Many single adult children in countries around the world live with their parents. Such coresidence has been thought to delay the transition to first marriage, although the exact reasons for the delay have not been sufficiently examined. Using panel data from Japan, we investigate whether changes in never-married adults’ residential status lead to alterations in their marital aspirations, courtship behaviors, romantic opportunities, and perceived obstacles to marrying. Our estimation of fixed-effects models helps address potential bias caused by single individuals’ selection into living in the parental home. The analysis indicates that living with parents is associated with a lower probability of forming romantic relationships, thereby decelerating the transition to first marriage. The never-married, however, do not desire marriage less, put less effort into finding romantic partners, or have fewer opportunities to meet potential partners when coresiding with parents. Overall, the findings suggest that living in the parental home increases never-married men’s contentment with their immediate social environment, whereas it decreases women’s psychological readiness to transition into adult roles, making both men and women less eager to settle into a romantic relationship.
Demographers and family researchers have long been interested in young adults’ living arrangements, with a special focus on their decision to leave or remain at their parents’ home (Billari and Liefbroer 2007; Buck and Scott 1993; Cordón 1997; Goldscheider and DaVanzo 1989; Goldscheider and Goldscheider 1999; Hughes 2003; Ravanera et al. 1998; Zorlu and Mulder 2011). Not only is leaving the parental home considered a major marker in the transition to adulthood (Billari and Liefbroer 2010; Furstenberg 2010), but also young people’s coresidence with parents is thought to affect their sense of independence, family views, relationships with parents, and psychological well-being (Aquilino 1997; Goldscheider and Waite 1991; Kins and Beyers 2010; Leopold 2012; Umberson 1992). In the United States, despite variations in the timing of home-leaving by race/ethnicity, class, gender, and historical time (Qian 2012; White 1994), relatively few individuals continue to live with parents past their mid-20s, and most leave home for reasons other than to form a union (Furstenberg 2013; Mulder et al. 2002). Similar trends are also observed in Western Europe and Scandinavian countries, with the latter having especially high rates of home-leaving before entering a union (Billari and Liefbroer 2010; Iacovou 2010).
In other parts of the world, such as East Asia and southern Europe, leaving the parental home before marriage is far less common. In Japan, for example, approximately 70 % or more of never-married men and women in their late 20s or 30s still live with parents, and the percentages have barely changed over the past several decades (Fukuda 2009; Raymo 2003b). Young people’s coresidence with parents is so prevalent in certain regions partly because their cultural norms consider marriage or union formation to be a natural point for children to leave home, if the children ever leave (Cordón 1997; Zeng et al. 1994). Until then, remaining at the parental home is socially acceptable, and parents may even prefer it (Billari and Tabellini 2011; Iacovou 2010; Raymo and Ono 2007). Because the countries in which young people’s coresidence with parents is especially widespread also tend to have greater delays in transitions to first marriage, the link between prolonged coresidence with parents and marriage postponement, which further results in reduced fertility, has received increasing attention (Billari and Rosina 2004; Billari and Tabellini 2011; Dalla Zuanna 2001; Raymo and Ono 2007; Yamada 1999). In Japan, in particular, the trend in delaying first marriage is more pronounced among men and women who live with parents than those who do not (Raymo 2003b), suggesting that extended coresidence with parents is conducive to later marriage.
In this study, we use panel data from Japan to further elucidate how premarital living arrangements may shape young adults’ transitions to first marriage. Despite some empirical evidence linking coresidence with parents and postponement of first marriage (Raymo 2003b; Raymo and Ono 2007; Sakamoto and Kitamura 2007), existing research provides little insight into how exactly coresidence may exert such an influence. Does living with parents make one desire marriage less? Does living with parents make it more difficult to meet potential marriage partners, because one spends less time socializing outside the home or because one is less likely to put effort into meeting potential partners? Alternatively, does living with parents make one less likely to form romantic relationships, either because a lack of residential independence makes one less attractive or because one becomes choosier given a lesser need for social companions while living with one’s natal family? This study answers these questions with unique data from Japan that provide detailed information on never-married people’s household compositions, marital aspirations, courtship behaviors, and self-identified reasons for remaining single at multiple time points. Using these longitudinal data, we can examine how the same individuals experience value and behavioral changes as they alter their living arrangements. This analysis enables us to overcome the difficulty of identifying the extent to which the different views and behaviors between two groups of never-married people—those living with and those living separate from their parents—are attributable to unobserved differences that may have made the two groups choose different living arrangements in the first place (e.g., personality).
Aside from the availability of unique longitudinal data, the case of Japan is particularly worthy of attention because it exemplifies both high prevalence of coresidence with parents and rapid declines in marriage and fertility. With the age at first marriage near and over 30 years old for women and men, respectively (Statistics Bureau, Japan 2014), Japan is among the countries experiencing the greatest delays in first marriage in the world. Late marriage is especially consequential for Japan’s fertility level because nonmarital births there remain rare (Hertog and Iwasawa 2011; Raymo et al. 2009). In this sense, examining the mechanisms linking premarital living arrangements to the timing of first marriage helps shed light on factors contributing to Japan’s exceptionally low fertility rate. Moreover, given that many other Asian countries share Japan’s extremely close connection between delayed marriage and low fertility as well as its cultural norms about coresidence (Jones 2005, 2007; Zeng et al. 1994), studying Japan is also useful for understanding the very low fertility in various parts of Asia as well as in similar contexts elsewhere.
Research on Premarital Living Arrangements
Most research on young adults’ living arrangements has focused on their decisions to leave their parents’ home, especially before entering marriage (Billari and Liefbroer 2010; Goldscheider and DaVanzo 1989; Goldscheider and Goldscheider 1999; Mulder et al. 2002; White 1994; Zeng et al. 1994). In Western societies, young adults’ paces of home-leaving depend on their parents’ education and resources, their own school enrollment, exposure to nonintact family structures, perceptions of parental expectations, and the local job and housing markets (Aquilino 1991; Billari and Liefbroer 2007; Buck and Scott 1993; Goldscheider and DaVanzo 1989; Hughes 2003; White 1994). Across countries, both cultural norms and macroeconomic conditions are thought to be important. In societies that feature strong family ties, frequent intergenerational exchanges, and normative expectations for parents’ continued care for children in early adulthood, young people coreside with parents longer, and their departure from the parental home coincides more with their entry into first marriage (Dalla Zuanna 2001; Manacorda and Moretti 2006; Zeng et al. 1994). At the same time, countries with greater delays in transitions out of the parental home tend to have experienced more severe deterioration in youth’s income and employment opportunities, and they often lack institutional support for youth’s access to independent housing (Aassve et al. 2002; Billari and Tabellini 2011; Cordón 1997). The extent to which colleges and universities are spread across the country, as opposed to concentrated in major population centers, also shapes the need for young people to leave home for higher education, leading to different timing of home-leaving across countries (Mulder et al. 2002).
In contrast to factors shaping young people’s premarital living arrangements, the consequences of continued coresidence with parents are less studied (Ward and Spitz 1992). Research addressing such consequences has often focused on parent-child relations. Coresidence, for example, has been shown to increase both closeness and tension between parents and adult children despite variations by the extent of children’s financial dependency and reasons for remaining at the parental home (Aquilino 1999; Aquilino and Supple 1991; Umberson 1992). On the whole, coresident young adults give, receive, and perceive more support from their parents than their peers who have left home (White and Rogers 1997). Over the long run, those who stay in the parental home longer maintain closer relations with their parents and are more willing to take in an aging parent or adult child in need of help (Goldscheider and Lawton 1998; Leopold 2012).
Although extended coresidence does not necessarily lead to parents’ or children’s dissatisfaction (Billari and Tabellini 2011; Manacorda and Moretti 2006), a delay in the departure from the parental home may have other consequences given that such departure is closely associated with other major life events in the transition to adulthood (Guzzo 2006). Specifically, young adults’ premarital living arrangements may play a role in their transitions to marriage and parenthood. An earlier study examining this connection found that people who left the parental home at a younger age want fewer children, leading to the argument that having a longer experience of living independently facilitates a taste for smaller families (Goldscheider and Waite 1991). More recent research, however, has generally observed a later entry into parenthood, a higher rate of childlessness, and reduced fertility among those having experienced extended coresidence with parents (Billari and Tabellini 2011; Schaffnit and Sear 2014). Coresident young adults’ higher likelihood of experiencing delays in marriage or union formation is thought to be the main explanation for the association between extended coresidence and lower fertility (Dalla Zuanna 2001; Kins and Beyers 2010; Miyamoto et al. 1997; Yamada 1999).
Perhaps because of the prevalence of unmarried adult children living in the parental home, researchers concerned about demographic changes in East Asia and Southern Europe are especially interested in the potential impact of extended coresidence on marriage formation (Dalla Zuanna 2001; Huang 2013; Raymo and Ono 2007; Yamada 1999). The modest amount of empirical research on how living with parents is associated with individuals’ timings of first marriage, however, has not produced entirely consistent findings. Analyses of data collected in the early 1990s, for example, showed that Japanese women coresiding with parents were likely to enter their first marriage at a faster pace than those living independently (Raymo 2003a). Nevertheless, Japanese adults living with parents have experienced greater increases in the age at first marriage than those in other premarital living arrangements over time, to the extent that coresidence is associated with later marriage among men of a more recent cohort (Raymo 2003b). A study using recent data from Taiwan also showed that men and women living with parents enter their first marriage significantly later (Yu et al. 2012). Perhaps in the past, stronger parental control over adult children’s marriage decisions in East Asia made coresident adult children—especially daughters—more subject to the pressure to marry. Living in the parental home could become a deterrent to marriage after parental control weakens, as it has in recent years (Tokuhiro 2010).
Not only is evidence on the relation between living with parents and the pace of entering first marriage inconclusive, but previous research has also provided little insight into how exactly remaining at the parental home may accelerate or delay the transition to first marriage. Even when studies have addressed the mechanisms linking premarital living arrangements with marriage postponement, they have offered no direct evidence on whether young adults’ marriage-related preferences, behaviors, and opportunities vary with their living arrangements (Raymo 2003b; Raymo and Ono 2007; Sakamoto and Kitamura 2007; Tsuya 2000). In the next section, we review research on coresidence and transitions to first marriage in Japan and provide details on what remains to be studied regarding the mechanisms connecting premarital living arrangements to the formation of first marriage.
Coresidence and Transitions to First Marriage in Japan
Japan is one of the most notable among the industrialized countries that have experienced drastic decreases in marriage in recent decades. Not only is Japan’s age at first marriage exceptionally high, at approximately age 30, but its trend of later and fewer marriages is also closely linked to fertility declines. Unlike in many Western industrialized countries that have undergone marriage declines, cohabitation and nonmarital childbearing remain relatively rare in Japan (Fukuda 2009; Hertog and Iwasawa 2011; Raymo et al. 2009). Therefore, delayed marriage is highly consequential for Japan’s birth rates (Jones 2007), ultimately driving the country into the lowest-low fertility level. Because of the significance of marriage declines in Japan, much recent research has been devoted to explaining this trend. One common focus of this research is increases in Japanese women’s educational attainment: such increases could contribute to marriage postponement by prolonging women’s years of school enrollment; by reducing their pool of marriage partners, assuming their persistent preference for educational homogamy; and by enhancing their long-term economic potential (Raymo 2003a; Raymo and Iwasawa 2005; Raymo and Ono 2007; Retherford et al. 2001). The continued difficulty that married women experience in balancing work and family demands is also thought to account for Japanese women’s avoidance of marriage (Boling 2008; Tsuya 2000).
Young adults’ extended coresidence with parents is another factor that is frequently cited in the literature on later and fewer marriages in Japan. Unlike in the United States and certain parts of Europe, young men and women in Japan rarely leave the parental home before marriage unless their jobs or schools are located far from their family of origin. Among the never-married who do live apart from their parents, most live alone rather than cohabit (Fukuda 2009). Because young Japanese adults’ premarital living arrangements are largely a result of geographic necessities related to employment or schooling, rather than their own values or resources, it is not surprising that the proportion of the never-married who coreside with parents has remained rather constant since the 1970s (Fukuda 2009; Raymo and Ono 2007). Given this lack of change, premarital living arrangements per se are unlikely to explain the trend of later and fewer marriages in Japan. Nevertheless, by further decreasing the appeal of marriage, extended coresidence with parents may heighten more recent cohorts’ tendency to postpone marriage. The argument for the detrimental effect of coresidence on the appeal of marriage typically emphasizes the “comfort of home,” contending that coresident adult children’s dependency on their parents for free room and board decreases their incentives to form their own family (Miyamoto et al. 1997; Sakamoto and Kitamura 2007; Tsuya 2000; Yamada 1999). For women, remaining at the parental home further shields them from the housework burden they must face upon entering marriage, making marriage comparatively unappealing (Raymo and Ono 2007). Not all empirical findings on premarital living arrangements and transitions to first marriage, however, support the comfort-of-home hypothesis. To begin, among men and women living with parents, those with wealthier parents do not postpone marriage more; in fact, women who are likely to benefit more from their parents’ economic resources marry faster (Raymo 2003b). Moreover, there is no evidence that coresident women who contribute more to the household income make a transition to their first marriage earlier (Raymo and Ono 2007). In addition, receiving financial resources from parents delays coresident women’s transitions to first marriage for only a particular cohort, and the father’s income relative to the potential husband’s is hardly relevant (Sakamoto and Kitamura 2007). By contrast, there is some support for the importance of the relief of the housework burden that young women may experience by remaining at the parental home. Among Japanese women living in the parental home, making a greater contribution to household work is associated with faster marriage transitions, but only for those with higher education (Raymo and Ono 2007). Regardless of whether their results are consistent with the comfort-of-home hypothesis, few studies addressing this hypothesis have directly examined the influences of premarital living arrangements on the desire for marriage because they have generally focused on marriage transitions—not intentions—as the outcome.1
Using marriage behaviors to approximate marriage intentions is problematic. Even with full intent to marry, people may enter first marriage at different paces—mainly because they may find marriage partners at different speeds. Discovering longer work hours to be detrimental to Japanese women’s chances of marriage, Sakamoto and Kitamura (2007) suggested that having the time to search for and meet romantic partners is critical for the transition to first marriage. Young adults living in the parental home may have less time available for partner seeking than those living independently. Although coresident adult children may spend less time on household work than their peers who maintain an independent residence, they might spend less time outside the home because their more frequent exchanges with their natal family could end up taking more of their time (White and Rogers 1997). In Japan, because most of the never-married who do not reside in their parental home live alone (Fukuda 2009), loneliness at home could plausibly drive noncoresident singles to spend comparatively more time outside. As a result, noncoresiding individuals may have more opportunities to meet and socialize with potential marriage partners, thereby marrying faster.
Living in the parental home may also affect adult children’s perceptions of and needs for social support, thus influencing their partner-seeking and marriage behaviors. Specifically, because coresident adult children perceive that they are getting more support and possibly feel more solidarity with their family of origin (Goldscheider and Lawton 1998; Leopold 2012; White and Rogers 1997), they may be less needy for social companions or other intimate ties. In support of this argument, research on Japan has found that coresident young adults are less likely than those living independently to think that marriage will improve their emotional security (Tsuya 2000; Tsuya et al. 2004). Having reduced social needs may make young people living in the parental home choosier in selecting romantic partners, even if they are no less interested in marriage.
Although much of the literature has emphasized coresident singles’ marriage postponements (Raymo 2003b; Sakamoto and Kitamura 2007; Yamada 1999), equally plausible is that the differing paces of entering marriage among individuals with different living arrangements are related to the ways in which residential independence facilitates the transition to marriage. Specifically, because young people who have left the parental home also report being more independent and capable in other aspects of life (Kins and Beyers 2010), they may be viewed as more attractive. Such people, therefore, may be able to find romantic partners more easily and thus marry earlier. In Japan, given the lingering cultural norm that encourages sons to continue residing with parents after marriage (Yasuda et al. 2011), independent living may also make men more attractive by signaling that their future spouses will not be compelled to live with their in-laws. Even if residential independence does not affect young adults’ perceived attractiveness, it may shape their psychological readiness for marriage by influencing their self-assessments of their maturity and capabilities. Finally, independent living could potentially accelerate young adults’ entry into marriage if those living apart from their parents are more likely to become pregnant before marriage. Because of the low social acceptance of nonmarital childbirth and an increase in premarital sexual activity, the number of “shotgun marriages”—that is, marriages resulting from nonmarital conceptions—has been rising in Japan (Raymo and Iwasawa 2008). Residential independence might enhance premarital sexual activity and unplanned pregnancies, thereby hastening the transition to first marriage.
In sum, although research related to premarital living arrangements enables us to suggest alternative reasons why coresidence with parents may be conducive to marriage postponement, few studies have examined reasons beyond the material comfort of home when considering the relevant mechanisms (Raymo 2003b; Raymo and Ono 2007; Sakamoto and Kitamura 2007). To better disentangle the various possible mechanisms, in this study, we directly analyze how premarital living arrangements affect marriage desires, romantic activities, and opportunities to meet potential partners in Japan. Perhaps equally important, we use longitudinal data to examine changes in individuals’ marital aspirations and courtship patterns according to alterations in their residential status. Much of the previous research addressing the mechanisms linking coresidence to marriage formation limited the sample to coresident women or relied on cross-sectional comparisons between coresiding and noncoresiding individuals (Raymo and Ono 2007; Sakamoto and Kitamura 2007; Tsuya 2000; Tsuya et al. 2004). Such analytical approaches are either insufficient to reveal the full influence of coresidence or unable to account for unobserved characteristics that might have led to unmarried adults’ differing residential choices in the first place. By using panel survey data that contain information for both men and women coresiding or noncoresiding with parents over time, we are able to better address the potential selection bias.
Data and Methods
Data used in this study are from the Japan Life Course Panel Survey (JLPS), conducted by the Institute of Social Science at the University of Tokyo. The JLPS is one of the rare and most prominent longitudinal surveys in Japan. It collected information from a nationally representative sample of men and women aged 20–40 in 2007 and has been following them every year since (for details of the sampling and data collection procedures, see Ishida 2013). The JLPS asked all single respondents to report their aspirations toward marriage, involvement in partner-seeking activities, opportunities to meet potential romantic partners, romantic relationship status, and reasons for currently being single at various waves. This rich information enables us to examine how individuals’ marriage intentions, romantic activities, and courtship opportunities vary with their residential arrangements over time. We use the data from Waves 1–5 in this study, but we exclude a supplemental sample that was newly added at Wave 5 because our analysis requires multiple observations from the same individuals by Wave 5. For the same reason, we also eliminate the cases that were interviewed at Wave 1 but did not complete any of the follow-up surveys (13 %). To conduct the statistical analysis, we convert the five waves of data into person-year observations, with time-varying information for each respondent. Because our ultimate focus is on the transition to first marriage, we further restrict the sample to person-years during which respondents were never married. After these restrictions, our analytical sample contains 1,061 men and 969 women.
Like all panel surveys, the JLPS experienced attrition at every wave. Aside from the 13 % of respondents who were not reinterviewed after the first wave, 15 % of the person-years during which respondents were likely to be never married were missing because some respondents were not interviewed at given waves.2 Following previous research diagnosing panel survey attrition problems (Fitzgerald et al. 1998), we fit logistic regressions by gender, predicting missing any wave before marrying with baseline characteristics, including respondents’ living arrangements and marriage-related views and activities.3 The pseudo-R2 values for the regression models for both gender groups were very small (≤.02), indicating that the variables related to our interests are of little relevance to the attrition. The model for women further passed the test proposed by Becketti et al. (1988) for ruling out selection bias. Even for men, those missing one or more waves were not statistically different from the rest in their likelihood of coresiding with parents, marriage aspirations, participation in partner-seeking activities, opportunities to meet romantic partners, and romantic involvement at the initial wave.4 Overall, our additional analysis suggests that attrition from the JLPS is unlikely to distort the results.
The statistical analysis in this study contains three parts. For each part, we fit separate models for men and women because the processes of transitioning to first marriage are likely to differ by gender. We begin with an event history analysis examining the relationship between coresidence with parents and the pace of entering first marriage given that prior research on this relationship has been inconclusive. We code the event of interest—entering first marriage—as 1 for time t if respondents changed their marital status from never-married to married between the previous wave of data collection (t – 1) and the current wave (t); otherwise, we code it as 0. To ensure that the individual characteristics used to predict the entry into first marriage precede the marriage transition, we use individual data collected at time t – 1 to create all the predictors for the first marriage occurring at time t. The analysis uses discrete-time logistic hazard models, which estimate the log odds of individuals entering their first marriage at a given time, conditional on the event not occurring before that time. Because our person-year observations are left-truncated, given that they were derived from panel survey data, we use conditional likelihood models to adjust for potential bias. Conditional likelihood models are virtually the same as discrete-time hazard models except that the duration of the spell is measured from the time respondents begin to be exposed to the risk of the event occurring, rather than from the time they enter the sample (Guo 1993). In this case, we assume that Japanese men and women are first exposed to the risk of first marriage when they turn age 18, the age at which most of them complete high school. Today, almost all Japanese complete high school (Yu 2009), and marrying while in high school or earlier is extremely rare (Raymo 2003a). Thus, we include the duration since age 18 in the conditional likelihood models, regardless of the age at which respondents enter the sample.
Although our use of conditional likelihood models helps adjust the bias caused by missing information on the years before the never-married entered the survey, our inability to observe the premarital experiences of those who had already been married at the initial wave can still be a problem if the processes of transitioning to first marriage were substantially different for the latter group. To assess the potential impact of this data limitation, we fitted logistic regressions on whether respondents were never married at the time of entering the panel survey and found age to have the most explanatory power for both men and women.5 Given this finding, we also estimated all the models with a much less selective sample: those who were age 28 and younger and never married at Wave 1. The results were generally similar, thus enhancing our confidence about the patterns reported in this article.
Turning back to the event history analysis, in the models predicting entry into first marriage, we also include desire for marriage, opportunities to meet potential partners, participation in partner-seeking activities, and romantic involvement to assess how taking into account these conditions affects the relation between coresidence and marriage transitions. To more directly identify the connections between premarital living arrangements and marriage-related intentions, behaviors, and opportunities, we estimate a series of fixed-effects models in the second part of the analysis, using the same set of variables—marital aspirations, partner-meeting opportunities, engagement in partner-seeking activities, and romantic involvement—as the outcomes, with coresidence with parents as the main predictor. Because fixed-effects models ultimately rely on variations in the same individuals over time to estimate the coefficients while dropping all time-invariant predictors from the models, they enable us to take into account unobserved characteristics that may simultaneously affect individuals’ residential choices and marriage-related outcomes—as long as these characteristics are constant over time. Personality, childhood family dynamics, and adolescent romantic experiences are examples of such characteristics.
We measure desire for marriage from the JLPS question on single respondents’ marriage aspirations. At each wave, these respondents were asked to select among the following options: (1) I definitely want to marry, (2) if possible I would like to marry, (3) I am okay with marriage or no marriage, (4) I do not want to marry, and (5) I never thought about the issue of marriage. We combine (5) with (3) because they both represent a rather neutral attitude toward marriage. We then create an index of marriage desire by reverse-coding the options—that is, coding the first option as 4, the second as 3, the third as 2, and the fourth as 1. We measure partner-meeting opportunities based on respondents’ answers to the question of how frequently they encounter members of the other sex whom they may want to date. Based on the four response categories—“very frequently,” “somewhat frequently,” “not frequently,” and “rarely”—we create an index ranging from 1 to 4, with 4 representing having the most opportunities. Unfortunately, the question about partner-meeting opportunities was not asked at Wave 2. Therefore, all models including this variable must rely on data from only four waves and thus have fewer observations.
With respect to engagement in partner-seeking activities, the JLPS asked respondents at each wave whether they had been using any of a list of ways to meet potential romantic partners since the last wave: for example, asking parents, siblings, or other family members for introductions; asking friends or acquaintances for introductions; asking coworkers or supervisors for introductions; participating in enrichment lessons and hobby meetings; taking courses or joining social club activities; using the Internet; taking part in arranged dates; attending match-making parties; and employing dating services. We first create a variable for the total number of methods respondents reported to have been using from the list of 13 available options. Because the more formal means of meeting romantic partners may be more effective than the informal ones, we create two additional variables for the numbers of formal and informal partner-search methods used, respectively. We consider asking parents or family members for introductions, taking part in arranged dates, attending match-making parties, and using match-making agencies or dating services to be relatively formal ways of finding romantic partners. All the other methods, such as asking friends to introduce and joining social club activities, are considered as informal. A separate factor analysis, not presented here, supported this division of partner-seeking methods into formal and informal ones.
For romantic relationship status, the JLPS asked respondents at each wave whether they were currently (1) engaged, (2) romantically involved with someone, or (3) not romantically involved at all. We use a binary variable to indicate whether respondents were romantically involved (i.e., those selecting one of the first two responses). In addition, because being engaged indicates a higher level of intensity of the romantic relationship, we create an indicator for intensity of romantic involvement, coding “engaged” as 3, “involved with someone” as 2, and “no involvement” as 1. In the JLPS section where respondents were asked to identify their reasons for remaining single, they also could report that they planned to marry their current romantic partner soon; thus, we also consider the small number of respondents who were not engaged but had plans to marry soon as having the highest intensity of romantic involvement.
The third part of the analysis takes advantage of the self-identified reasons for being single collected by the JLPS at Waves 1, 3, and 5. The questionnaires for those three waves allowed single respondents to choose any number of listed reason(s) indicating why they were not currently married. The list of reasons covers issues related to economic concerns (e.g., “I do not have sufficient funds to marry” and “marriage will lead to greater anxiety over economic conditions”); lifestyle choices and preferences (e.g., “I want to enjoy hobbies and have fun,” “I do not want to lose the freedom and comfort of being single,” and “marriage will lead to greater anxiety in life in general”); difficulties encountered during partner searches (e.g., “I am not good at socializing with the other sex” and “I have not met a suitable marriage partner”); and perceived conflict between marriage or relationships and other time demands (e.g., “I want to focus on my job/school” and “I do not have time for courtship”). Other reasons available are related to respondents’ sense of maturity and perception of appropriate timing for marriage (e.g., “I am too young to marry” and “I am too old to marry”). Respondents were also asked whether they remained single because their home is “warm and cozy.” We can therefore provide evidence on how leaving the parental home alters individuals’ assessment of the warmth of their current home, which could produce different needs for social companionship outside the home.6
Because the questions about reasons for being single were asked multiple times, we are able to estimate fixed-effects models predicting individuals’ probabilities of choosing each of these reasons, with coresidence with parents as the main independent variable. These models enable us to show whether changes in premarital living arrangements lead to changes in never-married adults’ perceptions of economic disadvantages brought by marriage, difficulties in meeting marriage partners, time availability for partner searching, and psychological readiness for marriage. Our focus on changes in perceptions is important because individuals are likely to notice changes in marriage-related views and experiences and hence to report such changes, even if they are not aware of all the reasons for their remaining single.
To simplify computations and interpretations, we fit linear regressions for all the fixed-effects models. For binary outcome variables, such as whether respondents were single for a certain reason, this approach means that linear probability models with individual fixed effects are estimated (Aldrich and Nelson 1984). In such models, the coefficients directly indicate increases or decreases in the probability of experiencing the outcome.7 In a separate analysis, we found the results to be similar when we instead used logistic regressions with fixed effects for all binary outcome variables. For dependent variables that have more than two values, such as the indexes of marital aspirations and partner-meeting opportunities, our use of fixed-effects linear regressions ultimately assumes these variables to be continuous. To be sure that this assumption does not bias the results, we also estimated fixed-effects ordered logit models via the “blow-up and cut” estimator, as proposed by Baetschmann and colleagues (2015),8 with these same dependent variables. We found no meaningful changes in the results when we treated marital aspirations, partner-meeting opportunities, and the level of romantic involvement as ordinal variables. For consistency, we also treat the indexes of marital aspirations and partner-meeting opportunities as continuous variables when we include them in the event history models predicting the entry into first marriage. Our additional check indicated that using categorical dummy variables instead of a single continuous measure for these variables did not alter the main results.
A similar set of independent variables is used in all parts of the analysis. For our primary predictor of concern—namely, coresidence with parents—we use the self-reported household compositions at each wave to construct a dummy variable representing living with at least one parent during the observed year.9 Other relevant time-varying sociodemographic predictors include age, educational level, school enrollment, employment status, personal income, and size of the city. We include age and age squared to capture the potentially nonlinear relationship between age and various marriage- and courtship-related opportunities and behaviors. We measure the level of education based on the highest level of school the respondents had completed by the survey time. We construct a dummy variable indicating that respondents were enrolled in school at the observed time. In addition to age and employment status, we control for school enrollment because many respondents attended school while holding jobs, and a few returned to school after starting their careers.10 For employment status, we divide respondents into four categories: (1) regular, full-time employees; (2) part-time or temporary employees; (3) persons engaged in family enterprise or self-employed; and (4) jobless persons. We measure annual personal income, in units of 100,000 yen, by taking the midpoint of the income categories provided at each wave of the survey. In addition, we introduce a time-varying variable indicating whether respondents lived in (1) any of 16 major population centers, (2) cities with more than 200,000 residents, (3) other cities, or (4) towns and villages. Controlling for the type of cities (or towns and villages) in which respondents lived is important because a move away from the parental home often involves moving to larger cities, where there are more jobs and universities, and residing in cities of different scales implies different opportunities to meet marriage partners.
Because individuals’ time available may affect their partner-search processes and opportunities, we further control for respondents’ reported work hours and commute time, also measured in hours, on a typical day. Given that these two variables are available only for those with jobs, we center both of them by subtracting the gender-specific sample mean, while coding those without jobs as 0.11 After this transformation, the estimated coefficients for each employment status in our models would represent the difference in a given outcome between a jobless person and a person with a certain employment status and average work hours and commute time, whereas the coefficients for work hours and commute time would indicate, among those with jobs, the extent to which one unit of change from the average work hours or commute time will alter the outcome. To illustrate the characteristics of the analytical sample, Table 1 presents descriptive statistics of never-married men and women at the first wave.
Aside from the time-variant predictors discussed earlier, we should note that prior research suggests that certain time-invariant factors, such as individuals’ family economic background and sibship positions, may also be relevant to the first part of our analysis (Raymo 2003b; Yu et al. 2012), in which we use conditional likelihood models to predict the occurrence of first marriage. In an exploratory analysis, we added various indicators of family economic status and sibship characteristics in the conditional likelihood models.12 Nevertheless, these indicators generally had no significant effects, perhaps because their effects are mediated by marriage desires and courtship behaviors, which are also included in the models. For parsimony and to conserve space, we do not include those time-invariant variables in the models presented.
To examine the relation between coresidence and marriage formation, Tables 2 and 3 show results from conditional likelihood models on the entry to first marriage for men and women, respectively. Model 1 for men indicates that living in the parental home is associated with a slower pace of entering first marriage, conditional on respondents not marrying at an earlier time point. Although having a stronger desire for marriage significantly accelerates the pace of transitioning to first marriage, adding marital aspirations in Model 2 reduces the negative coefficient of coresidence only slightly. Models 3 and 4 also show small reductions in the association between coresidence and entering first marriage in the subsequent year, with the association continuing to be negative and significant at the .05 level. In addition, Model 3 reveals that, surprisingly, using more partner-seeking methods slows down never-married men’s transitions to first marriage. Model 4, however, clarifies that only the use of informal partner-search methods is detrimental. A possible reason why active partner-seeking does not help speed up men’s transitions to first marriage is that those who already have a romantic partner—and hence are likely to marry sooner—are unlikely to be engaged in any partner-search activities. In Model 5, we add the variable of being romantically involved to account for the confounding effect of having a romantic partner. This addition alters the coefficients for formal and informal partner-search methods slightly. After we further differentiate those ready to marry from those in relationships in Model 7, however, engagement in formal partner-search activities becomes positively associated with the pace of entering first marriage, whereas using informal channels to seek partners remains negatively associated with this pace.13 Clearly, not all partner-search methods are equally effective. Perhaps because those trying to meet romantic partners through more informal routes are likely to be tempted by more possibilities, or because the behavior of trying more informal means reflects greater difficulties in attracting a partner, participating in more informal partner-search activities is linked to greater postponement of marriage.
Models 5 and 7 also demonstrate the strong impact of being romantically involved on the log odds of entering first marriage. The odds of Japanese men involved with a romantic partner becoming married during the subsequent year are 21 times greater than those of men who are not involved (exp(3.057) = 21). Not surprisingly, the chance of entering first marriage in the next year is even greater if a man is engaged (Model 7). Thus, finding a romantic partner ultimately is the key to transitioning to first marriage. Interestingly, taking into account the level of romantic involvement, as in Models 5 and 7, also leads to a substantial reduction in the coefficient of living with parents. It appears that the likelihood of romantic involvement largely accounts for the negative association between coresidence and the pace of entering first marriage.
Because one way in which residential independence can accelerate the transition to marriage is through increasing the likelihood of nonmarital conception for those with romantic partners, we introduce an interaction between being romantically involved and coresidence with parents in Model 6. Although a faster pace of entering first marriage among noncoresident singles who are romantically involved does not necessarily mean that they experience more premarital pregnancies, we should observe such a pace if residential independence indeed facilitates marriages preceded by pregnancies. The coefficient for the interaction, however, is only marginally significant, providing rather weak support for the argument. In a further analysis, we excluded the cases of those who were likely to have entered marriage as a result of premarital conceptions—that is, the small number of respondents who had a child in the same or previous year of entering first marriage. The results remained similar, suggesting that the likelihood of nonmarital conception explains, at most, a very small part of the difference in the pace of transitioning to first marriage between singles with different living arrangements.
Model 8 in Table 2 adds male respondents’ reports on their opportunities for meeting members of the other sex whom they would want to date. Because this variable is not available for one of the waves, the number of observations is lower in this model. Although Model 8 is not nested with the other models, the results for the other variables in this model are generally similar to those discussed earlier. Interestingly, for Japanese men, having more opportunities to meet potential romantic partners actually delays their entry into first marriage. This finding suggests that having more opportunities makes it more difficult for men to decide on a specific person to marry because they may want to spend more time “trying out” various options.
The results for Japanese women’s transitions to first marriage are generally similar to men’s. As Table 3 shows, living in the parental home is negatively associated with the log odds of entering first marriage in the following year. Adding marital aspirations actually makes coresidence slightly more negatively associated with the occurrence of first marriage, rather than explaining away the association. Like for men, introducing romantic relationship status leads to the greatest reduction in the association between coresidence and entry to first marriage in the subsequent year. Also similar to men, using more formal partner-search means accelerates women’s transitions to first marriage, whereas taking part in informal partner-seeking activities is conducive to a postponement of first marriage. Although being romantically involved is strongly associated with the entry into first marriage, among those with romantic partners, there is no difference by premarital living arrangement. This result suggests that similar to men, a higher likelihood of premarital conception is unlikely to explain why residential independence hastens the marriage transition for women. The only major gender difference is that women reporting that they have more opportunities to meet potential dating partners are no less likely to become married. Having more options appears not to prolong women’s search for marriage partners as it does men’s.
To be more certain about how premarital living arrangements are linked to never-married men’s and women’s marriage intentions and romantic opportunities and behaviors, Tables 4 and 5 present a series of fixed-effects models predicting these outcomes for men and women, respectively. The results show that contrary to the comfort-of-home hypothesis, Japanese men’s desire for marriage hardly changes with their living arrangements.14 Living with parents is also associated with neither changes in men’s opportunities to meet potential romantic partners nor their extents of participation in formal and informal partner-seeking activities. These findings suggest that time use, to the extent that it affects singles’ partner-search efforts and exposure to potential partners, is unlikely to explain the connection between coresidence and the transition to first marriage. Living with parents, however, is significantly and negatively associated with being romantically involved. That is, as men in Japan move away from their parents’ home, their chances of romantic involvement increase. The results for women are similar to those for men: living in the parental home is significantly associated only with the level of romantic involvement, not any other outcomes. Japanese women’s romantic involvement intensifies when they move from the parental home to an independent residence.
Because of our use of fixed-effects models, we can avoid the potential bias resulting from the possibility that young adults who were more romantically active in adolescence may be more likely to both leave the parental home before marriage and have success in forming romantic relationships in adulthood, given that such personal traits are time-invariant. The fixed-effects modeling approach, however, does not allow us to rule out reverse causality. Specifically, the negative association between living with parents and forming a romantic relationship in the fixed-effects models could mean that Japanese people tend to change their residential arrangements as soon as they become involved with someone, rather than that living in the parental home decreases the likelihood of forming a relationship. To gain confidence in the causal direction, we conducted a separate analysis and found that the majority of men and women (75.8 % and 83.1 %, respectively) who did not live with their parents during the years of transitioning into a relationship also did not live in the parental home in the previous year. Thus, most of these people moved out of the parental home before they entered a relationship. In addition, men’s and women’s chances of moving from the parental home to an independent residence during the year when they started a romantic relationship do not statistically differ from their chances in the years when they were not in a relationship.15 To further enhance our ability to make causal inferences, we fit separate logistic and ordered logit regression models using the previous year’s living arrangements to predict the current year’s romantic involvement, while including individual random effects to take into account unobserved heterogeneity.16 The models also control for the previous year’s romantic relationship status. Table 6 presents results from these additional models.
As illustrated in Table 6, Japanese men living with parents in the previous year are significantly less likely to be involved in a romantic relationship, after the level of romantic involvement in the previous year is controlled for. The results are similar regardless of how we measure romantic involvement—by using a dichotomous measure or by distinguishing different levels of romantic involvement. For women, the past year’s coresidence is similarly associated with a lower likelihood of being romantically involved in the current year. These results, along with the descriptive analysis showing that most people living independently moved out their parental home before entering a relationship, provide strong support for the argument that coresidence with parents reduces individuals’ likelihood of forming a relationship, leading to their slower paces of entering first marriage.
Why does living in the parental home decrease the chance of forming a romantic relationship if the former does not alter never-married adults’ desire for marriage, opportunities to meet potential partners, or participation in partner-seeking activities? Table 7, which summarizes the coefficients of coresidence on self-identified reasons for being single based on a series of fixed-effects models, helps shed light on this question. All 18 models on which the table is based include the same set of independent variables as in Tables 4 and 5. For simplicity, we report the actual numbers and standard errors only when the p value for the coefficient of coresidence is smaller than .10 (otherwise “N.S.,” for “nonsignificant,” is noted). Consistent with previous research showing little support for the notion that coresident adult children postpone marriage to take advantage of parents’ financial aid, living with parents does not increase men’s or women’s likelihood of identifying a probable increase in economic anxiety as the reason not to marry.17 Also contradictory to the comfort-of-home argument, living in the parental home does not make the never-married more likely to fear the loss of freedom and comfort with marriage, nor does it make them more likely to think that marriage will bring greater overall stress and anxiety in life. Moreover, coresidence is not associated with Japanese men’s and women’s probabilities of indicating not having time for courtship or wanting to focus on work as the reason for being single, suggesting that coresidence does not increase individuals’ time conflict with seeking and meeting a marriage partner. Therefore, time availability is unlikely to explain the negative association between coresidence and relationship formation, either. Likewise, because coresidence does not increase individuals’ probabilities of reporting difficulties finding suitable marriage partners or courting members of the other sex, living in the parental home appears not to make the never-married less attractive in Japan.
For Japanese men, the only reason for being single that is significantly associated with living with parents is “my home is warm and cozy.” Because a change in living arrangement is irrelevant to their probabilities of identifying the comfort of being single or the risks of increases in overall or economic anxiety with marriage as the reasons for remaining unmarried, it seems that by “warm and cozy,” respondents referred more to a sense of social connectedness and support felt at home, rather than the freedom or comfort enabled by the parents’ provisions of room and board and housework. For women, living with parents is significantly associated with the likelihood of feeling too young to marry, even after we control for age. This finding suggests that living in the parental home reduces Japanese women’s sense of maturity and psychological readiness for other major steps in the transition to adulthood. Both the greater contentment with the current social environment and the lack of psychological readiness for marriage could make never-married adults living with parents choosier—or less eager to settle—in the process of seeking romantic partners. Thus, even though coresidence does not reduce never-married individuals’ marital aspirations, partner-meeting opportunities, or partner-seeking effort in Japan, it makes them less likely to form romantic relationships, which constitutes a major impediment to their transitions to marriage.
Much previous research on demographic shifts in such regions as East Asia and Southern Europe calls attention to the influences of premarital living arrangements on young adults’ timings of first marriage (e.g., Dalla Zuanna 2001; Raymo 2003b; Raymo and Ono 2007; Tsuya et al. 2004; Yamada 1999), but the specific mechanisms linking the two are understudied. Using longitudinal data on never-married individuals’ marriage- and courtship-related views and activities in Japan, we shed light on how the prevalence of young adults’ coresidence with parents plays a role in shaping the demographic trends in that country. Despite some mixed findings in the past, results in this study demonstrate that Japanese men and women living in the parental home enter their first marriage at a slower pace than those living apart from their parents. Given that the data used in this study are much more recent than those used in the ones that show contradictory results (e.g., Raymo 2003a), our findings provide support for the argument that extended coresidence has become a deterrent to the transition to first marriage with the weakening of parental control over adult children’s marriage decisions in recent years. Whereas coresident adult children might have been subject to greater parental pressure and to marry sooner when parents had more influence over children’s marriage decisions, this is no longer the case in Japan today.
Because coresidence might not be conducive to marriage postponement without the weakening of parental influences on adult children’s marriage, and because the prevalence of coresidence among the never-married has not increased much with time (Fukuda 2009), we cannot argue that young adults’ extended coresidence alone explains Japan’s trend of later and fewer marriages. Our finding of a negative association between living with parents and the pace of transitioning to first marriage, however, indicates that after other forces leading to marriage postponement gain momentum, the prevalence of never-married adults’ coresidence with parents can further aggravate the declines in marriage and fertility.
Beyond clarifying the relation between premarital living arrangements and marriage timing, results from this study show that coresidence delays individuals’ transitions to first marriage mainly through decreasing their chances of forming romantic relationships. Interestingly, the decreased likelihood of romantic involvement is not a result of coresident adults’ having fewer opportunities to meet potential partners or trying less hard to find partners. A lower interest in marriage also cannot explain the negative association between living in the parental home and relationship formation because coresidence with parents is not relevant to never-married individuals’ desire for marriage. This absence of relevance is important because it directly contradicts the argument concerning the comfort of home, which expects young adults living with parents to desire marriage less. If we take the behavior of consciously engaging in partner-search activities as a further expression of the desire for marriage, then the lack of an association between coresidence and participation in partner-seeking activities also suggests that the comfort of the parental home is insufficient to dampen never-married adults’ interest in marriage. The analysis of self-identified reasons to remain single provides even more evidence against the typical arguments of the comfort-of-home perspective; living in the parental home does not make never-married Japanese adults more anxious about the possible downgrading of economic conditions or quality of life after marriage.
In a similar vein, our analysis provides no support for the argument about time availability for partner searches given that living in the parental home affects neither never-married adults’ perceptions of time conflict with relationships nor their actual engagement in partner-seeking activities. In addition, our results indicate that coresidence is unlikely to reduce never-married adults’ likelihood of forming relationships through lowering their appeal to potential romantic partners in Japan. Living in the parental home does not increase Japanese adults’ probabilities of acknowledging difficulties in finding or courting suitable marriage partners.
Ultimately, results from this analysis suggest that coresidence with parents obstructs the formation of romantic relationships for mostly psychological reasons. Living in the parental home increase Japanese men’s likelihood to report great contentment with their current home life. Because this contentment does not coexist with a greater concern about changes in economic conditions, quality of life, or personal comfort and freedom with marriage, we suggest that instead, this contentment is likely to derive from the stronger sense of social connectedness and support facilitated by living with the family of origin, or remaining in the social environment in which they grew up. In this sense, the comfort provided by the parental home that shapes marriage transitions is really emotional comfort, which is very different from the comfort typically stressed in the comfort-of-home hypothesis. Moreover, because living in the parental home does not lower men’s desire for marriage or effort put into seeking partners, coresident men’s greater contentment with their current life most likely interferes with their relationship formation by making them less willing to settle with any “acceptable” romantic partner.
In contrast to men, women in Japan are significantly more likely to think that they are too young to marry while living in the parental home. This finding is consistent with the psychological research contending that residential independence, as a major marker in the transition to adulthood, has spillover effects onto individuals’ self-perceived capability to manage other transitions in early adulthood (e.g., Kins and Beyers 2010). The lack of psychological readiness for marriage could explain our finding that living in the parental home mainly decreases Japanese women’s intensity of romantic involvement, making them especially unlikely to be engaged or plan to marry soon.
Aside from generating new knowledge about the mechanisms linking premarital living arrangements to marriage timing, a few findings from this study offer useful insights into the processes of transitioning to first marriage in Japan. First, we show that not all kinds of partner-seeking methods are the same. Although prior research has found that active engagement in partner-seeking activities provides little help to Japanese men’s and women’s transitions to marriage (Ishida 2013), our results reveal the importance of distinguishing formal from informal methods of seeking marriage partners. After we make this distinction, it becomes evident that formal partner-seeking methods do help, whereas informal ways of meeting potential partners tend to slow the transition to first marriage. Second, similar to the use of informal partner-seeking methods, having more opportunities to meet potential dating partners actually decelerates Japanese men’s transitions to first marriage, perhaps because having more possibilities makes it more difficult to decide on “the one and only.” Interestingly, a gender difference is clear: women are not distracted by having so many possibilities, suggesting that Japanese women’s partner searches may be more focused and less affected by the availability of options. Finally, we find that being able to form romantic relationships is the ultimate key to the transition to first marriage in Japan. Coresidence with parents is relevant only to marriage transitions because it impedes young adults’ romantic involvement. Thus, future research on demographic shifts in Japan should pay more attention to factors contributing to the difficulties in young adults’ relationship formation.
Although this research focuses on Japan, the results have general implications for many other Asian countries that share Japan’s norms regarding premarital arrangements and rapid declines in marriage, such as South Korea, Taiwan, and Singapore (Jones 2005, 2007). Even for countries where leaving the parental home before marriage is more common, coresidence with parents may still affect young adults’ mentalities and, therefore, their likelihood of relationship formation. As remaining in and returning to the parental home have become more popular for singles in the United States and other Western countries (Billari and Liefbroer 2010; Qian 2012), it is increasingly important to consider the potential consequences of this trend. By offering a new conceptual framework to link coresidence with other demographic behaviors, this study makes a general contribution to our understanding of the potential implications of changes in young adults’ premarital arrangements.
The authors thank the Institute of Social Science at the University of Tokyo for granting the access to the data from the Japan Life Course Panel Survey. The first author also gratefully acknowledges a research grant from the Sumitomo Foundation, as well as support from the Eunice Kennedy Shriver National Center for Child Health and Human Development Grant R24-HD041041, awarded to the Maryland Population Research Center.
One exception is Tsuya and colleagues’ (2004) analysis of the effect of coresidence on marriage desires among single men and women in Japan. Their use of cross-sectional data, however, makes it impossible to rule out the possibility that selection into residential independence explains the observed between-group differences.
When respondents were not interviewed at a wave, we have no way of telling their marital status unless they reentered the survey at a later wave, which allows us to retrospectively fill in their marital histories. If respondents became married in the same year they missed the survey, then the lack of information for that wave should hardly affect our analysis given that the person-year observations would have been excluded from the fixed-effects models anyway. The 15 % reported here is based on the assumption that respondents remained never-married during all the missing years, if they were never married at the last complete interview. Because attrition as a result of entering marriage is likely (as indicated by an analysis of respondents who missed a wave and then reentered the survey), this assumption is likely to lead to an overestimation of the amount of eligible person-year observations that are missing.
The models also include respondents’ father’s education and their own schooling, city size, employment status, work hours, and commute time.
We also conducted an additional analysis comparing those reentering the survey after missing a wave with those with all five waves completed. We found that the association between coresiding with parents and marriage chances was in the same direction and was stronger for those reentering the survey. If those missing some interviews and not reentering by Wave 5 were similar to those who did reenter, then excluding all person-years with no interviews should only weaken the association between coresidence and marrying found in our models. In this sense, the survey attrition leads to more conservative estimates, rather than distorting the results.
In addition to age, the logistic regressions also included father’s education; respondent’s educational attainment; current school enrollment; sibship size; whether the respondent’s first job was a full-time, regular one; size of the city in which the respondent lived; and whether the respondent had ever cohabited. Other than being older, living in smaller cities (or rural areas) and having cohabited before were the two variables that increased both men’s and women’s likelihood to have entered first marriage before the first wave. Women who were more educated and still in school were also more likely to be never-married, whereas men from smaller families and whose first job was a temporary or part-time one were more likely to remain never-married. In a further analysis, we tested whether the associations between coresidence and various outcomes differed by sibship size, first job’s employment status, and city size for men and schooling and city size for women; we found no significant differences, suggesting that our underselection of men and women of certain demographic characteristics might not alter the results in a meaningful way. Moreover, although we have no information about whether those who had been married lived with their parents before marriage, the finding that this group was more likely to have had cohabitation experience suggests that the rate of independent living was higher among this group. If we consider those ever-married at Wave 1 as having a faster pace of entering a romantic relationship and first marriage, then the associations between premarital living arrangements and entry to relationship and marriage for this group would be similar to those reported in the Results section, based on what we can infer about their living arrangements. This similarity should further alleviate concern about the data limitation.
We show a full list of the reasons provided in the selected waves of JLPS in the Results section, when we present the fixed-effects models predicting individuals’ probabilities of selecting each of the reasons.
Although linear probability models can be more intuitively interpreted, they may encounter the problem of predicted probabilities falling out of the 0–1 range. In our additional checks, however, we found no such problem for the fixed-effects linear probably models presented in this article.
We used two user-written programs for Stata, bucologit (version 11.2) and feologic_buc (version 10), to implement fixed-effects ordered logit models with the “blow-up and cut” estimator. Both programs yielded results similar to ours from fixed-effects linear regression models.
We also treat 27 person-year observations for those who lived with a grandparent, but no parent, as living in the parental home, with the assumption that the grandparents serve functions similar to those of parents. The results were virtually the same when we coded these observations as not living with parents.
An additional analysis estimating variance inflation factors indicated that including both age and school enrollment does not cause excessive multicollinearity.
The reason for coding the job characteristics of the jobless as 0 can be illustrated with a simplified equation: M = b0 + b1iXi + [b2 + b3j(Yj – j)]E, where M represents marriage intentions; Xi represents i predictors relevant to both employed and nonemployed people; (Yj – j) indicates j job characteristics centered on the group means; and E indicates whether observations are employed (coded as 1) or not (coded as 0). When an observation is not employed, M = b0 + b1Xi, whereas M = b0 + b1Xi + [b2 + b3j(Yj – j)] when an observation has a job. When an observation has average job characteristics (Yj = j), the difference between this observation and one without a job is b2. Compared with average workers, a one-unit increase in work hours or commuting time further contributes to b3j amount of change in marriage intention. The original equation can also be written as M = b0 + b1iXi + b2E + b3j(Yj – j)E, where b3j(Y – j)E = 0 when an observation is jobless. By coding mean-centered job characteristics (i.e., [Yj – j]) as 0 for the jobless, we can further modify the equation to be M = b0 + b1iXi + b2E + b3j(Yj – j), as b3j(Yj – j) would be 0 when one is jobless. In this study, we use a set of dummy variables for employment status, rather than just having a job or not, but the employment status dummy variables serve the same function as E in the illustrated equation.
Specifically, we used father’s education, a combination of father’s and mother’s education, or father’s occupation when respondents were 15 years old to approximate the economic advantages or disadvantages of respondents’ family of origin. For sibship characteristics, we tried sibship size, birth order rank, and whether respondents were the first or only son or daughter in the family.
We also considered the possibility that engaging in formal and informal partner-seeking activities has different implications for marriage transitions for those with and without romantic partners. An additional analysis including interaction terms between involvement in partner-seeking activities and relationship status, however, showed no significant interaction effects.
We contend that the comfort of the parental home, if it decreases one’s incentive to marry, as expected by the theory, should make young adults express a lower level of eagerness to marry. One may argue, however, that singles living with parents simply want to postpone marriage but still aspire to marry eventually. Because the question from Wave 1 of the JLPS asked respondents the age by which they expected to marry, we were able to conduct an ancillary analysis examining the association between coresidence and anticipated timing of entering first marriage with data from only that wave. We found no significant associations for either gender group. Thus, our results do not support the comfort-of-home hypothesis, regardless of how the hypothesis is conceptualized.
The results remained statistically nonsignificant when we compared the rates of home-leaving in the year of becoming engaged with those in any other year.
Because a fixed-effects modeling approach compares time-varying individual characteristics to the individual mean, rather than using characteristics of an earlier point to predict the current outcome, using fixed effects to control for unobserved heterogeneity is not possible for this part of the analysis.
Japanese men are marginally more likely to think they lack funds for marriage when they live with parents, but this association suggests more about men’s decreased concern about the funds needed to marry when starting to live independently than their worries about lower living standards after marriage.