This study explores how changes in sibship composition associated with fertility decline may, in conjunction with entrenched family norms and expectations associated with specific sibship positions, impact marriage rates and further reduce fertility. We evaluate this possibility by focusing on Japan, a society characterized by half a century of below-replacement fertility and widely shared family norms that associate eldest (male) children with specific family obligations. Harmonic mean models allow us to quantify the contribution of changes in both marriage market composition with respect to sibship position and sibship-specific pairing propensities to the observed decline in marriage rates between 1980 and 2010. One important finding is that marriage propensities are lower for those pairings involving men and women whose sibship position signals a higher potential of caregiving obligations, especially only-children. Another is that changes in marriage propensities, rather than changing sibship composition, explain most of the observed decline in marriage rates. We also found that marriage propensity changes mitigate the impact of the changing sibship composition to some extent. However, the limited contribution of changing sibship composition to the decline in first-marriage rates provides little support for a self-reinforcing fertility decline via the relationship between changing sibship composition and marriage behavior.
Research on the economic implications of very low fertility typically focuses on labor shortages (McDonald and Kippen 2001), slower economic growth (Bloom et al. 2010; Bloom and Finlay 2009), and the difficulty of sustaining public pension and health care programs (Bongaarts 2004). These economic ramifications may further reduce fertility via a self-reinforcing mechanism, the low-fertility trap, in which a complex array of economic, social, and demographic pathways combine to reduce desired fertility and thereby prolong periods of below-replacement fertility (Lutz 2008; Lutz et al. 2006).
However, conventional articulations of the low-fertility trap hypothesis do not explicitly recognize two other demographic processes that might support a self-reinforcing pattern of low fertility in some contexts. First, in most cases, the emergence of below-replacement fertility is associated with smaller sibship size (Präg et al. 2020) and an increase in only-children, eldest children, and single-sex sibships (Allendorf 2020; McHale et al. 2012; Pandian and Allendorf 2022). Second, declining marriage rates contribute to lower fertility in settings where the link between marriage and childbearing remains strong. As we will discuss, these two processes combined may play a particularly salient role in the reinforcement of low fertility in societies where sibship position is imbued with well-established and widely understood social meanings and normative obligations.
Our goal in this study is to integrate two strands of research: (1) the low-fertility trap literature and (2) work on the interplay of declining fertility, changing sibship size and composition, and social change in gender and family norms, expectations, and obligations (e.g., Allendorf 2012, 2020). Specifically, we evaluate the possibility that by altering the composition of the marriage market with respect to sibship characteristics, low fertility may contribute to lower marriage rates in Japan, where nearly all childbearing occurs within marriage. A key assumption underlying this novel scenario and our focus on Japan is that specific sibship characteristics are associated with established family norms, expectations, and assumptions about character traits that convey meaningful information in the marriage market.
Japan is of particular interest because it is characterized by (1) a half-century of below-replacement fertility due primarily to later and less marriage in combination with stable marital fertility and negligibly low levels of nonmarital childbearing (Raymo et al. 2015; Tsuya and Mason 1995);1 (2) relatively low public spending on families, reflecting and reinforcing normative expectations of intrafamilial provision of support (Esping-Andersen 1997; Peng and Wong 2008; Yoda 2022); and (3) patriarchal family norms associating family obligations with specific sibship positions (Raymo et al. 2015; Tsuya and Bumpass 2004; Yasutake 2010). Of central importance are long-standing and widely recognized expectations that firstborn sons (chōnan), or eldest daughters (chōjo) if the parents have no sons, coreside with parents, maintain the family lineage, and support parents in old age (Bumpass 1994; Kato 2013; Kurosu 1994; Martin 1990; Rindfuss et al. 2004; Taeuber 1958; Tsuya and Bumpass 2004).
Abundant anecdotal evidence suggests that many young women in Japan find the prospect of caregiving for aging parents-in-law to be increasingly unappealing.2 If so, men in sibship positions that signal potential caregiving obligations (i.e., eldest sons) may be evaluated less favorably in the marriage market, especially by those women most likely to have expectations of competing caregiving obligations for their own parents: eldest daughters with no brothers (Jones 2007; Yu and Hertog 2018). Alternatively, pairing behavior may have adjusted to reflect the new realities of the marriage market via an increasing propensity to form marriages between men and women whose respective sibship positions signal potential caregiving obligations.3
Supporting the possibility of this adjustment in pairing behavior, normative expectations of intergenerational coresidence and caring for older parents/parents-in-law are less rigid than in the past and increasingly involve strategic coordination and negotiation among siblings reflecting individual preferences and life circumstances (Kureishi and Wakabayashi 2010; Takagi et al. 2007; Wakabayashi and Horioka 2009). Assuming that siblings negotiate and share expected filial responsibilities for aging parents, individuals of the same sibship status (e.g., eldest sons) may be evaluated differently in the marriage market depending on their number of siblings. Thus, only-children are of particular interest because they are not only eldest sons and eldest daughters with no brothers, by definition, but also have no siblings with whom they might share or negotiate expected caregiving obligations. In this context, only-children may be viewed as the least desirable potential mates.
In this article, we examine these possibilities by addressing two empirical questions. First, to what extent have changes in the distribution of sibship position in the marriage market contributed to the observed decline in first-marriage rates in Japan? Second, to what extent have changes in spouse pairing behavior with respect to sibship position exacerbated or mitigated the contribution of compositional change in the marriage market to the decline in first-marriage rates? To answer these questions, we begin by locating our emphasis on changing sibship composition within the broader literature on implications of demographic change for the reinforcement or acceleration of fertility decline. Using data from the Japanese National Fertility Surveys (JNFS), we then provide a descriptive summary of patterns of spouse pairing by sibship position. Next, we use harmonic mean models of marriage to quantify the contribution of changes in both marriage market composition and pairing propensities (with respect to sibship position) to the observed decline in first-marriage rates between 1980 and 2010.
Results showed that both men and women are less likely to marry partners whose sibship position signals stronger normative expectations of future family care-giving obligations, especially only-children. We also found that changes in pairing propensities have played some small role in mitigating the potential impacts of changing sibship composition in the marriage market. However, the contribution of marriage market mismatches generated by changes in sibship composition to the declining first-marriage rates is not strong and is found only for women's marriage rates. Overall, results provide little support for a self-reinforcing fertility decline via the relationship between changing sibship composition and declining marriage rates.
Finally, we discuss what our results reveal about the possibility that low fertility may be self-reinforcing via its impact on the marriage market's changing sibship composition in societies where marriage and fertility remain closely linked and social meanings and obligations attached to specific sibship positions are widely recognized. Our focus on normative expectations associated with different sibship positions is specific to Japan, but similar relationships exist in other patrilineal East Asian societies. Indeed, these relationships may be even more relevant in countries such as South Korea, where fertility decline is both more recent and more rapid and where the social meanings associated with sibship position arguably remain stronger than in Japan.
Demographic Change, Marriage Market Mismatches, and Their Implications for Low Fertility
Efforts to understand prolonged periods of below-replacement fertility in wealthy countries have emphasized the low-fertility trap, a self-reinforcing mechanism in which demographic, social, and economic changes accompanying low fertility combine to further fertility declines (Lutz 2008; Lutz et al. 2006). Central to this hypothesis is social learning and the diffusion of social norms favoring the smaller family sizes that accompanied earlier reductions in fertility. Deteriorating economic conditions for younger cohorts may also play a key role in this low-fertility cycle by increasing the gap between consumption aspirations and expected income (Lutz 2008; Lutz et al. 2006), which is associated with later and less childbearing.
However, a central focus on declines in ideal family size is not universally relevant. For example, in Japan, the total fertility rate has long been below replacement level, but the ideal number of children has declined only slightly. Recently married women in the JNFS reported their ideal number of children to be 2.42 in 1977 and 2.11 in 2021 (NIPSSR 2022b).4 Nearly 50 years of below-replacement fertility in Japan has clearly not translated into the meaningful reductions in desired family size that the low-fertility trap framework emphasizes. Of course, other components of the low-fertility trap hypothesis—especially negative population momentum and declining economic prospects (Lutz 2008)—are of central importance in the Japanese case (Jones 2019). However, our goal in this article is to consider a novel extension of the notion of low fertility's self-reinforcing processes. Here, we focus on how changes in sibship composition and size accompanying fertility decline may contribute to marriage market mismatches and declining marriage rates and thereby to the continuation of low or declining fertility rates. This hypothesis retains the basic idea of self-reinforcing fertility decline from the low-fertility trap literature, integrates it with insights from research on social change due to sibship composition changes accompanying low fertility (Allendorf 2012, 2020), and situates it in the distinctive context of family, marriage, and fertility in Japan.
Sibship Position, Normative Expectations, and Changing Marriage Market Composition
The widespread desire to avoid normative expectations of living with and caring for parents-in-law may contribute to declining aggregate marriage rates via two mechanisms, one demographic (compositional) and one behavioral. First, in a context where filial norms and obligations associated with specific sibship positions (and attitudes toward those norms and obligations) remain relevant, the increasing prevalence of men and women in such positions may generate a marriage market mismatch in which pairings not involving (competing) normative expectations of coresidence or caregiving obligations become numerically more difficult. Second, behavioral responses to this compositional change in the marriage market may either exacerbate or mitigate the impact of the posited mismatch.
In Japan (and other patrilineal East Asian societies), the first mechanism may be especially relevant for understanding the implications of an increasing prevalence of eldest sons and eldest daughters with no brothers. Eldest sons have long been expected to live with and support their own parents (Atoh and Kojima 1983; Kojima 1992; Martin and Tsuya 1991) in tacit exchange for the inheritance of the family home and business (Wakabayashi and Horioka 2009). Numerous studies suggest that the material benefits of marriage for the wives of eldest sons (chōnan no yome) may be offset by the stress accompanying coresidence with their parents-in-law and the expectations of care provision (Campbell and Ingersoll-Dayton 2000; Furuya and Raymo 2022; Jenike 2003; Kamo 1990; Kurosu 1994; Park et al. 1999; Rindfuss et al. 2004; Traphagan 2003). Historically, second sons have followed eldest sons in this hierarchy of filial obligation (Rindfuss et al. 2004). In the case of no male offspring, eldest daughters were historically expected to marry a spouse who would carry on the family lineage as a son-in-law, who were sometimes adopted as sons (muko-yōshi) (Aruga 1954; Tsuya and Bumpass 2004; Tsuya and Choe 1991; Yasutake 2010).
To the extent that these family norms continue to hold salience and the prospect of caregiving obligations for in-laws is viewed unfavorably, only-children may be particularly disadvantaged in the marriage market by virtue of having no siblings who might assume or share caregiving responsibilities. Some empirical support for this conjecture can be found in recent analyses documenting lower rates of marriage among only-children (Kawata and Komura 2023) and showing that only-children (both men and women) registered with an online dating agency received significantly fewer responses from potential partners relative to those with siblings (Yu and Hertog 2018). Yu and Hertog (2018) also found that firstborn sons, but not firstborn daughters, were less likely to receive responses to date requests. These findings were based on data from a marriage-focused matching site, thereby providing an empirical basis for positing that unmarried men and women in Japan seek to avoid potential expectations of future support for aging parents-in-law, especially when they may expect to have competing care responsibilities for their own parents (Jones 2007). Of course, this emphasis on potential caregiving obligations does not preclude other possible mechanisms through which men and women of specific sibship status are treated less favorably in the marriage market. One possibility is negative perceptions of personality traits thought to be more prevalent among only-children (Cameron et al. 2013).
Figure 1 provides further empirical motivation for our focus on potential marriage market mismatches based on sibship position. The impact of declining fertility is clear from this summary of trends in the distribution of sibship positions by sex and birth cohort among JNFS respondents.5 For example, the percentage of eldest sons without brothers increased from 19% in the 1945–1949 birth cohort to 33% in the 1995–1999 cohort, and the percentage of eldest daughters with no brothers increased from 12% to 17% for the same birth cohorts. Only-children roughly doubled in prevalence, from 5% to 11% for men and from 6% to 9% for women. If we consider all sibship positions associated with normative caregiving expectations, the prevalence increased from 52% to 71% for men (all but younger sons) and from 18% to 27% for women (only-children and eldest daughters with no brothers). Equivalently, the prevalence of those in sibship positions that do not signal potential caregiving obligations decreased from 48% to 29% among men (younger sons) and from 82% to 73% among women (younger daughters with no brothers and daughters with brothers). If we counterfactually assume random pairing and universal marriage, the percentage of pairings potentially involving competing caregiving obligations (i.e., all pairings involving eldest sons and eldest daughters with no brothers) would double across the birth cohorts included in Figure 1, from 9% in the oldest cohort to 19% in the youngest cohort (not shown).6
Behavioral Response to Demographic Change
An implicit assumption in the preceding discussion is that the desire to avoid marriage to men (women) with potential caregiving obligations was relatively stable over time. Although a wide array of social and economic factors shape mate selection preferences (Kalmijn 1998; Schwartz 2013), we consider whether the shifting composition of potential mates with respect to sibship position has been accompanied by changes in how those characteristics are evaluated in the marriage market. Drawing on previous studies' findings, we suggest two possible patterns of change. The first is a stable or increasing propensity to avoid marrying partners whose sibship position signals potential obligations of intergenerational coresidence and caregiving. This scenario might result from changing attitudes or an increasing concern about potentially competing obligations to care for both the parents and the parents-in-law. The second pattern of change is an adjustment of pairing propensities to reflect changes in marriage market composition. Just as men might respond to a conventional marriage squeeze by increasingly marrying women who are the same age or older than themselves, a marriage market mismatch with respect to sibship position may induce a reduction in the propensity to avoid mates with potential caregiving obligations.
Increasing Propensity to Avoid Marriages With Eldest Sons and Eldest Daughters With No Brothers
A core tenet of writing on the second demographic transition is that attitudinal change in low-fertility societies typically involves a shift from collectivistic to individualistic values regarding family expectations and obligations (Lesthaeghe 2010). Such change has arguably been slower in Japan than in many countries, but it is happening nonetheless (Atoh 2001, 2005; Choe et al. 2014). If broad attitudinal changes include a growing distaste for intergenerational coresidence and associated caregiving obligations, we might expect the compositional changes in sibship position accompanying low fertility to result in the marriage market mismatch via the first behavioral response described in the previous section.
Direct measurement of such attitudes is rare (and often outdated), but some evidence hints at the potential relevance of this scenario. For example, the proportion of unmarried adults (aged 18–34) indicating a desire to coreside with their parents decreased from 71% to 65% for men and from 58% to 42% for women between 1982 and 1992 (NIPSSR 1983, 1989, 1994). More importantly, 35% of all unmarried women and 44% of women without siblings in that survey reported not wanting to live with their parents-in-law (NIPSSR 1989, 1994).7 Another nationally representative survey of married women younger than 50 revealed that the proportion agreeing that “it is a child's duty as a matter of course [to provide care to aged parents]” fell from more than 50% in the mid-1980s to 30% in the late 1980s through the early 2000s (Atoh 2001:12).8 These changes may reflect the diminishing economic advantages of marrying eldest sons or an increasing perception that coresidence with parents-in-law is unnecessary, detrimental to women's privacy and autonomy (Arai et al. 2000; Jenike 2003; Lebra 1984; Sugihara et al. 2004; Traphagan 2003), and stressful (Kamo 1990; Morgan and Hirosima 1983).
Adjustment of Pairing Propensities in Response to Changing Marriage Market Composition
Research on marriage squeezes and marriage market mismatches has long recognized that unmarried men and women facing a shortage of partners with preferred characteristics may respond by increasing the types of partners they are willing to marry. This casting of a wider net might involve an increased willingness to consider formerly married partners, partners with children, and nonnormative pairings with respect to age or educational attainment (e.g., female hypogamy) (Lichter et al. 1995; Qian et al. 2005; South 1991). There are several reasons to expect a similar response to a marriage market mismatch with respect to sibship composition in a setting like Japan, where specific sibship positions have been normatively associated with expectations of intergenerational coresidence and support obligations.
Of particular importance is evidence that intergenerational relationships and attitudes have changed over time and that decisions regarding coresidence and caregiving responsibilities have become increasingly motivated by strategic, rather than normative, considerations (e.g., Takagi et al. 2007). For example, intergenerational coresidence is no longer common soon after marriage and is increasingly initiated in response to an increased need for childcare support (Kureishi and Wakabayashi 2010), inheritance of land or a house from aging parents (Wakabayashi and Horioka 2009), or the death of a parent or parent-in-law (Takagi et al. 2007).9 Other factors contributing to a reduction in the immediate salience of coresidence and potential caregiving obligations include improvements in mortality (Ministry of Health, Labour and Welfare of Japan 2017) and healthy life expectancy (Tokudome et al. 2016; Yong and Saito 2009), as well as the implementation of the mandatory public long-term care insurance system and an observed decline in preferences among older Japanese for sons or daughters-in-law as caregivers (Long et al. 2009). These changes help explain evidence that the percentage of total caregiving for frail older parents provided by the spouse of a coresident child (typically the daughter-in-law) fell from 23% in 2001 to 8% in 2019 (Ministry of Health, Labour, and Welfare of Japan 2010, 2019).10 For all these reasons, we might expect that changes in marriage market composition have been accompanied by an increase in the propensity to marry a partner whose sibship position has traditionally been associated with potential caregiving obligations.
To evaluate the contributions of changes in sibship composition and specific sibship pairing propensities to declining marriage rates in Japan, we evaluate the following four hypotheses.
Hypothesis 1a: Women's (men's) propensity to marry eldest sons (eldest daughters with no brothers) is lower than for marriages to men (women) of other sibship positions.
Hypothesis 1b: Women's (men's) propensity to marry only-children is lower than for marriages to men (women) of other sibship positions.
These first two hypotheses articulate the foundational assumptions about marriage behavior behind our questions regarding changes in marriage market composition and pairing propensities.
Hypothesis 2: The decline in overall marriage rates between 1980 and 2010 is partly explained by an increasing prevalence of eldest sons and eldest daughters with no brothers (including only-children) in the marriage market, net of changes in pairing propensities.
If women's (or men's) desire to avoid marrying eldest sons (or eldest daughters with no brothers) has remained stable over time, we expect that some part of the marriage rate decline is explained (in a statistical accounting sense) by changes in marriage market composition (i.e., the increasing prevalence of eldest sons and eldest daughters without brothers, including only-children).
Hypothesis 3: The marriage rate decline between 1980 and 2010 is partly explained by a decline in the propensity for pairings that involve eldest sons and eldest daughters with no brothers (including only-children).
This hypothesis reflects the scenario in which changing attitudes and preferences (as reflected in pairing propensities, net of marriage market composition) exacerbate the contributions of changing marriage market composition expressed in Hypothesis 2.
Hypothesis 4: An increase in the propensity for pairings that involve eldest sons and eldest daughters with no brothers (including only-children) offsets the contribution of changing marriage market composition (with respect to sibship composition and size) to the marriage rate decline between 1980 and 2010.
This hypothesis reflects the scenario in which adjustments of pairing propensities in response to changing marriage market composition mitigate (to some degree) the contributions of changing marriage market composition expressed in Hypothesis 2.
We used pooled data from the 8th through 15th JNFS, conducted in 1982, 1987, 1992, 1997, 2002, 2005, 2010, and 2015. These surveys provide information on age, sibship position, and age at marriage for nationally representative samples of married women and unmarried men and women aged 18–34 in the 1982 and 1987 surveys and aged 18–49 in the 1992–2015 surveys.
To reconstruct the composition of national marriage markets over time, we used information on age, sibship position, and marriage year (if married) to create person-year observations for 15- to 49-year-old men and women. Because some years of the JNFS did not ask unmarried respondents about previous marriages, we begin by limiting the unmarried samples to never-married men and women. For each of these respondents, we constructed one person-year record for each year of age from 15 to age at the time of the survey. Additional restrictions and assumptions are required to construct similar person-year records for married men and women. Because only married women were surveyed, we used the information they provided about their husbands' age and sibship position to construct records for married men. Furthermore, because the JNFS did not ask remarried women about the timing of their first marriage or the age and sibship position of their first husband in some survey years, we restricted the sample to women in their first marriage. Similarly, because we have no information about remarried husbands' first marriage, we can include information about only those couples in their first marriage.11 After constructing individual records for each year of exposure to the risk of first marriage, we grouped observations into seven 5-year age groups (15–19, 20–24, . . . , 44–49) and used information on sibship size and composition for respondents and their husbands to classify men and women into four mutually exclusive and exhaustive categories of sibship position (as in Figure 1). For men, these categories are (1) only-child, (2) eldest son with no brothers, (3) eldest son with brothers, and (4) younger son. For women, the four groups are (1) only-child, (2) eldest daughter with no brothers, (3) younger daughter with no brothers, and (4) daughter with brothers. In these categories, “younger” does not mean “youngest,” given that this group includes both middle sons (or daughters) and youngest sons (or daughters). For both men and women, these categories are ordered from strongest to weakest expected caregiving obligations to one's own parents.
To examine change over time, we constructed marriage market data for three decades: 1980–1989, 1990–1999, and 2000–2009. We therefore constructed a total of 784 marriage rates (7 × 7 pairings by age group by 4 × 4 pairings by sibship position) for each of the three 10-year periods.
where is the number of first marriages for women (f ) who are age i, have sibship position k, and are paired with husbands age j with sibship position l in period t. Following Schoen (1988), the propensity for specific pairings is expressed as the force of attraction () and can be interpreted as the rate of encounters between men and women in the marriage market and the proportion of such encounters that lead to marriage (Qian and Preston 1993:483). Stated differently, reflects both the prevalence of different characteristics in the marriage market and how those characteristics are associated with marriage.
where is the first-marriage rate for women (f ) who are age i, have sibship position k, and are paired with husbands age j with sibship position l at period t. Marriage rates can thus be expressed as the product of the force of attraction and a measure of marriage market composition , which we will call an availability ratio (Raymo and Iwasawa 2005). Corresponding first-marriage rates for men () can be calculated by dividing in Eq. (1) by , resulting in an availability ratio of . Comparison of forces of attraction () for different pairings and their change over time allows for straightforward evaluations of Hypotheses 1a and 1b.
The first allows us to answer the counterfactual question, What would women's (or men's) age- and sibship position–specific marriage rates be in the 2000s if marriage market composition (with respect to sibship position) changed while pairing propensities (forces of attraction) remained constant since the 1980s? The second allows us to answer the counterfactual question, What would women's (or men's) age- and sibship position–specific marriage rates be in the 2000s if marriage market composition (with respect to sibship position) had not changed since the 1980s while pairing propensities changed as observed? Comparing the first set of counterfactual rates with observed rates for the 2000s and the second set of counterfactual rates with the observed rates for the 1980s allows us to evaluate Hypotheses 2–4. Given the marriage rates constructed for each period, we summarize the characteristic-specific pairings as total first-marriage rates (TFMR) representing the synthetic cohort probability of ever marrying by age 50.
Propensity to Marry by Sibship Pairings
Table 1 presents forces of attraction for specific sibship pairings. These values are calculated as the sum of the 49 possible age-specific forces of attraction for each of the 16 different sibship pairings. As described earlier, these forces of attraction can be interpreted as the propensity to form a particular marriage independent of the influences of marriage market composition.12 Looking first at forces of attraction in the 1980s, we see that for women in all sibship positions, the propensity to marry eldest sons is lower than that for younger sons. For example, eldest daughters without brothers had lower forces of attraction for pairings with eldest sons with and without brothers (.118 and .145, respectively) than for pairings with sons who have older brothers (.235). This finding is consistent with Hypothesis 1a. Further, consistent with Hypothesis 1b, forces of attraction for women in all sibship positions are lowest for pairings with men who are only-children.
An examination of changes over time shows waning support for Hypothesis 1a. The relatively lower forces of attraction for pairings with eldest sons are much less pronounced for the 1990s and 2000s. For women in all sibship positions, forces of attraction are similar for pairings with eldest sons (with and without brothers) and younger sons. This result reflects a combination of declining forces of attraction for pairings involving younger sons and small increases in forces of attraction for pairings involving eldest sons. However, forces of attraction for pairings with men who are only-children remain substantially lower than for other pairings. Net of marriage market composition (and changes therein), women have consistently lower propensities of marrying men who are only-children.
The patterns for men are somewhat different. Rather than looking across the rows of Table 1, we now look down the columns to compare forces of attraction across women's sibship positions for men of a given sibship position. In the 1980s, men had a lower propensity to marry women without brothers than women with brothers. For example, the forces of attraction among younger sons (column 4) were substantially lower for pairings with only-children, eldest daughters without brothers, and younger daughters without brothers (.196, .235, and .218, respectively) than for pairings with women who have brothers (.399).
Unlike our findings for women, however, these differences in men's pairing propensities by women's sibship position remain visible in the 1990s and 2000s. In all three decades, men are less likely (net of marriage market composition) to marry a woman whose sibship position signals a higher level of potential caregiving obligations to her parents. Consistent with our findings for women, forces of attraction for men's marriages to women who are only-children are lower than for all other pairings, regardless of men's sibship position. This finding holds across all three decades, although the magnitude of the difference with other sibship positions is lower than that for women. Therefore, Hypothesis 1b is supported in all three periods for men and especially for women.
Marriage Market Composition by Sibship Pairings
Table 2 presents availability ratios for specific sibship pairings in each of the three decades, separately by sex. The upper panel presents pairing-specific availability ratios for women (i.e., ). By definition, availability ratios are larger for pairings that involve larger groups of men (e.g., younger sons) and smaller groups of women (e.g., only-children). A comparison across decades shows relatively little change in pairing-specific availability ratios. The most important change is that suggested by the changing distributions of sibship position in Figure 1: a decline in the relative availability of men without potential caregiving obligations. Column 4 of Table 2 shows that for women's pairings with younger sons, the availability ratios are 3% (from .848 to .824 for only-child women) to 7% (from .403 to .373 for daughters with brothers) lower in the 2000s than in the 1980s. This relatively small change in marriage market composition reflects changes in the sibship position distribution and marriage behavior, especially the more pronounced delay in women's marriage relative to men's. Census data show that the increase in the proportion aged 25–34 who were never married has been more rapid for women than for men (not shown). This differential change in marriage timing, by definition, results in a decrease in the relative availability of potential mates for women (i.e., it increases the denominator of availability ratios for women).
The pattern of change in availability ratios for men in the lower panel of Table 2 thus differs from that for women. In particular, for all sibship pairings, availability ratios increased over time for men. For example, younger sons' availability ratios (column 4) between the 1980s and the 2000s show an increase of 5% (from .597 to .627) for marriages with daughters with brothers and 16% (from .301 to .349) for marriages with daughters without brothers. In other words, the marriage market became more advantageous over time for men of all sibship positions. This relative improvement in marriage market composition for men reflects the fact that the influence of the sibship composition changes described in Figure 1 (e.g., the increasing prevalence of eldest sons) was smaller than the impact of women's more pronounced marriage delay on increasing the numerator of availability ratios for men.
Careful evaluation of Hypotheses 2–4 requires three sources of information: (1) the period- and pairing-specific forces of attraction shown in Table 1; (2) the sex-, period-, and pairing-specific availability ratios shown in Table 2; and (3) the observed and counterfactual trends in TFMRs shown in Figure 2.13 As described earlier, comparisons of observed and counterfactual TFMRs allow us to evaluate the role of changing marriage market composition and changing marriage propensities in the overall decline in marriage across the three decades. More specifically, examination of pairing-specific contributions to differences in observed and counterfactual TFMRs allows us to evaluate the role of the specific changes of interest in Hypotheses 2–4.14
As in earlier studies (e.g., Fukuda et al. 2020), Figure 2 illustrates a large decline in observed TFMRs between the 1980s and 2000s for men and women. Over this 30-year period, the TFMR fell 20 points for men and 17 points for women. A comparison of observed TFMRs (red lines) and those that counterfactually hold forces of attraction constant while allowing availability ratios to change as observed (orange lines) shows that changing marriage market composition with respect to sibship position had similar implications. For women, the counterfactual decline in TFMR between the 1980s and 2000s (.945 – .921 = .024 points) is equal to 14% of the total observed decline (.945 – .779 = .166). For men, the counterfactual decline in TFMR (.985 – .966 = .019 points) is equivalent to 9% of the observed decline (.985 – .784 = .201).
The counterfactual TFMRs constructed by holding availability ratios (marriage market composition) constant at their 1980s levels while allowing forces of attraction to change as observed (green lines) show that changes in forces of attraction explain most of the observed decline in marriage. For example, for women, the counterfactual decline in TFMR (.945 – .808 = .137) corresponds to 83% of the observed decline in TFMR. An alternative interpretation of this counterfactual is that change in marriage market composition (availability ratios) accounts for 17% (100% – 83%) of the observed decline.15
The pattern for men is somewhat different: the counterfactual decline in TFMR when availability ratios are held constant is larger than the observed decline in TFMR (counterfactual decline of .985 – .730 = .255 is greater than the observed decline of .201). If forces of attraction had not changed as observed, holding marriage market composition constant at 1980s values would have resulted in an even more dramatic decline in men's TFMR. Stated differently, forces of attraction changed in conjunction with changing marriage market composition in ways that limited the decline in men's marriage rates as we discuss in more detail later. This pattern suggests possible support for Hypothesis 4.
To examine the contributions of changes in pairing-specific forces of attraction and availability ratios to the observed marriage decline, we turn to Table 3, which presents the contribution of each to the differences between the observed and counterfactual TFMRs shown in Figure 2. We begin this exercise by examining pairing-specific contributions to the difference between the counterfactual female TFMR calculated by holding availability ratios constant at their 1980s values (.808) and the observed TFMR for women in the 2000s (.779). Panel A of the table shows that this small difference of .029 is due to the availability ratio declines for pairings involving daughters without brothers, especially for their pairings with younger brothers (.052)—a pattern that is not consistent with the posited marriage market mismatch scenario we articulated. The negative contributions for all pairings involving potentially competing family obligations (i.e., pairings that do not include daughters with brothers or younger sons offset the observed decline in TFMR) indicate that the small declines in availability ratios for these pairings shown in Table 2 were offset by small increases in forces of attraction for these same pairings shown in Table 1. This pattern of interaction is consistent with Hypothesis 4.
In panel B (Table 3), where the difference between the counterfactual female TFMR is calculated by holding forces of attraction constant (.921) and the observed TFMR in the 2000s (.779), we can evaluate the contributions of pairing-specific change in forces of attraction. The zero or small and negative contributions for pairings involving potentially competing caregiving obligations reflect the small increases in forces of attraction for these pairings (Table 1). Nearly all the observed difference (.142) is due to the contribution of declining forces of attraction for all pairings involving younger sons. Without these changes in the propensity to marry younger sons, women's TFMR would have remained largely unchanged over this 30-year period. To summarize, results for women are largely inconsistent with Hypothesis 3 but provide some evidence consistent with Hypothesis 4.
Panels C and D (Table 3) present the corresponding contributions to differences between counterfactual and observed TFMRs for men. Panel C requires a somewhat different interpretation than the corresponding panel for women (panel A) because the counterfactual TFMR calculated by holding availability ratios constant at the 1980s levels (.730) was lower than the observed TFMR for the 2000s (.784). In panel C, negative values thus correspond to pairings that became numerically more feasible over time because of the more pronounced delay in women's marriage relative to men's. Changes in availability ratios for all pairings except those involving younger sons facilitated men's marriage. Without these changes in marriage market composition, the TFMR for the 2000s would thus be .054 lower than it was.
Panel D (Table 3) presents pairing-specific contributions to the difference between the counterfactual TFMR calculated by holding forces of attraction constant at their 1980s levels (.966) and the observed TFMR in the 2000s (.784). This difference of .182 demonstrates that declining forces of attraction fully explain the large observed decline in men's TFMR. As with the results for women, changing propensities to form pairings involving potentially competing caregiving obligations do not account for the decline in men's TFMR. Rather, it is the forces of attraction declines for all pairings involving daughters with brothers (and small changes in forces of attraction for pairings involving younger sons and eldest daughters without brothers). For men, we thus find no support for Hypothesis 3 and limited support for Hypothesis 4 (the small, negative values for the contributions of pairings involving potentially competing caregiving obligations).
Because our analytic sample is unmarried men and women aged 18–49, it is important to consider whether the family characteristics of relatively older (ages 35–49) unmarried people systematically differ from those of their younger counterparts in ways relevant to our questions. For example, some relatively older unmarried men and women may already be caring for aging parents. In this case, the direct signal provided by caregiving would be stronger than the indirect signal associated with sibship position and would thus influence how these people are evaluated in the marriage market.
To evaluate the possible implications of such differences, we estimated models restricting the sample to ages 18–34 (Tables A1–A3, online appendix). This restriction does little to change the overall patterns of forces of attraction (Table 1), availability ratios (Table 2), and pairing-specific contributions to changes in TFMRs (Table 3). One difference is in the contribution of each pairing to the observed TFMR and the counterfactual TFMR with forces of attraction held constant for women. Results in panel B of Table 3 show that declining forces of attraction for all pairings involving younger sons account for nearly all the observed decline in TFMR. Results based on the age-restricted marriage market (Table A3, panel B) show that the contribution of pairings involving daughters with brothers is not negligible. Table A1 shows that forces of attraction are lower across the board, reflecting that Nijkl (the number of marriages in the numerator) is reduced when we omit unmarried people aged 35–49. Importantly, the relative magnitudes of forces of attraction for different pairings are very similar to those found for the full sample. Table A2 shows that the availability ratios are also almost identical to those in Table 2. Taken as a whole, these results based on an age-restricted marriage market indicate that systematic differences with respect to age in the signal provided by sibship position (and how that signal is evaluated by potential partners) do not differ by age in ways that alter our findings.
Summary and Discussion
In this study, we investigated the potential relevance of one understudied consequence of lower fertility: the contribution of the changing sibship composition of the marriage market to declining marriage rates. In Japan and other patrilineal societies, being the eldest son (or the eldest daughter with no male siblings) may be associated with a marriage market disadvantage to the extent that women (or men) view sibship position as a signal of the potential likelihood of coresiding with parents or parents-in-law and caring for them after marriage. Because fertility decline implies a relative increase in the prevalence of eldest sons or eldest daughters without brothers (including only-children), we conducted the first empirical evaluation of how this compositional change may link fertility decline with lower marriage rates.
Focusing on Japan, we examined how changes in marriage market composition and pairing propensities with respect to sibship position contributed to the observed decline in first-marriage rates between 1980 and 2010. We used harmonic mean models of marriage that allow for separating changes in marriage rates into changes in pairing propensities and changes in marriage market composition. These models produced three important findings.
First, the propensity to marry an individual with a stronger normative expectation of future family caregiving obligations (i.e., eldest sons or daughters without brothers) is lower than for individuals without such expectations. This finding is consistent with our Hypothesis 1a. Further, consistent with Hypothesis 1b, the lower propensity of marriage is especially pronounced for only-children (both sons and daughters). Support for Hypothesis 1a has weakened over time for women, with differences in the propensity to marry eldest sons and younger sons converging in recent years.
Second, we did not find strong empirical support for Hypothesis 2, which posited that the observed decline in first-marriage rates is partly explained by a marriage market mismatch generated by changing sibship composition. Our counterfactual estimates showed that the changing composition of the marriage market with respect to sibship position accounts for approximately 17% of the decline in women's marriage rates. In contrast to the results for women and counter to our expectations, changes in marriage market composition limited the marriage rate decline for men. One possible explanation for this pronounced gender difference is that men's sibship characteristics and associated normative expectations are more important spouse selection criteria (for women) than are women's (for men) (Jones 2007). Previous studies suggest that these gender-asymmetric patterns may be rooted in the persistent gender-based division of labor in Japan, where men spend less time doing domestic work (Kan et al. 2022) and gender essentialist norms remain more salient (Brinton and Lee 2016) than in other low-fertility countries. In this context, it is not surprising that women primarily shoulder the burden of elder care (Long and Harris 2000; Traphagan 2003). It is therefore reasonable to assume that a woman who marries an eldest son will, on average, provide more care for her parents-in-law than would a man who marries a woman with care obligations to her own parents. However, marriage market mismatches generated by declining fertility and associated changes in sibship structure are less important than the relative improvement in marriage market composition for men because of women's more pronounced delay in the transition to marriage. This pattern of changing marriage behavior is poorly understood, with implications yet to be explored.
Last, and most importantly, we found that changes in pairing propensities helped mitigate the potential implications of changes in marriage market composition with respect to sibship position. Specifically, forces of attraction increased slightly for most marriages involving men and women with higher normative expectations of caregiving (eldest sons and daughters without brothers, including only-children) but declined markedly for pairings involving younger sons and daughters with brothers. These results are inconsistent with Hypothesis 3 but consistent with Hypothesis 4 and suggest that men and women may have adjusted their preferences in response to changes in sibship composition by casting a wider net (Lichter et al. 1995). They might also reflect ongoing changes in family norms, including a waning of sibship-based expectations in favor of a more strategic negotiation of caregiving arrangements (Kureishi and Wakabayashi 2010; Takagi et al. 2007; Wakabayashi and Horioka 2009). Other interpretations are certainly possible. For example, rather than casting a wider net, men and women whose sibship positions signal a higher likelihood of future caregiving obligations may display an increased likelihood of marrying with a commitment to face the shared challenge together. It is also possible that the increase in pairings involving potentially competing family obligations is driven primarily by the preferences of eldest sons, who may assume that their future spouse would prioritize care for his parents over hers. However, identifying whether and how the attitudes and behavior of men and women differentially impact pairing propensities is not possible.
Taken as a whole, our findings provide little support for a low-fertility trap scenario in which low fertility may be self-reinforcing via its influence on marriage market composition. At least in Japan, the long, gradual process of fertility decline has not resulted in dramatic changes in marriage market composition. Changes that have occurred have been offset by changes in pairing propensities (forces of attraction) and by gender differences in the pace of marriage delay (which has increased the availability of potential mates for men).
Our analyses and the data on which they are based have several limitations. First, and most importantly, we do not have information on marriage timing and sibship pairing patterns for marriages that ended in divorce. We estimate that about 8% of men and 10% of women in the latest survey (conducted in 2015) are excluded because of a first-marriage dissolution (see Tables A4 and A5, online appendix). As a result, our data underrepresent the number of first marriages and those at risk of first marriage. To the extent that the likelihood of marital dissolution is systematically related to specific sibship pairings, our results will reflect this underrepresentation. However, we are unaware of any empirical evidence with which to evaluate the association between sibship pairing and marital stability. Second, the age range of the unmarried sample differs across surveys, as noted in the Data section. As a result, the marriage markets we constructed from these data underrepresent older men and women in the 1980s, with potential implications for our estimates.16 Third, our initial period (the 1980s) is not ideal, given that the most dramatic impact of change in sibship composition on marriage market composition occurred before then. Ideally, we would use earlier cohorts who experienced the significant changes in sibship composition described in Figure 1, but data limitations preclude this approach.
Despite these limitations, our focus on an understudied but potentially important dimension of spouse pairing patterns sheds new theoretical and empirical light on the influence of fertility decline on the marriage market outcomes. We focused on Japan, but our questions and approach are of potential relevance in other “strong family” countries in which family expectations and obligations are associated with specific sibship positions. If perceived obligations to provide care to older parents are associated with specific demographic characteristics, such as sex and sibship position, the increases in single-sex sibships, eldest children, and only-children that accompany low fertility may produce marriage market mismatches that reinforce low fertility in these countries, at least in the short run. Importantly, these influences need not be limited to populations with histories of patrilineal social organization; they may also be salient in low-fertility countries where negative views regarding only-children are widely shared (Blake 1981; Poston and Falbo 1990).
The potential relevance of changing sibship composition in declining fertility may be particularly salient in East Asian countries, where the connection between marriage and fertility remains strong (Raymo et al. 2015). Empirical studies on sibship position and family expectations or obligations are limited, but research on South Korea suggests that eldest sons are more likely to expect to live with their parents and that their wives remain a primary source of informal care (Do et al. 2015). South Korea is another particularly interesting setting in which to reevaluate our hypothesis, given its rapid decline in fertility rates in the 1970s and 1980s. Evidence from mainland China is different, with only-child status signaling greater economic resources via inheritance. The tendency of only-sons and only-daughters to marry each other in China (Wen 2018) shows that the role of sibship status in shaping marriage formation depends on social context, even within East Asia. Thus, although our study did not find strong evidence for marriage market mismatches in Japan (especially for men), we see value in pursuing these questions in a range of societies in which obligations or understandings associated with sibship may provide meaningful signals in the marriage market.
An early version of this paper was presented at the 2019 annual meeting of the Population Association of America. This research is supported by the Nakajima Foundation, the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award P2CHD047879, and the Japan Society for the Promotion of Science KAKENHI grant JP22K01851. Permission to use data from the Japanese National Fertility Surveys was obtained through the research project of the Japanese National Fertility Surveys, the Department of Population Dynamics Research, the National Institute of Population and Social Security Research on the basis of the Statistics Act, Article 32(2020/6/10).
Replication code is available on Github: https://github.com/fumiyau/SibshipAssortativeMating.
The prevalence of nonmarital childbearing in 2020 was 2.38% (National Institute of Population and Social Security Research (NIPSSR) 2022a).
For example, numerous websites and online chat boards are devoted to discussing the downsides of marrying eldest sons, and popular media (e.g., television dramas) prominently feature conflict between daughters and mothers-in-law.
Throughout this article, we use the term propensity to indicate the likelihood of marriage independent (or net) of marriage market composition.
According to the same surveys, recently married women’s expected (as opposed to ideal) number of children remained stable over the same period, at 2.08 in 1977 and 1.95 in 2021. By contrast, unmarried women’s desired number of children declined from 2.28 in 1982 to 1.79 in 2021.
For consistency with our analytic sample, Figure 1 is based on never-married and first-married individuals.
These figures are calculated using the assumption that the number of marriages for each possible pairing reflects only sex-specific sibship position distributions and the total number of marriages (i.e., the number of marriages involving men and women who are both only-children would be the proportion of men who are only-children × the proportion of women who are only-children × the total number of marriages).
The JNFS did not ask these questions after the 1992 survey. After 1992, the JNFS also did not ask married respondents who were not currently coresiding with parents about their attitudes toward living with parents or parents-in-law, and the JNFS does not include married male respondents.
More recent data for this specific question are not available.
However, even though couples increasingly delay the initiation of intergenerational coresidence, the likelihood of eventually living with or close to parents or parents-in-law has changed little over cohorts (Kato 2013).
This decline in support provided by the daughter-in-law has not been offset by an increase in support provided by the spouse or one’s children but rather by caregiving services (9% in 2001 and 12% in 2019) and family living separately (8% in 2001 and 14% in 2019). The increasing role of family living separately hints at the shift from coresidence to proximate residence.
Limiting our focus to first-married couples reduces the sample size by 3% to 10% in each survey round (from 3.7% in 1982 to 7.5% in 2015 for men, and from 3.1% in 1982 to 9.6% in 2015 for women). Exclusion of remarried and formerly married respondents (and husbands) is more salient in the recent survey rounds, reflecting the increasing prevalence of marriages that involve at least one remarried spouse. One potential consequence of restricting the sample to first-married couples is thus underestimation of marriage rates in recent periods.
A mechanical interpretation of forces of attraction is the marriage rate that would be approximately observed for a given pairing if there were no other unmarried women (men) in the marriage market with whom to compete for a specific group of men (women).
The observed or counterfactual TFMR for a given period is calculated as the weighted sum of pairing-specific marriage rates, which are the product of the pairing-specific forces of attraction in Table 1 and the pairing-specific availability ratios in Table 2. In these calculations, weights are the proportion of the population (by sex) at risk of first marriage in each of the sibship groups.
Because TFMRs are the sum of pairing-specific marriage rates, differences in observed and counterfactual TFMRs are equal to the sum of differences in pairing-specific marriage rates. These are the differences we refer to as pairing-specific contributions to the difference in TFMRs.
The contribution of changing marriage market composition calculated by holding forces of attraction constant (14% of the observed decline) does not equal that calculated by holding availability ratios constant (17% of the observed decline) because these counterfactual calculations do not represent a formal decomposition analysis. More specifically, they do not account for interaction between changes in forces of attraction and availability ratios.
To assess the potential implications of differences in unmarried respondents’ ages across surveys, we examined data from the national census in 1980, 1985, and 1990. The percentage of the 15- to 49-year-old never-married population aged 35 and older was 6% in 1980, 10% in 1985, and 12% in 1990. Given the relatively low rates of marriage beyond age 35, we suspect that the absence of unmarried men and women aged 35 and older in 1982 and 1987 would have a negligible impact on our results. Nevertheless, our findings should be evaluated with this data caveat in mind.