MOTHER'S EMPLOYMENT: CULTURAL IMPRINT OR INSTITUTIONAL GOVERNANCE?: Belgium, West and East Germany in comparison

In the public and scientific discussion on the political scope of action for improving employment opportunities for mothers, cultural aspects of female labour market participation often remain ignored. This study argues that employment decisions of mothers vary not only with the institutional settings but also with the dominant representative culture of a region. In order to study the effects of culture and institutions on maternal employment, Belgium, East, and West Germany are selected for empirical analysis. In a first step the article demonstrates that these regions differ considerably in terms of family political measures but also in terms of the cultural orientations with regard to female employment. While both Belgium and East Germany (but not West Germany) have well-developed public childcare systems for the under-3s, East Germans have a more favourable opinion towards female employment than Belgians. In a second step the study shows by means of panel analyses that these cultural differences strongly determine female employment behavior. Results indicate that both East and West German mothers have significantly higher employment probabilities in the long-run than Belgian mothers. The well-developed childcare infrastructure and the short parental leave lead only temporarily – right after childbirth – to higher employment rates among Belgian mothers. With increasing age of the child the significance of cultural dispositions result in higher overall employment probabilities among German mothers. These empirical findings support the theoretical approach of Pfau-Effinger in treating both institutional structures and culture as (equally) crucial factors in explaining cross-national differences in mothers' employment.

Female employment has risen in the past decades in almost every developed country. However, in most countries starting a family turns out to be a restrictive factor for women's labour force participation while men's employment behaviour is hardly affected. But there are great differences in the employment patterns of mothers in Europe: the magnitude of the child effect on women's labour force participation is comparatively low in the Northern and strong in the Southern countries, while the other European states range in between. These variations in maternal employment patterns in Europe are often traced back to different family policies (cf. Esping-Andersen 1999: 179).

Even though we agree that family policy measures are important factors in explaining national differences in employment behaviour of mothers, we want to show in this paper that the efficiency of such measures is strongly dependent on the cultural dispositions prevalent in a country. Moreover, we claim that even in Western European countries the cultural dispositions supporting female and mothers' employment are not as uniformly prevalent as usually expected.

To test this hypothesis, we will examine the labour market participation of mothers in Belgium, West and East Germany. It will be shown that these regions – despite all other similarities – differ considerably in terms of religious values, attitudes towards work and family, and gender role orientations.

These differences are the result of idiosyncratic historical pathways typical of a variety of European countries: while Belgium is deeply rooted in a straight Catholic tradition, the eastern part of Germany was the ideological centre of the Protestant movement and later shaped by the socialistic institutional and cultural programmes of the German Democratic Republic (Froese and Pfaff 2005). West Germany represents a rather mixed history of both religious denominations. We argue that these historical origins matter in the present, even though all of these regions have been influenced by a general trend towards secularization. This is because cultural values that govern individual behaviour are path-dependent and persistent over long periods of time (Inglehart and Baker 2000), but also because the trends of secularization are by far not as uniform and sweeping as was stated in the past (cp. Casanova 1994: 17). Thus, in our opinion, the persistence of religion in Europe, its influence on cultural values, and its direct and latent impact on concrete social action has ultimately been underrated.

In this context the comparison of Belgium with West and East Germany is especially instructive, since cultural values and political measures are ‘inconsistent’ in Belgium (Leitner 2003b). On the one hand, women in Belgium are rather traditional, male-breadwinner oriented. On the other, they can rely – just like East Germans – on an excellent childcare infrastructure for toddlers under three years.¹ West Germany, in contrast, has a rather poor childcare infrastructure. The comparison of Belgium with East and West Germany should therefore give an answer to the question whether family policy arrangements are indeed the operative point in explaining differences in the employment behaviour of mothers between different countries.

With this analysis we follow Pfau-Effinger's approach in treating the relevance of cultural values ‘for the explanation of international differences with respect to social phenomena’ (Pfau-Effinger 2004: 1). However, we do not focus on the interplay of culture and policy measures, but try to separate both dimensions and estimate their independent relevance to the explanation of a mother's employment behaviour.

The paper consists of five sections. In the first part we sketch the framework of the three areas (Belgium, West and East Germany) in terms of family policies and culture. In a first step we describe the family policy measures that structure mothers' opportunities for entering or remaining in the labour market after the birth of a child. In a second step we analyze the dominant representative cultures of the three regions using the European Value Study (1999). We use cross-sectional regression models to examine each area's cultural characteristics and test whether the three areas differ considerably in terms of cultural orientations related to female employment.

In the second section we derive hypotheses, and set up the theoretical test design for our main analysis. The third part introduces the data sets and methods. Our data sources are the German Socio-Economic Panel (SOEP) and the Panel Study of Belgian Households (PSBH). By means of random effect regression models, we test in what way the employment probabilities for mothers (with children in different age groups) differ between the three areas. Section four presents the results of the analyses. We conclude with a summary of the main results.

The correlation between the design of family policies in a country and the employment rate of women with and without children is well known (cf. Lewis and Ostner 1994; Gornick et al. 1997, 1998; Meyers et al. 1999; Leitner 2003a; Geyer and Steiner 2007). In theory, it is in fact simple to argue – on a rational choice base – that measures of family policies provide incentives and opportunities that structure the decision-making process on the micro level (cf. Becker 1981; Bryant 1990). In some nations (especially in the conservative regime types, cf. Esping-Andersen 1990 ,1999) family policies aimed at the male-breadwinner model produce high costs for women's labour market activity: they explicitly support unpaid domestic work by granting tax relief if only one spouse is gainfully employed, and implicitly restrict labour force participation by providing only restricted public childcare opportunities for toddlers. In other nations (especially in the Scandinavian social democratic type) individual taxation and a fully developed public childcare system produce high costs for women who stay at home doing unpaid domestic work. Consequently, many analyses show that the labour force participation of women is low in the first and high in the second case (Ehrenberg and Smith 1994; Daly 2000; Korpi 2000; Gornick and Meyers 2003).

From this point of view we can state that the countries selected for this study both belong to the same type, that is, the conservative welfare state (Esping-Andersen 1990 ,1999). In fact, it is true that the family policies in Germany and Belgium have substantial similarity in generally supporting the male-breadwinner model. On the other hand a closer look reveals significant differences. They become apparent when analyzing the parental leave and childcare systems in more detail.

We will therefore provide a brief description of the family policy systems in Belgium, West and East Germany. We will limit our view on political measures that directly or indirectly shape the costs for women with and without children when entering or remaining in the labour market. Following Bettio and Plantenga (2004), we distinguish three relevant categories: ‘time off’, ‘services’, and ‘money’.

Time off comprises family policy components like maternity or parental leave that give parents the opportunity to interrupt their occupational career after the birth of a child for a certain period of time. Maternity and parental leave imply protection against dismissal during the leave period and (in the case of Belgium and Germany for the observation period) a flat-rate transfer payment. In most countries this measure goes along with a rather weak public childcare system (services), since it is regarded as the family's duty to bring up the children.

Examining the three areas in terms of ‘time off’ and ‘services’, we find that the Belgian system appears more ‘social democratic’ than ‘conservative’ in this respect (Table 1). Indeed, Belgium has a well established public childcare system. At the same time the opportunities for parents to interrupt their career for a longer period after childbirth are highly restricted. The official parental leave is only three months. An alternative is the career break which allows an interruption of up to 12 months or a reduction of working hours for a longer period. However, this measure was not primarily created for mothers with young children, but can well be used for other occasions, e.g., in cases of serious illness or care dependency of another family member.

In West Germany the situation is different. The restricted public childcare system for children under three years is compensated by providing the opportunity of a three-year parental leave: a family policy combination which explicitly supports women's (temporary) withdrawal from the labour market after the birth of a child.²

The case of East Germany is of particular interest because it combines the sufficient public childcare system carried over from the German Democratic Republic with the parental leave measures of the Western part.

The conservative base coat of the Belgian family policy system comes to the fore when examining the tax system (money). Even though Belgium has had a partially individualized tax system since 1989, the transfer of deductions between married partners is possible. Consequently, the tax load of a couple where only one partner is gainfully employed is on average 11 percentage-points lower than in cases where both partners earn about the same (cf. Table 1). The German ‘Ehegattensplitting’ has, in the end, the same effect, though it is produced by other means. In this case the partners can jointly assess their income for taxation. Because of the progressive tax rate system, couples are at an advantage if their incomes are unequally distributed. Almost like in Belgium, the tax load differs by an average of 12 percentage-points between a double-earner and a single-earner household (cf. Dingeldey 2002; OECD 2006).

Another important measure to treat within the category ‘money’ is child allowance. Since the payments depend on the number, and in Belgium additionally on the age, of children living in a household, the indicator shown in Table 1 gives an impression of the proportion of child allowance at an average wage (for a single-earner family with two children of two and seven years of age). It becomes obvious that the values on this indicator are very close (8.7 vs. 9.9 percent) which leaves us to conclude that in both countries this measure produces relatively equal outcomes.

Considering these facts, we find that of the three areas examined, West Germany has the strongest male-breadwinner-oriented institutional system: early childcare is seen as a family task. Thus, the state admits the possibility of a lengthy parental leave but does not engage in public childcare for under-three-year-olds.

Even though the territory of East Germany is embedded in the same policy framework, there is one remarkable difference: because of the socialist history, the East disposes of a well-developed childcare system that was taken over after the reunification. In this respect Belgium and East Germany are in fact very similar. The coverage rates of childcare provision are very close: 37 percent in East Germany and 30 percent in Belgium, compared to 3 percent in West Germany (see Table 1).

However, it also turns out that the institutional gap between West Germany and the two other regions (Belgium and East Germany) in terms of childcare is restricted to the time when the child is under three years of age. After this period the different systems converge. (The coverage rates for children aged between three and five years in Belgium, East and West Germany are very similar.)³ The same is true for parental leave, which is the second crucial difference between the areas’ childcare arrangements: in this case Germany (East and West) and Belgium differ. However, since parental leave has its origin in the Federal Republic of Germany it can be seen as a compensating measure for the insufficient public childcare provision.

To sum up, we can roughly distinguish between two different institutional stages that correspond with the age of the child: in the first stage (when a child under three years is living in the household), we have significant differences between the regimes in terms of childcare and parental leave (cp. Figure 1 ⁴ ). In the second phase (after the child reaches the age of three years), these differences vanish, because the provision of childcare places is nearly similar and parental leave opportunities expire in all regions.

Although the significance of shared cultural values has always been one of the central issues of sociology (cf. Durkheim 1977 [1893]), its explicit conceptualization within a broader theory of social action is the achievement of Talcott Parsons (1968a [1937], 1968b). For Parsons, social action and interaction is indeed impossible without people having a common cultural background (Parsons 1951: 11). Though an actor might be rational in the sense that he tries to choose the means that are most efficient to reach an end, the choice of the end itself is part of the cultural process (Parsons 1968a [1937], Vol. II: 699; Schimank 1996: 83). Following this trace, we argue that the tendency of women in general to enter or not to enter the labour market is in the first instance an end that is shaped by the cultural background of the society in which an individual grows up and lives. The means that are allocated by the welfare state will in this sense only be accepted and used by a majority if they correspond with the culturally defined ends in a society.

We define culture in accordance with Schwartz (1992) and Hofstede (1980) as a ‘system of shared values’, where values are ‘tendencies to prefer certain states of affairs over others’ (Hofstede 1980: 19). Moreover, values legitimate these preferences and therefore certain patterns of social behaviour (Kremer 2007: 62). The importance of values goes back to the fact that they are learned at an early stage in life. They are therefore normally unconscious to those who hold them and persistent over time (Bourdieu 1997: 105). In defining culture in this way, we distinguish sharply between culture and structure: culture is located in the personal system and describes a ‘collective programming of the mind’ (Hofstede 1996). Structure on the other hand refers to the material/institutionalized side of society (Rippl and Seipel 2008: 21).

The history of Belgium, East and West Germany is a typical European story: geographically close but culturally and politically developing for long periods in different directions. Belgium's historical course was certainly strongly influenced by its Spanish occupation at the end of the Dark Ages (1550–1700). It was isolated from the religious reformation process, and later on an outstanding example of the Counter-Reformation (Erbe 1993: 158). Just like the European Mediterranean countries, Belgium was continuously influenced by the Catholic Church, which sustainably formed its cultural character. Moreover, Catholicism was positively associated with Belgium's struggle for independence which was achieved in 1830 (Martin 1978: 230). As a result, to this day Belgium has resisted the rise of any significant Protestant movement (Bureau of Democracy, Human Rights and Labor 2007).

In contrast to Belgium, the eastern part of the current German territory was strongly influenced by the Reformation process, not least because the movement itself had its origin in Wittenberg (Moeller 1977: 53). A second major intervention reached the eastern part of Germany after World War II when Germany was divided and the socialist regime of the GDR established in the East. Both events are important when considering that East Germany today is the most secular region in Europe (cp. Davie 2007: 37/38).

In comparison to Belgium and East Germany, the religious development of West Germany is more complex. On the one hand it was also strongly influenced by the Reformation and later by the ‘secular religion of German nationalism under Hitler’ (Martin 1978: 231) where Christianity only played a subordinate role. On the other hand the Catholic Church did not lose its influence in the same way as in the eastern parts (mainly because the Counter-Reformation was more successful in the West). Today West Germany is therefore more diversified, with an almost equal distribution of Catholics and Protestants (Flora et al. 1983: 58).

We argue that these differences still play a central role in the development of domestic cultures. They are transmitted over generations and shape the socialization processes beneath the level of institutional programmes and therefore carry tradition into the modern context.⁵ This is true even though they are not part of the explicit educational strategies of contemporary societies which, in Europe, might be heading towards a more uniform model (Meyer 2005). But they are part of the implicit and unconscious knowledge and symbolic system and inscribe themselves in the daily routine activities (cp. Therborn 1995: 10; Fagot et al. 2000). Even the advocates of modernization and secularization theory have lately been forced to admit that path-dependencies are strongly visible in post-industrial societies (cp. Inglehart and Baker 2000) and undermine a uniform development.⁶ Moreover the persistence of religious orientations of individuals under institutionally secularized conditions becomes obvious within recent studies on religion. While Davie (2007) determines a tendency in Europe of ‘believing without belonging’, Campiche argues that there are two faces of religion: while the one is on a downtrend (the institutional side), the other is on the rise (universal component) (Campiche 2004). In this light the secularization of society and the irrelevance of religion for its inhabitants become at least ambiguous. By merely focusing on the institutional level of societies sociologists are clearly on the safe side in terms of operationalization; but they probably miss some of the central mechanisms of social life and the vigorous presence of the past which is conserved within the bodies of the individuals (Bourdieu 1997: 107).

The historical background of the regions opens up the field for a more detailed analysis. Since we are primarily interested in the dominant representative culture (Tenbruck 1989: 22) of the present society, we will have to take a closer look at the attitudes of the people living in these regions today. We argue that these attitudes are path-dependent (Schwinn 2006: 209), i.e., religious socialization, values towards family and work, and gender roles vary significantly between the three regions. We expect Belgium to be the most traditional country, with a strong community-oriented religious socialization (Durkheim 2006), a strong familial orientation, and the highest approval of the traditionally gendered division of labour. On the other side East German culture is expected to be characterized by a high degree of secularization, a strong labour market focus (of women), and as a consequence, an egalitarian gender role orientation. West Germany should range in between the scale with a moderate traditional/egalitarian set-up.

To test these hypotheses, we use the European Value Study (EVS) from 1999. We restrict the sample to our core interest group in the further analysis: females between 20 and 45 years of age.⁷

We select six items to picture three dimensions of cultural imprint. We admit that these items are not sensitive enough to depict the complexity of cultural orientations and differences between the regions. But they are certainly able to give an impression of the differences. Therefore the following aspects were chosen: religion, family/work orientation, and gender roles. These aspects should give an idea about the general cultural climate within a certain region which will likely influence the implicit ends and tendencies of individuals and groups. We assume that this is true even if certain individuals will deviate from the mainstream culture. Moreover, we expect that the cultural climate measured by these dimensions structures the orientations of women towards labour market participation. This is not for completely rational reasons in the sense of clear-cut ideals for good child care or explicit gender roles (even though they also play an important role). Rather the effectiveness of gender roles and gendered socialization is to a strong degree unconscious to those who hold them. Therefore, women in more ‘conservative’ regions will probably not make a clear decision against gainful employment from the beginning. But they will likely ‘indulge their gendered selves’ (Charles and Baker 2009) in the way prescribed in their culture; that is, tend to behave in a way that will produce these consequences.

The religious dimension is represented by church attendance at the age of 12 (socialization aspect) and the importance of religion in life. The field of family/work orientation is covered by two questions which ask for the importance of the family and the priority of work over spare time. The gender role dimension (cf. Pfau-Effinger 2004: 47) will be captured with the agreement/disagreement-level on two statements: (1) ‘Being a housewife is just as fulfilling as working for pay’ and (2) ‘A job is all right, but what most women really want is a home and children’.

These variables will be analyzed within separate regression models. In order to conduct a consistent and robust analysis for all variables, we recode the ordinal answer scales to dichotomous categories that can be used as dependent variables in logistic regressions. With the exception of two cases, the answer scale consists of four categories. For each of these variables we define the first category (strongly agree) as a positive outcome (1). All other categories are recoded to ‘0’. The indicator ‘priority of work over spare time’, which is available on a 5-ary scale, is aggregated for the categories ‘strongly agree’ and ‘agree’. Religious socialization refers to a scale with eight categories in which the frequency of attending religious services at the age of 12 was surveyed. In this case the first two scale points (once a week and more often) are defined as positive outcomes.

The country of origin of the respondents (Belgium, West and East Germany) constitutes the central independent variable within the regression models. It is operationalized as a set of dummies with Belgium as reference. We also check for a number of socio-demographic aspects: these are marital status (married vs. not married), having own children (yes vs. no), age in years (open), employment status (employed vs. not employed), level of education, and size of hometown (both 8-ary; for coding see EVS 1999).

Table 2 gives a first impression of the results. It provides the relative frequencies of a positive outcome on the recoded cultural items.

As expected, Belgium is the country with the strongest religious attachment in terms of socialization and importance of religion in life today. On the other side East Germany is the most secular region. These results are impressively confirmed in the logistic regression model when checking for socio-demographic factors (cp. Table 3, columns 1 and 2).

The findings relating to the dimension of the importance of family and work reveal a very similar picture. Although in all regions the importance of the family meets with majoritarian approval, significant differences occur. Nearly all Belgian women (94 percent), but only three quarters of the East German female population, between 20 and 45 years of age strongly agree to this statement. Bearing in mind the historically early and broad Protestant imprinting and the socialist female working tradition in East Germany, it is not surprising that nearly 61 percent of this group in comparison to only 22 percent of the Belgian sample agrees that ‘work should always come first, even if it means less spare time’. The tendencies detected in the cross-tabulation are confirmed again within the regression analysis. East Germany and West Germany differ (in the mentioned order and in comparison to Belgium) significantly negatively in terms of the importance of the family and significantly positive in terms of the importance of work (cf. columns 3 and 4 in Table 3).

In the third step the gender role conceptions are investigated. We find that results are consistent with previous findings: Belgian respondents agree far more often to both statements regarding the legitimacy of female absence from the labour market (‘Being a housewife is just as fulfilling as working for pay’ and ‘A job is all right but what most women really want is a home and children’). East and West Germans show a significantly lower degree of approval to these statements. These bivariate findings are again strongly supported by the regression analysis (cf. columns 5 and 6 in Table 3).⁸

Even though we cannot go into details of the results in this context, it can be clearly concluded that the three areas considerably differ in terms of the cultural dispositions of their inhabitants. The predicted differences between a more traditional, family-orientated culture in Belgium and a rather egalitarian, work-orientated one in East Germany, with West Germany in between, were strongly supported by the results in all three dimensions.

In the previous section we have extracted the differences and similarities between the three regions in terms of family policy and culture. We will now combine both perspectives and develop an integrated picture of the areas which allows us to deduce hypotheses.

We want to show that the selection of the areas adds up to a test design which enables us to decide whether cultural imprint and/or institutional set-up are independent factors that affect the probabilities of maternal employment.

In the section on family policy it has been argued that the institutional arrangements in the three areas can be divided into two stages. These stages relate to the age of the child. The line was drawn at the point where the policy measures converged. The first stage begins with the birth of a child and ends when the child is three years of age. In this stage we have different family policy arrangements in the three regions:

• -

  Belgian mothers are confronted with a highly restricted maternity leave system but a well-developed child care infrastructure.

• -

  In West Germany the official parental leave is granted for up to three years. On the other hand the public childcare opportunities are highly restricted in this time.

• -

  East Germany is a special case, since it combines the parental leave measure from the West with the well-developed public childcare system taken over from the former GDR.

In the second stage, which starts when the child reaches kindergarten age, we find that the institutional settings converge; first, because in all regions parental leave measures are no longer available and second, because kindergarten in contrast to crèche infrastructure is available in all areas to a similar degree.

Against the background of these information we can set up our first two hypotheses, which only take into account the probable effect of the institutional arrangements (while the cultural differences between the three areas remain unconsidered). The first hypothesis relates to the probabilities that a mother of an under-three-year-old will be employed; hence, the institutional systems between the three areas are considerably distinct:

The second hypothesis relates to the second stage when the child is three years of age or older. It takes into account that the institutional settings converge between the three regions. In this period the political measures themselves cannot directly be held responsible for regional differences in the labour market behaviour of mothers. However, differences in maternal employment can of course result from a long-term effect of the different institutional settings. Rational choice theory argues, for example, that a long employment break due to the usage of a three-year parental leave reduces the probability of returning to the workplace. This is because of amortization of human capital (cf. Becker 1993; Arun et al. 2004):Our third hypothesis takes into account the cultural differences between the three regions. In section two we have shown that the inhabitants of Belgium, West, and East Germany differ considerably in terms of their cultural orientations, and that these orientations can be ranked on a scale with the opposing ends ‘traditional’ and ‘liberal’. We have argued that these differences matter in terms of women's general orientations towards the labour market. We have come to the conclusion that East German women (on the liberal end) should have the highest, and women from Belgium (on the traditional end) the lowest orientation towards the labour market. This ranking differs considerably from the one we derived when only accounting for the family policy framework (which was: Belgium, East Germany, West Germany).

We will indeed juxtapose the cultural hypothesis in part in opposition to the institutional hypotheses. But in doing so, we cannot ignore the institutional setting that frames the decision process. This is because the family policy framework in the first stage is rather coercive: in West Germany a woman who decides to work right after maternity leave has indeed extremely high opportunity costs because of insufficient public childcare provision. In Belgium, on the other hand, parental leave is restricted to a period of three months: a mother who wants to avoid risking her job has no other choice than returning to the labour market right after this period.

These definite regulations can hardly be overcome by cultural values, which operate on a different level and by different means. Therefore we expect that the clear-cut restrictions and opportunities set by the family policy in Belgium and West Germany will dominate the cultural orientations in the time when the child is under three years of age. We expect that cultural orientations start to dominate the decision process in the second stage. This is mainly because the three areas are quite similar in terms of family policy interventions when the child is older than three years. In addition, the measures themselves lose their restrictive character, because the demand for childcare decreases with increasing age, i.e., increasing autonomy of the child. As a consequence, we expect that the cultural dispositions clearly come to the fore by the time when the child is three years or older:

In order to test the hypotheses, we use the Socio-Economic Panel (SOEP) for East and West Germany and the Panel Study of Belgian Households (PSBH) for Belgium. Both are representative longitudinal surveys of private households. Due to the fact that the PSBH was ceased in 2002, we restrict our analyses for Belgium to the period from 1992 to 2002 (11 waves). In Germany we start in 1992 and end in 2003 (12 waves).

To meet the requirements of our analytical subject, we select from both data sets only women (with and without children) who are between 20 and 45 years of age.⁹ Using these criteria, the total sample size is N=10,284, with N=5,884 women with children. They are distributed among the areas as follows: West Germany: N=5,605; East Germany: N=1,897; Belgium: N=2,782.¹⁰

To test our hypotheses, we use logistic regression models in which the employment status of a person is the dependent variable (see below). In doing so, we have to consider the fact that our data base is a panel in which information for one and the same person is available for different points in time. As a consequence, the regression model has to treat two kinds of variances: variance within and variance between persons. Two regression approaches are commonly used to handle this kind of data: fixed effects and random effects regression models (Greene 2002; Wooldridge 2002, 2003; Andreß et al. 2006; Baltagi 2008). Because of the restrictions of the fixed effects model in dealing with time constant covariates (cp. Andreß et al. 2006: 12), we choose the random effects model.

The dependent variable of the regression (labour force participation) is dummy-coded and has a positive outcome either if a person is in fact actively working part-time or full-time on the labour market or if a person is currently searching for a job.¹¹ Women who are housewives and women on maternity or parental leave are both treated as not working (0). In contrast to that, we treat unemployed women who are actively looking for a job as employed. This is because we have to control for female unemployment rates between the three areas. Such differences would certainly correlate with labour market participation probabilities of women but have nothing to do neither with cultural orientations nor with opportunities or restrictions produced by family policies. Indeed, in East Germany the unemployment rate of women is clearly largest (17.9 percent in 2002 [Destatis 2010]), while the rates in Belgium and West Germany are roughly the same (Belgium 8.7 percent Eurostat 2010,¹² West Germany 7.2 percent [Destatis 2010]).

The central independent variables consist of two major dummy variable sets. The first set simply indicates whether a person is living in Belgium, West or East Germany. In the analysis we use West Germany as reference.

The second major variable set addresses the question whether the respondent has a child living in the household or not. Besides the presence of a child, we have to account for the age of the youngest child in the household. For this purpose we use six age categories: birth phase (0 to <1 year), infant (1 to <3 years), kindergarten child (3 to <6 years), primary school child (6 to <10 years), schoolchild (10 to <13 years), and teenager (13 to <16 years). Women without children are in the reference category.

Since we are not only interested in the effect of a child on employment behaviour, but also in the differences of the child effect between the three regions, we use another variable set consisting of interactions between region and the child's age. By combining these independent variables, we can produce nested estimates for each region and for each age group of a child. In the end we are able to show in which way the three regions differ in terms of employment probabilities for mothers with children in different age groups.

Although these are the major effects we are interested in, we have to extend this basic model by including a variety of variables that control for socio-demographic composition. For this purpose we additionally enquired about age, education (3-ary), partner in household, partner gainfully employed,¹³ marital status (married vs. not married), number of children living in the household and an indicator for period effects (period 1: 1992–1995 [reference], period 2: 1996–1999, period 3: 2000–2003).

Figure 2 shows the results of the random effects logit model for our central independent variables (complete model cp. Appendix, Table A1). The predicted values are displayed as odds ratios.¹⁴

The x-axis of the diagram distinguishes between women with children in different age groups. The y-axis visualizes the odds ratios and thereby the negative or positive divergence in terms of probabilities of labour market activity in comparison to West German women without children (reference category). Values below 1 indicate that the likelihood of being active on the labour market is below that of childless women from West Germany. Values above 1 indicate that the likelihood is higher.

It can be seen that the coefficients develop with increasing age of the youngest child in two different ways. In Belgium the odds ratios remain on a relative constant level throughout the observed age span of the child. Only in the first year after the birth (OR ;= ;0.10) is a negative derivation from this constant course evident. This is probably because some respondents were interviewed within the three-month time span of maternity leave (which is obligatory for Belgian women).

The situation in both German parts, especially within the first three years of the child's life, is completely different. The birth of a child in these areas is an incisive event in terms of women's employment. The coefficients fall steeply within the first year in both German regions to an odds ratio of 0.01. With increasing age of the child the coefficients climb again and slowly converge to the level of the reference category. When examining the first stage, when the child is under three years of age, in more detail, one can also see that the maternal employment probabilities between women from West and East Germany drift apart very fast in the second and third year. While in both areas mothers steadily re-enter the labour market, the pace of this process is much higher in the East. This becomes obvious when comparing the odds ratios in the time when the youngest child is aged between one and three years, which is 0.10 for East and 0.03 for West Germany.

The results for the first stage prove that our institutional short-term hypothesis (1) is correct: Belgium has in comparison the highest, West Germany the lowest rate of working mothers in the time when the child is under three years of age. We can easily trace this result back to the differences in the institutional set-up: the sufficient public childcare system in Belgium and the short parental leave opportunity enable Belgian women to remain continuously on the labour market. On the other hand West German mothers, confronted with insufficient childcare and the possibility of a three-year parental leave, drop out of the labour market during this period. East German women who have the choice between both measures (parental leave and public childcare) drop out of the labour market to the same degree as West German mothers in the first year after childbirth. But afterwards they return significantly faster (yet without reaching the level of Belgian mothers in the first phase). This result indicates that East Germans frequently use both measures: parental leave to stay with their newborn child within the first year¹⁵ and public childcare to return to their job before the child reaches kindergarten age.

If we now take a look at stage 2, we are confronted with a new situation. First of all, it seems that by now almost all East German mothers have returned to the labour market. The odds ratios in the time when the child is aged between three and five years already correspond with the values for West German childless women (OR ;= ;0.96). By reaching this level, the probability that they are active on the labour market is higher than the probability that their Belgian counterparts are (OR ;= ;0.34). West German women still have the lowest probabilities of being gainfully employed at this point (OR ;= ;0.14). But the gap between Belgium and West Germany also closes when the youngest child reaches primary school age (6–10 years) and finally reverses completely when the child becomes 10 years or older. At the end of the observed age span, we therefore have a completely different picture than at the beginning of stage 1: Belgian mothers who initially had the highest working probabilities turn out to have the lowest probabilities when the child is over six years of age. East German women with children, though usually interrupting their occupational career for at least one year after childbirth, were able to return very quickly to the labour market. As a result, they have by far the highest probabilities of being gainfully employed from the time on when the child turns six.

If we now examine our competing hypotheses (2 vs. 3), we have to notice that the different cultural dispositions start to dominate the process with the beginning of phase 2. As soon as the family policy measures converge, the cultural dispositions come to the fore and modify the established order. Mothers from East Germany, the region with the most egalitarian cultural imprint, have the highest employment probabilities right at the beginning of phase 2. Belgium, the country with the most conservative background, has the lowest probability for mothers with children aged six years or older. Therefore, we have to accept hypothesis no. 3 and reject no. 2.

It seems that the employment break does not undermine the general career orientations of women. This of course does not mean that the factual career opportunities in terms of prestige and income are independent of career breaks. We are therefore far from believing that the hypothesis of amortization of human capital can be rejected. But it can be stated that a career break does not hinder women from participating in the labour market in general if egalitarian, occupationally focused, cultural orientations prevail in a society. Reversely – and more importantly – a sufficient public childcare infrastructure does not necessarily increase employment behaviour of mothers in the long run if traditional cultural values prevail.

In this context we have to take into account that the occupational behaviour of women in Belgium is relatively unaffected by the birth of a child. This becomes obvious when taking a look at the detailed regression table in the Appendix. It is indeed not the child but the country as such which produces the general negative effect on the dependent variable in comparison to German women without children. This means that Belgian mothers do have almost the same employment probabilities as women without children in Belgium. But because of the rather conservative cultural orientation, Belgian women (be they mothers or not) are in general less willing to participate in the labour market than women from Germany are. Hence, even if the Belgian family policy measures erase the child effect, they are in the long run unable to increase the employment rates of mothers in comparison to countries with a more egalitarian gender culture.

The aim of this study was to unravel the effects of cultural imprint and family policy on mother's employment behaviour. We have used the cases of Belgium, West and East Germany to set up a most similar test design. We have demonstrated that the family policies in the three areas are in general male-breadwinner-oriented, but significantly different in terms of public childcare provision (Belgium, East Germany) and parental leave (West, East Germany). On the other hand the cultural orientations of the inhabitants of these regions differ significantly, with Belgians on the traditional and East Germans on the ‘egalitarian’ end of the scale.

This set-up allowed us to test whether culture and policy are independent factors in explaining employment behaviour of mothers. Particularly the analysis of Belgium in comparison to East Germany seemed promising, since both countries have sufficient public childcare but rather divergent cultural orientations.

The main analysis – based on random effects regression models with the German SOEP and the Belgium PSBH panel data – showed that the Belgian policy system produced the highest probabilities of labour market participation for mothers with children under the age of three. In this age group the restricted parental leave allowance (three months) in combination with the public childcare system produced the best ‘effort’. However, the advantage of the system does not last for long: when the child is three years of age or older, mothers from East Germany show a significantly higher employment rate than their Belgian counterparts. And when the child reaches school age, the probabilities of being gainfully employed are even higher for West German mothers.

These results suggest that culture is an important factor in explaining the concrete employment behaviour of women: Even though Belgium has excellent reconciliation policy measures, it has the lowest employment rate of mothers in the long run. It seems as if this is the result of a cultural script in which the family and its derivates are conceptualized as highly important and as the central domain for a woman to engage.

Dina Hummelsheim is senior researcher at the Max Planck Institute for Foreign and International Criminal Law in Freiburg (Germany). Her research interests are focused on criminology (victimization, crime-related perceptions and attitudes), sociology of the welfare state and the family, and quantitative research methods. Her most recent publications are: ‘Social insecurities and fear of crime: A cross-national study on the impact of welfare state policies on crime-related anxieties’ (forthcoming), European Sociological Review (with H. Hirtenlehner, J. Jackson, and D. Oberwittler); and When Marriage Ends: Economic and Social Consequences of Partnership Dissolution, Cheltenham: Edward Elgar (with H.-J. Andress).

Jochen Hirschle is a post-doctoral researcher at the Department of Sociology in Hagen (Germany). His area of research is on economic sociology, religion, and sociology of the family. His most recent publications are: ‘Die Angst der Mitte vor dem Abstieg’, in Zeitschrift für Soziologie 38, 2009 (with H. Lengfeld); and ‘Informelle und institutionelle Partnermärkte’, in Soziale Welt 60, 2009.