1 – Microeconomic rationality

8From a microeconomic perspective, a partner’s decision to exit the labour force or take parental leave centres on maximizing a joint utility function. Resources and time are pooled and invested in the most cost-efficient manner (Mincer, 1962; Becker, 1991; Browning et al., 2014). Thus, the partner with the lowest opportunity cost (e.g. forgone income) of time spent on non-market activities will be more likely to exit the workforce (Mincer, 1962). Economic bargaining models similarly assert that the division of labour and household work is a product of intra-household negotiations (Lundberg and Pollack, 1996) in which both partners try to maximize their individual utility, and the partner with the greatest returns from employment has the most bargaining power. This power allows one partner to avoid housework and childrearing, which are assumed to be ‘unpleasant’ or less-rewarding activities (Geisler and Kreyenfeld, 2011).

9Concerning to the decision to opt out of paid work, we derive four potential mechanisms from microeconomic theory. First, individuals with higher hourly wages than their partner are less inclined to exit employment not only due to the high opportunity costs (forgone income), but also due to their higher bargaining power (Begall and Grunow, 2015). Second, the partner with higher pre-birth work hours is less likely to leave the labour force because having a stronger attachment to paid labour is likely to go hand in hand with greater experience, which in turn positively affects future access to paid work (Herrarte et al., 2012). Third, the partner with the most stable pre-birth employment is least likely to leave the labour force, as employment security yields utility and bargaining power. Fourth, workplace characteristics also determine work– family compatibility. In addition to numerous sector-specific factors, large workplaces are more likely to have in place routine practices that facilitate combining work and family, e.g. on-site day-care centres or greater access to parental leave, policies that are likely to stimulate parental employment (Bygren and Duvander, 2006).

10Similarly, couples’ parental leave uptake is expected to reflect pre-birth employment characteristics. First, the parent with the highest hourly wage will be less likely to use leave in order to limit the couple’s loss in income (Sundström and Duvander, 2002). Second, partners with more working hours are more likely to use parental leave because only full-time employees are eligible for the popular part-time parental leave schemes (Merla and Deven, 2010). What is more, individuals with fewer working hours need less leave to spend time with their children. Third, parents with stable employment patterns before childbearing are more likely to use parental leave because employment intensity is an eligibility criterion (Merla and Deven, 2010), and employees in stable employment positions potentially face less negative career consequences from taking leave (Kil et al., 2018). Fourth, large firms typically have routine parental leave practices and are more likely to offset the temporary loss of labour, which may result in fewer obstacles to leave-taking (Bygren and Duvander, 2006).

2 – Rationality in a normative and institutional context

11If women outperform men in the labour force (e.g. earn higher wages), the persistence of couples’ traditional gender specialization in response to childbearing constitutes a paradox, from a microeconomic point of view. Hence, it is necessary to go beyond microeconomic considerations and focus on normative and institutional contexts. In this regard, Duncan and Edwards (1997) coined the term ‘gendered moral rationalities’ to describe the behaviour of individuals who adhere to a collective understanding of which activities are best for men and women to perform. Thus, gendered labour force responses to childbearing reflect an expectation that men should act as financial providers, while women are considered caregivers (Haas et al., 2002). These gender norms are continuously reproduced by emphasizing gender in social interactions (West and Zimmerman, 1987; Lupton and Barclay, 1997; Coltrane, 2000), with deviations incurring social penalties (Heilman et al., 2004) or compensation behaviour (Brines, 1994). In addition, institutional factors such as longer maternity leaves versus paternity leaves in Belgium may reinforce gendered behaviours, as these systems introduce a gendered habitus among new parents.

3 – Previous research and hypotheses

12This article assumes that microeconomic rational theory is helpful for understanding the micro-level association between dual-earner couples’ relative pre-birth employment characteristics and subsequent parental employment strategies, while contextual factors can explain why parental employment strategies remain gendered in general (Schober, 2011). This study aims to answer two complementary questions, each of which frame two corresponding hypotheses.

13First: Are partners’ employment and leave-taking strategies linked to their relative pre-birth employment characteristics? According to microeconomic theories, gendered parental employment and leave strategies reflect gendered pre-birth employment characteristics. Regarding labour force exits, we expect that the partner with the highest wage, most working hours, most stable employment, and largest workplace will be less likely to exit the labour force compared to his/her partner (Hypothesis 1A). Previous results for the US (Sanchez and Thomson, 1997), UK (Kanji, 2011), Germany (Kühhirt, 2012), and the Netherlands (Begall and Grunow, 2015) indicate that, compared to their partners, women’s pre-birth relative wages and working hours positively relate to their employment following childbirth. Previous research also shows that mothers’ employment in the Netherlands is positively related to occupational status relative to their partners (Begall and Grunow, 2015); while in Spain it is negatively related to their partners’ pre-birth working hours (Herrarte et al., 2012). With respect to parental leave uptake, we assume that the partner with the lowest wage, most working hours, most stable employment, and largest workplace will be more likely than his or her partner to use parental leave (Hypothesis 1B). Although this mechanism is expected to hold for both male and female partners, support in available literature stems mostly from research indicating higher leave uptake by fathers whose female partners are higher earning and/or higher educated in Germany (Geisler and Kreyenfeld, 2011) or Sweden (Sundström and Duvander, 2002). Further evidence for the latter country also indicates a positive relationship between fathers’ workplace size and their leave uptake (Bygren and Duvander, 2006).

14The second question is: Would gendered parental employment and leave uptake strategies among employees be reversed by reversing the gendered distribution of pre-birth employment characteristics within couples? In line with contextual theories, we assume that, beyond pre-birth relative labour market positions, parents at the aggregate level will generally exhibit gendered strategies. Even in conditions where male and female partners have each other’s pre-birth employment position, gendered responses to childbearing persist due to gender roles being conformed to and reproduced (Brines, 1994; Duncan and Edwards, 1997; Coltrane, 2000) and to gendered institutions (e.g. longer maternity than paternity leave). This implies that female partners at the aggregate level are more likely to exit the labour market, beyond the distribution of labour market positions by gender (Hypothesis 2A), and that female partners are more likely to use parental leave, beyond the distribution of labour market positions by gender (Hypothesis 2B). Previous research similarly interprets persisting gendered labour force responses to childbearing when controlling for pre-birth labour force positions as a potential indication of gendered parenting norms and institutions (Kühhirt, 2012).

1 – Data

15We use data from the Belgian Administrative Socio-Demographic Panel (ASD Panel) that was constructed using longitudinal microdata from the National Register and the Crossroads Bank for Social Security. The ASD Panel is representative of the female population aged 15–50 years legally residing in Belgium between 1 January 1999 and 31 December 2010. Apart from the sampled women, the panel also includes all official members of the same household on 1 January in each year. Thus, the ASD Panel is representative of all co-residential heterosexual couples of whom the woman was of childbearing age and residing in Belgium in the period 1999–2010.

16This article exploits the longitudinal character of the ASD Panel by assessing how couples’ parental employment and leave use in the quarters after childbearing varies as a function of their pre-birth employment characteristics. As the potential reorganization of paid work in households is expected to differ by parity, analyses are performed separately for first and higher-order births. Parity is measured via female partners’ fertility histories. We observe couples from the moment of childbirth until the youngest child reaches age 6; they separate; a partner emigrates, dies, or becomes self-employed. When analysing one-child parents, couples are also censored 1 year before the next birth. The ASD Panel includes 12,546 couples who experience at least one birth in the 2001–2010 period and were a dual-earner household 1 year before birth. Couples displaying self-employment 1 year before birth (n = 2,355) are not considered in this study, since the pre-birth employment characteristics of interest are not applicable (e.g. size of workplace) and/or not available (e.g. working hours). After excluding couples that are not observed for 2 continuous years before childbearing (n = 1,366) and couples with missing values for the employment characteristics measured 1–2 years prior to birth (n = 434), this study relies on a sample of 8,391 couples, among which 11,083 births occur during 139,046 couple-quarters. For this sample, labour market positions as well as leave uptake are observed at the end of every quarter. Although it is not possible to distinguish parental leave from other types of leave within the broader system of career breaks, previous research shows that most leave-taking following childbirth is parental leave (Kil et al., 2018).

2 – Models

17The analytical strategy of this study is to use pooled multinomial logit models for relating partners’ relative pre-birth employment characteristics to the couples’ parental employment and leave-taking strategies after childbearing. To capture varying strategies over time since childbirth, the age of the youngest child is used as a time variable with three categories: younger than 1 year; below pre-primary school age (1–2.5 years); and below primary school age (2.5–6 years). The first multinomial logit model distinguishes (i) female labour force exit (base category) from (ii) male exit, (iii) two exits, or (iv) no exits, doing so separately for couple-quarters after the birth of a first child (Model 1A, n = 51,857) or a higher-order child (Model 1B, n = 87 189) (Table 1, upper panel). An exit implies leaving paid employment to enter either inactivity or unemployment. The uptake of parental leave is not considered an exit, as the employment contract remains intact, and such leave in Belgium is mostly used part-time and thus involves continuing in paid employment (Anxo et al., 2007). The second model estimates parental leave uptake versus employment only for couple-quarters after childbirth without a labour force exit after birth. It also excludes couples in which one or both partners are not eligible (n = 328). We distinguish (i) female leave uptake (base category) from (ii) male uptake, (iii) dual uptake, or (iv) no uptake for couple-quarters after the birth of the first child (Model 1A, n = 45,814) or a higher-order child (Model 1B, n = 75,982). Given that the gendered character of parental employment and leave strategies is central to this study, we focus on the probabilities of male versus female labour force exit and male versus female leave uptake.

18All independent variables of interest in this study are relative pre-birth employment characteristics (Table 1, lower panel). First, pre-birth relative hourly wage equals the ratio of the female partner’s hourly wage to the household hourly wage (= sum of partners’ hourly wages), for which we distinguish (i) male higher wage couples (in which the female partner’s wage is less than 45%) from (ii) female higher wage couples (in which the male partner’s wage is less than 45%) and (iii) similar wage couples. The wages used to construct this measure are gross wages divided by the employment percentage (a full-time contract is 100%). Second, we include an approximation of pre-birth relative working hours by distinguishing couples in which (i) the male partner has the highest employment percentage (the female partner’s employment percentage is less than 45% of the household employment percentage); (ii) the female partner has the highest employment percentage (the male partner’s employment percentage is less than 45%); and (iii) couples with similar employment percentages. Third, pre-birth relative employment intensity is measured as accumulated employment experience, which is the number of quarters in which a person was employed in the preceding 2 years. We distinguish couples with (i) higher male employment intensity (female partner’s employment intensity is less than 45% of the household employment intensity); (ii) higher female employment intensity (male partner’s employment intensity is less than 45%); and (iii) equal employment intensity. Fourth, based on a predefined indicator of workplace size in the register data (using the following ranges of values: less than 5, 5–9, 10–19, 20–49, 50–99, 100–199, 200–499, 500–999, and more than 999 employees), this study assesses pre-birth relative workplace size by distinguishing couples in which (i) the male partner has a larger workplace, (ii) the female partner has a larger workplace, and (iii) both partners are employed in the same size category.

Table 1

Distribution of couple-quarters by employment/leave strategies and births by relative pre-birth employment characteristics by parity, Belgium, 1999–2010

First births Second and higher-order births Couple-quarters (n) % Couple-quarters (n) % Model 1A Model 1B Employment strategy Neither exit 47,227 91.07 78,518 90.05 Male exits 1,210 2.33 2,037 2.34 Female exits 3,220 6.21 6,081 6.97 Both exit 200 0.39 553 0.63 Total 51,857 100 87,189 100 Model 2A Model 2B Leave strategy No leave uptake 36,954 80.66 55,334 72.83 Male leave uptake 802 1.75 2,274 2.99 Female leave uptake 7,784 16.99 17,654 23.23 Dual leave uptake 274 0.60 720 0.95 Total 45,814 100 75,982 100 First births Second and higher-order births Couples (n) % Couples (n) % Relative hourly wages Male’s higher 1,940 39.79 2,845 45.48 Equal 1,954 40.08 2,219 35.47 Female’s higher 981 20.12 1,192 19.05 Relative hours Male’s more 706 14.48 2,116 33.82 Equal 4,031 82.69 3,944 63.04 Female’s more 138 2.83 196 3.13 Relative workplace size Male’s larger 1,559 31.98 2,001 31.99 Equal 1,201 24.64 1,502 24.01 Female’s larger 2,115 43.38 2,753 44.01 Relative employment intensity Male’s higher 309 6.34 488 7.8 Equal 4,436 90.99 5,639 90.14 Female’s higher 130 2.67 129 2.06 Total 4,875 100 6,256 100

Distribution of couple-quarters by employment/leave strategies and births by relative pre-birth employment characteristics by parity, Belgium, 1999–2010

Note: The sample distribution of all covariates is available upon request.

19In addition, as previous research indicates that absolute employment characteristics at the individual and household levels also affect work–family strategies (Kühhirt, 2012; Schober, 2011), we control for the corresponding absolute pre-birth positions at the household level (see Appendix Tables A.1 and A.2) (household total wage, sum of working hours of both partners, sum of the employment intensity of both partners) and the female partner’s employer size. In order to control for numerous sector-specific characteristics that may impact parental employment and leave-taking in various undistinguishable ways, we control for both partners’ pre-birth sector of employment: agriculture, mineral extraction, industry; wholesale, retail; logistics, storage, energy distribution; education; public administration, extraterritorial organizations; health services, social care; art, leisure, recreation, other services; finance, estate; administration, support services, professionals; and hotel and catering. In addition, the models on labour force exit include a time-varying approximation of parental leave eligibility, which indicates continuous employment by the same employer for at least four of the preceding five quarters. This variable is used as a selection variable in the models for parental leave. Finally, as pre-birth employment characteristics may be influenced by pre-birth leave uptake, the models for couples with two or more children include a dummy for the use of pre-birth parental leave by both parents.

20The analyses also control for sociodemographic characteristics that have been found to impact labour force participation and parental leave use. As work–family strategies vary depending on the number of children, models for couples with two or more children control for parity: two, three, four, and five or more children. Assuming that couples who become parents at unconventionally young or old ages may adopt different parental employment strategies, we add a linear and squared term for the age of the father and the mother in years. As previous research indicates that parental employment and leave uptake varies by ethnic origin (Kil et al., 2017, 2018), we control for couples’ origin (based on partners’ and their parents’ country of birth): (i) couples with two Belgian origins, (ii) a male partner with a non-Belgian origin, (iii) a female partner with a non-Belgian origin, (iv) same-origin non-Belgian couples, (v) mixed-origin non-Belgian couples, and (vi) couples with unknown origin(s). Further, as unmarried cohabitation is associated with progressive values such as gender egalitarianism, we include a dummy variable indicating whether a couple is married. Finally, in line with research highlighting the importance of contextual correlates of couples’ employment patterns (Vitali and Arpino, 2016), we include female and male quarterly unemployment rates with a 1-year time lag and region of residence.

21In addition to using longitudinal micro-level models to assess whether couples’ parental employment strategies relate to pre-birth employment positions in the expected directions (Hypothesis 1), this study assesses whether pre-birth employment characteristics have a strong enough impact to considerably alter the gendered response to childbearing, particularly in cases where the distribution of pre-birth employment characteristics may be different (Hypothesis 2). Using parameter estimates obtained from Models 1 and 2, this study simulates parental employment strategies under a hypothetical scenario in which male and female partners in our sample have each other’s pre-birth employment characteristics. This simulation is performed in three steps. First, we estimate Models 1 and 2 by using the observed data at the micro level. Second, we substitute the male and female partners’ pre-birth employment characteristics (pre-birth wages, employment percentage, employment intensity, and size of workplace). Third, we combine the hypothetical distribution of pre-birth employment characteristics with the micro-level models by calculating out-of-sample predicted parental employment and leave-taking strategies for the hypothetical gender distribution of pre-birth employment characteristics, using the estimated associations in the observed data.

1 – Parental employment strategies

22Figure 1 displays the observed parental employment and leave-taking strategies for our sample, with particular attention to the constellation in which one partner takes leave or exits the labour force versus the continuation of the dual-earner model. Four observations are of particular interest. First, by definition, our sample of pre-birth dual-earner couples includes solely couples in which both partners were employed or on parental leave 1 year before birth. The infrequent uptake of leave 1 year before the first birth (Figure 1A), which is mostly leave to provide palliative care or take care of a severely ill relative, contrasts with a higher prevalence of leave uptake before a higher-order birth (Figure 1B), which is mostly related to childbearing (Kil et al., 2018). Second, our sample of dual-earner couples 1 year before birth exhibits a higher prevalence of non-employment 2 years before birth. This instability in labour market positions prior to childbearing may reflect anticipation behaviour in the sense that these couples prefer having a child while they are both employed (Wood et al., 2017). Third, after the birth of a child, the majority of our sample exhibits a continuation of the dual-earner model. The share of dual-earner couples—in which both parents work or use leave—hovers, respectively, around or slightly below 90% of our sampled couples for first and higher-order births. Fourth, the birth of a child is found to coincide with increasing gender differentials, as reflected in larger shares of female leave uptake and female labour force exits compared to their partners. Our multivariate analyses aim to assess the degree to which these gendered labour market responses to childbearing are related to pre-birth employment characteristics.

2 – The importance of pre-birth relative employment characteristics

23Controlling for the aforementioned covariates, partners’ relative pre-birth hourly wages significantly affect the probability of a male (versus a female) labour force exit among one-child couples (Table 2). Male higher wage couples have ((1 – 0.508) × 100) 49.2% lower odds of a male exit compared to equal wage couples. In contrast, when the female’s pre-birth wage is higher, a male exit is more likely (89.2% higher odds). This finding holds for first and subsequent births. With respect to working hours for higher parity couples, we find the expected negative effects of higher male working hours and positive effects for female working hours on a male labour force exit. For one-child couples, we similarly find a positive, though insignificant, relationship between higher female working hours and a male labour force exit; whereas the weak positive association between higher male working hours and a male employment exit is not consistent with our expectations. The effects of relative employment intensity indicate 78.6% lower odds of a male exit among one-child couples in which the male partner has higher pre-birth employment intensity and 81.1% higher odds in the opposite situation. The corresponding effects for higher parity couples are 65.1% lower odds when the male partner has higher pre-birth employment intensity and 569.2% higher odds when the female partner has higher employment intensity. The magnitude of the latter effect is likely to be affected by the low sample size for couples in which the female partner had higher employment intensity (Table 1). Finally, the effects of relative workplace size show that couples with a larger female workplace have 56.7% higher odds of exhibiting a male labour force exit after a first birth and 47.5% higher odds of a male exit after a higher-order birth. However, couples with a larger male workplace do not differ significantly from couples with equal workplace sizes. In sum, although some insignificant effects are not consistent with our expectations, these findings provide considerable evidence for microeconomic specialization or bargaining based on pre-birth employment characteristics. Partners with lower wages, fewer working hours, lower employment intensity, and employment in smaller workplaces are more likely to exit the labour force instead of their partner (Hypothesis 1A).

Figure 1

Distribution (%) of couple-quarters by parental employment strategy, Belgium, 2001–2010

Distribution (%) of couple-quarters by parental employment strategy, Belgium, 2001–2010

Note: Couples are censored in analyses for birth order x from four quarters before birth x + 1.

Table 2

Multinomial logit models of parental employment (Model 1: labour force exit vs. employment and leave uptake) (Model 2: leave uptake vs. employment)

Model 1. Labour force exits (base category: female exits) A. First births B. Second and higher-order births Male exits e(b) Both exit e(b) No exit e(b) Male exits e(b) Both exit e(b) No exit e(b) Relative hourly wages (Ref.: Equal) Male’s higher 0.508 *** 0.507 *** 0.665 *** 0.519 *** 0.656 *** 0.705 *** Female’s higher 1.892 *** 0.873 0.938 1.531 *** 1.307 1.223 *** Relative working hours (Ref.: Equal) Male’s more 1.166 0.663 0.731 *** 0.675 *** 1.252 0.991 Female’s more 1.394 0.663 1.002 1.467 * 3.976 *** 1.792 *** Relative employment intensity (Ref.: Equal) Male’s higher 0.214 *** 0.982 0.569 *** 0.349 *** 0.588 ** 0.575 *** Female’s higher 1.811 ** 0.672 1.598 ** 6.692 *** 3.041 *** 2.209 *** Relative workplace size (Ref.: Equal) Male’s larger 1.012 1.336 1.216 ** 1.169 0.440 *** 1.104 * Female’s larger 1.567 *** 1.169 1.085 *** 1.475 *** 0.659 ** 1.021 Model 2. Parental leave uptake (base category: female leave use) A. First births B. Second and higher-order births Male uses e(b) Both use e(b) No use e(b) Male uses e(b) Both use e(b) No use e(b) Relative hourly wages (Ref.: Equal) Male’s higher 0.718 0.394 *** 0.918 ** 0.557 *** 0.524 *** 1.017 Female’s higher 1.424 0.824 0.938 0.799 *** 0.840 1.011 Relative working hours (Ref.: Equal) Male’s more 3.007 *** 5.291 *** 1.424 *** 1.188 * 0.550 *** 1.460 *** Female’s more 2.742 ** 15.867 *** 0.770 * 0.695 ** 0.947 0.608 *** Relative employment intensity (Ref.: Equal) Male’s higher 1.379 * 0.765 0.873 ** 1.534 *** 1.653 *** 1.015 Female’s higher 0.701 *** 0.593 ** 0.935 0.682 *** 0.682 *** 0.914 *** Relative workplace size (Ref.: Equal) Male’s larger 1.504 13.681 ** 1.267 0.762 1.062 1.051 Female’s larger 0.272 * 8.223 * 0.840 0.382 ** 1.121 0.744 ** Model parameters −2LL 24,906.1 50,114.91 df 165 180 n couples 4,600 5,251 n couple-quarters 46,174 77,253

Multinomial logit models of parental employment (Model 1: labour force exit vs. employment and leave uptake) (Model 2: leave uptake vs. employment)

Note: The effects of all covariates are available in the Appendix.
Statistical significance: * p < .05, ** p < .01, *** p < .001.

24As for parental leave uptake, the odds of male (versus female) uptake among one-child parents are 42.4% higher for pre-birth female higher wage couples and 28.2% lower for pre-birth male higher wage couples. Although this pattern of effects is in line with our expectations, the differences are not statistically significant. For higher parity couples, the odds of male parental leave uptake are lowest for male pre-birth higher wage couples. However, female pre-birth higher wage couples also exhibit significantly lower odds of male leave uptake compared to equal wage couples, which is not in line with microeconomic theory. The effects of pre-birth relative working hours for higher parity couples are in line with our expectations, as couples with more male and more female working hours are respectively more likely to exhibit male and female leavetaking. In addition, the higher odds of male leave uptake among one-child couples with higher male pre-birth working hours are expected. However, one-child couples with higher female pre-birth working hours also exhibit higher odds of male leave-taking compared to couples with equal pre-birth working hours. The latter may be related to the small sample size of couples with higher female pre-birth working hours (Table 1). As expected, the occurrence of higher female employment intensity before birth is related to lower odds of male leave uptake. Finally, our hypothesis is supported by the significant negative effects of a larger female pre-birth workplace on male parental leave uptake, which further suggests that larger workplaces for women yield easier access to parental leave. In sum, these findings provide limited and mixed support for our hypothesis that partners with lower pre-birth wages, more work hours, higher employment intensity, and employment in a larger workplace will be more likely to use parental leave instead of their partner (Hypothesis 1B). This mixed support contrasts with the stronger support for our microeconomic expectations regarding gender differences among those exiting employment and entering inactivity or unemployment (Hypothesis 1B).

25Hence, these micro-level analyses indicate that, in line with microeconomic theory, gender differences in employment exits after childbearing are related to gender differences in pre-birth employment characteristics, implying that any embrace of a male breadwinner model is less likely to occur in cases where female pre-birth wages, working hours, employment intensity, and/or workplace size are all greater. When considering the predictive power of pre-birth relative employment characteristics towards the gendered uptake of parental leave, this seems to fall below our microeconomic expectations. However, both for parental employment and leave uptake strategies, questions persist. For example, despite there being a clear relationship between women’s relative pre-birth employment characteristics, subsequent parental employment, and leave uptake strategies, it remains unclear whether or not this relationship is strong enough a predictor to reverse gender inequality if we were to reverse the pre-birth gender distribution in employment characteristics.

3 – The aggregate-level impact of pre-birth labour force characteristics

26Would parental employment and leave uptake strategies be considerably different if the gendered distribution of pre-birth employment characteristics were reversed? Figure 2 presents parental employment strategies that were predicted based on the observed sample (i.e. predicted probabilities) and compares them to corresponding out-of-sample predicted probabilities of parental employment and leave-taking strategies for the hypothetical situation in which both partners switch pre-birth wages, working hours, employment intensity, and workplace size (i.e. simulated probabilities). The most common strategy—dual-earner couples without any employment exits or leave uptake (cf. Figure 1)—is not shown in Figure 2 so that we may focus on gendered parental employment and leave-taking strategies. As probabilities add up to unity, the remaining category can be calculated as the complement of the sum of the three represented strategies. The upper panel indicates that, under the condition of substituted pre-birth employment characteristics within couples, the probability of a female exit decreases, while that of a male exit increases; yet, the occurrence of a dual exit remains relatively stable. For one-child and especially higher parity couples, our findings indicate that the labour market response to childbearing becomes considerably more gender equal. However, even under the extreme and hypothetical situation that all male and female partners have each other’s employment characteristics, the probability of a female labour force exit clearly exceeds the probability of a male exit. These findings thus support the hypothesis that, on the aggregate level, female partners are more likely to exit the labour market, beyond the gender distribution of pre-birth employment characteristics (Hypothesis 2B).

27With respect to the gendered parental leave strategies that are displayed in the lower panel of Figure 2, two observations are particularly relevant. First, regardless of whether the observed or manipulated sample is used, the gender difference in parental leave uptake is much larger than the corresponding difference in labour force exits, especially for higher-order births. Hence, similar to other Western non-Nordic European countries (Geisler and Kreyenfeld, 2006; Anxo et al., 2007), leave uptake in Belgium is a strongly gendered phenomenon. Second, among one-child couples, the probabilities of female, male, or dual parental leave use are similar for the observed data and for the hypothetical simulation sample; whereas the probability of female leave uptake increases further for higher parity couples. These findings indicate that even in the extreme situation of all couples switching their pre-birth employment characteristics, the predicted gendered character of leave uptake remains relatively similar to that of the observed data; in other words, a wide gender gap in leave use. Hence, our hypothesis is confirmed that female partners on the aggregate level are more likely to use parental leave, beyond the distribution of labour market positions by gender (Hypothesis 2B).

Figure 2

Predicted parental employment and leave uptake strategies based on the observed data versus out-of-sample predictions on manipulated data, Belgium, 2001–2010

Predicted parental employment and leave uptake strategies based on the observed data versus out-of-sample predictions on manipulated data, Belgium, 2001–2010

4 – Robustness checks

28Three robustness checks were performed. First, regarding the extent to which women who anticipate childbearing will take jobs with particular characteristics 1 year before birth, there may be bias in the associations between these characteristics and parental employment strategies. However, descriptive analyses (not shown) indicate that relative pre-birth employment characteristics are relatively stable in the 2 years before birth, and analyses using 2 years before birth as the reference point yield similar results. Second, as the couple-quarters in our analyses are not independent, we included a random effect in additional models to control for clustering at the couple level. These models exhibit similar patterns of effects and substantive conclusions, but also more extreme odds ratios when controlling for clustering. As a result, simple multinomial logit models are presented in this article. Third, the analyses presented in this article do not include self-employment because the pre-birth employment characteristics are not applicable or available for this type of employment. Therefore, person-quarters exhibiting self-employment are also not considered in any of the parental employment strategies. Similar results are found by robustness checks that include quarters where one or both partners are self-employed in the analyses (4,076 person-quarters or a 2.9% increase in the sample).