Stable employment and income are often a prerequisite in the transition to parenthood. In the French context, employment of both partners is widespread, but is the female partner’s employment as determinant as that of the male partner? Do the characteristics of their jobs contribute to the decision to have a first child and its timing? Using rich, longitudinal administrative data, the author describes how the occupational situations and partners’ relative incomes accelerate or slow down the arrival of first births in France, highlighting original results on couples with similar incomes.
1 While men’s stable employment and income have consistently been positively correlated with childbearing in Western countries, the increase in female employment since the 1970s was initially perceived to have a negative effect on fertility (Ahn and Mira, 2002; Goldscheider et al., 2015). This was explained by the rise in female labour force participation without an equivalent shift towards increased male involvement in household tasks and childcare— implying that female employment would decrease the time available to raise children and thus increase the opportunity costs of childbearing (McDonald, 2000). In recent decades, however, female employment has increasingly become a prerequisite to family formation (Ahn and Mira, 2002; Goldscheider et al., 2015), especially in contexts that support the work–family combination. This raises the question whether both partners’ employment positions and income have now come to have an equivalent impact on the transition to parenthood.
2 The literature extensively documents the employment–parenthood and income–parenthood links for women and, to a lesser extent, men. But relatively little research has considered couple-level gender dynamics in childbearing decisions (Matysiak and Vignoli, 2008; Rondinelli et al., 2010; Pailhé and Solaz, 2012). Studies at the couple level in Finland and Italy found that the employment of both partners promotes parenthood; however, while in Finland the transition to parenthood was positively affected when either partner had a high income, in Italy the male partner’s economic position remained decisive (Vignoli et al., 2012; Jalovaara and Miettinen, 2013). In the Netherlands, it is the female partner’s education and earning potential that have been found to most strongly predict childbearing (Begall, 2013); and in Belgium, first-birth risk was found to be higher when the female partner had greater time availability and access to flexible work regimes than the male partner (Marynissen et al., 2020). In Great Britain, a positive association between the traditional division of labour and fertility has been weakening (Zhou and Kan, 2019), and the timing of married couples’ parenthood has been found to be predominantly dependent on the female partner’s labour market status (Inanc, 2015). In France, both men’s and women’s access to permanent employment has been found to positively affect the transition to parenthood (Landaud, 2019). Finally, Schmitt (2012) found a negative link between unemployment and family formation in Germany, the United Kingdom, and France that was similar for the male partner, but not the female partner. These studies demonstrate that the significance of partners’ labour market positions in childbearing decisions varies considerably between countries with different institutional and normative contexts. However, these studies rarely consider partners’ relative positions and do not address whether these associations change over the life course.
3 Using detailed longitudinal microdata from the French Permanent Demographic Sample (EDP in its French abbreviation) and discrete-time hazard models, this study examines the link between partners’ employment status and relative income and the transition to parenthood in France. This article focuses on first births, as studies identify the birth of the first child as a first critical event that sets in motion changes which rigidify the gender balance in paid and unpaid work within households, mostly towards more traditional gender role patterns (Baxter et al., 2008; Grunow et al., 2012; Solera and Mencarini, 2018). Partners’ employment statuses and relative income may thus play a more important role in couples’ transition to parenthood than in case of second or higher-order births.
4 This study makes three contributions to the literature. First, by examining the combination of the male and female partner’s employment position and their relative income, rather than each partner’s activity status or absolute income level separately—an approach rarely taken in previous research—this study is able to characterize gender dynamics in childbearing decisions. Secondly, this article takes the exploration of the link between partners’ labour market positions and the transition to parenthood a step further by exploring whether the association between partners’ employment status and relative income and the transition to parenthood varies with age over the life course. Considerations about partners’ labour market positions when deciding whether to have a first birth may play a different role at different ages. Thirdly, this study examines France, a country that combines high fertility and well-developed family policies predominantly focused on female employment, rather than on promoting gender equality (Morel, 2007; Saxonberg, 2013). The latter contrasts with Scandinavian countries, which have extensive, quasi-universal, and degenderized work–family policies; English-speaking countries, where limited, mostly cash support is provided to (particularly low-income and single-parent) families with children under age 3; Southern European countries, which have even more limited and highly fragmented assistance; and most Eastern European countries, which have only more recently started to invest in the dual-earner model (Thévenon, 2011). Couple-level employment–fertility dynamics in France could thus be expected to differ from all of these models. The focus on female employment, which in France is considered compatible with the responsibility of raising children, may render a dual-earner constellation more favourable to making the transition to parenthood, compared with situations where one or both partners are not in employment. Moreover, it could mean that both partners’ incomes have an equivalent impact on the transition to parenthood. However, in light of potentially unequal opportunity costs of parenthood as well as gender norms and family polices that do not actively target household-level gender equality, gendered associations could also prevail. Knowledge of couple-level gender dynamics in the employment–parenthood link in France may be of particular interest to policymakers not only in France but also in countries that have only more recently started to invest in work–family policies in view of supporting both the transition to parenthood and gender equality.
I – Theoretical framework
1 – Theoretical perspectives: partners’ employment positions and parenthood
5 Research shows that employment has an impact on fertility (Vikat, 2004; Matysiak and Vignoli, 2008; Vignoli et al., 2012; Jalovaara and Miettinen, 2013; Kreyenfeld, 2015; Wood et al., 2015), but there is still debate about the mechanisms that link them. In addition, couple-level gender dynamics are often overlooked in this context. Here, I discuss how couples’ employment status and income may affect whether and when they make the transition to parenthood. Mapping and accounting for these employment or income positions as such, including stability or change in them, lies beyond the scope of this study. In the following, I discuss how couples’ employment and income status at a given time and their fertility decisions may be related.
6 One strand of explanations suggests that couples make fertility decisions with imperfect information: the decision is made on the basis of anticipated economic (and non-economic) costs and opportunities (Gauthier, 2007). In this framework, the decision to make the transition to parenthood at a certain point in time is understood as the outcome of an assessment of the balance between costs and opportunities. Having and raising a child undoubtedly creates opportunities and adds value to individuals’ lives, but it also takes financial resources and time. A favourable financial position is more likely when both partners are active on the labour market. A favourable position in terms of time availability is more likely when at least one partner is less active or inactive on the labour market or when outsourcing of childcare is a possibility. A trade-off between these two resources is required: the more (less) time spent in employment, the greater (lesser) the financial resources available, but the less (more) time available to raise a child (Becker, 1991). Hence, a balance has to be struck, and couples have to decide at which point in their lives and careers they consider themselves ready to make this major life transition. Here, both the partners’ current employment positions and changes they anticipate in these labour market positions after the transition to parenthood may influence couples’ decisions as to whether and when to have a child (Liefbroer, 2005).
7 In this rational decision-making framework, anticipated opportunity costs resulting from the trade-off between income and time are, in principle, unrelated to gender, meaning that men’s and women’s labour market positions and income have equal weight in the decision to make the transition to parenthood. In this view, any situation in which the financial requirements and time availability preconditions are met at the couple level would be expected to be positively linked with the transition to parenthood, regardless of how much each partner contributes to these couple-level resources. However, other scholars argue that as pregnancy, giving birth, and breast-feeding, for example, are strictly female matters, the birth of a child more strongly impacts women’s labour market positions than those of men (at least around the time of childbirth), yielding different opportunity costs and hence different considerations with respect to the male and female partner’s employment status and income position in the transition to parenthood (Hochschild and Machung, 1989; Grunow et al., 2012).
8 In addition, different gender norms regarding parenthood and employment may reinforce these mechanisms. The ‘doing gender’ perspective suggests that men and women conform to and reproduce dominant gender norms (West and Zimmerman, 1987; Brines, 1993; Schneider, 2011). With respect to the transition to parenthood, this entails women’s greater involvement in childcare tasks, while men’s main responsibility lies in securing financial resources and establishing a solid labour market position. Empirical evidence demonstrates that even today and among dual earners, traditional gender divisions of paid work and childcare emerge after the birth of the first child (Grunow and Evertsson, 2016), which cannot be fully explained by the partners’ pre-birth employment positions (Wood and Marynissen, 2019). In this view, the employment statuses of male and female partners are likely to be differentially linked with the transition to parenthood, and the relative distribution of income between partners may also play a role.
9 However, some scholars have criticized these perspectives because they are based on a high degree of gender role specialization (Oppenheimer, 1994). First, it is argued that the unprecedented rise in women’s educational and labour market participation over the past decades has led women’s earning potential to become increasingly similar to that of men. Hence, gender specialization within couples may no longer yield the most favourable labour market preconditions for parenthood (Winkler-Dworak and Toulemon, 2007). In addition, the employment of both partners may represent not only a more efficient division of labour but also an economic necessity in view of parenthood. Research shows that many women opt to postpone their transition to parenthood until they have established a solid labour market position themselves (Andersson, 2000; Matysiak and Vignoli, 2008). Such a position may entail better access to work–family reconciliation policies, such as parental leave, as well as a higher income, which makes outsourcing childcare and household tasks affordable. Thus, the links of the male and female partner’s employment position and income with the transition to parenthood may be similar, especially in contexts where family policies facilitate work–family reconciliation.
2 – Institutional context of France
10 France is characterized by high fertility levels in the European context, particularly during the period under study (with a total fertility rate of 2.0 in 2010–2014 and 1.9 in 2015–2016) (Méda and Pailhé, 2008; Toulemon et al., 2008; OECD, 2021c). Research has shown that the proportion of childless women is very low in France (Méda and Pailhé, 2008; Pailhé and Solaz, 2012), which means that almost all French women make the transition to parenthood at some point in their lives. Furthermore, employment levels are high for both men and women, and maternal (full-time) employment levels are also relatively high (Toulemon et al., 2008; OECD, 2021a). As the combination of high fertility and high employment levels suggests, France has a long-standing history of family policies. The enrolment of children under age 2 in formal childcare has exceeded the Barcelona childcare target of 33% since the early 2000s (OECD Family Database, 2021), reaching 50% or more since 2011 (Motiejunaite-Schulmeister et al., 2019). Furthermore, nearly all children aged 3 to 6 attend pre-primary education (OECD, 2021b). In addition to outsourcing policies, other family policies providing monetary support (e.g. to employ a home-based childminder) or allowing leave (other than maternity or paternity leave) have been introduced to facilitate the work–family combination. Throughout the 1990s and 2000s, progressive changes in these family policies have taken place, which eventually resulted in the unification of all existing early childhood benefits into a single package in 2004, the prestation d’accueil du jeune enfant (PAJE) (Toulemon et al., 2008). A central feature of the PAJE is that whereas previously parental leave was only available from the second or third child onwards, this new policy also provides a parental leave allowance for parents having their first child, for a period of 6 months after the birth. During this leave, parents receive a very low, flat-rate child-rearing benefit  (Boyer and Fagnani, 2020). Parents can take part-time leave simultaneously and receive a reduced income-replacement benefit. Eligibility is conditional on employment before the birth of a child and on the number of children. Parents of a single child must have worked continuously for 2 years preceding the birth, of which at least 1 year for their current employer.
11 Although French family policies were initially shaped based on the male breadwinner model (Morel, 2007), today they are increasingly targeted towards dual earners (Fagnani, 2002). Moreover, the normative context in France increasingly supports the combination of employment and parenthood for both mothers and fathers from early on. Today, it is well accepted by both individuals and organizations that mothers of young children participate (even full-time) in the labour market and outsource childcare (Toulemon et al., 2008; Rossier et al., 2011). It seems plausible, then, that France’s high fertility, high (maternal) employment context, with a range of family policies and a favourable normative context, could entail gender-equal or gender-neutral labour market preconditions for parenthood. However, several factors may counteract this and entail gendered labour market preconditions. The expansion of work–family reconciliation policies has not necessarily involved promoting gender equality or modifying traditional gender roles (Fagnani, 2002; Morel, 2007). For one, the outsourcing of childcare predominantly alleviates formerly ‘female’ responsibilities. And while almost all mothers took maternity leave and two-thirds of fathers took paternity leave in 2016, parental leave uptake is highly gendered, at 98%–99% female, and has dramatically decreased over the last decade (Boyer and Fagnani, 2020). These trends are unlike those in many other European countries, where overall levels of uptake are rising and fathers are taking their parental leave (Koslowski et al., 2019). In addition to very low wage replacement benefits, suggested potential explanations include persistent gender-unequal distributions of child-rearing and domestic responsibilities within the family, traditional value systems, gender wage gaps, and workplace cultures in France (Boyer and Fagnani, 2020).
3 – Hypotheses
12 The theoretical perspectives discussed above suggest possible mechanisms for both gendered and gender-neutral or -equal associations between partners’ employment status and income and the transition to parenthood. The characteristics of the French context in particular suggest that either type of mechanism could prevail. I propose the following hypotheses on how partners’ combined employment status and relative income at specific points in time may relate to the transition to parenthood.
13 On the one hand, a context understood in terms of rational decision-making, unequal opportunity costs for men and women, and the ‘doing gender’ perspective, and where household-level gender inequalities are not actively targeted by public policies, higher first-birth probabilities should be expected if the female partner is not in employment and in male main-earner couples (Hypothesis 1). On the other hand, the French context is marked by women’s increasing labour market participation and earning potential, combined with strong work–family policies and a normative context that favours maternal employment. This context provides reason to expect higher first-birth probabilities if both partners are in employment, and similar first-birth probabilities regardless of whether it is the male or female partner who provides a larger share of the household income (Hypothesis 2).
II – Data and methods
1 – Data and sample
14 The Permanent Demographic Sample (EDP) is a panel of individuals constructed by the French national statistics office (Institut national de la statistique et des études économiques [INSEE]) from 1968 onwards using microdata from five different sources: (a) vital statistics since 1967; (b) population censuses (five censuses between 1968 and 1999; and annual micro-censuses since 2004); (c) the electoral register since 1990; (d) a panel of employed persons (PANACT) since 1967; and (e) social and tax data since 2011 (Direction des statistiques démographiques et sociales, 2019). The panel provides annual information on individuals born on certain days of the year (between 2 and 5 January, and 1 and 4 April, July, or October), henceforth referred to as EDP-persons. Individuals born on any of these days are followed from the moment they appear in a (micro-)census, a state document (other than a death certificate), the PANACT, or a tax return (Direction des statistiques démographiques et sociales, 2019). Altogether, the EDP allows the study of demographic events, socio-economic trajectories, and geographical, residential, professional, and social mobility.
15 I combined information on EDP-persons’ descent, births, and marriages from vital statistics and information from social and tax data which (in contrast to all other data sources) also include all members of EDP-persons’ households. This allowed me to identify the partners of the EDP-persons and retrieve information on both the EDP-persons and their partners’ employment status and income. I also retrieved information on EDP-persons’ level of education from census data. I could not retrieve information on their partners’ level of education because there are no unique person identifiers that identify data on non-EDP-persons in other data sources. The resulting subpanel spans the 2010–2016 period, as social and tax data were only available from 2011 onwards (data from fiscal year t provides information on year t – 1). From this subpanel, I observed all nulliparous women aged 18 and older with a co-resident partner. The sample was further restricted to couples for whom I had information on both partners’ income. Couples were followed until (a) their first child was born; (b) the female partner reached age 45, the presumed end of women’s reproductive lifespan; (c) the couple separated; or (d) censoring due to death, emigration, or the end of the observation period. This sample provided data on 581,591 couple-years for 200,213 couples, of which 60,962 had their first child between 2010 and 2016.
2 – Analytical strategy and variables
16 Measuring the event of having a first birth on a yearly basis, I estimated discrete-time hazard models of conception leading to a first birth using a complementary log–log link function. This link function allows the exponentiated parameter estimates e(b) to be interpreted in terms of hazard ratios  (Singer and Willett, 2003). In addition, I controlled for the clustering of observations or person-years within persons. I observed EDP-persons for multiple years, meaning that while observations are independent across clusters (EDPpersons), they are not independent within clusters (multiple observations per EDP-person). Therefore, the standard errors in these analyses allow for intragroup correlation, resulting in more robust standard errors, but not affecting the estimated coefficients.
17 I analysed conception leading to a first birth—the event lagged by one (calendar) year—rather than the first birth itself, to avoid reverse causation. The dependent variable takes the value 1 in the year of conception and 0 in all other years. I measured exposure (Tti) as the number of years elapsed since the year in which the female partner turned 18. I used a quadratic specification of the exposure variable, as the hazard of having a first birth typically peaks around certain ages. I considered alternative clocks, such as the number of years elapsed since graduation or start of cohabitation or marriage. Research demonstrates that the age at departure from (full-time) education—and thus the duration since—is an important predictor of the timing of first childbirth (Neels et al., 2017). During periods of enrolment in education, first-birth rates are substantially lower, while departure from education is generally followed by a series of other life course transitions: employment, financial independence, household formation, and transition to parenthood (Neels et al., 2017). Hence, the later an individual leaves education, the later they go through this series of transitions, including a first birth. Operationalizing these alternative clocks proved impossible due to limited information on age at leaving education and left censoring. To make up for this shortcoming, I included interactions between the female partner’s level of education (Eti) and the baseline hazard function. The female partner’s level of education is a time-constant variable with four categories, distinguishing (a) no education or primary; (b) secondary; (c) tertiary or higher; and (d) unknown. Appendix Table A.1 provides the distribution of all variables included in the models and the associated crude birth rates. The latter were calculated as the proportion of first births relative to the number of person-years observed in the category.
18 The main independent variable in the first two models was the partners’ combined employment status (EmplCti). This time-varying variable was derived from social and tax data and based on the male and female partners’ main income sources during the income reference year.  This indicates whether they were in employment (as employees or self-employed) or not (unemployed or inactive). The variable distinguishes couple-years where (a) both partners were employed; or (b) only the female partner was; (c) only the male partner was; or (d) neither was. In the third and fourth models, I included time-varying information on household income and its relative distribution between partners. Household income (hhIncti) is the sum of both partners’ incomes (including income from paid work, as well as individual unemployment benefits, public transfers, etc.) during the income reference year, divided into terciles. I defined relative income (relIncti) as the ratio of the female partner’s income to the household income, and included it in the analyses as a categorical variable, distinguishing (a) male main-earner couples, when the female partner earned less than 40% of the household income; (b) equal-earner couples, when the female partner earned between 40% and 60% of the household income, or when both partners have an income of zero; and (c) female main-earner couples, when the female partner earned more than 60% of the household income. Finally, I controlled for three sociodemographic characteristics shown to affect the probability of a first birth: the age of the male partner (AgeMti), a time-varying indicator of marital status (Mti), and a time-constant indicator of whether the female partner had foreign origins (Fi) (either she or at least one of her parents was not born in France) (Kulu et al., 2017).
19 The descriptive analyses represent couples’ combined employment status and relative income by the age of the female partner, as well as the observed conditional probabilities of making the transition to parenthood by the partners’ combined employment status and relative income. Next, I estimated a series of nested models to examine the relation between couples’ employment status and the transition to parenthood. Model 1 includes the baseline hazard function (exposure variables) in interaction with the female partner’s education, as well as information on the female partner’s origins and the couple’s marital status and combined employment status. Model 2 relaxes the proportional hazards assumption and includes the interaction of combined employment status with the baseline hazard function to examine whether the association between partners’ combined employment status and the transition to parenthood varies with the female partner’s age. Model 3 again builds on the previous model and includes partners’ relative income as well as household income. Finally, Model 4 includes the interaction of relative income with the baseline to examine whether the association between partners’ relative income and the transition to parenthood varies by age.
20 The following equation represents the calculation of the continuous risk accumulated within each time interval in the last, full model (Model 4):
III – Results
1 – Descriptive results: couples’ employment positions, relative income, and first births
22 Figures 1 and 2 show the distribution of couple-years and the observed conditional probabilities of a first birth by partners’ combined employment status and relative income. Figure 1A indicates that when the female partner is aged 20 and older, in the majority of couples both partners are in employment. This share fluctuates around a value of 70%, with the highest values when the female partner is in her late 20s. The share of couples where only the male partner is employed is highest when the female partner is in her late teens and early 20s, decreases to its minimum when she is in her late 20s to early 30s, and fluctuates between 16.0% and 18.5% beyond this age range. Apart from the youngest ages, the share of couples where either only the female partner is employed or neither partner is in the labour force is small and fairly stable. Figure 1B shows that at the youngest ages, observed conditional first-birth probabilities are higher for couples where both partners are not employed and where only the male partner is employed. From the age of 26–27, the highest probabilities of entering parenthood (up to 24.63%) are found in dual-breadwinner couples.
23 Figure 2A highlights the large share of male main-earner couples at young ages and relatively stable percentages of female main-earner couples across ages (around 12%). Similar to the rise in dual-breadwinner couples, the share of equal earners gradually increases with the age of the female partner. However, the share of equal-earner couples  reaches a maximum of 50%–53% between ages 25 and 30 and then decreases again in older couples. Equal-earner couples have high observed conditional first-birth probabilities between ages 27 and 34 (Figure 2B). Probabilities are second-highest in male main-earner couples, while female main-earner couples generally have the lowest probability of making the transition to parenthood until age 35.
Figure 1. Distribution of couples and conditional first-birth probabilities by age of the female partner and partners’ combined employment status, France, 2010–2016
Figure 1. Distribution of couples and conditional first-birth probabilities by age of the female partner and partners’ combined employment status, France, 2010–2016
Figure 2. Distribution of couples and conditional first-birth probabilities by age of the female partner and partners’ relative income, France, 2010–2016
Figure 2. Distribution of couples and conditional first-birth probabilities by age of the female partner and partners’ relative income, France, 2010–2016
2 – Multivariate results: age-specific associations between partners’ labour market characteristics and first-time parenthood
24 Table 1 presents the results of four nested discrete-time hazard models of conception leading to a first birth. The quadratic specification of the baseline indicates that the hazard function first increases, reaches a peak, and then gradually decreases. Consistent with well-established findings in the literature on the education–fertility link (Neels and De Wachter, 2010; Ní Bhrolcháin and Beaujouan, 2012; Neels et al., 2017), the results indicate that the pace of the increase and decrease, as well as the peak age of the hazard function, differ by level of education. However, the extent to which the shape of the hazard function differs by partners’ combined employment status and relative income remains an open question. Table 1 can be used to examine age-specific effects of partners’ combined employment status and relative income, while Figures 3 and 4 visualize the results of the interactions in Model 4 using predicted probabilities.
Table 1. Complementary log–log models of conception leading to a first birth, France, 2010–2016
Conception leading to a first birth (1) vs. no conception leading to a first birth (0) Model 1 Model 2 Model 3 Model 4 Exp(b) Exp(b) Exp(b) Exp(b) Constant 0.107 *** 0.070 *** 0.083 *** 0.073 *** Time (Years since age 18) Linear 1.254 *** 1.358 *** 1.345 *** 1.362 *** Quadratic 0.985 *** 0.982 *** 0.982 *** 0.982 *** Level of education (Ref.: None or primary) Secondary 0.379 *** 0.433 *** 0.432 *** 0.437 *** Tertiary or higher 0.114 *** 0.144 *** 0.140 *** 0.142 *** Unknown 0.669 *** 0.600 *** 0.600 *** 0.600 *** Education × Years since age 18 (Linear) Secondary 1.165 *** 1.135 *** 1.133 *** 1.131 *** Tertiary or higher 1.372 *** 1.312 *** 1.311 *** 1.309 *** Unknown 1.018 1.033 *** 1.030 ** 1.030 ** Education × Years since age 18 (Quadratic) Secondary 0.996 *** 0.997 *** 0.997 *** 0.997 *** Tertiary or higher 0.992 *** 0.994 *** 0.994 *** 0.993 *** Unknown 1.002 *** 1.002 *** 1.002 *** 1.002 *** Partners’ employment status (Ref.: Both employed) Only female employed 0.836 *** 1.580 *** 1.674 *** 1.764 *** Only male employed 0.903 *** 2.676 *** 2.782 *** 2.524 *** Neither employed 0.860 *** 3.433 *** 3.534 *** 3.537 *** Partners’ employment status × Years since age 18 (Linear) Only female employed 0.873 *** 0.878 *** 0.886 *** Only male employed 0.822 *** 0.827 *** 0.835 *** Neither employed 0.771 *** 0.779 *** 0.781 *** Partners’ employment status × Years since age 18 (Quadratic) Only female employed 1.006 *** 1.006 *** 1.005 *** Only male employed 1.007 *** 1.007 *** 1.007 *** Neither employed 1.010 *** 1.009 *** 1.009 ***
Table 1. Complementary log–log models of conception leading to a first birth, France, 2010–2016
Conception leading to a first birth (1) vs. no conception leading to a first birth (0) Model 1 Model 2 Model 3 Model 4 Exp(b) Exp(b) Exp(b) Exp(b) Relative income (Ref.: Equal earners) Male main earner 0.954 *** 1.157 ** Female main earner 0.929 *** 0.911 Relative income × Years since age 18 (Linear) Male main earner 0.977 ** Female main earner 0.978 Relative income × Years since age 18 (Quadratic) Male main earner 1.000 Female main earner 1.002 ** Household income (Ref.: Middle) Low 0.894 *** 0.894 *** High 1.027 ** 1.029 ** Married (Ref.: Cohabiting) 1.775 *** 1.769 * 1.780 *** 1.782 *** Migration background (Ref.: French origin) 1.019 1.028 *** 0.958 *** 1.043 *** Age of male partner 0.996 *** 0.995 *** 0.995 *** 0.995 *** No. of couple-years 581,591 581,591 581,591 581,591 No. of couples 200,213 200,213 200,213 200,213
Partners’ combined employment status
25 Model 1 (Table 1) shows that, controlling for level of education, compared to a situation where both partners are employed, the first-birth hazard is significantly lower if only the female or male partner is employed (16.4% and 9.7% lower, respectively), or if neither is employed (14.0% lower). However, further examination indicates that these associations vary with the female partner’s age. Adding an interaction with the baseline reverses the main effect of partners’ combined employment status (Table 1, Model 2), indicating a positive association between situations where one or both partners are not in the labour force and the first-birth hazard. In the first period, at age 18, the first-birth hazard is 58.0% higher in female breadwinner couples, 167.6% higher in male breadwinner couples, and 243.3% higher in couples where neither partner is employed, compared to dual-breadwinner couples. However, at age 30, for example, the first-birth hazard is 26.7%  lower in female breadwinner couples, 30.5% lower in male breadwinner couples, and 36.5% lower in couples where neither partner is employed, compared to dual-breadwinner couples (Table 1, Model 2). Figure 3 indicates that these reversed associations are mainly present when the female partner is particularly young. From the age of 25 onwards, it is dual-breadwinner couples that show the highest probability of a first birth, peaking at ages 28–29 at around 28%. Patterns among couples where only the male or female partner is employed are very similar. Whereas male breadwinner couples display slightly higher first-birth hazards than female breadwinner couples at young ages, from the age of 25 onwards, the two curves are practically identical. Furthermore, these curves reach their peak (around 23%) a little earlier (between ages 26–28) than dual-breadwinner couples. Finally, couples where neither partner is employed show the highest first-birth hazards at very young ages, but then consistently show the lowest first-birth probabilities among all types of couples. The finding of higher first-birth hazards in couples with only one partner employed or no employment at young ages is not in line with previous findings for France, which have indicated a delay in first-time parenthood in cases of male unemployment and of insecure employment positions for women (Meron et al., 2002; Pailhé and Solaz, 2012). However, Meron et al. (2002) also found that not participating in the labour force, especially at the start of life together as a couple, is associated with accelerated entry into parenthood. Moreover, potential selectivity of the analytical sample for these younger ages may also play a role. Women younger than 25 who are not in employment and who have a co-resident partner who also is not employed may have quite different fertility intentions and may perceive the opportunity costs of childbearing to be lower compared to their peers in employment, still enrolled in education, or without a co-resident partner.
Figure 3. Predicted first-birth probabilities by partners’ combined employment status in France (2010–2016), Model 4
Figure 3. Predicted first-birth probabilities by partners’ combined employment status in France (2010–2016), Model 4Note: These predicted probabilities were calculated assuming an average profile for all other covariates.
Partners’ relative income
26 Model 3 shows that—controlling for household income (positively related to the transition to parenthood) and still for combined employment status— the first-birth hazard is significantly lower in male (4.6% lower) and female (7.1% lower) main-earner couples than in equal-earner couples. Model 4 shows that this association varies significantly with age of the female partner. Figure 4 shows that up to the age of 25, first-birth hazards are similar for equal-earner and male main-earner couples. Thereafter, equal-earner couples consistently exhibit slightly higher first-birth probabilities than male main-earner couples, reaching their peak of 25.3% at age 27 of the female partner. Finally, female main-earner couples display a later peak and consistently lower first-birth hazards than those of equal and male main earners. However, their first-birth probabilities start to slightly exceed those of male main-earner couples from the age of 31 onwards, and those of equal-earner couples from the age of 35 onwards.
Figure 4. Predicted first-birth probabilities by partners’ relative income in France (2010–2016), Model 4
Figure 4. Predicted first-birth probabilities by partners’ relative income in France (2010–2016), Model 4Note: These predicted probabilities were calculated assuming an average profile for all other covariates.
27 I performed several sensitivity analyses with indicators to account for recent changes in partners’ combined employment status or relative income, and lagged the dependent variable by 2 years instead of 1 to rule out reverse causality. Two types of change variables were constructed and analysed separately. The first consisted of two dummies indicating whether partners’ combined employment status or relative income in calendar year t had changed in comparison to the preceding calendar year t – 1. The second type was similarly constructed but also indicates the direction of change (more or less gender-equal). The general pattern of associations between partners’ combined employment status and relative income and the transition to parenthood did not change when controlling for recent changes. Furthermore, the results indicated that a recent change in partners’ combined employment status is negatively associated with the transition to parenthood, whereas a recent change in partners’ relative income is not significantly associated with first-birth hazards. Additional analyses showed that the direction of change also matters: a shift to a more equal employment status yields higher first-birth hazards, while a shift to a more equal earner constellation yields lower first-birth hazards. While these analyses were only included as robustness checks in this study, these initial results suggest that examining change may be a fruitful path for future research.
28 The sensitivity analyses using a larger time lag show similar patterns of effects when the event of having a first birth is lagged by 2 years instead of 1. These results indicate that although the 1-year lag may in some cases yield overlap between the measurement of partners’ employment positions and income and the moment of (potential) conception, potential reverse causality does not appear to substantially bias the results. Hence, given the limited observation period in this study, I opted to use the 1-year lag in the main analyses to keep the sample size as large as possible.
29 Recent decades have seen an unprecedented increase in women’s labour force participation, earning potential, and income—without a corresponding shift to less gendered parenthood responsibilities. This study examined potential age-specific associations between partners’ employment status and relative income and the transition to parenthood in France. The results show significant age-specific associations between couples’ combined employment status and conception leading to a first birth. At very young ages, couples where neither partner is in the labour force are most likely to make the transition to parenthood, followed by couples where only one partner (male or female) is in employment. From the age of 25, these associations are reversed, with dual-breadwinner couples exhibiting the highest first-birth hazards and couples where neither partner is employed the lowest. With respect to partners’ relative income, equal-earner couples consistently exhibit the highest first-birth hazards up to the age of 35. Furthermore, male main-earner couples show a very similar pattern of first-birth probabilities to equal-earner couples, whereas those of female main-earner couples are generally lower. The findings that first-birth hazards are highest among couples where both partners are employed and among equal-earner couples suggest that in France, gender equality in couples’ labour market positions is considered the most favourable situation for the transition to parenthood. This provides support for Hypothesis 2. Also consistent with Hypothesis 2 is the finding that the probability of a transition to parenthood among couples with only one partner in employment is similar whether the employed partner is male or female. In contrast, the first-birth hazards of female main-earner couples are significantly lower than those of male main-earner couples. This finding is more consistent with Hypothesis 1, assuming higher opportunity costs for women and persistent gendered parenting norms. This suggests that gendered mechanisms may still be at play in relationship to partners’ incomes. Finally, the association between partners’ combined employment position and relative income and the transition to parenthood varies depending on the female partner’s age.
30 In the study period in France, gender-equal positions in couples’ employment and income yielded the highest first-birth probabilities. The relation between partners’ employment status and the transition to a first birth seems to have become increasingly gender-neutral. However, no similar shift towards gender-neutral patterns in the association between partners’ income and first-time parenthood has been observed thus far. These findings suggest that the French context of well-developed work–family support policies seems to provide a suitable environment to gradually reduce gender disparities in the employment–parenthood link. However, similar conclusions cannot be drawn with respect to gender equality in the income–parenthood link at the couple level, where gendered patterns continue to prevail.
31 Finally, this study has several limitations and thus offers related avenues for future research. First, the analysis of contemporary associations does not consider the stability of, or recent changes in, partners’ combined employment status or relative income. Consequently, potential self-selection mechanisms remained undetected. Childbearing-prone couples may self-select into combined employment statuses they consider most favourable to having a first child by the time they consider themselves ready for this transition. Hence, the examination of couples’ entire employment trajectories, rather than yearly measurements, in relation to the transition to parenthood should be a fruitful path for research. Secondly, while the EDP is a large and rich dataset, well suited to studying demographic events, it also has some limitations with respect to the availability of variables. Notably, the data did not allow me to determine whether individuals were enrolled in education in a given year. Furthermore, because social and tax data were only available from the 2011 fiscal year onwards, it was only possible to identify partners and retrieve information on their activity status and income from 2010 onwards. The observation period was thus relatively short, so employment and income stability over a longer period could not be operationalized. In addition, it was only possible to identify partners in the social and tax data, not in the other data sources (e.g. censuses). Information on variables that were available for many EDP-persons, such as education, employment regime, and organization (size, public vs. private, etc.), could not be retrieved for partners. One option for further study would be to look for EDP-couples (couples where both partners are sampled EDP individuals) and examine the selectivity and size of such a sample. If it were to prove suitable for analysis, this sample could offer an extremely valuable way to examine couple-level gender dynamics in the employment–parenthood link in more detail and over a longer period.
Table A.1. Summary statistics of the main variables
No. couple-years % couple-years Birth rate Conception leading to a first birth (tv) No 520,629 89.52 Yes 60,962 10.48 Partners’ combined employment status (tv) Both employed 428,663 73.71 10.70 Only female employed 39,822 6.85 10.18 Only male employed 89,411 15.37 9.63 Neither employed 23,695 4.07 10.28 Relative income (tv) Male main-earner couple 250,500 43.07 9.13 Equal-earner couple 258,021 44.36 11.79 Female main-earner couple 73,070 12.56 10.50 Household income (tv) Low 193,868 33.33 10.57 Middle 193,863 33.33 11.22 High 193,860 33.33 9.66 Age of the female partner (tv) 18–20 9,683 1.66 16.30 21–25 95,597 16.44 15.36 26–30 122,558 21.07 21.17 31–35 72,227 12.42 18.30 36–40 96,568 16.60 4.80 41–45 184,958 31.80 0.48 Age of the male partner (tv) ≤ 25 63,671 10.95 12.93 26–30 118,256 20.33 19.58 31–35 82,899 14.25 21.23 36–40 78,546 13.51 10.01 41–45 209,313 35.99 1.88 ≥ 51 28,906 4.97 0.63 Level of education, female partner (tc) None or primary 160,034 27.52 7.05 Secondary 105,065 18.07 10.38 Tertiary or higher 184,676 31.75 12.64 Unknown 131,816 22.66 11.70 Marital status (tv) Unmarried cohabitation 283,789 48.80 12.63 Married 297,802 51.20 8.44 Migration background, female partner (tc) French origin 482,165 82.90 10.04 Foreign origin 99,426 17.10 12.60 Total 581,591 100.00
Table A.1. Summary statistics of the main variablesNote: (tv) indicates time-varying variables, (tc) indicates time-constant variables.
The level was set at €397.20 per month when used full-time in 2020. Lower-income parents receive a supplementary means-tested allowance, increasing the benefit to €581.82 in the case of full-time leave uptake.
The hazard is the conditional probability that the event will take place between time t and t + Δt, given that it did not happen before t. The hazard ratio compares different groups’ hazard of experiencing the event.
This is the calendar year preceding that of the observed or potential birth of the first child.
Equal-earner couples also include couples where neither partner has an income, but this share is very small and relatively stable at around 1.9%.
((1.580 × 0.87312 × 1.006144) – 1)*100