INTRODUCTION

In the years following 1950, marriage represented a turning point for young people who decided to form a family and acquired economic and housing independence (Saraceno, 2017). Cohabitation, especially from the 1970s onwards and to an even greater extent thereafter, begins to constitute an alternative to marriage (Laslett, 1977) both for economic reasons and for individual choice (Toulemon, 1997; Kiernan, 2004). After the beginning of the 1980s, there is a transformation that first sees the postponement of marriage, the increase in cohabitation and divorces and the lowering of fertility (Liefbroer et al., 1994). This trend continued in Europe (Sassler and Lichter, 2020) with differences between countries based on their cultural and institutional contexts and productive systems. Differences that are also reflected in the opportunities and ways of supporting the autonomy of young people and union formation processes (Saraceno, 1997; Lesthaeghe and Surkyn, 2002; Sassler and Goldscheider, 2004; Saraceno, 2017).

Simultaneously, several changes occurred in Spain. Firstly, the rapid increase in the involvement of women in paid work. In the Fordist model, family members used to adapt to the working needs of the male breadwinner. Secondly, labor market flexibility and the economic crisis that began in 2007–2008 undoubtedly increased uncertainty about future decisions regarding family formation. Thirdly, the legal reforms undertaken in 1981 (legalization of divorce) to adapt the Civil Code to the 1978 Spanish Constitution based on the principle of equal rights for women and men, as well as among all children, affected family formation (Alberdi, 1999) because women and men were afforded equal rights in marriage and cohabitation (Tobio, 2001). Moreover, all children are now considered equal by the law-regardless of their parents’ marital status-and married and, since 2005, there are no legal distinctions between same-sex and heterosexual unions.

These changes, in a context of globalization, led to interest in studying trends across genders and generations regarding first union formation choices, based on its growing interdependence with the working conditions of partners (Saraceno, 2017).

Cohabitation in Spain has experienced a continuous increase over the last four decades. Here, a central role has been played by young generations who increasingly tend to choose cohabitation over marriage (Domínguez-Folgueras, 2011; García-Pereiro, 2011; Domínguez-Folgueras and Castro-Martín 2013). Recent studies have also identified differentiated behaviors of union formation across birth cohorts that can be linked to the socioeconomic characteristics of women (García-Pereiro, 2019). First, educational inequalities account for a profile change, with cohabitation starting among trendsetters towards the 1980s and expanding to women with lower educational levels thereafter. Second, the influence of women's experience in the labor market on the choice of cohabitation intensifies across generations. Results suggest the emergence of a socioeconomic gradient (SEG) of first union formation, where delaying union formation and cohabiting rather than getting married—observed among the youngest cohorts—could be strategies to address socioeconomic disadvantages without having to completely abandon the idea of starting a family (Perelli-Harris et al., 2010). These strategies are quite different from those implemented by the pioneers and might respond to the role of societal changes and growing economic uncertainty.

The main purpose of this paper is to address the relationship between socioeconomic conditions and first union formation in Spain by analyzing the influence of educational attainment and employment history on the transition to non-marital cohabitation and direct marriage, highlighting inter-generational and gender-specific trends over time. This paper contributes to relevant literature on the subject in several ways. First, it fills a gap in recent studies by analyzing changes in the SEG of first cohabitation and marriage, considering changes across cohorts and gender. As stated by Vignoli et al. (2016) and Schneider et al. (2019), little research has investigated whether and how socioeconomic aspects shape the decision to marry rather than cohabitate (or vice-versa). Some studies have emphasized the role of the economic resources of men in partnership formation (Bukodi, 2012) and marriage (Kim, 2017), others have focused on the educational gradient (Ní Bhrolcháin and Beaujouan, 2013; Vergauwen et al., 2017). Regarding the Spanish case, studies have focused on the timing of marriage (Gutiérrez-Domènech, 2008), testing the female economic independence hypothesis (Domínguez-Folgueras and Castro-Martín, 2008) and the role of trendsetters (García-Pereiro, 2019).

Second, considering the limitations faced by previous Spanish research on the subject (data from 2006 restricted to a female-only sample, i.e., Domínguez-Folgueras and Castro-Martín, 2008; García-Pereiro, 2019), the exploitation of the last available Fertility Survey (FS) conducted by the Spanish National Institute of Statistics (INE) in 2018 allows us to enrich our competing-risks event history analyses by adding several biographical years of observation to recent cohorts and by including men. These contributions boost the development of the specialized literature on partnerships and reproductive behaviors in Spain, stressing the longitudinal gender perspective.

Third, we analyze the effects of job instability, which having assumed greater importance since 1984 with the introduction of precarious and flexible forms of work, has had a greater impact on younger generations in their transition to first union formation. In fact, job uncertainty affects the choices of young people (Oppenheimer, 1988, 1997; Blossfeld et al., 2011; Bukodi, 2012), while regulatory cultural and local welfare systems influence the subjective perception of the same job uncertainty and related decisions.

Furthermore, delays in union formation seem to accompany a pluralization and increased complexity of family life paths of younger cohorts compared to those aged 60–70, for which there was a greater standardization of life paths, in general, and of the transition to first union formation, in particular (Settersten and Hagestad, 1996). This is also an understudied subject in Spain.

The remainder of the paper is organized as follows. Section 2 presents the theoretical background and research hypotheses. In Section 3, we describe data, methodology, and the analytical strategy implemented. Section 4 shows our main results in two subsections: the first presents differences regarding the quantum and timing of first unions, while the second is dedicated to changes in the role played by education and employment in first union formation across birth cohorts. Section 5 concludes.

Theoretical background

In recent decades, two micro-level explanations have been called upon to disentangle the influence of socioeconomic factors on union formation (Sassler and Lichter, 2020). One is Becker’s (1981) utility maximizing model, applied to mate-selection processes. First union formation decisions (marriage or cohabitation) are made based on the type of union from which individuals expect greater benefits, being particularly focused on the occurrence of this event. In this regard, and according to the concept of specialization within the couple, men holding a good socioeconomic position are more “marriageable” whilst highly educated and economically independent women see a reduction in the gains from getting married, given the mismatch between domestic and extra-domestic activities.

The reconciliation of these activities—in the absence of policies supporting the cost of children—together with work flexibilization, the spread of precarious work, and weak social protection mechanisms increase uncertainty about decision making (Barbieri et al., 2015). The perception of economic uncertainty challenges young people when deciding to enter their first union and which form of union to prefer. According to the International Social Survey Programme (ISSP, 2013), in 2005 43% of Spaniards between 18 and 33 years of age perceived employment uncertainty (compared to 10% in Nordic countries) and 32% declared that finding another job would be exceedingly difficult.

The subjective perception of job uncertainty also impacts first union transitions in terms of the occurrence and timing of the choice between cohabitation and marriage, as hypothesized by Oppenheimer (1988, 1997). Individuals’ economic conditions act either as an obstacle for union formation if men’s earnings deteriorate or as a facilitator if women’s economic strength increases. Then, the assortative mating of men with precarious positions in the labor market is limited, and with marriage being a more expensive form of union, it might be delayed until one’s economic position improves. In contrast, when achieving a higher position, women might delay union formation or prefer cohabitation over marriage, waiting to improve their careers before transitioning to marriage.

Oppenheimer’s uncertainty hypothesis holds in societies where young adults—and especially young men—feel unable to fulfill expectations linked to marriage because of their poor and unstable economic opportunities (Bukodi, 2012). This might also hold in societies transitioning to a dual-earner model, where women are called on to cope for the deteriorating employment conditions of their male counterparts while building their careers.

According to Esping-Andersen and Billari (2015), there is the need to consider the fundamental role played by the welfare state and levels of gender equality, specifically regarding fertility but indirectly related to union formation. In this sense, union formation rates are higher in countries with gender-egalitarian norms (Brodmann et al., 2007; Sevilla-Sanz, 2010; Myrskylä et al., 2011) and couples tend to be more stable when men contribute significantly to domestic tasks (Cooke, 2006; Sigle-Rushton, 2010).

Previous research confirmed that the diffusion of cohabitation has been heavily dependent on young adults’ preference for cohabitation over marriage (Castro Martín and Domínguez, 2008; Domínguez, 2011; García-Pereiro, 2011; Muñoz and Recaño, 2011; García-Pereiro et al., 2015).

It has been stated that the choice of entering a first union via cohabitation rather than marriage is associated with values, norms, and attitudes that favor independence, tolerance, and autonomy (Lesthaeghe and Surkyn, 2002; Thomson and Bernhardt 2010), the dissemination of which has been led by individuals with higher levels of education. In fact, better-educated women are more likely to enter into cohabitation than women with lower educational levels (Baizán et al., 2003; Castro Martín et al., 2008; Castro Martín and Domínguez, 2008; Domínguez, 2011; García-Pereiro, 2019). In spite of this, recent studies have identified a reversal in the positive educational gradient of cohabitation (Koytcheva and Philipov, 2008; Kalmijn, 2013; Ní Bhrolcháin and Beaujouan, 2013).

The relationship between women’s economic independence and union formation can vary. Although women’s economic resources positively influence entry into marriage, the effects on cohabitation are less clear. Regarding the former, empirical work suggests that the effect was negative for older cohorts and inverted for women in recent cohorts (Sweeney, 2002; Shafer and James, 2013; Addo, 2014; McClendon et al., 2014; Schneider and Reich, 2014; Kim, 2017). For the latter, several studies have found no association between employment and entering cohabitation (Carlson et al., 2004; Sassler and Goldscheider, 2004) whereas others have reported a positive association (Bracher and Santow, 1998; Clarkberg, 1999; Winkler-Dworak and Toulemon, 2007; Jalovaara, 2012), not only among women but also among men.

Research on the changing effects of employment status and education on union formation among men has also given mixed results. The likelihood of entering marriage for men is higher for the employed than for the unemployed, and for those with higher incomes and economic resources (Oppenheimer et al., 1997; Sweeney, 2002; Oppenheimer, 2003; Sassler and Goldscheider, 2004; Harknett and Kuperberg, 2011; Shafer and James, 2013; McClendon et al., 2014; Vignoli et al., 2016). However, some studies have found a declining trend in their influence on the likelihood of marriage over time (Sassler and Goldscheider, 2004), while others have reported no changes (Sweeney, 2002; Kalmijn and Luijkx, 2005; Bukodi, 2012).

For Spain, Gutiérrez-Domènech (2008) has demonstrated that employment increases the probability of getting married not only for women but also for men. Following this line, Domínguez-Folgueras and Castro-Martín (2008) have shown that the odds of entering cohabitation relative to marriage significantly increase among college-educated women and are higher among employed women.

When comparing women born between 1945–1954 with those of successive birth cohorts, authors report that college-educated women are more likely to enter cohabitation than marriage, although their models show smaller positive coefficients for the youngest cohort. García-Pereiro (2019) also found that the educational gradient of cohabitation is strongly dependent on birth-cohort effects.

Research hypotheses

Fewer first unions involve direct marriage, and cohabitation has become widespread (Sassler and Lichter, 2020). Previous research has pointed out birth-cohort differentials in the timing and prevalence of marriage and cohabitation for first union formation in Spain (Miret, 2007; Castro-Martín and Domínguez, 2008; Domínguez-Folgueras and Castro-Martín, 2008; Dominguez-Folgueras, 2011; Muñoz and Recaño, 2011; García-Pereiro et al., 2014; García-Pereiro, 2019). Thus, we expect to find significant inter-cohort differences in first-union transitions supporting Cherlin’s (2020) “retreat from marriage” as first cohabitation takes place over marriage as a conventional marker: the incidence of non-marital cohabitation will significantly increase over generations while the incidence of getting married directly will decrease for both women and men (Composition effects RH1).

The second research hypothesis regards compositional effects (RH2) of the SEG of first union formation in Spain. There will be a positive SEG of cohabitation because the incidence of cohabitation will be higher among better-educated and employed women and men (RH2 a: SEG of Cohabitation). For direct marriage a mixed gradient is expected, negative for education and positive for employment (RH2 b: SEG of Marriage).

In accordance with Spanish literature (Domínguez-Folgueras and Castro-Martín, 2008; Gutiérrez-Domènech, 2008; García-Pereiro, 2019), the SEG of first union formation might have changed across generations that lived their transition to adulthood in different contexts. Thus, as cohabitation spreads and becomes more accepted, the positive educational gradient of cohabitation is expected to get weaker for younger cohorts (RH3: Changing Educational Gradient of Cohabitation).

To further support our last research hypothesis, we consider the contextual setting experienced by generations, considering that employment instability in the Spanish labor market had an impact on young adults first union formation choices (Martínez-Pastor and Bernardi, 2011). The 1961–1970 birth cohort was raised during a complex and rapid process of modernization in Spain (Reher and Requena, 2019), where the expansion of the educational system was accompanied by an incipient welfare state, high unemployment, and the prevalence of the male-breadwinner model. The 1971–1980 cohort reached adulthood in a modern society where women had greater access to university education and entered the labor market in a context of scarcity and limited economic growth also affected by the increase in labor market flexibilization taking place after 1984. Although the male-breadwinner model still prevailed, democratic values, egalitarian gender roles, and secularization were being socially interiorized. The 1981–1990 cohort has lived their young adulthood with remarkable opportunities to access and complete higher education—even postgraduate degrees—but their high qualifications no longer guarantee equally relevant employment opportunities (in terms of status and earnings). The dual-earner model is prevalent, but-during their labor market entry-faced additional austerity measures that sharpened employment precariousness. The severe impact of the global financial crisis in 2008 pushed policy makers to reflect on the need to undertake deeper structural reforms to soften the labor market rigidity. Thus, in this context of economic precariousness, the relevance of couples’ economic resources (those of men and women) will increase, employment being important for union formation but even more important for entering first marriage than cohabitation, and for men more than for women (RH4: Changing Economic Gradient).

Data and methods

Data were drawn from the Fertility Survey (FS) conducted by the INE in 2018, which interviewed 17,175 individuals (14,556 women and 2,619 men) aged 18 to 55, collecting retrospective information on partnerships, childbearing, and work biographies.

In the FS2018, partnership biographies were collected retrospectively. Information on first union formation was derived by linking information on the timing and occurrence of the current partnership (at the time of the survey) and former partnerships. Timing regarding education, parental divorce, leaving the parental home, and working biographies were also collected.

To construct respondents’ childbearing biographies, the record of children ever had was merged to the main dataset and transformed into a person-month file—separately for women and men (this last being an independent sample). After excluding respondents with missing data, born between 1991-2000 and living with same-sex partners, the effective sample sizes were 11,332 women and 2,107 men.

To assess generational changes in terms of the socioeconomic disadvantages of first union formation, the following explanatory variables were included in the analyses.

Generation, disaggregated across three cohorts: 1961–1970, 1971–1980, and 1981–1990¹. We are aware that any classification is largely arbitrary. However, this categorization allowed us to compare different experiences of first union transitions over time considering the heterogeneous contexts in which each generation reached young adulthood and got ready to start living with a partner (Clemente and García-Pereiro 2020).

Educational attainment. Respondents were classified according to the highest level reached: primary school or less, lower secondary, upper secondary, or tertiary education.

Employment history. Respondents’ experience in the labor market was measured through a dummy variable that changes to 1 whenever the respondent starts working. These time-varying variables change over time. They were measured monthly and treated using episode splitting and, in the second case, summarizing employment episodes.

Controls. Several variables have been previously identified as determinants of first partnership formation. Previous research suggests that individuals who experienced the divorce of their parents are more likely to cohabit than marry (Axinn and Thornton, 1996; Amato and DeBoer, 2001; Domínguez-Folgueras and Castro-Martin, 2008). A time-varying variable was thus included to indicate a change whenever parental divorce occurred.

Living independently is also likely to affect partnership formation. Individuals living with their parents are more likely to get married than those who have been living independently and are more likely to cohabit (Liefbroer et al., 1994; Dominguez-Folgueras, 2011; García-Pereiro, 2019). Accordingly, we introduced a time-variant control variable that changes from living with one’s family to leaving the parental home.

Childbearing decisions strongly influence the choice between marriage and cohabitation. In Spain, the conception of a child and marriage remain highly interrelated events (Jurado Guerrero and Naldini, 1997; Baizán et al., 2003; García-Pereiro et al., 2015). Thus, information on fertility was included as a time-varying covariate that changes from childless to pregnant (8 months before the date of birth) to birth.

Other covariate included is education enrollment at the time of the survey, useful to differentiate between currently being in education and having already achieved the highest level (Coppola, 2004; Schneider et al., 2019). Moreover, the size of the family of origin (Kiernan, 2000; Winkler-Dworak and Toulemon, 2007), being foreign-born (Cortina et al., 2010), religious practice (García-Pereiro et al., 2014), urban residence and attitudes towards cohabitation (Domínguez-Folgueras and Castro-Martín, 2008) might positively influence first union transitions.

First union formation was operationalized in a competing-risk setting with two outcomes: cohabitation or marriage. The risk period started at age 15 and was censored at the time of the survey if the respondent remained single. Descriptive statistics for explanatory and control variables are shown in the Appendix (Table A).

By applying a competing-risk framework, the occurrence of direct marriage inhibits the occurrence of non-marital cohabitation, given that both are first-order events. Cumulative incidence curves (CICs) for cohabitation (event of interest) and direct marriage (competing event) were calculated to analyze timing and incidence. The cumulative incidence is computed as the estimate of the Kaplan–Meier of the overall survival function (considering all kind of possible failures) multiplied by an estimate of the hazard of the failure of interest (cohabitation in this case). The sum of all cumulative incidences equals 1- S (t), the complement of the overall Kaplan-Meier survival estimate (Coviello and Boggess, 2004).

Semiparametric Fine and Gray’s competing risks regressions (Fine and Gray, 1999; Pintilie, 2007) were run to estimate the cumulative incidence of first cohabitation in the presence of direct marriage, using birth cohort, educational attainment, and employment history as explanatory variables while controlling for other variables of interest. This model calculates the sub-hazard by applying the maximum likelihood approach, an accurate method to deal with competing risks (Bakoyannis and Touloumi, 2012; Pintilie, 2007). The estimated coefficients (sub-hazard ratios, SHRs) express the magnitude of the change (up or down) of the cumulative incidence function (CIF) through the effect of certain covariates while controlling for other covariates.

Our analytical strategy follows the biographic evolution of first union transitions, applying a generational perspective to composition and interaction effects. Composition effects allow identifying differences in the incidence and timing of first cohabitation vs. first marriage across birth cohorts. The first part of the results section is dedicated to the analysis of these compositional effects through: 1) a description of CICs computed by birth cohort, 2) an explanation of inter-generational differences in the incidence of cohabitation vs. marriage, and vice-versa, in models where birth cohorts are treated as an independent variable. Models with interaction terms between birth cohorts and educational attainment and birth cohorts and employment indicate whether and how socioeconomic disadvantages differentially impact first union formation across generations. Intergenerational differences were further developed by running models separately for each birth cohort, as a sort of robustness checks.

Results

Composition effects: the changing incidence of first marriage and cohabitation across generations

This section is dedicated to the analysis of the compositional effects. Figure 1 shows the CICs for entering first cohabitation vs. first marriage across female birth cohorts. Following the curves through generations, the increase in the incidence of first cohabitation among women and the simultaneous decrease in the incidence of direct marriage for each successive birth cohort must be noted. Differences in the timing of first union arise when taking a more detailed look: at age 25, nearly 32% of women born 1961–1970 got married directly, and this decreases to 13% for the next generation (1971–1980). Conversely, the incidence of cohabitation among women in the first birth cohort was almost negligible at age 25 but reached over 11% in the next cohort, while among the 1981–1990 generation the same-age-incidence expanded to 25%. In this generation, at age 35 more than 70% of women chose to cohabit rather than to get married directly.

The CICs computed for men’s first union transitions are displayed in Figure 2. For first unions of men born 1961–1970, marriage was the preferred choice: at age 25, 20% of men got married directly, and the incidence increased with age, almost reaching 80% right after 35 years of age. The incidence of cohabitation among this generation of men remained marginal—well below 5%—and although it increased with age, its value stabilized shortly before age 40 at around 10%.

When looking at the curves of men born 1971–1980, the distance between the CICs of cohabitation and marriage reduced markedly. The incidence of getting married directly diminished to 7% at age 25, while the incidence of first cohabitation at the same age increased to nearly 12%. The distance tends to increase with age, but figures regarding direct marriage do not reach those of the previous cohort, and cohabitation was on the rise.

Men in the 1981–1990 birth cohort chose cohabitation rather than marriage in their early youth. Figure 2 shows that all along their biographies, entering a first union through cohabitation prevails even at older ages (increasing the distance between CICs). At age 35, the incidence of cohabitation was reaching 70% while the incidence of marriage was around 10%.

Several differences arise when comparing the CICs of first union transitions of women and men. In the first birth cohort (1961–1970), the incidence of cohabitation at age 30 was 20% among men, while women held half of this incidence. In the second cohort, until the same age the CICs are similar: around 30% of men and women entered cohabitation instead of marriage. Instead, after age 35, the incidence of marriage and cohabitation among women reached similar values while among men diverged (increasingly) with age. Finally, first union transitions of the 1981–1990 generation show that more than 50% of women cohabited at age 30, while the incidence of men around this age was ten percentage-points lower.

Source: Own elaboration, EF2018.

Notes: The results of Pepe and Mori’s test (p-value<0.00001) show that differences across generations are statistically significant for cohabitation and marriage.

Figure 1. CICs for first union transitions (cohabitation vs. marriage) of women in Spain zby birth cohort.

Source: Own elaboration, EF2018.

Notes: The results of Pepe and Mori’s test (p-value<0.00001) show that differences across generations are statistically significant for cohabitation and marriage.

Figure 2. CICs of first union transitions (cohabitation vs. marriage) of men in Spain by birth cohort.

Regarding the timing, the choice of getting married directly has experienced little changes across cohorts², with women getting married earlier respect to men. Instead, the choice to enter cohabitation has been taking place at earlier ages across generations³, being women younger than men when starting their first cohabiting union.

Descriptive analyses were tested in a competing risks multivariate setting, confirming noticeable differences between cohorts in terms of first partnership formation. Table 1 presents the results of the regression model of competing risks for women’s first union transitions (M1: cohabitation vs. marriage, M2: marriage vs. cohabitation).

While the CIF of cohabitation increases, that of direct marriage significantly decreases in each successive cohort (Table 1). The SHRs of M1 confirm the rapid spread of first cohabitation in Spain reported in previous studies (Domínguez, 2011; García-Pereiro, 2019). In fact, the magnitude of change of the cohabitation CIF for women in the 1971–1980 generation is more than two times higher than that of the previous generation (1961–1970), while the SHR of cohabitation is more than four times higher among women in the 1981–1990 generation. Conversely, the SHRs of M2 illustrate a lower CIF of marriage across cohorts: 50% and over 80% lower for women born between 1971–1980 and 1981–1990, respectively, compared to the 1961–1970 birth cohort.

Women’s educational attainment exerts a positive effect on the CIF of first cohabitation and a negative effect on entering marriage. The SHR of cohabitation among women with tertiary education is 1.2 times higher than among women with primary education or less.

The effect of women’s employment on first union transitions is positive, increasing the CIF of both cohabitation and marriage (Table 1). The SHRs of entering cohabitation relative to marriage are 28% higher among employed women compared to the unemployed. At the same time, the SHR of direct marriage relative to cohabitation is 18% higher among employed women.

Results of the competing risks regressions for men are presented in Table 2. Regarding intergenerational differences, the SHRs associated with first cohabitation are, respectively, 1.8 and 2.8 times higher among men from the 1970s and 1980s cohorts than among the 1960s generation of men. Figures regarding men’s first marriage indicate that its CIF for the 1971–1980 generation is 47% lower than that of the previous generation (1961–1970). In the following generation, the SHR of marriage decreased even more, being around 79% lower. Intergenerational changes in the incidence of first cohabitation and first marriage seem to follow the same trend for men and women.

As shown in Table 2, having achieved tertiary education and being employed are associated with an increase in the CIF of cohabitation (M1-Tertiary: SHR=1.26, Employed: SHR=1.22). Although being employed also increases the CIF of first marriage (by 14%), the effect of educational attainment is reversed. In fact, the SHRs of entering marriage rather than cohabitation decrease as the level of education increases M2-Secondary I: SHR=0.84, Secondary II: SHR=0.78, Tertiary: SHR=0.58.

The SHRs of control variables that influence first cohabitation or marriage are shown in Table 1 (women) and Table 2 (men). For women only, being enrolled in education is important for entering cohabitation or marriage because it delays the entry into the first union, a result in line with previous studies (Domínguez-Folgueras and Castro-Martín, 2008).

In accordance with previous research, parental divorce (Amato and DeBoer, 2001; Domínguez-Folgueras and Castro-Martín, 2008; García-Pereiro, 2019), and leaving the parental home (Liefbroer et al., 1994; Domínguez, 2011; García-Pereiro et al., 2014) are associated with significant increases in the CIF of first cohabitation (Table 1—women: M1-SHR=1.26, SHR=2.59; Table 2—men: M1- SHR=1.38, SHR=2.67) and decreases in that of first marriage for women in both cases and for men only in the first (Table 1—women: M2-SHR=0.69, SHR=0.97; Table 2—men: M2-SHR=0.45, SHR=1.22).

Table 1. Results of competing risks regression models for the first union transitions of women in Spain (M1: cohabitation vs. marriage, M2: marriage vs. cohabitation).

	M1		M2
	SHR	sign.	SHR	sign.
Cohort
(1961–1970)
1971–1980	2.35	***	0.50	***
1981–1990	4.29	***	0.18	***
Education
(Primary or less)
Secondary I	1.07		0.95
Secondary II	1.11	*	0.79	***
Tertiary	1.24	***	0.60	***
Employed t	1.28	***	1.18	***
Controls
Educational enrolment	0.89	**	0.90	*
Frequent religious practice	0.57	***	1.44	***
Foreign-born	0.99		0.87	*
Siblings	1.00		0.99
Urbanization
(Full-urban)
Middle-urban	0.92	*	1.24	***
Rural	0.91	*	1.12	***
Parental separationt	1.26	***	0.69	***
Leave parental homet	2.59	***	0.97
Fertility-status
Birtht	1.02		0.21	***
Pregnancyt	1.11		4.86	***
Agree is all right to live together without getting married	1.57	***	0.79	***
N	11,332		11,332
Failures	4,570		5,533
Log - pseudolikelihood	-30,707		-48,139

Notes: * p<.05, **p<.01, ***p<.001. Reference Categories (RC); t time-varying covariate. Robust standard errors (RSE) are used.

Table 2. Results of competitive risks regression models for men’s first union transitions in Spain (M1: cohabitation vs. marriage, M2: marriage vs. cohabitation).

	M1		M2
	SHR	sign.	SHR	sign.
Cohort
(1961–1970)
1971–1980	1.80	***	0.53	***
1981–1990	2.78	***	0.21	***
Education
(Primary or less)
Secondary I	1.15		0.84
Secondary II	1.22	*	0.78	**
Tertiary	1.26	**	0.58	***
Employed t	1.22	***	1.14	***
Controls
Educational enrolment	0.98		0.90
Frequent religious practice	0.64	***	1.27	*
Foreign-born	0.98		0.86
Siblings	0.99		0.98
Urbanization
(Full-urban)
Middle-urban	0.94		1.15
Rural	0.86		1.02
Parental separationt	1.38	***	0.45	***
Leave parental homet	2.67	***	1.22	*
Fertility-status
Birtht	1.00		0.59	***
Pregnancyt	1.46	***	1.91	***
Agree is all right to live together without getting married	1.72	***	0.68	***
N	2,107		2,107
Failures	1040		745
Log - pseudolikelihood	-6,503		-5,211
* p<.05, **p<.01, ***p<.001
Notes: (RC); t time-varying covariate. RSE.

The effects of the rest of controls are consistent with those reported by previous studies (Domínguez-Folgueras and Castro-Martín, 2008; Domínguez, 2011; García-Pereiro, 2019). The SHR is lower for cohabitation and higher for marriage for women and men with a frequent religious practice (Table 1—women: M1-SHR=0.57, M2-SHR=1.44; Table 2—men: M1-SHR=0.64, M2-SHR=1.27) and vice-versa for favorable attitudes regarding cohabitation (Table 1—women: M1-SHR=1.57, M2-SHR=0.79; Table 2—men: M1-SHR=1.72, M2-SHR=0.68).

Regarding fertility, women’s results were not statistically significant but follow those of men, being pregnancy is positively associated with first cohabitation. The birth of a child is related to a lower SHR for marriage, while a pregnancy increases its incidence, being nearly 5 times and 1.9 times higher for women and men, respectively.

Generational interaction effects: the changing SEGs of first union formation

Models including interaction terms and separated by cohort examine if the relevance of SEG in first union transitions varies by generation. Table 3 shows the results of the inclusion of interaction effects between employment history, having achieved tertiary education, and birth cohort for women’s transitions to first cohabitation. Regarding the effects of education, having achieved tertiary education exerts a more pronounced positive effect on women’s CIF of first cohabitation for the 1961–1970 cohort than for 1981–1990 cohort (Table 3-M1). Results corroborate that the influence of women’s educational level reversed over time: for the 1981–1990 generation, the SHR of cohabitation was 20% lower for tertiary education than for primary or less (Table 3-M4).

Differences across birth cohorts also emerge when considering employment: the interaction term shows that the CIF of entering cohabitation rather than marriage among employed women is positive but lower in the last cohort (1981–1990) than in the 1961–1970 cohort (M1). This trend is also evident when examining the influence of employment separately by generation (M2-1961–1970: SHR=1.35, M3-1971–1980: SHR=1.27, M4-1981–1990: SHR=1.23).

Table 3. Results of competitive risks regression models for women’s first union transitions in Spain (cohabitation vs. marriage) with interaction terms (M1) and by birth cohort (M2, M3, and M4).

	M1		M2:1961–1970		M3:1971–1980		M4:1981–1990
	SHR	sign.	SHR	sign.	SHR	sign.	SHR	sign.
Interactions
(1961–1970)*Employed
1971–1980*Employed	0.95		-	-	-	-	-	-
1981–1990*Employed	0.90	**	-	-	-	-	-	-
(1961–1970)*Tertiary
1971–1980*Tertiary	0.94		-	-	-	-	-	-
1981–1990*Tertiary	0.60	***	-	-	-	-	-	-
Cohort
(1961–1970)
1971–1980	2.51	***	-	-	-	-	-	-
1981–1990	6.81	***	-	-	-	-	-	-
Education
(Primary or less)
Secondary I	1.08		1.08		1.03		0.98
Secondary II	1.22	*	1.19		1.19	*	0.84
Tertiary	1.47	***	1.49	***	1.37	***	0.80	*
Employed t	1.36	***	1.35	***	1.27	***	1.23	***
Controls
Education enrolment	0.90	*	0.95		0.92		0.87	*
Frequent religious practice	0.56	***	0.63	***	0.62	***	0.44	***
Foreign-born	0.97		1.44	**	1.01		0.86	*
Siblings	1.00		1.01		1.00		0.98
Urbanization
(Full-urban)
Middle-urban	0.92	*	0.89		0.89	**	0.97
Rural	0.93	*	0.96		0.88	*	0.94
Parental separationt	1.26	***	1.34	*	1.33	***	1.19	*
Leave parental homet	2.58	***	3.95	***	2.55	***	2.03	***
Fertility-status
Birtht	1.02		1.70	***	1.04		0.53	***
Pregnancyt	1.12		0.76		1.04		2.00	***
Agree is all right to live together without getting married	1.58	***	1.36	*	1.59	***	1.61	***
N	11,332		4,112		4,592		2,628
Failures	4,570		821		2,105		1,644
Log - pseudolikelihood	-39,688		-6,457		-16,757		-11,915
* p<.05, **p<.01, ***p<.001
Notes: (Reference categories); t time-varying covariate. RSE.

Table 4 displays the results of interaction terms on women’s direct marriage transition. It is evident that the positive influence of employment on marriage CIF has not changed across birth cohorts (M1-not statistically significant, M2-1961–1970: SHR=1.18, M3-1971–1980: SHR=1.14, M4-1981–1990: SHR=1.27) remaining equally important for subsequent generations.

Women’s negative educational gradient for marriage intensified over time, as the reduction of its SHRs for tertiary education is more pronounced for the 1981–1990 cohort than the 1961–1970 cohort.

Results for men are displayed in Table 5. Interactions between birth cohort and employment (M1) show that the SHR associated with cohabitation among employed men has not changed significantly. In cohort-separated models, the influence remains positive and significant and generational changes are not clear-cut (M2-1961–1970: SHR=1.22, M3-1971–1980: SHR=1.24, M4-1981–1990: SHR=1.20).

Interaction terms for men’s generational differentials by education show that having achieved tertiary education reduces first cohabitation CIF for the 1981–1990 cohort to a greater extent than for the 1961–1970 cohort (Table 5-M1). This result is confirmed by M2–M4, where the influence of having achieved a high level of education positively influences men’s CIF of entering cohabitation rather than marriage only for the 1960s cohort (M2-1961–1970: SHR=2.18), losing its strength over time with each successive generation.

Table 6 presents the results on men’s transitions to direct marriage. As shown, interactions between birth cohort and employment indicate that employment history is less relevant for marriage among the most recent cohorts with respect to the oldest (1961–1970). Models run for each generation confirm this, but the coefficients for the last cohorts are not statically significant (M3, M4).

Table 4. Results of competitive risks regression models for women’s first union transitions in Spain (marriage vs. cohabitation) with interaction terms (M1) and by birth cohort (M2, M3, and M4).

	M1		M2:1961–1970		M3:1971–1980		M4:1981–1990
	SHR	sign.	SHR	sign.	SHR	sign.	SHR	sign.
Interactions
(1961–1970)*Employed
1971–1980*Employed	0.99		-	-	-	-	-	-
1981–1990*Employed	0.94		-	-	-	-	-	-
(1961–1970)*Tertiary
1971–1980*Tertiary	1.02		-	-	-	-	-	-
1981–1990*Tertiary	0.76	*	-	-	-	-	-	-
Cohort
(1961–1970)
1971–1980	0.50	***	-	-	-	-	-	-
1981–1990	0.24	***	-	-	-	-	-	-
Education
(Primary or less)
Secondary I	0.98		0.97		1.00		0.77
Secondary II	0.78	***	0.77	***	0.82	*	0.75	*
Tertiary	0.61	***	0.60	***	0.63	***	0.58	***
Employed t	1.19	***	1.18	***	1.14	***	1.27	***
Controls
Education enrolment	0.90	*	0.95		0.88	*	0.80
Frequent religious practice	1.44	***	1.15	*	1.62	***	3.21	***
Foreign-born	0.87	*	0.71	**	0.76	**	1.44	**
Siblings	0.99		0.98	*	1.00		1.03
Urbanization
(Full-urban)
Middle-urban	1.24	***	1.12	*	1.40	***	1.36	**
Rural	1.13	**	1.08		1.27	***	0.96
Parental separationt	0.69	***	0.78	*	0.64	***	0.73	*
Leave parental homet	0.97		0.94		0.97		0.96
Fertility-status
Birtht	0.21	***	0.15	***	0.27	***	0.45	***
Pregnancyt	4.83	***	5.91	***	4.07	***	2.69	***
Agree is all right to live together without getting married	0.79	***	0.97		0.70	***	0.52	***
N	11,332		4,112		4,592		2,628
Failures	5,533		2,959		2,092		482
Log - pseudolikelihood	-48,135		-22,690		-16,774		-3,537
* p<.05, **p<.01, ***p<.001
Notes: (Reference categories); t time-varying covariate. RSE.

Table 5. Results of competitive risks regression models for men’s first union transitions in Spain (cohabitation vs. marriage) with interaction terms (M1) and by birth cohort (M2, M3, and M4).

	M1		M2:1961–1970		M3:1971–1980		M4:1981–1990
	SHR	sign.	SHR	sign.	SHR	sign.	SHR	sign.
Interactions
(1961–1970)*Employed
1971–1980*Employed	1.08		-	-	-	-	-	-
1981–1990*Employed	1.03		-	-	-	-	-	-
(1961–1970)*Tertiary
1971–1980*Tertiary	0.69		-	-	-	-	-	-
1981–1990*Tertiary	0.33	***	-	-	-	-	-	-
Cohort
(1961–1970)
1971–1980	2.25	***	-	-	-	-	-	-
1981–1990	5.92	***	-	-	-	-	-	-
Education
(Primary or less)
Secondary I	1.21		1.31		1.27		0.89
Secondary II	2.04	**	2.04	**	1.02		0.95
Tertiary	2.02	***	2.18	**	1.40	*	0.69
Employed t	1.17	*	1.22	*	1.24	***	1.20	**
Controls
Education enrolment	0.99		0.40	*	1.12		1.27
Frequent religious practice	0.62	***	0.96		0.42	***	0.87
Foreign-born	0.94		1.02		1.04		0.84
Siblings	0.99		1.00		1.02		0.95
Urbanization
(Full-urban)
Middle-urban	0.94		0.92		0.98		0.93
Rural	0.87		0.92		0.77		1.02
Parental separationt	1.40	***	1.19		1.32		1.60	*
Leave parental homet	2.70	***	3.01	***	2.81	***	2.33	***
Fertility-status
Birtht	1.01		1.20	*	0.95		0.84
Pregnancyt	1.47	***	1.20		1.42	**	2.50	***
Agree is all right to live together without getting married	1.71	***	1.28		1.98	***	1.81	*
N	2,000		679		835		486
Failures	933		213		436		284
Log - pseudolikelihood	-6,489		-1,298		-2,684		-1,563
* p<.05, **p<.01, ***p<.001
Notes: (Reference categories); t time-varying covariate. RSE.

Table 6. Results of competitive risks regression models for men’s first union transitions in Spain (marriage vs. cohabitation) with interaction terms (M1) and by birth cohort (M2, M3, and M4).

	M1		M2:1961–1970		M3:1971–1980		M4:1981–1990
	SHR	sign.	SHR	sign.	SHR	sign.	SHR	sign.
Interactions
(1961–1970)*Employed
1971–1980*Employed	0.82	*	-	-	-	-	-	-
1981–1990*Employed	0.70	*	-	-	-	-	-	-
(1961–1970)*Tertiary
1971–1980*Tertiary	1.19		-	-	-	-	-	-
1981–1990*Tertiary	1.39		-	-	-	-	-	-
Cohort
(1961–1970)
1971–1980	0.58	***	-	-	-	-	-	-
1981–1990	0.21	***	-	-	-	-	-	-
Education
(Primary or less)
Secondary I	0.83		0.81		0.86		1.45
Secondary II	0.69	*	0.68	*	0.95		2.13
Tertiary	0.54	***	0.53	***	0.69		1.61
Employed t	1.29	***	1.25	***	1.04		0.93
Controls
Education enrolment	0.87		1.16		0.72		0.53
Frequent religious practice	1.30	*	0.99		1.69	*	1.57
Foreign-born	0.89		0.77		0.81		1.89
Siblings	0.98		0.97		0.96		1.13
Urbanization
(Full-urban)
Middle-urban	1.16	*	1.17		1.18		1.44
Rural	1.02		0.94		1.35		0.63
Parental separationt	0.45	***	0.58		0.44	*	0.26	*
Leave parental homet	1.20	*	1.33	*	1.06		0.88
Fertility-status
Birtht	0.59	***	0.60	***	0.64	***	0.31	***
Pregnancyt	1.92	***	2.01	***	1.73	***	2.53	*
Agree is all right to live together without getting married	0.68	***	0.73	*	0.64	**	0.57	*
N	2,000		679		835		486
Failures	745		395		295		55
Log - pseudolikelihood	-5,206		-2,358		-1,875		-295
* p<.05, **p<.01, ***p<.001
Notes: (Reference categories); t time-varying covariate. RSE.

Men’s negative educational gradient for marriage reduces across generations, as the reduction of the CIF of marriage for tertiary education was less pronounced in the 1981–1990 cohort than the 1961–1970 cohort.