JCS本刊论文 | 夫妻教育匹配模式和生育行为

Welcome to follow

The Journal of Chinese Sociology

2025年10月13日，The Journal of Chinese Sociology（《中国社会学学刊》）上线文章Spousal educational matching patterns and fertility behavior（《夫妻教育匹配模式和生育行为》）。

| 作者简介

卿石松

华东师范大学社会发展学院教授、博士生导师。

主要研究方向：生育与人口发展战略、人口负增长及经济影响、劳动就业以及性别收入分配等。

Abstract

The rise in women’s educational attainment and shifts in gender roles have reshaped marriage patterns and childbearing in China. Drawing on household-level data, this study examines cohort changes in spousal educational matching and their implications for fertility behavior. The results show a decline in traditional gradient marriages alongside an increase in highly educated homogamous marriages and marriages where women marry down in education. Fertility outcomes vary across educational pairing types. Among homogamous couples, a negative gradient emerges: the higher the educational attainment, the fewer children they have, with highly educated couples exhibiting the fewest children and the lowest likelihood of having a second child. Additionally, as wives’ educational advantage over husbands increases, the probability of having a second child declines. However, over time, the fertility disadvantages associated with highly educated homogamy and female downward marriages have diminished. This study’s findings underscore how evolving gender roles, relative status within couples, and the gendered division of labor shape fertility decision-making, offering valuable insight into fertility trends and informing policy development.

Keywords

Reversed gender gap in education; Educational pairing; Fertility behavior; Gendered division of labor; Gender roles

Introduction

Since 1999, higher education in China has expanded rapidly, resulting in a significant increase in educational opportunities, particularly for women in younger cohorts. Beginning in 2007, the number of female college students (in both undergraduate and junior college programs) has consistently surpassed that of males each year. In 2021, women made up 61.04% of undergraduate admissions. Additionally, from 2009 to 2021, women constituted the majority of those admitted to master’s degree programs. This trend echoes the reversal of the gender gap in education observed in Western developed countries (Esteve et al. 2016). The proportion of the younger population receiving higher education has been steadily rising, and among those entering the age of marriage and childbearing, the number of women with higher education has already surpassed that of men. This emergent pattern has had a profound impact on marriage and fertility behaviors.

Educational expansion has long been recognized as a key driver of fertility transition. With longer years of schooling and the enhancement of human capital, educational expansion has contributed to delays in women’s age at first marriage and childbirth (Wu and Liu 2014). This has been identified as a major factor behind declining fertility rates and the trend toward smaller family sizes (Ge and Huang 2020; Wu et al. 2022). Furthermore, the expansion of women’s education—particularly the reversal of the gender gap—has led to transformations in the marriage market. The traditional pattern of hypergamy (or gradient marriage where women marry up) has gradually declined, while educationally homogamous marriages, in which both spouses are highly educated, and hypogamous marriages (where women marry down), have become more common (Esteve et al. 2016). At the same time, an increasing number of women now out-earn their husbands and take on the role of primary breadwinner (Klesment and van Bavel 2017). These structural changes affect both partners’ bargaining power in terms of labor force participation, household division of labor, and resource allocation, thereby influencing fertility decisions within the family. Consequently, against the backdrop of rising education levels and shifting gender roles, changes in educational assortative mating patterns and their impact on fertility behavior have become a significant topic of inquiry.

Although classical theories and a large body of individual-level empirical research have long supported a negative correlation between women’s education and fertility, recent studies have revealed a more complex relationship. Some research suggests that the relationship between women’s educational attainment and fertility may be U-shaped—fertility levels among highly educated women have shown a rebound (Hazan and Zoabi 2015). Similar studies have found that while highly educated women may delay their first birth, they are more likely to have a second or third child (Gerster et al. 2007; Klesment et al. 2014). A study conducted in China even suggests that education increases women’s fertility desires and how many children they have (Chen 2022). These findings suggest that classical theories may be outdated, or no longer fully applicable.

However, since most fertility decisions are made within the context of marriage and family, the characteristics and interactions of both spouses jointly shape fertility behavior. Focusing solely on one partner’s characteristics may lead to biased or misleading conclusions. In fact, unlike women, men’s educational attainment may be positively associated with fertility levels (Kravdal and Rindfuss 2008; Trimarchi and van Bavel 2017). The negative correlation between education and fertility among women is either insignificant or may even turn positive among highly educated women, partly due to the positive assortative matching effect. Furthermore, Kreyenfeld (2002) demonstrated that once a partner’s education level is controlled for, the positive correlation between women’s education and higher-order fertility becomes insignificant. These findings highlight the importance of considering both spouses’ educational levels and their matching patterns for a comprehensive understanding of the relationship between education and fertility.

Expanding our understanding of the relationship between education and fertility by using the couple as the unit of analysis holds significant theoretical and empirical value. Gary S. Becker’s neoclassical economic model of the family has long dominated related discussions. According to this model, under the traditional gender division of labor—where men are the breadwinners and women are the homemakers—a woman’s educational advantage and labor market participation lead to higher opportunity costs for childbearing (Becker 1960). As a result, low-education homogamy or the traditional gradient marriage pattern, in which the woman’s education is lower than the man’s, is generally considered more conducive to childbearing. In contrast, Valerie Kincade Oppenheimer’s dual-earner household theory, or the “resource pooling” model, offers a different perspective. This theory argues that having both spouses participate in the labor market can expand household income sources and enhance a family’s ability to cope with economic risks, thus forming a more realistic and optimal arrangement (Oppenheimer 1994). Within this framework, the wife’s higher educational attainment increases the household’s disposable resources, which in turn improves the capacity to raise more children. These contrasting theoretical perspectives underscore the importance of empirical research to test the relationship between educational assortative mating patterns and fertility outcomes.

However, empirical evidence remains limited and largely concentrated in European countries, offering no consistent support for either of the two theoretical frameworks. For example, some studies have found that trends in Central and Eastern European countries, such as Austria and Bulgaria, align more with the predictions of Becker’s theory—couples in which both partners have higher education tend to have fewer children (Osiewalska 2017). In contrast, research from other regions, particularly the Nordic countries, tends to support Oppenheimer’s dual-earner household model (Trimarchi and van Bavel 2020). This divergence may suggest that both theories are inadequate, or that the relationship between educational matching and fertility behavior is influenced by specific social contextual factors, such as prevailing gender norms and work-family support policies (Nitsche et al. 2018). Moreover, due to shifts in gender structures within education, changes in spousal matching patterns, and the evolving status of women, the relationship between educational assortative mating and fertility may be dynamic. Differences in findings could stem from variations in time periods or sample selection. Therefore, in addition to conducting cross-national comparisons, longitudinal comparisons within a specific country are also essential.

China’s marriage, family structures, and gender norms have traditionally been more conservative than those in Western societies, yet they have undergone rapid and profound transformations. Marriage and fertility patterns among younger cohorts may have already experienced significant changes, providing a unique case for exploring this issue. In this study, we use data from the China Family Panel Studies (CFPS) to analyze shifts in educational assortative mating patterns from a cohort perspective and examine their impact on fertility behavior. In doing so, we clarify the relationship between education and fertility, offering valuable insights for the development and optimization of fertility support policies.

This study’s contributions and significance are as follows. First, exploring how spouses’ educational levels interact with their fertility outcomes can advance fertility theories at both the couple and the family level. As shown by Qing (2020), the interaction between spouses is crucial in fertility decision-making. Adopting a spousal matching perspective, this study investigates the relationship between the educational characteristics of both partners—along with their pairing patterns—and fertility behavior. This approach contributes to a more comprehensive understanding of the complex relationship between education and fertility, encompassing not only the impact of women’s education on fertility, but also the role of men’s education, and particularly the influence of spousal educational matching on fertility outcomes. Secondly, the study’s findings reveal the dynamic relationship between educational matching patterns and fertility from a cohort perspective. Educational expansion and fertility transition have been among the most significant demographic shifts over recent decades. However, there has been a lack of systematic assessment of how fertility behavior under different educational matching patterns has changed over time, and across birth cohorts. This study addresses this gap by combining analyses of cohort change, educational assortative mating, and fertility trends, offering new insight into fertility transitions amid educational expansion. Finally, this research also supplements the existing literature on the evolution of educational matching among Chinese couples. Numerous scholars, from both China and abroad, have examined this topic (Dong and Xie 2023; Li 2008; Qi and Niu 2012; Shi 2019), but the conclusions have varied depending on both the period studied and the indicators and methods used. By leveraging the most recent data, this paper re-examines trends identified in prior research, and further develops understanding through detailed cohort-based analysis, laying a foundation for more in-depth future investigation.

Theoretical framework

and literature review

Theoretical foundation

Education, as an important predictor of income and social status, has received significant attention in fertility decision research. While numerous studies have examined the link between education and fertility from an individual perspective, research based on the interactive perspective of couples remains insufficient. In reality, both spouses’ educational characteristics determine the distribution of resources and opportunities within the family, and also shape spousal roles and the division of labor. This, in turn, influences the fertility decision-making process, as well as the outcomes, in multiple ways.

Becker’s (1960) neoclassical model provides a classic framework for understanding both the spousal division of labor and fertility decisions. Based on the traditional “male breadwinner, female homemaker” division of labor, the theory argues that a woman’s advantage in education and the labor market leads to higher opportunity costs for childbearing. Therefore, within this model, gradient marriages in which the woman is less educated than the man are more conducive to childbearing. At the same time, compared to couples with lower education levels or fewer resources, highly educated couples tend to place greater emphasis on investing in their children’s human capital and quality (Becker and Lewis 1973), which raises the costs of childrearing. This suggests that among all spousal educational pairing combinations, homogamous marriages in which both partners are highly educated may have the lowest fertility rates.

However, with the growing pursuit of a high-quality life and the increasing recognition of women’s economic roles within marriage and family, Oppenheimer (1994) offered an alternative analytical perspective. She emphasizes that dual-earner households represent an optimal arrangement for managing economic risks and maintaining living standards. They are also better equipped to bear the costs and challenges of childbearing. A wife’s participation in the labor market diversifies income sources and raises the family’s overall income, thereby better securing the household’s standard of living. From this perspective, either considering the family as a whole or from a resource accumulation standpoint, higher education can promote fertility due to greater economic resources and more stable expectations. This contradicts the traditionally observed negative relationship between education and fertility.

Gender equality theory offers a dialectical and integrated perspective on the two opposing views described above. The theory argues that although dual-earner families have a stronger economic capacity for raising children, their time available for childbearing and childrearing may be more limited. Because gender equality in the private sphere—such as household chores and childcare—often lags behind progress in the labor market, professional women frequently face the dual burden of work and family responsibilities. This has been considered one of the key reasons why fertility rates decline alongside women’s increased education and employment opportunities (McDonald 2000). Therefore, it is only when the gender revolution or gender equality advances to a certain stage—especially with men participating equally in childcare—that fertility levels among highly educated women or couples can recover, and even increase (Esping-Andersen and Billari 2015; Goldscheider et al. 2015). This process weakens or even reverses the negative correlation between education and fertility. In other words, in regions where the division of gender roles within the family is more equal or where there is institutional support for work-family balance, highly educated couples may have higher fertility levels.

Furthermore, the “negotiation” paradigm of fertility decision-making theory emphasizes that educational matching patterns influence fertility decisions through spouses’ relative status or bargaining power. Couples are not simply a unified whole but a collection of individuals with differing, and sometimes conflicting, interests or preferences. In terms of fertility, because the costs and benefits are not equally shared, spouses may have different reproductive intentions (Doepke and Kindermann 2019). Fertility choices often result from negotiation or bargaining between partners (Rasul 2008). Highly educated women have more opportunities in the labor and marriage markets than less-educated women, which typically translates into greater bargaining power. Thus, they exert a stronger influence on fertility decisions and outcomes. Considering that, especially in developing countries, women’s fertility intentions tend to be lower than men’s, an increase in women’s status or bargaining power may lead to a decline in fertility levels (Osiewalska 2018). This implies that when women’s educational attainment is high—particularly when it exceeds that of their spouse (i.e., hypergamy in reverse or female hypogamy or “marrying down”)—fertility levels may decrease.

In summary, fertility decisions are joint choices made by both spouses. The relationship between couples’ educational matching patterns and fertility cannot be solely attributed to the family’s overall resources or capabilities. Instead, a deeper understanding of fertility decisions requires distinguishing between the husband’s and the wife’s educational attainment, as well as considering their relative status within the couple. For example, in a heterogamous marriage, where one partner is highly educated and the other less educated, the couple’s overall educational level might seem irrelevant. However, from a gender and relational perspective, whether the husband or wife holds the higher education can have distinct implications for fertility decisions.

Literature review

The above-mentioned theories emphasize the impact of spousal interaction on fertility outcomes—that is, a partner’s educational characteristics can influence or moderate the relationship between education and fertility. However, due to the paucity of data at the couple and family level, only a few studies have incorporated spousal interaction factors to explore the specific relationship between couples’ educational matching patterns and fertility, and these studies have produced mixed or even conflicting conclusions.

Some studies provide partial support for Becker’s gender division of labor theory. For example, research in Austria and Bulgaria has shown that highly educated homogamous couples tend to have fewer children than those with lower education levels (Osiewalska 2017). Additionally, this study highlighted those marriages where women “marry down” educationally are the least favorable for fertility, although this finding does not hold in France, where public childcare services are more comprehensive. Later, using data from the Generations and Gender Survey (GGS) across ten European countries, the same researcher found results supporting Becker’s model, even after controlling for the household division of labor. Unlike homogamous marriages, educationally heterogamous marriages where the woman is less educated than her spouse are positively associated with the number of children, while “downward” marriages for women show a negative association with fertility. The negative effect of women marrying down on fertility is more pronounced in Central and Eastern European countries, possibly due to women in these countries being constrained by traditional family and social institutions, as well as lacking sufficient social support (Osiewalska 2018).

Other studies have supported Oppenheimer’s dual-earner family theory. Research based in Sweden shows that couples with higher education levels, abundant socioeconomic resources, and access to family-friendly employment policies have a higher probability of having a second or third child (Dribe and Stanfors 2010). Consistent with this, a study across 24 European countries found that couples where both spouses are highly educated tend to delay childbearing, but have the highest likelihood of having a second or third child (Nitsche et al. 2018). A recent study from Spain also found that compared to traditional marriages where women “marry up,” highly educated homogamous couples have a higher probability of having another child within the same observation period (across two survey waves), whereas low-education couples have a lower probability of childbearing (Bueno and Garcia-Roman 2021). However, Trimarchi et al. (2020) once again highlighted the heterogeneity across European countries. They found that the resource-accumulating dual-earner family model predominates in Belgium, while Central and Eastern European countries tend to follow Becker’s gender division of labor model in family patterns.

The heterogeneity and contradictions among these empirical findings may stem from differences in gender equality and family policy across countries or time periods. This underscores the necessity of considering specific social contexts and historical characteristics. In fact, the historical background of educational matching, fertility transition, and gender relations in China may differ significantly from that of Western countries. However, only a few studies have focused on Hong Kong and Taiwan, China, and have confirmed Becker’s theory of a gender-based division of labor—namely, that educational hypergamy (women marrying up) is conducive to fertility, while educational hypogamy (women marrying down) is detrimental to fertility (Sun and Zhao 2025; Tsou et al. 2011). Another limitation of existing research is the lack of consideration for temporal effects. The academic community generally emphasizes the impact of gender-based family division of labor on fertility, yet roles and divisions within the family also evolve over time. This suggests that relatively traditional but rapidly changing Chinese society offers a unique perspective and rich contexts for exploring the relationship between couples’ educational matching patterns and fertility behavior, as well as its changes over time.

In summary, the academic community has increasingly recognized the importance of couples’ educational matching when examining the relationship between education and fertility, while also emphasizing that research findings may vary across temporal and spatial contexts. This article is grounded in the unique cultural context of China to thoroughly investigate the influence of educational matching patterns in couples’ fertility decision-making processes, and to reveal how these patterns change over time from a cohort perspective. This study not only provides a significant case for understanding fertility behavior in a non-Western cultural setting, but also addresses the current gap in research regarding the temporal dimension. We hope that this work will inspire a new wave of scholarly reflection and discussion on related issues, thereby advancing theoretical development and policy innovation.

Data, variable measurement,

and methods

Data sources and sample selection

The data used in this study come from the China Family Panel Studies (CFPS), a large-scale national longitudinal survey conducted by the Institute of Social Science Survey, Peking University. The sample covers 25 provinces, municipalities, and autonomous regions, accounting for 95% of the national population (excluding Hong Kong, Macao, and Taiwan). In the survey, household serves as the sampling unit, and detailed information is collected on education, fertility, and other characteristics of the spouses and other family members, providing suitable and unique data for the research presented in this article.

Based on the research objectives, this study focuses on couples as the analytical unit. We use samples of couples in which the wife was born between 1950 and 1994 (i.e., aged 26 to 70) who have completed formal education and reached marriage and fertility age. The samples are divided into nine birth cohorts using five-year intervals. The cohorts are defined by birth year rather than marriage year because different birth cohorts may have experienced varying educational opportunities, fertility attitudes, and policy environments. Additionally, since women’s marriage and fertility behaviors are more directly influenced by biological age than men’s, we categorize birth cohorts based on the wife’s birth year rather than the husband’s.

To maximize the use of valid information, we integrate six waves of data from 2010 to 2020, covering a total of 20,177 couples (40,354 individual samples) to describe and analyze the patterns of couples’ educational matching and their cohort trends. To analyze the relationship between couples’ educational matching patterns and fertility, we employ the most recent 2020 survey data to ensure timeliness and accuracy. After filtering and processing, we obtained an initial sample of 13,167 couples meeting the conditions and possessing education information. Since the vast majority of Chinese couples have at least one child, we excluded any samples with no children or missing child information (502 couples in total, including 146 couples where the wife was born before 1980), any samples where the wife’s age at first birth was under 15 (52 couples), and any samples with missing control variables (90 couples). The final valid sample size was 12,523 couples.

Variable measurement

We employed two dependent variables: (1) the number of children ever born, and (2) the probability (or hazard) of having a second child, coded as “1” if the couple had two or more children and “0” otherwise.

The core explanatory variable is the couple’s educational matching pattern, that is, the pairing combination of the spouses’ educational levels. Following common practice in similar studies, the highest education level attained is categorized into four tiers: primary school and below (including primary school, illiterate or semi-illiterate), junior high school, senior high school and junior college and above (including junior college, bachelor’s, master’s, and doctoral degrees). Regarding the measurement of spousal educational matching types, some research suggests considering each spouse’s relative educational status within their gender and birth cohort to improve comparability across cohorts (Dong and Xie 2023). However, the absolute educational status is more strongly correlated with income potential, bargaining power within the couple, and fertility decisions. Therefore, we adopted a practice common to the vast majority of the studies in this field—measuring marital matching types by comparing the spouses’ nominal educational levels. The educational matching patterns are classified into three categories: homogamy (both spouses have the same education level); upward marriage or hypergamy, gradient marriage (the wife’s education is lower than the husband’s); and downward marriage for women, hypogamy (the wife’s education is higher than the husband’s). Homogamy is subdivided into four subcategories based on education level: both spouses having primary school educations and below, junior high school, senior high school, and junior college or above. Additionally, for gradient/hypergamous and downward/hypogamous marriages, we further distinguished the education gap between spouses (one-level difference versus two or more levels difference) to conduct robustness checks.

To better examine the relationship between education and fertility, this study includes a set of control variables in the model that are relevant to reproductive behavior. These variables are: age at first birth, gender of the first child (coded as “1” for daughter, “0” for son), ethnicity (coded as “1” for ethnic minorities, “0” for Han) and residence type (urban vs. rural). We selected these control variables because they are associated with both educational attainment and fertility. Additionally, due to variations in the enforcement of fertility control policies across provinces, as well as differences in the policy, economic, social, and cultural environments experienced by different birth cohorts, the model also controls for province and birth cohort fixed effects.

Model specification

and methodology

For the two dependent variables—number of children and the probability of having a second child—we employed Poisson regression and the Cox proportional hazards model, respectively. For the variable ‘number of children’, descriptive analysis (with a mean of 1.87 and a variance of 0.77) and an over-dispersion test indicated that the conditions for applying Poisson regression were met.

Building on this foundation, in order to address the issue of right-censoring in fertility events, we used event history analysis to examine the time from first birth to the birth of a second child (measured in months). For the second-birth event, the risk onset is defined as the time of first birth, and the endpoint is the time of the second birth. If a second child had not been born up through age 49 or at the time of the survey, it was right-censored. A Schoenfeld residual test revealed no clear trend in the residuals over the ordered survival time. Moreover, the log–log plots of survival functions across different educational matching patterns are approximately parallel, and the survival curves predicted by the Kaplan–Meier method and the Cox model closely align. These results indicate that using the flexible and intuitive Cox proportional hazards model to examine second-birth risk is appropriate. In addition, to correct for potential heteroskedasticity, we cluster the observations within primary sampling units during the regression analysis to obtain robust standard errors.

Descriptive analysis

Education matching patterns

and cohort changes

As the rate of educational attainment among women rapidly surpasses that of men, the historical pattern in which men have averaged higher education levels than women are changing, and the gender education gap is gradually narrowing. This trend suggests that the relative changes in education levels between men and women among both the overall and the marriageable population will be reflected at the couple level, leading to reversals in educational gaps between spouses. The data in Table 1 confirm this point. The table shows that the educational composition of both husbands and wives has been continuously upgrading, with the proportion of those with higher education (college degree or above) steadily increasing. Mirroring the trend of the narrowing gender education gap, the educational disparity between spouses has gradually decreased, even to the extent that wives’ education levels have surpassed those of their husbands. Starting with couples born after 1985 (“post-85” couples), the proportion of women receiving higher education has already exceeded that of their spouses. This demonstrates that significant changes have occurred in the relative educational status of men and women at both the population level and within couples.

Unlike the general narrowing and reversal of the gender gap in educational attainment, the narrowing or reversal of educational differences between spouses more directly reveals the profound transformation in educational matching patterns. Figure 1 illustrates the dynamic evolution of educational matching patterns across birth cohorts. From Fig. 1a, it can be observed that educational homogamy continues to dominate in Chinese society. In most birth cohorts, the proportion of educationally homogamous marriages reaches or exceeds 50%. This aligns with previous research findings that homogamy remains the dominant pattern in marital matching in China (Li 2008; Shi 2019; Qi and Niu 2012). However, the cohort trend also reveals some noteworthy new developments:

First, educational homogamy exhibits a curvilinear trend—initially declining, then rebounding, and subsequently declining again at a slower pace. Notably, there is a significant drop in the proportion of homogamous marriages among those born in the 1950s, which supports earlier research findings on the weakening intensity of educational homogamy (Raymo and Xie 2000). Starting from the birth cohort of 1965–1969, women’s hypergamy and homogamy show a trend of one rising while the other falls. That is, homogamy has rebounded and increased, which is consistent with the upward trend in educational homogamy observed among cohorts who had married in 1990 or later (Li 2008). However, among the youngest couples (wives born between 1990 and 1994), the proportion of homogamous marriages has declined slightly from the peak of 57.0% for those born between 1970 and 1974 to 52.6%.

Secondly, within heterogamous marriages, an apparent relative change has occurred between female hypergamy (marrying up) and hypogamy (marrying down). Since the 1965–1969 birth cohort, the proportion of women marrying down has been steadily rising and has even begun to surpass the more traditional pattern of hypergamy. In the youngest cohort, the share of hypergamous marriages has declined to 21.7%, while hypogamous marriages have increased to 25.8%. This aligns with findings from Qi Xu (2022), which reported that among couples married between 2014 and 2018, 25.3% were in hypogamous marriages—surpassing the 23.3% in hypergamous marriages. In fact, a study using census and survey data from 1960 to 2011 across 120 countries showed that in many countries—including France, Argentina, Indonesia, Kenya, South Korea, and the United States—younger cohorts of women were more likely to marry down than up, indicating a nearly global trend (with the exception of Africa) (Esteve et al. 2016). It is important to note that women born between 1990 and 1994 in this study were aged between 25 and 30 at the time of the survey, meaning that some—particularly those with higher education—had not yet married. Therefore, the estimates of educational matching patterns for this cohort may be subject to measurement error. However, compared with less-educated women, highly educated women who delay marriage tend to prefer either homogamous or hypogamous unions. This suggests that we may be underestimating, rather than overestimating, the trend of female educational hypogamy, and its continuous upward trajectory is credible. Of course, educational hypogamy among women may mask upward marriage patterns regarding occupational status and income (Xu 2022), as highly educated women are more likely to marry high-income men. Future research could explore this issue in greater depth.

Finally, Fig. 1b reveals significant changes within educationally homogamous marriages. With the implementation and enforcement of the Compulsory Education Law, the most notable shift has been the drop in the proportion of couples in which both spouses have an education level of primary school or below—from 57.5% in the oldest birth cohort to just 7.4% in the youngest cohort. Meanwhile, the proportion of couples in which both partners have attained junior high school education has steadily increased. Subsequently, due to the rapid expansion of higher education, the share of couples in which both partners have attained education at the college level or above has continuously risen since the 1975–1979 birth cohort.

In summary, the rapid increase in women’s educational attainment has outpaced that of men, leading to significant new trends and characteristics in marital matching patterns. The once-common homogamy among those with primary education or below has declined sharply, while homogamy among the highly educated has increased steadily. At the same time, the prevalence of traditional female-hypergamous marriages has decreased, whereas female-hypogamous marriages have become more common. These patterns align with trends observed in most developed countries and regions. This raises a question: in the context of changing gender structures in education and shifting educational matching patterns, has the negative gradient relationship between education and fertility also undergone new changes. How do marriage patterns and their transformations explain or predict fertility levels? Exploring these questions may help in assessing the applicability of existing theories within the Chinese cultural context, and offer fresh insight for improving and expanding fertility research, as well as for understanding broader trends in macro-level fertility change.

Fertility levels

across different matching patterns

Figure 2 reveals the fertility differences and cohort changes among various spousal educational matching patterns. Figure 2a shows that within homogamous marriages at different education levels, there is a distinct negative gradient relationship between education and fertility: as both spouses’ educational attainment increases, the number of children gradually decreases. Across all birth cohorts, couples in which both partners are highly educated (college degree or above) exhibit the lowest fertility levels, while couples with education at the primary school level or below have the highest fertility levels. Even when including the two types of heterogamous marriages in the comparison, the fertility level of highly educated couples remains consistently the lowest among all educational matching patterns, whereas the fertility level of low-educated couples (primary school/elementary or below) remains the highest.

At the same time, Fig. 2b reveals an interesting phenomenon: among heterogamous couples, there are significant differences in fertility patterns between traditional gradient marriages and female hypogamy (i.e., when the wife is more educated). In most birth cohorts, female hypogamous couples’ fertility levels are lower than those of gradient marriage couples. Meanwhile, the fertility levels of gradient marriages and female hypergamous couples (wife less educated) are generally comparable to those of homogamous couples. However, from a cohort perspective, the fertility disadvantage of female hypogamy is diminishing. Especially in the 1985–1989 birth cohort, the fertility level of female hypogamous couples is nearly on par with that of homogamous and hypergamous couples.

In addition, the rise in women’s educational attainment and structural changes in spousal educational matching patterns have jointly driven the fertility transition, marked by a decline in fertility rates. Throughout this process, both homogamous and heterogamous couples exhibit similar downward trends in fertility levels. Female hypogamous couples’ fertility levels have also decreased amid fluctuations. Data from this study’s sample show that the decline in fertility is pronounced among couples in educationally homogamous marriages: the average number of children dropped from 2.37 in the 1950–1954 birth cohort to 1.56 in the 1990–1994 birth cohort. This change stems not only from a decline in fertility across most educational levels, but is also intertwined with structural shifts in matching patterns, such as the drop in low-education homogamous marriages (elementary school or below) (see Fig. 1). Because fertility levels have declined across most matching patterns, the overall average number of children per couple in the sample decreased from 2.32 in the 1950–1954 birth cohort to 1.55 in the 1990–1994 birth cohort. At the same time, the proportion of couples with two or more children fell from 75.1 to 44.2%.

In summary, educational expansion and the reversal of the gender gap in education have brought about significant changes in spousal educational matching patterns. At the same time, there are strong associations between the educational attainment of both spouses, their matching patterns, and family fertility behaviors, providing these factors with substantial explanatory and predictive power regarding fertility decisions. Moreover, the relationship between educational matching patterns and fertility may differ across birth cohorts. These emerging changes and characteristics warrant further attention and discussion. The following sections provide in-depth regression analyses and tests to explore these issues.

Regression results

and discussion

Regression results

on the number of children

Building on the descriptive analysis, to further explore the fertility differences between homogamous and heterogamous marriages, as well as among homogamous marriages of different education levels, we conducted Poisson regression analysis on the number of children born. In this analysis, we used homogamous couples with higher education as the reference group.

The regression results in Table 2 indicate that educational marriage matching patterns influence couples’ fertility behavior. The results from the whole sample show that among all educational matching patterns, couples with higher education have the fewest children. However, the difference with homogamous couples with a high school education was not statistically significant. Meanwhile, homogamous couples with an education level of elementary school or below have the most children. This finding corroborates the existence of a clear negative gradient between spouses’ education level and fertility within the context of homogamous marriages. Additionally, compared to traditional female hypergamy (marrying up), female hypogamy (marrying down) is associated with a disadvantage in fertility. This conclusion is consistent with findings from studies conducted on the Chinese mainland, Taiwan, and Hong Kong (Sun and Zhao 2025; Tsou et al. 2011), as well as in Central and Eastern European countries such as Austria and Bulgaria (Osiewalska 20172018). Overall, these results support Becker’s theory of the gender division of labor in the family, and do not align with the assumptions of the dual-earner household model.

However, a more in-depth analysis from a cohort perspective reveals that the fertility gap between highly educated couples and those in other matching patterns is gradually narrowing. The regression results of the five birth cohort subsamples show that as the cohorts progress, some of the regression coefficients for spousal educational matching patterns lose their statistical significance. Moreover, those that remain significant also tend to decrease in magnitude. For example, the regression coefficient for homogamous couples with primary school education or below drops to 0.205 in the 1980–1989 cohort sub-sample. Although still significant, this is much lower than the corresponding coefficients in the 1960–1969 cohort (0.419) and the 1970–1979 cohort (0.331). By the 1990–1994 cohort, both the coefficient and its significance decline further. Other educational matching patterns exhibit similar trends. This suggests that the negative correlation between spousal education level and fertility is gradually weakening. The traditional fertility advantage of female hypergamy and the disadvantage of female hypogamy are also diminishing. Although a complete reversal of the negative correlation between education and fertility has yet to be observed, these changes point toward the potential emergence of new trends in the future, warranting further observation.

In addition to educational matching patterns, the regression results reveal other factors influencing fertility behavior. Among these, the postponement of a woman’s age at first birth significantly suppresses fertility levels. This may be due to a shortened reproductive window caused by delayed first childbirth, which in turn reduces total fertility. It may also reflect a broader trend among modern women who prefer later marriage and childbearing, or favor smaller families. Furthermore, consistent with expectations, families whose first child is a daughter exhibit higher overall fertility levels and a greater likelihood of having additional children. This partly reflects the “son preference” phenomenon in Chinese fertility culture, and may also be linked to the specific conditions for having a second child under strict birth control policies. At the same time, fertility levels in urban areas are lower than in rural areas, while women from ethnic minority groups and their families tend to have higher fertility. These results reflect not only differences in fertility policies between urban and rural areas and among ethnic groups, but also the diversity of fertility preferences, attitudes, and cultural norms. After controlling for the influencing factors of fertility behavior, as well as the fixed effects of birth cohorts and provinces, we can more accurately reveal the relationship between educational marriage matching patterns and fertility behavior.

In summary, fertility levels in Chinese families decline not only with the rise in women’s educational attainment, but also, compared to women marrying down, marriages in which women marry up (i.e., aligning with traditional gender roles and norms) are more conducive to childbearing. This finding supports the assumptions of Becker’s neoclassical model, but does not align with the predictions of the dual-earner family model. This discrepancy may be related to traditional gender culture and family division of labor in Chinese society. However, amid modernization, with the continuous transformation of women’s social status and roles, the educational matching patterns between spouses, gender norms and division of labor, and fertility decision-making mechanisms are also rapidly evolving. As a result, the negative correlation between women’s education and fertility across successive birth cohorts has weakened.

Regression results

on the risk of having a second child

The above analysis reveals the significant effect of spousal educational matching patterns on fertility behavior, though the extent of this effect has weakened. However, as previously noted, couples in the younger cohorts—particularly those who have received higher education—may be subject to right-censoring issues in terms of fertility behavior. To address this problem, we employ the Cox proportional hazards model to conduct a regression analysis of the probability of having a second child, and compare the differences across birth cohorts.

The Kaplan–Meier survival curves for the event of having a second child show that the survival curve is steep at the initial stage of the risk period, and then gradually levels off. The median survival time is 65 months, with an average of 75 months. Notably, among couples in homogamous marriages where both spouses have only primary school education or below, the probability of having a second child is significantly higher than in other educational matching combinations, with a much steeper survival curve. In contrast, highly educated couples show the lowest probability of having a second child, and their transition from first to second birth takes the longest and occurs at the slowest pace. On this basis, Table 3 further reports the Cox regression results for the probability of having a second child. The first column presents the regression results for the full sample, while the subsequent columns provide analyses for subsamples by birth cohort, revealing cohort differences in the relationship between couples’ educational matching patterns and fertility behavior.

Based on the full-sample regression results, the Cox proportional hazards model—which corrects for right-censoring in the sample—yields findings largely consistent with the earlier Poisson regression results. Among all spousal educational matching types, couples in homogamous marriages with high education levels have the lowest probability of having a second child, whereas those in homogamous marriages with an education level of primary school or below has the highest probability. Additionally, couples in female-hypogamous (woman marries down) marriages have a lower probability of having a second child compared to those in traditional gradient marriages (woman marries up). This suggests that highly educated couples are least likely to have a second child, and have the longest gap between the first and the second birth, resulting in a longer interval between births. This finding contrasts with some studies from other countries that indicate that highly educated women or couples tend to postpone the first birth but quickly proceed to a second birth (e.g., Nitsche et al. 2018). Instead, these results further validate the theoretical predictions of the neoclassical family economic model. Moreover, the findings underscore the significant role of women’s educational expansion and changes in marital matching patterns in promoting fertility transition.

However, it is important to note that the relationship between spousal educational matching patterns and the probability of having a second child shows some cohort-specific variation. In the subsample of the cohort born in the 1950s, the differences in the probability of having a second child across different educational matching patterns are minor. Seventy-five percent of these couples choose to have two or more children, which may be influenced by traditional fertility norms. In contrast, for birth cohorts who reached their childbearing years after the start of the reform and opening-up (such as those born in the 1960s, 1970s, and 1980s), economic rationality has played a more prominent role in fertility decisions, leading to more pronounced fertility differences across different educational matching patterns. Yet, as cohorts continue to shift, these differences tend to narrow, and the negative correlation between education and fertility is gradually weakening. Compared to couples with high educational attainment, the regression coefficients for homogamous marriages where both spouses have a senior high school education are no longer statistically significant in the “post-80 s” and “post-90 s” cohorts. Additionally, the coefficients for homogamous marriages at the junior high school and primary school or below levels have decreased. Moreover, in the “post-90 s” subsample, there is no significant difference in the probability of having a second child between women who married down and highly educated homogamous couples. This trend may be attributed to intensified social competition under a market economy and rising costs associated with childbearing and childrearing. As wives’ economic contributions to the household become increasingly important, highly educated couples may be better equipped to afford the costs and challenges associated with having children. In this context, the dual-income, resource-concentrated family model may be starting to influence younger couples’ fertility decisions.

When examining the fertility decisions of heterogamous couples in greater detail, we focus particularly on the changing trends in female-hypogamous marriages. Further regression results, using female-hypergamous (traditional gradient) marriages as the reference group, reveal that female-hypogamous couples face distinct disadvantages in terms of fertility, both in the full sample and within the subsample born in the 1960s (see Table 4). This disadvantage may arise from the higher opportunity costs and economic pressures associated with childbearing. Notably, within female-hypogamous marriages, women appear to have greater bargaining power in fertility decisions. When they perceive that their rights and interests are not adequately protected under traditional gender role norms—such as being expected to shoulder the majority of household and childcare responsibilities—they may choose to limit the number of children they have. Consequently, we observe that within female-hypogamous marriages, the greater the educational gap between spouses (especially when the wife’s education exceeds the husband’s by two or more levels), the lower the probability of having a second child. This pattern is consistent across the total sample and the “post-60 s” cohort, and aligns with findings from other scholars, such as Linhao He and Meng Chen (2021).

However, starting from the “post-70 s” cohort, a significant change emerges: the fertility disadvantage of female-hypogamous marriages is no longer statistically significant compared to (traditional) gradient marriages. This shift suggests that with economic progress and evolving gender roles and family division of labor, the fertility mechanisms in female-hypogamous marriages may be undergoing a transformation. The general rise in education levels often accompanies more egalitarian gender attitudes and perceptions of family roles. At the same time, women with higher education—especially those who are more educated than their husbands (i.e., in female-hypogamous marriages)—typically have stronger bargaining power in seeking to outsource household chores or spousal assistance. Multiple studies have also noted that female-hypogamous marriages can promote greater male involvement in housework (Niu 2017; Tong and Liu 2015). In recent years, Chinese families have also shown a trend toward more equal gender division of childcare time (Zhao 2018). Research from Taiwan further shows that husbands’ participation in childcare significantly increases women’s (and particularly highly educated wives’) willingness to have a second child (Cheng and Hsu 2020). These findings indicate that an egalitarian dual-earner family model can alleviate, or even offset, the increased opportunity costs associated with childbearing. Accordingly, over time, the fertility disadvantage of female-hypogamous marriages has gradually weakened, and may ultimately disappear altogether.

Despite progress in gender equality, traditional gender role norms persist. This may explain why the negative correlation between education and fertility has eased but has not fundamentally changed. China is currently undergoing a transition toward a more gender-equal society, but the transformation of gender attitudes has been uneven, with male spouses’ views lagging behind, which constrains the progress of gender equality. As women gradually gain advantages in education and income, they acquire more opportunities to outsource household chores, including purchasing services from the market and seeking intergenerational support. However, against the backdrop of rapidly changing parenting concepts and intensifying educational competition, the responsibilities mothers bear—especially in children’s extracurricular learning—have increased (Yang 2018). This phenomenon has led to a growing childrearing burden, particularly in the “nurturing” aspect. Data from CFPS 2020 show that wives still bear much more responsibility than husbands in activities such as preschool pick-up/drop-off and homework supervision. Moreover, there is a lack of support for an equal gender division of labor at both social and institutional levels. Some policies even implicitly reinforce women’s role as the primary caregivers for the family and children. For example, while professional women are entitled to relatively long maternity and parental leave, men’s paternity (care) leave and parental leave are generally shorter and often difficult to utilize, with many fathers not taking parental leave at all. In the absence of adequate institutional support for balancing work and family, gender equality in the family domain becomes increasingly crucial for fertility decision-making. Future research should explore this issue in depth, including through detailed analysis and validation.

This study provides a detailed and in-depth exploration of how couples’ educational matching patterns influence fertility behavior, with particular attention to the differences observed across birth cohorts. While previous research has touched on related topics, much of it has focused on Western countries, and especially European contexts. Moreover, it has insufficiently considered dynamic changes over time, failing to capture fertility behavior variations engendered by the passage of time or differences among birth cohorts. Using household-level survey data, this study reveals the specific effects of couples’ educational matching on how many children they have and the decision to have a second child. The results show that couples with higher education levels exhibit a significantly lower willingness to have a second child, alongside a longer interval between the first and second births. Additionally, we find that female hypogamy is associated with certain fertility disadvantages, but this disadvantage is gradually weakening.

Education is intertwined with multiple dimensions such as economics, culture, values, and individual preferences. The rise in women’s educational attainment has led to changes in marriage matching patterns, which, through the interaction of these factors, have exerted profound influence on fertility behavior. The findings of this study not only add new empirical evidence to the existing literature, but also highlight the need for further research. Future studies should delve deeper into the underlying mechanisms by which couples’ educational matching patterns drive differences and changes in fertility behavior. By uncovering these mechanisms, more solid and comprehensive theoretical and empirical support can be provided for relevant policy-making and academic discussions, thereby promoting high-quality population development and supporting the path of modernization with Chinese characteristics.

Conclusion and policy implications

In the wake of women’s educational attainment and the evolution of couples’ educational matching patterns, there have been complex effects on fertility decisions. In this study, we adopted a cohort perspective to conduct an in-depth analysis of the dynamic changes in couples’ educational matching patterns, and their intrinsic relationship with fertility behavior.

The findings indicate that with the continuous rise in women’s educational attainment, highly educated homogamous marriages and female hypogamy have emerged as new social phenomena. This shift has not only reshaped the structure of couples’ educational matching but also influenced family fertility decisions and overall fertility levels. Becker’s neoclassical household economic theory, particularly the gender division of labor model, still partially explains the fertility behavior patterns observed in China during the past few decades. The negative correlation between couples’ education and fertility predicted by this theory is confirmed in this study: highly educated couples tend to have fewer children and are less likely to have a second child. At the same time, we observed that, under certain social contexts, couples in female hypogamous marriages tend to have lower fertility than those in traditional gradient marriages or homogamous marriages. This may reflect the persistence of traditional values and gender divisions of labor in certain periods or in earlier birth cohorts within Chinese society, where women with higher education (or higher education than their husbands) are more likely to negotiate within the couple to fulfill the desire for fewer children.

This study also reveals a trend: among younger couples, fertility mechanisms seem to be shifting from the traditional gender division of labor model toward a more egalitarian resource-concentrated or dual-income family model. With the progress of time and cohort renewal, highly educated couples’ fertility disadvantage has gradually diminished, and the fertility disadvantage associated with female hypogamy has also declined, or even disappeared, among younger cohorts. This change is intertwined to the ongoing transformation of gender roles and gender relations. In modern society, with the escalation in housing, childbearing, and childrearing costs, women’s economic role within the family has become increasingly important. For many families, the wife’s economic contribution is indispensable to maintaining living standards and realizing fertility aspirations. Therefore, among younger cohorts, neither highly educated homogamous couples nor female hypogamous couples show significant fertility disadvantages, reflecting a transition in gender roles and family division of labor from traditional to more egalitarian patterns.

At the theoretical level, this study’s findings both challenge and complement existing theories. Although Becker’s theoretical model provides an important framework for understanding fertility decisions, it treats the couple as a unified entity lacking any internal conflicts of interest. It also implicitly assumes the woman’s role to be the homemaker. This assumption overlooks the significant role women now play in family economics and fertility decision-making. Therefore, for a comprehensive understanding of contemporary family fertility behavior, we must move beyond the limitations of Becker’s neoclassical theory. Meanwhile, Oppenheimer’s dual-earner family model offers a new perspective, emphasizing the positive impact of women’s rising economic role within the family on fertility behavior. However, the realization of this model depends on achieving an egalitarian division of labor within the household. In fact, as long as spouses cannot equally share household chores and childrearing costs, the bargaining framework continues to provide a powerful explanation for fertility decisions. Hence in this study, we have adopted a broader perspective to examine the influence of gender roles, educational matching patterns, and the family division of...