1Conjugal life and family life in France have undergone profound changes since the early 1950s. Childlessness and singlehood have increased, yet a large majority of individuals still cohabit and have children (Masson, 2013). Because cohabiting couples are increasingly likely to separate (Costemalle, 2015), a growing proportion of the French population is experiencing consecutive unions. Conjugal and family situations are becoming more diverse, as evidenced by the number of blended families (Bodier et al., 2015). Couplehood and family life nonetheless continue to be closely intertwined. The vast majority of children in France are born in the framework of a cohabiting union and, conversely, conjugal trajectories are affected by whether individuals have children (Martin, 1994; Saint-Jacques et al., 2009).
2Using data from the EPIC survey on individual and conjugal trajectories (Étude des parcours individuels et conjugaux), which was conducted by INED and INSEE in metropolitan France in 2013–2014, this short paper offers a new perspective on individuals who have repartnered after separation from or the death of their first partner. These new insights were obtained by studying the differences between the first and second partners’ ages, their social milieus, and the duration of second unions.
3Studies of second unions exist, but most focus on or adopt the perspective of specific subgroups, such as married couples. In 1983, François de Singly studied ‘second husbands’ from the women’s point of view (Singly, 1983), showing that homogamy (the tendency to partner with someone of a similar social status) was still strong in second marriages, though weaker among women with a first husband coming from an intermediate social class.
4Research on conjugal trajectories after first separations has shown that men enter a new union more quickly than women (Costemalle, 2015). Drawing on the 1986 Family Situations survey (Situations familiales), Catherine Villeneuve-Gokalp showed that people are less likely to have children in second than in first unions, that second unions seem to be less stable, and that they are less likely to be formalized through marriage (Villeneuve-Gokalp, 1990, 1991). Moreover, women do not remarry as often as men. In addition, individuals who are relatively slow to repartner after separation are more likely to get married than others.
5Eva Beaujouan’s more recent research, based on the 2005 ERFI study, has contributed new analyses on second-union fertility and the risk of separation. People who repartner have more children, on average, than people who have been in one union only, while those in the latter category have more children than people who remain single after a separation (Beaujouan, 2011, 2016; Beaujouan and Solaz, 2013). Finally, individuals’ family histories influence second-union fertility, regardless of age. Couples in which each partner had children during a previous union are less likely to have children together than couples where neither partner is already a parent at the start of the new relationship (Beaujouan, 2011).
6Several authors have analysed conjugal dynamics—particularly second-union formation and its characteristics—in countries other than France. Some of those studies concern second marriages only (Lampard and Peggs, 1999; Clark and Crompton, 2006); that is, they do not take into account non-formalized cohabiting relationships. Several of these authors note that second-union situations are more complex than first-union situations due to the partners’ conjugal, family, and family-founding pasts. Among second marriages in the United States (White and Booth, 1985; Teachman, 2008) and Canada (Desrosiers et al., 1995), having stepchildren increases the risk of separation; while the post-divorce trajectories of individuals in Flanders vary by whether the divorced partner is now in a non-formalized cohabiting relationship, married, has children, or has stepchildren (Vanassche et al., 2015). In Hungary and Finland, as in France, conjugal past affects second-union fertility: the probability of having children is lower when both partners have already had children in a previous union (Murinkó and Szalma, 2016).
7To observe the characteristics of second unions in France, this article draws on an EPIC  sample of 1,950 respondents (1,115 women and 835 men) who had been in a cohabiting couple relationship—hereafter called a ‘union’, whether formalized or not—at least twice in their lives, regardless of whether the second union was continuing at the time of the survey. The EPIC survey took into account same-sex couples.
8Detailed histories of past relationships allow for reconstructing the dynamics of respondents’ conjugal and family histories and for analysing differences between their first and second unions. The EPIC data on past relationships (where respondent and partner first met and partner’s social milieu) are more recent and, above all, more detailed than data from the French version of the Generations and Gender Survey (Étude des relations familiales et intergénerationnelles, or ERFI) (2005, 2008, 2011).
9Though the EPIC survey data are quite rich, they present some practical drawbacks. As respondents’ partners’ educational attainment is unknown, we cannot study homogamy for that variable or whether the second partner is more or less educated than the first. Furthermore, a considerable proportion of respondents and their partners did not work during the first union and, in some cases, the second union as well, thus making it impossible to ascribe a social category to them in order to study the social category of homogamy. 
10The survey was not conducted with a panel of individuals followed up over time and, therefore, it does not directly lend itself to studying union dynamics and trends across the generations. The further back in time we go, the less representative the sample is of the total population. Lastly, ongoing conjugal trajectories were truncated at the survey date. These constraints, which are inherent to retrospective data, made it necessary to adapt the analytic framework to the question under study. The Cox proportional hazard model was also used to analyse union duration and to account for censored trajectories.
I – The characteristics of first and second partners differ
11To study how second partner characteristics differ from first partner characteristics, first and second partners of all respondents who have been in at least two cohabiting unions were compared (partners of those who have been in only one union at the time of the survey are therefore not counted among the total of first partners).
1 – Conjugal and family-founding trajectories
12People who enter a second union have the experience of the first, including possible formalization (civil union,  marriage) and whether that union produced children. Often, the respondent’s new partner also has a conjugal and family-founding history: 64% of people who have been in at least two cohabiting unions repartner with someone who has also already been in one, and 35% form their second couple with someone who has already had children. At the start of the second cohabiting relationship, then, these people’s conjugal and family situations are more diverse than at the start of the first.
13Among EPIC respondents aged 26 to 65 who have repartnered, 56% belong or belonged to a couple in which they or their partner already had children. In approximately half of second unions (53%), either the respondent or their partner has already been married or in a civil union. In nearly 25%, both partners had already had children when their new union began (Table 1). Those proportions logically increase with age upon repartnering: the older people are, the more time they have had to formalize their first union or to have children. For example, 88% of individuals who start their second cohabiting relationship at age 37 or older have already had children or repartnered with someone who has.
Conjugal and family history of respondent and second partner
Conjugal and family history of respondent and second partnerInterpretation: Forty-four per cent of people aged 26 to 65 in 2013 who have been in at least two unions did not have children during their first union, and their second partner did not have children either.
Coverage: Individuals aged 26 to 65 living in metropolitan France and who have cohabited with at least two partners.
14There are also differences by social milieu, and they are measured here by educational attainment. The more education a person has, the less likely they are to have had children in their earlier relationship or to have repartnered with someone who is already a parent. This holds regardless of age at the start of cohabitation, and it reflects the fact that relatively highly educated individuals are more likely to have had childless first unions—for example, while they were students.
2 – Age differences between respondents and their consecutive partners
15In heterosexual couples, women are younger than their partner on average, though that gap began to close in the 1950s (Vanderschelden, 2006b) before stabilizing in the 1970s (Daguet, 2016). In the first unions of individuals aged 26 to 65 who have been in at least two unions, women were on average 3.3 years younger than their partner, and men were 1.2 years older than theirs.  In second unions, that difference falls to 2.3 years for women, while for men it increases a year and a half to 2.7 years.
16Thus, regardless of sex, people tend to have their second cohabiting couple relationship with someone younger than their previous partner. This trend is due to the fact that the later in life repartnership occurs, the age difference between the partners increases—regardless of whether we are considering the first or second union (at a given age, the union order variable has no effect on partner age difference; Figure 1). This explains in part why the second partner’s economic situation compared to the respondent’s is, on average, not as good as the first partner’s was: the partner’s situation is worse in 42% of second unions as against 33% of first unions.
3 – Where future partners first meet
17Age also explains differences in the place where partners meet between first and second unions. While individuals are still most likely to meet both their first and second partner at a friend’s house or during an evening with friends (25% of encounters), they are twice as likely to meet their second partner in the workplace as they are to have met their first partner there (25% as against 12%). On the other hand, they are less often students when they meet their second partner.
Age differences between partners by sex, respondent’s age at which union began, and union order
Age differences between partners by sex, respondent’s age at which union began, and union orderInterpretation: The partners of women who began their second union between the ages of 35 and 39 are on average 1.9 years older.
Coverage: Individuals aged 26 to 65 in 2013 living in metropolitan France and who have cohabited with at least two partners.
18The proportion of people who meet their second partner in a place outside work, school, or their circle of friends—such as at a dance or a nightclub, through family, in a public place, on vacation, in the neighbourhood, or at an association—is lower than that for the first partner. It may be that single people go to these places less often after separating than before they formed their first couple. Another possible explanation is that it is harder to find a partner there when one is older.
19Conversely, the use of marriage agencies or dating websites increases from first to second union. The proportion of individuals who meet their partners by way of these networks rises from 1% for the first union to 6% for the second. Some of this increase may be due to a time effect, as the use of Internet dating sites has grown considerably only in recent years.
4 – Social homogamy
20It has been shown that the proportion of socially homogamous couples is higher than it would be if people formed couples at random (Desrosières, 1978; Girard, 1981; Bozon and Héran, 1987; Vanderschelden, 2006a; Bouchet-Valat, 2015). The EPIC survey does not provide parents’ social category information, only that of the respondent and the respondent’s partner at the start of the relationship. Many respondents were still students themselves when they first cohabited or had student partners, meaning that we have no social category information on them. So, to compare respondent’s first and second partner regarding this variable, it was necessary to confine the study to respondents whose first two partners had or have already worked; that is, 57% of the EPIC sample.
21In nearly 60% of those cases, the second partner belongs to a different social category than the first. This change holds more consistently for respondents with at least a high school diploma (66%) than for those without (53%). Furthermore, to study trends in absolute homogamy (the proportion of couples in which the two partners belong to the same social category), it was necessary to restrict the analysis to people who themselves were working at the start of their first union, or 40% of the EPIC sample. In that subsample, absolute social homogamy falls slightly from 34% to 28% between the first and second unions.
22Does this mean that in choosing their second partner, individuals are more likely to break free of their own social milieu than they were in their first union? Does it mean that individuals whose first union ended in separation want to open themselves up to horizons other than the ones they were initially given? These questions are not easy to answer, especially since the second partner is more likely than the first to have grown up in a family context similar to the respondent’s. Respondents were asked to specify whether their consecutive partners were raised in contexts resembling their own in terms of material situation, religion, leisure and cultural activities, education, and politics; and 36% answered in the affirmative for the first partner in at least four of those five dimensions, while the figure rose to 40% for the second partner.  It is particularly in regard to education and politics that the second partner more closely resembles the respondent than the first partner did.
II – Second cohabiting unions depend on family and family-founding pasts
23The first cohabiting relationships of most people who have been in at least two unions were fairly short (seven years on average). Second unions generally last longer than first ones (this is the case for 61%, but in 20% of cases it cannot yet be determined which relationship will be longer). This is not surprising, since some first unions that ended in separation were quite short.
24If we change references and instead compare the duration of the second union to that of all first unions, regardless of whether they ended in separation or are still ongoing, we observe that the instantaneous risk of second-union breakup is no higher than the risk for all first unions. Beaujouan’s studies have shown that, in France between the 1970s and the 2000s, couple stability reversed in terms of union order. Whereas second unions formed in the 1970s and 1980s broke up more often than first unions over the same period,  the opposite has held since the 1990s. Indeed, when structure is maintained equal, second unions are more stable than first ones (Beaujouan, 2016).
25To examine the influence of particular variables on the risk of second-union breakup, a Cox proportional hazards model was used. Similarly to linear regression, it enables us to calculate effects with other variables kept equal by assigning coefficients to each independent variable (Table 2). Like first unions (Costemalle, 2015), second unions become more ephemeral from one generation to the next. People born in the 1970s or later who have repartnered are more at risk for separation (at equal durations of cohabitation time) than those born in the 1950s.
Relative risk of second-union breakup (odds ratios estimated by Cox model)
Relative risk of second-union breakup (odds ratios estimated by Cox model)Interpretation: A significant coefficient above (below) 1 indicates a higher (lower) relative risk of separation than in the reference situation.
Statistical significance: * p < .10, ** p < .05, *** p < .01; otherwise non-significant.
Proportionality test: Schoenfeld residuals test does not invalidate the proportional hazards assumption (test 5%).
Coverage: Individuals aged 26 to 65 living in metropolitan France and who have cohabited with at least two partners.
26For both sexes, a lower risk of separation is associated with the second relationship being in the form of marriage (rather than in a non-formalized relationship) or having a child under age 5 whose other parent is one’s second partner. That association does not necessarily reflect a cause-effect relationship. For example, partners who believe their new couple is made to last may more often choose marriage than others. The effect on risk of separation from having young children may be due to the fact that parents who would like to separate may nonetheless think it is preferable to stay together for the good of the child(ren), especially in the case of young children. Other hypotheses are possible, of course. The findings presented just above have been confirmed for first-union duration using the same model.
27What is more, as is the case for first unions, the second unions of individuals whose parents separated when they were minors more often end in separation than do those of people whose parents have cohabited continuously—which is a variation on the ‘intergenerational transmission of divorce’. Children of divorced parents are more likely than average to get divorced themselves (Traag et al., 2000; Engelhardt et al., 2002; Gähler and Härkönen, 2014).
28Taking individuals’ conjugal and family-founding pasts into account generates additional risk factors for second unions. Among men who have repartnered, the risk of separation decreases inversely to the duration of their first union. One explanation of this could be that men whose first relationship was brief are more ‘unstable’ than others and are therefore more at risk for a second breakup. Moreover, the risk of separation in second unions is higher when both the respondent and new partner have already had children than when neither has.
29This risk of separation also increases slightly for women and men whose new partner is younger than they are. That this effect is not reversed by sex may be due to structural effects, as 37% of women and 34% of men who have repartnered did so with someone who had not been in a union before. The set of repartnered women therefore does not correspond to the set of partners of repartnered men. As the statistics do not concern identical populations, the findings for the two sexes are not symmetrical.
30Other effects are significant only for women among the entire EPIC sample. A higher risk of breakup exists for the unions of highly educated women (five or more years of post–high school education) and, at the other end of the spectrum, of women with no diploma than for those of other women. However, the later the age at which they entered that second relationship, the lower the risk of separation. This may be because the older a woman is, the more likely she is to seek couple stability.
31Early in their second cohabiting union, most respondents have already been married, in a civil union, or had children with their first partner. This in turn implies that second unions do not appear to be a repetition of first unions and, moreover, that they involve more varied situations. Furthermore, in the majority of cases, the second partners of most respondents do not resemble their first. On average, second partners are younger than the first, and they are more likely than first partners to be from a different social category than that of respondents. They are also more likely to have grown up in a family context similar to that of the respondents, though, social homogamy seems to diminish slightly in second couples. Finally, the risk of separation in second unions is sensitive to the same factors as first unions: it increases among respondents whose parents separated while it decreases for respondents who have children with or are married to their second partner. However, some factors are by definition specific to second couples, such as having children from a previous union, a factor that puts new couples at greater risk of separation.
32While these trends are stable, the number of people who have several consecutive relationships in their lives continues to increase, meaning that family situations are certain to become still more complex. Based on the EPIC survey, this analysis of second unions enables us to better identify the changes under way.
33As a final note, many of the findings presented here are difficult to explain on the sole basis of the survey data. We would need to know the causes of separation for both first and second unions. Such breakups are probably not independent of other events that occur in people’s lives, such as work-related changes (job loss, job change, and change in place of work) or health problems. We would also have to collect more detailed information on how people perceive their union, for example on whether or not—and, if so, how—they project themselves into the future as a member of that couple, and what importance they attach to couplehood and family life. Many questions still need to be explored, not only so that we can better understand the dynamics of couple and individual trajectories, but also so that we may more clearly apprehend how behaviours, practices, and norms are changing in this area.
Description of respondents whose two partners had already worked when the relationship began (%)
Description of respondents whose two partners had already worked when the relationship began (%)Coverage: Individuals aged 26 to 65 living in metropolitan France who have consecutively cohabited with at least two partners.
See the introduction to the survey in this issue of Population (Rault and Régnier-Loilier, 2019).
Respondents’ parents’ and partner’s social category are also unknown in those cases.
France’s form of civil union, the pacte civil de solidarité, or PACS, was created in 1999 as an option for formalizing both same- and different-sex couples.
These age differences vary by sex. Men and women who have been in at least two unions do not always repartner with someone who has likewise already been in a cohabiting relationship.
The subsample on which these findings are based (respondents and both partners have already worked) is considerably different from the sample as a whole (all persons having formed at least two unions in their lives), namely in that it encompasses people who are, on average, older and less educated (see Table A.1).
That is, all first unions, whether or not they ended in separation.