PreprintPDF Available

Working Paper Series (E) Educational Assortative Mating in Japan: Evidence from the 1980-2010 Census Setsuya Fukuda 1

Authors:
Preprints and early-stage research may not have been peer reviewed yet.

Abstract

The existing results for Japan on the general trends in educational assortative mating are mixed largely due to data constraints. This study aims to describe the trends in educational assortative mating in Japan using four decennial population censuses conducted in 1980-2010 and to discuss their social and demographic implications. Our simple log-linear model shows that the strength of the association for educational homogamy remained constant between 1980 and 2000, and dropped significantly in 2010. From 1980 to 2010, the strength of the association for female educational hypergamy decreased, while, contrary to previous findings, both the share and the likelihood of female educational hypogamy increased. Our study is the first to use census data to provide clear findings on the trends in educational assortative mating in Japan and points to the emergence of new marriage behaviors which are, in fact, consistent with the observed shifts in educational assortative mating across the globe.
Working Paper Series (E)
Hibiya Kokusai Building 6F, 2-2-3 Uchisaiwaicyo, Chiyoda-ku, Tokyo 100-0011
http://www.ipss.go.jp
No.29
Educational Assortative Mating in Japan:
Evidence from the 1980-2010 Census
Setsuya Fukuda
Shohei Yoda
Ryohei Mogi
January 2019
http://www.ipss.go.jp/publication/e/WP/IPSS_WPE29.pdf
The views expressed herein are those of the
authors and not necessarily those of the
National Institute of Population and Social
Security Research, Japan.
1
Educational Assortative Mating in Japan: Evidence from the 1980-2010
Census
Setsuya Fukuda1
Shohei Yoda1
Ryohei Mogi2
Abstract
The existing results for Japan on the general trends in educational assortative mating are
mixed largely due to data constraints. This study aims to describe the trends in
educational assortative mating in Japan using four decennial population censuses
conducted in 1980-2010 and to discuss their social and demographic implications. Our
simple log-linear model shows that the strength of the association for educational
homogamy remained constant between 1980 and 2000, and dropped significantly in
2010. From 1980 to 2010, the strength of the association for female educational
hypergamy decreased, while, contrary to previous findings, both the share and the
likelihood of female educational hypogamy increased. Our study is the first to use
census data to provide clear findings on the trends in educational assortative mating in
Japan and points to the emergence of new marriage behaviors which are, in fact,
consistent with the observed shifts in educational assortative mating across the globe.
1 National Institute of Population and Social Security Research, Hibiya Kokusai Building 6th Floor, 2-2-3
Uchisaiwaicho, Chiyoda-ku, Tokyo, 100-0011, Japan
2 Centre d'Estudis Demogràfics, Universitat Autònoma de Barcelona, Carrer de Ca n'Altayó, Edifici E2, Universitat
Autònoma de Barcelona, 08193 Bellaterra / Barcelona
2
Introduction
Marriage plays a central role in forming a family in East Asia. The expansion of the
never-married population is one of the most dramatic demographic changes in the latter half of
the twentieth century in Japan. For example, Japanese census data shows that, in the period of
1970-2015, the proportions of never-married at age 25-29 is increased from 46.5% to 72.7% for
men and from 18.1% to 61.3% for women (NIPSSR, 2017). Vital statistics also shows that, in the
same period, average age at first marriage increased from 26.9 to 31.1 for men and 24.2 to 29.4
for women (NIPSSR, 2017). The trend toward later marriage and growth in the proportion of
men and women who never marry is not only a major cause of Japan’s low fertility rate, but is
also a driving force behind other major family and social changes.
Though correlates of the trend toward later and less marriage are complex, previous
studies showed that women’s education was negatively associated with marriage (Raymo, 2003;
Raymo & Iwasawa, 2005; Retherford, Ogawa, & Matsukura, 2001). The official statistics of the
School Basic Survey, conducted each year by the Ministry of Education, Culture, Sports, Science
and Technology, indicates that women’s tertiary education (Note 1) (ISCED2011 level 5 and
higher) very rapidly expanded since the 1960s in Japan. In the late-1970s, female enrollment
rates to tertiary education caught up with those of men and have been higher than men’s since
then. Although female enrollment to four-year universities among high school graduates is still
slightly lagged behind that of men (55.9% for men and 49.1% for women in 2017), it shows
rapid catch up since the mid-1980s (38.6% for men and 13.7% for women in 1985). The
expansion of female tertiary education is considered as one of the most prominent social changes
relating to the trends of later and less marriage in Japan.
Due to the rapid expansion of female tertiary education, education components in the
3
marriage market has been under tremendous shifts in Japan since the 1980s. Although these
shifts are expected to play an important role in determining the patterns of educational
assortative mating, evidence in Japan is incomplete and existing results are fairly mixed. For
example, previous studies mostly employ the framework of the status homogamy (e.g. Blau and
Duncan, 1967) and typically examine only the trends in educational homogamy as an indicator
of “openness” of social stratification (Miwa, 2007; Raymo & Xie, 2000; Shida, Seiyama, &
Watanabe, 2000; Smits & Park, 2009). Therefore, the trends in educational heterogamy is largely
ignored. Furthermore, previous studies show mixed results on the trends of educational
homogamy as some found no change in the strength of educational homogamy (Shida et al.,
2000), while others found weakening tendencies in educational homogamy (Miwa, 2007; Raymo
& Xie, 2000; Smits & Park, 2009). In addition to these limitations, there is no study which
examined over-time shifts in the strength of educational homogamy as well as educational
heterogamy by educational group. These incompleteness and inconsistency of the existing
findings are primarily resulted from the use of a small sample, which makes it difficult for the
previous studies to draw an accurate picture of time trends in educational homogamy as well as
educational heterogamy in Japan.
Using an extensive dataset of population censuses from over 120 countries (Note 2), the
recent study by Esteve et al. (2016) revealed that given the reversal of gender gap in education,
young people are increasingly more likely to form unions in which wives have higher
educational attainment than their husbands, while decreasing the likelihood of forming the
historical hypergamic pattern. However, in Japan, existing evidence suggests that there is a
strong tendency to avoid female educational hypogamy, particularly among university graduates
(NIPSSR, 2012; Suzuki, 1991). Raymo and Iwaswa (2005) also found that one-fourth to
4
one-third of the decline in marriage rates of junior college and university graduates between
1980 and 1995 could be explained by the increasing numerical difficulty of meeting a potential
husband with the same (or higher) educational qualifications as themselves, which, in turn,
contributing to the growing negative educational gradients in marriage rates in Japan. It is,
therefore, of prime interest to examine whether the patterns of educational assortative mating are
kept relatively unchanged in Japan or the new mating patterns have emerged around the turn of
the century as a result of the continuing expansion of female tertiary education, particularly that
of four-year university education.
The aim of our study is to provide evidence on the trends in educational assortative
mating by using for the first time census data to study educational homogamy/heterogamy in
Japan. For our study, we accessed the individual records of the full census data of four decimal
periods between 1980 and 2010. By relying on census data, we avoid any bias and complexity
that might result from the use of a small sample and are thus able to provide reliable evidence on
both the overall- and the education-specific trends in educational assortative mating in Japan.
In the following section, we first describe our data and methods. Then, we present our
descriptive results. In the final section, we summarize our findings and discuss their implications.
Data
Individual-level data from the decennial Population Census of Japan from 1980 to 2010 are used
to examine educational assortative mating in Japan (Note 3). The census covers all residents of
Japan as of October 1 in each census year. The analytical cases were selected using the following
steps. For the analysis, we first selected married men and women who were living together in a
private household and identified marital pairs using information about each individual’s marital
5
status and relationship to the household head (Note 4). Second, from these identified couples, we
selected only Japanese couples in which the wife was aged 30 to 39. Finally, we transformed the
individual records of these paired couples into dyad records. Using this procedure, we selected
over 25 million Japanese married couples for our analysis (8.6 million dyads in 1980, 6.8 million
dyads in 1990, 5.7 million dyads in 2000, 4.8 million dyads in 2010).
In the census, an individual’s educational attainment is defined as the highest educational
level he or she reached upon graduation, or the level in which he or she is currently enrolled. We
use directly the following four educational categories measured in the census: (1) primary school
/ junior high school (“JHS”: ISCED2011 level 1/2), (2) high school (“HS”: ISCED2011 level
3/4), (3) two-year junior college / technical school / vocational school (“JC/VS”: ISCED2011
level 5), and (4) university and graduate university (“UNI”: ISCED2011 level 6+). Because there
are only a few dyads (less than 0.5%) in which at least one of the spouses was still attending
school, the educational levels of those who were in school are treated as though they had
graduated. Because very few people in Japan marry before they finish their education, we also
assume that the couples’ educational attainment levels at the time of their marriage and at the
time of the census were identical.
However, there are at least two drawbacks to using census data for the analysis of
educational assortative mating in Japan. First, we cannot describe assortative marriage trends for
newlyweds, since there is no information about the date of marriage or the age at marriage.
Therefore, to analyze the trends in assortative mating, we use prevailing marriages observed at
each census rather than newlyweds or recently married. As has often been discussed, trends that
are based on the prevailing marriage patterns are subject to bias due to selective marital
dissolution, educational upgrading after marriage, and remarriage (e.g., Kalmijn, 1994; Mare,
6
1991; Qian, 1998; Raymo & Xie, 2000; Schwartz & Mare, 2005). As divorce is most frequent
among the least educated group in Japan (Raymo, Fukuda, & Iwasawa, 2013), our analysis may
include fewer couples in which the husband or wife has the lowest level of education than other
analyses based on newlyweds.
Nevertheless, prevailing marriages are typically used for other studies of educational
assortative mating using census data (Esteve & Cortina, 2006; Schwartz & Mare, 2005). For the
U.S., it has been shown that trends in educational assortative mating for prevailing and new
marriages are very similar, although trends for prevailing marriages lag those for newlyweds by
several years (Schwartz & Mare, 2005). The length of the lag depends on the size of the age
range and the marital duration used in examining the trends for prevailing marriages. As
marriage duration is longer for older couples, the broader age range results in the longer lag. In
our analysis, we chose couples in which the wife was aged 30 to 39 (Note 5). The average
marriage duration for married women aged 30-39, which was calculated from multiple rounds of
Japanese National Fertility Survey, was 10.5 years in 1982, 10.2 years in 1992, 8.9 years in 2002,
and 8.6 years in 2010. We chose this age range for two reasons. First, because most women in
this age group have completed their education, we can treat educational attainment as a fixed
measure in the analysis. Second, we confirmed that more than 80% of women who eventually
marry before age 50 are observed as “married” at age 30-39. Therefore, our analytical cases are
considered representative of the overall patterns of educational assortative mating for the
subjective female cohorts.
The second limitation associated with using census data is that we cannot distinguish
marriage order in the analysis. Thus, our analysis includes the assortative mating patterns of all
married couples, regardless of how many times the partners have been married. If the patterns of
7
educational assortative mating differ between couples in a first marriage and in a remarriage, the
period trends in educational mating would be affected by the share of couples who are remarried.
According to the vital statistics, the percentages of all registered marriages that were remarriages
increased between 1970 and 2010 from 10% to 20% for men and from 5% to 15% for women
(Ministry of Health, Labour and Welfare, 2017). Thus, the percentage of remarriages among all
marriages (i.e., couples in which at least one spouse is remarried) more than doubled from 11.1%
in 1970 to 25.6% in 2010. We, therefore, assume that in our data the share of marriages that were
remarriages was larger in the later periods. However, there is no evidence yet in Japan that how
educational mating patterns of remarried couples differ from those of couples in their first
marriage. Therefore, we have no way to assess the impact of the heterogeneities that may have
been caused by the increasing shares of remarried couples in the later periods.
Method
Our analysis is carried out in two steps. In the first step, we present a descriptive analysis of
trends over time in husbands’ and wives’ educational distributions and educational pairings
during the 1980-2010 period. In the second step, log-linear models are employed to examine the
strength of educational assortative mating patterns by controlling for the distributions of
husbands’ and wives’ educational components in each period. We are aware of the recent
critiques on the use of log-linear models in the study of assortative mating for the inability to
estimate people’s preferences for mates (Logan, 1996). But since the log-linear models are most
frequently used in previous studies examining patterns and strength of educational assortative
mating (Lichter & Qian, 2019; Schwartz, 2013), we present our results accordingly so that
readers can interpret the results in the same line with previous findings. The original data of the
8
analysis which is on the distributions of wives’ and husbands’ educational pairings in each period
are provided in the appendix. The reader is invited to review these data and use them in the
replication and/or modification of the model.
The log-linear model employed in the analysis is explained in the following. In the
saturated model, the expected value of each educational pairing in each period is described as


 =+
+
+
+

 +

 +

 +

,
where each character stands for W: wife’s educational level, H: husband’s educational level, and
P: census year. The most important parameter in this study is 
, which shows the periodical
change in the association of particular educational parings, which is independent of the
distributions of wives’ and husbands’ educational components in each period. Most of the
previous research on assortative mating relied on the most parsimonious model, which
approximates the observed cell with as few parameters as possible using likelihood statistics or
AIC/BIC. However, our purpose is to draw as precise a picture as possible of the trends in the
associations of educational assortative mating.
To achieve our analytical goal, we use both the simple model and the saturated model in
our analysis. In the simple model, we estimated the overall trends in the strength of educational
homogamy, hypergamy, and hypogamy, irrespective of educational levels, by inducing a design
matrix for the parameter 
, which is expressed as follows;
[Table 1]
9
With our interpretation of the simple model, we conclude our prolonged discussion about the
general trends in educational homogamy and heterogamy in Japan for the 1980-2010 period.
Next, we use the saturated model for our understandings of the education-specific trends
in educational matings. The saturated model can fully replicate the observed educational pairings,
while estimating 
 by controlling for marginal distributions of educational pairings. The
authors are aware that this approach is unusual in the research of assortative mating where
unsaturated log-linear models are preferable since they have simpler interpretations. But after
trying several design matrices for 
, to construct a parsimonious model, we found that there
is no design matrix that can satisfactorily approximate the trends in the strength of
education-specific homogamy, hypergamy, and heterogamy. Therefore, we decided to use the
saturated model and interpret the behavioral patterns of educational assortative mating over time.
Results
Descriptive Analysis
Figure 1 shows the distributions of husbands’ and wives’ educational attainment levels in our
analytical cases. Reflecting on the growing trends in tertiary education in Japan, the figure shows
a clear trend toward educational upgrading among both husbands and wives. The shares of
individuals with tertiary education (i.e., at the junior college / vocational school (JC/VS) or
university (UNI) level) increased over the study period, while the shares of those with primary or
secondary education (i.e., at the junior high school (JHS) or high school (HS) level) decreased in
the recent period. However, in line with the gender gap in university enrollment in Japan,
husbands are more likely than wives to have a university degree in all the study periods. After
1980, the percentage of wives with a university degree slowly started to catch up to that of men.
10
Thus, the gender gap among university graduates has been narrowing, particularly after 2000.
Wives are much more likely than husbands to have a JC/VS level of education. The
JC/VS category consists of different kinds of schools and the gender compositions of these
schools differ substantially. According to the School Basic Survey, only 1-2% of male HS
graduates enroll in junior college. We can, therefore, assume that the majority of men in the
JC/VS category received a degree from a vocational school (specialized training college) or a
college of technology. The National Fertility Surveys have found that 60% to 70% of women
with an educational qualification in the JC/VS category are junior college graduates. The JC/VS
category also encompasses a wide range of school types, including schools that offer
job-oriented training courses and schools that provide liberal arts education. Moreover, the types
of individuals enrolled in the specific courses offered by these schools differ by sex and by year.
It is thus important to keep in mind that the definition of educational homogamy or heterogamy
for men and women in this group is more ambiguous than for other education categories.
[Figure 1]
Next, we turn our attention to the patterns of educational mating. First, the educational
homogamy trends over time are presented in Figure 2. It is clear that the overall proportion of
couples in homogamous marriages decreased over the study period. The most frequent form of
educational homogamy is a marriage of two high school graduates. However, the proportion of
homogamous couples formed by high school graduates is decreasing rapidly by 2010. The
observed decline in homogamy among high school graduates is related to changes in the
educational composition of the study sample as shown in Figure 1. Although the share of junior
11
high school graduates who were in a homogamous marriage was relatively high in 1980, it
declined sharply after 1990 and had fallen to very low levels by 2000 and 2010. Similar to the
homogamous marriage patterns observed among high school graduates, the decline in the share
of junior high school graduates in such marriages is attributed in part to the compositional
change in the shares of wives and husbands with a junior high school degree only. However, the
decline in homogamy among these educational groups may also be influenced by changes in
their mating behaviors. We will examine whether this is the case using log-linear models in the
following section.
[Figure 2]
By contrast, the shares of JC/VS and university graduates in homogamous marriages increased
continuously from 1980 onward. In particular, we see that the percentage of university graduates
in homogamous marriages increased substantially from 2000 to 2010 (+ 5.4 % points). In the
next section, we examine whether this increase in homogamy among university graduates is
attributed to changes in educational attainment levels or to changes in behavioral patterns.
Next, we describe the trends in educational exogamy shown in Figure 3. Both
hypergamy and hypogamy are defined from the perspective of the woman, and their shares in
total marriages are shown by the wife’s education. The overall share of hypergamous marriages,
which is typically high in societies with a large gender gap in educational attainment (Blossfeld,
2009), increased from 1980 to 2000, but decreased after 2000. When we look at the educational
differentials in hypergamous marriages, we see that the share of hypergamous couples in which
the wife had a junior high school degree only decreased over time, while the share of
12
hypergamous couples in which the wife had a JC/VS degree increased. These trends seem to
reflect the changes in the educational attainment levels of wives and husbands as shown in
Figure 1.
On the other hand, forms of educational hypergamy in which the wife had a high school
degree increased moderately between 1980 and 2000, but had declined substantially by 2010.
These trends are likely to be affected by the changes in the educational components of wives
with a high school degree. However, as the proportion of men who were highly educated (JC/VS
+ university) increased continuously throughout the observation period (see Figure 1), the
availability of highly educated men for women with a high school degree should have been
greater in the later period. Hence, the decline in hypergamous marriage among women with a
high school degree might be related to changes in the strength of the matching. This question is
investigated further in the following section.
[Figure 3]
Next, we describe the trends in female educational hypogamy in Figure 3. Female hypogamy is
an atypical marriage pattern, particularly in gender-traditional societies where male bread-winner
ideology plays an important role in a society (Blossfeld, 2009). Contrary to our expectation
derived from previous findings (NIPSSR, 2012; Raymo & Iwasawa, 2005; Suzuki, 1991), our
descriptive analysis shows that the percentage of couples in hypogamous marriages increased
from 1980 onward, and had reached 20% by 2010. Especially from 2000 to 2010, the proportion
of women in hypogamous marriages who had a tertiary education (e.g., a JC/VS or a university
degree) increased substantially. This finding is consistent with the global trends in the
13
educational mating found by Esteve et al. (2016). Moreover, when we looked more closely at our
data, we found that this increase in the share of women in hypogamous marriages who were
JC/VS or university graduates was brought by men with a high school education. While the share
of husbands with a high school degree was declining in 2000-2010 (see Figure 1), this increase
in the share of female hypogamous marriages between a highly educated wife and a husband
with a high school degree is likely to be attributed to the increasing strength of the associations
for this match.
In sum, our descriptive analysis shows relatively large changes in the patterns of
educational mating between 2000 and 2010. However, these changes were influenced by changes
in the educational composition of our sample. Thus, these descriptive results do not provide a
full illustration of the changes in matching patterns regarding the partner’s educational level. In
the following section, we employ log-linear models to examine the strength of the association
between the educational assortative mating by controlling for the compositional changes in
educational distributions of wives and husbands.
Log-linear Models
Figure 4 shows the association parameters estimated by the simple model. The overall trends in
the strength of educational homogamy, hypergamy, and hypogamy are shown, irrespective of
educational level. The association parameters are calculated as 
 +

 in the model
equation (1). In the interpretation, the larger (or smaller) the association parameter is, the higher
(or lower) the odds of the given observed educational pairing are. When the association
parameter equals zero, the strength of the associations between the corresponding educational
pairing is considered as random, given the educational attainment levels of wives and husbands
14
at each period. The association parameter is estimated after controlling for the educational
distributions of wives and husbands at each period. Therefore, this parameter can be understood
as the strength of the matching or, at a lesser extent, the strength of the preferences for a
particular educational pairing.
[Figure 4]
As is shown in Figure 4, the association of educational homogamy is stronger than those of
educational heterogamy in Japan. Although Figure 2 indicates that its share declined, the strength
of educational homogamy was stable in the 1980-2000 period and had weakened only slightly by
2010. While the likelihood of educational hypergamy was as high as the likelihood of
educational homogamy in 1980, it had declined significantly by 1990 and by 2010. Although the
likelihood of educational hypergamy tends to be high in gender-traditional societies in which
women’s university enrollment lags behind that of men (Blossfeld, 2009), we find that this
preference was negative in 2010. This suggests that in Japan, the likelihood of hypergamy is now
lower than random matching. On the other hand, the likelihood of educational hypogamy was the
lowest among other pairings and was less observable than random marriages. However, this form
of marriage became less and less stigmatized after 1980. In 2010, the strength of educational
hypogamy got much close to that of educational hypergamy. Our analysis based on the simple
model indicates that the trends in the association of educational mating in Japan have been
surprisingly consistent with global trends in educational assortative mating, including with the
decline in female hypergamy and the increase in female hypogamy found by Esteve et al. (2016).
Next, the association parameters from the saturated model shown in Figure 5 are used to
15
examine the education-specific trends in the strength of educational mating. The parameter
estimates for Figure 5 are shown in the appendix. As can be seen in Figure 5, the trends in the
strength of educational assortative mating differ by educational level. The likelihood of
educational homogamy was highest among couples with JHS education, followed by among
couples with university education. However, in both groups, the strength of homogamy leveled
off or declined only slightly from 1980 to 2000, and then declined relatively sharply from 2000
to 2010. This pattern indicates that the increase in the share of university graduates in
homogamous marriages shown in Figure 2 is entirely due to the quantitative changes in the
educational distribution – e.g., to the increase in the proportion of married people who were
university graduates – rather than to changing patterns of educational mating. In fact, Figure 5
indicates that the increasing share of university graduates in homogamous marriages shown in
Figure 2 is offset by a declining strength of the association.
[Figure 5]
On the other hand, the association parameters of educational homogamy are much lower among
HS and JC/VS graduates than among JHS and UNI graduates. The shares of HS and JC/VS
graduates in homogamous marriages are low because the men and women in these mid-level
educational categories could choose to marry upward or downward. The trends in the association
parameters show that the strength of educational homogamy declined only slightly among HS
and JC/VS graduates (Note 6).
We now turn our attention to the education-specific trends in female hypergamy. The
highest likelihood this pairing can be found among the couples with a JC/VS wife and a UNI
16
husband and followed by the couples with a JHS wife and a HS husband. However, the strength
of the associations of both educational pairings was gradually declining after 1980, and nearly
reached the levels of random marriage in 2010. Among the hypergamous couples, these
educational pairs were the only ones showing positive associations, with the rest of the
educational pairs showing negative associations. The most distant educational pairings e.g., a
couple consisting of a JHS wife and a UNI husband displayed a particularly low tendency for
matching in the hypergamous situation. The strength of the association between a HS wife and a
UNI husband also became weaker after 1990. In other words, we find that husbands with a
university education were becoming more reluctant to marry a less educated wife.
Finally, we interpret the trends in the association parameters of female hypogamy shown
in the right panel of Figure 5. Surprisingly, we see that two educational pairs – namely, a couple
with a HS wife and a JHS husband and with a UNI wife and a JC/VS husband had positive
associations and that their associations even get stronger slightly after 1990. It is a new finding in
Japanese marriage studies that some educationally hypogamous pairs show stronger associations
than random matching. For other hypogamous educational pairs, all of the associations found
have been negative, but the trends in these associations are getting closer to those of random
marriages, which suggests that the social stigma of female hypogamy is weakening in Japan.
Summary and discussion
Japanese studies of educational homogamy have been largely hindered by small samples (Miwa,
2007). Because the trends in educational homogamy have not been clear, hypothesis testing of
the related theories on educational assortative mating has yet to be fully developed in Japan. Our
study uncovered the period trends in educational assortative mating in the 1980-2010 period
17
using micro-data of the Population Census of Japan. In the following, we summarize our findings
and discuss social and demographic implications of the results.
First, our descriptive analysis revealed the trends in educational assortative mating in
Japan in the 1980-2010 period. These trends can be briefly summarized as follows. The shares of
couples in homogamous marriages declined continuously in the 1980-2010 period. However, our
simple log-linear model shows that the strength of the association for educational homogamy
remained constant between 1980 and 2000, and had dropped significantly by 2010. Similarly, the
share of couples in female hypergamous marriages – which is a typical marital pairing in
gender-traditional societies where men are more educated than women (Blossfeld, 2009) –
increased from 1980 to 2000, and had decreased only slightly by 2010. But the estimated
strength of the matching declined constantly between 1980 and 2010. As a result, by 2010,
female hypergamous marriage had become a less preferred form of marriage than random
matching. On the other hand, over the observation period, both the share of couples in female
hypogamous marriages and the strength of the associations increased. For example, the share of
couples in female hypogamous marriages increased from 12% in 1980 to 21% in 2010. Although
the associations of female hypogamy were still negative and the lowest among the educational
pairings over the entire period, by 2010 they are at similar levels as those for female hypergamy.
This descriptive information is invaluable for scholars who work on related studies on trends in
educational pairings in Japan.
In our study, we also described the education-specific trends in the associations of
educational assortative mating, which, in fact, previous studies failed to investigate due to the
small sample size. Our analysis shows the recent decline in the strength of educational
homogamy among university graduates, the overall decline in the strength of female educational
18
hypergamy (except for the rare education pair of a UNI husband and a JHS wife), and the
increasing tendency to engage in educational hypogamy among women with a university degree.
The evidence presented in our paper may point to the emergence of new marriage
behaviors in Japan. Previous studies have shown that women with a university degree are
especially likely to engage in educational homogamy (NIPSSR, 2012; Suzuki, 1991) and that
their first marriage rates are the lowest across educational groups in Japan (Raymo, 2003;
Retherford et al., 2001). Therefore, female university graduates who did not find a similarly
educated male partner tended to stay unmarried rather than to marry a less educated man (Raymo
& Iwasawa, 2005). By showing that in recent years the trends in the associations for educational
hypogamy have been positive and increasing among university-educated women, however, our
results deviate from conventional assumptions about patterns of educational assortative mating in
Japan. One of the possible reasons behind the observed decline in educational homogamy and
the increase in educational hypogamy among university-educated women is a growing economic
inequality among university graduated men. The income gaps within the same age-education
groups have been growing since the mid-1990s (Kambayashi, Kawaguchi, and Yokoyama 2008)
and have been further increasing in the period beyond 2010 in Japan (Higuchi and Sato 2015).
Moreover, these within-group income gaps are particularly pronounced among male university
graduates (Ohta 2010). It, therefore, appears that the economic qualifications of university
graduates and of other less educated men are becoming less distinctive than they were in
previous periods. However, whether our finding that female university graduates are starting to
show a weakening preference for homogamy should require further research examining the trend
more carefully by using a dynamic model that takes into account the decisions of individuals
who have not married (Blossfeld, 2009).
19
Finally, we would like to discuss the social implications of our study. In many medium-
to high-income nations, more women receive a tertiary education than men. Esteve et al. (2016)
has pointed out that traditional female hypergamy is declining, while female hypogamy increases
once the gender gap in tertiary education turns over. In Japan, although the gender gap in tertiary
education has turned over already in the late-1970s, enrollment in four-year universities is still
higher among men than among women. Our analysis shows that, in fact, the similar changes as
the global trends in educational mating are already being observed in Japan at the behavioral
level. In western societies, it is evident that the gender reversals in tertiary education correspond
to changes in social norms, such as the increase in the share of households in which the woman
is the breadwinner, the diffusion of more gender-egalitarian attitudes, and the decline in the
divorce rates of female hypogamous couples (Esteve et al., 2016). Population scholars should be
prepared to investigate whether similar changes are taking place in Japan as the expansion of
female tertiary education continues.
Notes
1 Those include specialized training colleges, two-year junior colleges and four-year universities.
2 Data for Japan is not included in their study.
3 Permission to use the data on the Population Census of Japan was obtained through the Statistics Bureau,
Ministry of Internal Affairs and Communications, on the basis of the Statistics Act, Article 33 (2016.5.10).
4 If the spouses were not sharing a household, the couples were removed from the analysis, as the census did
not ask information about individuals who were living outside the household.
5 We changed the female age range to 25-34, but we got almost the same results as we did in this paper.
6 However, as we mentioned previously, the JC/VS category encompasses various kinds of schools, and their
compositions differ by period and sex. Therefore, a more precise examination is needed to determine whether
the homogamy those in the JC/VS category can really be considered “homogamous” in terms of educational
20
qualifications.
Acknowledgement
This research was supported by Grant-in-Aid for Scientific Research (A), “Marriage, Divorce
and Remarriage in Japan: Trends, Causes and Social Implications” (Principal Investigator: Miho
Iwasawa, JSPS KAKENHI Grant Number 25245061).
References
Blossfeld, H. P. (2009). Educational assortative marriage in comparative perspective. Annual
Review of Sociology, 35(1), 513-530.
Esteve, A., & Cortina, C. (2006). Changes in educational assortative among in contemporary
Spain. Demographic Research, 14(17), 405-428.
Esteve, A., Schwartz, C., van Bavel, J., Permanyer, I., Klesment, M., & García-Román, J. (2016).
The end of hypergamy: Global trends and implications. Population and Development Review,
42(4), 615-625.
Higuchi, Y. and Sato, K. (2015). The Japanese uniqueness compared with other developed
countries using employment and income statistics. Mita business review, 58(1), 15-36. (in
Japanese)
Kalmijn, M. (1994). Assortative mating by cultural and economic occupational status. American
Journal of Sociology, 100(2), 422-452.
Kambayashi, R., Kawaguchi, D., and Yokoyama, I. (2008). Wage distribution in Japan,
1989-2003. Canadian Journal of Economics/Revue canadienne d'économique, 41(4),
1329-1350.
Mare, R. D. (1991). Five decades of educational assortative mating. American Sociological
21
Review, 56(1), 15-32.
Ministry of Health, Labour and Welfare. (2017). Special report on the vital statistics in 2017:
The summary of statistics about marriage. Tokyo: Ministry of Health, Labour and Welfare. (in
Japanese)
Miwa, S. (2007). Long-term trends in status homogamy. In Y. Sato (Ed.), Deciphering
stratification and inequality: Japan and beyond (pp. 140-160). Melbourne: Trans Pacific Press.
National Institute of Population and Social Security Research. (2012). The process of marriage
and fertility among the married Japanese women – the fourteenth national fertility survey.
Tokyo: National Institute of Population and Social Security Research. (in Japanese)
National Institute of Population and Social Security Research. (2017). Latest demographic
statistics. Tokyo: National Institute of Population and Social Security Research. (in Japanese)
Ohta, K. (2010). Income inequality: Among individuals, scale of company, and industry. In Y.
Higuchi and Economic and Social Research Institute. (Eds.). Labour market and income
inequality (pp. 319-368). Tokyo: Keio University Press. (in Japanese)
Qian, Z. (1998). Changes in assortative mating: The impact of age and education, 1970-1990.
Demography, 35(3), 279-292.
Raymo, J. M. (2003). Educational attainment and the transition to first marriage among Japanese
women. Demography, 40(1), 83-103.
Raymo, J. M., Fukuda, S., & Iwasawa, M. (2013). Educational differences in divorce in Japan.
Demographic Research, 28(6), 177-206.
Raymo, J. M. & Iwasawa, M. (2005). Marriage market mismatches in Japan: An alternative view
of the relationship between women's education and marriage. American Sociological Review,
70(5), 801-822.
22
Raymo, J. M., & Xie, Y. (2000). Temporal and regional variation in the strength of educational
homogamy. American Sociological Review, 65(5), 773-781.
Schwartz, C. R., & Mare, R. (2005). Trends in educational assortative marriage from 1940 to
2003. Demography, 42(4), 621-646.
Shida, K., Seiyama, K., & Watanabe, H. (2000). The changes in marriage market. In K. Seiyama
(Ed.), The system of Japanese social stratification 4: The future of gender, generation, and
social stratification (pp. 159-176). Tokyo: University of Tokyo Press. (in Japanese)
Smits, J., & Park, H. (2009). Five decades of educational assortative mating in 10 East Asian
societies. Social Forces, 88(1), 227-255.
Suzuki, T. (1991). Social intermarriages in Japan. Journal of Population Problems, 46(4), 14-31.
(in Japanese)
23
Appendix
TABLE A-1: The distribution of educational matching between wife and husband from 1980 to
2010 in Japan
SOURCE: The Population Census of Japan in each period.
wife husband JHS HS JC/VS UNI Total
JHS 21.1 8.0 0.3 0.5 29.9
HS 8.6 35.9 1.9 9.9 56.3
JC/VS 0.4 2.4 0.8 5.7 9.3
UNI 0.0 0.3 0.1 3.9 4.4
Total 30.2 46.7 3.1 20.0 100.0
wife husband JHS HS JC/VS UNI Total
JHS 6.4 3.6 0.2 0.3 10.5
HS 7.0 36.8 2.8 10. 5 57.1
JC/VS 0.9 7.0 2.4 12.1 22.4
UNI 0.1 0.9 0.3 8.8 10.1
Total 14.4 48.3 5.7 31.6 100.0
wife husband JHS HS JC/VS UNI Total
JHS 1.6 2.0 0.2 0.3 4.1
HS 3.9 33.3 4.0 10. 6 51.8
JC/VS 0.8 9.5 4.8 16.2 31.4
UNI 0.1 1.3 0.6 10.7 12.7
Total 6.4 46.1 9.6 37. 9 100.0
wife husband JHS HS JC/VS UNI Total
JHS 1.0 1.5 0.2 0.3 3.0
HS 2.8 23.6 4.4 7.5 38.2
JC/VS 1.1 12.2 8.1 16. 4 37.9
UNI 0.2 2.8 1.8 16.1 20.9
Total 5.0 40.1 14.6 40.2 100.0
1980 (N = 8,636,518)
1990 (N = 6,785,605)
2000 (N = 5,712,160)
2010 (N = 4,815,036)
24
TABLE A-2: Association parameters estimated by the saturated model
NOTE: Educational heterogamy is based on the female perspective. The
subject cases are Japanese couples in which the wife is aged 30-39 in each
year.
Census Ye ar wife/husband JHS HS JC/VS UNI
1980 JHS 2.206 0.397 -0.634 -1.969
HS 0.109 0.547 -0.172 -0.484
JC/VS -1.004 -0.342 0.689 0.656
UNI -1.311 -0.603 0.116 1.797
1990 JHS 2.040 0.338 -0.606 -1.772
HS 0.104 0.485 -0.131 -0.459
JC/VS -0.893 -0.246 0.605 0.534
UNI -1.251 -0.578 0.132 1.696
2000 JHS 1.957 0.264 -0.687 -1.534
HS 0.239 0.477 -0.138 -0.578
JC/VS -0.768 -0.224 0.598 0.394
UNI -1.429 -0.517 0.227 1.718
2010 JHS 1.679 0.190 -0.600 -1.268
HS 0.223 0.434 -0.133 -0.524
JC/VS -0.623 -0.188 0.514 0.297
UNI -1.278 -0.436 0.219 1.495
25
TABLE 1. The design matrix for the parameter 
: The simple model
Wifes education
JHS HS JC/VS UNI
JHS 1 2 2 2
HS 3 1 2 2
JC/VS 3 3 1 2
UNI 3 3 3 1
Husband’s education
26
FIGURE 1: Proportions of wives’ and husbands’ educational attainment, 1980-2010
NOTE: The subject cases are Japanese couples in which the wife is aged 30-39 in each
year.
SOURCE: The authors’ calculation of the Population Census of Japan in each period.
27
FIGURE 2: The proportions of educational homogamy among all marriages, by educational level
NOTE: The subject cases are Japanese couples in which the wife is aged 30-39 in each
year.
SOURCE: The authors’ calculation of the Population Census of Japan in each period.
28
FIGURE 3: The proportions of hypergamy and hypogamy among all marriages by the wife’s
educational level
NOTE: The subject cases are Japanese couples in which the wife is aged 30-39 in each year.
SOURCE: The authors’ calculation of the Population Census of Japan in each period.
29
FIGURE 4: Overall trends in the strength of educational homogamy, hypergamy and hypogamy,
The simple model
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
1980 1990 2000 2010
Association Parameter
Homogamy
Hypergamy
Hypogamy
30
FIGURE 5: Trends in the association parameters of the full educational pairings,
The saturated model
NOTE: Educational heterogamy is based on the female perspective. The
subject cases are Japanese couples in which the wife is aged
1980 1990 2000 2010
−3
−2
−1
0
1
2
3
Year
Association Parameter
JHS−JHS
HS−HS
JC/VS−JC/VS
Univ−Univ
Homogamy
1980 1990 2000 2010
−3
−2
−1
0
1
2
3
Ye a r
JHS−HS
JHS−JC/VS
JHS−Univ
HS−JC/VS
HS−Univ
JC/VS−Univ
Hypergamy
Education Pair: Wife − Husband
1980 1990 2000 2010
−3
−2
−1
0
1
2
3
Ye a r
HS−JHS
JC/VS−JHS
JC/VS−HS
Univ−JHS
Univ−HS
Univ−JC/VS
Hypogamy
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
BACKGROUND Evidence of a negative relationship between educational attainment and divorce in Japan is not consistent with predictions derived from standard theoretical emphases on the costs of divorce. OBJECTIVE AND METHODS Using marital history data from a cross-sectional survey, we estimated educational differences in divorce for two marriage cohorts: 1980-89 and 1990-2005. We also used 14 years of panel survey data to evaluate four possible explanations for the observed negative educational gradient. FINDINGS Our results confirmed that educational attainment is inversely related to divorce in Japan, and showed that, in contrast to some previous findings, the negative relationship between women's education and divorce has not become stronger in recent years. Analyses of the panel data provided some support for hypotheses that focused on the role of economic strain and on cultural values regarding reputation or "face," but they also showed that the negative relationship between education and divorce remained strong even after controlling for a range of posited correlates. CONCLUSIONS Our failure to solve the theoretical puzzle motivating these analyses suggests that other types of contextual modification to standard theories of family change are required to explain the strong negative relationship between educational attainment and divorce in Japan. We discussed possible examples of such modifications, focusing on the patterns of selection into marriage and the central importance of investment in children's educational success in Japan's highly competitive educational system, while also offering more nuanced theorization regarding the role of reputation or "face".
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
In Japan, unlike in most other industrialized societies, the decline in marriage rates has been most pronounced among highly educated women. Theoretical interpretations of this distinctive pattern of change have typically emphasized increasing economic independence for women and reductions in the gains to marriage. In this paper, the authors develop and evaluate an alternative explanation that emphasizes women's continued dependence on men's economic resources and decline in the relative supply of highly educated men. Using data from four rounds of the Japanese National Fertility Survey, the authors decompose the observed decline in marriage rates into changes in the propensity to marry and changes in the educational composition of the marriage market. Results indicate that change in the availability of potential spouses accounts for one-fourth of the decline in marriage among university-educated women and explains a substantial proportion of the growing educational differences in marriage. The conclusion is that the relatively large decline in marriage among highly educated Japanese women likely reflects both increasing economic independence and continued economic dependence on men.
Article
We study trends in educational homogamy at six boundaries in the educational structure of 10 East-Asian societies and explain its variation using explanatory variables at the country, cohort and boundary level. Educational homogamy was higher at the higher boundaries in the educational structure. Since the 1950s it decreased at all but the lowest boundaries, indicating convergence to a relatively low level of homogamy. Educational homogamy is lower in societies that are more modern, have higher female employment and experienced less Confucian influence. Results support the general openness and the exclusivity hypothesis, which predict educational homogamy to decrease in modernizing societies and to be higher when the group of more highly educated is smaller. Findings suggest that the trend towards less educational homogamy is related to educational expansion.
Article
The abstract for this document is available on CSA Illumina.To view the Abstract, click the Abstract button above the document title.
Article
This study examines two micro-level hypotheses about status homogamy: (1) the cultural matching hypothesis (people prefer to marry someone of similar cultural status) and (2) the economic competition hypothesis (people prefer to marry someone of high economic status). Detailed occupations of newlyweds in the 1970 and 1980 censuses are analyzed. Scales of cultural and economic occupational status are developed, and log-linear models of scaled association are used to analyze 70 x 70 occupational marriage tables. It is found that assortative mating by cultural status is more important than assortative mating by economic status, the economic dimension of status homogamy is more important when people marry late, and economic status homogamy has increased between 1970 and 1980 at the expense of cultural status homogamy.
Article
Most studies of assortative marriage still rely on cross-sectional data and apply log-linear modeling of the contingency table of wives’ and husbands’ educational levels. However, these macro studies have provided quite ambiguous findings and interpretations. In comparison, the life course approach analyzes single individuals over the life course and explicitly recognizes the dynamic nature of partner decisions and the importance of educational roles and institutional circumstances. Based on life course studies, educational homogamy seems to be driven by three factors: (a) Individuals often prefer to associate with equally educated partners; (b) educational expansion increases contact opportunities for equally educated men and women at an age when young people start to look for partners and form couples; and (c) women's changing economic role in dual-earner societies increases the importance of women's education and labor force attachment.
Article
Data from the U.S. Census and Current Population Survey are used to examine trends in the propensity to marry or to cohabit by the age and educational attainment of potential partners. Marriage rates declined sharply across all age and educational combinations between 1970 and 1980 and declined more sharply for less-educated persons between 1980 and 1990. The rise in cohabitation compensated somewhat for the decline in marriage rates, but the compensation was unequally spread among age and educational combinations. Highly educated men were more likely, and highly educated women were no more or less likely, to marry than to co-habit with less-educated partners in 1970 and 1980. By 1990, however, educational assortative-mating patterns between these two types of unions were similar. In 1990, marriages and cohabitations involving women who were better educated than their partners were more common than those involving women who were less educated than their partners. In addition, men and women in their early 20s tend to have partners better educated than themselves, but persons in their 30s tend to cross the less-than-high-school/more-than-high-school educational barrier when partners differ in educational attainment.
Article
Trends in intermarriages between people of different social classes delineated by educational status, occupation of father, and nationality are examined for Japan for the period 1960-1987. (SUMMARY IN ENG)
Article
I use data from a large nationally representative survey to examine the relationship between women's educational attainment and the timing of first marriage in Japan. The results indicate that later marriage for highly educated women primarily reflects longer enrollment in school, that university education is increasingly associated with later and less marriage, and that the trend toward later and less marriage is occurring at all levels of educational attainment. These findings are consistent, albeit weakly, with the argument that higher education should be negatively associated with marriage only in countries in which gender relations make it particularly difficult for women to balance work and family.