Couple Formation is Prolonged not Postponed. New Paths to Union Formation in Contemporary France

Abstract

In France, as in other European countries, the age at first cohabiting union has risen over the past decades, as a result of longer school enrolment, structural economic changes, and new family norms. While the median age at first co-resident couple was 23.8 for men born in France in the beginning of the 1950s, it was 26.0 among the generation born in the beginning of the 1970s. This tendency is often referred to as a postponement of couple formation and as a part of a broader delay in the transition to adulthood. This article argues, on the contrary, that couple formation has not been postponed but prolonged. In fact, age at first couple formation has remained stable across generations born since the mid-twentieth century in France. Starting from there, we take on a biographical approach to examine the nature, duration, and articulation of the successive stages that make up young people’s conjugal trajectories in France. What are the different pathways into couple life, and how have these changed over time? In order to answer these questions, we use optimal matching methods to identify ideal typical trajectories and then logistic regressions in order to see how these relate to generational differences as well as sociodemographic characteristics. We observe three traditional, three timeless and five new paths to couple life. The main historical change is the increasingly gradual nature of union formation, a trend that reflects a dual pattern. First, unions are progressively institutionalized: the time laps between different relationship stages, such as “going out,” “settling in,” and eventually marrying, have expanded. Second, young people increasingly experience several relationships during youth: the different steps of couple formation are taken with different partners. We conclude that couple formation is not delayed per se; it is rather the material and institutional formalization of unions that is put off for the future. We discuss the scientific and methodological implications of this finding.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10680-022-09629-0.

Introduction

Young adults today settle in together at older ages than did former generations. This trend is clear from numerous European studies interested in the timing of union formation, which show an increase in the age at which men and women enter cohabitating relationships for the first time. In France, which is the focus of this article, this historical evolution is both well documented and clear (Galland, 1995, 2000; Robert-Bobée & Mazuy, 2005). Whereas for the generation born in the 1950s, the median age at first cohabitation was 23.8 for men and 21 for women, for the generation born in the early 1970s, it had risen to 26 for men and 23.7 for women (Prioux, 2003).

This article seeks to go beyond this general and well-known trend in order to examine the new pathways into couple life. Rather than studying the timing of cohabitation, it adopts a broad biographical approach to young people’s couple trajectories. This leads us, on the one hand, to consider the different stages of couple formation (from starting a relationship to eventually marrying) and, on the other hand, to take into account the complexity and reversibility of contemporary life course (separation and couple (re)formation).

This processual approach to couple formation is made possible by the innovative EPIC survey carried out in France in 2013–2014 (Rault et al., 2019). Based on a representative sample of individuals aged 26 to 65 years old, the survey allows for studying different stages of union formation and accounts for consecutive unions. Our study focuses on individuals aged 30 or more, and looks at the conjugal trajectories up to age 29. We use sequence analysis methods to describe the different pathways into couple life and then deploy logistic regressions in order to determine how these pathways relate to individual characteristics. We put special emphasis on generational differences in order to determine how pathways to couple life have changed over time.

The biographical perspective adopted in this study stands in sharp contrast to the static approach usually applied to the study of first unions, and it leads to a new understanding of this event. We argue that the change in young people’s couple formation is better understood in terms of a gradual entry to couple life rather than a deferral as is commonly claimed. Union formation has been prolonged, not postponed.

Background

The study of first union formation has mainly been carried out within the broader framework of research on the transition to adulthood. Over the past decades, the various events that traditionally characterize the transition to adulthood occur at increasingly older ages. A now well-documented demographic literature gives detailed information on the evolution of the transition to adulthood, the duration of the events involved in this transition, and the order in which they occur. In Europe, the process has been studied by Billari and Liefbroer (2010), who show that the transitional patterns have three characteristics. The first is the postponement of a certain number of events related to residential and financial independence as well as family formation. In France, Olivier Galland (1995, 2000) showed an overall delay of five events among generations born in the 1960s and 1970s: completion of education, departure from the parental home, first job, first cohabiting union, and first stable job. The delay in first union formation and marriage in France has also been stressed by Hoem et al. (2009) and Sobotka and Toulemon (2008) and has been shown to correspond to a greater dispersion of the median age at first marriage (Billari & Liefbroer, 2010). The second characteristic is the greater temporal distance between the first and the last event. In other words, where certain events (such as entry into parenthood) are postponed longer than others (such as residential autonomy, which comes at fairly early ages in France, for example), the transitional process lengthens, in contrast to the post-war years when the different demographic events often occurred concurrently (Pailhé et al., 2012). The third characteristic is the greater diversification of trajectories. Today, there is no longer one but multiple trajectories leading to adulthood in Europe (Billari, 2004; Buchmann & Kriesi, 2011), events occurring at different times and in a different order (Perelli-Harris et al., 2012; Shanahan, 2000). In the USA, Aassve et al. (2002) have shown that trajectories to adulthood are becoming increasingly complex. Robette (2020) recently demonstrated that this observation holds true for France, stressing an increase in the complexity of adulthood trajectories between generations.

It is possible to add a fourth characteristic to Billari's and Liefbroer's (2010) list, namely that the events themselves have changed. Some have become more widespread or, on the contrary, become rarer. One of the major changes is the decline of marriage (Prioux, 2005; Sobotka & Toulemon, 2008), which is no longer a key event in the transition to adulthood in Europe (Lesnard et al., 2016) as it is common today to form a couple without being married (Heuveline & Timberlake, 2004; Kiernan, 2002; Perelli-Harris et al., 2012) or even without living together (de Jong Gierveld, 2004; Haskey, 2005; Levin, 2004).

This broad historical movement has different sources, first of all institutional and material ones. School enrolment has been extended in Europe since the 1960s leading to an extension of the period identified as “youth” (Roser & Ortiz-Ospina, 2013). More recently, young adults have faced increasing difficulties entering the labor market, as well as greater job insecurity and high unemployment in many countries (Brunner & Kuhn, 2014; Kahn, 2010; Oreopoulos et al., 2012; Schwandt & von Wachter, 2019; Wachter, 2020). With regard to couple formation, studies show that education has a postponing effect on cohabitation. The effect of economic uncertainty appears more contrasted. Whereas in many countries, as in France, employment uncertainty appears to postpone union formation (Blossfeld et al., 2006; Bolano & Vignoli, 2021; Jalovaara, 2012; Landaud, 2021; Winkler-Dworak & Toulemon, 2007), some studies has found job instability and precariousness to delay marriage but actually favor cohabitation, as in Italy and in Australia (Bukodi, 2012; Vignoli et al., 2016). Other social forces are also at work however. Concomitant to economic uncertainty, many authors stress the rise in values of gender equality, associated with increased female education and labor force participation (Oppenheim Mason & Jensen, 1995), and a broad cultural shift toward greater individualization, favoring attitudes of self-fulfillment and individual choice (Sobotka, 2008). In Demography, these cultural and ideational transformations are commonly understood as key factors behind the “second demographic transition” and as an important element for understanding the changing patterns of couple formation (Lesthaeghe, 2010; De Kaa, 1987).

The broad framework of the transition to adulthood has thus given a theoretical frame for understanding the dynamics of first union formation and valuable information on the timing of this event. However, this framework also comes with limitations, both analytical and methodological. Indeed, studies on the transition to adulthood traditionally focus on thresholds, i.e. the timing of a certain event, such as the age at first cohabiting union. This fails to take into account the gradual nature of couple formation—in two ways.

First, the threshold perspective is a static one. This approach is undoubtedly appropriate for certain time-limited events such as leaving the parental home or completion of education that can indeed be conceptualized as an event. Yet other stages of the transition to adulthood rather take the form of a process. This is increasingly the case of labor market integration as access to stable jobs has become long and more uncertain (Blossfeld et al., 2008). It also holds true for couple formation. Whereas in the mid-twentieth century the beginning of a union often coincided with marriage and living under the same roof, we know that direct marriage has decreased in favor of premarital cohabitation (Kuperberg, 2019; Sobotka & Toulemon, 2008). Research on the respective ages at first cohabitation and first marriage indicates a gradual entry to couple life (Manning et al., 2014; Sassler et al., 2018) but, as it focuses solely on timing, it does not allow for describing what this new process might look like.

Second, the threshold perspective does not take into account the reversibility of events. Many of the stages of the transition into adulthood are not definitive. Young people may move back in with their parents, loose their job or separate for instance. With regard to union dissolution, we see an increase in separations among young adults followed by an increase in repartnering. In France, 19% of women and 16% of men born between 1978 and 1982 had experienced two couple relationships by the age of 30, compared with only 5 and 3% of women and men born in the 1950s (Rault and Régnier-Loilier, 2015). This has lead Marteau to speak of separation as a new stage in contemporary transition to adulthood (Marteau, 2021). Indeed, rather than a mere “failure,” union dissolution can be understood as a part of a sexual and romantic experimentation that hence characterizes youth.

This means that the gradual entry to couple life applies both at the union level (stages in a relationship, such as forming a couple, later moving in together, and eventually marrying) and at the life course level (stages in a trajectory, as in multiple consecutive unions). This progressive, and ultimately prolonged, entry to couple life is the main hypothesis of this article. Starting from there, we seek not only to prove, but also to describe, this gradual process in France.

Aim and Approach

Whereas the literature on young people’s union formation traditionally focus on the age at first cohabiting or married union, this study aims to go further in describing the nature and dynamics of young people’s relationship experiences. The overall aim is to show the different pathways to couple life and how these have changed over time. We argue that first couple formation has become prolonged rather than postponed. Also, we seek determine patterns in the new ways of entering couple life. Although many studies conclude that there has been a diversification of trajectories, very few describe what the new trajectories actually look like. Increasing individualization and diversification do not necessarily mean that there are no patterns at all, or no group-specific tendencies.

To that end, we adopt a life course perspective. Billari (2001b) was one of the first to highlight the value of using a longitudinal approach in order to study the transition to adulthood. In the same vein, Robette (2010, 2020) has shown the limitations of analysis based on survival curves (often used in the literature on the transition to adulthood in general, and on first union formation in particular) as this type of analysis does not allow for a correct account of the heterogeneity of individual transitions to adulthood. To bypass these limitations, we take a holistic perspective and focus on the diversity of young peoples’ trajectories. In order to do this, we take a twofold approach.

First, we adopt a broad definition of “couple relationship” that accounts for the various phases of union formation, starting from “going out” (starting a relationship) to “settling in” (beginning cohabitation), and eventually “getting married.” This means broadening the scope in order to include couples “living apart together” (Regnier-Loilier et al., 2009). To do so, we rely on a subjective definition of the couple: Young adults are considered to be in a couple when they declare to be so, independently of their living and marital arrangements. The analysis then focuses on sequences in the process of couple formation, by looking at the respective dates of entry into a relationship, entry into cohabitation, and entry into a civil union and/or marriage.

Second, we take into account consecutive relationships. More than just acknowledging the reversibility of life events, this is important to accurately capture the gradual entry to couple life. Indeed, the first relationship may not be the first cohabiting or married relationship: these different steps may be taken with different partners.

To analyze trajectories in their complexity, we use sequence analysis methods (Abbott & Hrycak, 1990). These methods are based on the calculation of a measure of distance between pathways allowing the observation of similarities and differences, from which a typology can then be deduced. As the number of typical trajectories is limited, regularities between trajectories can be identified and studied. Using logistic regressions, we then seek to determine the extent to which the ideal types of couple formation trajectories relate to the socioeconomic characteristics of the individuals, as well as the generation to which they belong. What are the different paths into couple life? How have these changed? We use the French EPIC survey to answer these questions. The main hypothesis is that entry to couple life has not so much been postponed as it has become prolonged, as couple institutionalization folds out progressively, in different relationship stages, and with different partners.

Data

The EPIC survey on individual and conjugal trajectories (Étude des parcours individuels et conjugaux) was carried out in France in 2013–2014 by the French institute for demographic studies (INED) and the national institute for statistics and economic studies (INSEE). The survey traces the trajectories of 7825 men and women aged 26 to 65 years old (see Rault et al. (2019) for a detailed presentation of the survey).

Historically, surveys on couple formation have focused on marriage (see for example Girard, 2012 in France). Over time, this definition has widened and other forms of union have gradually been addressed, starting with cohabiting couples (Bozon & Heran, 1989) and followed by those declaring themselves to be in a couple without living together (Rault et al., 2019). These definitions are often indicative of a historical context and the legitimacy of different forms of couple arrangements (Rault et al., 2019). The originality of the EPIC survey is that rather than using a pre-determined definition of what is a “couple,” which would likely exclude certain relationships, it relies on a subjective definition, the respondents being asked whether they were or had been in a couple or a “serious romantic relationship.” Those answering yes were asked at what date they entered the relationship, whether they had lived with this person, and whether they had entered a civil union or marriage with this person. Each time, the timing of these different relationships stages was registered. This makes it possible to account for different couple forms and sequences in the individual trajectories. Also, the survey does not focus on a specific union, respondents instead being asked about the full set of their “couple relationships or serious romantic relationships” over their lifetime and describing each of these relationships. Through this approach, intimate trajectories can be studied in their diversity and complexity.

The survey also has limitations, the first one being the risk of recall bias. For instance, it could be that certain relationships in the beginning of the life course are underreported by older generations who either forget to mention them or no longer consider them to be important, unlike younger generations for whom they are still significant because closer in time. However, we observe a similar age at first couple or serious romantic relationship across generations (see Sect. 6.1), and a significant minority of the oldest generation declare first experiences that were not cohabitant or not long-lasting.1 Although we cannot control for recall bias, which could be particularly common in the oldest generation, these observations are reassuring as there is no obvious indication of underreporting of early relationship experiences among the oldest generation.

To analyze the couple formation trajectories of young adults, we restricted our sample to relationships experienced before age 30, and focused on the trajectories between 18 and 29.2 In addition, we limited the sample to people aged 30 or over at the time of the survey with a view to obtaining comprehensive information on the respondents’ experiences before age 30. The size of our sample is 7222 people. We use weights calculated by INED and INSEE when analyzing the data.3

Methods

To describe the trajectories of couple formation among young adults, we use sequence analysis methods (Abbott, 1995; Abbott & Tsay, 2000; Billari, 2001a; Billari & Piccarreta, 2005). The aim is to go beyond traditional approaches focused on the occurrence of a single event to consider the trajectories as a whole, with their continuities and disruptions. This allows us to describe the diversity of trajectories and to produce a typology of pathways into couple life.

Optimal Matching

Sequence analysis is based on transcribing individual trajectories into sequences of ‘states’ (Abbott & Forrest, 1986). A state is the young person’s couple situation in a given month of their trajectory. We looked at the couple life of respondents from 18 to 29 years old, coding their experiences month by month into nine different states. These states take into account two variables: the type of relationship and the order of the relationship (first union or subsequent union). On the one hand, by taking into consideration all at once cohabiting couples, couples that are married or in a civil union, and couples that “live apart together,” we can be very precise in the description of the different types of relationship. On the other hand, while second unions were fairly common, third and fourth unions were less so before age 30,4 thus leading us to group together all consecutive unions that follow a first couple relationship.5 We also grouped together couples who were civilly married or in a civil union (called “civil solidarity pact,” or “PACS,” in France) as we were interested in the official formalization of the relationship rather than the specific form of this formalization. However, we isolated religious marriage as this form of marriage has declined in France and increasingly concerns a specific group in terms of demographics, behaviors and values (Maillochon, 2019). As it is extremely rare (and sometimes impossible) to religiously marry twice, no distinction was made between first and consecutive religious marriages. Thus, we obtained nine states: single, LAT 1 (for first union), LAT 2 (for subsequent union), cohabiting couple 1, cohabiting couple 2, civilly married 1, civilly married 2, religiously married, and separated. The succession of these states over time forms the couple life trajectory during youth.

The distance between two trajectories is calculated as the minimum number of operations required to transform one sequence into another. There are therefore two main types of operations: substitution (replacing one state by another) and insertion-deletion (inserting or deleting a state). The choice to use these two types of operations is a matter of time arbitrage. Insertion-deletion (indel) operations deform the temporal structure of the sequences to identify the succession of events composing the trajectories, while substitution operations preserve the temporal structure of the sequences in order to compare the events with a constant date (Lesnard & De Saint Pol, 2006). We use substitution costs based on the transitions between states (Rohwer & Pötter, 2005). This amounts to considering that the more frequent a transition between two states is in the data, the lower the substitution cost between these states will be (the transition having a higher probability of occurrence). As mentioned by Aassve et al. (2007), there is no clear rule to determine what should be the substitution cost between being in union or not being in union. For this reason, substitution costs based on the transition between these states are useful in our case. Moreover, we have chosen to set the cost of an indel operation at 1 as part of an intermediate approach, allowing a consideration of both the contemporaneity of sequences and the order of states. The same choice has been made in several articles on related subjects (Aassve et al., 2007; Widmer & Ritschard, 2009).

Clustering

After calculating the distances between sequences one by one, we used clustering techniques to identify and classify “groups” of individuals having similar trajectories. To choose the algorithm and the number of groups of the typology, we use a combined approach based both on statistical indicators and on a more in-depth exploration of the data.6 We carried out a statistical analysis to measure the quality of different algorithms and groupings (from 2 to 15 groups). By evaluating the clustering algorithms using the ASW, RHC and PBC (Studer, 2012) quality measures, the use of a partitioning around medoids (PAM) algorithm in 11 groups appears to be the best solution in statistical terms as it constitutes local maxima for each of the quality measures. A more qualitative analysis7 showed that by reducing the number of groups, all LATs are grouped together in the same group, whereas here we have a better consideration of the diversity of situations8 that is under-represented in the quantitative literature.

Multinomial Logistic Regressions

We use multinomial logistic regressions to identify which sociodemographic characteristics of young people explain the fact of being in one ideal typical trajectory of couple formation rather than another. As Fulda (2016) has pointed out on the basis of German data, partnership trajectories during the transition to adulthood are highly related to individual characteristics. We begin by assessing the generational effect in order to determine changes over time. We also include a series of other variables that may differentiate pathways into couple life. First, we seek to determine the effect of educational attainment that is an important factor for explaining differences in the transition to adulthood (Robette, 2010). Because credentials differ between generations, we use an indicator of relative educational level developed by Ichou (2014) that classifies the respondents in four groups: those belonging to the 25% (of their gender and generation) with the lowest educational level, those belonging to the 25% with the highest educational level, and two intermediate groups. Second, as the social background of young people has an influence on their transition to adulthood (Schoen et al., 2009; Sironi et al., 2015), we also consider the socio-professional category of the two parents. Parental occupation was coded following Erikson’s and Goldthorpe's typology, using 8 classes (Erikson & Goldthorpe, 1992), and then ranked according to the highest occupation.9 For instance, if the mother’s occupation belongs to the category “higher professionals and managers” (I) and the father’s to “lower professionals and managers” (II), it is the mother’s occupation that prevails. Third, we assess the effect of parenthood. As shown by Jalovaara and Fasang (2017), parenthood is a central component in the analysis of union trajectories (including the fact of establishing a partnership and the stability of cohabitation). Lastly, we also consider religious orientation, age at first sexual intercourse and migration status in our regressions, as we believe that these variables may be of importance to differentiate between trajectories. Because the sociodemographic characteristics of men and women impact their couple trajectory differently (Jalovaara & Fasang, 2015; Pailhé et al., 2012), we have opted for separate regressions by gender.10

The multinomial logit model for couple formation clusters is as follows:

$$\mathrm{Prob}\left(Y_i,=j|,w_i\right)=\frac{\exp (w_i^{{}^{\prime }}{\alpha }_j)}{1+{\sum }_{k=1}^J\exp (w_i^{{}^{\prime }}{\alpha }_k)},j=0,\cdots ,J$$ 1

where j is the couple formation trajectory follow by the individual i. We estimate the marginal effects of the different sociodemographic variables $w_i$ on the probability that the individual i follow the pathway $\mathrm{j}$ (see Greene, 2018 p. 830 for details):

$${\mathrm{\delta }}_{\mathit{ij}}=\frac{{\partial P}_{\mathit{ij}}}{{\partial w}_i}=P_{\mathit{ij}}\left[{\alpha }_j,-,\sum _{k=0}^J,P_{\mathit{ik}},{\alpha }_k\right]=P_{\mathit{ij}}\left[{\alpha }_j,-,\overline{\alpha }\right]$$ 2

The advantage of using marginal effects (instead of calculating coefficients in terms of multinomial log-odds or relative risk ratios) is to express the results in terms of differences in probabilities independent of the reference group. Marginal effects measure the effect that a change in an explanatory variable $w_i$ has on the predicted probability $Y_i=j$ while keeping all the other covariates constant.

Results

Typology of Trajectories

We begin by taking a general overlook at the evolution of couple trajectories between generations (see Fig. 1). To do so, we distinguish between individuals born between 1948 and 1959 (generation 1), those born between 1960 and 1969 (generation 2) and those born between 1970 and 1984 (generation 3). An initial observation pointing to a more gradual entry to couple life is the fact that the share of couples cohabiting over the period increased significantly between the first generation and the last one. In addition, a greater proportion of young people have had more than one relationship before the age of 30. We can also see that the share of initial singlehood, before entering a couple relationship, has remained relatively stable across generations. This is a first indication that first couple formation has actually not been postponed over time; it is rather the way of entering relationships that have changed. In fact, the proportion of young people who, at each age between 18 and 29, has experienced a couple relationship is strikingly similar across generations (see Fig. 2). However, consistently with the existing literature, the age at first cohabitation is different as it has increased over time (see Fig. 2). In other words, it is not the first union per se that has been delayed, but the moment when young adults decide to move in with their partner.

Fig. 1.

Transversal distribution of couple states among young people, by generation

Fig. 2.

Cumulative percentage of individuals who have experienced a couple relationship and cohabitation, by age and generation

When moving on to the cluster analysis, we first of all observe remarkably clear differences in pathways into couple life. The 11 clusters indeed have distinct features that we will briefly describe here. The next section will present each of these clusters in more detail and at the same time present the generational and sociodemographic characteristics associated with them.

Cluster 1 accounts for 14% of the sample and we named it Young married couples as these are individuals who experience a single romantic relationship characterized by marriage. The second cluster (12%) are Gradually married couples who are similar to those in the first cluster, except that the institutionalization is more progressive. The couples in the third cluster (5%) are Married in the eyes of God, meaning the trajectory is characterized by early and religious marriage. Cluster 4 (14%) is made up of Perpetual singles who spend their youth in singlehood. Cluster 5 (5%) are couple trajectories predominantly characterized by relationships Living Apart Together. Cluster 6 (7%) comprises Separated then married individuals, meaning individuals who settle in and rapidly marry with a second partner. Cluster 7 (10%) is made up Cohabiting first unions, whereas cluster 8 (8%) comprises Cohabiting second unions; both these groups are made up of young adults that spend most of their youth in cohabitation, but with different partners. Cluster 9 (7%) involves individuals who are Separated with a second gradual union, meaning that these individuals start over with a second partner and then take things slowly. Cluster 10 (10%) are Late bloomers; these individuals stay single for a long time (just as in cluster 4) before entering couple life. Lastly, cluster 11 (9%) comprises Singles with experience, referring to individuals who separate and then go back to long singlehood. Table 1 presents the descriptive statistics for all of these different groups.

Table 1.

Distribution of generations among cluster

Ideal typical trajectories	Generation			Total
	1948–1959	1960–1969	1970–1984
Traditional pathways of couple formation
Young married couples	30%	12%	3%	14%
Gradually married couples	18%	11%	8%	12%
Married in the eyes of God	6%	5%	3%	5%
Timeless trajectories of couple formation
Perpetual singles	13%	16%	13%	14%
LAT	5%	4%	5%	5%
Separated then married	7%	9%	6%	7%
New couple formation approaches
Cohabiting first unions	3%	11%	15%	10%
Cohabiting second unions	2%	7%	13%	8%
Separated with a second gradual union	3%	6%	11%	7%
Late bloomers	6%	11%	12%	10%
Singles with experience	6%	8%	11%	9%

Change in Couple Formation over Time

The results of the sequence analysis and the clustering11 (Fig. 3) allow us to carry out a fine-grained analysis of how the way young people enter couple life has changed over time. We identified three sets of trajectories depending on the generation. In Table 2, we present the distribution of generations among clusters. In the first group, the share of young people from generation 1948–1959 is much higher than their share in the sample. For example, 30% of the young people of the generation 1948–1959 have trajectories belonging to cluster 1, “Young married couples,” whereas the share of this cluster in the sample is only 14%. On the contrary, the clusters grouped in the second group have the particularity of being invariant in time and of characterizing all three generations. These trajectories of couple formation were common yesterday and are still so today. Finally, the last group is characterized by couple formation trajectories that are more common among young people of generation 1970–1984. For example, while “cohabiting first unions” represent 10% of the sample, they account for 15% of the trajectories of the youngest generation.

Fig. 3.

Transversal distribution of couple states among young people, 11 couple formation clusters

Table 2.

Descriptive statistics of the 11 ideal types of couple formation trajectories

	Traditional pathways of couple formation			Timeless trajectories of couple formation			New couple formation approaches
	Young married couples	Gradually married couples	Married in the eyes of God	Perpetual singles	LAT	Separated then married	Cohabiting first unions	Cohabiting second unions	Separated with a second gradual union	Late bloomers	Singles with experience	Total
Gender
Women	70%	46%	64%	34%	48%	62%	55%	55%	53%	41%	43%	51%
Men	30%	54%	36%	66%	52%	38%	45%	45%	47%	59%	57%	49%
Generation
1948–1959	68%	48%	43%	31%	31%	32%	9%	10%	14%	20%	23%	32%
1960–1969	25%	27%	30%	33%	27%	36%	33%	28%	24%	33%	27%	29%
1970–1984	8%	25%	26%	36%	42%	32%	58%	62%	62%	47%	50%	38%
First child before age of 30
Yes	94%	87%	92%	10%	39%	89%	71%	60%	35%	43%	21%	58%
No	6%	13%	8%	90%	61%	11%	29%	40%	65%	57%	79%	42%
Age at first sexual relation
< 18	50%	40%	36%	36%	40%	61%	58%	67%	55%	36%	48%	47%
≥ 18	50%	60%	64%	64%	60%	39%	42%	33%	45%	64%	52%	53%
Nationality at birth
French	87%	88%	77%	79%	84%	94%	92%	95%	92%	87%	90%	88%
Other	13%	12%	23%	21%	16%	6%	8%	5%	8%	13%	10%	12%
Religion
Catholic - religion declared important	22%	22%	32%	16%	17%	17%	14%	14%	17%	18%	19%	19%
Catholic - religion declared not important	50%	46%	34%	33%	38%	54%	41%	43%	42%	42%	39%	42%
Muslim	8%	5%	20%	16%	12%	2%	3%	3%	3%	6%	8%	8%
Other religion	4%	6%	7%	7%	4%	4%	4%	3%	4%	5%	5%	5%
No religion	16%	21%	6%	29%	28%	23%	38%	37%	34%	29%	29%	26%
Relative level of education
Low degree	30%	23%	28%	33%	18%	16%	29%	23%	12%	21%	21%	24%
Moderately low degree	37%	30%	27%	26%	29%	35%	32%	26%	25%	27%	27%	29%
Moderately high degree	20%	25%	22%	18%	19%	24%	21%	22%	24%	20%	21%	21%
High degree	14%	22%	23%	22%	34%	25%	18%	29%	39%	33%	30%	25%
Parent’s occupation
I Higher Professionals and Managers	6%	10%	15%	14%	16%	12%	13%	20%	24%	19%	19%	14%
II Lower Professionals and Managers	5%	8%	9%	8%	9%	14%	8%	11%	13%	11%	15%	10%
IIIa Higher routine non-manual	12%	14%	10%	11%	15%	13%	18%	16%	15%	14%	13%	13%
IIIb Lower routine non-manual	5%	3%	4%	4%	4%	4%	5%	5%	4%	4%	5%	4%
IV Proprietors, artisans, farmers	25%	26%	20%	24%	24%	23%	17%	16%	16%	18%	19%	21%
V Lower technicians & supervisors	6%	7%	3%	6%	4%	7%	7%	6%	8%	8%	7%	6%
VI Skilled manual	28%	21%	26%	20%	20%	18%	23%	18%	15%	18%	16%	21%
VII Unskilled manual	10%	9%	11%	11%	6%	7%	6%	4%	3%	4%	6%	7%
Inactive or missing	3%	3%	1%	4%	3%	1%	4%	3%	1%	3%	1%	3%
Observations	14%	12%	5%	14%	5%	7%	10%	8%	7%	10%	9%	100%

We use entropy to characterize each group. Figure 4 displays the distribution plot of the entropy of each class. Entropy represents the frequency and distribution of the durations of states of sequences. It is equal to 0 when the individual trajectory does not experience variability and equal to 1 when the individual trajectory experiences all states with similar durations.12 A group with low entropy is therefore a group whose trajectories have little variability (young people frequently remain in the same state throughout their trajectory). Conversely, a group with high entropy has trajectories that are characterized by a high diversity of states that make up the trajectories (young people often change their relationship status). Figure 4 displays entropy using box plot, which shows the median entropy level of the cluster in the center of the box, and the first and third quartiles of entropy at the ends of the box.

Fig. 4.

Entropy of each cluster

In the following section, we characterize each group using descriptive statistics (Table 2) and the results of multinomial logistic regression (Table 3).13 By reporting the results of the multinomial logit models in terms of marginal effect, we can express them in terms of probability according to couple formation clusters.

Table 3.

Regression results (multinomial logistic regression). Effects of sociodemographic characteristics on ideal types of couple formation trajectories

	Traditional pathways of couple formation						Timeless trajectories of couple formation
	Young married couples		Gradually married couples		Married in the eyes of God		Perpetual singles		LAT		Separated then married
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
Generation (Ref. 1948–1959)
1960–1969	− 0.168***	− 0.0401***	− 0.0701***	− 0.0900***	− 0.0166*	− 0.00746*	− 0.0223**	− 0.0232	− 0.0207	− 0.0329**	0.00762	− 0.0193*
1970–1984	− 0.342***	− 0.110***	− 0.0850***	− 0.142***	− 0.0454***	− 0.0133***	− 0.0345***	− 0.0648***	0.000197	− 0.0279*	− 0.00469	− 0.0378***
First child before age of 30 (Ref. No)
Yes	0.242***	0.0925***	0.136***	0.264***	0.0881***	0.0311***	− 0.176***	− 0.360***	− 0.0660***	− 0.0410***	0.120***	0.106***
Age at first sexual relation (Ref. ≥ 18)
< 18	0.0670***	0.00406	− 0.0626***	− 0.0298**	− 0.0101	− 0.00342	− 0.0400***	− 0.0551***	− 0.0243**	− 0.00458	0.0450***	0.0252**
Nationality at birth (Ref. French)
Other	0.0307	−0.00915	0.0736***	−0.0355	0.00256	0.00645	0.0240*	0.0338	0.0285*	0.00675	0.00243	-0.0168
Religion (Ref. No religion)
Catholic - religion declared important	0.0349*	0.0151	0.0210	0.0450**	0.0633***	0.0380***	− 0.0145	− 0.0211	− 0.0113	− 0.0102	0.000995	− 0.00387
Catholic - religion declared not important	0.0440**	0.0202***	0.0270	0.0175	0.0198	0.0317***	− 0.0240**	− 0.0332*	− 0.0173	− 0.00525	0.0267**	0.0170
Muslim	0.104***	0.0461***	− 0.0522	0.0683	0.123***	0.0417***	0.0618***	0.0846**	0.0400**	0.0426	− 0.0306	− 0.0437
Other religion	0.0529	0.0180	0.0210	0.0855**	0.0816***	0.0229	0.000591	0.0252	− 0.0386	− 0.00739	0.0145	0.0270
Relative level of education (Ref. High degree)
Low degree	0.0543**	0.0136	− 0.0678***	− 0.00158	− 0.0185	− 0.000376	0.0579***	0.134***	− 0.00524	− 0.0646***	− 0.0471***	− 0.0306*
Moderately low degree	0.0856***	0.0165**	− 0.0248	0.0131	− 0.00282	− 0.00529	0.0214*	0.0565***	− 0.00513	− 0.0170	0.000947	− 0.00743
Moderately high degree	0.0632***	0.0118	0.0156	0.0248	0.00494	0.00210	0.00559	0.0390*	− 0.0147	− 0.0306*	0.000528	− 0.00770
Parent’s occupation (Ref. I Higher Professionals and Managers)
II Lower Professionals and Managers	0.0158	0.0205	0.0400	0.0374	− 0.0165	0.0110	− 0.0319**	− 0.0243	− 0.0229	0.0163	0.0251	0.0467**
IIIa Higher routine non-manual	0.0311	0.0249*	0.0345	0.0540*	− 0.0352**	0.00367	− 0.0296*	− 0.00993	0.00601	0.0230	0.0121	− 0.00915
IIIb Lower routine non-manual	0.0991***	0.0114	0.0321	− 0.0373	− 0.0136	− 0.00588	− 0.0118	− 0.0354	0.0174	− 0.0432	0.0207	0.00505
IV Proprietors, artisans, farmers	0.0757***	0.00877	0.0396	0.0598**	− 0.0163	− 0.00537	− 0.00781	0.0280	0.0134	0.0125	0.0255*	− 0.0119
V Lower technicians & supervisors	0.0675**	0.0360**	0.0499	0.0513	− 0.0641***	− 0.00434	− 0.00587	− 0.0123	− 0.0460*	− 0.0233	0.0351*	− 0.0234
VI Skilled manual	0.104***	0.0265**	0.0261	0.0613**	− 0.00733	0.00621	− 0.0311**	0.0238	− 0.00278	0.0174	0.0102	− 0.0136
VII Unskilled manual	0.0666**	0.0341**	0.0600	0.0809**	− 0.00416	0.00675	− 0.00475	0.0593	− 0.0306	0.0286	0.0125	0.0218
Inactive or missing	0.176***	0.0163	0.218***	0.00696	− 0.00154	− 0.198***	0.0588*	0.110**	0.0856***	0.0142	− 0.855***	− 0.00998
Observations	763	282	459	429	220	101	389	595	193	162	355	186

	New couple formation approaches
	Cohabiting first unions		Cohabiting second unions		Separated with a second gradual union		Late bloomers		Singles with experience
	Women	Men	Women	Men	Women	Men	Women	Men	Women	Men
Generation (Ref. 1948–1959)
1960–1969	0.141***	0.111***	0.0733***	0.0692***	0.0382**	0.0121	0.0505***	0.0306	− 0.0134	− 0.0101
1970–1984	0.191***	0.142***	0.144***	0.101***	0.0939***	0.0720***	0.0742***	0.0468*	0.00780	0.0335*
First child before age of 30 (Ref. No)
Yes	0.0264*	0.0817***	− 0.0212*	0.0417***	− 0.107***	− 0.0225*	− 0.0858***	− 0.0280	− 0.157***	− 0.165***
Age at first sexual relation (Ref. ≥ 18)
< 18	0.0373***	0.0126	0.0529***	0.0606***	0.0277***	0.0250**	− 0.101***	− 0.0423**	0.00788	0.00770
Nationality at birth (Ref. French)
Other	− 0.00766	− 0.0115	− 0.112***	− 0.00907	− 0.0223	0.0174	0.00392	0.0446	− 0.0235	− 0.0269
Religion (Ref. No religion)
Catholic - religion declared important	− 0.0572***	− 0.0482**	− 0.0335**	− 0.0257	− 0.00893	− 8.88e-06	− 0.00347	− 0.00846	0.00871	0.0194
Catholic - religion declared not important	− 0.0269*	− 0.0368**	− 0.0160	− 0.00812	− 0.00557	− 0.00576	− 0.00564	− 0.00585	− 0.0221*	0.00853
Muslim	− 0.190***	− 0.0999**	− 0.0687	− 0.0800*	− 0.0401	− 0.0845	0.00518	− 0.0337	0.0472*	0.0585*
Other religion	− 0.0479	− 0.0680	− 0.0432	− 0.0486	− 0.0197	− 0.0613	− 0.0327	0.0272	0.0113	− 0.0206
Relative level of education (Ref. High degree)
Low degree	0.0577***	0.0155	− 0.0161	0.000525	− 0.0333*	− 0.0855***	− 0.0101	− 0.00301	0.0283*	0.0225
Moderately low degree	0.0148	0.0199	− 0.0336**	− 0.0215	− 0.0356**	− 0.0350**	− 0.0319*	− 0.0174	0.0112	− 0.00235
Moderately high degree	0.0201	0.0219	− 0.0318**	0.00115	− 0.0283**	− 0.0198	− 0.0201	− 0.0444*	− 0.0150	0.00177
Parent’s occupation (Ref. I Higher Professionals and Managers)
II Lower Professionals and Managers	0.0149	− 0.0727**	− 0.0286	− 0.0125	− 0.00274	− 0.0113	− 0.00746	− 0.0193	0.0142	0.00813
IIIa Higher routine non-manual	0.0351	0.0145	− 0.0198	− 0.00735	− 0.0110	− 0.00732	− 0.00270	− 0.0663*	− 0.0204	− 0.0201
IIIb Lower routine non-manual	0.00811	0.0455	− 0.0677*	0.0402	0.000154	− 0.00509	− 0.113***	0.0628	0.0287	− 0.0382
IV Proprietors, artisans, farmers	− 0.00757	0.0130	− 0.0253	− 0.0338	− 0.0360**	− 0.0113	− 0.0588***	− 0.0191	− 0.00253	− 0.0407*
V Lower technicians & supervisors	0.0203	0.00956	− 0.0430	− 0.00578	− 0.0340	0.0259	− 0.0189	0.00873	0.0390	− 0.0623**
VI Skilled manual	0.0175	0.0113	− 0.0274	− 0.0351*	− 0.0316*	− 0.0183	− 0.0444**	− 0.0280	− 0.0132	− 0.0515**
VII Unskilled manual	0.00267	0.0300	− 0.0109	− 0.0813*	− 0.0365	− 0.0495	− 0.0499*	− 0.0974*	− 0.00503	− 0.0332
Inactive or missing	0.157***	0.0581	0.0154	0.0399	0.0497	− 0.0529	0.0342	0.115**	0.0625	− 0.100
Observations	426	264	319	208	325	207	338	362	321	318

Reported coefficients are marginal effects at the means (MEM). MEM reports the partial effect of a given explanatory variable $w_i$ on the predicted probability $Y_i=j$ after setting all the other covariates at their means. We use “Young married couples” as reference group

*p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001

Traditional Pathways into Couple Life

Three ideal types of couple formation pathways generally apply to the oldest generation in the EPIC survey, born in the 1950s. The couple life trajectories of this generation, also known as the “baby-boomers,” are less diverse than those of the most recent generations. 54% percent of the individuals from the 1950s generation are in one of these three clusters, each of the latter being characterized by the experience of marriage and the fact of having had only one relationship throughout youth. Being in generation 1970–1984 reduces the likelihood of being in these three clusters (− 34.2% for women with regard to cluster 1 for instance). Having a child before the age of 30 increases the chances of being in these clusters both for women and men (up to + 26.4% for men with regard to cluster 2). This is consistent with the findings of Sobotka and Toulemon (2008) who showed that married couples enter parenthood at a much higher rate.

The first cluster, “Young married couples,” includes individuals who experiment a brief LAT period before getting married. It is a common ideal typical trajectory for the oldest generation since 68% of the individuals following this path were born in the 1950s (compared to only 8% being born in 1970–1984). The individuals in this cluster get married at a relatively young age (in their early twenties). The cohabitation period is inexistant or brief, as the partners move in together after marriage. It is a very largely feminized group (70% are women) in which having a low, moderately low, or moderately high educational degree increases the probability of belonging to this cluster (from + 6.3% to + 8.6%). Likewise, having parents from lower-level occupations (different from “higher professionals and managers” (I)) increase the likelihood to be in this cluster (ranging from + 2.5% to + 3.6% for men and from + 6.6% to + 10.4% for women).

“Gradually married couples” is also a cluster with a large share of people born in the 1948–1959 cohort (48% vs. 25% of individuals from the youngest generation). Unlike the previous one, individuals in this cluster get married more gradually. Having had sexual intercourse before the age of 18 decreases the probability of being in this cluster (− 3% for men and − 6.3% for women). For men, as in the previous cluster, having parents whose occupations do not correspond to “higher professionals and managers” increases the likelihood to be in this cluster.

The third cluster includes young people “married in the eyes of God,” meaning it is characterized by religious marriage. Here too, we find an over-representation of young people from the oldest generation (43% were born in 1948–1959 vs. 26% that were born in 1970–1984). As the smallest of the 11 clusters (together with cluster 5), it is a marginal situation across all generations, but especially among the most recent ones. Quite logically, since young people in this cluster predominantly marry religiously, having a religion increases the likelihood to be in this ideal typical trajectory (from + 3.8% for Catholic men to + 12.3% for Muslim women). This result echoes the work of Regnier-Loilier and Prioux (2008) who showed that religion generates different demographic behaviors, particularly in terms of fertility, marriage and (and a less likeliness to) separation followed by repartnering.

Timeless Trajectories

Three trajectories are equally frequent among all generations; the individuals following these pathways form a balanced group of people born in the 1950s, 1960s and 1970s/early 1980s.

The first cluster, “Perpetual singles,” is composed of young people who experience a long continuous period of singlehood throughout their twenties. This cluster has the lowest variability and the lowest entropy and is relatively balanced in terms of distribution by cohort (31% were born in the 1948–1959 cohort vs. 36% in the 1970–1984 cohort). It contains a majority of men (66%) and, as one could expect, having had sexual intercourse before age 18 decreases the probability to belong to this cluster (− 4% for women and − 5.5% for men). The same observation is made for having a child before age 30 (− 17.6% for women and − 36% for men). Of greater interest, this cluster has the highest proportion of young people having experienced same-sex relationships (33% compared with 0% to 3% in all other clusters). Having a low degree or a moderately low degree (from + 2.1% to + 13.4%) and being a Muslim (+ 6.2% for women and + 8.5% for men) increase the likelihood to be in this ideal typical trajectory. These observations indicate that staying single throughout youth depends on different contrasted factors. However, contrary to what is sometimes claimed, by journalist or social scientific scholars, “going solo” does not correspond to any “extraordinary rise and surprising appeal of living alone” among young generations (Klinenberg, 2012). In fact, remaining single throughout one’s youth is not something new, but a (male) condition that persists over time. Andersson and Philipov (2002) already showed that a substantial proportion of young adults remain single during the transition to adulthood.

The second cluster, “LAT” (Living Apart Together), also goes against a popular belief, namely that non-cohabiting relationships are a recent phenomenon (Levin, 2004). This cluster is composed of young people that experience a first long LAT relationship. At the end of their twenties, they experiment different couple states (cohabitation, marriage, separation and/or subsequent unions). Our study indicates that although this pathway into couple life is more common in the recent generation, it also applies to previous generations. Indeed, this is a relatively balanced cluster in terms of cohorts (31% were born in the 1948–1959 cohort vs. 42% in the 1970–1984 cohort) but also in terms of gender (48% are women). Having a child before age 30 decreases the likelihood to be part of this cluster (− 6.6% for women and − 4.1% for men). For men, having a relatively low educational degree is a factor that decreases the probability of belonging to this cluster (− 6.5%). This last result echoes that of Liefbroer et al. (2015) whereby LAT is more common among highly educated people. Lastly, this is a cluster with a small sample size, meaning that non-cohabiting relationships are still a minority form of unions during youth. LAT relationships remain mostly a prelude to cohabitation: it is a stage in couple formation rather than a full-fledged lifestyle (Regnier-Loilier et al., 2009). This is as true for the baby boom generation as it is for those born in the 1970s and early 1980s.

The third cluster of timeless trajectories is “Separated then married.” This cluster includes young people who are in a diversity of couple situations in early life course (first relationship, first separation, second relationship, marriage) and who end up marrying a subsequent partner. The great diversity of situations in this group is consistent with the relatively high level of entropy of this cluster. Once again, in this cluster, we do not observe a clear distinction according to generation (32% were born in the 1948–1959 cohort vs. 32% in the 1970–1984 cohort). This cluster contains a majority of women (61%). Having had a first sexual experience before the age of 18 (+ 2.5% for men and + 4.5% for women) or a child before 30 (+ 10.6% for men and + 12% for women) are factors that increase the likelihood to be in this ideal type of couple formation trajectory. Similarly, having a low educational degree decrease the probability to be included in this cluster (− 3.1% for men and − 4.7% for women).

New Ways of Entering Couple Life

The last set of ideal types of trajectories is characteristic of the new ways of entering couple life. These pathways are more common among individuals born between 1970 and 1984. Thus, belonging to the youngest generation increases the probability to be in one of these 5 clusters (up to + 19.1% for women with regard to cluster 7). The number of clusters indicates alone a greater diversity of ways of entering couple life among the youngest generation; the pathways have indeed become more complex. However, this is also true with regard to the nature of the trajectories that, as we will see, are more often characterized by separations and multiple relationships. Moreover, the greater complexity is verified in the entropy of these clusters, which is generally stronger in comparison to those associated with “traditional” and “timeless” pathways into couple life. These results are in line with research on the transition to adulthood that show a greater diversity in young people’s life trajectories among the most recent generations, as shown in France by Robette (2020).

The first cluster, “Cohabiting first unions,” includes mainly young people of the most recent generation (58% were born in 1970–1984), while the oldest generation accounts for only a small share (9% were born in 1948–1959). The individuals in this cluster spend the major part of their youth in a first cohabiting union. Having a religion decreases the probability to be in this cluster (down to − 19% for Muslim women), while having a child before the age of 30 increases the likelihood to take this path into couple life (+ 2.6% for women and + 8.2% for men). Finally, women with a relatively low degree have a higher probability to be included in this cluster (+ 5.8%).

Similar to the previous cluster, the second ideal type, “Cohabiting second unions,” is a form of cohabitating youth. The youngest generation represents 62% of the individuals in this cluster compared to only 10% of the oldest one. Young people who had sexual intercourse before the age of 18 are more likely to be part of this cluster. Having a child before age 30 decreases the probability to be in this cluster for women (− 2.1%) and increases it for men (+ 4.2%). The chance of being in this cluster also decreases for women with a relatively moderate low (− 3.4%) or relatively moderate high degree (− 3.2%). In short, while the previous cluster and this one are similar with regard to the living arrangements (cohabitation), they differ in terms of trajectories and social characteristics: educated individuals, especially women, more often settle in with a second partner.

The third cluster, “Separated with a second gradual union,” has together with the previous one, the highest proportion of individuals from the most recent generation (62% are born in 1970–1984 vs. 14% of individuals born in 1948–1959). It is made up by young people forming a first union and separating at a relatively early age before forming a subsequent union. This new relationship is characterized by a very progressive couple formation and institutionalization, going from LAT, to cohabitation, to civil marriage and then religious marriage. It has the highest concentration of individuals whose parents have occupations belonging to the category “higher professionals and managers” (24%) and individuals with the highest degrees (39%). Indeed, having a relatively low or moderately low degree reduces the probability of being in this cluster both for women (respectively − 3.3% and − 3.6%) and men (respectively − 8.6% and − 3.5%). Young people having had a child before 30 are less likely to be in this ideal typical trajectory (− 2.3% for men to − 10.7% for women), while those having a first sexual experience before 18 are more likely to be included in this group (+ 2.5% for men and + 2.8% for women). To enter couple life gradually, step-by-step, appears as the most emblematic pathway for socially privileged young people in the youngest generations.

Compared to previous clusters, the generation gap is smaller in the next one that we called “Late bloomers” (47% were born in 1970–1984 vs. 20% in 1948–1959). This cluster, which is predominantly composed of men (59%), is characterized by a relatively long period of singlehood before experiencing a first relationship when approaching thirty. To have a first sexual experience before 18 decreases the likelihood to be in this cluster (− 4.2% for men and − 10.1% for women), as well as having a child before 30 for women (− 8.6%).

The last cluster is made up by “Singles with experience,” in other words by individuals who experience a long period of singlehood after an early separation. Once again, this ideal typical pathway contains fewer differences in terms of generation than previous ones (50% are born in 1970–1984 vs. 23% in 1948–1959). These young people are similar to the “Perpetual singles,” in the sense that they spend most of their youth without a partner, the main difference being that they had a short relationship that did not last. Just like the “Perpetual singles,” having a child before 30 strongly decreases the probability of belonging in this cluster (− 15.7% for women and − 16.5% for men), and the group is mostly made up of men (57%). However, an important difference with the “Perpetual singles” is that women are more numerous in this group. An interesting pattern that stands out from the comparison of all clusters is that young women who experience singlehood more often do so after having had some form of relationship experience, whereas singlehood more often takes the shape of a “waiting game” for young men: love is yet to happen.

Conclusion

The main aim of this study was to identify the changing patterns of young adult’s conjugal trajectories in France: how have the pathways into couple life changed across generations? An initial central observation is that first union formation has not been delayed in France as is commonly said. In fact, over the period covered by this study, the timing of first “couple and serious romantic relationships” is strikingly similar for all generations. The age at first cohabitation, however, differs significantly across generations: young people today move in together with a partner at older ages, in part because their first relationship broke-up before they settled in. Hence, the main historical change does not lie in the postponement of first unions, but elsewhere—in the increasingly gradual entry to couple life.

Switching from a static to a biographical approach to first union formation reveals the great diversity of pathways into couple life. The sequence analysis carried out on the French EPIC survey shows 11 distinctly different conjugal trajectories during youth. Although the complexity of contemporary life course is a well-known fact, there are still relatively few studies that set out to describe the patterns behind this variation; diversity does not imply an absence of tendencies.

We identified three traditional, three timeless and five new couple trajectories among young adults in France. Among the oldest cohort (born in 1948–1959), known as the baby-boom generation, diversity is limited in several ways. First, these are trajectories that are predominantly characterized by marriage (civil and/or religious) and a single relationship that starts, for the most part, at a relatively young age. Second, there is generally little demographic differentiation between these clusters, except that Catholic and Muslim individuals tend, for obvious reasons, to marry religiously more often, and that women enter couple life earlier than men, as they are more numerous within the first cluster characterized by early marriage. In other words, young people of this generation follow a similar path to couple life, and marriage serves as a gateway.

The youngest generation (born in 1970–1984) take much more divers tracks to couple life. This is obvious all at once from the larger number of ideal typical trajectories, the higher entropy of each cluster, and the nature of the relationships that involve more distinct stages with different partners. Moreover, a strong pattern is the prolonged entry to couple life that follows a dual pattern.

First, we observe a slower and more gradual process of couple building, with unions progressively becoming institutionalized. Young adults increasingly proceed one step at a time, starting a relationship without living together, then settling down together under one roof, and eventually marrying. For the older generations, these events were much closer in time. For the most recent generation, these events are on the contrary spread out over a rather long period of time.

Second, couple formation has become more gradual in the sense that young people experience several relationships before living together and getting married. There is an increase in separations and the formation of new couples before the age of 30. Indeed, the two clusters that gather the largest number of individuals from the youngest generation are those where material commitment (as in cohabitation) and institutional formalization (as in marriage) are experienced with a second partner. Young people today proceed by trial and error more than older generations did. But there is also greater social differences; the gradual path to couple life is more common among educated individuals, and those with privileged family backgrounds, whereas less educated individuals, especially women, more often cohabit with their first partner.

Lastly, we also observe invariants. Some trajectories that are often described as new, or on the rise, have in fact been around for some time. Contrary to popular belief, staying single throughout youth is not something new but rather a common ideal type of trajectory for all generations, especially for men. Likewise, pathways marked by a long period of LAT relationships are admittedly somewhat more common among the youngest generation, but they were not absent from couple formation patterns of the older ones (Regnier-Loilier et al., 2009). Extended LAT is quite rare; this was already the case in the past and holds true today.

Discussion and Implications for Further Research

Academics within the population sciences who are interested in broad demographic change must be particularly attentive to the historical logics at work. Are the changing patterns we see in our data actual shifts in behavior (empirical change), or is it rather that our indicators have become less efficient in capturing what we set out to study (methodological lag). The transformations in the entry to couple life poses precisely this problem.

As a part of a broader framework of research on the transition to adulthood, the literature on couple formation has often taken for granted that first unions have been postponed, much in the same way as have others events traditionally associated with entering adult life. However, it is worth noticing that the delay in the timing of these events differs considerably. A close reading of Galland’s research on France shows that, while the age at completion of education, first job and first child has considerably risen over the last decades, the change in age at first cohabiting union is less pronounced (Galland, 1995, 2000). Similarly, Manning, Brown and Payne found that age at first cohabitation has remained stable in the USA between 1984 and 2010 (while age at marriage has risen significantly) (Manning et al., 2014).

Moreover, and more importantly, is cohabitation a good measure of first union formation? When marriage rates started to drop in Europe in the second part of the twentieth century, demographers progressively turned from age at marriage to age at cohabitation in order to establish the timing of couple formation. Undeniably, it no longer made much sense to define the entry to couple life by an event that had become more of a confirmation of union stability. The same question rises today with regard to cohabitation. It is possible that the profound changes that characterize youth, with longer school enrolment and later stable employment, has led young people to dissociate couple life and cohabiting life. This could be due to economic uncertainty, as moving in together often requires economic independency (Landaud, 2021), or to a desire to try the relationship out before settling down together. This would mean that cohabitation too, just as marriage, is rather a way of endorsing the relationship rather than a way of entering it.

In any case, it appears particularly clear from our analysis that young people have not postponed their entry into couple life, as they define it, but delayed the material and institutional formalization of their union. Another study carried out on the EPIC survey suggests that the pattern observed for generational change applies also to educational differences. Whereas age at first cohabitation differs considerably in France according to educational attainment, in line with the existing literature, this is not the case for the age at first couple formation per se, which is remarkably similar across educational groups (Bergström, 2017). In other words, young people from different generations and with different degrees enter unions approximately at the same time; what differs are their aspirations and/or their material possibilities to rapidly (or more slowly) settle down with their partner. Changing the indicator thus radically changes our understanding of this demographic evolution. Society is changing fast, but not always the way we expect. We need to make sure that our tools are sharp and up to date in order to grasp it.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

The EPIC survey was carried out by INED and INSEE. EPIC benefited from the financial support of the French National Research Agency (CECHIC Project, ANR-12-CORP-0016–01), the Caisse nationale des allocations familiales (CNAF), the research and statistics department in the Ministry of Health (DREES), and the iPOPs Laboratory of Excellence (Individuals, Populations, Societies).

Data Availability

Data are available to researchers via Quetelet PROGEDO Diffusion.