Explanations for the substantial decline in rates of marriage in East Asian countries often emphasize the role of rapid educational expansion for women in reducing the desirability of marriages characterized by a strong gender-based division of labor. Focusing on South Korea, we consider a very different scenario in which changing educational composition of the marriage market reduces the demographic feasibility of such marriages. Analyses of 1% microsamples of the 1990 and 2010 Korean censuses show that changes in the availability of potential spouses accounted for part of the decline in marriage rates over a period of 20 years (1985–1989 to 2005–2009) for highly educated women and less-educated men. We also show that growth in international marriages played a role in preventing an even more dramatic decline in marriage among low-educated men. These findings support the general relevance of marriage market mismatches in gender-inegalitarian societies and highlight the declining feasibility of marriage for low-educated men in such contexts. Findings also hint at important implications for inequality in a society such as Korea, where marriage remains a symbol of social success and is closely related to women’s economic well-being and men’s health and subjective well-being.
Social scientists have written extensively about the long-term decline in marriage rates in the United States and other wealthy countries. Prominent explanations include increasing economic independence for women, increasing employment insecurity during young adulthood, shifting attitudes, and the emergence of alternative lifestyles, including nonmarital cohabitation (Becker 1981; Bumpass et al. 1991; Oppenheimer et al. 1997; Thornton 2001). Although the trend toward later and less marriage is universal, its implications depend on context. Changing marriage behavior is more consequential for trends in union formation, fertility, and population aging in societies where the spread of nonmarital cohabitation has been slow and childbearing remains tightly linked with marriage.
This is particularly true in East Asian societies where marriage remains central to family life, intergenerational relationships, and social organization more generally (Park and Lee 2017; Raymo et al. 2015). In Japan, Korea, and Taiwan, pronounced decline in marriage rates is the driving force behind low fertility and rapid population aging—demographic trends that have emerged as critical policy issues (Byoun 2017; Jo 2015; Retherford and Ogawa 2006; Suzuki 2006). Specific policy concerns include (but are not limited to) the viability of public pension and healthcare programs, reduced family support at older ages, projected labor force shortages and the roles of immigration and mechanization in manufacturing and service sectors, and the promotion of work-family balance to support marriage and childbearing.
Research on later and less marriage in East Asian societies has also provided important insights into how theories of marriage timing may need to be adapted or modified to reflect contextual differences, especially with respect to gender inequality. Theoretical and empirical emphases on rapid improvements in women’s economic independence and reduction in the gains to marriage are compelling in the East Asian context (e.g., Park 2007; Park et al. 2013; Raymo 1998, 2003). The same is true of emphases on the tension between expanding opportunities outside the home and limited change in spouses’ division of domestic labor (McDonald 2009). In both cases, gender-asymmetric division of labor within marriage is thought to present unmarried women with an either-or choice between marriage and motherhood or career and other individual pursuits. The increasing tendency for women to choose the latter is seen as evidence of the declining desirability of marriages characterized by gender-based specialization.
However, there is good reason to believe that a comprehensive explanation for later and less marriage in East Asia is more complex than suggested by these commonly referenced theoretical frameworks. Of particular importance is evidence that change in the desirability of marriage is relatively limited and that expectations (and acceptance) of a clear gender-based division of labor within marriage remain strong. Tension between rapid social change and relative stability in norms and expectations around marriage sits at the core of one alternative explanation that emphasizes reductions in the feasibility (rather than the desirability) of marriages characterized by a strong gender-based division of labor. In this “marriage market mismatch” scenario, rapid relative improvements in women’s educational attainment, combined with limited change in normative desires and expectations regarding educational homogamy and female educational hypergamy, make it numerically more difficult for highly educated women (and less-educated men) to find a suitable partner (Raymo and Iwasawa 2005). This scenario provides a framework for reconciling dramatic decline in marriage rates with the continued centrality of men’s economic prospects as a spouse-selection criterion and maintenance of expectations for a clear gender-based division of labor within marriage.
In this article, we evaluate the marriage market mismatch scenario in South Korea (Korea, hereafter), a society where decline in marriage and associated reduction in the total fertility rate (TFR) have been particularly rapid (Park and Lee 2017; Park et al. 2013; Yoo and Sobotka 2018), women’s relative educational attainment has increased dramatically (Park 2007), and historical norms and patterns of spouse pairing show a strong tendency for marriages to be educationally homogamous or female hypergamous (Park and Smits 2005; Smits and Park 2009). This combination of characteristics, along with growth in international marriages, makes Korea a particularly interesting setting in which to evaluate the implications of changes in marriage market composition. We are not aware of any previous effort to quantify the role of increasing international marriage, especially the relatively rapid growth of marriages involving native-born men of low socioeconomic status (SES) and foreign-born women (Kim 2017), in offsetting mismatches in the domestic marriage market.
We seek to answer three specific research questions about the relevance of marriage market mismatches for understanding the marked decline in marriage rates in Korea.
(1) To what extent have changes in the educational composition of the marriage market contributed to declining marriage rates among highly educated women, net of changes in specific pairing propensities?
(2) To what extent have changes in the educational composition of the marriage market contributed to declining marriage rates among low-educated men, net of changes in specific pairing propensities?
(3) To what extent has international marriage offset the impact of mismatches in the domestic marriage market for men?
Women’s Educational Attainment and Marriage in East Asia
The pace and magnitude of the trend toward later and less marriage has been particularly pronounced in Japan, Korea, and Taiwan (Chen and Chen 2014; Park and Lee 2017; Park et al. 2013; Raymo 2013; Raymo et al. 2015). Mean age at first marriage has increased substantially over time for both men and women in these societies: from about age 27 for men in 1980 to about age 31 in 2010, and from less than age 25 for women in 1980 to age 29 in 2010 (Raymo et al. 2015). It is also clear that an increasing proportion of men and women in the region will never marry (Raymo et al. 2015:476). In contrast to the United States and many European countries, where nonmarital childbearing is common, the continued strength of the relationship between marriage and fertility in East Asia accentuates the demographic importance of later and less marriage. All three societies are characterized by period fertility rates that are well below replacement, and it is clear that the long-term decline in TFR primarily reflects reductions in the proportion of women who are married (Choe and Park 2006; Tsuya and Mason 1995).
Efforts to understand declining marriage rates in East Asia commonly emphasize the role of educational expansion and changing occupational structure in contributing to women’s growing economic independence. Standard economic models of marriage are built on the assumption that the gains to marriage derive from spouses’ pooling of complementary specializations in market and domestic labor, and thus imply that women’s growing economic independence should make such marriages less beneficial and less attractive (Becker 1981; Oppenheimer 1997). Gender equity theories of low fertility suggest that this “independence effect” associated with women’s expanding education and employment opportunities should be particularly pronounced in societies where the gender division of domestic labor remains highly unequal (McDonald 2000a, b, 2013). In these frameworks, women’s higher educational attainment and earnings should be associated with later and less marriage as newfound economic independence is used to “buy out” of marriage or at least to postpone transition to the “onerous status of wife and mother” (Tsuya and Mason 1995:156). Past research on East Asia has provided ample evidence of this posited negative relationship between women’s SES and marriage (Ono 2003; Park et al. 2013; Raymo 1998, 2003).
Marriage Market Mismatches
Empirical support for the role of women’s growing economic opportunities in delaying marriage does not mean that this theoretical emphasis provides a complete explanation of marriage trends in gender-inegalitarian countries. Findings from an earlier study of Japan are consistent with a very different explanation in which relative improvements in women’s educational attainment, combined with relatively stable spouse-pairing preferences, generate a marriage market mismatch that results in lower rates of marriage for highly educated women (Raymo and Iwasawa 2005). In societies where entrenched norms and expectations support status homogamy and female status hypergamy, convergence in men’s and women’s educational attainment will, all else equal, result in a growing number of highly educated women competing for a relatively smaller pool of well-educated men.
This focus on the implications of shifting marriage market composition is not novel: it is a straightforward extension of a long line of research on the ways in which changes in the relative prevalence of men and women of different ages (or other characteristics) can result in marriage squeezes (e.g., Crowder and Tolnay 2000; Schoen 1983). The main contribution of Raymo and Iwasawa’s (2005) study was to suggest that widely observed trends toward higher educational attainment for women (in both absolute and relative terms) can play an important role in shaping patterns of family formation in gender-inegalitarian societies characterized by established norms and expectations that the husband’s SES (e.g., education) is at least as high as the wife’s. Stated differently, relative improvements in women’s educational attainment may contribute to lower marriage rates in gender-inegalitarian societies even in the absence of marked decline in the desirability of marriages characterized by gender-based specialization. Raymo and Iwasawa’s (2005) findings demonstrated that the substantive role of marriage market mismatches in Japan is not small, with shifting marriage market composition due to changes in educational attainment accounting for one-fourth to one-third of the decline (between 1980 and 1995) in the proportion married among female junior college and university graduates.
Importantly, the same combination of relative improvements in women’s educational attainment and relatively stable preferences for female educational hypergamy may also generate marriage market mismatches detrimental to less-educated men. Growth in the relative supply of women at the top of the educational distribution implies a reduction in the relative supply of women at the bottom of the distribution. Despite the potential relevance of changing marriage market composition for marriage trends among men at the lower end of the educational distribution, we are unaware of any effort to empirically assess its relevance. Raymo and Iwasawa (2005) mentioned the potential relevance of such mismatches for men, but they limited their focus to women because of data limitations.
The large body of research on declining rates of marriage and the growing concentration of lifelong singlehood among men of lower SES in the United States and other countries has typically focused not on marriage market composition but rather on the declining economic prospects of men with lower levels of education. This work has emphasized the decline in low-skill jobs that used to provide a pathway to the middle class (Cherlin 2014), the rise in “bad jobs” that pay little and provide low employment security (Kalleberg 2009), and the associated difficulty of the transition to stable employment and economic independence in young adulthood (Oppenheimer et al. 1997). A similar focus on declining economic prospects is also central to recent research documenting lower rates of marriage among Korean men with low levels of education relative to their higher-SES counterparts (Park and Lee 2017).
Response to Marriage Market Mismatches
Assuming that marriage markets are geographically bounded, mismatches generated by shifting market composition will result in a decline in marriage rates unless those who are numerically disadvantaged choose to cast a wider net. For example, in classic depictions of marriage squeezes generated by a baby boom, numerically disadvantaged women can respond by marrying younger men. Similarly, low-SES women faced with a shortage of marriageable men generated by high levels of unemployment, poverty, and incarceration (Lichter et al. 1992; Sawhill and Venator 2015; Wilson 1987) might respond by marrying men of lower status than they might choose in a more favorable marriage market. Research on responses to shifting marriage market composition is mixed. Earlier research in the United States found little evidence that women alter their pairing behavior in response to unfavorable marriage markets (Lichter et al. 1995), but more recent cross-national research has shown that relative improvements in women’s educational attainment have been accompanied by an increase in the proportion of women who marry men with less education than themselves (Esteve et al. 2012, 2016).
An alternative response to marriage squeezes or mismatches rarely considered in past research is to expand the boundaries of the marriage market. Just as the emergence of online dating has expanded marriage markets beyond relatively narrow geographical boundaries, international marriage can change the composition of marriage markets at the national level. This may be particularly important in East Asia, where international marriages are increasingly common (Lee 2014; Raymo 2013; Yang and Lu 2010). Anecdotal and empirical evidence describes the role of marriages to women from China and Southeast Asian countries in offsetting the difficulties that native-born Japanese, Korean, and Taiwanese men of low SES, especially those living in rural areas, face in the marriage market (Lee 2008; Liaw et al. 2010). To our knowledge, however, there has been no rigorous empirical evaluation of the extent to which growth in international marriage has offset the impact of shifting domestic marriage market composition with respect to educational attainment in gender-inegalitarian societies in East Asia. What we do know about growth in international marriage suggests that offsetting of mismatches in the domestic marriage market is far more likely for low-educated men than for highly educated women (Liaw et al. 2010).
The Korean Context
Table 1 clearly shows the dramatic expansion of higher education in Korea, especially for women. The first four columns present the distribution of educational attainment among native-born men and women aged 25–44 in the 1990 and 2010 Korean censuses, respectively. In 1990, 26% of men had not attended high school, but the corresponding figure in 2010 was only 3%. During the same period, the percentage of men who attended postsecondary education increased from 5% to 22% for junior college and from 24% to 42% for university. In 1990, almost one-half of women had not attended high school, but two decades later, this figure was only 3%. The percentage of women who attended junior college increased from 3% to 23%, and the share of women who attended university rose from 11% to 35%.
Importantly, the relatively larger increase in educational attainment for women resulted in marked changes in the sex ratio within each educational category (shown in the last two columns of Table 1). In 1990, there were only 6 men for every 10 women at the lowest level of education, but 22 men for every 10 women at the highest level. Two decades later, the ratio of men to women at the lowest level of education increased to 0.9, and the sex ratio at the highest level of education shrank to 1.2. It is clear that the rapid relative improvements in women’s educational attainment depicted in Table 1, if combined with relatively stable, gender-asymmetric spouse-pairing norms, should reduce the availability of potential mates for highly educated women and low-educated men in Korea.
There are several reasons to believe that focusing on the feasibility, in addition to the desirability, of marriage may be particularly warranted in the Korean context. First, there is little evidence that young Korean men and women intend to remain unmarried, although views of marriage as an obligation have waned (Lee et al. 2015). For example, data from the 2016 Korean General Social Survey (KGSS) showed that 43% of respondents strongly agreed or somewhat agreed with the statement that “you should get married no matter what,” and data from 2012 showed that 56% of respondents agreed or strongly agreed that “married people are generally happier than unmarried people” (authors’ tabulations). Importantly, a traditional division of labor within marriage—with the husband in the role of primary earner and the wife primarily responsible for childcare and domestic labor (especially when children are young)—is also supported both normatively and in practice. For example, in the 2012 KGSS, more than one-third of respondents (37%) agreed (at least somewhat) that “it is more important for a wife to help her husband’s career than to pursue her own career,” and more than one-half (56%) agreed that preschool children suffer when their mother is employed. Similarly, according to a recent survey of Seoul residents, 8 of 10 household heads in their 30s and 40s (and thus likely to have a child) reported that wives are primarily responsible for household work with little or no help from husbands (Seoul Metropolis 2018). Recent qualitative research described how this gender-asymmetric division of labor is widely understood and accepted by both men and women (Brinton and Oh 2019). To be sure, support for gender specialization within marriage and the prevalence of breadwinner-homemaker marriages have declined over time in Korea and other East Asian societies (Han 2014; Raymo and Fukuda 2016), but this change is more limited than in the United States and other Western societies.
Second, recent data on marriage attitudes showed that higher education is associated with more positive attitudes toward marriage among unmarried Korean women (Lee et al. 2015). This pattern is not consistent with expectations derived from conventional theoretical emphases on educational expansion, economic independence, and the declining desirability of marriage characterized by gender-based specialization.
Third, cross-national studies have shown that educational homogamy in Korea remains higher than in many other countries (Smits and Park 2009; Smits et al. 1998). Some studies documented a recent decline in the odds of educational homogamy (Park and Kim 2011, 2012), but the share of educationally homogamous marriages has continued to increase, from 51% in 1980 to 79% in 2015, as men’s and women’s educational distributions have become more similar (Shin et al. 2017). Importantly, preferences for, and the prevalence of, female status hypergamy are also particularly strong in Korea (Choe 1998; Park and Smits 2005), and women remain far more likely than men to view economic prospects as an important criterion in choosing a spouse (Lee et al. 2015).
Focusing on the role of international marriages in offsetting marriage market mismatches may also be particularly relevant in Korea. The percentage of all marriages involving a Korean native and a foreign-born spouse was less than 5% during the 1990s but began to increase in the early 2000s, peaking at 13.5% in 2005 before stabilizing at around 10% in 2010 (Kim 2017). At that time, 7 of 10 international marriages involved a Korean groom and a foreign bride (Statistics Korea 2018). The majority of these brides are from East and Southeast Asian countries (especially China, Vietnam, and the Philippines), and their Korean husband is often relatively old and has low educational attainment (Kim 2017; Lee 2008). In contrast, Korean women are more likely to marry men from countries such as Japan and the United States (as well as China), and educational differences in the likelihood of international marriage are small (Kim 2017).
Data and Method
Marriage Rates From Retrospective Census Data
To examine changes over time in first-marriage rates by education and age, we use 1% microsample data from the 1990 and 2010 Korean censuses. In both years, census forms ascertained the age and educational attainment of all household members as well as the age of first marriage for ever-married men and women. We begin by using information on each individual’s relationship with the household head to identify married couples within each household. In the simple case of a nuclear household, for example, the married couple consists of the household head and his/her spouse. Because the census does not include information with which to identify whether the current marriage is the first marriage, we define first marriages as those in which the reported year of first marriage (calculated from information on age at first marriage) is the same for husband and wife. The census data provide only age at first marriage, rather than the exact year and month of marriage, so we also treat as first marriages those in which the calculated years of first marriage for husbands and wives differ by one year.1
We focus on two 5-year periods—1985–1989 and 2005–2009—to examine change in first-marriage rates over two decades. Data on individuals’ current age and age at first marriage (if ever married) allow us to construct person-year records for each five-year period that include information on respondents’ sex, age, and educational attainment as well as whether they married at a specific age. We classify individuals into six 5-year age groups: 15–19, 20–24, 25–29, 30–34, 35–39, and 40–44. Because the number of reported marriages at ages 15–19 is small, and even 0 for some age-education combinations, we do not report the marriage rates for this age group. In these person-year (person-age) data, an individual can contribute up to five observations: for instance, if a man was 25 years old and never married in the 1990 census, he contributes five observations: 20 years old in 1985, 21 in 1986, 22 in 1987, 23 in 1988, and 24 in 1989, with no experience of marriage. If a man was 26 years old and currently married in the 1990 census and reported his age at first marriage as 23, he contributes three observations: never married at 21 years old in 1985, never married at 22 in 1986, and married at 23 in 1987. First marriage is a censoring event, so this man is not observed in the data after 1987. In calculating first-marriage rates, we assume that marriages occur in the middle of the year.
Korea did not receive a substantial influx of immigrants until the early 2000s, a fact that is reflected in information provided about nationality in the 1990 and 2010 census data. The 1990 census microsample does not include foreign residents, but the 2010 census includes both native-born Koreans and foreign-born residents. We therefore limit our focus to Korea-born men and women in the 2010 census to maintain comparability with the 1990 census. To calculate observed first-marriage rates for men in 2005–2009, we assume that foreign-born women married to a Korean man entered the Korean marriage market at the time of marriage. This assumption is based on evidence that many foreign-born wives do not live in Korea long before marriage but rather migrate to Korea from their country of origin for the purpose of marriage, often arranged by brokers (Onishi 2007). Importantly, this means that the duration of exposure to the domestic Korean marriage market for foreign-born women is, by definition, 0 in our analyses. We know less about the characteristics of foreign-born men married to a Korean woman, so for the sake of consistency and simplicity, we assume that they too arrive in Korea at the time of marriage and thus have no exposure to the domestic Korean marriage market. Because Korean women married to a foreign-born man contribute only 0.08% of the total person-year (person-age) records in 2010, this simplifying assumption has no effect on our results.
We classify individuals into four educational groups: (1) did not attend high school; (2) attended high school but not a postsecondary institution; (3) attended junior college; and (4) attended university. Our decision to use the highest level of education attended, rather than completed, reflects the facts that we examine relatively young ages at which people may still be enrolled in postsecondary education and that the census does not provide information on the timing of school entry and completion. Because we treat this measure of educational attainment as time-invariant, values at ages 15–18 for those who attended tertiary education (categories 3 and 4) will reflect future attendance at junior college or university. This has no impact on our results given the very low levels of teenage marriage in Korea and the facts that very few students drop out of secondary school and that dropout rates in postsecondary educational institutions are also relatively low (Byun and Park 2017; Chung et al. 2015).
Because availability ratios are calculated at the national level using information on the total number of men and women of a given age and education level at risk of marriage, it is important to recognize that the composition of marriage markets at the local level will vary in ways that reflect patterns of population distribution within the country. For example, if less-educated men in rural areas are relatively isolated from the shrinking population of similarly educated women who may be more likely to move to urban areas, the extent of marriage market mismatches faced by these men at the local level will be greater than suggested by the availability ratios in Eq. (1a). Because differences in marriage market composition at the local level are not reflected in availability ratios (by definition), they contribute to the forces of attraction by shaping the rate at which men and women of specific ages and education levels encounter each other in the national marriage market. The fact that we have neither the knowledge nor the data to define subnational marriage markets is a limitation of our study.
Counterfactual Marriage Rates 1 (Constant Availability Ratios)
Counterfactual marriage rates calculated by holding forces of attraction constant at their 1985–1989 values lead to similar conclusions, so for the sake of brevity, we do not present these results.
Counterfactual Marriage Rates 2 (Assuming No International Marriage)
These calculations produce a total of 4,032 different first-marriage rates—that is, 24 × 24 = 576 possible pairings by age and education for each of the seven combinations of sex and period or counterfactual scenario (i.e., the four sets of observed marriage rates described by Eqs. (1a) and (1b) and the three sets of counterfactual marriage rates described by Eqs. (2a), (2b), and (3)). To summarize this large number of marriage rates in a way that allows us to clearly and intuitively address our research questions, we begin by summing across values of i and k (male rates) and j and l (female rates). This produces 24 age-education specific rates for each combination of sex and period (or counterfactual scenario). We then use these rates to address our research questions in two ways. We begin by examining ratios of the rates. Ratios of observed rates in 2005–2009 to observed rates in 1985–1989 describe changes in marriage rates over a 20-year period. We expect most of these ratios to be less than 1.0, reflecting the pronounced trend toward later and less marriage in Korea. Ratios of counterfactual 2005–2009 rates to observed 1985–1989 rates describe what change would have looked like if marriage market composition had remained constant at its earlier values and if there was no international marriage in the latter period. Counterfactual ratios greater than corresponding observed ratios indicate that the component held constant (i.e., availability ratios or international marriage) contributed to lower marriage rates, whereas counterfactual ratios less than the observed ratio indicate that the factor held constant contributed to higher marriage rates.
Because marriages rates are not expressed in a particularly intuitive metric, we also use life table techniques to describe the cumulative probability of first marriage for synthetic cohorts of men and women assumed to follow the observed or counterfactual marriage rates between ages 15 and 44. As in Raymo and Iwasawa (2005), we compare these synthetic cohort marriage curves to calculate the contributions of changes in availability ratios and international marriage to the observed change in marriage between the two periods. We expect the observed marriage curve for 2005–2009 to fall below that of 1985–1989 for all combinations of sex and education. Counterfactual marriage curves above (below) the observed 2005–2009 curve indicate that the component held constant contributed to lower (higher) marriage rates. The size of the gap between the observed and counterfactual marriage curves for 2005–2009 reflects the magnitude of that contribution to lower (or higher) marriage rates.
Changes in First-Marriage Rates
Table 2 presents ratios of both observed marriage rates in the two periods and ratios of counterfactual marriage rates in 2005–2009 to observed marriage rates in 1985–1989 for women and men. To evaluate the statistical significance of change in observed marriage rates and differences between observed and counterfactual rates, we estimate bootstrapped standard errors for these ratios. The first column, “Observed Ratio,” shows the ratios of the observed marriage rates in 2005–2009 to the corresponding rates in 1985–1989 for women. These ratios are significantly less than 1.0 at younger ages for all educational groups (except 20- to 24-year-old junior college graduates) but greater than 1.0 in some cases for the oldest two age groups. Marked decline in women’s marriage at younger ages is thus offset, to some extent, by a shift in marriage to older ages. This pattern of recuperation in marriage at older ages among Korean women has not, to our knowledge, been documented previously. It is also clear that decline in marriage is more pronounced for women who did not attend college.
For men (column 3), ratios are (with one exception) less than 1.0, reflecting the large decline in marriage rates for men of all ages and education levels. Almost all these ratios are significantly different from 1.0; and in some cases, the declines are particularly pronounced, especially those involving men with lower levels of education. For example, the marriage rates of 30- to 39-year-old men with less than high school education are 79% to 87% lower in 2005–2009 than in 1985–1989. This is dramatic change in marriage behavior over a relatively short period.
The second and fourth columns of Table 2, “AR Constant,” present ratios of marriage rates for 2005–2009 in which availability ratios (marriage market composition) have been counterfactually held constant at their 1985–1989 levels to observed marriage rates for 1985–1989. In most cases, these ratios are smaller but broadly similar to the observed ratios, indicating that the observed decline in marriage is explained primarily by reduction in the propensity to marry, independent of marriage market composition (i.e., decline in forces of attraction). This pattern is not surprising and is similar to that described in Raymo and Iwasawa’s (2005) study of Japan. The fact that these ratios are, in many cases, lower than the observed ratios (in columns 1 and 3) indicates that changes in marriage market composition have been conducive to marriage (i.e., marriage rates would have been even lower if marriage market composition had not changed). This is especially true for women with lower levels of education, a pattern that we will demonstrate more clearly shortly.
Results for men are similar to those of women with one important exception. The fact that the counterfactual ratios in the fourth column are lower than 1.0 but often slightly higher than the observed ratios indicates that reductions in men’s marriage rates are attributable to both declining forces of attraction and changing marriage market composition. Moreover, in some cases, counterfactual ratios are not different from 1.0, but observed ratios are significantly less than 1.0 (i.e., at ages 40–44 for men who did not attend high school and at ages 35–44 for men who attended high school only), indicating a substantial role for changing marriage market composition in the decline in marriage rates for older men with lower levels of education.
The last column in Table 2, “No International Marriage,” presents ratios of counterfactual marriage rates in 2005–2009 to observed marriage rates in 1985–1989, with the counterfactual rates calculated by assuming that international marriage was not an option (i.e., that Korean men who married a foreign woman counterfactually remained unmarried). Comparison of these ratios with ratios of observed marriage rates in the third column reveals an interesting pattern. With a few exceptions, these ratios are almost identical to the observed ratios, indicating little or no role for international marriage in boosting men’s marriage rates. The exceptions are for relatively older, less-educated men. For example, the counterfactual ratios for 35- to 39-year-old and 40- to 44-year-old men who did not attend high school were 0.08 and 0.17, respectively. In the absence of international marriage, the marriage rates for these men in 2005–2009 would have been only 8% and 17% of the observed marriage rate in 1985–1989, a much greater decline than indicated by the observed ratios of 0.21 and 0.40 in the first column. A similar pattern is observed for 35- to 44-year-old men who attended high school. Counterfactual ratios are also smaller than observed ratios for older men who attended junior college or university, but differences in these ratios are much smaller, suggesting that growth in international marriage was largely irrelevant for highly educated men. These findings based on large-scale population data are consistent with earlier, smaller-scale research (cited earlier) describing growth in international marriages among older men of low SES.
Cumulative First-Marriage Probability for Synthetic Cohorts
To more clearly describe the substantive impact of changes in marriage market composition (i.e., availability ratios) on marriage rates, we use the marriage rates summarized in Table 2 to calculate cumulative probabilities of first marriage (i.e., age-specific percentages who have ever married) for synthetic cohorts of women and men, by educational level. Counterfactual cumulative probabilities of first marriage are based on the assumption that synthetic cohorts experience age-specific marriage rates constructed by holding availability ratios constant at their 1985–1989 values or by eliminating the possibility of Korean men’s marriage to a foreign bride in 2005–2009. To convert age-specific marriage rates to age-specific probabilities, we assume that marriage rates are constant within five-year age groups (Preston et al. 2001:46).
Figures 1–4 show cumulative probabilities of first marriage for three different synthetic cohorts of Korean women at each level of education. Looking first at women who did not attend high school (Fig. 1), comparisons of marriage trajectories for two synthetic cohorts based on observed marriage rates in 1985–1989 (solid black line) and 2005–2009 (dashed black line) show the large decline in marriage rates. The cumulative percentage ever married by exact age 45 in 2005–2009 (74%) is well below the corresponding value for 1985–1989 (98%) and clearly shows that marriage is no longer universal for Korean women with low education.4 When we counterfactually assume that marriage market composition did not change (i.e., only forces of attraction changed), the cumulative proportion percentage married by exact age 45 is a bit more than half (58%) of the observed value for 1985–1989 (solid gray line). The fact that this counterfactual marriage curve is lower than the observed curve for 2005–2009 indicates that changes in marriage market composition were conducive to marriage for women with the lowest level of education. Stated differently, change in marriage market composition worked to prevent the marriage rates of these low-educated women from falling even further. Inspection of pairing-specific changes in availability ratios indicates that the most important change in marriage market composition for these women was the increasing availability of unmarried men with a high school education or less, especially men over the age of 30.
The pattern of change for women who attended high school (Fig. 2) is qualitatively similar. The observed decline in marriage is less than for women who did not attend high school, but the roles played by changing forces of attraction and changing marriage market composition are essentially the same. The cumulative percentages ever married would have been lower at all ages if composition of the marriage market had not become more conducive to marriage. The cumulative percentage married by exact age 45 is 86% in 2005–2009 but only 74% when we hold availability ratios constant at their 1985–1989 values (solid gray line). For these women, the increasing availability of university-educated men (of all ages) and high school–educated men over age 30 was particularly important in keeping marriage rates from falling further.
The patterns for women who attended junior college (Fig. 3) and university (Fig. 4) differ markedly from their less-educated counterparts but are quite similar to those of the highly educated Japanese women examined by Raymo and Iwasawa (2005). For both groups, changing forces of attraction (holding availability ratios constant) account for most, but not all, of the relatively small observed decline in marriage. The fact that the solid gray curves (holding availability ratios constant) are higher than the observed marriage curves for 2005–2009 indicates that marriage rates would have been higher than actually observed in 2005–2009 if marriage market composition had not changed. In other words, changing marriage market composition contributed to the observed decline in marriage for women in the two highest educational groups.
The counterfactual curve calculated assuming constant availability ratios for women who attended junior college is 4–9 percentage points above the observed 2005–2009 curve beyond age 30. One interpretation of these differences between counterfactual and observed curves is that changing marriage market composition accounts for 39% to 57% of the overall decline in marriage between 1985–1989 and 2005–2009 (calculated as 1.0 minus the ratio of the counterfactual difference to the observed difference, as in Raymo and Iwasawa 2005:815). Differences between counterfactual curves calculated assuming constant availability ratios and the 2005–2009 observed curves for women who attended university range from 1 to 4 percentage points beyond age 30. Stated differently, 8% to 17% of the observed decline in marriage between 1985–1989 and 2005–2009 can be attributed (in a demographic accounting sense) to changing marriage market composition for Korean women in the highest educational category.5 Examination of forces of attraction and availability ratios for specific pairings shows that the decline in marriage for highly educated Korean women reflects both declining forces of attraction and declining availability for marriage with men who attended university.
Figures 5–8 show cumulative probabilities of first marriage for Korean men. Figure 5 presents figures for four synthetic cohorts of Korean men who did not attend high school. Comparison of the marriage trajectories based on observed marriage rates in 1985–1989 (solid black line) and 2005–2009 (dashed black line) shows a substantial decline in marriage. The share of these low-educated men who had married by their 45th birthday fell by half, from 92% in the earlier period to 46% in the later period. Importantly, we see that the counterfactual curve calculated by holding the availability ratio constant (solid gray line) is higher after age 35 than the observed 2005–2009 marriage curve (dashed black line). If marriage market composition had not changed, the marriage rates of these low-educated men would have been higher than observed. The absolute magnitude of the implications of shifting marriage market composition for Korean men with the lowest level of education is much larger than that observed for highly educated women (in Figs. 3 and 4). In relative terms, change in availability ratios for men in the lowest educational category accounts for 14% and 37% of the observed decline in marriage by ages 40 and 45, respectively (compared with 8% and 17% among university-educated women). Finally, the counterfactual marriage trajectory based on the assumption of no international marriage (dashed gray line) shows that the cumulative percentage ever married by exact age 45 is 10 percentage points lower than the observed value for 2005–2009. Stated differently, for men who did not complete high school, the rise in international marriage has kept marriage from falling even further than it would have if potential mates for these men were restricted to native-born Korean women.
The pattern for men who attended high school (Fig. 6) is similar to that for the least-educated group. Through age 30, changes in marriage market composition account for none of the reduction in cumulative percentages married, but they play an important role at older ages. Indeed, counterfactually assuming no change in availability ratios accounts for 73% and 86% of the observed decline in the percentage ever married by ages 40 and 45, respectively. As in Fig. 5, we again see in Fig. 6 that international marriage has kept the percentage ever married among older men who attended high school from falling further than it might otherwise have. The observed percentages ever married beyond age 40 (dashed black line) are 6–8 percentage points higher than the values calculated by counterfactually assuming that no international marriages took place (dashed gray line). Inspection of pairing-specific availability ratios for men with a high school education or less shows that, not surprisingly, the decline in the availability of women in the two lowest educational groups is of particular importance.
A similar but less extreme pattern holds for men who attended junior college (Fig. 7). Beyond age 30, changes in marriage market composition account for a nontrivial proportion of the decline in marriage, with a counterfactual decline in the proportion married (solid gray line minus dashed black line) equal to 33% to 55% of the observed decline in marriage (solid black line minus dashed black line). A small role for international marriage is seen only at older ages, where the observed marriage curve is slightly higher than the counterfactual curve based on the assumption of a closed marriage market.
For men who attended university (Fig. 8), the pattern is quite different. There is no evidence that changes in marriage market composition have contributed to reductions in the percentage ever married. If anything, changes in marriage market composition have contributed to slightly higher levels of marriage, as indicated by the fact that the counterfactual marriage trajectory (holding availability ratios constant) is somewhat lower than the observed marriage trajectory for 2005–2009 between exact ages 25 and 40. Decline in marriage among the most highly educated Korean men is due primarily to decline in forces of attraction. Pairing-specific forces of attraction show that decline was particularly large for educationally homogamous pairings. There is also no evidence that international marriage has played a role in the marriage behavior of highly educated men. The counterfactual marriage curve calculated by assuming no international marriage is identical to the observed marriage curve for 2005–2009 (dashed black line).
Our primary goal in this study was to examine the contribution of shifting marriage market composition to observed decline in first-marriage rates in Korea. In doing so, we extended previous research on marriage market mismatches and declining rates of marriage for women in gender-inegalitarian societies in two important ways. First, we also evaluated the extent to which declining rates of marriage among low-educated men reflect changing marriage market composition. Second, we examined the extent to which the implications of such mismatches for low-educated men may be offset by an increase in international marriage. Results demonstrated large declines in marriage (especially among the less-educated), showed that changes in marriage market composition account for nontrivial proportions of the decline in marriage among highly educated women and low-educated men, and indicated that the rise in international marriage prevented the marriage rates of low-educated men from falling even further.
Taken as a whole, our results provide empirical support for Raymo and Iwasawa’s (2005) speculative claim that marriage market mismatches contribute to lower marriage rates in gender-inegalitarian societies. Not only are our results for highly educated women similar in magnitude to those in the earlier study of Japan, but we also find an even stronger role for market mismatches in shaping the marriage outcomes of men with lower levels of education. These findings have important implications for our understanding of family formation and social/economic inequality in gender-inegalitarian societies (at least those in East Asia). Marriage market mismatches generated by a combination of rapid relative improvements in women’s educational attainment and entrenched norms of status homogamy and female status hypergamy have likely exacerbated the implications of increasingly tenuous economic circumstances among those with a high school education or less for growing inequality in marriage prospects (especially for men).
To the extent that marriage is associated with well-being (e.g., social status, economic security, and physical and emotional health), marriage market mismatches appear to play an underappreciated role in shaping patterns of inequality in countries like Korea and Japan. Further investment in developing a broad theoretical and empirical understanding of linkages between rapid relative improvements in women’s educational attainment and processes of stratification and inequality that focus on associated shifts in population composition is a potentially valuable direction for subsequent research.
Our study is not without limitations. We examined only two time points separated by 20 years and did not consider trends beyond 2010, a time during which highly educated women in Japan appear to have responded to marriage market mismatches by changing their pairing behavior (Fukuda et al. forthcoming). Second, we focused only on first marriages, which may be limiting in light of recent increases in divorce and remarriage in Korea, despite a relatively low prevalence of divorce and remarriage during the period we examined (Park and Raymo 2013). Third, we also were not able to account for growing heterogeneity within broad educational groups. Korea is now the world’s most highly educated society (OECD 2018), and analyses of educational assortative mating should arguably focus not on simple measures such as “attended college” but rather on more refined classifications of higher education (e.g., elite vs. non-elite colleges). This limitation is also relevant to our definition of educational categories. Assuming that educational categories have the same meaning in 1985–1989 and 2005–2009 is potentially problematic given the rapid expansion of education over this 20-year period (Shavit and Park 2016). For example university education can mean something quite different when more than half of the population attends college than when college education is relatively uncommon. Attention to how the changing meaning of specific educational levels affects marriage behavior and marriage market dynamics is critically important for research on educational assortative mating, but we recognize that this is beyond the scope of currently available data in most societies, including Korea.
Finally, as noted earlier, the harmonic mean model is attractive in that it facilitates straightforward assessment of the role of changing marriage market composition. However, it is limited in that it provides no information about specific social and economic forces of change in marriage behavior. In this model, forces of attraction represent a “black box” of sorts that encompasses all factors that shape marriage propensities independent of marriage market composition. Although our results showing that changes in the force of attraction account for most of the observed decline in marriage rates are thus potentially consistent with the various theoretical explanations articulated in previous research (e.g., women’s increasing economic independence, men’s declining economic prospects, changing attitudes), they shed no light on the relative importance of each.
Despite these limitations, our study extends research on educational expansion and marriage trends in wealthy societies in important new directions. Most critical is recognition that growth in women’s educational attainment can contribute to lower levels of marriage even in the absence of marked declines in the desirability of marriages characterized by a strong gender-based division of labor, and our generation of relevant empirical evidence with which to evaluate this insight. This is not to say that decline in desirability of such marriages is irrelevant: indeed, our results show that the large majority of the decline in marriage for most groups, including highly educated women, is accounted for by decline in forces of attraction. However, it does mean that in societies where the desire for gender-asymmetric division of labor (or at least the normative acceptance of such arrangements) remains prevalent, relative growth in women’s education can, by reducing the feasibility of such marriages, contribute in substantively meaningful ways to overall declines in marriage. This is an important message that social scientists should keep in mind as they seek to document and understand rapid decline in marriage in other relatively gender-inegalitarian societies around the world.
James Raymo acknowledges support provided by a center grant from the Eunice Kennedy Shriver National Institute of Child Health & Human Development (P2C HD047873). Hyunjoon Park acknowledges support from the Laboratory Program for Korean Studies through the Ministry of Education of Republic of Korea and Korean Studies Promotion Service of the Academy of Korean Studies (AKS-2016-LAB-2250002).
Because we identify married couples by the fact that they reside in the same household and report the same year of first marriage, we are not able to include information on the first marriages of those who experienced divorce or widowhood. The absence of information on previous spouses of the formerly married prevents us from including information on first marriages for these men and women. The absence of information on the age/year at which the previous marriage dissolved and the age/year of remarriage also necessitates that we limit our focus to first marriages. The relatively low prevalence of remarriages in Korea makes this less of a limitation than it would be in other societies. Of the 104,414 men and 106,859 women aged 15–44 in the 1990 census 1% microsample, 0.9% of men and 2.3% of women were either divorced or widowed. In the 2010 data, 2.3% of 97,170 men and 3.7% of 97,888 women aged 15–44 were either divorced or widowed (authors’ tabulations).
More specifically, the availability ratio represents the proportion of random encounters that would involve men of a given age and education for women of a given age and education (and vice versa), and the force of attraction reflects both the rate of encounters and the proportion of those encounters that result in marriage (Qian and Preston 1993:483). Because encounters in the marriage market are not random—for example, because spheres of activity are segregated by age and educational attainment and because international marriages are often arranged by brokers—it is important to recognize that the force of attraction reflects these features of the marriage market in addition to the propensity for a given type of encounter to result in marriage.
Given the very small number of international marriages involved, we do not calculate the corresponding counterfactual rates for Korean women.
It is important to keep in mind that these synthetic cohort measures reflect changes in both the quantum and tempo of first marriage.
Although the substantive magnitude of these differences appears small, observed marriage rates for 2005–2009 and counterfactual rates are statistically different between ages 30–44 for all but two groups (35- to 39-year-old junior college attendees and 40- to 44-year-old university attendees; results not shown).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.