. The analysis of educational homogamy has been modelled more or less self-consciously along the lines of
traditional studies of intergenerational occupational inheritance captured by the cross-classification of a
hierarchically ordered distribution of occupations held by a generation of adults with that of the parental
generation (usually fathers). The questions of interest in social mobility studies concern whether or not societies
are more ‘fluid’ or ‘open’ than in the past, whether some class boundaries are more permeable than others, or
whether some societies are more open than others. From the outset, researchers recognized that, where the
marginal distributions of such tables were not identical, some degree of mobility would be displayed of
necessity; it is simply not arithmetically possible for all cases in the table to fall in the diagonal (Goldthorpe,
2000). Similarly, as some occupations decline and others expand, the absolute amount of mobility will also
change. To address this issue, mobility researchers turned to odds ratios and log-linear analysis to distinguish
between absolute mobility rates (the total rates shown in a percentage table) and relative mobility rates, the
odds ratios that define the association between occupational origins and destinations, net of changes or
differences in the marginal distribution of occupations. Both the methods (odds ratios, log-linear analysis) and
metaphors (absolute versus relative rates, ‘openness,’ ‘permeability’) of social mobility studies have become
standard fare in analyses of marital homogamy.
2. Since intermarriage occurs primarily between adjacent education levels, a decline in intermarriage at the top of
the educational hierarchy does not necessarily correspond to a decline in intermarriage at the bottom of the
educational distribution.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 7 -
Are the less educated more likely than in the past to marry within their own educational class or
to marry up? Finally, we consider the gender-specific features of these trends. Since the rapid
increase in women’s educational attainment relative to men’ implies declining opportunities for
women to marry better-educated men but rising opportunities for men to marry more-educated
women, we consider how these trends differ for men and women.
Compared to previous studies, our approach is most similar in spirit to that of Schwartz and
Mare (2005) who address the first two of our questions in the context of the United States but do
not test for gender asymmetries in homogamy/intermarriage trends. Paradoxically, we find that
the assumption of gender symmetry provides a reasonable fit for the Canadian but not the U.S.
data, a finding that changes our understanding of the overall trends. Though not a major focus of
our analysis, we also cast some doubt on their claim that U.S. homogamy rates fell over the
period from the 1940s to the 1960s.
The paper is organized as follows. Section 2 provides an overview of recent debates on relative
rates of educational homogamy and the empirical results that have stimulated these discussions.
In Section 3, we sort out a series of methodological issues related to the measurement of
educational homogamy and present the rationale for the methodological strategies adopted here.
Section 4 describes our data and Section 5 presents our results. Section 6 concludes.
2. Literature review: Has educational homogamy changed?
As Halpin and Chan (2003: 473) observe, “education has always been a factor in the choice of
mate selection in modern societies.” Schools and universities provide contexts in which young
people meet one another and produce common understandings of desirable lifestyles and cultural
tastes. There are many reasons why post-industrial societies are likely to reinforce such patterns.
Rising levels of post-secondary education among women in recent decades have augmented the
opportunities for well-educated men to meet well-educated women. Since workplaces are
typically characterized by employees with similar levels of education but growing gender
diversity, they too provide new sites for meeting potential spouses with similar educational
qualifications (Oppenheimer, 1994). Moreover, the rise in women’s labour force participation,
combined with a growing earnings gap between more- and less-educated workers, has arguably
increased the economic incentives for men (and women) to choose a highly educated partner
(Kalmijn, 1998).
Studies specifically related to the United States and Canada more or less agree on a rising level
of educational homogamy from the early in the 20th century through the 1960s (but see Schwartz
and Mare, 2005). Studies of trends since then, however, are far from consistent (see Table 1).
Kalmijn (1991a, 1991b) finds that educational homogamy increased steadily over the period
from the 1930s to the 1980s in the United States. Ultee and Luijkx (1990) find a slight increase
in educational homogamy in the United States, and a slight decrease in Canada up to the end of
the 1970s, though in both cases the changes are not statistically significant. Qian and Preston
(1993) and Qian (1998) suggest that educational homogamy increased in the 1980s in the United
States. Schoen and Cheng (2006) find a rise in educational homogamy in North Carolina and
Wisconsin in the 1970s and 1980s. Raymo and Xie (2000) find that educational homogamy was
stable in the United States from the early 1970s to the late 1980s. Mare (1991) concluded that
educational homogamy increased from the 1930s to the 1970s but then stabilized or even
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 8 -
declined in the 1980s. Schwartz and Mare (2005), in contrast, conclude that homogamy in the
United States declined from 1940 to 1960 but then rose from the 1960s to the turn of the century.
Table 1 Overview of major studies on changes in educational homogamy related to United States
and Canada, 1990 to 2006
Studies Educational
groupings
Target
population
Cohorts of
change
Modelling
strategies
Data sources Sample size Conclusion
Ultee and Luijkx,
1990
4 levels. United
States: < Grade
9, 9 to 11, 12,
college. Canada:
< Grade 9, 9 to
13, some
university,
university degree
United States: all
marriages
involved males
aged 20 to 64.
Canada: all
marriages
Overlapping
marriage cohorts
Log-linear
models: step
(distance)
parameters
1971 and 1981
Canadian
Census; 1962
and 1973 U.S.
CPS1, 1982 to
1985 U.S. GSS2
Re-scaled to
1,000 marriages
per
period/country
Slight fall in
homogamy in
Canada, rise in
the United States
Kalmijn, 1991 5 levels: < Grade
9, 9 to 11, 12,
college 1 to 3,
college 4+
Couples married
within 10 years
Real marriage
cohorts
Log-linear
models: distance
and diagonal
parameters
1962 and 1973
OCG3
2,400 to 5,000
couples per
period
Rise in
homogamy
Kalmijn, 1991 4 levels: <
Grade 12, 12,
college 1 to 3,
college 4+
Married couples
in their first
marriage
Overlapping
marriage cohorts
Log-linear
models: quasi-
symmetry
parameters
1955 GAF4, 1965
NFS5, 1972 to
1989 GSS
450 to 4,100 per
decade
Rise in
homogamy
Mare, 1991 5 levels: <
Grade 10, 10 to
11, 12, college 1
to 3, college 4+
Couples aged 16
to 34, first
marriage within a
year
Real marriage
cohorts
Log-linear
models:
crossings
parameters
1940, 1960 to
1980 U.S.
Census, 1985 to
1987 CPS
4,000 to 13,100
couples per
period
Rise in
homogamy until
the 1970s
Qian and
Preston, 1993
3 levels: <
Grade 12, 12,
college
Newly married
couples with
women aged 18
to 44
Real marriage
cohorts
Harmonic mean
function
1973, 1980 and
1988 CPS
< 1,000 couples
per year
Rise in
homogamy
Qian, 1998 4 levels: <
Grade 12, 12,
college 1 to 3,
college 4+
First married
within 5 years,
aged 15 to 39
Real marriage
cohorts
Log-rate models:
asymmetry and
crossing models
1970, 1980
Census, 1988,
1990 and 1992
CPS
9,600 to 146,200
couples per
period
Rise in
homogamy
Raymo and Xie,
2000
4 levels: <
Grade 12, 12,
college 1 to 3,
college 4+
Newly married
couples aged 18
to 34
Real marriage
cohorts
Log-linear
models: diagonal
and distance
parameters
1970 U.S.
Census, 1985 to
1987 CPS
1,300 to 2,000
couples per
period
Slight fall in
homogamy
Smits, Ultee and
Lammers, 2000
4 levels: <
Grade 12, 12,
college 1 to 4,
college 4+
Married couples
aged 20 to 52
Cross-sectional
age groups
Log-linear
models: step
(distance)
parameters
1976 Canadian
Census, 1980
U.S. Census
About 3,500 per
age
cohort/country
Significant rise
in both Canada
and the United
States
Smits, 2003 2 levels: < Grade
12, 12+
Married couples
aged 18 to 52
Cross-sectional
age groups
Log-linear
models: odds
ratio for 2-by-2
table
1976 Canadian
Census, 1980
U.S. Census
About 15,000
each age
cohort/country
Slight fall in
homogamy in
both countries
Schwartz and
Mare, 2005
5 levels: < Grade
10, 10 to 11, 12,
13 to 15, > = 16
Newly married
and prevailing
marriages aged
18 to 40
Real and
overlapping
marriage cohorts
Log-linear
models:
homogamy and
crossings
parameters
1940 to 2000
U.S. Census and
1962 to 2003
CPS
Large total
sample
Fall from 1940 to
1960, rise
afterwards
Schoen and
Cheng, 2006
4 levels: < Grade
12, 12, 13 to 15,
16+
Recent marriages Real marriage
cohorts
Harmonic mean
function
1970-to-1990
marriage records
in two U.S. states
Over 50,000 in
each state
Rise in
homogamy
1. Current Population Survey.
2. General Social Survey.
3. Occupational Change in Generation.
4. Growth of American Families.
5. National Fertility Survey.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 9 -
Like Mare (1991), Smits and his associates (Smits, Ultee and Lammers, 1998; Smits, Ultee and
Lammers, 2000; and Smits, 2003) report an inverted, yet asymmetric U-curve trend in
educational homogamy across a number of nations. They attribute the inverted U-shape pattern
to changing preferences associated with modernization. They argue that educational homogamy
first rises with industrialization as education becomes increasingly important in deciding
individuals’ socioeconomic status. People with higher levels of education have greater potential
to maximize family socioeconomic status through marriage and thus are more attractive in the
marriage market (the status attainment thesis). However, in the later stages of industrialization,
continued modernization favours greater societal ‘openness’ and individualization as people
become increasingly able to afford the luxury of mate selection on the basis of other desirable
criteria. As highly educated women become increasingly able to support themselves, for
example, the incentives to seek out a ‘good provider’ as a mate may well decline. A growing
literature suggests that women’s marital preferences have changed as their position in the labour
market continues to improve while the opposite is occurring among young men (Duncan, 2003;
Oppenheimer, 1997; Sweeney, 2002; and Sweeney and Cancian, 2004). Equipped with potential
or realized economic resources, a highly educated woman may be willing and able “to marry a
man who is unlikely to be a great provider but who is highly desirable in other respects.”
(Oppenheimer, 1994: 315).
Mare’s (1991) ‘life course’ thesis attributes the trend reversal to changes in the marital
opportunity structure. He suggests that people who are married while they are at school or
shortly after leaving school are more likely to have similar levels of education. But as the time
gap between leaving school and age of marriage rises, the pool of potential spouses becomes
increasingly heterogeneous and is likely to lead to a decline in homogamy. From the early part of
the 20th century until the 1970s, the age gap between leaving school and marriage narrowed but
has been rising since then. According to the life-course argument, the U-shape trend in the time-
gap would lead to a rise in educational homogamy until the 1970s but a decline or stabilization
thereafter (Mare 1991; Halpin and Chan 2003).
The inconsistent findings among existing studies are in part attributable to large differences in
study populations and methodologies taken up in the following section. However, they also
reflect differences in analytical focus. Some studies (e.g., Kalmijn, 1991b; Raymo and Xie,
2000; Smits, Ultee and Lammers, 2000; and Smits, 2003) examine the overall trends without
attending to trends among men and women at different education levels. The issue here is that
the overall trend may obscure large changes, even in opposite directions, among the underlying
components (Wong, 2003). Still other studies draw their conclusions from changes in one or
other of the underlying components. Mare (1991), for example, focused mainly on changes in the
difficulty of intermarriage between adjacent education levels. When it becomes more difficult for
marriage to cross one educational boundary but easier to cross another barrier, it is not clear
what the overall trend is. Similarly, a rise in homogamy between highly educated, more affluent,
individuals would support the status attainment hypothesis over the ‘modernization’ thesis but
the preferences of the highly educated do not necessarily dominate the overall trend (Smits,
2003). In this respect, the recent U.S. study by Schwartz and Mare (2005) provides an important
corrective to earlier research by considering both overall change and trends by education level.
Disentangling the various components of educational homogamy requires a careful
reconsideration of the measurement and modelling strategies used in the literature. We turn to
these methodological issues in the next section.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 10 -
3. Now you see it, now you don’t: Measuring changes in educational homogamy 3
In this section, we organize our discussion around three issues: (1) the choice of educational
grouping; (2) first marriage versus marriage stock, and real marriage cohorts versus synthetic
cohorts; and (3) measuring overall and heterogeneous changes in homogamy with log-linear
models. From the discussion of the potential impact of these differences on detecting changes in
educational homogamy, we can either choose the most appropriate approach or compare results
from different approaches in our subsequent analyses.
3.1 Education groupings
Wong (2003) shows that arbitrary and inconsistent classification of education levels is a vital
weakness in studies on temporal trends and cross-country differences in educational homogamy.
He demonstrates that combining Mare’s (1991) five categories of education into four categories
results in differential loss of association between wife’s and husband’s education at various time
periods. Aggregation to four categories exaggerates the degree of homogamy at earlier periods
and deflates the upward trend. Similarly, the conclusions drawn by Smits, Ultee and Lammers
(2000) and by Smits (2003) are quite different, although the only difference between the two
studies is that the former uses four educational categories and the latter, two.
The method of aggregation also affects results. Aggregation across education levels at the lower
end of the educational distribution eliminates heterogeneity in education levels prevalent in
earlier historical periods and inflates the homogamy estimates for the beginning of the time
series. Conversely, aggregation at the top eliminates heterogeneity prevalent in later periods,
inflating estimates of homogamy at the end of the period. To illustrate, Table 2 shows the effects
of aggregating changes in the absolute rate of educational homogamy among couples where both
are aged under 35 in the United States (1940 to 2000) and Canada (1971 to 2001). The first
column shows the change in homogamy rates when measured with nine educational classes.
Educational groupings 9, 7, 6, 5, 4a, and 3a show the effect of aggregation into fewer categories
by combining the lowest with the next-lowest educational categories. When aggregation is done
from the bottom up, the time trend towards rising levels of homogamy gradually disappears and
is then reversed. For example, educational homogamy in the United States increases from 1940
to 1960 with nine educational categories but decreases with more aggregate groupings. In 1940,
about one half of married young women and 43% of married young men had less than Grade 10
education, and much of the variation in the educational distribution was among those with few or
no years of schooling, those that had completed Grade 8 and those with some high school. By
1960, in contrast, variation in education levels at the bottom of the distribution had all but
disappeared. Hence, by grouping together all those with less than Grade 10, Schwartz and Mare
(2005), in their estimates for 1940, conceal the actual diversity in education levels and inflate the
estimated level of homogamy relative to later periods.
3. This title emulates that of Wong (2003).
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 11 -
Table 2 Couples with same education levels among all couples aged 34 or under, United
States and Canada
Groupings of educational categories
9 7 6 5 4a 4b 3a 3b
percent
United States
1940 36.4 44.1 47.3 51.0 66.6 53.3 86.5 69.0
1960 38.5 40.3 41.5 44.9 53.6 50.0 76.9 58.6
1970 45.0 46.0 46.4 48.8 53.5 55.9 71.9 60.5
1980 48.9 49.3 49.5 51.0 53.2 60.8 65.7 63.0
1990 … 51.9 52.0 52.9 53.9 63.7 62.6 64.8
2000 … 53.9 54.1 54.7 55.8 67.4 62.7 68.6
Canada
1971 … 35.4 36.3 41.8 54.0 46.3 73.5 58.4
1981 … 41.7 41.9 44.4 48.2 52.5 59.4 56.2
1991 … 49.0 49.1 50.6 52.1 59.8 60.5 61.4
2001 … 54.7 54.7 55.2 55.9 70.8 59.6 71.4
… not applicable
Notes: The various groupings of educational categories are defined as the following: 9: Grades 0 to 4, 5 and 6, 7, 8, 9, 10 and 11, 12, college 1 to
3, college 4 and over; 7: Grades 0 to 4, 5 to 8, 10 and 11, 12, college 1 to 3, college 4 and over; 6: less than Grade 9, Grades 9, 10 and 11,
12, college 1 to 3, college 4 and over; 5: less than Grade 9, Grades 9 to 11, 12, college 1 to 3, college 4 and over; 4a: less than Grade 12,
Grade 12, college 1 to 3, college 4 and over; 4b: less than Grade 9, Grades 9 to 11, 12, and college; 3a: less than Grade 12, Grade 12,
college 1 to 3, college 4 and over; 3b: less than Grade 12, Grade 12, college. For Canada, Grade 12 is replaced by high-school graduation;
college 1 to 3 is replaced by some post-secondary education without a university degree. Results in this table are derived from data from
the sources below.
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample
microdata files; University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Similarly, from 1970 on, educational homogamy in both the United States and Canada
unambiguously rises with 7, 6, and 5 educational categories, changes little with the 4a groupings,
but declines with the 3a groupings. In 1970, the highest level of education for most young people
was high-school graduation, and crossing barriers of intermarriage for the majority of people
occurred at or below high-school graduation. Thus, a fine distinction across education levels at
the lower end is critical.
By 2000, in contrast, most young people had at least some post-secondary education, and
crossing barriers occurred mainly at levels beyond high-school graduation. Hence, aggregating
from the top down eliminates heterogeneity in educational attainment prevalent in later but not
earlier periods, producing inflated levels of homogamy at the end of the series. The differences
between grouping 5 and 4b and between 4a and 3b reflect the effects of aggregation at the upper
end by combining “some college” with “college graduation.” Thus, for later periods, a fine
distinction across education levels at the upper end is important.
A suitable grouping of education levels for examining temporal changes should adequately
reflect the main sources of educational heterogeneity at both the earlier and later periods. Ideally,
the more detailed grouping the better. However, too much detail creates many empty cells in the
cross-tabulation of wives’ and husbands’ education and complicates modelling. A practical
solution is to choose a grouping that closely resembles more detailed groupings in revealing the
trends in homogamy. In our analysis we choose the five-level classification that distinguishes
between elementary school only, some high school, high-school graduation, some post-
secondary (some college in the United States) education, and university (college in the United
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 12 -
States) graduation. The five-category grouping reveals the same trend as more detailed
groupings and captures the main sources of educational heterogeneity at both the beginning and
the end of the period.
3.2 First marriage versus marriage stock, and real versus synthetic cohorts
In previous studies on trends in educational homogamy, some researchers favour newly formed
first marriages while others use marriage stock (see Table 1). From newly formed first
marriages, husbands’ and wives’ educational attainment can be measured at the time close to
marriage formation. This advantage makes them appropriate subjects for studying the role of
education among those entering marriage for the first time. In comparison, the level of
educational homogamy among marriage stock (prevailing marriages) reflects the combined
effects of assortative entry into the first marriage, assortative dissolution of marriage, assortative
entry into subsequent marriages, and the tendency that partners grow alike in educational
attainment after marriage (Gelissen, 2004; Rogers, 2004). Kalmijn (1991b) shows that the
percentage of homogamous marriages rises as a marriage cohort ages. Given the large
prevalence of union dissolution and remarriage in contemporary western societies, focusing on
first marriages will not reveal the overall picture of educational homogamy among prevailing
marriages.
A related issue concerns the use of real versus synthetic marriage cohorts. Newly formed first
marriages from repeated cross-sections with sufficient time intervals represent distinct marriage
cohorts. This may not be the case for prevailing marriages. In the studies by Smits, Ultee and
Lammers (2000) and Smits (2003), differences between younger and older couples from one
cross-sectional data set are used to infer changes in educational homogamy. The obvious
problem with such an approach is that older couples have stayed much longer in marriage than
younger ones. If homogamy increases or decreases with the length of marriage, then differences
between younger and older couples at least partially capture these attrition effects. When
prevailing marriages from multiple cross-sectional data are compared, problems occur due to
overlapping of marriage cohorts. As shown in Table 3, the trend toward rising educational
homogamy among younger couples is clear in both countries. The trend is not as clear among
older couples or among all marriages.4
4. Older and younger couples are also differentially sensitive to the grouping of education levels. In 1971, young
Canadian couples (both the husband and wife are aged from 15 to 34) had a lower level of educational
homogamy than older couples. This is primarily because most of the older couples were married before 1960
and the five educational categories did not adequately reflect the important barriers to intermarriage in the
earlier decades. When we split those with elementary education into three categories (no education, 1 to 4
years, 5 to 8 years), older couples had a lower level of educational homogamy than the younger group in 1971
(results not shown here).
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 13 -
Table 3 Couples with same educational levels by age of wives and husbands, based on five
educational groups, United States and Canada
Age of husbands and wives
Both from 15 to 34 Either over 34 All marriages
percent
United States
1970 48.8 46.2 46.9
1980 51.0 46.8 48.0
1990 52.9 47.3 48.5
2000 54.7 49.8 50.5
Canada
1971 41.8 50.0 47.6
1981 44.4 45.2 45.0
1991 50.6 49.7 49.8
2001 55.2 50.8 51.2
Notes: The five educational groups in the United States are as follows: less than Grade 9, Grades 9 to 11, Grade 12, college 1 to 3, college 4 and
over. In Canada, they are as follows: less than Grade 9, some high school, high-school graduation, some post-secondary, undergraduate
degree and over. Results in this table are derived from data from the sources below.
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample
microdata files; University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
In this study, our main conclusions are based on analyses of prevailing marriages among young
couples in which both partners are under 35 for two reasons.5 First, our effort to analyze first
marriages (reported in Appendix A) suffers from problems of data sparsity (small sample).
Second, prevailing marriages among younger couples provides a more complete picture of trends
since they also take account of assortative trends in marital dissolution and remarriage. By
focusing on younger couples, we also can minimize the impact of cohort overlap.
3.3 Measuring heterogeneous and overall changes in homogamy with log-linear models
Most studies of educational homogamy rely on log-linear modeling of the contingency table of
wives’ and husbands’ education levels. For a two-way table, the log-linear model takes the
general form
Log (FijWH) = λ0 + λiW + λjH + λijWH ,
where, FijWH refers to the expected frequency of the (i,j) cell consisting of wives with education
level i and husbands with education level j. Both i and j ranges from 1 to k. λiW captures the
marginal effect of wives’ educational distribution, λjH the marginal effect of husbands’ education,
and λijWH captures the association between wives’ and husbands’ education.
The saturated model includes all the linearly independent effects: k1 factors for wives’
marginal effect, k1 for husbands, and (k1)( k1) for their interactions. The saturated model fits
the data perfectly but does not have any extra degree of freedom for testing specific hypotheses.
Accordingly, researchers use more parsimonious forms of the interaction terms. Previous studies
on educational homogamy often use some variations or combinations of the following forms:
quasi-independence (diagonal) parameters, crossings parameters, distance parameters, and quasi-
5. We also conducted a separate analysis for couples in which only the wife’s age is confined to be under 35, to
make sure our results are robust to sample selection given that men and women differ in the mean marriage
ages. Results for couples in which the husband’s age is not restricted are reported in footnotes 13 and 14.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 14 -
symmetry parameters (see Table 1). Specifications of these parameters are well documented in
the literature (e.g., Haberman, 1979; Hout, 1983).
Using five educational categories as an example, we show in Table 4 how the above four types
of parameters correspond to the log odds ratio of intermarriage (or homogamy when assigned the
opposite sign) between a given pair of educational categories.6 From this table, it is clear that the
four commonly used log-linear models differ only in their assumptions about the relationships in
log odds ratios of intermarriage among different pairs of education levels.7 The choice of models
is typically determined empirically by goodness-of-fit statistics. As shown in Tables 12 and 13,
when applied to all married couples aged under 35, the quasi-independence model fits the data
poorly. Among the remaining three models, the quasi-symmetry model always has the smallest
log-likelihood ratio chi-square statistic (L2), and often the smallest BIC (Bayesian Information
Criterion) value which penalizes less-parsimonious models. Accordingly, we report results based
on the quasi-symmetry model since it is typically the best-fitting model when, as in our study,
the sample size is sufficiently large. Furthermore, a given parameter from the quasi-symmetry
model directly corresponds to a log odds ratio of intermarriage between two education levels as
shown in Table 4.
6. We want to stress that it is crucial to interpret the above parameters in terms of log odds ratio of intermarriage
(homogamy) between any given pair of educational categories. This interpretation will help us interpret
changes in these parameters over time. For instance, the log odds ratio of intermarriage between high-school
(level 3) and university graduates (level 5) based on a quasi-independence model can be expressed as
)(log 53
55
53
33
35
qq
F
F
F
F
λλ
+=
×.
Similarly, the log odds ratio of intermarriage for the same combination of education levels becomes 2(λc3 + λc4)
in a crossings model, 2λd2 in a distance model, and 2λs53 in a quasi-symmetry model. In the situation when some
of these models are combined in the estimation, the log odds ratios can be derived in a similar fashion.
7. The quasi-independence model assumes a general tendency of intermarriage for each education level (λqi for
level i), e.g., university graduates have a higher tendency to marry other university graduates than high-school
graduates to marry other high-school graduates. In this model, the log odds ratios of intermarriage between
education level i and j is the sum of the λqi and λqj. The crossings model assumes a unique barrier to
intermarriage for crossing two adjacent education levels and the log odds ratio of intermarriage between two
education levels depends on the selection and number of barriers crossed. The distance model assumes that the
difficulty of intermarriage is the same among pairs of education levels that have the same relative distance. For
instance, the relative distance between less than high school (level 1) and high-school graduation (level 3) is
considered the same as the distance between high-school graduation (level 3) and university graduation
(level 5). In contrast, the quasi-symmetry model does not assume any relationship among pairs of education
levels. Since the quasi-symmetry model imposes few restrictions on the parameter estimates, it is less
parsimonious than other models but often fits the data better.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 15 -
We further add a hypergamy parameter to examine the possibility that, with a given education
level, women are more likely than men to marry up (or down) the educational hierarchy.8 Rather
than assume women and men have the same tendency to marry up (or down), the hypergamy
parameter essentially allows us to test for asymmetry in male–female rates of intermarriage.
In sum, we use quasi-symmetry parameters and a hypergamy parameter to capture the
association between wives’ and husbands’ education. Building on this base model, we examine
how such an association has changed over time in terms of overall trends, and trends specific to
the highly educated and less well educated, and trends specific to women and men.
To answer our first research question regarding trends in the overall rate of educational marital
homogamy, we can choose from various modelling strategies including Yamaguchi’s (1987)
uniform layer effect model, Xie’s (1992) log-multiplicative layer effect model, and the
regression-type approach proposed by Goodman and Hout (1998). We use the log-multiplicative
layer effect model since it allows greater parsimony than the regression approach and defines
layer (time period) effects by directly comparing the pre-defined pattern of the two-way
association between husbands and wives’ education (in our case, the quasi-symmetry pattern
plus hypergamy), which is not the case in the uniform layer effect model.9
8. In the log-linear model, the hypergamy variable is coded simply as 1 when a woman marries a better-educated
man and 0 when a woman marries a man with the same or less education.
9. We estimated both the log-multiplicative layer model and the uniform layer effect model. The results showed
similar trends.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 16 -
Table 4 Parameters and their interpretations in some log-linear models of educational
assortative marriage
1. Quasi-independence model
Wifes
1
(1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) λ
q1
000 0
(2) 0 λ
q2
00 0 -(λq1+λq2)
(3) 0 0 λ
q3
00-(λ
q1
+λ
q3)
-(λ
q2
+λ
q3)
(4) 0 0 0 λ
q4
0-(λq1+λq4) -(λ
q2
+λ
q4)
-(λ
q3
+λ
q4)
(5) 0 0 0 0 λ
q5
-(λq1+λq5) -(λ
q2
+λ
q5)
-(λ
q3
+λ
q5)
-(λ
q4
+λ
q5)
2. Crossings model
Wifes
1
(1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0 λ
c1
λ
c1+
λ
c2
λ
c1+
λ
c2+
λ
c3
λ
c1+
λ
c2+
λ
c3+
λ
c4
(2) λ
c1
0λ
c2
β
c2+
β
c3
λ
c2+
λ
c3
+λ
c4
2λ
c1
(3) λ
c1+
λ
c2
λ
c2
0λ
c3
λ
c3+
λ
c4
2(λ
c1+
λ
c2
)2λ
c2
(4) λ
c1+
λ
c2+
λ
c3
λ
c2+
λ
c3
λ
c3
0λ
c4
2(λ
c1+
λ
c2+
λ
c3
)2(λ
c2+
λ
c3
)2λ
c3
(5) λ
c1+
λ
c2+
λ
c3+
λ
c4
λ
c2+
λ
c3
+λ
c4
λ
c3+
λ
c4
λ
c4
02(λ
c1+
λ
c2+
λ
c3+
λ
c4
)2(λ
c2
+λ
c3
+λ
c4
)2(λ
c3
+λ
c4
)2λ
c4
3. Distance model
Wifes
1
(1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0 λ
d1
λ
d2
λ
d3
λ
d4
(2) λ
d1
0λ
d1
λ
d2
λ
d3
2λ
d1
(3) λ
d2
λ
d1
0λ
d1
λ
d2
2λ
d2
2λ
d1
(4) λ
d3
λ
d2
λ
d1
0λ
d1
2λ
d3
2λ
d2
2λ
d1
(5) λ
d4
λ
d3
λ
d2
λ
d1
02λ
d4
2λ
d3
2λ
d2
2λ
d1
4. Quasi-symmetry model
Wifes
1
(1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0 λ
s21
λ
s31
λ
s41
λ
s51
(2) λ
s21
0λ
s32
λ
s42
λ
s52
2λ
s21
(3) λ
s31
λ
s32
0λ
s43
λ
s53
2λ
s31
2λ
s32
(4) λ
s41
λ
s42
λ
s43
0λ
s54
2λ
s41
2λ
s42
2λ
s43
(5) λ
s51
λ
s52
λ
s53
λ
s54
02λ
s51
2λ
s52
2λ
s53
2λ
s54
1. Wifes education level.
Parameters
Log odds ratios of intermarriage between two
education levels
Husband’s education level Husband’s education level
Husband’s education level
Husband’s education level Husband’s education level
(4) Some post-secondary; (5) Undergraduate degree and over.
Husband’s education level Husband’s education level
Notes: Education levels in the United States are as follows: (1) Less than Grade 9, Grades 9 to 11; (2) Grade 12; (3) College 1 to 3;
(4) College 4 and over. In Canada they are as follows: (1) Less than Grade 9; (2) Some high school; (3) High-school graduation;
Husband’s education level
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 17 -
To answer our second questions regarding changes across education levels, we rely on the
interaction terms between quasi-symmetry parameters and time periods. The change in a given
quasi-symmetry parameter can be expressed as the difference of two log odds ratios or the log of
the ratio of odds ratios. For instance, the change in λs53, which is one half the log odds ratio of
intermarriage between high-school graduation and university graduation, can be expressed as
×÷
×=
×
×=
1
55
53
33
35
2
55
53
33
35
1
55
53
33
35
2
55
53
33
35
1.532.53 log
2
1
loglog
2
1
tttt
tsts F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
F
λλ
.
Thus, a positive change in λsij indicates an increase in the tendency of women and men with
education levels i and j to marry across their own education levels (or a decrease in homogamy).
Alternatively, a negative change in λsij indicates a decrease in the tendency of women and men
with education levels i and j to marry across education levels.
To answer our third question regarding gender-specific trends, we further add the interaction
terms between the hypergamy parameter (λh) and time periods. In this model, where the change
in hypergamy is controlled, the change in a given quasi-symmetry parameter over time
λsij.t2 λsij.t1, assuming i is a higher education level than j, is the change in the log odds ratio of
marring down the education hierarchy among women. Meanwhile, the change in the log odds
ratio of marrying down the education hierarchy among men is (λsij.t2 λsij.t1) + (λh.t2 - λh.t1). 10
4. Data sources
The data for prevailing marriages among young adults aged under 35 were derived from the
1970, 1980, 1990, and 2000 U.S. Census public use sample from the Integrated Public Use
Microdata Series (Ruggles et al., 2003). The U.S. Census does not explicitly differentiate legal
marriage from common-law unions. Beginning in 1980, couples who lived together but were not
legally married were allowed to report the marital status they considered the most appropriate.
Therefore, marriages from the U.S. Census include an increasing proportion of common-law
unions starting from 1980. For Canada, the data are derived from the 1971 Census 33% sample
microdata file, and 20% sample microdata files for the decennial census years from 1981 to
2001. Beginning in 1981, legal marriages and common-law unions are distinguished in the
Census of Canada. We estimated our Census of Canada results separately for all unions (legal
marriages and common-law marriages) and legal marriages only, to see whether they differ in
the trend in educational homogamy. Since differences were invariably trivial, we only report
results for all unions. For both countries, we only consider marriages among the white
population since racial intermarriages may involve unique pattern of educational matching
(Kalmijn, 1991a).
10. For example, λs53.t2 - λs53.t1 is the change in log odds of intermarriage between university educated and high-
school graduates )()(loglog
2
1
1.2.1.532.53
1
55
53
33
35
2
55
53
33
35
ththtsts
tt F
F
F
F
F
F
F
F
λλλλ
+=
×
×, when
university-educated women marry men with high-school graduation, both λh.t2 and λh.t1 equal 0 since the
hypergamy variable equals 0 when women marry down the educational hierarchy. But for university-educated
men marrying women with high-school graduation, the change in λh.t2 has to be taken into account since the
hypergamy variable equals 1 when women marry up the educational hierarchy.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 18 -
As shown in Tables 5 and 6, the sample size ranges from 109,630 to 611,080 marriages in the
United States and from 100,740 to 386,720 in Canada. Following common practice in the
literature (e.g., Raymo and Xie, 2000), we scaled down the sample size to about 100,000 (the
smallest yearly sample size in our data) for each year in the subsequent modeling.11 The rescaled
sample is small enough that our chosen parsimonious quasi-symmetry model can fit the data
reasonably well (with negative BIC [Bayesian Information Criterion] statistic) for any given year
(as shown in Tables 12 and 13).
5. Results
5.1 Changes in absolute rates: Educational homogamy among the married
Tables 5 and 6 show the percentage distribution of wives’ and husbands’ level of education and
changes in absolute rates of homogamy and intermarriage among prevailing marriages for young
adults aged under 35 in the United States (from 1970 to 2000) and Canada (from 1971 to 2001).
Average educational attainment rose for both sexes over the three decades, particularly for
women (the column totals). By 2000 and 2001, wives had higher average education levels than
their husbands in both countries. Starting from a lower base, the gains were larger among
Canadians, and particularly among Canadian wives.12 By 2001, the gender gap in the share of
husbands and wives with university degrees was 5 percentage points in Canada (23.7% for
women and 18.8% for men) compared to a 2.5-percentage-point gap in 2000 in the United States
(27.9% for women and 25.4% for men). Canadians were less likely to have completed a
university degree than Americans but much more likely than Americans to have some post-
secondary schooling.
Among prevailing marriages, the percentage of educationally homogamous couples (the sum of
diagonal cells in Tables 5 and 6) increased steadily in both countries, and by 2000 and 2001,
young couples had a similar level of educational homogamy in the two countries. In the United
States, some 55% of marriages consisted of couples with the same level of education in 2000, up
from 49% in 1970. In Canada, 54% of couples had the same level of education in 2001, up from
42% in 1971.13 The U.S. rate increased by about 2 percentage points per decade. In Canada, the
rate rose by 2.6 percentage points in the 1970s and then accelerated to 6 percentage points in the
1980s and about 5 percentage points in the 1990s.
11. We also use both larger and smaller rescaled sample sizes to test the sensitivity of our results. The parameter
estimates based on different rescaled sample sizes are very similar in value, but are more likely to be significant
with larger sample sizes.
12. Change in the marginal distributions of education for husbands and wives as indexed by the index of
dissimilarity are as follows: U.S. wives, 0.37; Canadian wives, 0.55; U.S. husbands, 0.22; Canadian husbands,
0.38.
13. Among legal marriages the corresponding change was from 42% in 1971 to 44% in 1981, 51% in 1991, and
55% in 2001. Among unions in which the wife’s age is under 35 while the husband’s age is not restricted, the
change was from 42% in 1971 to 43% in 1981, 49% in 1991, and 53% in 2001. See footnote 5.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 19 -
Table 5 Assortative mating on educational attainment for all married couples aged under
35, United States, 1970, 1980, 1990 and 2000
Husband’s years of schooling
Wife’s years of schooling < 9 9 to 11 12 13 to 15 > = 16 Total Sum of
diagonals
percent
1970
< 9 2.7 1.7 1.3 0.2 0.1 6.0
9 to 11 2.9 6.5 7.3 1.2 0.3 18.2
12 2.5 7.4 26.9 8.8 4.6 50.2
13 to 15 0.2 0.7 3.7 5.1 5.3 15.0
> = 16 0.1 0.2 1.0 1.7 7.7 10.6
Total 8.5 16.4 40.1 17.0 18.0 100.0 48.8
Sample size 109,632
1980
< 9 1.4 0.9 0.8 0.2 0.1 3.4
9 to 11 1.3 3.9 5.3 1.2 0.2 12.0
12 1.3 5.0 26.0 10.2 4.6 47.1
13 to 15 0.2 0.7 5.5 8.0 6.9 21.2
> = 16 0.1 0.1 1.5 2.9 11.7 16.3
Total 4.3 10.7 39.1 22.5 23.5 100.0 51.0
Sample size 611,076
1990
< 9 1.0 0.5 0.5 0.2 0.0 2.1
9 to 11 0.6 2.6 3.6 1.0 0.1 7.9
12 0.7 3.9 22.1 9.1 2.3 38.0
13 to 15 0.2 1.3 10.1 14.7 6.4 32.7
> = 16 0.0 0.2 2.1 4.5 12.5 19.2
Total 2.5 8.5 38.3 29.4 21.3 100.0 52.8
Sample size 496,373
2000
< 9 1.4 0.5 0.6 0.2 0.0 2.8
9 to 11 0.6 2.1 2.6 0.8 0.1 6.2
12 0.8 2.9 16.7 6.9 1.6 28.8
13 to 15 0.3 1.4 10.4 16.6 5.8 34.4
>=16 0.1 0.2 2.8 7.0 17.9 27.9
Total 3.2 7.1 33.0 31.4 25.4 100.0 54.7
Sample size 354,061
Source: University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 20 -
Table 6 Assortative marriage on educational attainment for all married couples aged under 35,
Canada, 1971, 1981, 1991 and 2001
Legal marriages and common law
husband’s years of schooling
Legal marriages
husband’s years of schooling
Wife's years of schooling (1) (2) (3) (4) (5) Total (1) (2) (3) (4) (5) Total
1971
(1) < grade 9 8.8 4.9 1.6 1.3 0.1 16.6 8.8 4.9 1.6 1.3 0.1 16.6
(2) Some high school 7.3 15.1 7.2 4.5 1.0 35.0 7.3 15.1 7.2 4.5 1.0 35.0
(3) High-school graduation 2.7 8.2 9.9 6.5 2.5 29.7 2.7 8.2 9.9 6.5 2.5 29.7
(4) Some post-secondary 0.9 2.2 2.6 5.1 3.8 14.5 0.9 2.2 2.6 5.1 3.8 14.5
(5) University degree + 0.1 0.2 0.3 0.7 3.0 4.2 0.1 0.2 0.3 0.7 3.0 4.2
Total 19.7 30.5 21.5 18.0 10.3 100.0 19.7 30.5 21.5 18.0 10.3 100.0
Sum of diagonals 41.8 41.8
Sample size 386,723 386,723
1981
(1) < grade 9 1.8 1.4 0.8 0.9 0.1 5.0 2.0 1.4 0.8 1.0 0.1 5.2
(2) Some high school 2.5 10.9 4.5 7.6 0.6 26.0 2.4 10.3 4.4 7.6 0.6 25.3
(3) High-school graduation 1.2 5.1 8.4 9.7 1.4 25.8 1.2 4.8 8.5 10.0 1.5 26.1
(4) Some post-secondary 0.9 5.1 5.4 17.5 5.4 34.2 0.8 4.8 5.4 17.6 5.6 34.2
(5) University degree + 0.0 0.3 0.5 2.5 5.7 9.0 0.0 0.3 0.5 2.5 6.0 9.3
Total 6.4 22.7 19.6 38.2 13.1 100.0 6.4 21.6 19.6 38.7 13.7 100.0
Sum of diagonals 44.3 44.4
Sample size 318,036 275,674
1991
(1) < grade 9 0.7 0.7 0.3 0.3 0.0 2.0 0.7 0.6 0.3 0.3 0.0 1.8
(2) Some high school 1.1 9.4 3.4 4.8 0.3 19.0 1.0 8.6 3.0 4.8 0.3 17.7
(3) High-school graduation 0.6 4.8 9.3 7.9 0.9 23.5 0.6 4.5 9.4 8.5 1.0 24.0
(4) Some post-secondary 0.6 6.2 8.0 23.6 4.7 43.1 0.6 5.6 7.9 24.2 5.1 43.4
(5) University degree + 0.0 0.4 1.0 4.1 6.9 12.4 0.0 0.4 0.9 4.2 7.5 13.1
Total 3.1 21.5 21.9 40.7 12.9 100.0 2.9 19.6 21.6 42.0 13.9 100.0
Sum of diagonals 50.0 50.5
Sample size 264,339 197,607
2001
(1) < grade 9 0.6 0.3 0.2 0.2 0.0 1.3 0.7 0.2 0.1 0.2 0.0 1.2
(2) Some high school 0.6 5.5 1.9 3.3 0.2 11.5 0.4 4.3 1.5 3.1 0.2 9.4
(3) High-school graduation 0.4 2.5 5.4 5.2 0.6 13.9 0.3 1.9 5.1 5.3 0.7 20.8
(4) Some post-secondary 0.6 5.8 8.1 29.9 5.3 49.6 0.4 4.9 8.0 30.2 6.0 42.0
(5) University degree + 0.1 0.6 1.4 8.9 12.7 23.7 0.0 0.6 1.5 9.6 14.9 26.6
Total 2.2 14.6 16.9 47.5 18.8 100.0 1.8 12.0 16.1 48.4 21.7 100.0
Sum of diagonals 54.0 55.2
Sample size 173,179 100,744
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 21 -
Tables 7 and 8 show the detailed trends in homogamy and intermarriage for men and women
separately by education level for the United States and Canada respectively. The total
homogamy rate rose among women in both countries but the trend was driven mainly by the
sharp increase among women with some post-secondary education. The average trend, however,
was offset by a decline in homogamous marriages among university-educated women—a decline
of 8 percentage points in the United States and almost 17 percentage points in Canada. By 2001,
only 54% of young Canadian university-educated married women had partners with university
degrees. Less- educated women, in contrast, were somewhat more likely to marry up in 2001
than in 1971.
Not surprisingly the trends for men tend to be the mirror image of those for women. Rates of
homogamy and marrying ‘up’ rose sharply for better-educated men, and by 2000 and 2001, well-
educated men were more likely to have a highly educated partner than were highly educated
women, reversing the situation of 1970 and 1971. Among less-educated men, rates of homogamy
actually declined (except for those with less than nine years of schooling in the United States)
because of a dramatic increase in the share marrying better-educated women. The share of male
high-school graduates marrying better-educated women, for example, rose from 12% to 40% in
the United States and from 13% to 56% in Canada over the three decades. In 1970 and 1971,
women with high-school completion were much more likely than men to marry up. By 2000 and
2001, the advantage had turned decisively to men. Overall, rising education levels have
improved the marriage market for men much more than for women.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 22 -
Table 7 Changes in upward, downward and homogamous marriage, by sex and education level,
United States, 1970, 1980, 1990 and 2000
Wife’s
years of schooling
Husband’s
years of schooling
< 9 9 to 11 12 13 to 15 > = 16 Total < 9 9 to 11 12 13 to 15 > = 16 Total
% of marriage within an education level in a given year % of marriage within an education level in a given year
Up
1970 54.6 48.2 26.8 35.6 29.3 67.7 49.8 11.8 9.9 20.5
1980 59.9 56.4 31.4 32.4 28.1 67.5 54.5 17.8 13.0 20.0
1990 55.1 59.5 29.8 19.6 22.0 61.9 63.5 31.8 15.1 23.4
2000 47.6 55.4 29.4 16.8 20.4 54.6 63.2 39.9 22.2 24.5
Homogamous
1970 45.4 35.8 53.5 33.7 72.3 49.9 32.3 39.7 66.9 29.7 42.6 51.7
1980 40.1 32.9 55.2 37.6 72.0 52.2 32.5 36.9 66.5 35.5 49.8 52.5
1990 44.9 33.3 58.0 45.0 65.0 53.2 38.1 30.9 57.6 50.1 58.6 52.8
2000 52.4 34.4 57.9 48.2 64.1 54.2 45.4 29.7 50.5 53.0 70.4 55.5
Down
1970 16.0 19.7 30.7 27.7 20.7 10.5 21.3 60.4 57.4 27.7
1980 10.7 13.4 30.0 28.0 19.7 8.6 15.7 51.5 50.2 27.5
1990 7.2 12.2 35.4 35.0 24.8 5.6 10.6 34.8 41.4 23.8
2000 10.2 12.7 35.0 35.9 25.4 7.1 9.6 24.8 29.6 20.0
… not applicable
Source: University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Table 8 Changes in upward, downward and homogamous marriage, by sex and education level,
Canada, 1971, 1981, 1991 and 2001
Wives
years of schooling
Husbands
years of schooling
< 9 9 to 11 12 13 to 15 > = 16 Total < 9 9 to 11 12 13 to 15 > = 16 Total
% of marriage within an educational level in a given year % of marriage within an educational level in a given year
Up
1971 47.1 36.1 30.1 26.0 … 31.4 55.3 34.6 13.3 3.9 … 25.8
1981 63.3 48.8 43.3 15.7 … 29.6 71.0 45.9 30.1 6.5 … 25.5
1991 63.6 44.4 37.5 10.9 … 24.2 77.0 52.9 40.9 10.0 25.5
2001 53.7 47.4 41.4 10.8 … 20.2 71.4 60.4 55.9 18.7 24.9
Homogamous
1971 52.9 43.0 33.4 34.9 71.2 44.3 44.7 49.5 46.1 28.0 28.7 43.7
1981 36.7 41.8 32.6 51.1 63.3 46.1 29.0 47.9 42.9 45.8 43.2 44.5
1991 36.4 49.6 39.5 54.8 55.9 49.9 23.0 44.0 42.5 58.0 53.9 50.5
2001 46.3 47.6 38.5 60.2 53.7 53.2 28.6 37.3 31.6 62.9 67.6 56.3
Down
1971 … 20.9 36.5 39.1 28.8 24.2 15.9 40.6 68.1 71.3 30.5
1981 9.4 24.1 33.2 36.7 24.4 6.2 27.0 47.7 56.8 30.0
1991 6.0 23.0 34.3 44.1 25.9 3.1 16.6 32.0 46.1 24.0
2001 5.0 20.1 29.0 46.3 26.6 2.3 12.5 18.4 32.4 18.8
… not applicable
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample micro-data files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 23 -
5.2 Changes in relative rates
Given the larger gains in wives’ educational attainment, changes observed in the absolute rates
may not appear especially surprising. In 1970 and 1971, husbands were on average much better
educated than wives while in 2000 and 2001, wives were better educated than husbands. The
convergence of wives’ education on that of husbands increased the possibility of forming
homogamous unions. As wives surpassed the educational attainments of husbands, the likelihood
that more women would marry down increased as did the likelihood that husbands would marry
better-educated women. Testing the modernization thesis, for example, requires estimation of
changes in relative rates, net of changes in the educational distributions of men and women.
Table 9 summarizes the model goodness-of-fit estimates for various steps of the log-linear
estimation. The starting model (M0) includes only the marginal values for husbands’ and wives’
education and period but assumes no association between the education of husbands and wives,
and no association between time period and the association between husbands’ and wives’
education. Model one (M1) adds parameters for the quasi-symmetry model of association
between husbands’ and wives’ education and a hypergamy (the tendency for wives to marry up)
parameter, but assumes that the associations do not change over time. Introduction of the
hypergamy parameter essentially allows the model to estimate different parameters for husbands
and wives, whether men and women at the same education level have the same tendency to
marry within or to marry out. For both countries, the model (M1) with quasi-symmetry and
hypergamy parameters significantly improves the model fit relative to the baseline model that
assumes no association between husbands’ and wives’ education.
Table 9 Goodness-of-fit results for models of all marriages among young adults, Canada and
United States
United States Canada, legally married and
common law
Canada, legally married
d.f. L2BIC d.f. L2BIC d.f. L2BIC
M0: Baseline model 64 199833 199007 64 150566 149829 64 153072 152246
M1: M0 + Quasi symmetry + hypergamy 53 1252 568 53 2104 1420 53 2748 2064
M2: M1 + log-multiplicative layer effect 50 966 321 50 1502 857 50 2006 1360
M3: M1 + Quasi symmetry*period 23 254 -43 23 220 -77 23 257 -40
M4: M3 + hypergamy*period 20 134 -124 20 204 -54 20 221 -37
Notes: The baseline model contains the main effects of wives’ education, husbands’ education, and period; and two-way interactions between period
and education for each sex. L2 is the log-likelihood ratio chi-square statistic; d.f. is the degree of freedom; BIC is the Bayesian Information
Criterion. BIC = L2 – (d.f.)ln(N), where N is the rescaled sample size (401,282 for the United States and 402,499 for Canada).
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample
microdata files; University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 24 -
Change in relative rates of homogamy
Model two (M2) adds the log-multiplicative layer effect to the model (M1) with quasi-symmetry
and hypergamy parameters, and provides the parameter values (the normalized Φ parameter) that
answer the question of whether or not the relative rate of homogamy has changed over time. The
log-multiplicative layer effect model (M2) has the best fit based on the BIC (Bayesian
Information Criterion) statistic in Canada and improves the model fit by the standard of the log-
likelihood ratio chi-square statistic in both countries.
The log-multiplicative layer effect model (M2) tests for the ‘average’ or ‘overall’ change in the
relative homogamy rate over time. The results confirm a steady increase in the relative rates of
educational homogamy. The normalized Φ parameter for the United States rose from 0.48 in
1970, to 0.49 in 1980, to 0.51 in 1990, and to 0.53 in 2000. In Canada, Φ for all unions increased
from 0.45 in 1971, to 0.47 in 1981, to 0.53 in 1991, and to 0.54 in 2001.14 In effect, in both
countries, the relative rate of marital homogamy increased unambiguously over all three decades.
Importantly, the predicted values for marital homogamy, net of changes in the marginals,
indicate that most of the increase was due to change in the association between husbands’ and
wives’ education rather than to changes in the distribution of husbands’ and wives’ education. In
the United States, the change in association, net of changes in the marginals, accounts for
4 percentage points of the 6-percentage-point increase in educational homogamy from 1970 to
2000. In Canada, the change in association accounts for almost 10 percentage points of the 12-
percentage-point increase. In other words, most of the increase in educational homogamy over
the three decades is not a result of changes in the relative supply of husbands and wives with
different education levels.
Changes in the odds of crossing educational boundaries
The heterogeneous change model (M3) generates parameters required to answer questions
concerning where in the educational hierarchy the rise in homogamy is being produced (i.e., are
the changes at different education levels the same or different.). M3 improves the model fit
further by the standard of the log-likelihood ratio chi-square statistic and BIC (Table 9). Finally,
Model 4 (M4) asks whether the change in the subcomponents differs among husbands and wives
by testing for change in the hypergamy parameter. The inclusion of changes in the hypergamy
parameter also improves the model fit further, although it is not as parsimonious as M3 for the
Canadian data. The parameter estimates for both M3 and M4 are presented in Table 14.
To facilitate interpretation, we transform the parameter estimates in M3 and M4 into the odds of
intermarriage relative to the odds of a homogamous marriage by period and present them in
Table 10 for the United States and in Table 11 for Canada. Each table contains three panels titled
separately as “Overall” (first panel), “Wives marrying down” (second panel), and “Husbands
marrying down” (third panel). The odds ratios in the first panel are derived from M3 which
assumes change in intermarriage is symmetrical with respect to sex. The odds ratios in the
14. Among legal marriages the corresponding change was from 0.44 in 1971 to 0.47 in 1981, 0.54 in 1991, and
0.55 in 2001. Among all unions in which the wife’s age is under 35 while the husband’s age is not restricted,
the change was from 0.46 in 1971, to 0.47 in 1981, 0.53 in 1991, and 0.54 in 2001.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 25 -
second and third panels are derived from M4 which specifies that men and women experience
different changes in intermarriage.
In both countries, intermarriage across education levels occurs primarily between adjacent
education levels, and the odds ratios for intermarriage are very small when wives and husbands
are separated by more than one education level. Since the odds of intermarriage across more than
one education level are very small, their changes over time have minimum impact on the overall
trends in educational homogamy. Hence, our summary of findings will focus primarily on
changes in intermarriage across adjacent education levels.
Declining intermarriage at the top of the educational hierarchy
In both Canada and the United States, declining intermarriage among university graduates was a
major source of the overall rise in educational homogamy. In Canada, the relative rate of
intermarriage between the university educated and those with some university fell by 38% (from
0.201 to 0.125) and in the United States by 45% (from 0.249 to 0.136). The largest declines
occurred in Canada during the 1970s and in the United States during the 1980s. In Canada, the
downward trend was arrested in the 1990s, and rates of intermarriage by the university educated
actually rose slightly by 2001.
Table 10 Odds of intermarriage relative to odds of homogamy among prevailing marriages,
United States, 1970, 1980, 1990 and 2000
Overall
Spouse’s years of schooling
Wives marrying down
Husband’s years of schooling
Husbands marrying down
Wife’s years of schooling
< 9 9 to 11 12 13 to 15 < 9 9 to 11 12 13 to 15 < 9 9 to 11 12 13 to 15
1970
9 to 11 (years) 0.282 0.199 0.287
12 0.043 0.304 0.029 0.212 0.042 0.304
13 to 15 0.003 0.025 0.248 0.002 0.018 0.169 0.003 0.026 0.243
> = 16 0.0003 0.001 0.021 0.249 0.0002 0.001 0.015 0.168 0.0003 0.001 0.022 0.242
1980
9 to 11 (years) 0.218 0.170 0.220
12 0.031 0.259 0.024 0.201 0.031 0.259
13 to 15 0.004 0.027 0.270 0.003 0.021 0.208 0.004 0.028 0.268
> = 16 0.0002 0.001 0.021 0.219 0.0002 0.0004 0.017 0.168 0.0003 0.001 0.022 0.217
1990
9 to 11 (years) 0.106 0.093 0.106
12 0.017 0.243 0.015 0.213 0.017 0.243
13 to 15 0.003 0.032 0.283 0.002 0.029 0.249 0.003 0.032 0.283
> = 16 0.0001 0.001 0.017 0.155 0.0001 0.001 0.015 0.137 0.0001 0.001 0.017 0.156
2000
9 to 11 (years) 0.106 0.110 0.105
12 0.019 0.212 0.020 0.220 0.019 0.211
13 to 15 0.002 0.029 0.259 0.002 0.032 0.268 0.002 0.030 0.257
> = 16 0.0001 0.0005 0.014 0.136 0.0001 0.001 0.016 0.142 0.0001 0.001 0.015 0.136
not applicable
Note: Results were derived by the authors from the source below.
Source: University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 26 -
Whereas Canadian trends were quite similar for men and women, the U.S. decline was almost
entirely driven by declining intermarriage among university-educated men (from 0.242 to 0.136,
or 44%). The decline in intermarriage among U.S. university-educated women (from 0.168 to
0.142, or 15%) was modest by comparison and, as in Canada, stabilized and even rose slightly
during the 1990s.
Trends in the middle
In the United States, the odds ratios of intermarriage between those with some post-secondary
education and high-school graduates rose from 1970 to 1990 (from 0.248 to 0.283), falling back
to 0.259 by 2000. The U.S. trends were very different for men and women however. The odds
ratios of intermarriage among women rose by 59% (from 0.169 to 0.268). Among men, the odds
ratios of intermarriage rose moderately (from 0.243 to 0.283) from 1970 to 1990, before falling
back to 0.257 in 2000. In Canada, intermarriage rose during the 1970s but then declined
substantially over the following two decades for a net decline of 12%, and changes were quite
similar among women and men.
Table 11 Odds of intermarriage relative to odds of homogamy among prevailing marriages, Canada,
1971, 1981, 1991 and 2001
Overall Wives marrying down Husbands marrying down
Spouse’s years of schooling Husband’s years of schooling Wive’s years schooling
< 9 9 to 11 12 13 to 15 < 9 9 to 11 12 13 to 15 < 9 9 to 11 12 13 to 15
1971
9 to 11 (years) 0.268 0.251 0.269
12 0.047 0.392 … 0.044 0.366 … 0.047 0.393 …
13 to 15 0.025 0.130 0.339 0.023 0.122 0.313 … 0.025 0.130 0.336
> = 16 0.0002 0.003 0.021 0.201 0.0002 0.003 0.019 0.186 0.0002 0.003 0.021 0.199
1981
9 to 11 (years) 0.179 0.168 0.179
12 0.057 0.249 … 0.053 0.234 … 0.056 0.249 …
13 to 15 0.025 0.201 0.363 0.023 0.189 0.339 … 0.025 0.202 0.362
> = 16 0.0001 0.003 0.013 0.137 0.0001 0.002 0.012 0.128 0.0001 0.003 0.013 0.137
1991
9 to 11 (years) 0.114 0.116 0.113
12 0.028 0.184 … 0.029 0.187 … 0.028 0.183 …
13 to 15 0.010 0.133 0.288 0.010 0.136 0.295 … 0.010 0.133 0.288
> = 16 0.0001 0.002 0.014 0.118 0.0001 0.002 0.014 0.120 0.0001 0.002 0.013 0.118
2001
9 to 11 (years) 0.057 0.059 0.057
12 0.019 0.163 … 0.020 0.170 … 0.019 0.162 …
13 to 15 0.006 0.116 0.264 0.007 0.122 0.276 … 0.006 0.116 0.263
> = 16 0.0001 0.002 0.012 0.125 0.0001 0.002 0.012 0.131 0.0001 0.002 0.012 0.125
not applicable.
Note: Results were derived by the authors from the source below.
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 27 -
From 1970 to 2000, the odds ratios of intermarriage between high-school graduates and those
with some high school fell from 0.304 to 0.212 (30%) in the United States but the overall trend
was entirely driven by high-school men. Among high-school women, the odds ratios of
intermarriage with men with some high school actually rose slightly from 0.212 to 0.220 in the
same period. In Canada, the decline in intermarriage between high-school graduates and those
with some high school was even more precipitous, falling by 58% (from 0.392 to 0.163) from
1971 to 2001, and trends were very similar for men and women.
Intermarriage at the bottom
Declines in the odds ratio of intermarriage between those with some high school and those with
less than high school were uniformly large in both Canada and the United States for both men
and women. In the United States, the odds ratio of intermarriage between those with less than 9
years of schooling and those with 9 to 11 years of schooling decreased 62%, from 0.282 in 1970
to 0.106 in 2000. The decline in the odds ratio of intermarriage among the least educated was
much larger among husbands than among wives. The odds ratio of marrying down among those
with 9 to 11 years of schooling declined 63% among husbands (from 0.287 to 0.105), compared
with a decline of 45% among wives. Thus, the chance of the least-educated wives marrying up
declined faster than that of least-educated husbands.
In Canada, the odds of intermarriage between those with 9 to 11 years of schooling and those
with less than 9 years of schooling decreased 79% from 1971 to 2001, and the decline was
similar in magnitude among men and women.
6. Conclusion
Our analysis of prevailing marriages among young adults reveals several clear trends in
educational homogamy and intermarriage in the United States and Canada. First, the overall
level of both absolute and relative rates of educational homogamy have unambiguously
increased in both countries over the three decades, and we find no evidence for the asymmetric
inverted-U shape reported in some previous studies. Second, the overall trend at the national
level appears to have been driven mainly by changes in the association of husbands’ and wives’
education rather than by changes in the relative supply of more- and less-educated partners.
Declining odds of intermarriage at both the top and the bottom of the educational hierarchy were
major drivers of the rising rate of marital homogamy in both countries, but two important
differences stand out. Declining rates of intermarriage between high-school graduates and those
with both higher and lower levels of education were much larger in Canada. Indeed the odds
ratios of intermarriage between those with some university education and high-school graduates
actually increased in the United States. Second, while changing patterns of intermarriage were
quite similar among Canadian men and women, changes among U.S. men and women differed in
important ways. The declining odds of marrying down among university graduates were
predominantly a male phenomenon in the United States, while rising intermarriage with high-
school graduates by those with some post-secondary education occurred mainly among women.
Finally, the odds of marrying down among female high-school graduates rose slightly over the
period, but fell among male school graduates. Put somewhat differently, in the United States, the
prospect of marrying up the educational hierarchy improved substantially for male high-school
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 28 -
graduates and males with less than high school and, compared to women, declined only modestly
for men with some post-secondary education.
There were two important exceptions to the general trend towards rising homogamy. The first
was the aforementioned increase in intermarriage between those with some post-secondary
education and high-school graduates in the United States. The second was the stabilization
and/or small gains in intermarriage in the 1990s among male and female university graduates in
Canada and among female university graduates in the United States.
Extrapolating from either the overall trend or the exceptions to it in order to draw strong
conclusions concerning prevailing theoretical accounts of trends is, however, a hazardous
exercise. We are skeptical about drawing strong conclusions concerning the signal that changes
in relative rates provide with respect to changes in marital preferences versus changes in the
marital opportunity structure. Changes in relative rates (the odds ratios of homogamy or
intermarriage), controlling for changes in the marginal distributions at the national level, tell us
little about changes in the marital opportunity structure at the ‘local’ level, whether defined
geographically or organizationally (i.e., in schools and workplaces), where most future partners
meet.
The increased tendency of U.S. women with some post-secondary education to marry down and
small increases in intermarriage by female university graduates in the 1990s in both countries are
seemingly consistent with Oppenheimer’s (1994: 315) observations concerning changing
preferences, the expectation that more-educated women are more willing and able “to marry a
man who is unlikely to be a great provider but who is highly desirable in other respects.”
However, it is important to point out that the rising intermarriage in the United States occurred
predominantly between women with some college education and men with high-school
graduation. Women with some college education were not the ones with the greatest economic
resources or potential. Indeed, even by 2000 they still had a much lower labour force
participation rate than men with high-school graduation. Among those who participated in the
labour force, women with some college education still earned less than 60% of what male high-
school graduates did.15 Similarly, university-educated women had a much lower labour force
participation rate and lower average earnings than men with some college education. Therefore,
the large increase in intermarriage among women with some college education over the three
decades and the small increase in intermarriage among women who finished college education
may primarily reflect the possibility that those well-educated women who are not successful in
the labour market marry men who are less educated but manage to be in a superior financial
position. More empirical studies are needed to confirm this possibility.
15. Based on the 2000 U.S. Census public use sample, we find that among married young women (under age 35)
the labour force participation rate (defined here as the percentage with positive employment earnings in the
previous year) was about 82% among the university educated, 72% among those with some college education,
and 66% among high-school graduates. Among those with positive employment earnings, the average earnings
were $34,000 among the university educated, $21,100 among those with some college education, and $17,300
among high-school graduates. Among married young men, the labour force participation rate was 96% among
the university educated, 94% among those with some college education, and 91% among high-school
graduates. Among young husbands with positive employment earnings, the average earnings were $55,800
among the university educated, $36,700 among those with some college education, and $31,700 among high-
school graduates.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 29 -
The overall downward trend in intermarriage, especially from 1970 to 1990 in the United States
and from 1971 to 1991 in Canada, does not support Mare’s (1991) life-course hypothesis, which
postulates changes in the opportunity structure as the causal mechanism—that the rising time
gap between school completion and marriage increases the probability that men and women with
different educational backgrounds will meet one another. In particular, the life-course hypothesis
suggests that marriage across barriers at the university level is particularly sensitive to the time
gap between school leaving and marriage. However, the time gap for the university educated
rose continuously over three decades in Canada while intermarriage by the university-educated
people decreased significantly from 1970 to 1990 and only increased slightly in the 1990s.16
It is clear, however, that the U-turn in marital homogamy postulated by modernization theorists
(Smits, Ultee and Lammers, 1998; Smits, Ultee and Lammers, 2000; and Smits, 2003) has yet to
have had large impacts on marital patterns in Canada and the United States. Were it so, this
would be good news for policy-makers and those concerned with rising inequality in family
incomes. The rise in educational homogamy, along with increased selection into marriage based
on education, has been a potent force underlying rising inequality in family earnings. Were it the
case that these trends in absolute levels of marital homogamy were driven mainly by the
revolution in women’s educational attainment (changes in the ‘marginals’) there would be reason
for optimism since that trend is undoubtedly now reaching maturity. But that is not the case:
most of the gains in marital homogamy are the result of changes in the association between
husbands’ and wives’ education rather than by changes in their levels. The exceptions to the
trend provide some evidence that this upward trend may be abating, but scant hope for any large-
scale reversal in the proximate future.
16. In the United States, the time-gap between age at marriage and age at school leaving increased from 1970
through the late 1980s, but may have stabilized in the 1990s (Schwartz and Mare, 2005). In Canada, the gap
between the two life-events has been rising continuously since the earlier 1970s. Median age at first marriage
has increased by about six years, while median years of schooling among young adults aged from 20 to 30 has
only increased from 12 to 14 years.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 30 -
Table 12 Parameters and interpretations of common log-linear models of educational assortative
marriages for couples aged under 35, United States, 1970 and 2000
1970 2000
Parameters Log odds ratios of
intermarriage between two
education levels
Parameters Log odds ratios of
intermarriage between two
education levels
1. Quasi-independence model 1. Quasi-independence model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 2.2 0.0 0.0 0.0 0.0 … … (1) 3.7 0.0 0.0 0.0 0.0
(2) 0.0 1.0 0.0 0.0 0.0 -3.2 (2) 0.0 1.8 0.0 0.0 0.0 -5.5
(3) 0.0 0.0 0.6 0.0 0.0 -2.8 -1.6 (3) 0.0 0.0 1.0 0.0 0.0 -4.7 -2.8
(4) 0.0 0.0 0.0 0.7 0.0 -2.9 -1.7 -1.3 (4) 0.0 0.0 0.0 0.3 0.0 -4.0 -2.1 -1.3
(5) 0.0 0.0 0.0 0.0 2.7 -4.9 -3.7 -3.3 -3.4 (5) 0.0 0.0 0.0 0.0 2.4 -6.1 -4.2 -3.4 -2.7
2. Crossings model 2. Crossings model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.8 -1.6 -2.6 -3.5 (1) 0.0 -1.2 -2.1 -2.9 -4.1
(2) -0.8 0.0 -0.8 -1.8 -2.7 -1.6 (2) -1.2 0.0 -0.9 -1.8 -2.9 -2.4
(3) -1.6 -0.8 0.0 -1.0 -1.9 -3.2 -1.6 (3) -2.1 -0.9 0.0 -0.8 -2.0 -4.2 -1.9
(4) -2.6 -1.8 -1.0 0.0 -1.0 -5.1 -3.5 -2.0 (4) -2.9 -1.8 -0.8 0.0 -1.1 -5.9 -3.5 -1.7
(5) -3.5 -2.7 -1.9 -1.0 0.0 -7.0 -5.4 -3.9 -1.9 (5) -4.1 -2.9 -2.0 -1.1 0.0 -8.2 -5.8 -3.9 -2.3
3. Distance model 3. Distance model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.6 -1.8 -3.4 -4.2 (1) 0.0 -0.8 -2.0 -3.5 -4.2
(2) -0.6 0.0 -0.6 -1.8 -3.4 -1.2 (2) -0.8 0.0 -0.8 -2.0 -2.9 -1.7
(3) -1.8 -0.6 0.0 -0.6 -1.8 -3.6 -1.2 (3) -2.0 -0.8 0.0 -0.8 -2.0 -4.0 -1.7
(4) -3.4 -1.8 -0.6 0.0 -0.6 -6.7 -3.6 -1.2 (4) -3.5 -2.0 -0.8 0.0 -0.8 -6.9 -4.0 -1.7
(5) -4.2 -3.4 -1.8 -0.6 0.0 -8.5 -6.7 -3.6 -1.2 (5) -4.5 -3.5 -2.0 -0.8 0.0 -9.1 -6.9 -4.0 -1.7
4. Quasi-symmetry model 4. Quasi-symmetry model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.6 -1.6 -2.9 -4.3 (1) 0.0 -1.1 -2.0 -3.2 -4.6
(2) -0.6 0.0 -0.6 -1.9 -3.7 -1.3 (2) -1.1 0.0 -0.8 -1.8 -3.8 -2.3
(3) -1.6 -0.6 0.0 -0.7 -2.0 -3.1 -1.2 (3) -2.0 -0.8 0.0 -0.7 -2.1 -4.0 -1.6
(4) -2.9 -1.9 -0.7 0.0 -0.7 -5.7 -3.7 -1.4 (4) -3.2 -1.8 -0.7 0.0 -1.0 -6.3 -3.5 -1.4
(5) -4.3 -3.7 -2.0 -0.7 0.0 -8.5 -7.5 -3.9 -1.4 (5) -4.6 -3.8 -2.1 -1.0 0.0 -9.2 -7.6 -4.2 -2.0
not applicable
1. Wifes education level.
Notes: Education levels are as follows: (1) Less than Grade 9, Grades 9 to 11; (2) Grade 12; (3) College 1 to 3; (4) College 4 and over. Goodness of fit for 1970: quasi-
independent model: d.f. =11, L2 = 17783, BIC=17656; crossings model: d.f. =12, L2=2398, BIC=2260; distance model: d.f.=12, L2=705, BIC=567; quasi-
symmetry model: d.f.=6, L2=270, BIC=201, n =100470. Goodness of fit for 2000: quasi-independent model: d.f. =11, L2 = 10145, BIC=10018; crossings model:
d.f. =12, L2=1005, BIC=867; distance model: d.f.=12, L2=946, BIC=808; quasi-symmetry model: d.f.=6, L2=1 , BIC=-57, n =100180. 2
Source: University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 31 -
Table 13 Parameters and interpretations of common log-linear models of educational assortative
marriages for couples aged under 35, Canada, 1971 and 2001
1971 2001
Parameters Log odds ratios of
intermarriage between two
education levels
Parameters Log odds ratios of
intermarriage between two
education levels
1. Quasi-independence model 1. Quasi-independence model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 1.8 0.0 0.0 0.0 0.0 … … (1) 4.6 0.0 0.0 0.0 0.0
(2) 0.0 0.5 0.0 0.0 0.0 -2.2 (2) 0.0 2.0 0.0 0.0 0.0 -6.6
(3) 0.0 0.0 0.6 0.0 0.0 -2.3 -1.0 (3) 0.0 0.0 1.2 0.0 0.0 -5.8 -3.2
(4) 0.0 0.0 0.0 0.8 0.0 -2.3 -1.2 -1.3 (4) 0.0 0.0 0.0 -0.1 0.0 -4.5 -1.9 -1.2
(5) 0.0 0.0 0.0 0.0 3.1 -4.9 -3.6 -3.7 -3.9 (5) 0.0 0.0 0.0 0.0 2.4 -7.0 -4.4 -3.6 -2.4
2. Crossings model 2. Crossings model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.8 -1.4 -2.0 -3.2
………
(1) 0.0 -1.7 -2.4 -3.0 -4.2
………
(2) -0.8 0.0 -0.6 -1.3 -2.5
-1.6 ………
(2) -1.7 0.0 -0.7 -1.3 -2.5
-3.4
(3) -1.4 -0.6 0.0 -0.7 -1.9
-2.7 -1.2 (3) -2.4 -0.7 0.0 -0.6 -1.8
-4.9 -1.5
(4) -2.0 -1.3 -0.7 0.0 -1.2
-4.1 -2.5 -1.4 (4) -3.0 -1.3 -0.6 0.0 -1.2
-6.1 -2.7 -1.2
(5) -3.2 -2.5 -1.9 -1.2 0.0
-6.5 -4.9 -3.8 -2.4 (5) -4.2 -2.5 -1.8 -1.2 0.0
-8.4 -5.0 -3.5 -2.3
3. Distance model 3. Distance model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.6 -1.3 -2.2 -4.1
………
(1) 0.0 -0.9 -1.5 -3.0 -4.9
………
(2) -0.6 0.0 -0.6 -1.3 -2.2
-1.1 ………
(2) -0.9 0.0 -0.9 -1.5 -3.0
-1.8
(3) -1.3 -0.6 0.0 -0.6 -1.6
-2.7 -1.1 (3) -1.5 -0.9 0.0 -0.9 -1.5
-3.0 -1.8
(4) -2.2 -1.3 -0.6 0.0 -1.3
-4.5 -2.7 -1.1 (4) -3.0 -1.5 -0.9 0.0 -0.9
-6.0 -3.0 -1.8
(5) -4.1 -2.2 -1.3 -0.6 0.0
-8.2 -4.5 -2.7 -1.1 (5) -4.9 -3.0 -1.5 -0.9 0.0
-9.8 -6.0 -3.0 -1.8
4. Quasi-symmetry model 4. Quasi-symmetry model
Husbands education level Husband’s education level Husbands education level Husband’s education level
Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4) Wifes1 (1) (2) (3) (4) (5) (1) (2) (3) (4)
(1) 0.0 -0.7 -1.5 -1.8 -4.4
………
(1) 0.0 -1.7 -2.3 -2.9 -5.2
………
(2) -0.7 0.0 -0.5 -1.0 -3.0
-1.3 ………
(2) -1.7 0.0 -1.0 -1.1 -3.1
-3.4
(3) -1.5 -0.5 0.0 -0.5 -1.9
-3.0 -0.9 (3) -2.3 -1.0 0.0 -0.6 -2.2
-4.6 -2.0
(4) -1.8 -1.0 -0.5 0.0 -0.8
-3.7 -2.0 -1.1 (4) -2.9 -1.1 -0.6 0.0 -1.0
-5.9 -2.2 -1.3
(5) -4.4 -3.0 -1.9 -0.8 0.0
-8.8 -6.0 -3.9 -1.6 (5) -5.2 -3.1 -2.2 -1.0 0.0
-10.5 -6.3 -4.3 -2.1
not applicable
1. Wifes education level.
Notes: Education levels are as follows: (1) Less than high school, (2) Some high school, (3) High-school graduation, (4) Some post-secondary, (5) University degree
and over. Goodness of fit for 1971: quasi-independent model: d.f.=11, L2=12756, BIC=12629; crossings model: d.f.=12, L2=1899, BIC=1761; distance model:
d.f.=12, L2=2029; BIC=1901; quasi-symmetry model: d.f.=6; L2=70, BIC=1, n =100590. Goodness of fit for 2001: quasi-independent model: d.f. =11, L2 =
5593, BIC=5466; crossings model: d.f.=12, L2=1725, BIC=1587; distance model: d.f.=12, L2=3498, BIC=3360; quasi-symmetry model: d.f.=6, L2=13, BIC=-
6; n =100640.
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 32 -
Table 14 Changes in the parameters in quasi-symmetry and hypergamy models for
Quasi-sy mmetry
λ
s51,
1970
1
-4,22 *** -4,26 *** -4,37 *** -4,37 *** -4,37 *** -4,37 ***
1980/1970 -0,04 -0,01 -0,08 -0,09 -0,10 -0,12
1990/1970 -0,57 *** -0,49 ** -0,56 ** -0,53 ** -0,72 *** -0,69 ***
2000/1970 -0,53 *** -0,29 * -0,55 *** -0,48 ** -0,89 *** -0,76 ***
λ
s52,
1970 -3,75 *** -3,78 *** -2,99 *** -2,99 *** -2,99 *** -2,99 ***
1980/1970 -0,12 -0,10 -0,01 -0,01 0,00 -0,01
1990/1970 -0,07 -0,01 -0,18 *** -0,14 ** -0,19 *** -0,15 **
2000/1970 -0,16 * 0,04 -0,23 *** -0,17 ** -0,18 *** -0,07
λ
s53,
1970 -2,02 *** -2,09 *** -1,95 *** -1,97 *** -1,95 *** -1,97 ***
1980/1970 0,01 0,05 -0,23 *** -0,23 *** -0,24 *** -0,24 ***
1990/1970 -0,11 *** -0,01 -0,21 *** -0,17 *** -0,22 *** -0,18 ***
2000/1970 -0,19 *** 0,01 -0,29 *** -0,23 *** -0,24 *** -0,13 **
λ
s54,
1970 -0,79 *** -0,89 *** -0,82 *** -0,84 *** -0,81 *** -0,84 ***
1980/1970 -0,06 ** 0,00 -0,19 *** -0,19 *** -0,19 *** -0,20 ***
1990/1970 -0,24 *** -0,10 *** -0,27 *** -0,22 *** -0,27 *** -0,22 ***
2000/1970 -0,30 *** -0,09 ** -0,24 *** -0,18 *** -0,23 *** -0,13 ***
λ
s41,
1970 -2,94 *** -3,04 *** -1,86 *** -1,88 *** -1,85 *** -1,88 ***
1980/1970 0,04 0,11 -0,01 -0,01 -0,03 -0,04
1990/1970 -0,14 0,00 -0,46 *** -0,41 *** -0,53 *** -0,48 ***
2000/1970 -0,33 *** -0,08 -0,71 *** -0,64 *** -1,11 *** -0,99 ***
λ
s42,
1970 -1,94 *** -2,01 *** -1,03 *** -1,05 *** -1,03 *** -1,05 ***
1980/1970 0,04 0,09 * 0,22 *** 0,22 *** 0,21 *** 0,20 ***
1990/1970 0,13 *** 0,24 *** 0,01 0,06 * -0,01 0,04
2000/1970 0,08 * 0,29 *** -0,06 ** 0,00 -0,07 *** 0,02
λ
s43,
1970 -0,79 *** -0,89 *** -0,55 *** -0,58 *** -0,55 *** -0,58 ***
1980/1970 0,04 ** 0,10 *** 0,03 * 0,04 0,03 0,03
1990/1970 0,07 *** 0,19 *** -0,08 *** -0,03 -0,07 *** -0,02
2000/1970 0,02 0,23 *** -0,12 *** -0,06 * -0,10 *** -0,01
λ
s31,
1970 -1,66 *** -1,76 *** -1,54 *** -1,56 *** -1,53 *** -1,56 ***
1980/1970 -0,16 *** -0,10 * 0,09 ** 0,09 * 0,07 * 0,07
1990/1970 -0,45 *** -0,32 *** -0,26 *** -0,21 *** -0,39 *** -0,34 ***
2000/1970 -0,42 *** -0,20 *** -0,45 *** -0,39 *** -0,79 *** -0,69 ***
λ
s32,
1970 -0,69 *** -0,78 *** -0,48 *** -0,50 *** -0,48 *** -0,50 ***
1980/1970 -0,08 *** -0,03 -0,23 *** -0,23 *** -0,24 *** -0,25 ***
1990/1970 -0,11 *** 0,00 -0,38 *** -0,34 *** -0,42 *** -0,37 ***
2000/1970 -0,18 *** 0,02 -0,44 *** -0,38 *** -0,56 *** -0,47 ***
λ
s21,
1970 -0,72 *** -0,81 *** -0,67 *** -0,69 *** -0,67 *** -0,69 ***
1980/1970 -0,13 *** -0,08 * -0,20 *** -0,20 *** -0,22 *** -0,23 ***
1990/1970 -0,49 *** -0,38 *** -0,43 *** -0,39 *** -0,55 *** -0,50 ***
2000/1970 -0,49 *** -0,30 *** -0,77 *** -0,72 *** -1,05 *** -0,97 ***
Hypergamy, 1970 0,18 *** 0,36 *** 0,03 * 0,07 ** 0,02 0,07 **
1980/1970 -0,11 ** 0,00 0,01
1990/1970 -0,24 *** -0,09 ** -0,10 **
2000/1970 -0,41 *** -0,12 ** -0,19 ***
marriages among young adults, United States and Canada
United States
Canada, legally married and
common law Canada, legally married
Model 4Model 3 Model 4 Model 3 Model 4 Model 3
not applicable
* significance at the 10% level
** significance at the 5% level
*** significance at the 1% level
1. For Canada, the years are the following: λs51, 1971; D1981/1971, D1991/1971; D2001/1971.
Notes: Results for Models 1 and 2 are not presented here; see Table 9 for the specifications. All models in this table contain the main effects
of wives’ education, husbands’ education, and period; and two-way interactions between period and education.
Sources: Statistics Canada, 1971 Census of Canada, 33% sample microdata file and 1981-to-2001 decennial Censuses of Canada, 20% sample
microdata files; University of Minnesota, Minnesota Population Center, 1970-to-2000 U.S. Census public use microdata files.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 33 -
Appendix A Trends in educational homogamy among newly formed first marriages
The data on recent first marriage for the United States were derived from the 1970 U.S. Census
public use 1% sample microdata file and the 1980 U.S. Census public use 5% sample microdata
file, as well as the 1990 and 1992 June Current Population Survey. All these files were
downloaded from Integrated Public Use Microdata Series (Ruggles et al., 2003). The most recent
data were from the Survey of Income and Program Participation, 2001 Panel, Wave 2 (United
States Department of Commerce, 2004). For Canada, the data were derived from the 1971
Census 33% Public Use Sample Tape of Families (Statistics Canada, 1975), the 1984 Canadian
Fertility Survey (Balakrishnan, Krotki and Lapierre-Adamcyk, 1988), and the 1990 and 2001
General Social Survey (Statistics Canada, 1991; 2002).
Recent first marriages were defined as persons who are married for the first time within the
previous two years in United States and within the previous three years in Canada. We use a one
year longer time span for Canada to increase sample size. We only include those who are aged
under 35 at the time of survey to make it compatible with our analysis on prevailing marriages
where including those who are older ages would create too much overlap over a 10-year period.
We only consider marriages among the white population since racial intermarriages may involve
a unique pattern of educational matching (Kalmijn, 1991), and we cannot obtain a large enough
sample size from recent surveys to study non-white population groups separately.
As shown in Tables A.1 and A.2, the sample size for different years ranges from 450 to 76,850
married couples in the United States and from 520 to 4,350 in Canada. Following the common
practice in the literature (e.g., Raymo and Xie, 2000), we rescale the sample size to about 2,000
for each year. We also use both larger and smaller rescaled sample sizes to test the sensitivity of
our results. The parameter estimates based on different rescaled sample sizes are very similar in
value, but are more likely to be significant with larger sample size.
Tables A.1 and A.2, show the percentage distribution of newlyweds by wives’ and husbands’
level of education in the United States from 1970 to 2001, and Canada from 1971 to 2001. The
row and column totals reveal that the average educational attainment increased dramatically for
both sexes over the past four decades, particularly for women (the column total). For instance, in
2001, 45.9% of wives and 37.3% of husbands finished university education in the United States,
compared with 14.2% and 20.2% in 1970. Now newly married wives had higher average
education levels than their husbands, while the opposite was true 40 years ago. Canada also
experienced a similar change.
These tables also show some overall changes in educational homogamy among recent first
marriages. A crude measure of educational homogamy is the percentage of marriages involving
men and women of the same education level (the sum of diagonal cells in the 5 by 5 tables).
Judged from this measure, the United States and Canada seem to have experienced different
trends in educational homogamy. In the 1970s, 51% of recent first marriages are educationally
homogamous in the United States. The percentage of first marriages involving spouses in the
same schooling categories in the United States rose steadily to 53% in 1980, 55% in the early
1990s, and since then fell to 52% in 2001. In Canada, the level of educational homogamy among
newlyweds changed little from 1971 to 1990, but rose abruptly from 44% in 1990 to 57% in
2001. Since the sample size for the United States in 2001 and for Canada since 1984 was small,
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 34 -
although still in the order of some published studies (e.g., Kalmijn, 1991a; 1991b; Halpin and
Chan, 2003), we should exercise caution in interpreting these trends.
Results from the log-multiplicative layer effect models show that the level of educational
homogamy among newlywed first marriages changed little in both the United States and Canada
from 1970 to 1990. Since then educational homogamy decreased in the United States, but rose
substantially in Canada. The normalized Φ parameter showing the relative strength of
educational homogamy (Xie, 1992) was 0.52 in 1970, 0.54 in 1980, 0.52 in 1990, and 0.41 in
2001. The corresponding numbers in Canada are 0.49, 0.44, 0.45, and 0.60. Again, the
examination of change during the 1990s was based on a small sample in both countries.
Since our sample size for the 1990s is too small for first marriages, we do not conduct further
analyses on heterogeneous change by education level.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 35 -
Table A.1 Assortative marriage on educational attainment for newlyweds, United States,
1970, 1980, 1990 and 1992, and 2001
Husband’s years of schooling
Wife’s years of schooling < 9 9 to 11 12 13 to 15 > = 16 Total Sum of
diagonals
percent
1970
< 9 1.7 1.2 0.9 0.1 0.1 7.4
9 to 11 1.7 5.0 5.9 1.1 0.2 10.4
12 1.8 6.5 26.1 9.8 3.9 48.0
13 to 15 0.2 0.8 4.7 8.0 6.2 19.9
> = 16 0.1 0.2 1.3 2.8 9.9 14.2
Total 5.4 13.7 38.9 21.8 20.2 100.0 50.8
Sample size 16,697
1980
< 9 1.2 0.9 0.7 0.2 0.1 6.1
9 to 11 1.0 4.6 5.6 1.0 0.2 9.2
12 1.0 5.5 26.5 8.5 3.4 44.8
13 to 15 0.2 0.8 6.6 9.0 6.0 22.7
> = 16 0.0 0.1 1.7 3.5 11.9 17.3
Total 3.4 12.0 41.1 22.1 21.5 100.0 53.1
Sample size 76,848
1990-92
< 9 1.5 0.5 0.6 0.2 0.1 2.8
9 to 11 0.6 3.8 4.5 0.6 0.1 9.6
12 1.2 3.2 23.2 4.8 3.1 35.6
13 to 15 0.6 1.2 8.9 10.3 5.8 26.8
> = 16 0.1 0.3 4.0 4.9 16.1 25.3
Total 4.0 8.9 41.2 20.7 25.2 100.0 54.9
Sample size 1,596
2001
< 9 2.1 0.5 0.4 0.0 0.0 3.0
9 to 11 0.2 3.5 4.4 2.0 0.4 10.5
12 0.4 1.9 11.9 4.4 4.3 22.7
13 to 15 0.0 1.3 6.3 6.1 4.1 17.8
> = 16 0.0 1.0 8.9 7.4 28.6 45.9
Total 2.7 8.1 31.9 19.9 37.3 100.0 52.2
Sample size 447
Sources: University of Minnesota, Minnesota Population Center, 1970 and 1980 U.S. Census public use microdata files, and 1990 and 1992 June
Current Population Survey; United States Department of Commerce, Survey of Income and Program Participation, 2001 Panel,
Wave 2.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 36 -
Table A.2 Assortative marriage on educational attainment for newlyweds, Canada, 1971,
1984, 1990 and 2001
Husband’s years of schooling
Wife’s years of schooling (1) (2) (3) (4) (5) Total Sum of
diagonals
percent
1971
(1) < Grade 9 6.8 5.1 0.8 0.5 0.2 13.5
(2) Some high school 7.0 18.6 7.6 3.4 1.8 38.5
(3) High-school graduation 2.5 9.6 9.7 4.5 2.9 29.0
(4) Some post-secondary 0.6 2.1 2.6 4.3 3.5 13.0
(5) University degree + 0.1 0.3 0.4 1.0 4.2 6.0
Total 17.0 35.6 21.2 13.6 12.6 100.0 43.7
Sample size 4,346
1984
(1) < Grade 9 0.6 0.8 1.0 0.0 0.2 2.5
(2) Some high school 1.7 3.1 10.3 2.3 0.7 18.2
(3) High-school graduation 0.8 0.9 23.6 6.9 4.4 36.6
(4) Some post-secondary 0.2 0.9 9.7 9.5 6.5 26.8
(5) University degree + 0.2 0.4 3.7 3.1 8.6 15.8
Total 3.5 6.2 48.3 21.8 20.3 100.0 45.4
Sample size 514
1990
(1) < Grade 9 0.7 0.1 0.4 0.8 0.0 1.9
(2) Some high school 0.5 3.8 4.1 2.3 0.5 11.1
(3) High-school graduation 0.9 4.9 8.4 10.4 1.5 26.0
(4) Some post-secondary 0.3 6.5 8.1 19.6 6.1 40.6
(5) University degree + 0.9 0.4 1.4 6.2 11.6 20.4
Total 3.1 15.7 22.3 39.3 19.7 100.0 44.0
Sample size 450
2001
(1) < Grade 9 0.1 0.2 0.0 0.3 0.0 0.6
(2) Some high school 0.1 1.7 0.6 0.6 0.0 3.0
(3) High-school graduation 0.3 1.4 7.2 8.3 0.5 17.6
(4) Some post-secondary 0.6 2.8 11.1 24.9 6.1 45.4
(5) University degree + 0.1 0.3 3.4 6.4 23.3 33.4
Total 1.2 6.3 22.2 40.4 29.9 100.0 57.2
Sample size 522
Sources: Statistics Canada, 1971 Census of Canada Public Use Sample Tape of Families and 1990 and 2001 General Social
Survey; University of Western Ontario, Population Studies Centre, 1984 Canadian Fertility Survey.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 37 -
References
Balakrishnan, T.R. (Principal investigator), Karol J. Krotki, and Evelyne Lapierre-Adamcyk.
1988. Canadian Fertility Survey, 1984. (Machine readable data file.) London, Ont.:
Population Studies Centre, Department of Sociology, University of Western Ontario.
August 1988.
Duncan, Roderick. 2003. “Does sex and the city predict the future of marriage?” Challenge.
46, 3: 73–88.
Esping-Andersen, Gøsta. 2007. “Sociological explanations of changing income distributions.”
American Behavioral Scientist. 50: 639–658.
Fortin, Nicole M., and Tammy Schirle. 2006. “Gender dimensions of changes in earnings
inequality in Canada.” In Dimensions of Inequality in Canada. Jonathan R. Kesselman
and David A. Green (eds.). Vancouver, B.C.: UBC Press.
Gelissen, John. 2004. “Assortative mating after divorce: a test of two competing hypotheses
using marginal models.” Social Science Research. 33,3: 361–384.
Goldthorpe, John H. 2000. On Sociology: Numbers, Narratives, and the Integration of Research
and Theory. Oxford: Oxford University Press.
Goldthorpe, John H. 1987. Social Mobility and Class Structure in Modern Britain. Oxford:
Clarendon Press.
Goodman, Leo A., and Michael Hout. 1998. “Statistical methods and graphical displays for
analyzing how the association between two qualitative variables differs among countries,
among groups, or over time: A modified regression-type approach.” Sociological
Methodology. 28, 1: 175–230.
Haberman, Shelly J. 1979. Analysis of Qualitative Data: New Developments. Vol. 2. New York:
Academic Press.
Halpin, Brendan, and Tak Wing Chan. 2003. “Education homogamy in Ireland and Britain:
Trends and patterns.” British Journal of Sociology. 54, 4: 473–495.
Hout, Michael. 1983. Mobility Tables. Newbury Park, CA: SAGE Publications.
Hyslop, Dean R. 2001. “Rising U.S. earnings inequality and family labor supply: The covariance
structure of intrafamily earnings.” American Economic Review. 91, 4: 755–777.
Kalmijn, Matthijs. 1998. “Intermarriage and homogamy: Causes, patterns, trends.” Annual
Review of Sociology. 24: 395–421.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 38 -
Kalmijn, Matthijs. 1991a. “Shifting boundaries: Trends in religious and educational homogamy.”
American Sociological Review. 56, 6: 786–800.
Kalmijn, Matthijs. 1991b. “Status homogamy in the United States.” American Journal of
Sociology. 97, 2: 496–523.
Kenworthy, Lane. 2004. Egalitarian Capitalism. New York: Russell Sage Foundation.
Mare, Robert D. 1991. “Five decades of educational assortative mating.” American Sociological
Review. 56, 1: 15–32.
Oppenheimer, Valerie K. 1997. “Women’s employment and the gain to marriage: The
specialization and trading model.” Annual Review of Sociology. 23: 431–453.
Oppenheimer, Valerie K. 1994. “Women’s rising employment and the future of the family in
industrial societies.” Population and Development Review. 20, 2: 293–342.
Qian, Zhenchao. 1998. “Changes in assortative mating: The impact of age and education, 1970-
1990.” Demography. 35, 3: 279–292.
Qian, Zhenchao, and Samuel H. Preston 1993. “Changes in American marriage, 1971 to 1987:
Availability and forces of attraction by age and education.” American Sociological
Review. 58,4: 482–495.
Raymo, James M., and Yu Xie. 2000. “Temporal and regional variation in the strength of
educational homogamy.” American Sociological Review. 65, 5: 773–781.
Rogers, Stacy J. 2004. “Dollars, dependency, and divorce: Four perspectives on the role of
wives’ income.” Journal of Marriage and Family. 66, 1: 59–74.
Ruggles, Steven, Matthew Sobek, Trent Alexander, Catherine A. Fitch, Ronald Goeken, Patricia
Kelly Hall, Miriam King, and Chad Ronnander. 2003. Integrated Public Use Microdata
Series: Version 3.0. (Machine readabale database.) Minneapolis, MN: Historical Census
Projects, University of Minnesota, Minnesota Population Center (producer and
distributor), 2004. Download from http://www.ipums.org.
Schoen, Robert, and Y. Alice Cheng. 2006. “Partner choice and the differential retreat from
marriage.” Journal of Marriage and Family. 68, 1: 1–10.
Schwartz, Christine R., and Robert D. Mare. 2005. “Trends in educational assortative marriage
from 1940 to 2003.” Demography. 42, 4: 621–646.
Smits, Jeroen. 2003. “Social closure among the higher educated: Trends in educational
homogamy in 55 countries.” Social Science Research. 32, 2: 251–277.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 39 -
Smits, Jeroen, Wout Ultee and Jan Lammers. 2000. “More or less educational homogamy? A test
of different versions of modernization theory using cross-temporal evidence for 60
countries.” American Sociological Review. 65, 5: 781–788.
Smits, Jeroen, Wout Ultee and Jan Lammers. 1998. “Educational homogamy in 65 countries: An
explanation of differences in openness using country-level explanatory variables.”
American Sociological Review. 63, 2: 264–285.
Statistics Canada. 2002. General Social Survey, Cycle 15: Family History. (Machine readable
data file.) Ottawa: Statistics Canada.
Statistics Canada. 1991. General Social Survey, Cycle 5: Family and Friends. (Machine readable
file.) Ottawa: Statistics Canada.
Statistics Canada. 1975. Census of Canada, 1971. Public Use Sample Tape of Families. National
File (machine readable data file). Ottawa: Statistics Canada.
Sweeney, Megan M. and Maria Cancian. 2004. “The changing importance of white women’s
economic prospects for assortative mating.” Journal of Marriage and Family. 66, 4:
1015–1028.
Sweeney, Megan M. 2002. “Two decades of family change: the shifting economic foundations of
marriage.” American Sociological Review. 67, 1: 132–147.
Ultee, Wout C., and Ruud Luijkx. 1990. “Educational heterogamy and father-to-son
occupational mobility in 23 industrial nations: General societal openness or
compensatory strategies of reproduction?” European Sociological Review. 6, 2: 125–149.
United States Deptartment of Commerce, Bureau of the Census. 2004 Survey of Income and
Program Participation (SIPP) 2001 Panel. (Wave 2 Topical Module Microdata file)
(Computer file). ICPSR release. Ann Arbor, MI: Inter-university Consortium for Political
and Social Research (distributor). 2004.
Wong, Raymond S. 2003. “To see or not to see: Another look at research on temporal trends and
cross-national differences in educational homogamy.” Taiwanese Journal of Sociology.
31: 47–91.
Xie, Yu. 1992. “The log-multiplicative layer effect model for comparing mobility tables.”
American Sociological Review. 57, 3: 380–395.
Yamaguchi, Kazuo. 1987. “Models for comparing mobility tables: Toward parsimony and
substance.” American Sociological Review. 52, 4: 482–494.
Analytical Studies – Research Paper Series Statistics Canada – Catalogue no. 11F0019MIE, no. 299
- 40 -
... As of 2021, 93% of residents in Canada used the internet (including through smartphones), which was comparable to the corresponding rates in countries such as Germany (91%), the United States (92%), and the United Kingdom (97%) (International Telecommunication Union, 2023). In recent decades, similar to many other Western countries (Qian, 2017;Van Bavel et al., 2018), the gender gap in education has reversed from favoring men to favoring women in Canada, and it has become more common for women to marry down rather than up in education (Guppy & Luongo, 2015;Hou & Myles, 2008). Meanwhile, Canada is an immigrant-receiving country, with immigrants accounting for about 25% of the population according to the 2021 Census (Statistics Canada, 2023). ...
... which is consistent with the increasing prevalence of women paired with less-educated men in Canada and many other countries (Hou & Myles, 2008;Van Bavel et al., 2018). By contrast, when two partners differ in the other three characteristics, the man rather than the woman tends to have a higher status. ...
Article
Full-text available
Objective: This study examines, for the first time in Canada, the relationship between how different-sex couples meet and assortative mating on education, race, nativity, and age. Background: Extending research on how the likelihood of heterogamy differed between offline and online dating, this study disentangles the implications of institutional and third-person influences from those of online dating for configuring the patterns of heterogamy and gender asymmetry in assortative mating. Method: Data from a 2018 national survey are analyzed using (multinomial) logit models. Results: Educational heterogamy and nativity heterogamy are higher, but age heterogamy appears lower, in online than offline dating. Next, specific channels of offline dating - formal institutions, social ties, and other channels - are distinguished and compared with online dating. Online dating tends to entail higher educational and nativity heterogamy (vs. meeting through formal institutions), higher racial and nativity heterogamy but lower age heterogamy (vs. meeting through social ties), and higher educational heterogamy (vs. meeting through other offline channels). Further considering gender asymmetry shows that online dating is associated with higher educational hypergyny (more-educated man, less-educated woman) than meeting through other offline channels; higher nativity hypogyny (immigrant man, native-born woman) than meeting offline (overall, formal institutions, social ties); and lower age hypergyny (older man, younger woman) than meeting off-line through social ties. Conclusion: The findings help untangle the roles of institutional, social, and digital forces in shaping assortative mating. They illustrate the importance of leveraging theoretically informed comparisons to understand how online and offline dating configures assortative mating and its gender-asymmetric patterns.
... The relative distributions (RD) framework described below draws on Handcock and Morris' description (Hou & Myles, 2008;Hout, 2012). Let Y 0 be a random variable representing a given measurement for a population (e.g., income). ...
... This suggests that educational attainment matters for earnings (leads to an improvement in earnings) and that illnesses in childhood may inhibit educational potential. This finding builds on the literature which has emphasized education as an important pathway for future socioeconomic outcomes as cognitive skills are developed and professional credentials are acquired, all of which influence employment and earnings in later life (Hou & Myles, 2008;Hout, 2012). In addition, the literature finds that childhood conditions do impact early as well as later educational attainment, thus suggesting that childhood conditions might indirectly impact future outcomes by imposing a limit on educational attainment (Case et al., 2005;Contoyannis & Dooley, 2010). ...
Article
Full-text available
Research on the long-term effects of health in early life has predominantly relied on parametric methods to assess differences between groups of children. However, this approach leaves a wealth of distributional information untapped. The objective of this study was to assess distributional differences in earnings and mental health in young adulthood between individuals who suffered a chronic illness in childhood compared to those who did not using the non-parametric relative distributions framework. Using data from the Panel Study of Income Dynamics, we find that young adults who suffered a chronic illness in childhood fare worse in terms of earnings and mental health scores in adulthood, particularly for individuals reporting a childhood mental health/developmental disorder. Covariate decompositions suggest that chronic conditions in childhood may indirectly affect later outcomes through educational attainment: had the two groups had similar levels of educational attainment, the proportion of individuals with a report of a chronic condition in childhood in the lower decile of the relative earnings distribution would have been reduced by about 20 percentage points. Findings may inform policy aimed at mitigating longer run effects of health conditions in childhood and may generate hypotheses to be explored in parametric analyses.
... It has, for example, been shown that women's and men's education is an important factor in the division of domestic work (Hersch & Stratton, 1994). However, the question arises as to whether and, if so, how the role of educational attainment has been changing given the increasing shares of women with tertiary education (in the population of women) and of better educated women who marry men with a lower level of education than theirs (Hou & Myles, 2008;Nitsche et al., 2018). Gender norms have also attracted economists' attention in recent years, following the theoretical foundations established by Akerlof and Kranton (2010) and Kranton (2016). ...
Article
Full-text available
We examine the relationship between female contribution to household income and the division of housework between the partners, while accounting for their attitutes towards gender roles. We use data from the “Generation and Gender Survey” for Poland: a country where both employment rates of women and their involvement in housework are high, men and women work long hours, and labour market regulation and policies are unsupportive of work− family balance. We find that the female share of total household income is negatively related to women’s heavy involvement in housework. The direction of this relationship does not change when women earn more than their partners, so there is no support for the gender deviance neutralization hypothesis. We also find that individual gender norms matter for women’s involvement in unpaid work at home, and the uncovered link between the female share of household income and inequality between the partners in the division of housework. Women from less traditional households are more likely to share housework equally with their partners. Among couples with traditional gender attitudes, the female contribution to household income is not related to the division of housework. We conclude that narrowing gender pay gaps may be an important step towards more equality not only at work but also at home.
... It has, for example, been shown that women's and men's education is an important factor in the division of domestic work (Hersch & Stratton, 1994). However, the question arises as to whether and, if so, how the role of educational attainment has been changing given the increasing shares of women with tertiary education (in the population of women) and of better educated women who marry men with a lower level of education than theirs (Hou & Myles, 2008;Nitsche et al., 2018). Gender norms have also attracted economists' attention in recent years, following the theoretical foundations established by Akerlof and Kranton (2010) and Kranton (2016). ...
Article
Full-text available
We examine the relationship between female contribution to household income and the division of housework between the partners, while accounting for their attitutes towards gender roles. We use data from the “Generation and Gender Survey” for Poland: a country where both employment rates of women and their involvement in housework are high, men and women work long hours, and labour market regulation and policies are unsupportive of work− family balance. We find that the female share of total household income is negatively related to women’s heavy involvement in housework. The direction of this relationship does not change when women earn more than their partners, so there is no support for the gender deviance neutralization hypothesis. We also find that individual gender norms matter for women’s involvement in unpaid work at home, and the uncovered link between the female share of household income and inequality between the partners in the division of housework. Women from less traditional households are more likely to share housework equally with their partners. Among couples with traditional gender attitudes, the female contribution to household income is not related to the division of housework. We conclude that narrowing gender pay gaps may be an important step towards more equality not only at work but also at home.
Thesis
Full-text available
This research combines three studies analyzing son preference, educational assortative mating patterns, and intergenerational co-residence patterns among Indian immigrant families in the United States of America, United Kingdom, Canada, and Spain, using IPUMS datasets and United Kingdom labor force data (UKLFS). The results of Study 1 suggest that lower educational attainment is associated with a stronger tendency to favor sons among the Indian immigrant groups studied in the USA, Canada, and Spain. The findings of Study 2 indicate that Indian male immigrants in the USA and Canada tend to marry hypogamously, use their "immigration status" to attract more educated female partners from their home country, while this is not the case in Spain. Finally, the results of Study 3 suggest that in all four countries (USA, Canada, UK and India); age, education level and occupation significantly impact patrilocal residence. The findings provide insight into some of the key patriarchal cultural norms and practices of the Indian population both in India and abroad, and highlight the importance of education and economic autonomy in promoting gender equity and empowerment among immigrant populations.
Thesis
Full-text available
This research combines three studies analyzing son preference, educational assortative mating patterns, and intergenerational co-residence patterns among Indian immigrant families in the United States of America, United Kingdom, Canada, and Spain, using IPUMS datasets and United Kingdom labor force data (UKLFS). The results of Study 1 suggest that lower educational attainment is associated with a stronger tendency to favor sons among the Indian immigrant groups studied in the USA, Canada, and Spain. The findings of Study 2 indicate that Indian male immigrants in the USA and Canada tend to marry hypogamously, use their "immigration status" to attract more educated female partners from their home country, while this is not the case in Spain. Finally, the results of Study 3 suggest that in all four countries (USA, Canada, UK and India); age, education level and occupation significantly impact patrilocal residence. The findings provide insight into some of the key patriarchal cultural norms and practices of the Indian population both in India and abroad, and highlight the importance of education and economic autonomy in promoting gender equity and empowerment among immigrant populations. Keywords: Indian immigrants, son preference, educational assortative mating, post-marital co-residency patterns, patrilocal residence, matrilocal residence, cultural norms, gender roles.
Chapter
In this chapter, we briefly review methodological issues in research on assortative mating. The motivation for this chapter is recognition that our understanding of trends or consequences of assortative
Chapter
In this chapter, we discuss two important conceptual issues regarding assortative mating, one about why marriages tend to be homogamous and the other about why patterns of assortative mating may change.
Article
In this paper, we propose a new bootstrap procedure for the empirical copula process. The procedure involves taking pseudo samples of normalized ranks in the same fashion as the classical bootstrap and applying small perturbations to break ties in the normalized ranks. Our procedure is a simple modification of the usual bootstrap based on sampling with replacement, yet it provides noticeable improvement in the finite sample performance. We also discuss how to incorporate our procedure into the time series framework. Since nonparametric rank statistics can be treated as functionals of the empirical copula, our proposal is useful in approximating the distribution of rank statistics in general. As an empirical illustration, we apply our bootstrap procedure to test the null hypotheses of positive quadrant dependence, tail monotonicity, and stochastic monotonicity, using U.S. Census data on spousal incomes in the past fifteen years.
Article
Examines educational attainment as a dimension of assortative mating. Barriers to marriage between persons with unequal amounts of formal schooling increased between the 1930s and the present. These increases may be the result of trends in average educational attainment, age at leaving school, and age at marriage. The degree to which schools affect the selection of marriage partners is dictated by the degree to which leaving school and marriage occur closely together and by the educational attainments of marriage partners. Variation in the average age at leaving school and marriage and in educational attainment induce variation in educational assortative mating. Trends in age at marriage affect both the structure of marriage and inequality within and between generations.
Article
According to classical works on social stratification, status homogamy can be regarded as an indicator of the "openness" of society. In contrast to previous approaches, this article treats marriage choice as a multidimensional phenomenon and makes a distinction between ascriptive- and achievement-oriented characteristics. Ascriptive status homogamy is measured by the similarity of spouses with respect to their fathers' occupational class, while the achieved dimension of status homogamy is measured by the similarity of spouses' educational attainment. Multivariate log-linear models are used to explore the relative importance of these factors for the choice of a spouse, and the article tests the hypothesis that there has been a transition from ascription to achievement in patterns of marriage selection. This study first demonstrates empirically that previously conducted synthetic cohort analyses of educational homogamy suffer from selection biases and then, using the Occupational Change in a Generation (OCG...
Article
We describe recent changes in propensities to marry according to the age and educational attainment of potential spouses. Relating actual marriage to the population of men and women at risk enables us to distinguish between changes in the availability of eligible partners and changes in the force of attraction between men and women in particular categories. The procedure is applied to data from 1973, 1980, and 1988 Current Population Surveys. Multivariate analysis suggests that the sharp declines in marriage rates between 1972 and 1979 were not highly differentiated by age or education for either men or women, but that the smaller declines between 1979 and 1987 were highly concentrated among younger women. Age and education homogamy increased during the latter period. Including cohabiting unions in the definition of marriage reduces the magnitude of the declines but does not alter their essential patterns.
Article
I assess whether intermarriage between Protestants and Catholics has increased over the course of the twentieth century and, if it has, whether the declining salience of religious boundaries has been accompanied by a rising importance of educational boundaries in spouse selection. By analyzing a set of national surveys that were conducted between 1955 and 1989 and using a research design that separates the effects of period and duration of marriage, I examine changes over a longer period of time than any previous study has done. Multivariate loglinear analyses show that intermarriage between Protestants and Catholics has increased dramatically since the 1920s, while intermarriage between different educational groups has decreased. Currently, the social boundaries that separate educational groups seem to be stronger than the boundaries that separate Protestants and Catholics. In addition, there is some evidence that interfaith marriages have become increasingly homogamous with respect to education, suggesting that education has replaced religion as a factor in spouse selection.