1 Introduction

The aim of this article is to analyze the relationship between the characteristics of parental leave and paternal childcare time. To that end, we exploit micro-level time-use data from the Multinational Time Use Study (MTUS, Gershuny and Fisher 2010) from eight industrialized countries from 1971 to 2005, and merge them with macro-level parental leave characteristics collected from a variety of sources. During the last decades, economists and other scientists have identified numerous effects of paternal engagement in childcare. Among these are children’s well-being (Bonke and Greve 2012; Carlson and McLanahan 2004; Palkovitz 2002), a gain in social competences and work-life balance for fathers (see overview in Hook 2006), higher fertility (Buber 2003; de Laat and Sevilla-Sanz 2011; Duvander and Andersson 2006; Lappegård 2010; Oláh 2003), and increased marital stability and satisfaction (Greenstein 1995; McHale and Crouter 1992; Oláh 2001; Sanchez and Gager 2000; Sigle-Rushton 2010; Wengler et al. 2008). Moreover, fathers’ childcare may enable their partners to further engage in paid work, mitigating income loss due to family duties (Boll 2011; Light and Ureta 1995; Mincer and Polachek 1974). To date, there has been great cross-national variation in the time fathers spend with children (Gimenez-Nadal and Sevilla 2012; Stancanelli 2003). All European and most other industrialized countries have established parental leave for fathers, but these policies vary considerably between countries, providing different incentives and disincentives for men to assume childcare tasks (Sullivan et al. 2009). Questions thus arise as to whether children benefit from parental leave policies for fathers, in terms of their spending time together, and how these policies should best be constructed in order to promote the time investment of fathers in their children.

From a theoretical point of view, parental leave policies are expected to have a societal effect on paternal childcare through several mechanisms. Firstly, paid parental leave can stimulate fathers’ uptake of parental leave resulting in an extension of paternal time spent with the baby. According to micro-economic theories, the time allocation of a father depends, inter alia, on the opportunity costs of the foregone alternative (mainly paid work) and the relative resources of his partner (see overview in Reich 2011). Drago (2011) suggests that if fathers had the opportunity to take paid parental leave or to reduce working hours, they would devote about 70 % of the reduction in working time to childcare. Secondly, paternity leave may have long-lasting effects on a father’s time with his child. Existing empirical literature often finds positive effects of fathers who have taken leave after the birth of the child, in various childcare indicators (Haas and Hwang 2008; Nepomnyaschy and Waldfogel 2007; Schober 2012; Tanaka and Waldfogel 2007). Thirdly, positive within-family and out-of-family spill-over effects may evolve: within the family, fathers’ take-up of leave may also result in the higher childcare involvement of a father with his other children, thereby enhancing paternal childcare productivity (Chalasani 2007; Frenette 2011). Out-of-family positive effects are likely to arise because even fathers who do not use parental leave themselves may become more engaged with their children if such policies actively promote paternal engagement with children, thereby reducing the negative stigma of men performing childcare tasks (Akerlof and Kranton 2000; de Laat and Sevilla-Sanz 2011; Hook 2010; West and Zimmermann 1987).

The behavioral response of fathers to parental leave policies is likely to depend on their educational level. Sociological theories (Blossfeld and Drobnič 2001) as well as empirical results for fathers’ use of parental leave (see survey and results in Reich 2011) and their childcare time (Borra and Sevilla-Sanz 2011; Chalasani 2007; Guryan et al. 2008; Ishii-Kuntz and Coltrane 1992; Ramey and Ramey 2010) point to the strong involvement of highly educated fathers in childcare work compared to their lesser educated counterparts. For example, Borra and Sevilla-Sanz (2011) use data from twelve countries and find a childcare gradient of approximately between 10 and 20 min per day between highly and lowly educated mothers and of approximately 12 min for fathers.

We use multivariate models to estimate the correlation between several parental leave indicators and the time fathers spend with their children. The main independent variables are three key factors for the availability and use of parental leave by fathers: the duration of parental leave for both parents (family right), the duration of exclusive parental leave for fathers, and parental leave benefits. Our results show that the duration of parental leave available to both parents, and of exclusive parental leave for fathers, as well as a high benefit level (compared to none) are generally positively related to paternal childcare time. We also show that the positive correlations are dependent on fathers’ educational level, with the strongest positive correlation for highly educated fathers regarding exclusive parental leave for fathers and high benefit rates. In particular, our estimates indicate that an increase of one extra exclusive parental leave week for fathers increases a highly educated father’s time with his child by roughly 7 min per week. Receiving a high parental leave benefit rate compared to none is associated with an increase in all paternal childcare time by nearly 1 h per week, and approximately 1 h and 10 min per week for highly educated fathers.

The study at hand contributes to the empirical literature on the correlation between parental leave policies for fathers and paternal childcare time in directly addressing distinct parental leave characteristics in cash and kind instead of dealing with policy packages or indices. Few articles directly evaluate parental leave policies in terms of their effect on paternal childcare time, or account for all of the outlined mechanisms. Some literature analyzes the effect of parental leave after a policy reform in a given country. For example, Rege and Solli (2013) show that Norwegian fathers were more involved with their children after the introduction of four exclusive parental leave weeks for fathers. This article however does not offer an analysis with a cross-national perspective and is limited to the particular institutional setting of the reform, restricting the number of parental leave indicators that can be used to those that changed during the policy reform. Other papers have looked at parental leave policies in a cross-national setting. Bygren et al. (2011) show that an index consisting of fiscal and cash child benefits supporting the traditional male breadwinner model has a negative effect on the time fathers spend with children, and their index based on earner-carer policies has a positive impact on paternal childcare time. Smith (2001) and Smith and Williams (2007) show that the score in a father-friendly policy index is positively correlated with paternal childcare. In Hook’s (2006) analysis, weeks of parental leave are negatively correlated and parental leave available for men is positively correlated with the number of minutes per day fathers spend in unpaid work, including household chores and childcare. Consequently, no conclusion about whether children are direct beneficiaries of these policies could be made. In addition, Hook does not test for parental leave benefits and exclusive weeks for the father, two policy measures that are frequently assumed to have an effect on paternal behavior (OECD 2007).

We contribute to the international comparative analysis of parental leave policies by providing a more pronounced distinction of parental leave entitlements, distinguishing between transferable rights between the parents and individual rights for the father. Furthermore, we use interaction effects to find out whether the paternal educational level plays a role in the relationship between parental leave characteristics and childcare time. We also enlarge the temporal and geographical scope of the analysis, using surveys from the 1970s to 2005 from eight industrialized countries, and use very detailed information on fathers’ time spent with children through the use of diary data, whereby individuals recorded their activities in a 24-h-period. This information is more detailed and reliable than that used in most studies, which use rather broad categories of childcare time based on retrospective hours per week, as opposed to detailed diary data. It also allows us to distinguish childcare time from unpaid work, which is crucial, as incentives and rewards from childcare time differ from those of other housework chores (Buber 2003; Kimmel and Connelly 2007; Sundström and Duvander 2002).

From a broader perspective, our article contributes to the body of literature on the evaluation of parental leave policies. Fathers’ parental leave has, for example, been analyzed in terms of fertility outcomes (Duvander and Andersson 2006; Duvander et al. 2010; Lappegård 2010; Oláh 2003) as well as employment and wage effects (Cools et al. 2011; Johansson 2010; Rege and Solli 2013). We add insights into the question of whether children are the beneficiaries of such policies by considering the increase in the time their fathers spend with them.

The article is structured as follows. In Sect. 2, the institutional background of parental leave policies is discussed. In Sects. 3 and 4, the data, the variables as well as the econometric specification are presented. Section 5 provides the results and robustness checks, and Sect. 6 concludes.

2 Institutional background

Parental leave is defined as national legislation that allows parents to stay at home to care for their child subsequent to maternity and, if available, paternity leave.Footnote 1 During the last decades, all industrialized countries analyzed in this study introduced parental leave policies, but the motivation and pace of the introduction, as well as the accessibility for fathers, vary considerably. Political decisions on parental leave policies depend on fiscal, demographic and economic rationales as well as attitudes towards ‘traditional’ and ‘modern’ types of families (Morgan 2009). While national governments used to be the sole decision-making authorities with regard to parental leave, policymakers of EU member states have been admonished to follow the guidelines of the European Commission (EC) in recent decades. In order to support paternal engagement with children, the European Union legislation demands a minimum of 3 months parental leave for both parents and the European Commission states that “men should be encouraged to assume an equal share of family responsibilities, for example they should be encouraged to take parental leave” (UNICE et al. 1996) in the 1996 framework agreement on parental leave.

In Nordic countries (Sweden, Finland and Norway), which are classified by sociologists as social-democratic welfare states, the introduction of parental leave was driven by labor market and gender equality considerations. It was intended to give mothers the opportunity to continue participating in the labor market in the long term, ensure their income during the leave period, and reduce gender differences in paid labor and childcare work. Thus, the ability of fathers to take leave was introduced relatively early, i.e., in the 1970s and 1980s, and with distinct incentives (Neyer et al. 2006). In order to further promote paternal uptake of leave, exclusive parental leave weeks for fathers were introduced in the 1990s. In Sweden, 1 month of exclusive parental leave for fathers was introduced in 1995, and was expanded to 2 months in 2002 (Björklund 2006). In Finland, 3 weeks of parental leave exclusively for fathers were introduced in 1993. In Norway, four parental leave weeks became exclusively reserved for fathers as of 1993, and this entitlement was extended to 6 weeks in 2005 (Haataja and Mattila-Wiro 2006; Neyer et al. 2006; Rønsen 2004).Footnote 2 As parental leave benefit is related to the wages paid before the leave and hence the opportunity costs of staying at home, the system is comparatively attractive even for well-paid fathers (and mothers). In Sweden, the benefit amounted to between 70 and 90 %, in Norway between 80 and 100 % of the wage throughout the period under study (Björklund 2006; Gauthier and Bortnik 2001; Haataja and Mattila-Wiro 2006; Neyer et al. 2006; Rønsen 2004). In Finland, it equaled 45 % of the former wage until 1982. Since then it has varied between 66 and 80 %, with higher percentage rates for poorer parents and lower percentage rates for better-off parents (Haataja and Mattila-Wiro 2006; Neyer et al. 2006; Rønsen 2004). Compared to no benefit or a flat-rate payment, this system reduces the incentive gap between partners with different income levels.

The Nordic parental leave systems have been praised for their high wage compensation and the individual entitlements for fathers. Moreover, the relatively short and well-paid leave helps mothers to care for their children in the first months after birth, and then quickly return to the labor market. Consequently, a mother’s time constraints as well as her bargaining power arising from market work call for the further long-term involvement of fathers (or other carers than the mother) in child-raising duties. Nevertheless, the Nordic parental leave systems have been criticized for continuing to promote the gendered division between paid and unpaid labor. Firstly, as most mothers earn less than their husbands, the parental leave benefit has a gender bias as long as it is below 100 % of their former wage. Secondly, mothers tend to use their exclusive parental leave weeks plus most of the leave that can be shared, while fathers tend to use only paternity and parental leave that cannot be taken by mothers. This leads to specialization gains of the mother in childcare work (Datta Gupta et al. 2008; De Henau et al. 2007). Therefore, from an institutional point of view, it remains unclear whether these policies can impact fathers’ childcare time in the long run.

In Germany and Italy, two countries labeled as conservative welfare states, family policies were introduced based on the normative predefinition that mothers should be the primary caretakers of young children. As a consequence, parental leave was designed so that it was comparatively more accessible and attractive for mothers than for fathers. In Germany, from 1986 to 2006, the ‘child-raising leave’ was comparatively long (between eight and 36 months, with several amendments in this period) and the benefit rate was a flat-rate payment, which was means-tested and thus only paid to parents in need (Bundesgesetzblatt 1985; Juris Das Rechtsportal 2012; Kreyenfeld 2004; Merz 2004). Since 1991, fathers were eligible to use the leave, but due to the lengthy, ill-paid leave less than one to three per cent of the users were fathers between 1991 and 2006 (Federal Statistical Office 2009).Footnote 3 In Italy, parental leave was comparatively short (6 months) at first, but due to the lack of public childcare slots and their short opening hours for children younger than three, it was difficult for mothers to return to work earlier (Del Boca et al. 2005). Although parental leave could be taken by the father instead of the mother (before 1999), it was more economically viable for the secondary earner to take the leave, because the benefit amounted to 30 % of the former wage only, and thus the leave was mostly taken by mothers who were normally the secondary earners in a household. Following the guidelines of the EC directive on parental leave, this was extended to 6 months per parent in 1999, but their sum of leave months could not exceed 10 months. However, if fathers took at least 3 months consecutively, an additional month was granted to them (Clauwaert and Hager 2000; Columbia University 2011; Eurofund 2009a; Giovannini 2004, 2005).

The other countries in this study—the United Kingdom, the Netherlands and Canada—have been largely ‘liberal’ regarding family policies, following the view that the market would satisfy parental demand for solutions to the reconciliation of work and family life, and that governmental aid should mainly be directed to families in need. Therefore, these systems have little impact on the relationship of resources between the father and the mother, and thus tend to promote maternal involvement in childcare more than paternal, given that such interspousal decisions are based on opportunity costs, and that most mothers earn less than male partners. In Canada, 10 weeks of parental leave were introduced in 1990. In 2001, it was extended to 35 weeks. This can be shared between partners, but no exclusive father weeks have so far been introduced. Parents receive 55 % of the wage, low earning parents up to 80 % (Columbia University 2011; Gauthier and Bortnik 2001; Marshall 2003). Parental leave was most recently introduced in the Netherlands and the United Kingdom. In both countries, it remained unpaid and of comparatively short duration throughout the period under study. In the Netherlands, unpaid parental leave first came into effect in 1991. Parents received individual leave entitlements equaling 13 times the number of weekly working hours (Groenendijk 2005). In 2009, the entitlement was expanded to 26 times the weekly working hours per parent (Clauwaert and Hager 2000; Groenendijk and Kreuzenkamp 2009). The United Kingdom has long been reluctant to introduce parental leave policies, and they even blocked the parental leave proposals of the European Economic Community in 1983 and 1991. It then agreed to abide by the 1996 EC Directive, which directed all EU member states to introduce a minimum of three months parental leave for each parent. Since 1999, parents in the United Kingdom have had the right to three months unpaid leave after the birth of a child (Haas 2003; Pronzato 2009; UNICE et al. 1996).

In sum, a wide variety of leave systems are evident through the countries and decades under study. The parental leave systems differ between countries and points in time with respect to the duration of the leave that is available to both parents, the duration of weeks exclusively available for the father, and the benefit rate. This implies different incentives for fathers to spend time with their children.

3 Data

Time-use data are usually either derived from time-use diaries or from stylized questions in questionnaires referring to a certain time span before the interview (e.g., the preceding week) (Gershuny 2012). Stylized questions tend to lead to inaccurate estimations of time use because of incomplete recall, a lack of clarity about the definitions of time-use categories, and the risk of social desirability bias (Gershuny 2012; Kan 2008; Monna and Gauthier 2008; Niemi 1993; Plewis et al. 1990; Presser and Stinson 1998). Time-use diaries are therefore a more accurate and reliable source of data about people’s time allocations, especially with regard to unpaid work (Frazis and Stewart 2012; Monna and Gauthier 2008). For example Gershuny (2012) shows that the use of information derived from stylized questions does not increase the explanatory power of models for activities that are generally undertaken every day, such as childcare.

The time-use data and other micro-level data on which this study is based stems from the Multinational Time Use Study (MTUS) versions 5.52, 5.53 and 5.80 (Gershuny and Fisher 2010). The MTUS provides representative samples of individual data with diary records from 20 countries from the 1960s until the 2000s. The analyses are restricted to countries with at least two surveys per country at different points in time in order to capture changes over time. The following countries are analyzed: Canada, Finland, Italy, Germany, the Netherlands, Norway, Sweden, and the United Kingdom. The number of surveys ranges from two (Italy, Germany, Sweden) to seven (Netherlands). The earliest survey was conducted in Canada in 1971; the most recent surveys are from the Netherlands and the United Kingdom in the year 2005. The sample consists of fathers who live with their partner in the same household (whether married or not), who are between 20 and 55 years old and who have at least one child below the age of 18 in the household. The overall sample size amounts to 58,864 fathers.Footnote 4 The number of observations throughout the countries ranges from 2,897 (4.9 %) in Norway to 16,208 (27.53 %) in the Netherlands. The largest share of observations stems from the German survey in 2001 (8.1 %), the smallest from the United Kingdom in 1995 (0.33 %). The macro-level parental leave data is collected from international databases and reports, as well as articles on individual countries or country groups (see Appendix for a detailed documentation of the variables and the references from which they originate).

Our dependent variable is the number of childcare minutes on the survey day. Childcare is one of 69 different main activities recorded in the MTUS. It includes the following activities with/for children: preparing meals, feeding, putting to bed, medical and bodily care, looking after the child, helping with homework, reading to the child, playing. Thus, all kinds of activities primarily done for or with the child are considered to be childcare. The average number of childcare minutes of all fathers is presented in Fig. 1 for all surveys. It reveals that this ranges from 10 min in the United Kingdom in 1974 to 67 min in the United Kingdom in 2005. Over time, childcare time increased in all countries but Sweden. In the cross-country comparison, average minutes prove to be particularly low in Italy (1989: 18 min) and quite high (≥50 min) in Sweden (2000), Canada (1998) and the United Kingdom (1995, 2005). The two British surveys particularly stand out, but analysis of the distribution of variables reveals that the patterns are similar to those in the other British surveys. Thus, we included these surveys, but also checked the robustness of results without these surveys.

Fig. 1
figure 1

Fathers’ average minutes of childcare on the survey day. Sources: MTUS (Gershuny and Fisher 2010); own calculations and illustration

Full size image

The average minutes per day are calculated on the basis of all fathers, as they are all considered in the analysis, regardless of their participation in childcare. Amongst the 58,864 fathers in the sample, 26,435 fathers participate in childcare, i.e., they spend more than zero minutes with their children. In relative terms, the percentage of participants varies from 18 % in the United Kingdom in 1974 to 60 % in Sweden in 2001. Generally, the participation rate has gradually increased over time in all countries except for Sweden and Canada.

The main independent variables are three key factors for the availability and use of parental leave of fathers: the duration of parental leave for both parents (family right), the duration of exclusive parental leave for fathers, and parental leave benefits. The duration of the two types of parental leave are measured in weeks. The family right captures all weeks of parental leave that may be transferred from one partner to another. A transfer to the mother is not possible for exclusive father weeks. As to the benefit category, we distinguish between no benefit, moderate benefit (i.e., flat-rate benefit or less than 55 % of the wage), and high benefit (at least 55 % of the wage). The value assignment of parental leave variables depends on the parental leave scheme at the time of the birth of the youngest child. A range of further characteristics could be consulted to shape differences in parental leave legislation between countries. However, due to a lack of data and appropriate harmonization, as well as for theoretical considerations, this information is not suitable for a cross-country comparison.Footnote 5 Hence we restrict our analysis to the three key parameters presented above and use country and survey fixed effects in the models to capture remaining policy variation across countries and over time.

As the data does not provide information about the use of parental leave, we refer to whether a father has been eligible for certain parental leave measures after the latest birth. The age of the youngest child is available in three categories: less than 5 years, 5–12 years, and 13–17 years. This means that we assign the father eligibility for the measure if it has come into force before the birth of the child. If a particular leave arrangement was available for the majority of years of a certain age category, fathers are coded to have been eligible for this measure. For example, parental leave for 13 weeks was introduced in the United Kingdom in 1999. In the 2005 survey, the coding is as follows: if the youngest child is younger than 5 years, the duration variable has the value 13. If the youngest child is between five and twelve, the duration takes the value zero, because most fathers (age of youngest child below six) could not have taken parental leave. Fathers whose youngest child is between 13 and 17 are also assigned the value zero, because no parental leave was available in the birth years of these children.

Table 5 in the appendix depicts the parental leave legislation for fathers in our sample. The duration of parental leave as a family right varies from 0 to 156 weeks, exclusive parental leave weeks for fathers from 0 to 30 weeks. ‘Daddy weeks’ are available in only eight of the 30 surveys. There is no parental leave benefit in a large number of surveys, especially in the earlier ones.

The choice of independent variables at the individual level is made according to their relevance to fathers’ involvement in childcare as presented in related theoretical and empirical literature (see Reich 2013 for an overview). The models account for the following individual-level variables: number of children, age of the youngest child (0–4, 5–12, 13–17 years), age of the father and its square, his educational level, his work status, and whether the diary refers to a weekday or a weekend day. Educational level is coded in three categories: lower than completed secondary education (not completed ISCED level 3), completed secondary education (ISCED level 3 or 4), and above secondary education (ISCED level 5 or higher). As for work status, four categories are available in the data: not employed, part-time employed, full-time employed and employed with unknown working hours. These categories refer to the general employment status, not to the amount of work on the diary day. Since the number of surveys has clearly been prioritized over the inclusion of certain independent variables, some potential independent variables could not be included in the regression because they were missing in many of the surveys, e.g., marital status and partner characteristics.

In several models, dummies for the countries and the decades (1971–1979, 1980–1989, 1990–1999, 2000–2005) are included, in order to assess the impact of parental leave policies net of time and country fixed effects. In addition, further macro-level factors that account for country- and time-specific differences potentially related to fathers’ participation and minutes of childcare are included in the model. Paternal involvement is likely to depend on prevailing time cultures for paid and unpaid work in the country in the survey year. Hence, workload variables for different kinds of unpaid work have been used. The workload is defined as the sum of the average number of minutes men spend on an activity and the average number women spend on the same activity. The workload for childcare captures the time- and country-specific childcare time culture. The workload for housework (not including cooking) accounts for time-flexible, the workload for cooking for time-inflexible housework time-use culture. This distinction is important as time-inflexible housework such as cooking limits time opportunities for other activities more than time-flexible tasks (Hook 2010).Footnote 6 The female employment rate, as reported in the OECD Statextracts (2011), is used as an indicator for the prevalence of modern gender roles and for women’s bargaining power. This indicator is especially useful as the deployed micro-data for many surveys lacks individual information on the mother’s work status, which is usually assumed to have an impact on a father’s involvement with his children. These macro-level variables are assumed to capture major differences in time cultures and gender roles. There may be other factors influencing paternal childcare time, but these are not available for all countries and surveys from 1971 to 2005, e.g., the share of children in public childcare.Footnote 7 As such factors are related to gender roles and time-use cultures, they are at least partially captured by the macro-level variables included.

4 Econometric specification

A wide variety of models have been used to analyze time-use data, but the main discussion is between Tobit and Ordinary Least Squares (OLS) models (Craig and Mullan 2010; Foster and Kalenkosky 2012; Stewart 2009). According to recent analyses, the OLS model is superior to the Tobit model since the OLS model generates consistent and unbiased estimates, even if the fraction of zero-value observations is high on the diary day (Frazis and Stewart 2012; Stewart 2009). Thus, OLS models have been successfully applied in time-use studies (e.g., Craig and Mullan 2010; Kendig and Bianchi 2008). We use OLS models for the main estimations. Subsequently, Tobit models are used as part of the robustness checks. We estimate four different models. The first model—OLS1 (O1)—contains parental leave characteristics as well as individual-level variables, hence the equation is defined as

$$ Y_{i,c,t} = \alpha_{i} + \beta_{1} PL_{i,c,t} + \beta_{2} IL_{i,c,t} + u_{i,c,t} $$
(1)

where Y denotes the number of childcare minutes on the survey day, PL a vector of parental leave variables, IL a vector of individual-level variables, and u the error term. The indices i, c, and t show that the data is individual-level data from different countries and at different points in time. As paternal childcare minutes are likely to depend on country-specific norms and traditions, we additionally control for country fixed effects in the second model, O2:

$$ Y_{i,c,t} = \alpha_{i} + \beta_{1} PL_{i,c,t} + \beta_{2} IL_{i,c,t} + \beta_{3} CT_{i,t} + u_{i,c,t} $$
(2)

CT denotes the country dummies. Next, we introduce dummies for the decade of the survey, in order to control for the overall long-term trend that fathers’ childcare time has increased in most countries. Therefore, the equation for model O3 is

$$ Y_{i,c,t} = \alpha_{i} + \beta_{1} PL_{i,c,t} + \beta_{2} IL_{i,c,t} + \, \beta_{3} CT_{i,t} + \, \beta_{4} DC_{i,c} + u_{i,c,t} $$
(3)

with DC as the survey decade. This model shows the effect of parental leave characteristics net of time (decade) and country fixed effects.

As other factors on the macro-level can affect ‘time-use cultures’ and hence paternal childcare time, we control for additional macro-level variables (ML) in the fourth specification (O4). This specification is our preferred specification:

$$ Y_{i,c,t} = \alpha_{i} + \beta_{1} PL_{i,c,t} + \beta_{2} IL_{i,c,t} + \, \beta_{3} CT_{i,t} + \, \beta_{4} DC_{i,c} + + \, \beta_{5} ML_{i,c,t} + u_{i,c,t} $$
(4)

This specification controls for time- and country-specific female employment rates as well as a couple’s average daily minutes for three categories of unpaid work in a given year and country: childcare, housework and cooking.

The results for the main independent variables are expected to become less significant as more independent variables are introduced from O1 to O4, as unobserved heterogeneity between countries and points in time as well as across time-use cultures could influence the size of the coefficient as well as the significance of parental leave variables. Table 1 depicts the summary statistics of the dependent variable, childcare minutes of fathers per day, and all independent variables.

Table 1 Summary statistics
Full size table

5 Results

5.1 Main results

The results for the main models of fathers’ childcare time are presented in Table 2. The results from model O1, which does not control either for country or time fixed effects or for other macro-level indicators, are shown in column 1. In this model, weeks of parental leave as a family right, as well as for fathers exclusively, and moderate or high parental leave benefits (compared to none) are positively associated with paternal minutes of childcare. An increase of 1 week of transferable parental leave and one exclusive ‘daddy week’ are associated with an increase in paternal childcare time of less than 1 min. The linkage between transferable leave and paternal childcare behavior is weaker than that of leave earmarked for the father, suggesting that policies which explicitly focus on fathers are likely to be more successful in boosting fathers’ time investments in their children than more generally oriented instruments. Furthermore, a change from no benefit to a moderate benefit is related to a 3.2-min change in childcare time, and a change from no to high benefit is related to a change in childcare minutes of 10.4 min per day.

Table 2 Fathers’ childcare time–main models
Full size table

In the next step (model O2), we include country fixed effects. Unobserved permanent country specific characteristics might have a direct effect on paternal childcare and also be correlated with parental leave policies in a systematic way. For example, fathers in more egalitarian countries may do more childcare. If more gender egalitarian countries also have more permissive parental leave policies, failure to account for country fixed effects would bias the correlation upward, as the PL coefficient will be capturing the higher country egalitarianism.

The results for model O2 in column 2 of Table 2 show that taking country fixed effects into account does not alter the significance of most parental leave variables. The coefficient of moderate benefits loses significance in model O2, however, suggesting that most variation in this policy instrument stems from cross-country differences. However, the linkage between transferable parental leave weeks, exclusive father weeks as well as high benefits on the one hand and paternal childcare time on the other hand has even been reinforced, indicating a reasonable within-country variation in those measures over time.

In model O3, we additionally control for time fixed effects, in order to capture the correlation between parental leave characteristics and paternal childcare time, net of time (decade) trends. As parental leave became more generous over the past decades (see Table 5 in the appendix), and fathers’ average childcare time has risen at the same time (Chalasani 2007; Hall 1998; Maume 2011; Sandberg and Hofferth 2001; Sayer et al. 2004; Sullivan et al. 2009), regressions without time fixed effects would overestimate the pure measurement effect in recent surveys. Indeed, all parental leave coefficients that are positive and significant in model O2 are still significant but smaller in O3. This applies to transferable parental leave, exclusive ‘daddy weeks’ and high benefit rates.

The correlation between parental leave policies for fathers and their childcare time may depend on cultural values that are not captured by country and time fixed effects. For example, where it is common that mothers engage in paid labor, parental leave policies may be more generous for fathers, and childcare may be distributed more equally between the parents. Thus, the estimates would still be biased upward. That is why model O4 additionally includes country level variables that capture the time-use cultures of three kinds of unpaid work—childcare, housework, cooking—as well as the female employment rate. This is the model that comes closest to the ideal of capturing the pure effect of the parental leave characteristics. As seen from column 4 in Table 3, it supports the positive relationship between transferable parental leave weeks and paternal childcare minutes. An increase of 1 week is related to an increase of less than a minute of paternal childcare time per day. Weeks of exclusive parental leave for fathers are not significant in this model. The loss of significance of the ‘daddy weeks’ could be related to its correlation with the female employment rate and workload for childcare time since the coefficients of these two newly introduced variables are positive and significant. The relationship with the female employment rate is especially plausible, since mothers are expected to have a higher labor market attachment in countries where parental leave for fathers is explicitly promoted.

Table 3 Fathers’ childcare time–interaction between fathers’ parental leave duration and educational level
Full size table

The coefficient for high benefit rates is, again, significant. Net of country, time and several cultural fixed effects, there is a positive association of high benefits with paternal childcare time. In particular, the existence of high benefits (compared to none) increases paternal childcare time by 7.4 min per day. In contrast, the results for moderate benefit rates remain insignificant, qualitatively supporting the result of model O3. In sum, it seems that the number of parental leave available to both parents and the number of weeks earmarked to the father, as well as high leave benefits, are related to paternal childcare time net of country and decade effects, and that the result for parental leave for both parents and high benefits even holds when other macroeconomic factors capturing time cultures and female labor market participation are taken into account.

Regarding individual-level variables, most results remain stable across the model specifications. Taking model O4 as a basis, a father’s age and his educational level affect paternal childcare time positively. An additional year of age increases a father’s childcare time by round about 2 min, but the negative effect of the squared age implies that the magnitude of the positive effect of his age declines with each additional year. Highly-skilled fathers spend 11.6 min more time with their children than their more lowly educated counterparts. Furthermore, fathers spend 12.5 min more on childcare on weekends than on weekdays. Negative effects are found for children’s age and paternal employment status. If the youngest child is between 5 and 11 years old instead of younger, fathers spend 31.3 min less on daily childcare. The negative impact amounts to 42.4 min if the child is a teenager, i.e., between 13 and 17 years of age. Compared to non-employment, part-time employment results in a decrease of childcare minutes by 12.7 min, full-time employment by 18.2 min, and work with ‘unknown’ work hours in a decrease by 23.0 min. Furthermore, the number of children does not affect a father’s childcare time.

As expected, the inclusion of the time dummies in model O3 shows that paternal minutes of childcare have significantly increased over time. However, taking additional macro-level variables into account (model O4), the sign is even reversed, concluding that changes in time-use cultures of paid and unpaid work are the driving forces behind the temporal change. Model O4 reveals that the female employment rate and a couple’s average daily childcare time in a specific country and survey are positively related to a father’s childcare time.

These results are in line with earlier findings from Bygren et al. (2011) in that measures supporting the dual-earner/dual-carer family model (here: parental leave for fathers exclusively) are positively related to paternal childcare. They also support the results of Smith and Williams (2007) in that wage compensation can be positively related to paternal childcare time. We go one step further by showing that the results from the authors that rely on categorized information originating in retrospective time use data endure even when metric information from paternal self-reported childcare minutes arising from time-use diaries are directly taken into account. In contrast to Hook’s (2006) results, not only is the mere availability of parental leave for fathers important, but also the exclusive right to take the leave. Hook finds that fathers who are eligible for parental leave devote 19 more minutes per day to unpaid work, but it remains unclear which parent in fact takes the leave and whether it is really childcare that is increased. We extend the results from Hook in a twofold way: We directly address childcare as the component of unpaid work that we are interested in, and we relate this behavior to a policy exclusively targeted at fathers. To this end, our empirical findings give particular evidence for the hypothesized effect of father- oriented policies. We furthermore go beyond the investigation of Rege and Solli (2013) since our results confirm their finding of the crucial importance of a father’s exclusive right to take leave in a multi-national context.

5.1.1 Heterogeneous effects of parental leave policies by fathers’ educational attainment

As the literature has pointed to the different behavior of fathers with different educational levels, we test whether the impact of exclusive parental leave weeks for fathers differs according to their educational level. According to our specification presented above, highly educated fathers are those with more than a secondary education, medium educated fathers those with completed secondary education, and lowly educated fathers those with less than secondary education.

Table 3 shows that parental leave weeks available to be taken by both parents are negatively related to the childcare minutes of fathers with a medium or high educational level in almost all of the models O1a–O4a. In particular, each additional week of transferable leave reduces paternal weekly childcare minutes by approximately 1 min, if he has at least a secondary education. We conclude that the positive association of transferable leave to paternal childcare is solely driven by lowly educated fathers.

The coefficients for the interaction between exclusive parental leave weeks for fathers and a high educational level are positive in all four OLS models, the corresponding interaction with moderate educational level is not significant. Compared to fathers with a primary educational level, those with secondary education show similar behavior. For both groups, there is no relationship between ‘daddy weeks’ and their childcare behavior. On the contrary, fathers with a tertiary degree do more childcare when exclusive leave is available. Thus, exclusive parental leave weeks for fathers are only positive and significant in all models if combined with a father’s high educational level. In this case, an increase in 1 week of exclusive parental leave is associated with an increase of approximately 1 min of childcare per day. Hence, the correlation between exclusive parental leave weeks for fathers and paternal childcare time is solely driven by highly educated fathers.

A similar pattern is observed regarding high parental leave benefits. The coefficients of the interaction effect between those and a high educational level are positive and significant in all four models. In model O4a, high benefits are associated with almost ten more childcare minutes per day among highly educated fathers compared to lowly educated fathers. The interaction between high benefit rates and a medium education is only significant in models O3a and O4a, and the coefficients are smaller (than for highly educated fathers), indicating a weaker relationship. In contrast, for moderate benefits, the association is positive for fathers with a medium educational level, but not significant for those with a high educational level. That means that whereas fathers with medium education respond to any kind of benefit, fathers with graduate education are particularly sensitive to a high benefit level.Footnote 8

5.2 Robustness checks

Several variations of the models have been estimated in order to test the robustness of the results. These include Tobit models as well as variations in the sample, the parental leave variables and the dependent variable. The results for the parental leave variables are summarized in Table 4.

Table 4 Fathers’ childcare time–robustness checks
Full size table

In the Tobit models T1 to T4, the set of variables are the same as in the OLS-models O1 to O4. Qualitatively speaking, i.e. referring to the sign of the coefficient and its level of significance, model T1 supports the results of model O1 for all parental leave variables. Comparing model T2 to O2, the positive significant results for parental leave weeks (transferable and exclusive weeks) as well as high benefit rates are supported by both models. Moderate benefits show a non-significant coefficient in O2, and a negative significant coefficient in T2. The coefficients that are positively significant in models O3 and O4 become insignificant in models T3 and T4, respectively. Similarly, the coefficients that are not significant in O3 of O4 are negatively significant in T3 and T4. Thus, no coefficient changes its sign from positively significant to negatively significant or vice versa when comparing the OLS model to the corresponding Tobit model.

This pattern is also observed when comparing the interaction effects between exclusive parental leave weeks for fathers and the educational level in models T1a-T4a. In particular, the coefficients for moderately educated fathers remain insignificant or become negatively significant, while the former positive coefficients for highly educated fathers are insignificant in the Tobit models. The OLS results for the interaction effects with moderate benefits levels are supported in the Tobit models, while the results for those with high benefits lose their positive significance. The negative correlation between transferable leave weeks and paternal childcare time for fathers with a medium or high educational level are supported in the Tobit models. Thus, we conclude that Tobit models partly support the results derived from OLS models, bearing in mind that OLS models are the superior specification (see Sect. 4).

Next, variations in the sample are tested. The OLS models are re-estimated without the surveys from the United Kingdom from 1995 and 2005 because of their unusually high number of childcare minutes (models O1b–O4b). They are also estimated without the Swedish surveys due to the striking decrease in average childcare minutes from the first to the second survey (models O1c–O4c). Moreover, we test whether the results remain the same for fathers whose youngest child is 11 years old or younger, instead of 17 or younger, as the main results have shown that fathers devote significantly more childcare time to younger children (models O1d–O4d). All of these analyses qualitatively mainly support the results presented in the main models above.

We also test different definitions of parental leave characteristics. We first substitute the number of exclusive parental leave weeks for fathers with a dummy variable (as in Hook 2006), indicating whether or not a father had access to exclusive parental leave weeks at the time of the youngest child’s birth (models O1e–O4e). The results show that the relationship between the dummy specification of fathers’ exclusive parental leave and paternal childcare time loses significance once time fixed effects are taken into account (model O3e). It is concluded that it is the duration of exclusive parental leave weeks for fathers that is more important for paternal childcare minutes than the mere existence of such a policy.

6 Conclusion

This article presents cross-national and cross-time analyses that contribute to the empirical evidence on the relationship between parental leave policies and paternal childcare time. We find that parental leave weeks available to both one’s parents as well as the existence and number of fathers’ exclusive parental leave weeks are positively related to their childcare time, even when taking country and time fixed effects into account. Due to associations with the female employment rate and the average workload of parental childcare time in a country, the significance of exclusive parental leave weeks is lost when these covariates are incorporated into the model. A clear positive relation throughout models is found between a high benefit level (compared to none) and paternal childcare time.

Furthermore, we find that the correlations between the parental leave variables and paternal childcare time reasonably depend on fathers’ educational level. While more highly educated fathers are found to respond less to parental leave weeks for both parents than their lowly educated counterparts, the coefficients are in turn most positive for highly educated fathers in their interaction with exclusive parental leave weeks for fathers as well as with high benefit rates. These findings are robust throughout most model specifications.

The fact that we find positive effects for parental leave policies on a father’s childcare time, particularly for highly educated fathers, is particularly important in the current institutional framework. Several studies which compare parental leave systems and their effect on the gendered pattern of paid and unpaid work conclude that there is no current scheme which leads to an equal division of parental leave between the partners (De Henau et al. 2007; OECD 2007). These studies point out that the duration of non-transferable leave entitlements for fathers should be extended at the expense of parental leave as a family right, which mothers tend to take up additionally to their individual entitlement and maternity leave. Furthermore, benefit rates should be designed in such a way that it does not matter financially which parent uses the leave, which would be the case if the benefit amounts to 100 % of the former wage (De Henau et al. 2007; OECD 2007). As benefit rates are frequently high but less than 100 % of the wage and combined with leave as a family right, they are also attractive to highly qualified mothers, hampering their positive effect on paternal childcare time. Our study provides a more optimistic view of parental leave policies and, as parental leave policies expand in industrialized countries, sets up the basis for further research on this topic.