## Abstract

We investigated the conditions under which married children live with their older parents in Japan. We focused on how needs and resources in each generation are associated with whether married couples live with their parents in parent-headed and child-headed households, and we also investigated difference in power relations between older and younger generations and between children and their spouses. We analyzed a nationally representative sample of older parents (n  =  3,853) and their married children (n  =  8,601) from the 1999 Nihon University Japanese Longitudinal Study of Aging (NUJLSOA). Mutinomial regression revealed that married children with relatively affluent parents tended to live with them in parent-headed households and that married children with parents who are in relatively poor health or who are widowed tended to live with them in child-headed households. We also found that less-educated married children tended to live in the households of their higher-income parents, suggesting that parents may be “purchasing” traditional arrangements with less-affluent children. In addition, children with an educational advantage over their spouses were more likely to have parents living with them in child-headed coresident households. We conclude that traditional multigenerational coresidence has become a commodity negotiated within families based on relative resources and needs within and across generations.

## Introduction

Multigenerational stem-family households are often depicted as living arrangements in which older parents and adult children commit significant amounts of help and support to each other in order to ensure the well-being of one or both generations (Wilmoth 2000). In Japan, such households are also formed against the backdrop of strong cultural expectations of filial piety on the part of adult children: mainly, married sons as the primary successors in the family. However, the prevalence of multigenerational households has declined steeply in Japan over the past three decades, suggesting that traditional living arrangements have become increasingly optional as filial piety has lost some of its historical coercive force.

The standard Western explanation for declining rates of multigenerational households is that older adults will purchase privacy by choosing to live independently of their adult children when their resources permit. In general, studies demonstrate that higher income of older parents is associated with an increased likelihood of an independent living arrangement from children (e.g., Englehardt et al. 2005; Ruggles 2007). However, at the same time, Ruggles (1996) noted that only about 20%–30% of the increase in independent living among the aged in the United States over the first half of the twentieth century can be explained by rising income of the older population. Recent research by Ruggles (2007) suggests that resources available to both older and younger generations need to be taken into account when assessing the decline of multigenerational households. We would further suggest that resource differences between generations within families have consequences for the type of coresident household formed, and especially for which generation maintains headship status. Further, because multigenerational households in which elders reside often consist of married children in Asian societies, it is also important to consider resource differences between adult children and their spouses. Few studies on multigenerational households in which older individuals reside have systematically examined how within-generation resource variation in the marital unit structures influences residential decisions with respect to older parents. We extend this line of research by considering resource advantages—both between and within generations—as a factor that is instrumental in dictating whether married children will live together with elderly parents in their own or in their parents’ household.

Specifically, we suggest that in Japan, older people with greater resources relative to their children will purchase piety by leveraging their intergenerational advantage to induce their married sons to live in their households. Similarly, we expect that married sons with greater resources relative to their wives will leverage their spousal advantage to have their own parents live in the married couple’s household. Differentiating parent-headed and child-headed coresidence yields a meaningful contrast in the Japanese context because household headship carries a unique cultural status in Japan and reasonably reflects who in the household wields the most power over household affairs and decisions.

## Intergenerational Coresidence, Filial Piety, and Power Resources

Japan remains one of the only developed countries where a substantial proportion of older adults live with their adult children in multigenerational households. In 2009, 43% of people aged 65 and older lived with a child (Cabinet Office 2011); comparatively, figures in Western countries were less than 5% (Maeda and Ishikawa 2000). The relatively high rate of coresidence of older parents in Japan is generally understood in ideological terms as adherence to the Confucian ideal of filial piety, mandating that adult children respect and meet the needs of their aging parents. Intergenerational coresidence in East Asian societies reflects a distinctive family culture in which adult children are obligated to repay their parents’ earlier sacrifices by taking care of them when the parents become old (Koyano 1996; Takagi and Silverstein 2006).

As a fundamental norm in Japanese society, filial piety represents a culturally coercive force that requires adult children to acquiesce to the demands of their older parents. Support for the norm of filial piety from older parents influences family coresidence decisions. Studies in China and Japan have shown that older parents who embrace the ideal of intergenerational coresidence tend to live with their adult children—a finding that persists after adjusting for possible endogeneity in the relationship (Logan and Bian 1999; Takagi and Silverstein 2006). The demands imposed by filial piety are not gender-neutral. In strongly patrilineal societies in East Asia, older adults favor living with a married son over a married daughter (Logan et al. 1998; Takagi and Silverstein 2006).

However, filial piety is not an absolute in its strictures, and there is some malleability in the way that it can be fulfilled. The linkage between multigenerational households and filial norms has grown increasingly loose; coresident households are formed less in response to cultural dictates and more in response to the needs of elders (or adult children) for support (Martin and Tsuya 1994; Takagi et al. 2007). Scholars have emphasized negotiations among family members with regard to maintaining or modifying shared living arrangements in contemporary Japan (Izuhara 2000, 2004; Traphagan 2003), suggesting that filial piety has become more of guide than a rule, and arguably more of a commodity than an ideology in Japanese society.

Multigenerational households in which older adults reside are not only declining in prevalence but are also shifting away from lifelong coresidence in the parents’ home and toward residential reunification later in their lives (Takagi et al. 2007). There are distinctive differences between these two living arrangements, with older adults having poorer health and fewer economic and social resources—and, consequently, less power—in the latter type of household than in the former. These differences are not subtle; some empirical evidence suggests that the distribution of consumption corresponds to the distribution of resources in multigenerational households in Japan. Hayashi (1995), for instance, found that the balance of food expenditures within extended multigenerational households in Japan tended to favor older parents who made relatively larger contributions to the household income.

In Western social science literature, intergenerational coresidence has increasingly come to be viewed as the outcome of negotiations between pertinent family members based on mutual resources, needs, and preferences across and within generations (Lundberg and Pollack 1994; Pezzin and Schone 1997, 1999). Seen from this perspective, the decision to coreside by adult children and their parents is based on the relative distribution of needs and resources in the family. However, in Japan, the cultural ideal of filial piety—even though it has become more violable—still shapes preferences, making it plausible that resource advantages of the elderly will be used to purchase intergenerational piety rather than intergenerational privacy.

Coresidence of a married child with his or her parents is a decision made jointly between spouses, requiring compliance on the part of the child-in-law. Thus, relative conjugal power held by children and children-in-law may be an indicator of their likelihood of coresiding with their parents. Daughters-in-law may be reluctant to accept coresidence with their husband’s parents because such a living arrangement expects them to assume caregiving duties with respect to their parents-in-law as prescribed by cultural tradition (Harris and Long 1993; Hashizume 2000; Jenike 1997; Sodei 1995; Traphagan 2003).

## Other Factors Affecting Intergenerational Coresidence in Japan

In both East Asian and Western societies, coresidence of older people with their adult children is based on needs, interpersonal factors, and family structure, as well as resources and cultural preferences. Needs experienced by older adults that are associated with coresidence include poor health (Brown et al. 2002; Crimmins and Ingegneri 1990; Silverstein 1995), lack of economic and housing resources (Izuhara 2000; Logan et al. 1998; Mutchler and Burr 2003), and loss of a spouse (Logan et al. 1998; Martin and Tsuya 1994; Raymo and Kaneda 2003).

Coresidence can be initiated for the benefit of adult children as well, particularly those who need assistance with childcare (Morgan and Hirosima 1983). Housing also represents a resource that may be mobilized by older parents or by adult children to establish a coresident household (Logan and Bian 1999; Raymo and Kaneda 2003). Rural residency also enhances the likelihood of coresidence, presumably because of the relatively strong traditional local culture prevalent in such areas (Raymo and Kaneda 2003).

Structural factors regarding family size and distribution may also influence whether older parents coreside with their children. Parents with children living nearby may have less desire and incentive to live with their children because social support can be provided by that proximate child (Logan et al. 1998; Martin and Tsuya 1991). In addition, larger families may afford greater opportunities for elders to coreside with adult children but reduce the likelihood of any one child living with the parent.

## Hypotheses

In this research, we focus on how the amount and balance of resources in inter- and intragenerational relationships influence whether adult children coreside with their aging parents in Japan and whether they or their parents claim household headship. Relationships we consider include those between older parents and (1) their married children; (2) married children and their spouses; and (3) married children and their unmarried siblings.

We hypothesize that married children and older parents with relatively greater socioeconomic resources will tend to go against tradition and live independently of each other. This pattern is likely to be stronger among the sons than among the daughters of older parents. However, daughters with more socioeconomic resources may have greater power to defy traditional dictates and be more likely to coreside with their parents than those with fewer such resources.

Examining the resources of parents and children in combination, we expect older parents with relatively greater resources than their adult children to “purchase” filial piety by inducing their children to be members of their households. Specifically, we expect that parents with greater income will better exercise their culturally legitimated power to accommodate children, who have less education and possibly greater need for parental resources, in their own household.

Further, we hypothesize that the relative imbalance in resources between married children and their partners will make it more likely that the generations will coreside in the household of the more advantaged generation. We propose that such an advantage will increase the partner’s decision to acquiesce to the less-advantaged spouse (see Elder and Rudolph 2003) and result in coresidence with the parents of the higher-educated child. Because of gender expectations, this dynamic should be more strongly evident when the son is the advantaged spouse than when the daughter is the advantaged spouse.

Much of the literature finds that unmarried children are more likely to live with their parents than married children because they ostensibly have fewer competing family obligations (Logan and Bian 1999; Rindfuss and Raley 1998; Ward et al. 1992). Thus, we expect that unmarried children, when available, will crowd out their married siblings as residential partners for older parents. (This is not to imply that in Japan, older adults are overall more likely to live with unmarried children, who are relatively rare in the cohort we are studying.) Because it is more likely that unmarried children will benefit from their parents’ housing than the reverse, we hypothesize that the presence of unmarried children in the family will suppress the prevalence of married children living in parent-headed households.

## Method

### Sample

We analyzed data from the initial wave of the Nihon University Japanese Longitudinal Study of Aging (NUJLSOA) conducted in 1999. The NUJLSOA is a national longitudinal study that focuses on the health and social conditions of the Japanese population aged 65 years and older (USC/UCLA Center on Biodemography and Population Health 2004). The baseline sample consisted of 4,997 primary respondents, including 606 proxies, for a response rate of 74.6%. The survey asked respondents about sociodemographic backgrounds of up to 10 biological/adoptive children and children-in-law. Because we were interested in parents with married children, we restricted our analysis to the intergenerational relationships of 3,853 self-reporting respondents who had at least one married child. This resulted in an operational sample of 8,601 parent–married child dyads.

About 20% of the sample that met the selection criteria had missing data on relevant items. To retain all observations and avoid sample selection bias, we used multiple imputation, a technique that derives predicted values to replace missing data (Schafer and Olsen 1998). Predicted values are averaged across the five imputed data subsets. Robust standard errors incorporate sampling variability so that corrected confidence intervals are obtained for estimated coefficients. In comparing the distributions of variables using listwise deletion of missing values (2,656 parents and 5,451 children) with distributions using imputed data (3,853 parents and 8,601 children), we found few differences (see Tables 1 and 2). Analyses based on imputed and listwise deleted samples produced similar results, improving our confidence in using the imputed data set.

### Dependent Variable

The outcome variable was represented by three residential outcomes that combined living arrangements of children with respect to their parents and household headship: (1) coresides with parent in household headed by parent; (2) coresides with parent in household headed by married child; (3) does not coreside with parent. The last group served as the reference group in subsequent analyses. The person designated as the household head was identified through older respondents’ answers to the question, “Who is the head of your household?” Household headship indicates who in the home assumes primary responsibility for running household affairs (MHLW 2009). This position is not trivial in Japan because household heads are the principal figures in the preparation of tax reports, registration of residences, and determination of eligibility for social benefits. As such, headship entails both symbolic and economic connotations with regard to who primarily represents the living unit shared by household members (Winther 2008).

The sociodemographic characteristics of adult children captured in this study included age (in years), gender (1 = male, 0 = female), and rural residency (1 = lives in a rural area, 0 = lives in an urban area). Educational achievement was assessed by using a four-point scale (1 = junior high school, 2 = high school, 3 = vocational school or junior college, and 4 = four-year university and/or graduate school). Occupational status was measured with three dichotomous variables: professional, blue-collar, and clerical (reference = unemployed or not working). Because of the relatively high incidence of self-employed individuals among people coresiding with parents or children in Japan (see Raymo et al. 2004), a dichotomous variable was constructed to indicate whether adult children were self-employed in their main occupation (1 = yes, 0 = otherwise). Age of grandchildren was measured by using three dichotomous indicators based on an assessment of the educational status of the offspring of each adult child: preschool age, junior high school age, or high school age or older (reference = no children).

### Parent’s Characteristics

Sociodemographic variables included age (in years), gender (1 = male, 0 = female), and marital status (1 = widowed; 0 = other status). Socioeconomic status of older parents included educational achievement (measured on a four-point scale with 1 = junior high school, 2 = high school, 3 = vocational school or junior college, and 4 = four-year university and/or graduate school), household income (13-point scale ranging from 1 = <500,000 Japanese yen [approximately US$4,160] to 13 = >15 million Japanese yen [approximately US$125,000), homeownership (1 = lives in an owned home, 0 = lives in a rented home), and the respondent’s self-employment status (1 = mostly self-employed in the past, 0 = otherwise). Health was measured with self-reported ability to perform seven instrumental daily activities (see Table 1) and represented as a dichotomous variable (1 = has difficulty performing at least one activity, 0 = is able to do all activities without difficulty).

Filial attitudes were measured by agreement with following statement: “A child should be expected to support and take care of his or her aged parents, as the child should feel a sense of gratitude to the parents for raising him/her.” Responses ranged from disagree (1) to agree (4).

Two dichotomous variables were constructed to measure characteristics of children that may influence coresidence: number of children, presence of a proximate child (1 = at least one married/unmarried child lives in the same town/city, 0 = otherwise), and having an unmarried child (1 = at least one unmarried child, 0 = otherwise).

### Intergenerational and Spousal Resource Differences

Because the data available on the social status of children are limited, we measure relative resource differences between older parents and adult children using non-equivalent measures of income for parents and education for children. An interaction term between these two measures is used to represent how education of children changes coresidence likelihood at different levels of parental income.

To measure relative achievement between adult children and their spouses, we created a dichotomous variable indicating whether each married child held a higher level of education than his or her spouse (1 = yes, 0 = otherwise).

### Statistical Model

We used multinomial logistic regression to estimate the log odds that married children were coresident with their parents in either a parent-headed or a child-headed household relative to being noncoresident. Because our interest is in coresidence likelihood of certain married children, we “reshaped” the data such that the units of analysis were the married children of each responding parent, with the parent’s information (either father’s or mother’s) repeated for each married child in the family. To adjust for the nonindependence of sampling units due to family clustering, we used robust standard errors.

The first model (main-effects model) examined the effects of all predictors. Because coresidence in Japan implies practice of the traditional patrilineal culture that favors sons over daughters as a coresident child, we investigated the differences between married sons and daughters in the way that educational status influences the likelihood of living with parents. Specifically, we added interaction terms between married children’s gender and (1) their own educational level and (2) the variable signifying that they had greater education than their spouse. We expect that coresident outcomes will be more sensitive to the education and the inter-spousal advantage of married sons than of married daughters.

To test our hypothesis that older parents who are advantaged will purchase piety by accommodating relatively disadvantaged children in their households, we include an interaction term between older parents’ income and married child’s educational level. The use of interaction terms is an efficient strategy to examine the effects of differences in two variables that are measured on different metrics.

## Results

Characteristics of older parents are described in Table 1. The average age was 75 years; 43% were male, and 39% were widowed. The large majority (90%) owned a home, and their average household income was between 1.5 million Japanese yen (about US$12,500) and 3 million Japanese yen (about US$25,000). Slightly less than one-third of the parents (29%) had up to a high school education, and slightly less than one-fifth (19%) reported having been predominantly self-employed in the past. In terms of health, slightly more than one-tenth (12%) reported having some difficulty in performing at least one of the instrumental daily activities. A little less than one-half of the older respondents (46%) had at least one child living nearby, and about one-third (30%) had at least one unmarried child. Regarding normative attitudes, almost two-thirds of these older parents (65%) agreed that children should have filial responsibility.

Table 2 presents the characteristics of married children included in the sample. The average age was 47 years, and sons and daughters were almost equally represented. Almost three-quarters (73%) had a child of junior high school age or older, with smaller proportions having a child of preschool age (13%) or elementary school age (23%). Married children were almost equally distributed across blue-collar, clerical, and professional occupations, representing one-quarter to one-third in each category, and about one-sixth (15%) were self-employed. In terms of education, one-half of them had at least a high school education (52%), and about one-quarter received a college or a higher education (24%). Overall, slightly less than one-fifth (19%) had a higher educational status than their spouse. Our preliminary analysis also showed that it was more common for sons to have a higher educational status than their spouses (28.5% of sons) than daughters (10.4%). About one of four children (24%) lived in rural areas.

Table 3 presents the cross-classification of living arrangements by gender of adult children. Overall, 7% of married children coresided with their parents in a parent-headed household, and 11% of the adult children coresided with their parents in a child-headed household. (Although these percentages may seem low, consider that they were calculated based on the number of children represented in the data and not on the number of parents.) Sons were more likely than daughters to coreside with parents in both household types, reflecting gender norms in Japan (Koyano 1996; Takagi and Silverstein 2006).

Table 4 presents the results of the multinomial logistic analysis in which we show logistic coefficients and odds ratios (OR) associated with living in each of the two types of coresident households, relative to living independently from parents. The main effect equations for each outcome reveal gender differences consistent with the pattern observed in our earlier cross-classification analysis. Sons were more than six times more likely than daughters to live with parents in a child-headed household (OR  =  9.3) or in a parent-headed household (OR  =  6.4). In addition, both types of coresident patterns were more common among adult children who were older, had a child of elementary school age, were self-employed, and lived in rural areas, compared with their counterparts. In terms of socioeconomic factors, adult children in professional occupations were about one-third less likely to coreside with their parents in either of the two living arrangements considered. However, greater education lowered the likelihood of coresidence only of the type in which the child is the household head (almost 30% lower relative likelihood with every increase on the educational scale). Having more education than one’s spouse increased the relative likelihood of living in a child-headed household by 50%.

Among characteristics of older parents, widowhood and homeownership produced the largest effects. Homeowners were more likely to be living in child-headed (OR  =  3.6) and parent-headed (OR  =  5.2) coresident households. We note that legal ownership of the self-owned house is not clearly defined in the survey, which suggests both possibilities: that is, some children may be heads of households that their parents own, or they may accommodate their parents in their self-owned homes. (It is plausible to assume that legal homeownership mostly coincides with household headship, but we did not have empirical data to test this supposition.) Widows were more than 3.5 times more likely than the married parents (OR  =  3.6) to live in child-headed coresident households and less than half as likely as the married parents (OR  =  0.46) to live in parent-headed coresident households.

Although married children were less likely to live with older fathers than older mothers in child-headed households (OR  =  0.79), they were more than 50% more likely to live with older fathers in parent-headed households (OR  =  1.55). These children were also more likely to live in the households of parents who were younger (OR  =  0.96) and who more strongly endorsed the value of filial responsibility of adult children (OR  =  1.12). With regard to parents’ socioeconomic variables, children whose parents had relatively low income tended to live with them in child-headed households (OR  =  0.94), while those whose parents had relatively high income tended to live with them in parent-headed households (OR  =  1.08). Children tended to live in parent-headed households if their parents were previously self-employed (OR  =  1.35). Parents’ education did not predict whether children lived in either type of coresidential arrangement.

The availability of siblings exerted a consistent influence on coresidence outcomes of married children. Having more siblings, having an unmarried sibling, and having a married or unmarried sibling living in the same city as the parent reduced the married child’s likelihood of living with parents in both types of households (OR ranging from 0.49 to 0.62).

Focusing on the interaction terms added to the second equation, we found significant interactions between children’s gender and their educational status with respect to living with parents in each type of household arrangement. In Figs. 1 and 2, we plot these interactions as probabilities predicted from estimated coefficients along with other variables held constant at their means. Figure 1 shows that increasing education level of sons reduced their likelihood of living with parents in a child-headed household. The education status of daughters had minimal impact on this type of coresidence, with low percentages across all educational levels. Similarly, Fig. 2 shows that advancing education of sons decreased the likelihood of their living in a parent-headed household. However, an upward trend among the more highly educated daughters reveals a positive interaction term, suggesting a modest disordinal pattern whereby daughters with greater education are increasingly likely to live in parent-headed households. We note that our preliminary analysis conducted separately for sons and daughters revealed that daughters with no male siblings were more likely than those with brothers to coreside with their older parents (results available upon request). This is consistent with the strong son preference in Japan and suggests that coresidence of older adults with their married daughters is strongly conditioned on the absence of a son in the family.

We found no interaction between the adult child’s gender and the presence of an education advantage over the spouse, even when we ran the same model separately for sons and daughters (results available upon request). Having a higher educational status than the spouse predicted a higher likelihood of living in a child-headed coresidence household. (Although the effect size of this variable was consistent by gender, it was not significant among daughters because of the smaller number of daughters having an educational advantage over their husbands.) This potentially suggests that conjugal power in this arena is symmetric with respect to gender (see Table 4).

Finally, the interaction term between income of parents and education of children was close to statistical significance (p  <  .06) with respect to residing with parents in a parent-headed household. Examining predicted probabilities depicted in Fig. 3, we note that less-educated children are more likely than highly educated children to live with parents belonging to the highest income category. These higher-income parents are presumably able to purchase piety from their less-educated children. If the children are highly educated, parental income is not consequential.

Interaction terms in nonlinear models such as ours require careful examination because the magnitude and direction of the interaction effect, as well as its statistical significance, may be misleading for the whole sample without using the cross-derivative (that incorporates the main effects) to calculate estimated values (see Ai and Norton 2003). We used the command inteff in STATA 10.1 to examine the magnitude and significance of interaction effects across individuals in the sample. These “corrected” interaction effects were always in the same direction (i.e., negative) as the interaction logits in our models, and their statistical significance tended to be mostly robust.

To examine the potential drawbacks and particulars of differentiating coresidence types, we also ran a logistic regression analysis in which we combined the two types of coresidence as a single dichotomous outcome (results available upon request). All common predictors for parent-headed and child-headed coresidence were detected in the logistic model. Some factors that were significant in predicting only one type of coresidence (such as having an older parent with traditional norms) were significant in the logistic regression, whereas other variables (such as the older parent’s income) were not. We contend that, on balance, differentiating coresidence into child-headed and parent-headed provides important information that allows us to identify different power dynamics in each type of household.

## Discussion

This study investigated social, economic, and family factors associated with coresidence of married children with their older parents in Japan. Our findings exclusively reflect the experiences of married adult children. Coresidence with unmarried children entails a very different situation because conventionally many adult children in Japan coreside with their parents only until marriage (Raymo 2003). Coresidence of adult children who live with older parents after marriage typically involves the prospect of long-term commitment to the needs and demands of either or both generations and the added interpersonal complexity of a child-in-law in the household.

Our analysis focused on the distribution of resources within and between generations to represent the power dynamics that differentiate coresident parent-headed and child-headed households from living independently. Japan represents an intriguing context within which to examine how a highly modernized society still incorporates strong cultural norms of intergenerational responsibility, as guided by traditional dictates of filial piety. In general, our findings suggest that intergenerational coresidence of the aged is not simply the manifestation of filial piety, but is often a conditional and negotiated arrangement based on the resources and needs of each generation.

In particular, several needs of the adult children and their parents elevated their propensity to live together in a common household. Adult children who were not working (compared with professionals) and who had school-aged children were more likely than their counterparts to live with parents in either type of household, suggesting that coresidence served the needs of offspring as much as it did the needs of their elderly parents. Elevated needs among older parents (widowhood, low income) increased the likelihood of child-headed coresidence but not parent-headed coresidence, underscoring the utility of differentiating household type when interpreting the meaning of coresidence in terms of generational benefits.

Self-employment of both older parents and married children also plays a role in determining whether they will reside with each other. Presumably, coresidence enhances the effectiveness or compensates for instability in the family business (Raymo et al. 2004). However, past self-employment status of older parents was associated with coresidence only in a parent-headed household, suggesting that this may be a long-term residential arrangement. Variability in the kinds of occupations and socioeconomic statuses among the self-employed will require more detailed analyses to better clarify the motivations for coresidence in this group of families.

Consistent with previous findings (e.g., Raymo and Kaneda 2003), we found rural residency to be an important predictor of both types of coresidence. Urban and rural areas in Japan are strikingly different in terms of housing availability, social norms, and social service availability (e.g., Jenike 2003; Knight 2003). This result suggests that regional-level socioeconomic context plays a significant role in structuring living arrangements between generations in the family (Takagi et al. 2007).

Our findings also suggest that the son preference is still strong in Japan as sons are overwhelmingly more likely than daughters to coreside with their older parents. However, there are variations in the propensity of sons to coreside. As heads of households, highly educated sons had the lowest propensity to coreside with parents. Although it is possible that more advantaged sons end up providing monetary assistance to their parents in lieu of housing, their residential independence from parents can be interpreted as a clear shift in the manifestations of filial obligations in Japan. Our finding that daughters with greater education were more likely to live in the more traditional parent-headed households is striking and suggests, paradoxically perhaps, that advantaged daughters have greater latitude to practice filial duties that were once the sole right of the sons.

The findings show that parents’ normative expectations of their children are significantly associated only with the parent-headed coresidence and not with child-headed coresidence. Married children are more likely to abide by the cultural family norm if their parents maintain a relatively strong traditional view toward filial responsibility. However, the diminished power of traditional norms in Japanese society and increased role of negotiations based on needs and resources across generations highlight the contingent nature of intergenerational processes over the acquiescence of children to rigid norms on which older parents could formerly rely.

Our analysis suggests that child-headed coresidence is associated with resource dynamics that extend beyond the parent-child relationships to include similar dynamics between adult children and their spouses. Adult children with greater education than their spouses are more likely to accommodate their parents in a shared household arrangement than those who do not hold such an educational advantage. These findings imply that neo-traditional coresidence in Japan is partly contingent on the distribution of education within the conjugal unit of adult children. Children-in-law—who are generally reluctant to coreside with their spouse’s parents—have less power to resist the preferences of more highly educated spouses. However, there is no evidence that spousal privilege based on educational advantage is any less influential in the unusual situation in which married daughters invite their parents into their households.

Because multigenerational households are still largely formed by sons, our findings are largely driven by the gender composition of families. However, traditional family culture has become more egalitarian as gender norms appear to be shifting in Japanese society. As for socioeconomic power, a growing number of women are pursuing a higher level of educational training in Japan (Raymo and Kaneda 2003), which may grant daughters more power in negotiating with their husbands to create a household shared with their own parents. Such an arrangement may also be favored by aging parents themselves because they increasingly prefer to have a daughter rather than a daughter-in-law as a primary caregiver (Cabinet Office 2003). Furthermore, the steadily decreasing fertility rates in Japan imply that there may be increasing cases in which daughters have to play a critical role in assuring the well-being of their own parents in the absence of sons in the family.

Alternatives to coresidence also constitute a part of the negotiation for married adult children in the family. Our findings show that married children’s coresidence with their parents can be either delayed or substituted by siblings who are either still single or living separately but close to their parents. Because unmarried children are likely to have few family responsibilities of their own, they may more readily assist their aging parents than married children. And because an increasing number of adult children in Japan either delay marriage or never get married (Raymo 2003), single adult children may play a growing role in intergenerational household arrangements of the elderly. Another alternative to coresidence for married children appears to be the presence of a child residing close to the parent, which is an increasingly accepted alternative to coresidence in Japanese society (Martin and Tsuya 1994). Underlying these various nontraditional intergenerational arrangements—coresidence with daughters, unmarried children as coresident partners, and proximity as an alternative to coresidence—is the understanding that filial piety itself has diminished in Japan and has come to be practiced in a variety of ways.

Second, in examining the likelihood of coresidence of adult children, our study controlled for family-wide factors but did not formally take into account how coresidence and other forms of support to older parents were divided among siblings. Family size is relatively small in Japan, but there are likely cross-sibling trade-offs with respect to time, money, and space considerations involving older parents in multiple-child families.

Third, the data set for this study lacked detailed information on educational institutions, which imposed limitations in more precisely delineating educational differences between adult children and their spouses. Because social position in Japan tends to be directly linked to the prestige of particular schools, there is likely to be variability in socioeconomic status even within the group of university graduates. Nevertheless, we believe that our four-level scale of education captures essential, if overly broad, differences in the social status of adult children and their spouses.

Lastly, because sociodemographic and occupational data on adult children in this study were provided by their older parents, the nonresponse rate concerning children is relatively large. Selection bias would be problematic if the propensity to not report on children were correlated with the dependent variable (Aquilino 1999). The assumption that data are missing at random (conditioned on observed variables) is buttressed by preliminary analyses that found no significant difference in the results between original and imputed data sets. Further, nonresponse rates among noncoresident parents with regard to their children’s characteristics are similar to those of coresident parents, providing evidence that selection bias is minimal. Nevertheless, our results need to be cautiously interpreted in light of these issues.

In conclusion, this study demonstrates that filial piety may be better viewed as a code of conduct that has transmogrified rather than as one in demise. Although it is true that coresidence has declined dramatically over the past few decades, its manifestation has also changed from a socially regulated living arrangement based on strict lines of responsibility to a more flexible arrangement that serves the needs, and is enabled by the resources, of each generation. Intergenerational coresidence arrangements in Japan may be governed by an evolving normative paradigm different from the one to which the Japanese society has traditionally adhered. Demographic changes in the larger Japanese society—lower fertility and marriage rates—will likely cause change, rather than crisis, in how traditional expectations come to be redefined with regard to older adults and their adult children. Already, the predictable code governing intergenerational coresidence has given way to a new, more flexible framework that acknowledges the role played by transfers in reducing resource inequality between generations. This framework also accepts the possibility that solutions are negotiated between and within generations, with each generation leveraging its power resources to achieve desired ends, and also that alternative arrangements can substitute coresidence with married sons who are expected fulfill piety requirements, arguably generating greater uncertainty about whether and how (i.e., parent- or child-headed) coresidence (and its alternatives) will manifest itself.

## Acknowledgments

This research was partially supported by Grants AG021656 and AG021609 from the National Institute on Aging. We would like to thank the Nihon University Center for Information Networking for use of the Nihon University Japanese Longitudinal Study of Aging data. We especially would like to thank Dr. Yasuhiko Saito at Nihon University for his assistance in accessing the data. We also appreciate the thoughtful comments and suggestions provided from the anonymous reviewers of Demography.

## References

Ai, C., & Norton, E. C. (
2003
).
Interaction terms in logit and probit models
.
Economic Letters
,
80
,
123
129
. 10.1016/S0165-1765(03)00032-6
Aquilino, W. S. (
1999
).
Two views of one relationship: Comparing parents’ and young adult children’s reports of the quality of intergenerational relations
.
Journal of Marriage and the Family
,
61
,
858
870
. 10.2307/354008
Brown, J. W., Liang, J., Krause, N., Akiyama, H., Sugisawa, H., & Fukaya, T. (
2002
).
Transitions in living arrangements among elders in Japan: Does health make a difference?
.
Journal of Gerontology: Social Sciences
,
57B
,
s209
s220
. 10.1093/geronb/57.4.S209
Cabinet Office. (2003). Korei-sha Kaigo ni Kansuru Seron Chosa (Population Survey of Caregiving for Older Individuals) (in Japanese). Retrieved from http://www8.cao.go.jp/survey/h15/h15-kourei/index.html
Cabinet Office. (2011). Korei-Shakai Hakusho (Annual Report on the Aging Society 2011) (in Japanese). Tokyo, Japan: Author. Retrieved from http://www8.cao.go.jp/kourei/whitepaper/w-2011/zenbun/23pdf_index.html
Crimmins, E. M., & Ingegneri, D. G. (
1990
).
Interaction and living arrangements of older parents and their children: Past trends, present determinants, future implications
.
Research on Aging
,
12
,
3
35
. 10.1177/0164027590121001
Elder, H. W., & Rudolph, P. M. (
2003
).
Who makes the financial decisions in the households of older Americans?
.
Financial Services Review
,
12
,
293
308
.
Englehardt, G. V., Gruber, J., & Perry, C. D. (
2005
).
Social Security and elderly living arrangements: Evidence from the Social Security notch
.
Journal of Human Resources
,
40
,
354
372
.
Harris, P. B., & Long, S. O. (
1993
).
Daughter-in-law’s burden: An exploratory study of caregiving in Japan
.
Journal of Cross-Cultural Gerontology
,
8
,
97
118
. 10.1007/BF00971563
Hashizume, Y. (
2000
).
Gender issues and Japanese family-centered caregiving for frail elderly parents or parents-in-law in modern Japan: From the sociocultural and historical perspectives
.
Public Health Nursing
,
17
,
25
31
. 10.1046/j.1525-1446.2000.00025.x
Hayashi, F. (
1995
).
Is the Japanese extended family altruistically linked? A test based on Engel curves
.
Journal of Political Economy
,
103
,
661
674
. 10.1086/261998
Izuhara, M. (
2000
).
Changing family tradition: Housing choices and constraints for older people in Japan
.
Housing Studies
,
15
,
89
110
. 10.1080/02673030082496
Izuhara, M. (
2004
).
Negotiating family support? The “generational contract” between long-term care and inheritance
.
Journal of Social Policy
,
33
,
649
665
. 10.1017/S0047279404008049
Jenike, B. R. (
1997
).
Gender and duty in Japan’s aged society: The experience of family caregivers
. In J. Sokolovsky (Ed.),
The cultural context of aging: Worldwide perspectives
(pp.
218
238
).
Bergin & Garvey
:
Westport, CT
.
Jenike, B. R. (
2003
).
Parent care and shifting family obligations in urban Japan
. In J. W. Traphagan, & J. Knight (Eds.),
Demographic change and the family in Japan’s aging society
(pp.
177
201
).
New York
:
State University of New York Press
.
Knight, J. (
2003
).
Repopulation the village?
. In J. W. Traphagan, & J. Knight (Eds.),
Demographic change and the family in Japan’s aging society
(pp.
107
124
).
New York
:
State University of New York Press
.
Koyano, W. (
1996
).
Filial piety and intergenerational solidarity in Japan
.
Australian Journal on Ageing
,
15
,
51
56
. 10.1111/j.1741-6612.1996.tb00203.x
Logan, J. R., & Bian, F. (
1999
).
Family values and coresidence with married children in urban China
.
Social Forces
,
77
,
1253
1282
.
Logan, J. R., Bian, F., & Bian, Y. (
1998
).
Tradition and change in the urban Chinese family: The case of living arrangements
.
Social Forces
,
76
,
851
882
.
Lundberg, S., & Pollack, R. A. (
1994
).
Noncooperative bargaining models of marriage
.
The American Economic Review
,
84
,
132
137
.
Maeda, D., & Ishikawa, H. (
2000
).
Ageing in Japan: Retirement, daily lives, pensions and Social Security
. In D. R. Phillips (Ed.),
Ageing in the Asia-Pacific Region: Issues, policies, and future trends 2000
(pp.
113
132
).
New York
:
Routledge
.
Martin, L. G., & Tsuya, N. O. (
1991
).
Interactions of middle-aged Japanese with their parents
.
Population Studies
,
45
,
299
311
. 10.1080/0032472031000145456
Martin, L. G., & Tsuya, N. O. (
1994
).
Middle-aged Japanese and their parents: Coresidence and contact
. In L. Cho, & M. Yada (Eds.),
Tradition and change in the Asian family
(pp.
153
178
).
Honolulu
:
University of Hawaii Press
.
Ministry of Health, Labour, and Welfare (MHLW). (2009). Graphical review of Japanese household. Tokyo, Japan: Author. Retrieved from http://www.mhlw.go.jp/english/database/db-hss/grjh-index.html
Morgan, S. P., & Hirosima, K. (
1983
).
The persistence of extended family residence in Japan: Anachronism or alternative strategy?
.
American Sociological Review
,
48
,
269
281
. 10.2307/2095111
Mutchler, J. E., & Burr, J. A. (
2003
).
Living arrangements among older persons: A multilevel analysis of housing market effects
.
Research on Aging
,
25
,
531
558
. 10.1177/0164027503256692
Pezzin, L. E., & Schone, B. S. (
1997
).
The allocation of resources in intergenerational households: Adult children and their elderly parents
.
The American Economic Review
,
87
,
460
464
.
Pezzin, L. E., & Schone, B. S. (
1999
).
Intergenerational household formation, female labor supply and informal caregiving: A bargaining approach
.
Journal of Human Resources
,
34
,
475
503
. 10.2307/146377
Raymo, J. M. (
2003
).
Premarital living arrangements and the transition to first marriage in Japan
.
Journal of Marriage and Family
,
65
,
302
315
. 10.1111/j.1741-3737.2003.00302.x
Raymo, J. M., & Kaneda, T. (
2003
).
Changes in the living arrangements of Japanese elderly: The role of demographic factors
. In J. W. Traphagan, & J. Knight (Eds.),
Demographic change and the family in Japan’s aging society
(pp.
27
52
).
New York
:
State University of New York Press
.
Raymo, J. M., Liang, J., Sugisawa, H., Kobayashi, E., & Sugihara, Y. (
2004
).
Work at older ages in Japan: Variation by gender and employment status
.
Journal of Gerontology: Social Sciences
,
59B
,
s154
s163
. 10.1093/geronb/59.3.S154
Rindfuss, R. R., & Raley, R. K. (
1998
).
Intergenerational contact in the United States and Japan
. In K. O. Mason, N. O. Tsuya, & M. K. Choe (Eds.),
The changing family in comparative perspective: Asia and the United States
(pp.
193
213
).
Honolulu
:
University of Hawaii Press
.
Ruggles, S. (
1996
).
Living arrangements of the elderly in America, 1880–1980
. In T. K. Hareven (Ed.),
Aging and generational relations over the life course: A historical and cross-cultural perspective
(pp.
254
271
).
Berlin, Germany
:
de Gruyter
.
Ruggles, S. (
2007
).
The decline of intergenerational coresidence in the United States, 1850 to 2000
.
American Sociological Review
,
72
,
964
989
. 10.1177/000312240707200606
Schafer, J. L., & Olsen, M. K. (
1998
).
Multiple imputation for multivariate missing-data problems: A data analyst’s perspective
.
Multivariate Behavioral Research
,
33
,
545
571
. 10.1207/s15327906mbr3304_5
Silverstein, M. (
1995
).
Stability and change in temporal distance between the elderly and their children
.
Demography
,
32
,
29
45
. 10.2307/2061895
Sodei, T. (
1995
).
Care of the elderly: A women’s issue
. In K. Fujimura-Fanselow, & A. Kameda (Eds.),
Japanese women: New feminist perspectives on the past, present, and future
(pp.
213
228
).
New York
:
The Feminist Press
.
Takagi, E., & Silverstein, M. (
2006
).
Intergenerational coresidence of the Japanese elderly: Are cultural norms proactive or reactive?
.
Research on Aging
,
28
,
473
492
. 10.1177/0164027506287788
Takagi, E., Silverstein, M., & Crimmins, E. M. (
2007
).
Intergenerational coresidence of older adults in Japan: Conditions for cultural plasticity
.
Journal of Gerontology: Social Sciences
,
62
,
s330
s339
. 10.1093/geronb/62.5.S330
Traphagan, J. W. (
2003
).
Contesting coresidence: Women, in-laws, and health care in rural Japan
. In J. W. Traphagan, & J. Knight (Eds.),
Demographic change and the family in Japan’s aging society 2003
(pp.
203
226
).
New York
:
State University of New York Press
.
USC/UCLA Center on Biodemography and Population Health. (2004). Nihon University Japanese Longitudinal Study of Aging [Data code book and questionnaire]. Retrieved from http://www.usc.edu/dept/gero/CBPH/nujlsoa
Ward, R., Logan, J. R., & Spitze, G. (
1992
).
The influence of parent and child needs on coresidence in middle and later life
.
Journal of Marriage and the Family
,
54
,
209
221
. 10.2307/353288
Wilmoth, J. M. (
2000
).
Unbalanced social exchanges and living arrangement transitions among older adults
.
The Gerontologist
,
40
,
64
74
. 10.1093/geront/40.1.64
Winther, J. (
2008
).
Household enumeration in national discourse: Three moments in modern Japanese history
.
Social Science History
,
32
,
19
46
. 10.1215/01455532-2007-012