The influence of formal and informal support systems on the labour supply of divorced mothers

Most studies focussing on the economic consequences of divorce find that women suffer from a decreased level of prosperity following this event (e.g., Burkhauser et al.1991; Smock 1993, 1994; Bianchi et al.1999; Poortman 2000; McKeever and Wolfinger 2001). This is mainly related to their weaker attachment to and position in the labour market, and to the fact that the care responsibility for any children usually falls on the shoulders of the mother. In general, women experience a net income deterioration of about 20 per cent in the first few years after divorce (for an overview see Uunk 2004; Andreß et al.2006). As previous studies have shown that post-divorce employment acts as an important barrier against economic deprivation (Poortman 2000; Dewilde 2006), the comparative study of female employment patterns following this life event has important and policy-relevant implications.

The main research question of this contribution concerns the facilitating influence of childcare support systems on the post-divorce employment of women. Following Gunnarsson and Cochran (1993), we furthermore think that any study of support systems should take account of both formal and informal provisions. Or as Beck (1994): 106) states: ‘private and political strategies for solutions must be seen as connected’. Therefore, in this article we evaluate how formal and informal types of childcare support influence the labour market participation of women following divorce.

Concerning the impact of formal support systems, our main premise is that the post-divorce employment situation of women is largely dependent on welfare state policies supporting lone mothers with children to stay in or enter the labour market (Beck 1994). Cross-national comparisons focussing on female employment show that an important institutional arrangement promoting women's economic independence is the provision of childcare facilities (Gornick et al.1998; Drobnic 2000; Uunk et al. 2005). However, in a recent publication on the impact of institutional arrangements on the gender wage gap, Mandel and Semyonov (2005) conclude that while mother-friendly interventions enable women to become economically independent, these policies are also rather costly in terms of their later occupational and economic attainment.

Studies examining the influence of welfare state arrangements on the economic consequences of partnership dissolution for women are relatively scarce and most of them only analyse a limited number of countries (e.g., Burkhauser et al.1991; Dewilde 2002; Andreß et al.2006). In a recent study based on ECHP data, Uunk (2004) evaluates the impact of income- and unemployment-related welfare provisions on the income loss following divorce for a wider range of countries. While both types of measures temper the economic consequences of divorce for women, the impact of income-related policies seems to be stronger. Nevertheless, the provision of public childcare facilities also has a strong negative effect on the income loss following divorce.

Next to formal childcare arrangements, informal support networks also play an important role in facilitating the care burden after marriage dissolution (Leira 1990; Bromer and Henly 2004; Raeymaeckers et al.2008). Informal care providers like relatives and friends often facilitate the balance between work and care by providing practical help. Furthermore, from a social policy-perspective, it is crucial to evaluate the optimal mix of support systems (Motel-Klingebiel 2005). Therefore, we do not only pay attention to the influence of formal regulations, but also scrutinise the interplay between formal arrangements and informal care.

Whereas most authors measure the impact of formal childcare provision on individual outcomes by utilising only one indicator (Uunk 2004; Uunk et al.2005), in this article we consider the provision of childcare facilities and the coverage rate as two separate measures of day care policies (Stier 2006). This way, we are able to disentangle the effects of both supply and demand indicators. Furthermore, in order to test whether changes in post-divorce employment are driven by the general female employment-supportive climate rather than by specific institutional arrangements, we contrast the effects of informal support networks and formal childcare facilities with the index constructed by Mandel and Semyonov (2005) charting the whole of a country's family policies.

To conclude, this article focuses on the impact of formal and informal childcare provision on the labour supply of divorced mothers, using different types of macro-level indicators. As the focal point of our analysis concerns the effects of institutional arrangements, a comparative approach is needed. Using data from seven waves (1995–2001) of the European Community Household Panel (ECHP), we include 13 countries in our analysis: Denmark, The Netherlands, Belgium, France, Ireland, Italy, Greece, Spain, Portugal, Austria, Finland, Germany and the United Kingdom.

The outline of this article is as follows. In the first section, we discuss different ways to measure the influence of formal childcare support on female post-divorce employment. Secondly, we take a closer look at the role of informal networks. Next, we introduce the data, method and operationalisations, followed by the analysis. We end with an overview and discussion of our results.

The modernisation process and the dynamic of individualisation have made it possible for both sexes to liberate themselves from their ascribed male/female gender roles (Beck 1994). Whereas in the former industrial society men occupied an important breadwinning role and women were responsible for household tasks, the increasing female employment rate from the second part of the twentieth century onwards facilitated the independency of women. According to Beck (1994): 111), a woman's independence is determined by two factors: economic security and children. In his view, as long as women bear children and feel responsible for them, ‘children remain obstacles in the occupational competition, as well as temptations to a conscious decision against economic autonomy and a career’. Indeed, numerous studies have found that the presence of children has a detrimental effect on female employment (Joshi 1998; Drobnic 2000; van der Lippe and van Dijk 2002; Sorensen and Hill 2004; Vlasblom and Schippers 2005). The difficulties women experience when trying to reconcile work and family are often referred to as the ‘negative child-effect’ (Uunk et al.2005) or the ‘child penalty’ (Gornick et al.1998). The effect of children on the labour career of their mothers is however influenced by institutions at the macro-level. The way states regulate their family policy, or more specifically, the childcare arrangements they provide, influences the restraining impact of pre-school children on the labour market participation of their mothers (Stier et al.2001). Equal gender roles can only be created in institutional structures that support equality. In the following sections, we address the influence of welfare state arrangements on female employment.

A classic study focussing on the influence of macro-level institutions on individual outcomes is the work by Esping-Andersen (1990). His ‘Three Worlds of Welfare Capitalism’ reflect the different ways states and markets combine in order to produce and allocate (economic) welfare, resulting in different patterns of decommodification (or market-independence) and stratification. Esping-Andersen's typology is however criticised for its lack of attention given to the Southern-European welfare model (e.g., Ferrera 1996; Bonoli 1997), and by feminist authors because it neglects the care-giving role of women (e.g., Sainsbury 1994, 1996). Concerning the last issue, Sainsbury (1994, 1996) distinguishes between countries supporting an individual policy model promoting women's economic independence and countries supporting a male-breadwinner model. Applying the female independency issue to Esping-Andersen's (1999) typology, especially in social-democratic welfare states (Sweden, Denmark, Norway, Finland) female employment rates are particularly high (van der Lippe and van Dijk 2002). The liberal countries (Canada, Australia, United Kingdom, United States) occupy an intermediate position, while in conservative regimes, female employment is low (Germany, France, Belgium, Italy). Next to differences in female employment rates, different types of welfare states provide for different types of childcare provision (Gustaffson 1994). Social-democratic regimes provide universally available, state-subsidised public childcare. In liberal welfare states, the provision of childcare arrangements is a private responsibility and therefore mainly market-oriented. Conservative countries support the male-breadwinner model and provide limited public childcare. Care arrangements in these countries are neither market-oriented nor well-expanded.

Although Esping-Andersen's welfare regime typology inspired many authors, we notice that when taking a closer look at the way childcare arrangements are regulated, countries belonging to the same regime cluster actually show interesting variations. Norway for example differs from the other social-democratic countries in providing rather limited formal childcare arrangements (van der Lippe and van Dijk 2002). Gornick and Meyers (2003) find that conservative welfare states are characterised by distinct patterns of care arrangements. While Germany, The Netherlands and Luxembourg provide only limited public childcare, provisions are more extended in France, Belgium and Italy. We thus conclude that Esping-Andersen's welfare typology neglects specific country differences in the provision of public childcare (see also Stier et al.2001; Uunk 2004).

A possible way to take account of country-specific variations in policies is provided by Dewilde (2002) and Burkhauser et al. (1991). Both authors evaluate the effect of different institutional environments on the economic consequences of divorce for women by focussing more in-depth on specific countries. Dewilde (2002) combines the typologies of Bonoli (1997), Siaroff (1994) and Esping-Andersen (1990) and uses a sample of five countries (Denmark, Belgium, Germany, Italy and the United Kingdom). She finds that in social-democratic welfare states (Denmark), the high accessibility of the labour market results in a high proportion of divorced women being able to work full-time. This can be attributed to the extensive provision of publicly funded childcare arrangements (Dewilde 2002). In the United Kingdom, an example of a liberal Beveridgean high spending welfare state, the limited access to the labour market for women with care responsibilities results in a lower post-divorce female employment rate. Concerning the conservative welfare states, Dewilde (2002) finds that the financial consequences of divorce are more limited in Belgium, where childcare provision is ample, than in Germany. Burkhauser et al. (1991) compare Germany and the United States and find the German welfare state to be less able to protect women from reductions in relative economic status following marriage dissolution.

The above studies show that institutional arrangements influence female employment outcomes, which in turn has consequences in terms of the economic impact of important life events, such as partnership dissolution. An important drawback is however that these authors neglect the interplay between macro- and micro-level determinants. Because post-divorce employment is an outcome of both a country's institutional features and individual human capital factors, it is important to include the combined effect of both determinants (van der Lippe and van Dijk 2002; Stier 2006). We thus have to use a so-called macro/micro or multilevel approach.

In the existing literature on the effects of institutions on female employment we can distinguish two methodological approaches. The first considers a specific institutional environment as more than the sum of the parts and uses an index that represents a country's policy package (Sainsbury 1994, 1996; Mandel and Semyonov 2005). The second approach measures the influence of ‘domain-specific’ policy characteristics such as public childcare facilities (e.g., Uunk et al. 2004). We discuss both approaches in the next section.

Only recently, studies examining the influence of a country's macro-level arrangements have focused on the combined effect of indicators at the micro-level and structural characteristics at the macro-level. Fuwa (2004) uses both macro- and micro-level determinants to explain the division of household labour in 22 countries. Mandel and Semyonov (2005) adopt this approach to examine if country differences in family policies and wage structures influence the earnings gap between men and women.

Gornick et al. (1998) utilise a sample of 14 countries to study the effect of an index based on different indicators of child-support policy (maternity leave, paternity benefits, tax relief for childcare, guaranteed care coverage, percentage of children in publicly funded childcare) on female employment. In addition to supply indicators of employment-friendly policies, they also include demand indicators such as childcare and paternity leave coverage rates. Their results show a strong ‘negative child-effect’ in countries with low support of female employment (United Kingdom, United States). In countries with high support of female labour market participation (Belgium, France, Denmark, Finland), the child effect is positive. While a high score on employment-friendly policies results in high employment rates of women with both young and old children, countries with a low score tend to favour the employment of women with older children.

The index used by Mandel and Semyonov (2005) is an indicator of the family-friendly features of a country's policy environment. Their measure is based on the number of weeks of fully paid maternity leave, the percentage of children (0–6) in publicly funded childcare and public employment as a share of total employment. In this article, we use their index to test the effect of family-supportive policies on female employment following partnership dissolution. We hypothesise that family-supportive policies have a positive effect on female post-divorce employment (Hypothesis 1).

The disadvantage of using a composite index when measuring the influence of institutional characteristics is the impossibility to gauge the exact causal effect of specific welfare state arrangements on individual outcomes (Uunk 2004; Stier 2006; Pettit and Hook 2005). Therefore it is important to include separate indicators of specific welfare state policies. In the literature on female employment, ‘public childcare provision’ seems to be the most used macro-indicator (van der Lippe and van Dijk 2002). For instance, Gornick et al. (1998) find that the provision of public childcare arrangements facilitates female labour supply. In general, public childcare explains around one-third of the observed country differences in the reduction of work hours following the birth of the first child (Uunk et al.2005). We thus hypothesise that the provision of childcare facilities has a positive effect on female post-divorce employment (Hypothesis 2).

A shortcoming of the study by Uunk et al. (2005) is that they neglect the demand side of childcare arrangements. Pettit and Hook (2005) however also include a demand measure for childcare in terms of the percentage of children (0–2 years old) that actually uses publicly funded childcare. These authors find a positive relation between the demand side indicator and female employment. In order to observe a more methodologically correct relationship between childcare provision and female labour supply, it is thus necessary to include both the provision of childcare places and the coverage rate for each country. Or as Stier (2006) states: ‘measures of institutional arrangements should include both direct (e.g., specific policies and regulations) and indirect indicators (e.g., behavioural outcomes at the macro-level)’. Therefore, we not only include a direct indicator of childcare institutions (provision), but also an indirect indicator (coverage rate), as both can vary in a different way across the countries in our sample. Different factors influence this relationship. Firstly, the financial costs of childcare vary across countries and may lead to differing coverage rates. Secondly, the cultural values of a country can influence the use of childcare arrangements (Gornick et al.1998). Next to the effect of the provision of childcare facilities, we thus hypothesise that the childcare coverage rate has a positive effect on the post-divorce employment participation of women (Hypothesis 3).

Studies focussing on the social consequences of partnership dissolution find mixed evidence. When comparing single parents with couples, Gunnarsson and Cochran (1993) find that the number of networks of the latter category exceeds the first. However, while earlier studies observe that divorce has a decreasing effect on the social networks of the ex-spouses (Milardo 1987), recent studies using a broader operationalisation reject the negative divorce-effect. Terhell et al. (2004) find that a divorce can signify both a negative and a positive evolution. Much depends on the sex of the former partners, as especially men experience chronically impeded personal networks. However, a network loss after a divorce can be compensated at the long term (Terhell et al.2004). Kalmijn and Broese van Groenau (2005) also refute the devastating effect of divorce on networks and find that after a partnership dissolution, women tend to increase their contacts with kin and non-kin relatives. Social contacts with neighbours, however, tend to decrease. These recent studies of the social consequences of partnership dissolution show that divorced mothers do not always experience a reduction of their social network, and can still rely on support from their social environment. Leslie and Grady (1985) find that kin-ties in particular seem to provide divorced mothers with emotional and practical support when dealing with the negative consequences of divorce. Therefore, the influence of informal networks on the post-divorce labour market participation of divorced mothers is an interesting question that needs further examination.

Concerning the impact of social networks on the economic consequences of divorce, a conceptual distinction must be made between social activity and social support (Hughes 1988). Social activity is defined as the extent of social interaction, while social support is often referred to in terms of material or practical assistance. In this article, we evaluate the effect of social support networks on the female labour market participation following divorce. When women experience difficulties combining work and care, their existing social network could facilitate this combination by providing specific practical aid (Leira 1990). Jarvis (1999) points at the importance of the web of coping strategies of households in selecting their preferred childcare options. According to Leira (1990), women's lateral self-organisation of childcare has been as important as welfare state interventions or services in restructuring the balance between female employment and childcare. This leads us to the formulation of our next hypothesis: informal care networks have a positive effect on female post-divorce employment (Hypothesis 4).

It is thus possible that the assumed positive effect of social support networks on female employment after divorce counterbalances the state's inability to provide formal care and facilitates the combination of paid work and childcare. Jarvis argues that individuals and institutions are situated in networks which are often haphazardly interconnected and reflexively reproduced, resulting in an adaptable web of social relations which evolve in the context of changing demands. In Britain, for instance, where the formal childcare coverage is very low, it is not surprising that working mothers have to rely on help from family and friends (Borchost 1990). On the other hand, in a country like Belgium, where formal childcare facilities are widespread, the existence of informal support networks could be rather absent, and have no bearing on post-divorce female employment (Ghysels 2004).

The thesis that modern welfare states take over the functions of informal social networks is referred to as the crowding out-hypothesis. The trade-off between a country's policy arrangements and the material and practical assistance provided by informal support networks was already put forward by Parsons (1956), who argued that the expansion of state arrangements has a negative effect on the functions provided by the family and other networks. More recently, the crowding out-thesis was taken up by Putnam (1995) and others, who found that economic development has a negative impact on the importance of social networks.

The trade-off between formal and informal support systems is also present in the more recent work of Esping-Andersen on familialisation and defamilialisation (1999). Defamilialisation is defined as the degree to which households’ care responsibilities are relaxed by welfare state provision. According to Esping-Andersen, familialistic regimes are influenced by Catholic social teachings, meaning that public interference only takes place when family and social networks fail to provide certain functions – the so-called principle of subsidiarity. Especially in the Southern-European countries, families and social networks are considered as the most important source of social support; public interference in family matters is thus restricted. While in the social-democratic regime cluster, care tasks of households are to a large extent considered as a social responsibility, in the liberal welfare states the gaps in the public and family support system are in principle filled by the market. However, as markets only rarely substitute for public services or family self-servicing, in the liberal welfare states one might speak of a ‘concomitant welfare state and market “failure”’ (Esping-Andersen 1999: 64).

At a more disaggregate level, the trade-off between formal and informal childcare provision seems to be confirmed by Gunnarsson and Cochran (1993), who compare social networks of lone mothers in the United States and Sweden. They find that in Sweden, a country promoting extensive public childcare, lone mothers have less social support networks at their disposal. Hansen (2005) observes that in countries with less family-supportive public policies, the most common caregivers are the relatives outside the family.

Following the above line of reasoning, we hypothesise that there exists a crowding-out effect between formal and informal childcare support networks. We thus expect that when a state provides extensive childcare services, divorced mothers rely less on their social network to combine work and care (Hypothesis 5). We test the crowding-out effect at both the individual and the aggregate level.

As stated in the introduction, the analyses in this article are based on the European Community Household Panel (ECHP). The ECHP is a cross-nationally comparative longitudinal survey containing data from the 15 former European Union Member States. It is a so-called household panel study (HPS), that is a standardised questioning of an initial sample of households and individuals at regular–in this case annual–points in time. The sample of households and individuals is representative of the population in each of the participating countries. Cross-national comparability is guaranteed through a standardised design and common technical and implementation procedures (EUROSTAT 2005).

In this article, data from 13 countries are analysed: Denmark, The Netherlands, Belgium, France, Ireland, Italy, Greece, Spain, Portugal, Austria, Finland, Germany and the United Kingdom. Luxembourg is left out because of the small sample size and Sweden due to the lack of panel information.

ECHP data is collected in a prospective way, allowing us to use a longitudinal research design. This way, problems of selectivity are avoided. Indeed, the group of divorced women in a cross-sectional survey consists of women who experienced relationship dissolution over a longer period of time. Some women divorced many years ago, while others only recently split up with their partner. Furthermore, since most women eventually repartner, it is possible that the group of divorced women remaining single is quite a selective subsample of all women experiencing relationship dissolution.

Starting point for our analysis is the construction of a sample of women experiencing partnership dissolution between 1995 and 2001 (due to reasons of data availability, wave 1 is not taken into account). Although the number of divorces on a yearly basis is rather small, the ‘pooling’ of events in a prospective way allows us to construct a sample large enough follow for multivariate analysis. Divorce is defined as the transition from marriage or cohabitation in one year (t) to not living as a couple or living with a new partner in the subsequent year (t+1). Our sample thus contains both married and cohabiting respondents at time (t) and allows for the analysis of short-term changes in the labour supply of women following divorce. Restricting our analysis to women aged 16–64, living in a heterosexual couple at time (t) and caring for children below 12 years old, the number of divorce events amounts to 941, of which we have valid information for 836 cases. The number of events is rather small for Greece (n=20) and Ireland (n=23). The United Kingdom and Germany contribute the largest number of divorced women (n=134 and n=173), while sample sizes for the other countries range from 50 to 90. Comparable small numbers of divorce events have been reported in numerous studies for different European countries (e.g., Burkhauser et al.1991; Jarvis and Jenkins 1999; Poortman 2000; Uunk 2004).

Our main research question concerns the change in the labour supply of mothers following divorce. To this end, we compare the employment situation of the women in our sample in the year before partnership dissolution (year t) with the employment situation after this event (year t+1). Our dependent variable is operationalised as the difference between post-divorce and pre-divorce (weekly) working hours.

Micro-level control variables (measured at time t) are age, number of children below five years old, education, weekly working hours, employment and the use of formal and informal childcare. Education is measured as the highest level of education achieved. The ECHP only distinguishes three categories: (1) less than second stage of secondary education (ISCED 0-2); (2) second stage of secondary education (ISCED 3); and (3) recognised third-level education (ISCED 5-7). Respondents still at school are assigned to the lowest level of education. The effect of employment is measured with a dummy indicating whether the respondent is employed or not. Both indicators of childcare use are dummy variables (use/no use) and are derived from the following questions: ‘Are any of the children under 12 in the household looked after on a regular basis by someone other than their parent or guardian whether at home or outside such as at a crèche or garden?’ and ‘Does your household have to pay for any of the children to be looked after on a regular basis?’. We thus assume that paid childcare corresponds to formal childcare, while unpaid childcare is assumed to be provided by informal caregivers.

Our macro-level variables are the childcare coverage rate (demand), the provision of childcare facilities (supply), the informal childcare coverage and the extent of family-friendly policies. The childcare coverage rate is calculated on the ECHP data, and refers to the proportion of households with children below 12 years old using formal childcare (average figure for the years 1995–2001). Data on the provision of public childcare are taken from Uunk et al. (2005) and refer to the period between 1990 and 1995. Figures for Germany, however, refer to the situation during the late 1980s. Country scores for the availability of informal childcare are again derived from the data, and refer to the proportion of households with children younger than 12 using formal childcare (average figure for the years 1995–2001). The extent of family-friendly policies is operationalised as the ‘Welfare State Intervention Index’ (WSII) constructed by Mandel and Semnyonov (2005). Scores for Greece, Spain and Portugal are based on ILO data (ILO 2005) and Gauthier (1999) and are calculated using the formula of Mandel and Semyonov (2005).¹

The most appropriate method to evaluate the impact of macro-level arrangements on the difference in pre- and post-divorce working hours of divorced mothers is a multilevel analysis, as this method allows us to control for country-level differences in individual characteristics. The importance of taking human capital variables into account has been demonstrated by several authors (Smock 1993; Poortman 2000; McKeever and Wolfinger 2001; Drobnic 2004), who agree on the fact that women with high human capital in terms of substantial educational and professional resources seem to experience less severe economic consequences following divorce.

The Proc Mixed procedure in SAS allows us to assess the effect of both micro- and macro-level determinants for individuals nested within higher-level structures, and thus produces less biased estimates for the macro-level covariates, as standard errors are corrected accordingly (Snijders and Bosker 1999). By using a random intercepts model, we assume that the intercept varies between countries, but that the slope is constant.

In a first step, we calculate the explainable variation in the change in working hours before and after divorce. Next, the individual-level control variables are included. In the third model, the effect of the family-friendly policy-index on the change in working hours is calculated. The next step consists of including both the formal childcare coverage rate (demand) and the provision of childcare facilities (supply). In a fifth model, we contrast the provision of informal childcare with our composite index and finally in Model 6 we show how the individual-level use of formal and informal childcare are related.

Table 1 displays the country scores on our macro-level determinants. Looking at the availability of informal childcare support, we note that scores are lowest in the Scandinavian countries and highest in the Southern-European countries. The provision of informal help is however highest in the United Kingdom. This is in line with previous research by Wheelock and Jones (2002), who find that grandparents are important care providers for British mothers. Informal childcare provision is also relatively high in Belgium and France. Concerning our index of family-friendly policies, the Scandinavian countries display the highest values. Somewhat surprisingly, scores are furthermore relatively high in the Southern-European countries. These countries have relatively generous and/or long maternity leave arrangements and extensive public service employment (cf. Table 5 in Appendix). For instance, almost 80% of Spanish women are employed in the public sector (Simo 2005). Ireland and the United Kingdom have the least family-friendly policies. As we saw earlier, public childcare provision is high in Denmark, Finland, Belgium, France, and to a lesser extent, Portugal. The coverage rates are more or less in line with this trend, although based on this indicator we note a rather low score for Belgium, while the values for Germany and Italy are comparatively high.

Table 2 shows mothers’ average weekly working hours before and after partnership dissolution. In line with other research, women work relatively long hours in the Scandinavian welfare states (i.e. Denmark and Finland). Conservative welfare states such as Germany and The Netherlands have the lowest scores. Somewhat surprisingly, divorced mothers’ working hours are also relatively high in the Southern-European countries. This might be related to the fact that in these countries divorce is still more of an ‘elite’ phenomenon and thus less widespread. Women with high levels of human capital are more at risk of experiencing partnership dissolution, as they are the only ones who can afford to divorce (Goode 1970).

In general, with the exception of Ireland, mothers tend to increase their labour market participation after experiencing partnership dissolution. Employment can therefore be considered as an important coping strategy for women in order to deal with the financial losses caused by a divorce. Large differences are observed in Austria and Spain, where mothers work approximately seven hours more after dissolution. Next to Denmark (3.68 hours), the increase in working hours is also relatively elevated in Germany, France and Greece (about 2 hours).

Before examining the effects of our individual-level independent variables, we estimate our baseline model which serves as the starting point for our further analyses (see Table 3). In this model we include the pre-divorce working hours as a control variable because the employment situation prior to divorce is an important determinant of differences in pre-and post-divorce working hours. The variance components of Model 1 show that the between-country variation is significant. The explainable variance at level 2 is however low [=0.08=(19.427/(237.97 + 19.3272)]: 8 per cent of the variation in the change in working hours before and after divorce can be explained by country-level differences.

Looking at Model 2, which contains the other individual-level variables, we see that women with children below 5 years of age experience more difficulties to increase their working hours after divorce compared to mothers with older children. Furthermore, women who were already employed before divorce have a significantly higher chance of increasing their working hours compared to non-working mothers. Looking at the explained variance at the macro-level, we notice that R² reaches a value of 10 per cent [=0.099 = 1−(17.5092/19.4272)]. It thus seems that individual characteristics explain a reasonable amount of variation. Looking at the country residuals (Table 4), we however see that the differences in observed and expected country residuals, in comparison with the baseline model, have not declined for all countries in our sample.

Model 3 explains the change in mothers’ working hours following divorce by including the index constructed by Mandel and Semyonov (2005) that attempts to measure the extent of family-friendly policies in a country. This effect is significant at the 0.05 level. The positive sign indicates that female employment after divorce is facilitated in countries with more family-friendly policies. The explained country-level variation in comparison to the baseline model amounts to 42 per cent [=0.42 = 1−(11.2721/19.4272)]. Table 4 shows that especially in Denmark, a country characterised by a high WSII-score (93), the country residuals decrease substantially from 1.4 (Model 1) to −0.2. In other words, in Denmark family-friendly policies explain a substantial amount of variation in the dependent variable. Our first hypothesis predicting a positive influence from family-friendly policies is thus confirmed. These results are in line with Mandel and Semyonov's (2005) finding that, combining several components, the WSII-index ‘measures a broad phenomenon that represents more than the unique effect of each one of them’.

In Model 4, we additionally include our two indicators of formal childcare (supply and demand). First of all, we notice that our index of family-friendly policies no longer has a significant impact on the change in mothers’ working hours following divorce. Furthermore, both indicators of formal childcare arrangements turn out to be insignificant. The same is true when both are introduced separately, and they remain insignificant when the family-friendly policy index is removed from the model (results not shown). Both Hypothesis 2 and Hypothesis 3 are thus rejected. These results seem to contradict studies finding a positive effect of formal childcare arrangements on female employment (e.g., Uunk et al. (2005) on the changes in working hours following the birth of the first child). Additional analyses regressing the working hours of mothers on the provision of formal childcare reveal a positive and clearly significant effect. However, the change in working hours following divorce does not seem to be related to the provision of public childcare. The country-level variance explained by this model is 27 per cent [0.27 = 1−(14.1952/19.4272)], which is significantly lower compared to Model 3.

In Model 5, we include our macro-level indicator of informal support together with our index of family-friendly policies. While the effect of the composite index remains significant, we find no significant effect of informal childcare provision on the change in mothers’ working hours following divorce. The country-level variance explained by this model amounts to 32 per cent [0.32 = 1−(13.1686/19.4272)]. Although informal social networks do not exercise an independent influence, this model seems to explain more variance compared to Model 4. However, we have to conclude that Hypothesis 4 is not confirmed.

Finally, in Model 6 we evaluate the impact of our individual-level variables indicating the use of formal and informal childcare. Controlling for all macro-level indicators and other individual-level determinants, both indicators of childcare use are highly significant. However, when we exclude the use of formal childcare from the model, the estimate for the use of informal support becomes insignificant (results not reported), indicating that both support variables are interrelated. Hypothesis 4 predicting an independent effect of informal support networks is therefore also rejected at the micro-level, as we find that informal networks seems to fulfil a complementary role, enabling divorced mothers to use paid childcare in order to increase their working hours. Our crowding-out hypothesis (Hypothesis 5) is thus not confirmed. In this model, the family friendly-policies indicator stays significant at the 0.05 level. The explained country-level variance amounts to 30 per cent [0.30 = 1−(13.575/198.83)].

In this article, we evaluate the effect of both formal and informal childcare support systems on the post-divorce labour supply of divorced mothers. We model the change in working hours before and after divorce by using both micro- and macro-level determinants. In order to assess the influence of formal support systems at the macro-level, we chart the whole of a country's family policy environment by using the ‘Welfare State Intervention Index’ developed by Mandel and Semyonov (2005) and additionally include two variables that represent the formal childcare provision in terms of supply and demand. Next to formal support systems, we also look at the impact of informal childcare support systems.

Concerning the micro-level determinants, we find that being employed and the presence of young children are important indicators. Both variables are highly significant and influence divorced mothers’ possibilities to enter the labour market, the first in a positive way, the second in a negative way.

The combined strength of the three components included in our index of family-friendly policies results in a positive effect on female post-divorce employment. The country-level residuals show that particularly in Denmark – a country where family policies can be defined as highly ‘woman friendly’ – the model fits the data very well. This finding is in line with earlier research that policies supporting an individual model of social rights for both sexes facilitate women's independence. In other words, female employment after divorce is positively influenced by a policy environment that is characterised as highly woman-friendly.

Including the provision and coverage of formal childcare does not result in a better model. The absence of any significant effects seems to contradict studies reporting that female employment is positively influenced by extensive public childcare arrangements. A possible explanation is that our indicators are rather crude, and are thus no good measures of a country's facilitating influence on the ability of divorced mothers to combine work and care. We do however find evidence that the availability of formal childcare does influence mothers’ working hours as such, but not necessarily the change in working hours following divorce.

In contradiction to our expectations, our macro-level indicator referring to the extent of informal childcare support also does not exercise an independent influence on the change in divorced mothers’ working hours following partnership dissolution. However, compared to the model with only the composite index of family-friendly policies, the increase in the explained country-level variance does point at the facilitating role social networks can play in combining work and care.

This latter finding is confirmed when we look at the way the individual-level use of formal and informal childcare are interrelated. More specifically, we find that formal childcare solutions reinforce the effect of the use of informal support networks on the labour supply of divorced mothers. This is in accordance with earlier research stating that informal networks are of complementary importance. In this line of reasoning, it is for instance possible that in some countries working hours do not match the closing hours of childcare or schools and that divorced mothers have to rely on both kin and non-kin networks to bridge the gap between formal childcare and the demands of the labour market.

To summarise, we find empirical evidence for our main hypothesis that a country's institutional environment plays an important role in facilitating employment as a strategy for mothers to cope with their financial losses following partnership dissolution. However, contrary to our expectations we find that the change in labour supply is more responsive to the whole of a country's family policies rather than to so-called ‘domain-specific’ indicators of formal and informal childcare provision. In this sense, our results open up avenues for further research, as we need to look in other directions in other to explain country-level differences in the labour supply of mothers following divorce. Furthermore, our empirical evidence suggests that at the micro-level formal and informal strategies are connected. The complementary role of informal support systems, facilitating the use of formal childcare, is an important finding from a policy point of view. However, further research has to make clear which conditions have to be fulfilled in order to help divorced mothers to combine work and care, thus enabling them to mitigate the economic consequences of partnership dissolution.

ECHP data was made available through the Research Centre for Longitudinal and Life Course Studies (CELLO), University of Antwerp. EUROSTAT bears no responsibility for the analyses and interpretations presented here.

⁰⁰⁰⁵

Peter Raeymaeckers is a Research Assistant associated with the Research Unit on Inequality, Social Exclusion and the City (OASeS) at the University of Antwerp

Caroline Dewilde is a Post-Doctoral Research Fellow (Research Foundation-Flanders) associated with the Research Unit on Inequality, Social Exclusion and the City (OASeS) at the University of Antwerp

Laurent Snoeckx is a Research Assistant associated with the Research Centre for Longitudinal and Life Course Studies (CELLO) at the University of Antwerp

Dimitri Mortelmans is Associate Professor and head of the Research Centre for Longitudinal and Life Course Studies (CELLO) at the University of Antwerp