1Over recent decades, France has seen major transformations in at least two dimensions of conjugality. First, discontinuity now constitutes a widespread feature of conjugal trajectories as separations become increasingly more frequent and the durations of first unions decrease (12% of first unions formed between 1970 and 1978 ended within eight years, with this figure rising to 29% for those formed between 1997 and 2005; Costemalle, 2015). Repartnering after separation has become an ordinary life-course event (Wu and Schimmele, 2005). Among those who separate between the ages of 25 and 50 years, half of individuals cohabit again with a partner within five years (Costemalle, 2015). More generally, the number of serious intimate relationships over the lifespan has been increasing (Rault and Régnier-Loilier, 2015).
2Second, the couple—which was previously a synonym for marriage—is now a more fluid concept. In 2011, among all individuals living as part of a couple (all ages combined), 73% were married, 23% were in a free union (union libre), and 4% were in a civil union, or PACS  (Buisson and Lapinte, 2013). However, this cross-sectional statistic does not capture the full scope of recent changes. Being part of an unmarried couple has come to represent a fully-fledged form of union in itself, so much so that France now has one of the highest rates of unmarried cohabitation (Prioux, 2009). Each year, the number of registered PACS draws closer to the number of marriages celebrated (192,000 versus 228,000, respectively, in 2017; Papon and Beaumel, 2018). Lastly, nearly 1 in 10 adults reports being in a ‘stable intimate relationship’ even though they do not live with their partner (Régnier-Loilier et al., 2009).  However, this situation covers highly variable realities depending on the timing of the relationship in the life cycle, as has been shown through the construction of typologies based on both qualitative (e.g. Duncan and Philips, 2010; Duncan et al., 2013; Stoilova et al., 2014) and quantitative data (Régnier-Loilier et al., 2009; Reimondos et al., 2011; Pasteels et al., 2015; Coulter and Hu, 2017). While for some couples not moving in together is the involuntary result of circumstances (such as geographical distance), for others it is a choice, particularly after a separation.
3These two concomitant trends suggest a need to investigate the link between past conjugal history and the form of unions. Insofar as the configuration of couple relationships  is not fixed and changes over time (more than 9 out of 10 cohabiting relationships are preceded by a period of non-cohabitation; see Costemalle, 2015), this link needs to be examined from a dynamic, process-based perspective. However, the available data is generally limited to an inventory of cohabiting relationships, thus prompting research that focuses on repartnering solely from the perspective of those who are living together. As such, the data excludes not only non-cohabiting relationships but, more generally, the points in the relationship when each individual is living at home.
4The EPIC survey on individual and conjugal trajectories (Étude des parcours individuels et conjugaux) retraced respondents’ full histories of intimate and conjugal relationships, including periods of being in a nonco-residential intimate relationship. This article, which focuses on the post-separation period of a cohabiting couple, looks at the probability of moving in with a new partner and how this varies depending on a set of characteristics.
5Based on certain known determinants of the formation of a new cohabiting relationship, as well as on the results of studies on couples who do not live together, we begin by formulating several initial research hypotheses. We then present a description of the conjugal situations of individuals at the time of the cross-sectional survey, after which we take a longitudinal perspective to look at the probability of their cohabiting based on their conjugal history. Finally, we model the hazard (instantaneous risk) of moving in with a new partner over time, based on a set of factors: social characteristics, the partners’ conjugal and reproductive histories, and the characteristics of both the separation and current relationship. The period of non-cohabitation sheds light on the conditions of repartnering, as it reveals not only the ways in which actors enter into a new relationship but also the differentiated expectations that they may have at different moments in their romantic and conjugal trajectories.
I – Examining the repartnering process
1 – Factors associated with the formation of a new partnership
6Within the substantial research literature on cohabiting repartnerships, several trends repeatedly appear. First, the greater the age at separation, the lower the chances of forming a new couple; and this is particularly true for women (Cassan et al., 2001; Wu and Schimmele, 2005; Beaujouan, 2009, 2012; Costemalle, 2015; Schimmele and Wu, 2016). In conjunction with age, marital status also has an effect. One Canadian study notably showed that widowers and widows take longer to repartner than divorced individuals, who themselves take longer to do so than those who were never married (Wu and Schimmele, 2005). Some divorcees avoid forming a new cohabiting relationship to avoid returning to a situation of dependence similar to that of their marriage (Levin, 2004), while widows and widowers express less interest in finding a new partner (Carr, 2004). For women, the death of a spouse sometimes marks the end of a period in which they have acted as their husband’s caregiver; then, following the grieving process, they acquire a sense of independence and re-establish links with their social network (De Jong Gierveld, 2002).
7Having children also seems to impede women from forming new cohabiting unions (Bernhardt, 2000), as they are generally the parents given residential custody after separation. The presence of young children in the home limits opportunities to go out and therefore the opportunities to meet a new partner (Ivanova et al., 2013; Botterman et al., 2014). Moreover, in anticipation of the possible difficulties associated with being a step-parent, potential partners may be dissuaded from forming a union with a woman who has dependent children (Stewart et al., 2003) while, on the other hand, mothers may fear that a new cohabiting relationship could provoke problematic interactions between their children and the new partner (Martin, 1994). In contrast, fathers with custody of their children apparently repartner more quickly than those without custody (Goldsheider and Sasster, 2006). Even when the separation happens at age 45 or later—at a time when young children are less likely to be in the home—the difference between men and women nonetheless persists (Schimmele and Wu, 2016). It is only when the separation happens early in the life cycle (before the age of 25 or before the couple has children) that women and men form new unions within the same amount of time (Beaujouan, 2012; Costemalle, 2015).
8While the probabilities of re-forming a new cohabiting relationship are well documented, quantitative studies have generally failed to shed light on part of the repartnering process: the construction phase of the intimate relationship, when the new partners see each other regularly before potentially moving in together.  It may be hypothesized that this process varies according to individual conjugal histories. The article by Nathalie Beltzer and Michel Bozon (2006) is among the few statistical studies that go beyond the narrowly defined scope of re-forming a cohabiting union, and they look more closely at the emotional and sexual aspects of life following a separation. According to their observations, women are more likely to know (without necessarily having previously had sexual relations with) their new partner at the time of the separation, with this new partner being the one with whom they are most likely to form a cohabiting union. In contrast, among men, ‘the degree of partner turnover’ (p. 467) after separation is higher. Beltzer and Bozon also show that access to a post-conjugal sex life is less frequent among women aged 35 or older than for men of the same age or for younger women. Being older corresponds to a longer average period of living as part of a couple before separation, a sustained conjugal commitment that may leave more pronounced subjective traces in women than in men, thus reducing their aspirations for meeting a new partner. However, the data used in their study  did not allow the authors to analyse the role of the individuals’ parental status at the time of separation, nor to examine the length of the pre-cohabitation relationship.
9More recently, Sofie Vanassche, Martine Corijn, and Koen Matthijs (2015b) identified different post-divorce conjugal trajectories in Flanders, including those of individuals who remain in a long-term non-cohabiting relationship (women and men in this category tend to be older and more educated than the average at the time of separation). Evidence from individuals in a stable, noncohabiting, intimate relationship followed up over several years has shown that divorcees in France maintain such relationships for longer durations, which is in contrast to young people who tend to treat non-cohabitation as a ‘trial period’ that leads rapidly to either moving in together or breaking up (Régnier-Loilier, 2016, 2017). As such, the significance is very different for non-cohabitation, often referred to as living apart together (LAT), depending on whether it occurs in a first union or as part of a repartnering process; thus, it should not be treated as a single, simple phenomenon.
2 – Avenues for research
10Carried out in 2013–2014 by INED and INSEE, the EPIC survey retraces in detail the full intimate and conjugal relationship trajectories of individuals aged 26–65 years, including the distinction between the beginning of the relationship and the beginning of cohabitation, and it provides a number of details on the last separation (see Box). As such, it covers the full conjugal history and can be used to study the dynamics of the formation of a new cohabiting relationship following a breakup. Four principal hypotheses are suggested by the combined research results on repartnering and on LATs.
Box. The EPIC survey
All ‘couple relationships or serious intimate relationships’ (with no restrictions on duration or co-residence criteria) were detailed. The various stages of each relationship were recorded: start date of the relationship; any dates of moving in together, marriage, PACS; children’s dates of birth; and the dates indicating the end of the relationship, end of cohabitation, divorce or PACS dissolution. As such, we have the respondent’s age at the start of each of their relationships, their conjugal situation (whether or not they have been married or registered in a PACS), their parental status (already have children, their ages, and whether the children live with them), the duration of each relationship, and even the length of time between each relationship.
Moreover, a detailed description was taken of the last separation from a cohabiting union. This included a few questions aimed at capturing the respondents’ feelings: ‘Was this separation disruptive for you on an emotional level?’ with the possible answers being ‘Yes, very.’ / ‘Yes, somewhat.’ / ‘No.’ ‘And on a material and financial level?’ We also know whether the ex-partners ‘continued to live together after deciding to separate’ and how they got along (‘At the time of the separation, would you say that your relationship with [person’s name] was …’ ‘Excellent.’ / ‘Good.’ or ‘Neither good nor bad.’ / ‘Antagonistic, tense, or violent.’ / ‘Nonexistent or almost non-existent.’).
A few pieces of information on the conjugal past of the respondent’s various partners were also gathered: ‘Had [person’s name] ever lived with anyone previously?’, ‘Had he/she been married?’, ‘Had [person’s name] had children before the beginning of your relationship?’, ‘Did [person’s name] live with his or her children (from another relationship)?’
EPIC therefore offers the possibility of two types of analysis. Cross-sectionally (Section II), it can be used to study the respondents’ conjugal situations at the time of the survey by distinguishing among those in cohabiting relationships, those in non-cohabiting relationships, and those who report not having a partner. Among the latter two, it can further distinguish between those who are in a ‘serious intimate relationship’ (non-cohabiting) and those who are not in a relationship. The second type of analysis is longitudinal (Sections III and IV), as the survey can also be used to study various sequences in the process of forming a new relationship, most notably the time period between the beginning of the current relationship and the start of any cohabitation.
11First, we might expect the time spent in a non-cohabiting relationship to be longer for women with dependent children, who are much more likely than men to be given residential custody.  Given this status quo, mothers are less likely to form another cohabiting union (Beaujouan, 2012; Vanassche et al., 2015a). Claude Martin (1994) notes, moreover, that they are more likely to choose a non-cohabiting relationship. Most qualitative studies on LATs emphasize that when mothers repartner, they ‘choose’ not to cohabit with their new partner in order to avoid disturbing their children by imposing a stepparent on them (Bawin-Legros, 2001; Levin, 2004; Duncan et al., 2013). In contrast, there is effectively no difference between the parents of adult children and non-parents in terms of forming new cohabiting unions (Vanassche et al., 2015a), and we would therefore not expect the existence of adult children to affect time spent in a non-cohabiting relationship.
12Second, we hypothesize that a cautious attitude towards a new relationship may develop following a long previous relationship such as marriage or some other form of cohabiting union that is followed by divorce or other form of separation that incurs emotionally or materially disruptive effects. This may result in a delay in moving in together, as time is taken to ensure the solidity of the relationship while maintaining separate personal space (Villeneuve-Gokalp, 1997). This may even diminish any desire to live together as a couple in order to avoid the risk of experiencing another painful breakup (Levin, 2004; Duncan et al., 2013).
13The third relates to the previous hypothesis in that the period following separation may be understood as a more or less prolonged period of self-reconstruction, a period in which the individual (re)learns to live alone while experiencing a certain degree of freedom on a personal level as well as in the form of an expanding friendship network (Villeneuve-Gokalp, 1994). We might thus hypothesize that the more time an individual spends alone before beginning a new relationship, the more difficult they find it to surrender this new—or renewed—freedom by moving in with a new partner. Non-cohabitation may therefore allow individuals to maintain a certain distance from shared conjugal life and thus maintain their own identity while on their own (Charrier, 2008).
14Lastly, our fourth hypothesis postulates that the time taken to move in with a new partner will be shorter among individuals at the lower end of the social hierarchy, and this is for two reasons. First, greater economic constraints can push couples to move in together in order to benefit from certain economies of scale associated with cohabitation (one rent rather than two, etc.) (Haskey and Lewis, 2006).  In turn, we can expect those at the top of the social ladder to be more inclined to remain in a non-cohabiting couple. This is, however, in contrast with what was observed for previous cohorts—namely, that level of education is no longer a distinguishing factor in marital status (Prioux, 2009). Thus, the choice of one or another form of union varies with social status. For example, Bailly and Rault (2013) showed that those who opt for a PACS tend to be more educated than those who choose to marry. Moreover, Liefbroer et al. (2015) found a slight positive correlation across Europe between level of education and the probability of being part of a non-cohabiting couple. This may reflect socially differentiated conceptions of living as a couple, with lower social classes being less open to alternatives to the ‘standard’ form of relationship, while those who are more educated are more resistant to traditional family norms and potentially become the pioneers of a new form of conjugality, one in which the partners each have their own home (Liefbroer et al., 2015).
15In order to test these hypotheses, we will begin with a descriptive analysis (first cross-sectional, then longitudinal using Kaplan–Meier survival functions), followed by modelling (Cox models) to measure the specific effects of different factors on the relative likelihood of partners moving in together over the course of the relationship, versus remaining in a non-cohabiting union.
3 – Coverage of the study
16The study covers individuals who had been in a previous cohabiting relationship (cases where the responses to the EPIC survey include a set of details on the last separation ) and were in a new ‘serious couple or intimate relationship’, whether a cohabiting union or not (configurations a, b, and c in Figure 1). The objective is to analyse the duration of the non-cohabitation period of ongoing relationships at the time of the survey (represented in Figure 1 by the capital letter A), based on the characteristics of the previous cohabiting relationship and the separation process.  In configuration a, either the current relationship is a cohabiting union (possibly preceded by a period of noncohabitation), or the relationship is still non-cohabiting at the time of the survey (right-censored, which will be taken into account in the analyses). In configuration b, which is uncommon, the current relationship began while the respondent was still living with their previous partner. In a quarter of cases, respondents had continued to cohabit (taking time to get organized, etc.), despite their relationship being over; in the other cases, the period of overlap between the previous and current relationship was most often brief (less than eight months in half of the cases). In configurations c and d, one (or more) non-cohabiting relationships took place after the last cohabiting relationship,  but they were over by the time of the survey. As the study concentrates on relationships that were ongoing at the time of the survey and configuration d does not include any such relationships, it is out of the scope of this paper and therefore not analysed here.  In configuration c, only the duration of the non-cohabiting period of the current relationship is analysed. Nonetheless, the analyses will take into account the occurrence of a non-cohabiting relationship between the end of the previous cohabiting relationship and the current partnership.
Cases present in the EPIC data and duration studied (A)
Cases present in the EPIC data and duration studied (A)
17A comment is merited here on representativeness in light of these methodological choices: this study focuses on relationships that were ongoing at the time of the survey; therefore, it is not representative of separated individuals. Individuals who were single when surveyed may have experienced a non-cohabiting romantic episode (configuration d); while others may have gone on to have a future relationship. Furthermore, sexual orientation is not taken into account in the analyses. Although women and men in same-sex couples are proportionally more likely to be in a relationship with someone who lives in a different home (Buisson and Lapinte, 2013; Régnier-Loilier, 2018), the corresponding subsamples in EPIC are too small to study the specificities of their behaviours following separation (19 men and 18 women). In addition, regarding people whose last cohabiting relationship ended due to their partner’s death, the survey sample does not include enough of this population to include them in the study.
II – Identifying non-cohabiting couples
1 – A configuration that is difficult to identify accurately from the surveys
18Until recently, surveys usually collected information on the situations of cohabiting couples only. Individuals not living with a partner were sometimes asked about any stable intimate relationships with a person living elsewhere, as in the case of the ERFI survey on family and intergenerational relationships (Étude des relations familiales et intergénérationnelles).  Here, nearly 10% of adults (aged 18–79 years) in France reported being in this situation in 2005. However, this notion of a ‘stable intimate relationship’ covers a wide variety of situations (Régnier-Loilier et al., 2009), and only a third preferred to instead describe their relationship as ‘more of a couple relationship’ (Régnier-Loilier, 2018). In France, it was only in 2011  that a question item was introduced via the Family and Housing Survey, or EFL (Enquête Famille et Logements), which aimed to directly identify non-cohabiting ‘couples’: ‘Are you currently in a couple? Yes, with someone who lives in the same home; Yes, with someone who lives in a different home; No, but I have been part of a couple in the past; No, I have never been part of a couple.’  Only 3% of adults (aged 18–79 years) in the EFL reported being in a relationship with someone who did not live in the same home (Régnier-Loilier, 2018).
19In the EPIC survey, respondents were asked the same question at the beginning of the questionnaire to ensure that the response would not be influenced by other questions in the survey. Later in the questionnaire, respondents were specifically asked to describe all of their ‘couple relationships or serious intimate relationships’. This broad definition of relationship may have affected how respondents understood the question, leading more of them to report being in a relationship ‘with someone who does not live in the same home’. Despite these two precautions (formulation and position in the questionnaire), a comparison of the two sources reveals large discrepancies in frequency at all ages (Figure 2). On average for the age range that is common to the two surveys (26–65 years), the proportion of individuals who reported being in a couple with someone living in another home was more than three times greater in EPIC (a little less than 6% versus the 2% reported in EFL).
Proportion of men and women in a non-cohabiting union, by age
Proportion of men and women in a non-cohabiting union, by ageNote: Moving averages of order 5 (the proportion at age a corresponds to the weighted mean of the proportions observed at ages a − 2, a − 1, a, a + 1, a + 2).
Coverage: Respondents aged 26–65 (metropolitan France, ordinary housing).
20There are at least two possible explanations for these discrepancies. First, it was clearly explained to respondents that the general theme of the survey covered conjugal histories, and this may have led more respondents to report being ‘in a relationship with someone who does not live in the same home’ if they were in an intimate relationship yet not living with a partner. Secondly, the data collection method may have played a role: the EFL questionnaire was self-administered (with no interviewer), while the EPIC questions were asked by an interviewer. As such, the interviewer could provide further details on what was meant in the survey by ‘being in a couple’. Although no precise definition was given in the instructions for data collection, the interviewers were nonetheless given some guidance indicating the broad interpretation aimed at in the EPIC questionnaire. In contrast, the EFL questionnaire served as a supplement to the population census questionnaire and asked respondents whether they ‘lived’ in a couple. It could be that, for the sake of consistency, some individuals who responded in the census that they did not ‘live in a couple’ also indicated in the EFL that they were not ‘in a couple’. Whatever the reasons, these differences illustrate the difficulty of precisely identifying ‘non-cohabiting couples’. While marriage and even cohabitation are quite clearly identifiable events, the context of non-cohabiting relationships tends to obscure the boundary between being and not being in a couple, thus leaving the interpretation up to the respondents and resulting in large variations.
2 – A situation that is more common in the second half of a relationship trajectory
21Regardless of the source, the proportion of individuals who reported being in a relationship with someone living in another home is higher at the youngest ages (Figure 2). At these ages, non-cohabitation is often associated with a first intimate relationship and can be understood as a temporary phase leading fairly quickly to the partners either moving in together or breaking up (Régnier-Loilier, 2016). Non-cohabitation then becomes rarer as cohabiting conjugality becomes commoner. The slight rebound seen between the ages of 40 and 55 years in the EPIC survey (but not in the EFL) may reflect post-separation situations. After a separation, a repartnering period begins. And, as mentioned above, the probability that individuals will re-establish a cohabiting union declines as the duration of their preceding union increases; and, further, this depends in particular on their parental status at the time of the separation. At these ages, children are often present and may lead to individuals forming new relationships without moving in with their new partner.
22A respondents’ relationship situation (cohabiting or not) at the time of the EPIC survey thus seems to be closely tied to their romantic and conjugal history (Figure 3). Four situations are distinguished in the survey: individuals forming part of a cohabiting couple, those in a couple living separately, those who do not consider themselves to be in a couple but report being in a ‘serious intimate relationship’ (non-cohabiting), and those who are not in an intimate relationship.
23The proportion of individuals not in a relationship (neither in a couple nor in a serious intimate relationship) at the time of the survey is higher among those with experience of previous relationships: 30% of men and 39% of women who had previously been part of a cohabiting couple were not in a relationship, while 15% of men and 7% of women had never been in a serious intimate relationship previously. Similarly, the proportion of respondents in a ‘noncohabiting couple’ is markedly higher among those who had previously been in a cohabiting couple (around 10% after a first cohabitation) than among those who had never been in an intimate relationship before the one under way at the time of the survey (less than 2%).
Relationship situation at the time of the survey, by relationship and conjugal history, and sex
Relationship situation at the time of the survey, by relationship and conjugal history, and sexInterpretation: At the time of the survey, 83% of men who had never previously had a serious intimate relationship were cohabiting with a partner.
Coverage: Respondents aged 26–65 (metropolitan France, ordinary housing).
24These differences may be due to the combination of two effects. First, after separation, couples may choose a form of relationship where each partner maintains a separate home (exercising caution before living with anyone again, for example). And second, after separation, individuals may find themselves in situations that delay or restrict the partners from moving in together (the presence of children, for example).
III – Analysing the dynamics of the formation of a new union according to past conjugal experience
1 – Propensity to live together based on age and relationship history: A descriptive approach
25Let us now look more specifically at the dynamics of forming a new intimate relationship after a separation by taking a longitudinal approach. The observed cross-sectional differences suggest that past conjugal experience influences not only the timing of cohabitation in new relationships but also the choices in the form of the relationship (whether living together or apart). In order to take a dynamic look at the time spent in a non-cohabiting relationship according to past relationship history while also taking into account that some couples were not yet cohabiting but may do so at a later date (right-censoring), we used a survival function (Kaplan–Meier method) to measure—for each month since the beginning of the current relationship—the probability that the partners would continue to live separately (Figure 4).  As no significant difference was found between men and women, the two sexes are grouped together here.
Probability of remaining in a non-cohabiting couple by relationship history (Kaplan–Meier survival curve)
Probability of remaining in a non-cohabiting couple by relationship history (Kaplan–Meier survival curve)Interpretation: Eighteen months after the start of the relationship, 51% of individuals with no previous relationship were still not cohabiting.
Note: The coloured ranges represent the limits of the 95% confidence interval.
Coverage: Respondents aged 26–65 years in a couple or serious intimate relationship (metropolitan France, ordinary housing).
26Generally speaking, a lasting relationship moves fairly quickly towards the establishment of a shared home. After 15 months, the probability of still living apart is below 50%. After two years, it is only one-third. Conjugal experience has various effects on the timing of cohabitation. While in a proportional sense very few people with no previous experience of a serious intimate relationship are in a non-cohabiting relationship (2%), the probability that they will move in with a partner during the initial months of their first relationship is also the lowest. After 18 months in a relationship, half of individuals are still in a non-cohabiting relationship versus one-third of individuals who have experienced at least one previous cohabiting union.
27This is partly due to an effect of age at the start of the relationship (as Figure 5 confirms) and is explained by the various situations in which people tend to be at various ages. Some of the youngest individuals have no past experience and are still pursuing their education when they meet their partner, a situation that is not particularly compatible with cohabitation. Limited resources and uncertainties about future place of residence (which will be determined on entry into the labour market) prompt young people to delay cohabitation (Giraud, 2017). Moreover, young adults often prefer to avoid becoming ‘adults’ too soon (Singly, 2000), particularly in terms of getting tied down in a cohabiting relationship when their friends of the same age are ‘going out clubbing’ (Giraud, 2017). After five years (60 months), however, the probability that those in their first relationship will still be living apart from their partner is less than 10% versus 17% for those who have experienced at least one cohabiting relationship, which rises to 27% for those having experienced at least two periods of cohabitation. Having experience of cohabitation therefore means that an individual is less likely to cohabit with a new partner.
28The data reflect two different patterns: first, the probability of being in a non-cohabiting relationship is negligible at the end of eight years among young people (under age 30) who have never previously lived with a partner; and, second, individuals who have previously lived with a partner are less likely to end up moving in with a new partner. These results corroborate those obtained using panel data that show low survival in non-cohabitation for the youngest groups, for whom living apart represents a test period that leads rapidly (within a few years) either to separation or to cohabitation, while for those who have previously lived with a partner, living apart together can be a long-term choice (Régnier-Loilier, 2016). The next part of the analysis focuses on this particular population of individuals—those who have cohabited in the past—to assess the effect of conjugal history characteristics on the repartnering process.
Probability of remaining in a non-cohabiting couple by age at the start of a relationship (Kaplan–Meier survival curve)
Probability of remaining in a non-cohabiting couple by age at the start of a relationship (Kaplan–Meier survival curve)Interpretation and note: See Figure 4.
Coverage: Respondents aged 26–65 years in a couple or serious intimate relationship (metropolitan France, ordinary housing).
2 – Dynamics of new union formation according to conjugal experience
29Conjugal history is not limited to the number of past experiences. Studies on cohabiting repartnerships suggest that each partner’s marital experience (having been married or not, duration of the relationship) and parental experience (having children when the new relationship begins, the children’s age and place of residence) influences the repartnering process and the form of the new relationship. Other factors may also come into play, such as sex, social class, age at the beginning of the current relationship, specific characteristics of the individual’s conjugal history (length of time since the separation, intimate relationships since the separation, stability of the relationship) or of the separation (quality of the relationship with the ex-partner at the time of separation, continuing to cohabit after deciding to separate, materially or emotionally disruptive separation). 
30Some of these characteristics are partly interlinked (for example, long relationships are also more likely to result in marriage and children). We use a semiparametric proportional hazards model (Cox model) to reveal the net effect of each factor on the repartnering process. In other words, we estimated the hazard of having moved in together in the form of a time-dependent function (time since the start of the current relationship, in months) for individuals who had previously been in a cohabiting couple and who were in a relationship, cohabiting or not, at the time of the survey.
31We present three models that include all the variables listed above (Table 1).  The first model combines both sexes, then a separate model for each sex. As the sexual and conjugal behaviours of women and men after a separation are different (Beltzer and Bozon, 2008), our aim is to see whether differences by sex arise in the repartnering process.
Little variation between men and women in their propensity to cohabit
32The first model (‘All’, Table 1) reveals no significant difference by sex in the hazard of moving in with the partner in a new relationship. At first glance, this result may seem to contradict earlier studies on repartnering, most of which have found that women are less likely to form a new cohabiting union. However, this finding is a product of the different approach adopted here: the study concerns a selected population of individuals who are already in a new relationship (whether cohabiting or not). The lack of any difference between the sexes suggests that sex does not particularly affect the decision to cohabit or not, nor does the length of time spent not cohabiting with a partner, but that it does have an effect on whether an individual enters into a new relationship at all. Beltzer and Bozon (2008) observed that, following a separation, women aged 35 or over are less likely to meet a new partner. As a complement to the models, Figure 6 shows the probability of beginning a new relationship (cohabiting or not) over time and by sex. The EPIC data do in fact show that women have a lesser tendency to repartner. For example, five years after separation, 55% of women have not begun a new intimate relationship, versus 47% of men. These differences between women and men persist even after controlling for other characteristics that are typically included in studies on the subject (results not presented here). 
Hazard of moving in versus continuing to live apart (hazard ratios, Cox model)
Hazard of moving in versus continuing to live apart (hazard ratios, Cox model)Interpretation (e.g. women model): Women with no educational qualifications are 1.43 times more likely than holders of a baccalauréat (Ref.: French high school diploma) to move in with their partner. Ref. = reference situation.
Statistical significance: *** p < .01, ** p < .05, * p < .10; otherwise non-significant.
Note: The ‘under 30 years * duration in months’ interaction term is not a risk ratio but a ratio of risk ratios.
Coverage: Individuals in a relationship (cohabiting or not) at the time of the survey and having previously lived with a partner (excluding widowers and widows).
Probability of remaining single after a separation, by respondent’s sex (Kaplan–Meier survival curve)
Probability of remaining single after a separation, by respondent’s sex (Kaplan–Meier survival curve)Interpretation: Eighteen months after the end of the relationship, 73% of women were not in a new relationship.
Coverage: Individuals aged 26–65 who had previously been in a cohabiting relationship (metropolitan France, ordinary housing).
Note: The drop observed at duration 0 reflects individuals who were in a new relationship as soon as they separated from their previous partner.
33As suggested by most of these studies, the difference between the sexes in terms of the timing and intensity of entering a new serious relationship (cohabiting or not) mainly relates to the presence of children from a previous union. No difference is seen among individuals who have not had children; but when the previous union produced children, the propensity of women to repartner is lower than that of men (Figure 7). Here again, the difference is confirmed after controlling for other factors.  Above all, it is the propensity to meet a new partner and begin an intimate relationship that differentiates women and men who have children from a previous relationship—much more so than the form that this relationship takes (cohabiting or not).
Probability of remaining single after a separation, by respondent’s sex and the presence of children (Kaplan–Meier survival curve)
Probability of remaining single after a separation, by respondent’s sex and the presence of children (Kaplan–Meier survival curve)Interpretation: See Figure 4.
Coverage: Respondents aged 26–65 who had previously been in a cohabiting relationship (metropolitan France, ordinary housing).
For women, presence of children lowers the probability of a new cohabiting partnership
34Among individuals in a new relationship, modelling reveals a specific effect of parental status at the time of repartnering for women (Table 1): having one or more minor children, whether they reside with her on a full-time basis or not, decreases a woman’s likelihood of moving in with her new partner. Women are also less likely to move in with a partner when the latter is already a parent himself.  For men, family situation does not appear to have any significant effect on the hazard of moving in with a new partner. This confirms that a man’s parental history has a lesser impact on their subsequent conjugal life, or it may reflect different mechanisms that produce the same effect. On the one hand, fathers who do not live with their children may be ‘freer’ to repartner; on the other hand, those who do live with their children may find that a new cohabiting partner also provides parenting support, or they might be more oriented towards family life, which could explain why no difference is seen between these two scenarios. The parental status of their new partner does not have a clear effect: if she lives with children, the hazard of moving in together is lower but not significantly so (it becomes significant when the respondent’s family situation is not included in the model,  which is similar to the findings for women). Having adult children who live elsewhere, however, has no impact on the probability of moving in with a new partner, neither for women nor for men, all else being equal. In this situation, the partners may judge that the children are less likely to be disturbed by a new cohabiting union and/or that step-parent relationships are less likely to disturb the new intimate relationship. As such, the hypothesis that women spend longer not cohabiting when they have dependent children is confirmed.
Differing effects of a previous marriage by sex
35The effect of previous marital status on the length of time spent not cohabiting with the current partner is partly confirmed, but it is complex to isolate precisely. First, we see no symmetry in the results between the men and women models (Table 1). Although, for men, the hazard of moving in with a partner is lower when the latter has already been married (men model), no significant effect of marriage is seen in the women model. This may be related to taking into account more of the respondent’s characteristics than those of the partner. The partner’s previous conjugal history is limited to a single indicator (whether or not they have previously been married), whereas a set of other variables are taken into account for the respondent (duration of previous relationship as well as the context and consequences of the separation ).
36For men, having previously been married produces the opposite effect to that seen in women: for any given parental status, duration of previous relationship, or age at the start of the new relationship, it is more likely that they will cohabit in their new relationship. This result is consistent with the study conducted by Francine Cassan et al. (2001), who reported that divorced men are much more likely to live with a new partner. This may be related to a selection effect in previously married individuals, as the choice to marry may correspond to a preference for or attraction to being part of a couple, thus prompting the same individuals to form a new union more readily (Bernhardt, 2000).
37No significant effect of duration of previous relationship is observed once other characteristics (in particular, past marital situation) have been introduced.  However, an analysis stratified by past marital situation shows the following: while the duration of the previous relationship has no significant effect among those who have previously married, we can see that among those who have never married it is negatively correlated with the likelihood of moving in with the new partner.  In other words, the experience of marriage overrides the duration of the relationship, although in the absence of marriage the duration of the relationship has a negative effect on the process of entering a new cohabiting relationship.
38With regard to the experience of separation, perceived negative consequences of the separation (material or emotional disruption) have no effect for either men or women. Two factors are nonetheless negatively associated with cohabitation in a new relationship. Women who describe their relationship with their ex-partner at the time of separation as ‘antagonistic, tense, or violent’ are less likely to cohabit with their new partner. Among men, the negative effect is associated with having continued to cohabit with their ex-partner after deciding to separate. This situation may reflect difficulties in moving on from the previous relationship, hesitation about separating at all, or financial and material difficulties that had to be overcome before being able to cease cohabitation.
39Overall, the effects from the previous relationship’s characteristics (duration, marital status) and the experience of separation run in the expected direction. Women seemingly tend to be more tentative about entering cohabitation with a new partner if they have experienced past marriage, a long previous relationship, and/or difficulty with the ex-partner at the time of separation. For men, the likelihood of moving in with a new partner is reduced only when the process of ceasing cohabitation at their last separation was protracted, whereas having been previously married has the opposite effect.
Length of time between separation and current relationship has no effect
40The length of time between the separation and the beginning of a new relationship does not affect an individual’s propensity to move in with their new partner. The data do not confirm the hypothesis that the longer individuals take to begin a new relationship, the less likely they will be to move in with their new partner.
41Likewise, women and men who begin their new relationship before (or immediately after) separating from their ex-partner do not show a significantly different hazard of cohabiting with the new partner. Note that in this type of situation (beginning of the current relationship preceding the end of the previous relationship: see Figure 1, case b), we considered that individuals were exposed to the chance of moving in from the outset of the new relationship, even if they were still living with their ex-partner. Taking the separation of the previous relationship (and not the beginning of the current relationship) as the starting point, the likelihood of cohabitation is significantly higher for those whose relationships overlapped (result not presented ). In this scenario, cohabitation happens more quickly from the time of the separation, but ultimately the total time spent in a non-cohabiting relationship does not differ from that observed for successive relationships without an overlap.
Higher probability of cohabiting among those less educated
42A marked effect of education level (which correlates with social class) is observed in women, confirming our fourth hypothesis. The hazard of moving in together is significantly higher among the least educated, and the opposite trend is seen among women with higher education qualifications. The interpretation of this result remains open, however, and a number of elements may be at play. On average, a high level of education is accompanied by a higher standard of living and thus by a greater possibility of residential independence, thereby imposing less economic pressure to move in together. As such, the most educated may be less sensitive to the potential economies of scale obtained through living together rather than separately.  Another explanation, which may be complementary or competing, is the possibility of socially differentiated concepts of being part of a couple, with greater attraction towards cohabitation among those lower on the social scale and, conversely, a greater penchant for residential independence among the most educated.
Hazard of living together decreases with age
43Finally, various characteristics of the current relationship have an impact on the propensity to live together. First, age at the start of the relationship significantly influences the repartnering process. For those who began a new relationship before age 30, the effect is not significant at the start of the relationship when compared to those aged 30–44 years, but the interaction with the dependent variable (time before repartnering) is significant and positive. This indicates an increasing likelihood of moving in together as the relationship persists over time among the youngest individuals. In contrast, those aged 45 to 65 years at the start of their new relationship are less likely to share a home.
44Second, being in a relationship that has been interrupted one or more times (without information on the reasons or the duration) decreases the likelihood of quickly moving in together.  While this may be a sign of more unstable relationships and, as such, be accompanied by less haste to share the same home, the effect is also a ‘mechanical’ one: temporary breakups correspondingly extend the time considered here as forming part of the time spent in a noncohabiting relationship. 
45Lastly, having been in another non-cohabiting intimate relationship between the last cohabiting relationship and the current relationship (Figure 1, case c) decreases the propensity to cohabit. This may reflect less desire to cohabit after separation. Likewise, the individual may have considered the intermediate relationship to have been a failure and might therefore exercise a degree of caution towards a new cohabiting relationship.
46With most available demographic surveys and data, repartnering following a separation can be studied only by looking at the formation of a new co-residential union. However, conjugality cannot be reduced to cohabitation alone, particularly after a separation. In this situation, non-cohabiting conjugality is common: cross-sectionally, nearly a fifth (18%) of women and men who are in an intimate relationship following a previous cohabiting union do not live with their new partner. However, this ‘snapshot’ suffers from certain limitations, particularly in terms of accurately identifying those who are part of a couple with someone living in a separate home. Although it does not eliminate all the difficulties involved in identifying periods of non-cohabitation, EPIC does provide the possibility of distinguishing different phases in the formation of a relationship. As such, it can be used to perceive repartnering not as a state but as a process, specifically by studying the duration of the non-cohabiting period and how it varies depending on individual situations. In terms of the hypotheses we formulated, four main results emerge.
47First, once a new relationship has begun, the average timing and probability of moving in with the new partner is equivalent for women and men. While most studies have shown that women are less likely to return to living with a partner, this seems primarily related to their being less likely to begin a new intimate relationship. It may not be that cohabiting as such is less attractive or possible for them, but that they have fewer opportunities to meet partners or less interest in beginning a new intimate relationship.
48Second, as many qualitative studies have emphasized, it seems important to examine the role of individuals’ past conjugal histories in the repartnering process, particularly for women. Having been married or in a long relationship and having experienced conflict or tension with their ex-partner at the time of separation decreases a woman’s propensity to move in with her new partner.
49Third, age at the start of the new relationship and parental status are determinants of the form taken by the new union: the older the individual, the lower their tendency to form a cohabiting relationship. Similarly, the presence of young children (minor and/or co-resident at the time of separation) from a previous relationship limits cohabitation, although this result is seen only in women.
50Fourth, whereas women and men in lower social classes are slower to resume a cohabiting relationship (Beaujouan, 2009), once a new relationship has begun, the effect of social class is reversed. The propensity to establish a shared home decreases with increasing levels of educational qualification (and income), and vice versa, with a more marked effect among women. This may relate to the economies of scale associated with living together and/or social differentiation in aspirations regarding life as a couple.
51While the non-cohabiting couple is sometimes considered a possible consequence of the developing values of individualism and autonomy that are characteristic of modern societies, the current state of knowledge warrants a stipulation to this line of explanation. While it cannot be denied that those who are in a non-cohabiting union tend to describe their relationships in a way that focuses more on notions of self-fulfilment, independence, and autonomy, such relationships remain uncommon (between 2% and 6% of individuals aged 26–65 in France, depending on the source). Furthermore, they seem to be specific to certain periods in the life cycle. Among the youngest individuals who are in their first serious intimate relationship, non-cohabitation appears to be a transitory step linked to the conditions of the period (being a student, one of the partners lacking employment, the relationship has begun too recently). However, after a few years, the probability of persisting in this noncohabiting relationship becomes insignificant. If the non-cohabiting couple can be seen as a sign of the ‘second demographic transition’, characterized by a growth in individualism, this occurs mainly in the second phase of the relationship trajectory—after individuals have tried out the ‘standard’ form of relationship (cohabitation) and, in particular, have had children. In this case, non-cohabitation reflects different, non-exclusive types of logic: a cautious attitude, a way of reconciling new personal aspirations, and protecting not only the children from a previous relationship but perhaps even one’s new partner from the potential difficulties associated with being a step-parent. Moreover, at this point in the life cycle, there is likely to be less pressure to cohabit. First, partners are generally less likely to want to have children (some already have all the children they want, or they are no longer able to do so). In a context where living together is one of the prerequisites for having a child, not planning to become parents reduces the pressure to live together. Second, unlike the youngest adults, partners sometimes have their own home, of which they may be owners: this too could reduce the economic pressure to move into a shared home.
AcknowledgementsI would like to thank Marie Bergström, Wilfried Rault, Marc Thévenin, and Sandra Zilloniz for their helpful feedback on this article, as well as the anonymous reviewers and the editorial board for their suggestions.
The EPIC study was financed by INED, INSEE, the Agence nationale de la recherche (the ANR’s CECHIC project on the ‘Corpus for the study of a hundred years of couple formation in France’), the Caisse nationale des allocations familiales (CNAF), the Direction de la recherche, des études, de l’évaluation et des statistiques (DREES), and the LabEx iPOPs (Individuals, Populations, Societies).
The pacte civil de solidarité (PACS) is a French contract of civil union created in 1999 and intended for both same-sex and opposite-sex couples.
The figures are similar in other countries such as Germany, Russia (Liefbroer et al., 2015), Australia (Reimondos et al., 2011), Italy (Régnier-Loilier and Vignoli, 2018), Canada (Turcotte, 2013), Great Britain (Haskey, 2005), and the United States (Strohm et al., 2009).
Here, ‘configuration of the couple relationship’ refers both to the type of partnership (free union, PACS, marriage, etc.) and to cohabitation status (whether the partners live together or not).
Statistical data generally contain no information on the period preceding possible cohabitation, which explains why most studies focus on the formation of a new cohabiting relationship. A recent study by Ingmar Rapp (2018), on the other hand, uses the German Socio-Economic Panel to look at repartnering in Germany from a broader perspective that includes non-cohabiting relationships.
The Knowledge, Attitude, Belief and Practice survey on sexuality, ORS, 2001.
A year after a divorce, 76% of minor children live primarily with their mothers (Bonnet et al., 2015).
A competing hypothesis can, however, be advanced: if moving into a shared home might lead to the loss of some social benefits, this could encourage partners to remain in a non-cohabiting relationship (De Jong Gierveld, 2002; Connidis et al., 2017). Similarly, the occupational instability that is more common at the lower end of the educational scale could increase the probability of noncohabitation (Castro-Martín et al., 2008).
Relationships at the time of separation, whether they continued to cohabit after deciding to separate, and emotional and financial consequences of the separation.
Only the characteristics of the most recent separation of a cohabiting union are described in the survey.
It may have begun, in some cases, before the separation.
Note that, by construction, if they had cohabited, this period would not have been studied; it would have corresponded to the ‘last cohabiting relationship’.
The ERFI study (INED–INSEE, 2005, 2008, 2011) is the French version of the Generations and Gender Survey.
Previously, the Family Situations survey (INED–INSEE, 1985) had included the following question in the description of the respondent’s marital history: ‘Did you share a home or have you always maintained two separate homes?’ This question was asked, however, only in reference to periods of ‘being part of a couple’. It may be assumed that the periods where individuals had a partner without sharing the same home were not always reported as periods of being part of a couple, and they may sometimes have been reported as periods of living alone.
Survey carried out by INSEE on a sample of 359,669 women and men aged 18 years or above.
The window of observation is limited to eight years: beyond this range, events become negligibly rare.
See Box for the formulation of the questions.
The assumption of proportionality of hazards (which aims to measure how constant the risk ratios are across the entire observation period) was tested for each variable in two steps. The first test was performed on the standardized Schoenfeld residuals. Then, for the few variables indicating the possibility that the assumption was not met, a linear interaction with elapsed time was introduced. The only variable where this yielded a significant interaction was age at the start of the relationship (for the category ‘before age 30’). This interaction was therefore kept in the models.
Control variables: age at separation, presence of children from the previous union, past marital status, duration of the previous relationship, level of education, having continued to live with the ex-partner after deciding to separate, material and emotional consequences of the separation, and the order of the previous relationship.
The same model as described above, but including an interaction term between sex and the presence of children (negative effect, significant at the 1% level for women) while stratifying by both the presence of children (negative effect, significant at the 1% level for women) and the absence of children (no significant difference between men and women).
The uncommon situation where the father lives with his children has a negative but nonsignificant effect. Note that this effect becomes significant when the woman’s family situation is not included in the model (model not presented).
Model not presented.
It is not possible in a survey to ask a respondent to describe their partner’s previous relationship and the context of the corresponding separation.
These characteristics are interrelated: the mean duration of the previous relationship is much longer among those previously married (12 years) than among those not previously married (5 years).
An interaction term between the duration of the previous relationship and past marital situation is significant at the 5% level for past relationships that lasted at least 10 years (model not presented).
In this case, the assumption of proportionality in the Cox model is not verified for this category, thus reducing the quality of the model.
Although the level of income is known only at the time of the survey, we tested its effect. The same tendency emerged: women with the lowest incomes have a higher hazard.
The information comes from the question item: ‘Have you ever broken up with [person’s name] and then subsequently resumed your relationship? Yes, once. / Yes, multiple times. / No.’
To be fully rigorous, the duration of the breakup(s) should be subtracted from the relationship, but this information is unknown; as such, this variable is included as a control.