Do economic resources play a role in bargaining child care in couples? Parental investment in cases of matching and mismatching gender ideologies in Germany

Family formation and family life have undergone pronounced changes in the late twentieth and early twenty-first century. Educational expansion has led to a rise in educationally homogamous couples and dual-earner families (Blossfeld and Timm 2003; Domański and Przybysz 2007; Kollmeyer 2012). Couples thus more often attain similar levels of labour market resources and postpone parenthood while in education (Balbo et al. 2013). Mothers have increased their labour market attachment (Cipollone et al. 2014) and pursue occupational stability (Aisenbrey et al. 2009; Abendroth et al. 2014). At the same time, institutional shifts in gender culture and family policies foster fathers’ more active engagement in childcare and more equitable divisions of unpaid work within families (Bünning 2015; Edlund and Öun 2016). Yet, domestic work divisions in families remain gendered, and we know that the polarization towards gendered work distributions appears to be ignited by the birth of the first child (Bianchi et al., 2012; Grunow et al. 2012). Studying childcare work, how couples organize it, and which micro-and macro-level factors contribute to persistent gendered childcare divisions in families is therefore central to advancing our understanding of gendered work–family divisions today (Bianchi et al., 2012; Grunow et al. 2012). However, childcare divisions, in contrast to housework divisions, have received surprisingly little attention in this debate to date.

Raising children remains a time and resource intensive endeavour, leading to tensions between parents’ gainful employment and family life (McGinnity 2014). These tensions are acerbated by the apparent rise in cost- and time-intensive approaches to parenting, especially among highly educated parents (Wall 2010). Families thus need to carefully weigh the potential costs and benefits of the time mothers and fathers spend caring for their children at the expense of gainful employment and vice versa. From this point of view, it seems likely and reasonable that the partners’ relative labour market resources; i.e. earnings and education, and thus opportunity costs of own childcare time matter, and that parents bargain for the division of childcare and paid work. Alternatively, parents could choose work-care schemas based on own and joint gender ideologies as a decision criterion over relative economic resources and family work divisions. In this latter case, the couples’ ideological pairings would be central predictors of childcare divisions and potentially enhance or suppress economic bargaining among couples. In particular, childcare divisions and the impact of relative resources should vary between couples in which both partners (1) are supportive of maternal employment (indicating egalitarian gender ideologies), (2) are unsupportive of maternal employment (indicating traditional ideologies or ideologies of intensive mothering) and (3) have mismatching ideologies concerning maternal employment. However, this association has rarely been tested empirically and the studies that have done so found mixed evidence (Kimmel and Connelly 2007; Bloemen and Stancanelli 2014).

We suggest that the lack of empirical evidence and the inconsistency of previous findings may partly be due to a lack of adequate data; a problem we seek to overcome in this study, by using the Panel Analysis of Intimate Relationships and Family Dynamics (pairfam). First, most previous studies that addressed parental investment and bargaining relied on cross-sectional data, thus measuring partners’ individual labour market resources (i.e. income or earnings) and childcare divisions at the same point in time. Partners’ resources have frequently been shown to be endogenous though, as relative income may affect divisions of unpaid work and vice versa (for a summary see Bloemen and Stancanelli 2014; Carlson and Lynch 2017). The causal interrelatedness between the dependent and independent variable studied is thus clouded by endogeneity (also referred to as reverse causality), implying that longitudinal measures of parents’ income and time allocation need to be applied (and the former lagged) to account for endogeneity. Second, a parent’s allocation of time to paid work and childcare is linked to the time allocation of relevant others; most importantly the partner’s (Elder 1994; Lundberg and Pollak 2008). Couple data are thus indispensable to correctly assess partners’ simultaneous decision-making to avoid biased estimates. Third, decision-making is considered to be based on partners’ well-defined preferences (e.g. preferences related to maternal employment) in the economic bargaining framework (Lundberg and Pollak 1996). Because preferences and gender ideologies are interlinked (cp. Eastwick et al. 2006), sociological studies of unpaid work commonly refer to either gender ideologies (Evertsson 2014) or preferences (Bonke and Esping-Andersen 2011) when assessing the bargaining framework. We use the sociological concept of gender ideologies in this paper (Davis and Greenstein 2009). While few studies employ direct measures of gender ideologies (Evertsson 2014; Nitsche and Grunow 2016), educational level is often used as a proxy. Education is however also associated with labour market resources and thus potentially conflates them with gender ideologies. We thus need direct measures of both aspects, gender ideologies and economic resources, to separate these dimensions, and ideally for both partners.

Our contribution to the literature lies in addressing all three aspects by using longitudinal couple data and separate measures of labour market resources and gender ideologies to assess potential bargaining in partners’ childcare divisions in Germany. The German case is particularly interesting for this type of research, as Germany has undergone rapid family policy changes over the last decade, thereby transforming the German institutional system from supporting a male breadwinner and female homemaker family model to supporting dual earning and caring among parents. From a European perspective, the German policy shift is interesting because it is in line with the European Commission’s political objective of fostering women’s and men’s equal economic independence, regardless of the life and family models they choose (European Commission 2016). It is thus likely that more European countries will follow the German turn in family policy and strengthen support for dual-earner/dual-carer families.

In this paper, we measure both partners’ ideologies concerning maternal employment, and whether the individual ideologies are matched or not, as representing one particular dimension of gender ideologies in a multidimensional gender ideology framework (Grunow et al. 2018; Knight and Brinton 2017). We use the term gender ideologies as suggested by Davis and Greenstein (2009) to refer to the attitudes captured in surveys and representing ‘individuals’ levels of support for a division of paid work and family responsibilities that is based on this notion of separate spheres’ (Davis and Greenstein 2009: 87). While our main indicator of gender ideologies is partners’ support of maternal employment for pre-school children, we present evidence of the robustness of this indicator using two additional indicators (and dimensions) of gender ideologies. Mismatching gender ideologies have been found to impact marital quality and household divisions of labour (Ogolsky et al. 2014; Oláh and Gähler 2014), and are thus assessed in this paper. Labour market resources are measured in terms of own absolute and relative income. We attempt to answer the following research questions: Is there a significant association between partners’ ideological pairings and the gender division of childcare among them? Do we also find a significant effect for partners’ relative labour market resources and the gender division of childcare among them? Is there a differential effect of labour market resources on the division of childcare contingent on the partners’ agreement on and support of maternal employment?

While mothers continue to perform much of childcare throughout the late twentieth and early twenty-first century, fathers have nevertheless increased their time with children, concomitant with the massive rise in maternal employment (Hook 2006; Sayer 2010; Goldscheider et al. 2014). It is yet unclear how strong these associations between the work and family spheres are, as findings remain ambivalent and cross-national variation high. For instance, the studies by Gracia (2014), Gracia and Esping-Andersen (2015) and Raley and colleagues (2012) point to strong positive associations between female employment and paternal childcare time while other studies suggest weak associations. In particular, Craig and Mullan (2011) report that mothers’ and fathers’ work arrangements relate modestly to parental shares of childcare, and that the association furthermore varies by educational level and across European countries (see also Kitterød and Pettersen 2006). Differences by country and educational level have also been found in the comparative study by Esping-Andersen (2009) on the ‘incomplete revolution’. The study suggests that the participation of women in the labour market is driven by the gender roles adopted in the domestic sphere and stratified in most European countries, with highly educated women displaying a stronger labour market attachment than low-educated women.

On the level of couples and parents surveyed individually, research considering the impact of economic and cultural factors on childcare simultaneously is scarce, both for Germany and other national contexts. A review over roughly two decades of research by Davis and Greenstein (2009) suggests that especially father’s gender ideologies are associated with paternal childcare. More recently, research from Sweden suggests that couples display more gender-equal divisions of childcare when the female and/or the male partner hold strong egalitarian gender ideologies (Evertsson 2014). Effects of couples’ educational parings have been found, suggesting that highly educated couples invest more time in their children than low-educated couples (Bonke and Esping-Andersen 2011). Despite issues of endogeneity (or reverse causality) between resources, ideologies and care work, longitudinal studies on the association between gender ideologies, relative resources and parental divisions of childcare are lacking (Bonke and Esping-Andersen 2011; Kühhirt 2012).

On the societal level, economic and cultural factors also appear to matter for couples’ divisions of childcare. In cross-national perspective, fathers spend less time with their children in countries with prevailing traditional gender roles (Raley et al. 2012). In line with this research, fathers’ relative contribution to childcare is more equal to the mothers’ contribution in societies that offer extensive public day care (Craig and Mullan 2011; Gracia and Esping-Andersen 2015). In most western societies, both mothers and fathers have increased the time they spend with their children, suggesting rising cultural standards of parenting and a spread of time-intensive caring practices despite dual earning (Sayer 2010; Wall 2010).

In Germany, the trend towards rising parental time for children has been documented as well (Grunow 2007; Neilson and Stanfors 2014). Still, parenthood predominantly leads to a long-term increase in women’s (not men’s) absolute and relative time for childcare and housework (Kühhirt 2012; Dechant et al. 2014). In line with these shifts in unpaid (care) work, paid work shifts in the opposite direction, strengthening paternal earner roles and maternal carer roles. Germany is thus marked by rather large gaps between women’s labour force participation and maternal employment.

This pattern corresponds with the German institutional system (see also Lueck and Ruckdeschel 2018). Germany has a joint taxation model, resulting in higher marginal tax rates for secondary earners and thus lower incentives for mothers to work compared with individual taxation models (Dingeldey 2001). German parental leaves are among the longest in Europe, with a maximum duration of 34.2 months of job protection and 12.2 months of payment (Keck and Saraceno 2012). While maternal employment had been a norm and practice in the eastern part of Germany pre-reunification, the western part adapted slowly to the idea that mothers of small children should continue working. Ten years ago (in 2007), the government introduced two so-called daddy moths (Schober 2014). This reform was associated with increased childcare time of fathers, especially during the child’s first year, while mothers tended to return to work sooner than before (Schober 2014). In addition, publicly subsidized childcare provision for children below age three increased markedly as a result of two federal laws (enacted in 2005 and 2008) and since 2013 children aged one year have a legal right to a day-care slot (Schober 2014).

In line with the historical division of Germany, notable and persistent differences concerning gendered divisions of paid and unpaid work have been found between the eastern and western part. Both spheres are being more equally balanced among couples living in the East (Cooke 2007; Hofäcker et al. 2013). Available evidence points to the impact of structural factors (i.e. remaining differences in regional childcare and unemployment rates) and to persistent cultural differences between eastern and western German mothers (Drasch 2012). Research further suggests that unequal divisions of childcare in the western part of Germany decrease as children reach the primary school age, reflecting the shift from predominant maternal home care for pre-school-aged children to universal institutionalized care for school-aged children (school-entry age in Germany is 6–7 years; Hofäcker et al. 2013), rather than fathers’ increased care time for older children (Cooke 2007; Neilson and Stanfors 2014).

On the couple level, partners’ relative resources have been found to play a minor role for the shift towards more unequal divisions of childcare and housework during family formation in Germany. Even for couples in which the female partner earned more than the male partner, parenthood frequently resulted in more traditional divisions of unpaid work, suggesting that bargaining power cannot make up for couples’ gender ideologies (Kühhirt 2012). Previous longitudinal research investigating this association for Germany, however, looked at childcare and housework as one joint dependent variable and did not apply direct controls for gender ideology. In the present study, we are able to overcome these shortcomings and thus contribute new evidence to the question of childcare divisions and the mechanisms that might be driving them.

Research on couples’ specialization in household production usually refers to two main strands of theory, economic and sociological, and several associated theoretical mechanisms (recently reviewed in Evertsson 2014). Within the theoretical setup of both, economic and sociological theories, it has been argued that housework and childcare have different analytical implications, even though gendered contributions to both spheres might be linked (Bianchi and Raley 2005; Bonke and Esping-Andersen 2011; Evertsson 2014).

Economic resource bargaining theory assumes that partners are self-interested and that the partner with the greater earnings potential can refrain from performing unpaid work (which is considered to be unpleasant), due to his/her greater bargaining power within the couple. According to bargaining models, the division of unpaid work is repeatedly renegotiated. Bargaining models consider housework to be unpleasant. Childcare is arguably demanding but also considered desirable and a display of love and care (Ferree 1990; Bonke and Esping-Andersen 2011). However, spending time caring for children creates opportunity costs for missed time and energy invested in paid employment and missed development of job-related skills (Bonke and Esping-Andersen 2011). Thus, according to resource bargaining, it is reasonable to expect partners’ relative labour market resources to influence parental divisions of time for childcare.

Sociological theory has argued that gender ideology is a relevant driver of human behaviour (West and Zimmerman 1987; Greenstein 1996; Krüger and Levy 2001). Krüger and Levy suggest that even though egalitarian ideologies are spreading, individuals possess an inherent ‘core’-ideology, the so-called gendered ‘master status’ (Krüger and Levy 2001), which may be more traditional than acquired outer layers of egalitarian ideology. Important life events may trigger an activation of this ‘core-ideology’, driving gendered behaviour. Others have discussed more explicitly how gendered ideology may affect work divisions (Greenstein 1996). According to Greenstein, gender ideology directly relates to the division of unpaid work among couples. Moreover, he finds that partners’ ideologies interact in predicting housework divisions; his egalitarian ideology has an increased effect when she is egalitarian as well (p. 591). Through the lens of the doing gender concept, gendered behaviour is understood as a situated achievement (West and Zimmerman 1987). Gendered behaviour in the sphere of unpaid labour is thus considered the product of gendered displays, with both men and women adjusting their behaviour to what they believe is expected of them according to prevailing normative ideas. Hence, individuals ‘do gender’ by engaging in activities prescribed as gender appropriate, for example through an increased engagement in childcare and housework for women. According to the doing gender perspective, gender displays become especially meaningful when individuals behave in gender-atypical ways otherwise, as an act of compensation. For instance, women who out-earn their male partner in the labour market may ramp up their engagement in domains prescribed as female to compensate (Brines 1994).

In addition, one might expect some type of interaction between resources and gender ideologies, as gender ideologies may impact market productivities over time and ideological support for maternal employment (or for maternal home care) may vary both within and between couples. This aspect of gender ideology is of special relevance in the German context, where full-time maternal homecare of pre-school children has long been the norm and institutional support for the employment of mothers of children below the age of three has only recently increased. As the time spent with children is limited by other responsibilities and against the background that gendered patterns of care persist despite mothers’ employment, it is important to assess the directionality of such potential interaction effects. Hypotheses on how these may look like in light of the theoretical concepts are discussed below.

In line with our theoretical framework, we first address the questions of whether partners’ relative resources (H1) and gender ideologies (H2) may affect their division of childcare when both predictors are modelled simultaneously but not (yet) interacted with each other. Following resource bargaining theory, we would expect a positive effect of an increase in her (lagged) share of the household income on his share of childcare time and vice versa (Hypothesis 1). Following the sociological theories, ideological pairings will influence the gendered division of childcare. His share of childcare will be largest among couples in which both hold egalitarian ideologies, and lowest among couples in which both hold traditional ideologies (Hypothesis 2).

Second, we investigate whether there is a differential effect of partners’ relative resources contingent on the ideological pairing of the couples. For couples with mismatching ideologies, bargaining theory allows to derive specific expectations, while no clear expectations can be derived for couples with matching ideologies. For the sociological gender ideology concepts, the reverse is true. Clear hypotheses arise for couples with matching ideologies, but not for those who are mismatched.

According to bargaining theory, we expect a differential effect of relative income on childcare divisions among couples with mismatching ideologies but not (necessarily) among couples with matching ideologies. Under the condition of mismatching ideologies, incentives may be higher to use own resources to negotiate (or promote) ones’ preferred division of childcare. We would thus expect a positive effect of her relative resources on his contribution to childcare in cases where she has more positive attitudes towards maternal employment than him (Hypothesis 3a). Conversely, we would expect a negative effect of her relative resources on his contribution to childcare in cases where she has less positive attitudes towards maternal employment than him (Hypothesis 3b). We do not expect bargaining to take place when partners have matching ideologies, in this case, they might opt for divisions of paid work and care that are most in line with their desired work divisions for the family unit.

According to sociological gender ideology concepts and the doing gender approach, we expect a differential effect of relative income on childcare divisions, particularly among couples with matching ideologies. The doing gender reasoning implies that compensation in the private sphere takes place specifically among couples in which she out-earns him (Brines 1994). We test a simplified version of this, and hypothesize that her share of childcare increases as her income increases among couples in which both are unsupportive of maternal employment (Hypothesis 4a). While compensation behaviour may also take place among couples with conflicting ideologies, the predictions are less clear, and we refrain from formulating specific hypotheses for these.

According to the ‘master-status’ theory, the remaining group of couples, namely couples with two egalitarian partners, may be special (Krüger and Levy 2001). We argue that such couples, in which both partners maintain egalitarian ideologies after the life-altering transition to parenthood, may represent an ‘egalitarian island’, relatively immune to traditional outside gendered expectations. Hence, they are ‘more free’ to teamwork fair solutions to both partners, in terms of time and energy investments in paid and unpaid work. According to this argument, if her share of income increases, his share of childcare would increase, and vice versa (Hypothesis 4b).

We draw on data from the release 7.0 of the Panel Study of Intimate Relationship and Family Dynamics (pairfam), a panel study from Germany launched in 2008 (Brüderl et al. 2016). A detailed description of the study can be found in Huinink et al. (2011). Waves have been collected yearly, our analyses are based on all currently available waves (waves 1–7). The original sample features 12,402 individuals, referred to as anchor persons because their partners and family enter the study only through association with this anchor. Anchor persons were randomly selected from the population in Germany, and stem from the birth cohorts 1991–1993, 1981–1983 and 1971–1973. Partners of anchor persons were also interviewed, but with a rather low response rate. In wave 7, approximately 50% of partners have filled in the questionnaire. Pairfam has also been subjected to attrition. In wave 7, about 48% of the original sample of anchors could be re-interviewed. Despite these drawbacks, pairfam is a rich and unique data source. It includes relationship, fertility and educational histories, and features a large set of measurements on family life and relationship dynamics. An overview of sample sizes, attrition and response rates for anchor and partners in each wave can be found in the pairfam documentation reports, which are all available online (Fuss et al. 2016, link provided in the references).

Our analytic unit is the couple. We select all couples living with children, and condition further on couples for whom both interviews (anchor and partner) are available. Measures on partners’ characteristics such as education and income are collected via the anchor person. Ideology items of partners are available through the partner interview only. We exclude same-sex couples due to very low case numbers of these types of couples in the pairfam; they would require a distinct conceptual framing and analysis as gendered dynamics in same-sex couples are likely distinct. After excluding same-sex couples and after list wise deletion due to missing values, our analytic sample contains 1660 couples (4369 observations). Table 1 depicts descriptive statistics of the analytic sample, at the first observed wave for each couple. On average, couples have 1.5 children, and the youngest child is about 4 years old. Anchors report that she spends more time with childcare than him (with an average of 3.658), and she contributes about 33% to the household income. Couples with both partners supportive of maternal employment are most common (35.2%), followed by dually unsupportive couples (29.8%). About one fifth disagree about their attitudes on maternal employment with her being supportive but not him, while 9.6% of couples disagree with her being less supportive than him.

The dependent process of interest is the division of childcare between mothers and fathers among co-residential couples with children. The items measuring the division of time spent with caring for (own) children have been collected in every wave from both partners. Either partner has been asked whether caring for children is taken care of (1) (almost) completely by their partner, (2) mostly by the partner, (3) split about 50/50, (4) mostly by the respondent or (5) (almost) completely by the respondent. For our analyses, we use the responses provided by anchors, and have recoded this variable such that 1 stands for him taking care (almost) completely of the children and 5 her taking care (almost) completely of the children. We control for the sex of the anchor in the models, to account for the well-known fact that both men and women usually over-report their share of unpaid work (Bonke 2005). Couples have been observed for up to seven waves. Some were already living together in the first wave, other relationships started during the observation time. Since we lose the first observation due to lagging of the covariates, the observation time used in the models ranges from one to six waves per couple (the average is about 2.6).

The pairfam contains four questions addressing gender ideologies, which are referring to the domains of work and family. These are:

1. Val1i3: Women should be more concerned about their family than about their career

2. Val1i4: Men should participate in housework to the same extent as women

3. Val1i5: A child aged under 6 will suffer from having a working mother

4. Val1i6: Children often suffer because their fathers spend too much time at work

The Cronbach’s alpha for these items does not result in acceptable values to create a composite index (alpha <.3), neither for all four nor for subgroups of these items. We believe this is because these items address distinct dimensions of gender ideologies and that attitudes concerning housework are not merely mirroring attitudes towards paid work or care (cp. Grunow and Baur 2014; Grunow et al. 2018; Knight and Brinton 2017). Due to low alphas and the separate dimensions of gender ideologies these items address, we decided to refrain from building an index. The items Val1i5 (‘A child aged under 6 will suffer from having a working mother’) measures ideology towards childcare and working mothers, and directly relates to our research question (Maternal Employment Ideology). Therefore, we chose this single item as the main ideology predictor for our analysis. We thus operationalize gender ideology as ideology towards childcare and working mothers, using the Val1i5 indicator as the main predictor in our analyses.

To test the validity of this operationalization, we use two alternative ideology items available in the dataset for robustness checks, namely Val1i3 (Work–Family Roles Ideology) and Val1i4 (Housework Division Ideology). These checks are presented in the appendix (Table A1, Figures A1 and A2): Here, we calculate the analogue models to Table 2, using the items Val1i3 (model 1 and 2) and using the item Val1i4 (models 3 and 4) as alternative operationalizations for gender ideology, instead of item Val1i5.

While responses to the items Val1i3–Val1i5 are straightforward in terms of representing an egalitarian or traditional response, Val1i6 is hard to interpret as both egalitarian and traditional parents might agree with the statement ‘Children often suffer because their fathers spend too much time at work’. In particular, it is unclear whether ‘too much time at work’ is understood by respondents in terms of long overtime hours, implying that fathers in a traditional breadwinner family hardly see their child during the week, or whether ‘too much time at work’ is understood in terms of working full-time rather than part-time, as would be an option in egalitarian dual-earner families. For this reason, this variable is not included in our robustness checks.

The gender ideology item we use measures support for maternal employment when children are young: ‘A child aged under 6 will suffer from having a working mother’. Agreement to this statement from anchor persons and partners alike was measured on a 5-point scale, ranging from ‘disagree completely’ to ‘agree completely’. We have used his and her responses to create five ideological pairings: (a) both partners are supportive of maternal employment (either partner responded 1 or 2), (b) both partners are moderate (either partner responded 3, or one partner responded 3 the other one 2), (c) both partners are unsupportive of maternal employment (either partner responded 4 or 5, we also include combinations of 3 and 4 here), (d) mismatch: she is supportive he unsupportive (partners are two steps or more away from each other) and (e) mismatch: she is unsupportive, he supportive (partners are two steps or more away from each other). This item is time constant, we use the first available measurement, to avoid a possible feedback loop between resources and later measured ideologies over the course of the panel. The two other ideology items for the robustness checks are coded in the exact same fashion. Please note that the ideological pairings are relatively evenly distributed across the sample for the ‘Maternal Employment’ item and the ‘Work–Family Roles’ item (see Table 1). However, with respect to ‘Housework Division Ideology’, the large majority of couples consist of two partners in favour of an equal division. Less than one percent of couples fall into the group of two unsupportive partners of equal housework divisions; hence, the model results for this item and category are estimated on a small cell size. We emphasize that this estimation serves as a robustness check of our main results and cannot stand alone.

Income resources are measured in terms of (a) her absolute monthly amount of net income (logged), and (b) the share her income provides to the monthly household income (from 0% to 100%). Since our theoretical focus lies on investigating the differential effect of relative income by gender ideology, the educational measures serve as control variables. We estimate models with and without controlling for education, in order to ascertain that the income effects do not hinge upon educational background. His and her education is measured in four groups: (a) currently enrolled, (b) low education, (c) medium education and (d) high education (tertiary college or university). A dummy variable expresses whether her educational attainment is higher than his (couples with one or two partners enrolled in education are always coded 0 on this variable). Income and education measures are time-varying, and the income variables are lagged by one wave.

All models control for the number of children, the age of the youngest child, the age of the mother, marital status (married or cohabitating), whether the couple lives in West or East Germany, and the sex of the responding anchor person. All control variables are time-varying (except sex of anchor person, as this doesn’t change over time), meaning that they represent the current ages and statuses in each wave, and (can) change over panel waves.

Our main findings are presented in Table 2 and Figure 1. In reading the results, positive coefficients indicate a change towards a more traditional division of childcare (increasing her share); negative coefficients point to an increase in his share. Model one shows our first key finding. Relative resources are significantly associated with the division of childcare among couples, net of gender ideology and control variables. While her absolute monthly net income is not significantly related to the division of childcare trajectories, relative income significantly predicts changes in the division of childcare. An increase in her contribution to the household income is associated with him spending more time with caring for the children in the next wave. The coefficient indicates that a 10% change is associated with a shift of 0.0199 in the dependent variable, which is a very small change of about 2% of a unit. The magnitude of the coefficient is −.199, which indicates the change in the dependent variable from 0 to 1, signifying a 100% change in income (from 0% to100%), as the variable is coded in decimals from 0 to 1. The magnitude of this shift is modest, yet this predictor is highly significant and remains so throughout all tested specifications. His education also significantly predicts the division of childcare. The more highly educated he is, the more traditional the division of childcare becomes, net of gender ideology and other controls. While we control for income and income share, work statuses are not reflected in the models. Hence, this education effect may partly reflect paid employment divisions. Additionally, ideological pairings are significantly associated with the division of childcare: As expected, couples with two partners unsupportive of maternal employment display the most traditional division of childcare, dually supportive couples have the most egalitarian division, and dually moderate couples are situated in the middle. The difference between moderate and unsupportive couples is also significant when tested (χ² = 4.2, p. = .034). While all types of couples except those mismatched with a supportive woman and an unsupportive man differ significantly from couples with two partners supportive of maternal employment, not all groups differ significantly when tested against each other. Notably, couples in which she is supportive of maternal employment but he is not do not differ statistically form dually supportive couples (reference group), while couples with an unsupportive woman and a supportive man do not differ significantly from dually unsupportive couples when tested against each other (χ² = 2.44, p = .119). This suggests that it is mainly her attitude towards maternal employment which may be driving how the couples share caring for children. However, an alternative model (not shown) modelling his and her ideology separately indicates that his attitude towards maternal employment is marginally significant (p = .065) in predicting the division of childcare, but only after controlling for resources.

Significant interaction effects between relative income and the ideological pairings are present in model two. Thus, the association between relative income and childcare divisions is contingent upon the ideological pairing of the couple, as hypothesized. The interactions between the indicator for a hypogamous couple (she has more education) and the ideology variable are largely insignificant, further underscoring the relevance of the relative income/ideology interaction. Figure 1 illustrates our finding visually, displaying marginal effects based on the coefficients in Table 2. Among couples who are dually supportive of maternal employment, a 10% increase in her contribution to the household income is associated with an increase in his participation in childcare in the next wave. The same applies to couples with conflicting attitudes when she is the partner supportive of maternal employment. If both fall into the moderate group, there will also be an increase in his share with her increasing income, but much more modestly in size as compared to the two groups with supportive women (the difference between both supportive/both moderate is significant, p. = .009). In contrast, small negative effects of relative income on the division of childcare are present among the two pairings including a woman unsupportive of maternal employment. For both, her share of childcare increases with her share of income. The difference in the association between income share and childcare divisions between the two pairings with unsupportive women versus the other pairings is statistically significant, as can be seen in the test values listed underneath Figure 1.

The control variables show the signs that are expected based on our literature review. The more children there are in a household, the more couples’ divisions of childcare tend to shift towards the mother. With rising age of the youngest child, childcare divisions become more equal again. Couples in the eastern part of Germany report more equal divisions of labour than do their western German peers. Male anchors report more equitable divisions of childcare than do female anchors. Controls for the female partners’ age and union status are not statistically significant.

Table A1 and Figures A1 and A2 present the results from our robustness checks. The presentation of the models and figures is analogue to the main results from Table 2 and Figure 1. Models 1 and 3 in Table A1 show main effects without interactions for the two additional ideology items. They further clearly confirm our previous results. An increase in her income share, controlling for these other gender ideology items, is further significantly associated with an increase in his time spent with childcare. Additionally, dual-egalitarian couples (the reference category) with regards to both housework and work–family role ideology have the most egalitarian share of childcare. In model 1 (housework ideology), the group of dual-traditional couples is very small (as shown in the descriptives in Table 1), leading to an insignificant p value on the coefficient. However, coefficient sizes make clear that the share of childcare is more gender traditional in all the other ideological pairings, compared to ‘egalitarian island’ couples. Graphs A and B illustrate the interaction effects between relative income and ideological pairings from models 2 and 4 in the Table A1. These confirm our main findings: an increase in her relative income is strongly associated with an increase in his childcare contribution among ‘egalitarian island’ couples, while this is not the case among the other types of couples.

Longitudinal evidence on the factors potentially driving how partners in heterosexual couples divide childcare is still scant, especially with respect to investigating the role of partners’ own and relative economic resources. In times of changing gender regimes in educational systems, labour markets and family lives, however, this research gap deserves serious attention. Addressing it will help us understand more deeply how and why couples share childcare the way they do, which may in turn have important implications for both men’s and women’s engagement in families vs. paid employment and, in extension, fertility. In addition, we may gain insights into the permeability of gendered behaviour in societies at large.

We extend the literature on childcare divisions using novel data and methods, focusing on examining whether (1) relative income predicts childcare divisions (while controlling for couples’ gender ideology pairings), (2) whether couples’ gender ideology pairings predict childcare divisions (while controlling for partners’ relative income) and (3) whether the effect of relative income may hinge upon the partners’ ideological pairings (and vice versa). We have argued that Germany is a particularly relevant case for testing our theoretical framework. First, the rapid family policy changes during the last decade might have triggered wider cultural change in the German society (Gangl and Ziefle 2015). This will perhaps make it easier for couples with egalitarian gender ideologies to maintain their egalitarian ideals and change traditional gendered master statuses in the future (Krüger and Levy 2001). Second, egalitarian ideologies have been spreading everywhere in Europe, suggesting that similar household dynamics as the ones under study here might occur in other European countries (Grunow et al. 2018; Knight and Brinton 2017). In this sense, the German policy shift might foreshadow cultural and political changes in Europe more widely.

We test six hypotheses. Our first hypothesis (H1) states that an increase in her (lagged) share of the household income will have a positive effect on his share of childcare, and vice versa. Our findings support this hypothesis. This finding is robust throughout specifications and while controlling for educational resources, thus supporting resource bargaining theory.

Our second hypothesis (H2) states that his share of childcare will be largest in couples in which both are supportive of female employment and lowest in couples in which both are unsupportive of female employment. This hypothesis is also supported, suggesting that partners’ ideology pairings are relevant for couples’ divisions of unpaid work as well.

In addition, we tested for potential interaction effects between our ideology measures and partners’ relative resources, arguing that resource bargaining will perhaps be more likely to occur among couples with mismatching gender ideologies while gender ideologies should be especially effective mediators of relative-income effects on childcare in couples with matching gender ideologies. In particular, we tested complementing hypotheses, expecting a positive effect of her relative resources on his contribution to childcare in cases where she has more positive attitudes towards maternal employment than him (H3a), and a negative effect of her relative income on his share of childcare in cases where she was more unsupportive of maternal employment than him (H3b). Our findings support hypothesis 3a but not hypotheses 3b. Couples in which she is more supportive of maternal employment than him apply the logic of relative resources in quite similar ways as do dually supportive couples, namely relating her increased income share to his increased childcare share. In the opposite case, in couples with a woman unsupportive and a man supportive of maternal employment, couples act much like dually unsupportive couples. This significant difference in the effect of relative income between the two mismatched ideological pairings is no longer present in our robustness check models, which test two alternative gender ideologies. Still, the coefficients point in the same direction in the models testing work–family gender ideology (and are virtually zero in the other model testing housework ideology). Our hypotheses 4a and 4b pertain to couples with matching ideologies towards maternal employment. We tested expectations on whether couples with matching unsupportive ideologies towards maternal employment compensate gender-‘atypical’ increases in her relative earnings by increasing her share of childcare time (H4a). In the same vein, we tested whether ‘egalitarian island’ couples in which both partners are supportive of maternal employment might be the ones most free to make work–family decisions based on their own convictions and thus increasing his share of childcare most when her share of income increased and vice versa (H4b). For ‘egalitarian-island’ couples supportive of maternal employment, we find a significant positive relative-income effect on his childcare share, while the reverse is true for couples unsupportive of maternal employment. The relative-income effect for the dually supportive ‘egalitarian island’ couples is strongest as compared to the effects for all other ideology pairings, and it is also robust. We conclude that economically efficient divisions of childcare depend upon both, the couples’ ideological pairing and on mothers’ own ideologies towards maternal employment.

Of course, individual and couples’ gender ideologies, gendered cultural contexts, gendered family–work arrangements, and educational, employment and income trajectories and the resulting relative-resource structures between partners are all interlinked as part of an endogenous system. In other words, relative income in couples could indeed be a bargaining chip, women (or/and men) potentially use to bargain childcare divisions with their partner. But our theoretical framework and analyses suggest that an alternative pathway is likewise plausible, namely that relative income is a by-product of gendered work–family life arrangements partners have planned and created jointly, based on their underlying gender ideology structures. Our findings even suggest that perhaps a strong united couple ideological conviction, rooted in the egalitarian beliefs of both partners, is necessary to break open classical gender arrangements – that only such ‘egalitarian-island’ couples may be outfitted appropriately to make arrangements work over the long haul which may be at least partly contradictive to traditional work–family arrangements. The hypotheses developed and tested in this paper merely address two possible pathways of how relative resources, gender ideologies and childcare divisions may be associated, without being able to truly assess causal directionality of ideology and relative resources or the significant interactive differences between these two elements. Yet, our results are insightful and have important implications for future research. Our stylized examples (hypothesizing selective ideology arrangements versus bargaining power) speak to the economic theories, including classical bargaining theory (Lundberg and Pollak 1996) and New Home Economics (Becker 1981), and also enhance our understanding of the sociological theories discussed in this paper (West and Zimmerman 1987; Greenstein 1996; Krüger and Levy 2001). However, in reality, more and more complex mechanisms may be at work. Our findings enhance our understanding of these processes by illustrating that combinations of these theoretical approaches, for example scenarios in which bargaining power may be used differentially based on ideological pairings and on whether the partners have worked out a specific work–family combination plan or not, appear plausible. Future research, therefore, should address these deeper mechanisms, empirically as well as, perhaps even more importantly, theoretically. The era of dual breadwinner/dual caretaker families is still relatively young, and their realities are still evolving and changing in high speed as social scientists are analysing them. Future questions to be addressed would include the following: Are egalitarian couples ‘better’ at strategical joint planning? Are traditional couples less apt to do the same, as single breadwinner wages rarely support a family today? Are women engaging in ‘gatekeeping’ behaviours in the family, contingent on their own (traditional) beliefs, perhaps to their own detriment? And what would be the role of national cultural and family policy context in this latter case? Our findings have inspired these questions and invite future research to explore the emerging and changing mechanisms driving family life further.

This paper uses data from the German Family Panel pairfam, coordinated by Josef Brüderl, Karsten Hank, Johannes Huinink, Bernhard Nauck, Franz Neyer, and Sabine Walper. Pairfam is funded as long-term project by the German Research Foundation (DFG).

No potential conflict of interest was reported by the authors.

Natalie Nitsche is a research scientist at the Wittgenstein Centre (IIASA, VID/ÖAW, WU). Her research interests include family formation, fertility, and social inequalities. Much of her work focuses on couples and investigates how partners' joint resources and preferences relate to their childbearing behavior and work-divisions. Her research has been funded by the National Science Foundation (NSF), the European Commission, and the Austrian Science Fund (FWF).

Daniela Grunow is a Professor of Sociology at the Goethe University Frankfurt, Germany. Her research focuses on the interaction of market work and gender relations in different welfare regimes. She is co-editor of the Kölner Zeitschrift für Soziologie und Sozialpsychologie and holds a Goethe-Fellowship, which promotes research-excellence in the humanities and social sciences at the Forschungskolleg Humanwissenschaften in Bad Homburg.