1In countries where fertility is low, the effects of family policies on fertility, female employment and the work-life balance have been widely studied, but findings diverge widely from one country to another. Given the broad diversity of national contexts and policies, comparison is very difficult. Anna Matysiak and Ivett Szalma overcome this difficulty in their comparison of Hungary and Poland, two countries which share many demographic, cultural, political and economic characteristics, but which have implemented very different family policies. They analyse employment entry and second births among mothers who already have one child in order to compare the effects of different parental leave measures on employment and birth timing.
2Parental leave entitlements are important policy instruments aimed at providing parents with time off work to care for a young child. They include maternity leave, which is usually the share of leave granted to women immediately before and after a birth to protect the health of mother and child. On the one hand, these leave entitlements enable employed parents, and most often women, to avoid the role incompatibility that comes with attempting to raise young children while in paid employment (Berger and Waldfogel, 2004; Thevenon and Solaz, 2013), so should have a positive effect on fertility and women’s employment. On the other hand, it has often been argued that generous parental leave entitlements may lead to greater losses in women’s human capital and may jeopardize women’s employment and earning prospects as a consequence (OECD, 1995; Ruhm, 1998; Evertsson and Duvander, 2011). Against this background, the effects of parental leave entitlements on fertility and women’s employment re-entry have been the subject of extensive empirical research. [1]
3This research has led to conflicting findings, however, partly because researchers are usually not able to account for a series of country-specific and individual-specific factors that may confound the real effect of a policy on individual behaviour, such as an interaction of the policy with a country’s cultural or economic setting, or differences in individual propensities to make use of the policy and adopt a given behaviour. Two methodological solutions have been proposed in the literature to deal with these problems. First, Neyer and Andersson (2008) suggested disentangling the effects of country or region specificities on policy effects by comparing the potentially most similar contexts which display well-recognized differences. Using examples from Nordic countries, the authors argue that we are better able to assess whether cross-country differences in individual behaviour are due to family policy interventions in a setting which is characterized by a strong “cross-country similarity over a considerable number of institutional, economic, and cultural factors” (p. 709) than in a setting consisting of dissimilar countries. Second, a number of statistical methods have been developed in the field of policy evaluation research with the aim of accounting for individual differences in policy uptake and in the adoption of a given behaviour (Venetoklis, 2002; Blundell and Costa-Dias, 2009; Khander et al., 2010). However, they often require an experimental or a quasi-experimental design or some other special research set-ups which are rarely available.
4This article contributes to the discussion on the effects of parental leave programmes on fertility and women’s employment. More precisely, following the approach proposed by Neyer and Andersson (2008), we look at how the parental leave payments affect second birth risks and women’s job take-up after first birth. To do so, we compare Hungary and Poland after the breakdown of communism until around the year 2005. These are two countries which differ strongly in parental leave programmes but display a number of similarities in their socio-cultural, economic and family-policy related settings. More specifically, the parental leave mandate in Hungary is universal and provides much higher financial compensation than the means-tested programme in Poland. As far as the similarities are concerned, the two countries share the experience of state socialism and a centrally planned economy which organized the thinking and functioning of both societies over nearly five decades following the Second World War. In both Poland and Hungary, the socialist states encouraged women’s labour force participation while at the same time supporting the traditional gender division of labour at home. This resulted in a dual earner – female double burden model (Pascall and Manning, 2000). Both countries underwent a relatively successful transformation to a market economy in the 1990s (Wyplosz, 2000), and despite a temporary economic slowdown, living standards improved substantially. In comparison to other post-socialist countries, Hungary and Poland also stand out for their particularly strong attachment to family values and their traditional perception of gender roles, especially regarding mothers’ involvement in the labour market. Already in the past, the two countries proved relatively resistant to public policies promoting women’s employment, and female labour force participation was relatively modest by Eastern European standards. This continued to be a characteristic feature of Hungary and Poland in the post-1989 period. Overall, the comparison of Hungary and Poland creates a very good setting for testing the effects of differences in parental leave payments on mothers’ fertility and employment behaviours.
5The first section of this article provides a review of empirical studies on possible effects of parental leave programmes on fertility and women’s employment. In the following section we describe the parental leave regulations in the two countries and elaborate on other aspects of the country contexts that might be relevant to fertility and women’s employment choices. After presenting our research hypotheses, we describe our data and method and then present and discuss our empirical findings.
I – Review of studies on the effects of parental leave
1 – Effects of parental leave on the transition to second birth
6Parental leave was introduced to enable working parents to take care of an infant without having to give up employment. It often goes in hand with parental leave benefits that compensate the parent for the income lost during the career break. From this perspective, paid and short parental leave periods should lower both the opportunity costs of staying at home and the direct costs of having children as they give a mother an employment guarantee, protecting her from job loss while providing her with additional income. However, longer leave periods lead to a depreciation of human capital and reduce women’s promotion opportunities, thereby increasing the indirect costs of children. The opportunity costs might be particularly high for highly qualified women who have much more to lose in terms of income and advancement opportunities than the low-educated. From this perspective, one might expect paid parental leave to encourage fertility more especially among low-educated women. However, if leave payments are small, this may strongly discourage low-educated women from enlarging their family while on leave, since they may not be able to afford a prolonged reduction in their income.
7Empirical research on the impact of parental leave mandates on fertility is still quite scarce and provides somewhat conflicting findings. Studies which apply a macro-level approach and investigate how developments in the total fertility rate (TFR) in most developed countries were affected by changes in family policies and economic conditions show that the effects of parental leave entitlements on TFR range from negative (D’Addio and d’Ercole, 2005) through insignificant (Gauthier and Hatzius, 1997; Hilgeman and Butts, 2009) to positive (Adsera, 2004; Luci-Greulich and Thevenon, 2013). Luci-Greulich and Thevenon (2013) suggest that the insignificant or negative estimates might result from a failure to control for the development of childcare services for under-threes, but this explanation does not apply, for instance, to the study by Hilgeman and Butts (2009).
8The micro-level research on the topic is no more conclusive than the macro-level research. In a study for Norway, Aassve and Lappegard (2010) investigated the role of the cash-for-care benefit which was introduced for parents who chose to stay at home to care for a child aged 12-36 months. Applying propensity score matching, they found that the benefit increases the risk of having a child and is particularly influential on second births. The same authors also showed that the main beneficiaries are low-educated women. In line with Aassve and Lappegard (2010), Vikat (2004) suggested that a similar cash-for-care benefit increases third birth risks in Finland. Somewhat less conclusive findings were provided for Norway by Lappegard (2008), who analysed effects of parental leave payments on subsequent births. Her findings suggest that the use of parental leave, at least for mothers, increases second birth risks, but lowers third birth intensities. Different conclusions were reached by Gerber and Perelli-Harris (2012) for Russia (1985-2000). The authors found no elevated risk of second conception for women on leave (paid fully for the first six months and partially up to the eighteenth month) but, remarkably, they observed that women who returned to work after taking more than six months leave had a significantly higher second conception risk than women who took shorter leave.
9These micro-level studies do not account for the selective use of parental leave entitlements. It is likely that women who take longer leave are more family-oriented and would probably have another child even if they were unable to take paid leave. This problem has been resolved by Lalive and Zweimueller (2009) who applied a regression-discontinuity design to investigate the effects of the prolongation of parental leave from one to two years in Austria in 1990. They concluded that the extension led to the birth of 12 additional children per 100 women and that the effect was stronger among the low-educated women. They further showed that the policy change affected not only the tempo but also the quantum of fertility. The latter finding was questioned by St’astna and Sobotka (2009), however. Another remarkable finding of Lallive and Zweimueller (2009) was that a policy reversal in 1996, when leave was shortened again by six months, had no further long-term repercussions on fertility.
10Likewise, there are very few studies which look into the fertility effects of parental leave entitlements by level of education, and their findings are inconsistent. For instance, Aassve and Lappegard (2010) find that among women claiming cash-for-care benefits in Norway, acceleration of the transition to a subsequent child is faster among highly educated than among low-educated women. According to the authors, this can be explained by a selection of highly educated women who claim the benefits into the group of family-oriented women. Other studies on the topic focus on the effects of a speed premium – that rewards parents who have another child soon after a previous birth – on childbearing decisions across social groups, but they again yield inconsistent findings. For Sweden, no educational differences in the transition to a subsequent child were found after the speed premium had been introduced (Andersson et al., 2006). By contrast, for Austria, low-educated and low-wage women were shown to react much more strongly to a speed premium than better educated and high-wage women (Lalive and Zweimueller, 2009).
2 – Effects of parental leave on employment entry
11The impact of maternity and parental leave on women’s employment is also ambiguous; it depends on the duration of leave and the related benefit entitlements. Generally, well-paid but short leave entitlements are considered to speed up women’s entry to employment after birth. There are several reasons for this: first, paid leave shields mothers from a potential job loss and hence reduces the time spent on a job search (Pylkkanen and Smith, 2003); second, short career breaks have limited negative impact on women’s human capital (Baker and Milligan, 2008; Evertsson and Duvander, 2011); and third, paid leave gives women time to search for an optimal childcare solution after they resume work (Thevenon and Solaz, 2013). Empirical research conducted in the US in the late 1980s and 1990s, when not all new mothers were entitled to leave, provides evidence consistent with these arguments, by showing that mothers with a leave entitlement returned to work faster than women whose employment contract was terminated due to a birth (Berger and Waldfogel, 2004; Hofferth and Curtin, 2006; Klerman and Leibowitz, 2000; Waldfogel et al., 1999). However, being entitled to leave may not be independent of the women’s opportunities or motivations to resume employment, and this factor was not taken into account in these studies.
12By contrast to short leave, longer paid leave may decrease women’s employment and earning prospects due to substantial losses in their human capital (Mincer and Polachek, 1974). This may, in turn, lower the demand for women’s labour as well as women’s interest in returning to paid work. In this vein, using a difference-in-difference approach, Schøne (2004) estimated that the introduction of a cash benefit for care of preschool children in Norway led to a 4% reduction in women’s labour force participation. Ronsen (2009) estimated the effect of the same benefit on women’s working hours and found that their working time fell by 3.75 hours (i.e. about 20%) over the four years following the benefit introduction. Lallive and Zweimueller (2009) investigated how the prolongation of leave in Austria in 1990 affected women’s employment and found that it delayed women’s labour market entry even after the leave period had expired, but had no effects on women’s cumulative work experience and earnings. A closer examination conducted by Lallive et al. (2011) shows that the decline in women’s employment is due mainly to the extension of the benefit. The importance of the parental leave benefit for delaying labour market return was confirmed by Pronzato (2009) in a pooled analysis of ten European countries. Her study shows, however, that cash transfers lead to a postponement of mothers’ employment entry mainly in the first year after birth. Finally, some studies have investigated whether the effect of parental leave entitlements depends on the duration of leave. Although they generally confirm the non-linear relationship between leave duration and women’s employment (Baker and Milligan, 2008; Evertsson and Duvander, 2011; Pettit and Hook, 2005; Rønsen and Sundström, 2002; Thevenon and Solaz, 2013), there is no agreement as to the maximum length of leave which is not detrimental to women’s employment. While some studies suggest that child-related career breaks lasting no more than four to five months (corresponding in practice to maternity leave) do not have a negative effect on women’s occupational careers (Baker and Milligan, 2008; Evertsson and Duvander, 2011), others yield much longer durations of two or even three years (Pettit and Hook, 2005; Thevenon and Solaz, 2013).
13The effects of parental leave payments on the employment rates of women may also vary with educational level. In general, highly educated women have higher opportunity costs of not working, such as lost earnings and promotion opportunities. On the other hand, low-educated women with little or no paid parental leave entitlements may not be able to afford the serious reduction in their earnings and standard of living if they stop working. Given these arguments, one might expect well-paid parental leave to have a stronger negative impact on the employment of low-educated women than leave with poor compensation. Even though education is the main factor determining women’s employment continuity after motherhood (Dex et al., 1998; Gutierrez-Domenech, 2005), relatively few studies have addressed the impact of parental leave entitlements on mothers’ employment by educational level and they yield mixed findings. In general, international comparisons show that highly educated women return to paid employment more quickly after birth than the low-educated, but they also demonstrate that educational differences are weaker in countries where policies are more generous (Gustafsson et al., 1996; Gutierrez-Domenech, 2005). Single country studies yield contrasting findings, however. The study by Lallive and Zweimueller (2009) shows that the effects of leave prolongation is stronger for highly educated women, but only in the short run, with no differences in the long run. By contrast, Aassve and Lappegard (2010) reveal that highly educated mothers are considerably less likely to take the cash-for-care benefits than those with only primary education because a longer break has stronger effect on the career of better educated mothers. Finally, Pronzato (2009) demonstrates that parental leave payments lower the hazard of returning to work most strongly for medium-educated women, with no significant differences between high- and low-educated women.
II – Hungary and Poland: background
14Given the inconsistency in previous findings, this article seeks to provide new empirical evidence of the effects of parental leave entitlements on fertility and women’s employment entry or re-entry. To this end we compare Hungary and Poland, which differ substantially in parental leave entitlements, while displaying similarities in socio-political and cultural terms. We first describe the parental leave regulations in the two countries and then present other characteristics that are potentially relevant for employment and childbearing decisions of women who have already given birth to one child. Our description of the regulatory frameworks covers the period beginning in the early 1990s and ending in the mid-2000s (the last years covered by our empirical analyses). More recent changes in parental leave entitlements are not considered.
1 – Parental leave policies
15An important feature of the Hungarian parental leave scheme is its universality (Korintus, 2010; Szelewa, 2012; Velluti, 2014): leave and benefits are granted to all parents, irrespective of their previous working history. However, the level of the cash benefit depends on whether or not the parent worked before the birth and on the wage earned. Women who worked for at least 365 days in the two years preceding the birth are “insured”, and are entitled to an earnings-related benefit amounting to 70% of their earnings over the previous two-year period. There is no ceiling on the payment for the first 24 weeks after birth (maternity benefit). Beyond 24 weeks, a ceiling was set at the level of the minimum wage in the 1990s and was doubled in 2000 (Table 1). The leave can be prolonged until the child’s third birthday, but in the third year the cash benefit becomes a flat-rate payment equal to the minimum old-age pension. Hungarian mothers who worked less than 365 days in the two years preceding the birth are entitled to a cash benefit equal to the flat-rate payment which insured mothers can claim in the third year of their leave. It is paid until the child’s third birthday. An important feature of the Hungarian leave system is also the fact that it creates an insurance relationship. This means that women on leave build up eligibility for subsequent parental leave if they give birth to another child during the leave period; women who did not work before the leave become entitled to health insurance and the time spent on leave counts for them as pensionable service. This two-tier parental leave system, for insured and non-insured women, had already been introduced under state socialism and remained largely unchanged over the 1990s and 2000s, with the exception of the years 1996-1999 when the earnings-related parental leave benefit was abolished and the flat-rate benefit became means-tested (Szelewa, 2012; Szikra and Szelewa, 2010).
16In terms of leave duration, the Polish parental leave system is just as generous as the Hungarian one since a woman can remain on a leave for around 3.5 years in total. During this time the woman’s job is secured and she is guaranteed to return to a position similar to the one she held before the birth. The parental leave payments are much more limited than in Hungary, however. First of all, only women who worked before giving birth are entitled to cash transfers (Velluti, 2014). This means that women who did not work before birth or whose work record is less than six months (we will call them non-insured) are not entitled to any parental leave payments in Poland, in contrast to the situation in Hungary where non-insured women are granted flat-rate benefits. Second, the cash benefit paid to insured women is related to their pre-birth earnings only for the first 16 weeks after first birth, which correspond to maternity leave. In this period, the compensation is even higher than in Hungary and amounts to 100% of the previous earnings. After this period, however, the benefit becomes a flat-rate payment (Table 1) and is paid exclusively to women whose per capita family income does not exceed a threshold which qualifies them for receiving social assistance. The flat-rate parental leave benefit is paid for two years, even though a woman can stay on leave for three years following the maternity leave. As in Hungary, women become eligible for parental leave payments for the subsequent child if they give birth while on leave. The Polish parental leave system was in place practically throughout the 1990s and 2000s. The only modifications concerned the duration of maternity leave payments (which were prolonged from 16 weeks to 26 weeks over the years 1999-2001, then shortened to 16 weeks again in 2002); the level of the flat-rate benefit; and the income criterion for benefit entitlement, which decreased slightly over time, mainly because of changes in the indexation scheme (Balcerzak et al., 2003; Kotowska et al., 2008).
Parental leave entitlements in Hungary and Poland in the mid-2000s
Parental leave entitlements in Hungary and Poland in the mid-2000s
2 – Other country-specific conditions relevant to fertility and employment choices
17Apart from differences in parental leave provision, Hungary and Poland share many similarities with respect to the other country-specific conditions which might be considered relevant for employment and further childbearing choices following first birth. These include childcare provision for the youngest children, the role of relatives in providing care, flexibility of working hours, gender role attitudes or attachment to family and work values.
18Childcare provision for the under-threes, which is crucial for a decision to (re-)enter employment after a first birth, is very poor in both countries. In Poland, only 2% of children aged under three were enrolled in public day-care in 2008, and 6% in Hungary (OECD, 2013). Furthermore, in both countries, the working hours are long and largely inflexible, while part-time jobs are scarce. The working week in both countries is slightly above 40 hours, versus 37 hours in EU-15. Part-time employment is not widespread either; since the collapse of the communist system, the proportion of part-time workers has not exceeded 6% in Hungary and 10% in Poland. Furthermore, studies show that individuals, including women, are not interested in shortening their work time (Matysiak, 2007; Hobson et al., 2011). Moreover, working hours in both countries are relatively rigid: only 14% of female employees work on flexible schedules (i.e. staggered working hours, flexitime or working time banking) versus at least 50% in countries such as Germany, Denmark or Sweden (Plantenga and Remery, 2009).
19Hungary and Poland, like many other post-socialist countries, have very traditional gender attitudes (Oláh, 2011; Hobson and Fahlén, 2009; Saxonberg and Sirovatka, 2006). The gender division of tasks is heavily asymmetric, and social disapproval of mothers who work when their children are young is widespread (Muszynska, 2007; Philipov, 2008; Blaskó, 2005). This traditional perception of women as major care providers is a paradox because women are expected to withdraw from the labour force when their children are very young, but then to resume employment when their children are slightly older in order to contribute to the household budget (Luck and Hoffacker, 2003; Treas and Widmer, 2000; Szalma, 2010).
20Finally, empirical studies show that Poles and Hungarians are strongly attached to family (Stankuniene and Maslauskaitė, 2008) and value children just as highly (Fokkema and Esveldt, 2008). Despite family-oriented values, women in both countries are also attached to paid employment: according to the data from the module on Work Orientations from the International Social Survey Programme (1997), around one in two women aged 18-54, agree with the statement that work is a person’s most important activity; this is one of the highest proportions in Europe, after Bulgaria, Slovakia and Portugal (authors’ own calculations). Under state socialism, men and women were forced to take up paid employment (Frejka, 2008), and the high value placed on work by women in Hungary and Poland remains a typical feature of post-socialist countries.
21Despite their numerous similarities, the two countries also display some differences, in addition to those related to parental leave programmes. Both countries underwent economic transformation in the 1990s, with a decline in GDP in the early 1990s and massive transformations in the labour market. However, compared with Hungary, the economic reforms were implemented more radically in Poland, thus resulting in higher social costs (Szikra and Szelewa, 2010). While unemployment increased sharply in both countries, in Hungary it peaked at 10.2% in 1995 and started to decline thereafter, while in Poland it rose continuously up to 20.2% in 2002, falling slightly in the following years. Furthermore, Hungarian family policy offers a much better buffer against poverty. The Hungarian system of family allowances, which was introduced in the 1940s, has remained universal, except for the short period 1995-1998 when means-testing was introduced. The amount of the benefit depends on the number of children and changed over time: in the mid-1990s the family allowance was around 13.5% of the national average wage, but fell to 7.8% in 2003 (Speder and Kamaras, 2008). Additionally, workers with more than two children enjoy tax deductions that offset the costs of raising children. These were available over the whole period under study, apart from the years 1995-1998. In Poland, means-testing of universal family allowances was introduced in 1995 (Balcerzak-Paradowska et al., 2003). Since then, only families in the highest need, whose income does not exceed the threshold qualifying for social assistance, have been eligible for the benefits. The level of benefits has varied over time and is much lower than in Hungary: it amounted to 2.6% of the national average wage in 1998 and 1.9% in 2003 (author’s calculations).
3 – Developments in fertility and women’s employment
22Hungary and Poland share the experience of communism, and both were exposed to the ideology of full-time employment and the forced integration of women, including mothers, into the labour market. Remarkably, the two countries were exceptionally resistant to this ideology, and women’s employment rates at that time were the lowest of the whole region. At the end of the 1980s, the employment rate of women aged 20-44 was 75.3% in Hungary and 76% in Poland, while in the majority of other socialist countries it exceeded 90% (authors’ calculations on data from ILO (2008)). After the collapse of state socialism, Hungary and Poland maintained their status as countries with relatively low female labour force participation with respect to other countries in the region, with a rate oscillating around 60% in Hungary and between 56% and 67% in Poland over the study period.
23Both Hungary and Poland are also low fertility countries, with a total fertility rate (TFR) of 1.3-1.4 in the 2000s. Both experienced strong declines in first- and second-order fertility in the 1990s and early 2000s, with the first-order TFR falling from 0.81 (Hungary) and 0.85 (Poland) in 1990 to slightly less than 0.6 in the early 2000s in both countries, and second-order TFR dropping from 0.7 to somewhat less than 0.4 in the same period (Figure 1A). The years 2003-2007 brought only small improvements in fertility. In parallel to changes in fertility, mean age at first and second birth also increased in both countries (Figure 1B). The mean age at first birth increased particularly strongly in Hungary and moved closer to the mean age at second birth.
First- and second-order total fertility rate, Hungary and Poland, 1989-2007
First- and second-order total fertility rate, Hungary and Poland, 1989-2007
Mean age at first and second birth, Hungary and Poland 1989-2007
Mean age at first and second birth, Hungary and Poland 1989-2007
III – Research hypotheses
24Given the differences in parental leave policies in Hungary and Poland, we expect to find clear differences in the two countries in the way these policies shape women’s behaviours after first birth, namely the transition to second birth and employment (re)entry.
25First, we anticipate that Polish women on leave will experience stronger financial pressure due to the fact that the wage compensation during parental leave in Poland is very modest, and limited to certain social groups. This pressure may discourage them from giving birth again while on leave. By contrast, more generous parental leave entitlements granted to insured women in Hungary should create favourable conditions for continued childbearing during leave. Given these considerations our Hypothesis 1 is:
H1 Long and paid parental leave speeds up transition to second birth
27Additionally, parental leave payments may have differential effects on different social groups due to the way in which benefit entitlements are calculated. In Hungary, where the level of parental leave benefit for the insured depends on previous earnings, it is more equally distributed across social groups than in Poland, where the highly educated are very unlikely to receive any financial compensation while on leave (apart from the first few months of maternity leave). Hence, provided that the low-educated women are able to make ends meet while on leave, one should expect educational differences in parental leave to have larger effects on second birth risks in Poland. Hypothesis 2 is thus:
H2 Non-universal, means-tested and low parental leave payments increase differences in the effect of parental leave on second birth risks across educational groups, facilitating the transition to second birth among the low-educated compared to the highly educated.
29Third, given that public childcare provision for youngest children is very scarce in both countries, but that the Hungarian parental leave system is far more generous, we expect Polish women to have higher opportunity costs of remaining on leave than Hungarian women. Consequently, Polish mothers are anticipated to enter employment after birth much more quickly than Hungarian mothers. This gives rise to Hypothesis 3:
H3 The lower the parental leave payments, the faster the employment entry among mothers.
31Fourth, although highly educated women should return to paid work more quickly than the low-educated in both countries, we expect educational disparities in the rates of return to employment to differ between the two countries. More specifically, provided that the low-educated women are able to make ends meet while on leave, the educational disparities in the intensity of employment entry are anticipated to be greater among insured women in Poland than among their counterparts in Hungary, due to much larger differences in wage compensation during parental leave across social groups in Poland. We thus formulate Hypothesis 4:
IV – Data and method
33We used two longitudinal datasets to test our research hypotheses. For Hungary, we employed the first two waves of the Generations and Gender Survey (GGS). The GGS is a panel survey of a nationally representative sample of the resident population in Hungary aged 18-79, conducted by the Demographic Research Institute of the Hungarian Central Statistical Office. The first wave was conducted in 2001 and the second between November 2004 and June 2005. The Hungarian GGS sample comprises 13,450 individuals and includes full fertility and employment histories of respondents since 1989, recorded on a monthly basis. For Poland, we used the Employment, Family and Education Survey (EFES) conducted in October and November 2006 on a representative sample of 3,000 women born in 1966-1981. The dataset contains women’s full education, employment and fertility histories since the age of 15, recorded on a monthly basis.
34For the purpose of our analyses, the sample was restricted to women born between 1971-1981 who had their first child before the survey date. These were the only cohorts for which we had full employment histories in the Hungarian dataset. The women in these cohorts all made their fertility and employment choices exclusively under the capitalist system. Our final sample thus covers 842 Hungarian women and 1,494 Polish women.
35Using this data, we created two monthly event history files for each individual, including the events of second conception and employment entry after first birth. The conception was measured nine months before birth. “Employment entry” signifies taking up a job, whether or not the woman was working before birth, and whether or not she made use of a parental leave after the birth. Each woman was observed from the first birth until the event of interest (employment entry or second conception) or until the survey date if it came first. Women who did not experience the event before the survey were censored.
36We then specified two main continuous time hazard models, namely a model for a second conception and a model for employment entry after first birth, with time since first birth as a process time in each of them. The baseline hazards are approximated using piecewise linear spline functions. Piecewise linear splines are usually applied to approximate continuous non-linear functions by fitting linear functions to data within each predefined interval. They were proposed by Lillard (1993) for modelling hazards and have been used quite extensively since then in demographic studies (Brien et al., 1999; Le Goff, 2002; Baizan et al., 2004; Kulu and Vikat, 2007), as they offer very efficient and flexible ways of approximating patterns of duration dependence. Piecewise linear splines are formally vectors of n+1 spline variables whose slope coefficients vary across intervals separated by n nodes. The nodes separating the time intervals in our study were set in a way that ensures the best possible fit to the data. This means that the nodes may differ across the two countries.
37In order to test Hypotheses H1 and H2, two specifications of the hazard model of second birth were proposed. We refer to them as Model B1 and Model B2. Model B1 aims at testing Hypothesis H1 on the role of paid parental leave for the transition to second birth. The main explanatory variable in this model is thus woman’s labour market status, which is time-varying. This variable assumes five categories in Hungary: working, on parental leave and insured, on parental leave and non-insured, [2] unemployed, and inactive (non-employed and not on leave). In Poland four categories are distinguished since a person cannot be on parental leave when non-insured. H1 is tested by comparing the effect of being on parental leave among insured persons in the two countries as well as the effect of being on parental leave among the non-insured in Hungary and the effect of being non-employed (unemployed and inactive) in Poland. We do not compare the effect of being on parental leave among the non-insured in Hungary with the effect of being on parental leave in Poland, as the situation of the two groups is completely different: by contrast to the former, the latter have a job they can return to after the leave. Similarly, we do not compare the fertility behaviours of non-insured and insured women in Hungary as the two groups may differ in their childbearing behaviours with respect to factors we are unable to measure (e.g. work orientation or family orientation). Our findings are controlled for a set of time-constant and time-varying woman’s characteristics. Among the time-constant characteristics we included age at first birth to control for a woman’s biological ability to conceive the second child, and parents’ education level to capture her social background. The time-varying covariates cover time since first birth and woman’s educational level. We distinguish three educational levels: high for women with a university degree, medium for women with post-secondary, general secondary or vocational secondary education, and low for women who completed basic compulsory or basic vocational education. [3] For the parents’ education, the high and medium levels are merged together due to the small number of births to women with highly educated parents. The distribution of person-months and events by women’s characteristics is presented in Table 2.
Distribution of person-months and occurrences by the explanatory variables and control covariates included in the study
Distribution of person-months and occurrences by the explanatory variables and control covariates included in the study
38Alongside Model B1 we also proposed Model B2 which aims at testing the Hypothesis H2 on the educational differences in the effects of parental leave on second birth risk. This model differs from B1 by the fact that the woman’s labour market status variable is interacted with woman’s educational level.
39Hypotheses H3 and H4 are evaluated using the hazard model of employment entry after first birth. Three model specifications are proposed: Models E1, E2 and E3. In Model E1 the baseline hazard of employment entry, which is the function of time since first birth, is interacted with the labour market status of a woman. In this model a woman can be on parental leave (again two categories are distinguished in Hungary to differentiate between the insured and non-insured mothers), she can be unemployed or inactive. In Hungary, during the first three years all mothers are on leave (unless they decide to enter employment) and they become unemployed or inactive after the leave expires, provided that they do not have another child or take up a job. In Poland, a mother is on leave only if she worked before birth for at least six months. The interaction between the baseline hazard of employment entry and mother’s labour force status thus allows us to approximate the hazard of employment entry separately for women on leave and for women who are not employed and are not on leave. With Model E1 we thus test the Hypothesis H3 on the role of parental leave entitlements for women’s employment entry in two steps. First, we compare patterns of employment entry in Hungary and Poland among insured women who are on leave. Second, we juxtapose patterns of employment entry of non-insured women on leave in Hungary with those of unemployed and inactive women in Poland. We do not compare the Hungarian women who are on leave but are non-insured with Polish women who are on leave since the labour market situations of the two groups are substantially different. For the same reasons, and as we cannot control for unobserved characteristics, we do not compare insured women’s risks of entering employment with those of non-insured women. Our findings are controlled for a set of woman’s time-constant and time-varying characteristics. Among time-constant characteristics we introduce work experience acquired by the time of first birth, and the time-varying characteristics cover woman’s educational attainment and calendar year. The distribution of person-months and events by women’s characteristics is presented in Table 2.
40In the next step, the baseline hazard in Model E1 is additionally interacted with woman’s educational level (Model E2) to address our fourth research hypothesis on educational differences in the rates of employment entry (H4). Finally, we introduced a control for the birth and age of a second child to our model E1 (Model E3). Taking these variables into account allows us to better understand our findings on the patterns of women’s employment entry after first birth in the two countries yielded by Model E1. Namely, we can see whether, and to what extent, the pattern of employment entry we obtained from Model E1 is affected by the birth of a second child while on leave. The age of the second child is introduced with the use of the conditional linear spline which starts operating at the conception of the second child and captures the effect of the time elapsed since then on a woman’s risk of employment entry.
V – Results
1 – Transition to second birth
41It is clear from our data that Hungarian women decide to have their second child while on parental leave with the first child. Hungarian women who are on leave and who are insured are 70% more likely to conceive the second child than women who have already entered employment after a first birth (Model B1 in Table 3). Similarly, a high second conception risk is observed for non-insured Hungarian women on leave, namely women who did not have a job before they gave birth. Non-employed Hungarian women are just as likely to conceive a second child as working ones. In Poland, no significant relationship between parental leave entitlements and second conception risk is observed. Polish women on parental leave are as likely to have their second child as those who work. The same holds for women who are inactive. Only unemployed women in Poland are less likely to progress to the second child than working ones.
Effects of women’s labour market status on second conception risks in Hungary and Poland, general and by woman’s education level (relative risks)a)
Effects of women’s labour market status on second conception risks in Hungary and Poland, general and by woman’s education level (relative risks)a)
a) The interaction effect between unemployment and high education could not be estimated because no events (second births) were observed among highly educated and unemployed women in the Hungarian sample.Significance levels: * < 0.1, ** < 0.05, *** < 0.01.
42These findings highlight two clear differences in the behaviours of the Hungarian and Polish women which speak in favour of hypothesis H1 whereby paid parental leave encourages the transition to second birth. First, being insured and on parental leave encourages further childbearing in Hungary but not in Poland. Given that insured mothers in Hungary receive much more generous parental leave payments than most mothers in Poland, this finding is consistent with Hypothesis H1. Second, parental leave payments to non-insured women without a job (the case of Hungary) also seem to create favourable conditions for childbearing. While these payments are lower than those received by insured women (for most of the leave), we also interpret this finding in favour of our Hypothesis 1, because we find that non-insured women in Poland, who are not entitled to parental leave payments, do not have a lower second conception risk.
43Model B2 does not reveal any significant educational differences in second birth risks among women on parental leave relative to working women in either of the analysed countries. The interaction coefficients are insignificant and the Akaike information criterion (AIC) suggests that Model B1 should be preferred over Model B2. Hence, our data leads us to reject the hypothesis H2 whereby means-testing increases social disparities in the effects of parental leave on the transition to second birth. It seems that the role of parental leave for the transition to second child does not differ by educational groups. However, it is also possible that the number of observations at our disposal is too low to yield significant results. Further research on larger datasets and longer fertility histories of the youngest cohorts are needed to better investigate educational and labour market status differences in second birth rates.
2 – Employment entry after first birth
44Hungarian and Polish women also differ substantially in the patterns of employment entry after first birth. These differences are visible in Figure 2, which illustrates the log hazard rates of employment entry of women on parental leave (stemming from Model E1). It is clear that women in Hungary stay on the leave longer than women in Poland (Figure 2A). The risk of employment entry for the insured Hungarian women is very low and starts to increase only after the child’s second birthday, i.e. when the earnings-dependent benefit is replaced by the flat-rate one. Despite the fact the benefit is reduced when the child reaches age 2, the hazard of taking up a job does not peak before the child’s third birthday, i.e. when the benefit is withdrawn. A similar pattern of employment entry is observed for the non-insured women. Altogether these findings show that Hungarian women tend to make use of their full parental leave entitlements and enter employment after the leave ends. Some of them remain on leave beyond the statutory period and their intensity of resuming employment is even lower. As will be shown later, these women have their second child during the leave and hence become entitled to a further period of parental leave.
Employment entry after first birth by women’s labour market status, without controlling for the birth of the subsequent child (Model E1). Log hazard rates
Employment entry after first birth by women’s labour market status, without controlling for the birth of the subsequent child (Model E1). Log hazard rates
Note: Results are standardized for the woman’s educational level, her work experience and calendar period. The reference category is low-educated women with more than 6 years’ work experience observed between 1999 and 2004. Full estimates are presented in Appendix Tables A.2 and A.3. The absolute log hazards range from 0 to 6, corresponding to hazards ranging from 0.001 to 0.4 per month (0.012 to 5.0 per year).45In Poland, the intensity of entering employment after a birth differs strongly by women’s labour market status. Women who were employed before the birth largely return to their jobs within six months (Figure 2B), when the maternity leave with full wage compensation ends. The remaining women who were employed before birth and choose not to return to their jobs after the end of the maternity leave stay out of work for a longer time. Their intensity of entering employment one year after birth is relatively low and it grows steadily to reach its maximum 3.5 years after the birth, i.e. when the parental leave expires. The pattern of employment entry for the unemployed and inactive women is different from that of insured women. The risk of entering employment does not change much with the time elapsed since birth, but for the first two years it is clearly higher than the risk of entering employment for non-insured women in Hungary.
46In sum, comparing the patterns of employment entry of insured women in Hungary and Poland and those of non-insured women in the two countries we conclude that women in Poland take up employment much more quickly after the first birth than women in Hungary. As all Hungarian women obtain a certain level of financial compensation for the first three years after birth while Polish women usually are not paid (with an exception of insured women during the first six months of the leave and some specific cases of economic hardship) these findings are consistent with our Hypothesis H3 whereby better paid parental leave entitlements discourage employment entry after first birth.
What is shown by the analysis by educational level?
47In the next step we investigated educational differences in employment entry for women on leave (Figure 3). The AIC statistics for Models E1 and E2 suggest that the model fit improves for both countries once an interaction between women’s labour force status and education is introduced into the model. In fact, in both countries substantial differences in the patterns of employment entry are observed across educational groups. In Hungary, these differences are most visible for insured mothers within the first two years after birth. Even though the intensity of entering employment peaks three years after the birth for all insured mothers, regardless of their educational level, the highly educated insured mothers are clearly most likely and the low-educated clearly least likely to enter employment in the first two years (Figure 3A). The differences by education between the non-insured mothers are less pronounced and tend to emerge after the third year of the child’s life (Figure 3B). [4]
Employment entry after first birth by women’s labour force status and educational level in Hungary and Poland, without controlling for the birth of the subsequent child (Model E2). Log hazard rates
Employment entry after first birth by women’s labour force status and educational level in Hungary and Poland, without controlling for the birth of the subsequent child (Model E2). Log hazard rates
Note: Results are standardized for calendar period, woman’s employment status at first conception, her current employment status (inactive, unemployed or on parental leave) and her work experience. The reference category is low-educated women with more than 6 years’ work experience observed between 1999 and 2004. Full estimates are available from the authors on request.48In Poland, the educational differences in employment entry are present for women on leave as well as for women not entitled to formal leave (unemployed and inactive). Regardless of labour force status, the highly educated display a higher intensity of entering employment than the low-educated (Figures 3C, 3D, 3E). These differences are most pronounced among mothers on leave: in this group mothers who did not enter employment within six months of the first birth are still quite likely to do so until the child is one year old, whereas such behaviour is very unlikely among the medium and low-educated women on leave. Altogether, however, our findings suggest that educational differences in employment entry emerge to a similar extent in both Hungary and Poland and lead us to reject our Hypothesis H4 which predicts that non-universal and means-tested parental leave entitlements amplify differences in the rates of employment entry across social strata.
Second birth during parental leave?
49Our findings presented so far suggest that Hungarian women tend to stay much longer on parental leave than Polish women. Furthermore, Figure 2 also shows that some women tend to stay on leave for even longer than the three-year statutory leave period. Figures 4A and 4B shed more light on these findings. They present patterns of employment entry after first birth for women on leave obtained from our last model E3 (green lines) in comparison to model E1 (black lines). The two models differ in only one respect, namely model E3 controls for the fact of giving birth to the second child and the child’s age. We can see that Model E3 yields substantially higher hazards of employment entry around the end of the parental leave for all groups of women, the insured and non-insured mothers in Hungary as well as the insured mothers in Poland. This result suggests that women who use their full parental leave entitlements tend to conceive the second child while on leave. It also complements our story: better paid parental leave in Hungary than in Poland means that Hungarian women can afford to use their parental leave entitlements fully, to postpone employment entry and conceive the second child before resuming economic activity.
Employment entry after first birth in Hungary and Poland, after controlling for the birth of the second child (Model E3). Log hazard rates
Employment entry after first birth in Hungary and Poland, after controlling for the birth of the second child (Model E3). Log hazard rates
Note: All results are standardized for woman’s educational level, calendar period, her employment status at first conception, her current employment status (inactive, unemployed or on parental leave) and her work experience. The reference category is low-educated women with more than 6 years’ work experience observed between 1999 and 2004. Full result estimates are presented in the Appendix Tables A.2 and A.3.Conclusion
50This article aimed at evaluating the effects of parental leave entitlements on the progression to the second child and women’s employment entry after the first birth. In particular, we investigated how universal parental leave payments affect these two events in comparison to poorly paid means-tested benefits. We also examined whether means-testing leads to an increase in socioeconomic differences in fertility and labour market behaviours of one-child mothers. To this end, we followed the approach proposed by Neyer and Andersson (2008) for investigating policy effects on individual behaviours. Namely, we compared Hungary and Poland in the period from the fall of communism up until the mid-2000s. These two countries share many similarities in their institutional, cultural and economic frameworks, but differ in parental leave regulations, with Hungary granting universal entitlements to parental leave payments and parental leave benefits which are strongly tied to pre-birth earnings, while Poland offers flat-rate benefits only to parents in the greatest financial need.
51Consistent with our expectations, we found that well-paid and universal parental leave encourages progression to a second child, but leads to substantial delays in women’s entry into employment. We base the first conclusion on the finding that Hungarian women who are on leave are more likely to conceive the second child than working women; a relationship which has not been found for Poland. The second conclusion is drawn upon the observation that Polish women display a much higher propensity to enter employment shortly after birth than Hungarian women who, for their part, tend to use their full parental leave entitlements. Note that the difference between the two countries in the rates of employment entry cannot be due to better childcare provision in Poland. Indeed, formal childcare facilities for the youngest children (below age 3) are very scarce in both countries, and women who resume employment usually have to make their own childcare arrangements.
52However, our findings do not support the hypotheses whereby means-tested and non-universal parental leave entitlements increase social disparities in the timing of second births as well as the intensity of entering employment. We did not find any educational differences in the effects of being on parental leave on second birth risks in either of the analysed countries. We did, however, find clear educational differences in the timing of entry to employment in both countries, with highly educated women entering employment much more quickly than the low-educated women, regardless of their labour force status. These differences were evident in both countries.
53Overall, this study contributes to the literature on the effects of parental leave payments on fertility and women’s employment, which has produced somewhat conflicting findings. Furthermore, the empirical studies conducted so far have usually focused on single countries and looked at how country-specific regulations affected women’s behaviours. In this article, we have looked, instead, at how certain mechanisms built into the parental leave systems influence fertility and employment entry.
54The present study also has some shortcomings, however. First, we do not control for selection of women into parental leave programmes due to unobserved characteristics such as family orientation. Second, even though we selected two fairly similar countries for the comparison, we cannot exclude that there are also some other differences between them, unrelated to parental leave entitlements, which moderate the effects of parental leave on second conception risks or the risks of employment entry. In fact, not only the parental leave payments but also the whole system of family allowances in Hungary is more generous than in Poland and it might enable women on leave to stay out of paid employment for longer in order to have a second child. At the same time, the pressure of unemployment was higher in Poland than in Hungary in the analysed period and it might have speeded up women’s employment entry. In our view, however, it is unlikely that the direction of the effects of parental leave we have found would have changed if we had been able to account for the differences in family allowances and unemployment in the two countries. The findings are very clear-cut, in particular with regard to the timing of employment entry, which corresponds very closely with the withdrawal of parental leave payments. However, the magnitude of the effects of parental leave payments on fertility and women’s employment may plausibly be reinforced by differences in family allowances and unemployment pressure between the two countries. All in all, there is no doubt that in the two countries under study, parental leave entitlements affect women’s decisions to have a subsequent child during the leave period and to enter employment after they give birth to another child.
55Finally, one may wonder about the long-term consequences of differences in parental leave regulations on fertility in the two countries. Does the generous parental leave system in Hungary encourage more women to have a second child than the means-tested and low paid system in Poland, or does it only accelerate the second birth? On the basis of our findings, we cannot provide a firm answer to this question. One can hypothesize, however, that the differences in parental leave payments mainly affect the timing of second births with no long-term consequences for completed fertility. This hypothesis is supported by the observation that there are hardly any differences in second-order total fertility rates in the two countries, but that the difference between mean age at first birth and mean age at second birth is one year greater in Poland than in Hungary.
Acknowledgements
This article has been published thanks to the European research project RECWOWE (Reconciling Work and Welfare in Europe), 2006-2011, co-funded by the European Commission, under the 6th Framework Programme for Research - Socio-economic Sciences and Humanities (contract no. 028339-2) of the Directorate-General for Research.Transition to second birth in Hungary and Poland, full model estimates, coefficients (standard deviation)(a)
Transition to second birth in Hungary and Poland, full model estimates, coefficients (standard deviation)
Entry to employment in Hungary, full model estimates, coefficients (standard deviation)
Entry to employment in Poland, full model estimates, coefficients (standard deviation)
Notes
-
[*]
Wittgenstein Centre, Vienna Institute of Demography, Austrian Academy of Sciences, Austria; Institute of Statistics and Demography, Warsaw School of Economics, Poland.
-
[**]
Swiss Centre of Expertise in the Social Sciences (FORS), Switzerland.
-
[1]
For example, for the effects on women’s employment see Evertsson and Duvander, 2011; Berger and Waldfogel, 2004; Ronsen 2009; Lalive and Zweimüller, 2009; Lalive et al. 2011. For the effects on women’s fertility see Lalive and Zweimüller, 2009; St’astna and Sobotka, 2009; Lappegard, 2008; Lappegard and Aassve, 2010).
-
[2]
Insured women are women who worked at least 365 days in the two years preceding birth. This information is not directly available in the data and was retrieved from the women’s employment histories.
-
[3]
In both countries at the time periods covered, it was compulsory to complete eight years of primary education and then to choose between basic vocational (2-4 years), secondary general (4-5 years) or secondary vocational education (4-5 years). Most people in the analysed cohorts completed one of the three post-primary education levels. However, those who graduate from basic vocational education have the most basic vocational skills and cannot continue education at the university level without first completing secondary general or secondary vocational education. We therefore categorized them as low-educated.
-
[4]
We do not display the pattern of employment entry for the highly educated non-insured mothers as the precision of the estimates is very low due to the small number of mothers in this group.
