|Home | About | Journals | Submit | Contact Us | Français|
This article documents the prevalence, duration, and marital outcomes of cohabiting unions in Japan. It then examines the correlates of cohabitation experiences and also describes differences in the family-formation trajectories of women who have and have not cohabited. Cohabitation has increased rapidly among recent cohorts of women, and cohabiting unions in Japan tend to be relatively short in duration and are almost as likely to dissolve as to result in marriage. Life table analyses demonstrate that the cumulative probabilities of marriage and parenthood within marriage are roughly similar for women who did and those who did not cohabit. The most notable difference is in the pathways to family formation, with women who cohabited more likely both to marry subsequent to pregnancy and to delay childbearing within marriage. Taken as a whole, these results suggest that cohabiting unions in Japan are best viewed as an emerging prelude to marriage rather than as an alternative to marriage or singlehood. We conclude with speculation about the likelihood of further increases in cohabitation in Japan and the potential implications for marriage and fertility.
The prevalence of nonmarital cohabitation has increased rapidly in most low-fertility societies (Kiernan 2004), but cross-national studies demonstrate that the nature of cohabiting unions depends on context. Substantial regional variation in the prevalence of cohabiting unions, their duration, and their relationships to marriage and childbearing (e.g., Heuveline and Timberlake 2004; Kiernan 2001) has been linked to variation in policies, economic circumstances, and normative environments (Lesthaeghe 1995). Comparative studies have also provided important insights into the role that cohabitation plays in the family-formation process, suggesting that cohabiting unions tend to function as an alternative to marriage in some societies but as a precursor to marriage in other societies (Heuveline and Timberlake 2004).
Japan has not been included in international comparative studies of cohabitation, although it is likely a rich source of insight with respect to the emergence of cohabiting unions and their role in the family-formation process. As one of the few non-Western countries characterized by very late marriage and lowest-low fertility, Japan provides an excellent opportunity to evaluate the generality of increasing nonmarital cohabitation. Until recently, available evidence suggesting that nonmarital cohabitation remains uncommon in Japan (Atoh 2001b) has presented an important challenge to the generality of a “package” of family changes associated with the second demographic transition (Lesthaeghe 1995). However, more recent studies demonstrating substantial increases in the prevalence of cohabitation in Japan (Iwasawa 2005; Tsuya 2006) provide seemingly powerful evidence for the general importance of nonmarital cohabitation in low-fertility, late-marriage societies.
Japan also presents a valuable opportunity to examine how the role of cohabitation may depend on linkages between marriage and childbearing. Unlike many low-fertility societies in which nonmarital childbearing is common and often occurs within cohabiting unions, the proportion of children born to unmarried mothers in Japan is negligible. Furthermore, limited socioeconomic differences in earlier family change (Hodge and Ogawa 1991; Raymo 2003) and the relative homogeneity of the Japanese family life course (Brinton 1992) provide an important opportunity to evaluate the generality of observed sub population differences in the role of cohabitation in the United States and elsewhere (Manning and Landale 1996; Raley 1996).
Our objective in this article is to advance our understanding of premarital cohabitation in Japan within the limitations of newly available, nationally representative data. We begin by describing the prevalence of cohabitation experience, the mean duration of cohabiting unions, and the proportion of cohabiting unions that end in marriage. This basic information will allow for initial comparison of cohabiting unions in Japan with those in other countries and provide a basis for further theorization regarding contextual influences on the prevalence and characteristics of nonmarital unions in late-marriage, low-fertility societies.
We then estimate multivariate models for cohabitation experience, duration, and outcomes as a function of established covariates of cohabitation experience in the United States and elsewhere. Like cross-national studies, analyses of within-country heterogeneity have provided valuable insights into the meaning of cohabitation. In the United States, racial, ethnic, and socioeconomic differences in the prevalence and nature of cohabiting unions suggest the importance of cultural differences (e.g., Manning and Landale 1996) and highlight the role of economic stability as a prerequisite for marriage (Smock, Manning, and Porter 2005). Do we observe similar patterns in Japan, where socioeconomic differences in the family life course have been relatively limited but economic inequality among young adults has increased in recent years (Ohtake 2005)?
Finally, we describe and compare the marriage and marital childbearing trajectories of women who have and those who have not cohabited. Studies exploring the relationships between cohabitation, marriage, and childbearing in the United States have demonstrated that cohabitation is associated with a higher likelihood of premarital conception (and subsequent marriage) and that an increasing proportion of conceptions and childbirths occur within cohabiting unions (Kennedy and Bumpass 2008; Wu, Bumpass, and Musick 2001). Relationships between cohabitation and marriage timing may be particularly important in Japan, where the trend toward later marriage, combined with negligible levels of nonmarital childbearing, has contributed to very low period fertility rates (Tsuya and Mason 1995). Recent increases in the proportion of marriages preceded by pregnancy (Raymo and Iwasawa 2008) also suggest that nonmarital conceptions may be an important mechanism linking cohabitation to marriage and fertility, particularly for women at the lower end of the socioeconomic spectrum.
The prevalence, duration, and marital outcomes of cohabiting unions vary substantially across countries. The prevalence of cohabitation experience is particularly variable, ranging from more than 80% in France to less than 10% in Italy.1 Median duration of cohabiting unions ranges from more than four years in France to just over one year in the United States; the proportion resulting in marriage exceeds 75% in several countries (e.g., Austria, Belgium, and Finland) but is less than one-half in others (e.g., France and New Zealand).2 This cross-national variation in the prevalence and nature of cohabiting unions is the basis for Heuveline and Timberlake’s (2004) classification of countries according to the role of cohabitation in the family-formation process. They characterized cohabitation as an alternative (or equivalent) to marriage in societies such as France and Sweden, where prevalence is high, duration is long, and the proportion resulting in marriage is relatively low. In countries like Austria and Finland, where prevalence is high but average duration is short and a large proportion of unions result in marriage, they argued that cohabitation is better described as a precursor to marriage or a trial marriage. In other countries such as Italy, Poland, and Spain, the role of cohabitation is described as “marginal,” with normative sanctions and institutional penalties combining to limit cohabiting unions to a relatively small minority of couples (Heuveline and Timberlake 2004:1216). These broad classifications are informative and important, but substantial within-country heterogeneity in the nature of cohabiting unions makes the process of classifying countries into a single category difficult and somewhat arbitrary. In the United States, for example, most studies of cohabitation recognize that nonmarital unions are heterogeneous in nature, with many serving as a precursor to marriage; others are an alternative to singlehood or dating, and some are an alternative to marriage (Casper and Bianchi 2002; Heuveline and Timberlake 2004).
Cross-national comparisons of cohabiting unions have not included Japan because the necessary data have not been available. Until recently, data on current and past cohabitation experiences were collected only from unmarried respondents of the Japanese National Fertility Surveys. These data indicate that the prevalence of cohabitation at any point in time has been very low, ranging from less than 1% of respondents in 1987 to about 2% in 2005. Similarly, relatively low proportions of unmarried men and women report having ever cohabited, ranging from 3% in 1987 to 8% in 2005 (Kaneko et al. 2008). The actual proportion that has cohabited is obviously higher, but without data from married respondents, it has been impossible to ascertain how much higher. Nevertheless, these incomplete data suggest that cohabitation plays a very limited role in the family-formation process in Japan. This is somewhat unexpected in light of the growing gap between earlier initiation of sexual activity (Japanese Association for Sex Education 2000) and later marriage (Raymo 2003; Retherford, Ogawa, and Matsukura 2001), increasing marital instability (Raymo, Iwasawa, and Bumpass 2004), changing attitudes about family and gender relations (Atoh 2001a), and widespread skepticism about the benefits of marriage (Tsuya, Mason, and Bumpass 2004). Japanese scholars have suggested several possible explanations for the relatively low prevalence of cohabiting unions, including the highly asymmetric gender division of household labor; limited evidence of an increase in individualistic attitudes; relatively limited opportunities for meeting potential partners; the high cost of establishing a temporary residence; and minimal financial, emotional, and normative incentives to leave the parental home before marriage (Atoh 2001b; Miyamoto, Iwakami, and Yamada 1997).
However, recent survey data that include married respondents indicate that the prevalence of cohabitation in Japan is substantially higher than previously believed. For example, one in five 25- to 34-year-old female respondents of the first rounds of the Japanese Generations and Gender Survey and the National Survey on Population, Family, and Generations in Japan reported that they had lived in a cohabiting union (Iwasawa 2005; Tsuya 2006).3 This figure is twice as high as the proportion reported by respondents just 10 years older, despite the longer exposure time for the latter. This rapid increase in cohabitation represents a potentially valuable source of insights relevant to discussions of the second demographic transition—a widely cited organizational framework for describing and understanding a “package” of family changes (including the emergence of nonmarital cohabiting unions) that has accompanied extended periods of below-replacement fertility in many industrialized countries (e.g., Lesthaeghe 1995; van de Kaa 1987).
For example, the rapid emergence of cohabitation in Japan, in conjunction with long-term declines in marriage and fertility and more recent increases in divorce (Raymo et al. 2004), suggests that the notion of the second demographic transition may be more broadly generalizable than posited by some who have argued that the associated package of family changes is not likely to be observed outside northwestern European and English-speaking countries (e.g., Coleman 2004). Distinctive features of the Japanese social context make this question of generalizability particularly important. In contrast to most low-fertility Western societies, Japan has yet to experience substantial change in some of the key mechanisms thought to underlie family changes associated with the second demographic transition, including increasing emphasis on self-fulfillment or individuation, growth in women’s economic independence, and normative acceptance of nonmarital childbearing (Atoh 2001a, 2001b; Atoh, Kandiah, and Ivanov 2004). In this context, should evidence of family change, including the emergence of premarital cohabitation, be considered evidence of a second demographic transition in Japan? Should evidence of rapid family change in Japan be discussed and interpreted with reference to frameworks other than the second demographic transition? If so, which frameworks are most relevant? Does the Japanese experience suggest a need to rethink or extend the social, economic, and ideational changes thought to underlie the package of family change associated with the second demographic transition? These are questions with both theoretical relevance for assessing the generality of the second demographic transition as a framework for understanding family change and practical relevance for understanding the implications of family change in non-Western, low-fertility societies where the link between marriage and fertility remains strong. Careful documentation of trends in cohabitation, the characteristics of cohabiting unions, and the role of cohabitation in the family-formation process in Japan is an essential first step in addressing these questions.
Like cross-national comparisons, studies of within-country variation in the prevalence and nature of cohabiting unions have provided important insights into the role of cohabitation in the family-formation process. Several studies of cohabiting unions in the United States have shown that prevalence is higher among racial minorities; those with lower levels of education, less-favorable economic circumstances, and lower social class backgrounds; those who are less religious; and those who have more liberal attitudes toward marriage and family (see Seltzer  and Smock  for reviews). Other studies have shown that cohabiting unions are longer in duration and less likely to result in marriage for blacks (Manning and Smock 1995) and for couples with less-stable economic circumstances (Smock and Manning 1997). Together, these results have been interpreted as evidence that cohabiting unions are more likely to function as alternatives to marriage among blacks and those with more limited economic resources but as precursors to marriage or trial marriages among whites and those with more favorable economic prospects. Because the large majority of Americans express a desire to marry eventually, these findings suggest that cohabitation may play a role in the differential realization of family goals. Furthermore, because cohabiting unions are less stable than marriage (Manning, Smock, and Majumdar 2004), racial and socioeconomic differences in the nature of cohabiting unions may have implications for variation in the well-being of women and children (Avellar and Smock 2005; Raley, Frisco, and Wildsmith 2005).
In one of the only studies to examine the correlates of cohabitation in Japan, Tsuya (2006) found that, as in the United States, cohabitation is more common among those with less education. Significant socioeconomic differences in the prevalence and nature of cohabiting unions, in conjunction with evidence of increasing differences in other family behaviors, including marriage preceded by pregnancy (bridal pregnancy; Raymo and Iwasawa 2008), divorce (Raymo et al. 2004), and married women’s labor-force participation (Kohara 2001) would represent an important departure from the homogeneity that characterized Japanese family behavior in the second half of the twentieth century (e.g., Brinton 1992). This pattern is also potentially consistent with a more general emphasis on growing socioeconomic differences in family behavior in industrialized countries (McLanahan 2004). Attitudes are another established correlate of cohabitation in the United States and elsewhere, but similar relationships have yet to be examined in Japan. Evidence of substantial change and variation in family-related attitudes (Atoh 2001a) suggest that more liberal attitudes toward marriage, childbearing, and family behavior more generally may be associated with a higher likelihood of unmarried cohabitation in Japan as well.
Efforts to understand the role of cohabitation and how it may differ across countries or population subgroups have focused on linkages between cohabitation and marriage and childbearing. Studies in the United States have found an association between cohabitation and nonmarital conception (Brien, Lillard, and Waite 1999; Manning and Landale 1996) and demonstrated that first births are increasingly likely to occur within cohabiting unions (Carlson, McLanahan, and England 2004; Kennedy and Bumpass 2008). The fact that nonmarital conceptions are less likely to lead to marriage among black and Hispanic women suggests that cohabitation more closely resembles an alternative to marriage among racial and ethnic minorities (Manning 2001; Manning and Landale 1996). Similar differences with respect to economic well-being have been interpreted as further evidence that the role of cohabitation as an alternative to marriage is more pronounced among those in less stable economic circumstances, whereas nonmarital unions more closely resemble a stage in the marriage process for those with higher socioeconomic status (e.g., Loomis and Landale 1994).
In Japan, the implications of delayed marriage and very low fertility for population aging and population decline are the subject of great social and political concern and have motivated several high-profile policy efforts to promote marriage and increase fertility rates. It is thus surprising that little scholarly or political attention has been devoted to premarital cohabitation and its potential impact on marriage and fertility. In the only previous study to address these relationships, Tsuya (2006) found that cohabitation is associated with higher rates of marriage and bridal pregnancy. In this article, we extend Tsuya’s (2006) research by using life table methods to simultaneously describe differences in marriage and childbearing patterns with respect to cohabitation experience and to examine how these relationships differ by educational attainment.
Research on the role of cohabitation in the family-formation process in the United States and elsewhere, in conjunction with recent work on marriage and childbearing in Japan, suggests several ways in which family-formation trajectories may differ by cohabitation experience. Of particular importance is the strong relationship between marriage and childbearing in Japan. Recent vital statistics data indicate that only 2% of births are registered to unmarried mothers, while approximately one in four first marriages is preceded by pregnancy (Raymo and Iwasawa 2008). It may be that cohabitation hastens marriage by increasing the risk of pregnancy via more-frequent sexual activity (Bachrach 1987) or less-vigilant contraception. Alternatively, it may be that unanticipated pregnancy increases the likelihood of forming a cohabiting union prior to marriage and childbirth. The concentration of bridal pregnancy at the lower end of the educational spectrum (Raymo and Iwasawa 2008) suggests that early marriage via cohabitation and premarital pregnancy should be more common among these women.
At the other end of the socioeconomic spectrum, distaste for the highly asymmetric gender division of labor and the associated opportunity costs of marriage have been linked to later marriage among women with higher education and higher earnings (Raymo 2003; Raymo and Ono 2007). This suggests that cohabitation may be associated with later marriage and fertility among these women to the extent that nonmarital unions are perceived as relatively egalitarian arrangements (South and Spitze 1994) that offer many of the benefits of marriage (e.g., coresidential intimacy, sexual access, and economies of scale) without either the normative assumptions of long-term emotional and financial commitment or the legal barriers to dissolution that accompany marriage.
Documenting differences in the family-formation patterns of those who have and those who have not cohabited is straightforward, but it is very difficult to evaluate causal effects of cohabitation on marriage and childbearing. This difficulty reflects both the complexity of the family-formation process and the fact that men and women who form cohabiting unions differ systematically from those who do not with respect to characteristics such as religiosity, family attitudes, and relationship skills that are not typically ascertained in surveys (e.g., Clarkberg, Stolzenberg, and Waite 1995). Recognizing that cohabitation, marriage, and childbearing are interrelated components of an increasingly complex and heterogeneous family-formation process in the United States and Europe, recent studies have modeled the three processes simultaneously (Baizán, Aassve, and Billari 2003, 2004; Brien et al. 1999; Musick 2006). These studies have found that observed relationships between cohabitation, marriage, and childbearing partly reflect unobserved factors, with those who are more likely to cohabit also more likely to bear children and to marry. Data limitations prevent us from adopting a similar approach, but we evaluate results with these possibilities in mind.
We use data from the 1st National Survey on Population, Family, and Generations in Japan (SPFG), a nationally representative survey of 2,421 women aged 20–49. The SPFG was conducted in April 2004 by the Mainichi Shimbun Population Problems Research Council and had a response rate of 61% (Mainichi Shimbun Population Problems Research Council 2005). The survey contained three questions about cohabitation experience. The first asked respondents whether they had ever cohabited, with response options of “no,” “yes, in the past,” and “yes, currently cohabiting.” The second asked the length of the most recent (or ongoing) cohabiting union, with 10 categorical response options ranging from “less than one month” to “five years or more.” The third asked respondents whether they married their most recent cohabiting partner or intend to marry their current partner.
Using these data, we first construct measures of prevalence, duration, and outcomes similar to those presented by Heuveline and Timberlake (2004). The percentage ever in a cohabiting union refers to both current and past unions, mean duration is a weighted average of the midpoints of the 10 duration categories, and the percentage ending in marriage refers to completed unions.4 We provide tabulations of each of these characteristics by respondents’ birth cohort to describe change over time. This basic information extends our empirical understanding of cohabitation in Japan and allows for comparison with other industrialized countries.
We then examine the correlates of cohabitation by estimating logistic regression models for cohabitation experience and cohabitation ending in marriage and a linear regression model for the duration of cohabiting unions.5 In these models, we use available information to measure sociodemographic and attitudinal correlates identified in studies of cohabitation in the United States and elsewhere. We describe change over time using a six-category measure of birth cohort (1954–1959, 1960–1964, 1965–1969, 1970–1974, 1975–1979, 1980–1984), and evaluate socioeconomic differences with a four-category measure of respondent’s education (high school or less, vocational school, junior college, university) and a dichotomous indicator of nonstandard employment (part-time employment and self-employment) in the first job after completing school. Differences with respect to socioeconomic background are evaluated by using a measure of father’s education (years of schooling plus a dichotomous indicator of missing values). Research on the United States suggests that cohabiting unions should be more common, longer, and less likely to result in marriage among women with lower education, nonstandard employment, and less-educated fathers.
The SPFG included several attitudinal questions, and we include four of these in our models. We measure family attitudes with responses to questions about the degree to which respondents (dis)approve of (a) married couples having different last names and (b) same-sex couples.6 We measure attitudes toward marriage and childbearing with responses to questions about the degree to which respondents (dis)agree that (a) couples should marry if they get pregnant and (b) married couples should have children. For each of these questions, we treat responses as categorical. Related research on cohabitation in the United States suggests that women with more-liberal family attitudes should be more likely to cohabit (Clarkberg et al. 1995), have longer unions, and be less likely to marry.7
Our ability to estimate relationships between cohabitation experience and marriage timing and subsequent parenthood is limited by the fact that the SPFG did not collect information on the beginning and end dates of cohabiting unions. Without this information, we are unable to model transitions into cohabiting unions and first marriages as competing risks or to treat cohabitation as a time-varying covariate in models of first-marriage rates. In the absence of information required to examine entry into cohabiting unions, first marriage, and parenthood simultaneously, we treat cohabitation experience as a fixed characteristic and use life table methods to describe the family-formation trajectories of women who did and women who did not live in a cohabiting union. Because it is not possible to link cohabiting unions to a specific marriage, we limit the sample to the 2,211 women who were never married or were in their first marriage. Excluding cases with missing data on the year and month of first marriage or first birth (n = 81) and educational attainment (n = 7) as well as those who reported marrying before age 16 (n = 5) or having a premarital birth (n = 57) leaves us with an analytic sample of 2,061 respondents.8
Because our sample size is small (especially for women who have cohabited), we begin by estimating event-history models to generate predicted probabilities of first marriage and first marital birth. We do this by expanding the cross-sectional data into person-month form and estimating two sets of discrete-time hazard models via logistic regression. We first estimate the following competing-risks model for first marriage separately for women in each of the four educational categories:
where pijt is the probability that the ith woman experiences first marriage type j (j = marriage prior to pregnancy, marriage subsequent to pregnancy) at time t, and pi0t is the probability of remaining never married. Exposure begins in the month of the respondent’s 16th birthday, and censoring occurs in the month of first marriage or the survey date. We define bridal pregnancy as cases in which the reported month of first birth occurs eight or fewer months after the reported month of marriage. Based on the results of preliminary analyses, duration dependence is measured as a linear spline (DUR) with knots at ages 26 and 35.9 COHAB is the dichotomous indicator of cohabitation experience, ENROLL is a time-varying indicator of school enrollment included to control for the rarity of marriage while still enrolled in school, and COHORT is a continuous measure of birth cohort.10 We use the estimated coefficients from these models to calculate age-specific probabilities of marriage prior to pregnancy and marriage subsequent to conception, by cohabitation experience and educational attainment.11 We then use these predicted probabilities to construct a total of eight first-marriage tables (i.e., 2 categories of cohabitation experience × 4 categories of educational attainment).
We add information on postmarital transition to parenthood in two steps. For women who married subsequent to pregnancy, we assume that first birth occurred at the same age as marriage. For all other women, we estimate a separate set of discrete-time hazard models for first birth following marriage. In these models, exposure begins in the eighth month after marriage and censoring occurs at first birth or the survey date. As in the marriage models, duration-dependence is specified as a linear spline. Based on preliminary analyses, we use a spline with a single knot at four years after first marriage. These models also include the dichotomous indicator of cohabitation experience, control for age at marriage and birth cohort, and are estimated separately for women in each of the four educational categories. Estimated coefficients from these models allow us to generate predicted marriage duration–specific, first-birth probabilities by age at marriage, cohabitation experience, and educational attainment. We then apply these predicted probabilities to the predicted number of age-specific marriages from the marriage tables described earlier to generate the predicted numbers of first births by duration since marriage, age at marriage, cohabitation experience, and educational attainment. We present results as age-specific synthetic cohort proportions to experience bridal pregnancy, first marriage and first birth, and first marriage but not first birth.
In Table 1, we describe the prevalence, duration, and outcomes of cohabiting unions reported in the SPFG. The first row shows that 15% of the sample ever cohabited, that the mean duration of cohabiting unions was slightly less than two years, and that slightly more than one-half (58%) of completed unions resulted in marriage.12 Tabulations by birth cohort show a substantial increase in cohabitation experience among women born after 1965. One in five women born in the 1970s (i.e., women aged 25–34 at the time of the survey) reported ever living in a cohabiting union.13 This figure is far lower than in the United States (Kennedy and Bumpass 2008) and most European countries (Kiernan 2001) but represents substantial change in the family-formation process in Japan. The low prevalence of cohabitation in the youngest cohort (born in 1980–1984) reflects the fact that these women were only aged 20–24 at the time of the survey. Iwasawa (2005) found that, controlling for exposure, the likelihood of cohabitation is highest among the youngest cohorts in this survey.
Mean duration is highest for women born in the late 1960s, and the apparent decline in the length of cohabiting unions for cohorts born after 1970 reflects the relatively limited opportunity to experience cohabiting unions of five or more years.14 The sharp decline in the proportion of completed unions resulting in marriage for cohorts born after 1975 likely reflects two different processes. The first is a matter of timing, with some in these younger cohorts reporting that they intend to marry, but have not yet married, their former cohabiting partner. It is also possible that there has been a real decline in the likelihood that cohabiting unions result in marriage.
In sum, the figures in Table 1 indicate that cohabiting unions in Japan are less common and shorter in duration but have a similar likelihood of resulting in marriage when compared with other countries included in recent cross-national studies. The classification scheme proposed by Heuveline and Timberlake (2004) suggests that cohabitation was a marginal experience for earlier cohorts of Japanese women but can be described as a prelude to marriage for women born after 1965. Given the high cost of establishing a temporary residence and the prevalence of extended coresidence with parents in Japan (Nishi and Kan 2007), this change provides powerful evidence of the widespread emergence of nonmarital cohabiting unions in low-fertility, industrialized countries.
In Table 2, we describe relationships between experience, duration, and results of cohabiting unions and the sociodemographic and attitudinal characteristics described earlier. Along with the distribution of covariates in column 1, we present exponentiated coefficients (i.e., odds ratios) from a logistic regression model of cohabitation experience (column 2), coefficients from a linear regression model of cohabitation duration (column 3), and odds ratios from a logistic regression model of cohabitation resulting in marriage (column 4).
Looking first at the likelihood of ever cohabiting (column 2), odds ratios for birth cohort describe the increase in cohabitation over time already presented in Table 1. Like previous studies (Iwasawa 2005; Tsuya 2006), we find a strong relationship between educational attainment and cohabitation experience. There was no significant difference between those who completed high school or less and vocational school graduates, but the odds of cohabitation for women with at least a two-year college degree were less than half as high as for those in the lowest education category. The large (and statistically significant) difference between vocational school graduates (odds ratio = 0.77) and junior college graduates (odds ratio = 0.43) also suggests that the common practice of collapsing these two groups into a single educational category may not be appropriate in studies of cohabitation.
The likelihood of cohabiting was positively related to nonstandard employment following school completion but was not significantly related to father’s education. These differences with respect to both educational attainment and nonstandard employment are similar to patterns observed in the United States and suggest that cohabitation in Japan may be emerging as an alternative type of union for young couples who do not yet have the economic resources and stability that are normatively necessary for marriage. Attitudinal variables were also related to cohabitation experience, with more liberal attitudes associated with a higher likelihood of having lived in a cohabiting union.15 For example, women who were not opposed to spouses having different last names were much more likely to have cohabited relative to those who were strongly opposed. Similarly, the likelihood of cohabitation experience was higher for those not at all opposed to same-sex couples and those who disagreed that couples should marry if they become pregnant. Contrary to our expectations, the odds of cohabitation were lower among those who somewhat disagreed that married couples should have children.
The coefficients in column 3 indicate that among those who have cohabited, union duration was longer for daughters of more highly educated men and longer for women with more-liberal attitudes toward different last names for spouses and marriage in response to pregnancy. Interestingly, cohabiting unions were shorter, on average, for the small proportion of women who strongly disagreed that married couples should have children. The results in column 4 indicate that cohabiting unions were less likely to result in marriage for women with more-educated fathers and more-liberal attitudes toward family arrangements and marriage in response to pregnancy. One exception was the relatively high likelihood that cohabiting unions result in marriage for those with neutral attitudes toward the importance of children for married couples.
Figures 1 and and22 present results from the life table analyses of first marriage and first childbirth. In Figure 1, we present the synthetic cohort proportions experiencing bridal pregnancy, both marriage and parenthood, and marriage but not parenthood by ages 25, 30, and 40.16 These figures are based on the full sample. In Figure 2, we present cumulative probabilities of experiencing each outcome by age 30 for women in each of the four educational categories.
Figure 1 shows that the cumulative probabilities of marriage and first birth were very similar for women with and without cohabitation experience. The proportion married was 3–5 percentage points higher for women who cohabited, and the proportion with a first birth was 1–2 percentage points higher. The only notable differences are that women who cohabited were more likely to experience bridal pregnancy and also somewhat more likely to marry but not yet have a child. This suggests potential heterogeneity in the nature and role of cohabiting unions. To further explore this possibility, we compare results across educational categories.
Figure 2 shows that the proportions married by age 30 are, in most cases, somewhat higher (3–13 percentage points) for women who reported cohabiting. The exception is vocational school graduates, for whom the cumulative probability of marriage is 3 percentage points lower for those who cohabited. For those who completed high school or less and for junior college graduates, similarity in the proportion married by age 30 masks differences in the pathway to marriage, with those who cohabited more likely to marry while pregnant. This is especially true for high school graduates: 20% of those who cohabited had experienced bridal pregnancy by age 30, compared with 11% of those who did not cohabit.
For all educational groups, women who cohabited were more likely to marry but remain childless by age 30. This is particularly true among vocational school graduates, for whom similarity in the proportion married reflects very different patterns of family formation through age 30. Among women who cohabited, the proportion of parents (i.e., bridal pregnancy plus marriage and child) was substantially lower (.44 versus .61 among those who did not cohabit), and the proportion married but without children was twice as high (.27 versus .13). To some extent, the relatively high proportion married but without children reflects later marriage among cohabitors. The proportion of vocational school graduates married by age 25 (not shown) was .25 for those who did cohabit and .31 for those who did not. Among junior college and university graduates, the higher proportion married among those who had cohabited primarily reflects the higher likelihood of being married but without children. For university graduates, cohabitors were slightly more likely to experience each of the three family-formation pathways, and the difference in the cumulative probability of marriage by cohabitation experience was largest for this group (13 percentage points).
These patterns continue through age 40 (results not shown), with the cumulative probability of marriage higher among those who had cohabited for all educational groups and the largest difference (10 percentage points) observed among university graduates. Differences in the cumulative probability of first birth are generally smaller than those for marriage. For high school and junior college graduates, the most notable difference is the higher likelihood of bridal pregnancy among those who cohabited. For vocational school and university graduates, the most notable difference is the relatively high proportion married but not yet having children among women who cohabited.
In contrast to the rapidly growing body of research on cohabitation in the United States (Seltzer 2000; Smock 2000) and Europe (Kiernan 2002), nonmarital unions have received very little attention in studies of family formation in Japan. This likely reflects the relatively low prevalence of cohabiting unions as well as the lack of suitable data with which to construct even the most basic demographic measures of cohabitation. In the absence of adequate information, it has been commonly assumed that cohabitation is of limited importance in Japan. This is an important limitation given policy emphases on promoting marriage and fertility in Japan as well as the potential relevance of the Japanese experience for evaluating the generality of findings from research on cohabitation in other low-fertility Western societies.
This article begins to fill this gap in our understanding of the family-formation process in Japan. Using one of the very few nationally representative sources of data on cohabitation, we have been able to examine trends in cohabitation and to provide evidence on the duration of cohabitation, the extent to which cohabiting partners marry each other, and linkages between cohabitation and family formation. We demonstrated that cohabitation has increased rapidly among recent cohorts of women, and that cohabiting unions in Japan tend to be relatively short in duration and are almost as likely to dissolve as to end in marriage.
We have also shown that the correlates of cohabitation in Japan resemble those in the United States, with lower educational attainment and more-liberal family attitudes both strongly associated with cohabitation experience. Educational differences suggest that, as in the United States, cohabitation may serve as an alternative to marriage for couples without the economic resources deemed necessary for marriage. Unlike the United States, however, the strong link between marriage and childbearing results in a relatively high proportion of bridal pregnancies rather than nonmarital births among less-educated Japanese women in cohabiting unions. The fact that highly educated women are less likely to cohabit suggests that the perceived opportunity costs of nonmarital unions are high, either because they are seen as precursors to marriage or as relationships that are similar to marriage in terms of expectations regarding the division of household labor. The strong relationships between liberal family attitudes and the likelihood of cohabitation suggest the relevance of linkages between values and family behavior emphasized in discussions of the second demographic transition.
Our life table analyses indicated that family-formation trajectories do not differ substantially with respect to cohabitation experience. Cohabitors married and became parents somewhat earlier than those who did not cohabit, but the magnitude of these differences was small. The most notable difference was in the pathways to family formation, with women who cohabited more likely to experience bridal pregnancy and more likely to delay childbearing within marriage. These two different patterns of family formation appear to be related to education, with the former most common among women at the lower end of the educational spectrum and the latter more pronounced among vocational school and university graduates.
Taken as a whole, these results suggest that cohabiting unions in Japan may be best viewed as a prelude to marriage rather than as an alternative to marriage or singlehood. It is important to keep in mind, however, that our analyses have not accounted for selection into cohabitation. To the extent that cohabitation is associated with unobserved characteristics that are also associated with the pace of marriage and childbearing, our life table analyses may obscure important influences of cohabitation on the family-formation process. For example, if cohabitors are selectively composed of women who are more likely to postpone marriage and childbearing for unobserved reasons, our results will understate the extent to which cohabitating is associated with earlier marriage and childbearing. Similarly, if cohabitors are selectively composed of women who are more likely to marry and have children early, as appears to be the case in the United States and other low-fertility societies (e.g., Baizán et al. 2003, 2004; Brien et al. 1999; Musick 2006), our results will overstate the extent to which cohabitation is associated with earlier family formation. It is important to keep this issue in mind when interpreting results for a country such as Japan, where cohabitation has increased but can still be described as a nonnormative or innovative family arrangement.
In addition to the empirical regularity with which cohabitation has spread in other late-marriage, low-fertility societies, there are several reasons to expect further increase in the prevalence of cohabiting unions in Japan. The first is the powerful role of social networks and diffusion processes. Based on evidence that half of Japanese adults know someone who has cohabited and that disapproval of cohabitation appears to be weaker than previously believed (Rindfuss et al. 2004), we suspect that normative barriers to cohabitation are waning. The SPFG provides more direct evidence of the potential for further increases in cohabitation. When asked whether they would like to cohabit in the future, less than half of never-married respondents said no. One-third said yes, and another one-fourth were unsure.
There are many limitations to this study that we hope can be addressed in subsequent research as the necessary data become available. The most important is the limited information collected on cohabitation in the SPFG, especially the absence of beginning and ending dates of all nonmarital unions. As a result, we were unable to model cohabitation and marriage as competing risks or to treat cohabitation as a time-varying covariate in models of first-marriage timing. By treating cohabitation experience as a fixed characteristic and using life table methods to compare the family-formation trajectories of women who have and have not cohabited, we were able to work around these data problems but obviously have not resolved them. Without information on the beginning and ending dates of all cohabiting unions, it was impossible to determine the temporal ordering of pregnancy and union entry. Furthermore, because we have information only on respondents’ current marriage, we could not examine the relationship between cohabitation and first-marriage timing for women who have separated from their first spouse. Finally, the absence of information on aborted pregnancies prevented an examination of the extent to which women in cohabiting unions are more likely than their unpartnered counterparts to legitimate premarital pregnancies. The collection of these missing but essential data should be a priority. Another important subject for future study is to extend research from the United States and elsewhere on relationships between cohabitation experience and other marital outcomes, such as divorce. Divorce in Japan has increased substantially in recent years (Raymo et al. 2004), and the role of cohabitation (if any) in this trend remains unexplored. We are not aware of any source of currently available data appropriate for this kind of analysis.
We conclude by emphasizing that the results in this article, along with evidence from related research on divorce (Raymo et al. 2004) and bridal pregnancy (Raymo and Iwasawa 2008), highlight not only that conditions are “ripe for substantial change in family behavior in Japan” (Rindfuss et al. 2004:854) but also that substantial change is already underway. By taking advantage of the distinctive social, political, and economic context in which these changes are taking place, research on ongoing family change in Japan may provide a rich source of insight into the generality of patterns observed in the United States and other Western societies, as well as the ways in which local contexts shape the pace and nature of family change.
We would like to express our appreciation to the Population Problems Research Council of the Mainichi Shimbun for providing the data.
The research reported herein was supported by grants from the National Institute of Child Health and Human Development (R03-HD049567) and the Graduate School of the University of Wisconsin–Madison. Additional support was provided by the Institute of Social and Economic Research at Osaka University. Computation was carried out using facilities of the Center for Demography and Ecology at the University of Wisconsin–Madison, which are supported by Center Grants from the National Institute of Child Health and Human Development and the National Institute on Aging. Earlier versions of this article were presented at meetings of the Population Association of America, the Population Association of Japan, and the International Union for the Scientific Study of Population.
1.These figures refer to the life table proportion of women who have ever been in a cohabiting union by age 45 (Heuveline and Timberlake 2004)
2.Median duration is the life table estimate of the duration at which 50% of cohabiting unions end in either marriage or dissolution (Heuveline and Timberlake 2004).
3.Both surveys were conducted in 2004.
4.For the open-ended duration category (five or more years), we arbitrarily assign a value of seven years to calculate mean union duration. Because only 7% of women with cohabitation experience reported unions of five or more years, results change little when we use alternative assumptions, such as 5 or 10 years.
5.The number of observations lost as a result of nonresponse to the cohabitation questions was 23 in the model for cohabitation experience, 30 in the model for cohabitation duration, and 31 in the model for marital outcomes.
6.Japanese law requires spouses to have the same last name; in the vast majority of cases, women take their husband’s surname. The pros and cons of revising this law have received a good deal of political and media attention in recent years.
7.Of course, we recognize that the experience of cohabitation may also influence family-related attitudes, but the cross-sectional nature of the available data does not allow us say anything about the direction of relationships between cohabitation experience and attitudes.
8.The proportion of nonmarital births is higher for women who report ever cohabiting (.05) than for those who do not (.02), suggesting that cohabitation may be an emerging context for childbearing (or childrearing). Unfortunately, the small number of cases prevents us from examining change over time in the relationship between cohabitation and premarital childbearing or treating premarital births as a separate outcome in our analyses. These relationships should be examined when appropriate data become available.
9.For university graduates, the first knot in the duration spline is placed at 27 years old, the age at which observed marriage rates peak.
10.In the absence of enrollment history data, we estimate the month of school completion by utilizing the formulaic determination of age at school entry and assuming no grade skipping or grade retention (Raymo 2003).
11.We evaluate these probabilities for the 1965 birth cohort.
12.The proportion of completed cohabiting unions resulting in marriage will be biased upward to the extent that respondents have experienced multiple cohabiting unions. We suspect that serial cohabitation is uncommon in Japan but have no way of evaluating this assumption. It is also higher than would be the case if we limited the tabulations to women currently in their first marriage. It may be that postdivorce cohabitation is relatively common and that these unions have a relatively high likelihood of resulting in remarriage, but we are not able to evaluate this possibility with the available data.
13.This figure replicates Iwasawa’s (2005) tabulations of these same data and is very similar to the figure calculated by Tsuya (2006) based on the first Japanese Generations and Gender Survey. Both are substantially higher than the corresponding figure of 12.8% from the 24th National Survey on Family Planning (conducted in 1998 by the Mainichi Shimbun Population Problems Research Council).
14.Median duration is 15 months for all cohorts born between 1965 and 1979.
15.Recognizing that educational attainment and attitudes are correlated, we have estimated models with and without the attitudinal variables. Because the estimated coefficients for education and other variables are similar in both sets of models, we present results for the full model only.
16.To adjust for educational differences in age at marriage and childbearing, we also examined the proportions experiencing marriage and parenthood by duration since school completion. Because conclusions regarding educational differences in the relationship between cohabitation and family-formation trajectories were similar for both measures of time, we have chosen to present results by age. Results based on time since school completion are available on request.