THE IMPACT OF OCCUPATIONAL SEX-COMPOSITION ON WOMEN'S FERTILITY IN SPAIN

In parallel to a trend towards a progressive postponement of major family events from the 1970s until now, Spain has witnessed an important growth in education and women have entered into the labour market in record numbers. The greatest increase in female labour force participation occurred between 1981 and 1991; much later than in the other European countries and has been the result of the progressive entry of younger cohorts into the labour market as they reached working age. The so-called ‘baby boom’ and ‘baby bust’ generations (those born from the mid-1960s until the mid-1970s) were successfully involved in the profound changes undergone by the Spanish educational system during the 1980s, aimed at promoting both secondary and university studies (Boyd-Barret and O'Malley 1995).

These female cohorts stay longer in the educational system than women of previous generations and are those with more career opportunities in recent Spain. Nowadays, the proportion of women with a university degree at certain ages has even surpassed that of men. In addition, 77.2 percent of Spanish women aged 25–34 are economically active, a proportion that is slightly above the EU-25 average for this age group (75.7 percent). It has also become more common for young married women, including those with young children, to participate in the labour market. Previous studies show that women with a young child aged 0–2 still reach relatively high participation rates in Spain (68 percent), but ‘the rate decreases steadily with the number of children and particularly after the third child when the youngest is aged 0–2’ (González 2006: 196). Moreover, the proportion of less educated women in the labour market is low from a comparative viewpoint. In Southern Europe, only 20/25 percent of less educated women are in the labour market (40/45 percent in the other advanced societies), and this figure decreases dramatically if they are less educated and also have children (15 percent) (OECD 2001c: Table 4.I, cited in Esping-Andersen et al.2002: 30, footnotes 12 and 44).

Therefore, the increasing levels of educational attainment and labour participation among the youngest female generations have not eradicated gender differences in the attachment to the labour market at childbirth and/or when there is a presence of children in the family, particularly for the less educated women. Sex roles and gender relations are changing in Spain, but very slowly. The domestic division of labour is relatively traditional and it is still the mother who accounts for most of the opportunity costs of childbearing (Tobío 2005). Neither has the occupational-sex composition in labour market disappeared. Men and women continue to choose different fields of education and occupy gender-typical employments later on (Bradley 2000), which may have significant consequences on their labour integration and demographic behaviour. Most occupational groups nowadays have higher female representation because women are more evenly distributed across occupations than in the past, but occupational segregation persists because the sectors in which women predominantly work are different from those in which men are concentrated. Recent data show that in EU-25 as a whole, ‘women make up almost 80 percent of those employed in health and social work, over 70 percent of those employed in education and over 60 percent of those working in retailing’ (Eurostat 2008: 56).

Research on the effect of sex segregation on labour market outcomes (the gender wage gap, occupational mobility, etc.), has been very prolific both within sociology and economics (Polavieja 2005). However, the extent to which occupational sex-segregation impacts on women's demographic behaviour is poorly explored, at least in Spain. In order to fulfil this gap, this paper tries to give an answer to the specific research question whether there are differences among women regarding their demographic behaviour according to their occupational category. My main purpose is to examine in how far occupational gender segregation (and particularly the selection of typical female occupations) affects the transition to first, second and third births in Spain.

It is a well-stylized fact that there is not a straight linear negative relationship between women's employment and fertility. Previous studies show that, at the macro level, the correlation between female participation rates and total fertility rates was negative and significant up to the early 1980s. However, by the late 1980s the correlation had become positive and equally significant (Ahn and Mira 2002). At the individual level, there generally exists a negative relationship between fertility and female employment, but there are important differences across countries. For the Scandinavian countries, for instance, some scholars demonstrate that this relationship is positive (Hoem 1993), while it is negative for the Southern European countries (Baizán 2005; Solera and Bettio 2007).

The human capital literature predicts that the higher the level of women's job opportunities, the higher the value of their time and therefore, the higher the opportunity cost of children (Becker 1981). This is because apart from the income which is forgone during the time spent by women out of the labour market, they are also likely to pay a higher price in terms of wage losses and career advancement (Blau et al.1992: 41). Hence, we will expect employed women to have a lower probability of becoming mothers and having a higher number of children than women out of the labour market (‘human capital hypothesis’ (H₁)). However, as noted, research demonstrates that the negative effect of female employment on motherhood is more evident in traditional family systems with strict gender-specific divisions of labour and substantial institutional constraints (Fagnani 2007). Mother-friendly policies, and particularly better day-care facilities, are shown to be essential in eradicating the incompatibility problem. In Spain, the institutional context imposes serious constraints to the combination of family and work. Paid maternity leave is relatively short and formal childcare is insufficient.

Labour market institutions in each specific institutional setting are also a key factor in explaining women's labour market participation and the opportunity cost of children. Unemployment rates, part-time employment and the availability of jobs in the public sector are viewed as some of these explaining factors. In principle, a reduction of fertility is expected to apply while unemployed because a birth could make it difficult for a woman to be ready for work and also because unemployment may reduce the resources available. However, in some cases, unemployed women have lower opportunity costs in comparison to employed women and as a result, unemployment may turn out to be a period to have children. The existing literature demonstrates that the effects on fertility differ across countries according to its duration, the extent of the unemployment benefits, the woman's human capital, and the partner's characteristics (Baizán 2005). In Spain, female unemployment increased to record numbers during the eighties. In 1980, the female unemployment rate was 12.8. By 1985, it had raised to 25.1. From that time onwards, it remained at high levels (30.6 in 1996) (INE 2003). Unemployment benefits in Spain are low from a comparative viewpoint. Hence, we will expect unemployed women to postpone and lower fertility due to their uncertain situation in the Spanish labour market (‘uncertainty hypothesis’ (H₂)).

In contrast, part-time jobs are expected to influence positively fertility by decreasing the work/family conflict (‘conciliation hypothesis’ (H₃)). Countries differ considerably regarding the quantity and the quality of part-time employment but existing research confirms that part-time jobs permit an easier re-entry in the labour market after a more or less long interruption when a/another child is born. The availability of part-time work is rare in Spain (less than 16 percent of employed women aged 25–54 in the 1990s). Working women mostly work full-time with higher working hours in Spain than in other European countries (39.5 vs. 38.8 h (EU-25 average)). In addition, part-time employment is characterized by two dimensions in Spain. First, part-time jobs mostly offer low level occupations requiring little in terms of qualifications and offering limited chances of upward mobility. Secondly, a high proportion of part-time work is only temporary employment and, therefore, is associated to instability and precariousness.¹

Finally, differences among women employed in the public vs. the private sector are also pointed out in previous studies as important determining factors of employed women's fertility due to the different degrees of stability and security that they imply. In the public sector, good employment conditions with respect to pay, security, and time flexibility are shown to encourage fertility (Solera and Bettio 2007). Research already carried out has shown that female public jobs and employment rates in ‘women-friendly jobs’ accounts for a much smaller figure in Spain than in other societies (Esping-Andersen et al.2002: 76–7, Table 3.2). Unfortunately, the FFS data do not permit us to test here whether the public administration is the main sector in which women find it easier to combine family and employment responsibilities. A proxy will be suggested in the next section to grasp this issue from a different angle.

It is clear that conditions and desires towards motherhood have changed over the past decades. Women face nowadays a wider spectrum of choices in all spheres of life, and value more their time, their individual/economic independence and their self-fulfilment. However, in any given cultural environment, women constitute a heterogeneous group in assimilating and accepting these new values, and each woman displays her own fertility and work preferences (Hakim 2003). Economists usually assume that women prioritize family activities and responsibilities because they are those who give birth. However, Hakim sets out three preference groups, as ideal types. She argues that nowadays few women are ‘uncommitted’ to the labour market or maintain the traditional housewife role. Fully ‘committed’ or work-centred women are also very minoritarian. Hence, she says, the vast majority of women are ‘adaptive’ and maintains a dual-role, that is, choose to combine motherhood with paid employment. In spite of this being the general norm in all modern societies, Hakim has also recently accepted that there may be ‘social constraints and contextual influences that help to determine the relative sizes of the three groups in any particular social setting’ (Hakim 2000: 4). For instance, Spanish adults show one of the highest levels of support in the EU for the ‘egalitarian’ model of symmetrical roles in the family. However, a quite different picture is viewed with regard to the actual size of the above-cited three groups, simply because the proportion of home-centred women is still higher in Spain than in other societies (Hakim 2000, 2003).

Previous studies have successfully emphasized this idea that employed women cannot be treated as a homogenous group (Bernardi 1998). Numerous empirical evidence also demonstrates that women ‘choose’ as Hakim argues, but they are also constrained in their choices (González 2006). The reconciliation issue is crucial. This means that each woman displays a varied range of responses to the structural dilemmas facing all women (Gerson 1985: 11), that is, there are individual effects that stem from heterogeneity in women's personal preferences and constraints regarding family formation and the entry into motherhood. Therefore, we should reconsider the general assumptions proposed by the microeconomic theory and Hakim's preference theory. The working hypothesis proposed here is that family-orientation and work-orientation are not necessarily opposites.

Some women may have ‘high aspirations for both a family and a work career’ (Lappegård 2006: 65), so that they will attempt to ‘reach a reasonably high level of both employment and family life goals without maximizing one to the substantial exclusion of the other’ (Chafetz and Hagan 1996: 201). In this respect, occupations will be an additional important explanatory variable for the fertility behavior of women since a choice of a specific type of employment may be subject to individual preferences about a desired life style, in which women show a particular orientation towards family life and motherhood, irrespective of their employment career. Recent studies cast doubt on the assumption that a woman's accumulation of human capital and labour participation per se must result in lower fertility. In fact, the inclusion of the field of study has proved to be both theoretically and empirically relevant for women's fertility in Spain (Martín-García and Baizán 2006). An increasing number of studies have also recently shown that the type of education is equally significant for women as one of the most important determinants of first birth timing, as well as for overall fertility levels in Norway and Sweden (Hoem et al.2006; Lappegård 2006). All these studies show a positive relationship between academic studies concerned with the care of individuals and/or which emphasize interpersonal skills and fertility decisions, irrespective of women's educational level.

At an individual level, therefore, would the picture be different regarding occupational choice and their effect on fertility? Existing literature points out that the aspirations, attitudes and behaviour of the group of women who choose certain fields of study during education, and who decide on so-called ‘mother-friendly employment’ later in life, may be in sharp contrast to the aspirations, attitudes and behaviour of other female groups (Jonsson 1999: 394). Thus, it seems reasonable to predict an interrelationship between gender-typical occupation decisions and anticipated future family responsibilities. In other words, ‘a young woman with strong family preferences may pursue education and even a career, but she is most likely to select herself into the kinds of studies and jobs that are most easily compatible with motherhood’ (Esping-Andersen et al.2007: 27).

Undoubtedly, the woman's knowledge of labour market returns and outcomes explain sex-typical choices. ‘In some fields, as in science and engineering, technological change progresses rapidly. A woman returning from a labour force interruption will not only have to contend with her depreciation of skills over the interim but also with the advancement of the field during her absence. On the other hand, in such other fields as teaching, the pace of technological progress is slower. A woman returning from a work force interruption is likely to find that her earnings fall less steeply. Women anticipating traditional roles are, therefore, expected to avoid fields where the rate of technological change is rapid and to concentrate in fields where the cost of work force interruptions is lower’ (Blau et al.1992: 196). Many women anticipate their role as worker and mother, and thus choose an occupation that increases the productivity in both spheres.

In addition, ‘male-dominated occupations typically pay comparatively higher wages, but may have time-demanding work norms or a male-dominated work environment, such as standard and longer working hours, less consideration for employees’ needs to give care, and often a working culture dominated by “masculinity”’ (Hoem et al.2006: 339). In this respect, it is expected that labour-market environments with a higher proportion of female employees provide more concerned and accustomed employers to the needs of working parents with children. Women employed in teaching and health professions, traditionally viewed as feminine and with more favourable employment conditions, may then present higher fertility (‘mother-friendly occupation hypothesis’ (H₄)).

The data used here is taken from the Spanish Fertility and Family Survey, a retrospective survey conducted from November 1994 to October 1995. This survey uses a monthly time scale and provides individual-level data on family dynamics for the birth cohorts born between 1945 and 1977, as well as their employment and educational histories and some social background characteristics. The sample design and general results are fully described and commented on in Delgado and Castro (1999). The survey covered a total of 4,021 women. Our sample includes women who did not have children at age 15. The dependent variable is taken as the time of birth minus 9 months, in order to measure as closely as possible the moment when the decision to have a child was taken and to avoid changes that may occur between conception and birth, such as the woman's departure from the labour market. Observation begins at the age of 15 (analysis of first birth) or when the first/second child was born (second and third birth) and ends with the conception of the first, second and third birth, respectively, or for right-censored cases, with the date of the interview.

The main independent variable in this study is the woman's type of employment. In order to construct this variable, we first make use of the traditional Erikson and Goldthorpe's (1987) class classification based on occupational status. Our variable consists of six categories since ‘farmers’ and ‘farmer workers’ are herein grouped in a single category due to small sample sizes. In a second step, we differentiate those women who work as ‘health and teaching professionals’ in the service class. All other professionals are grouped together in the category ‘others’. Table 1 presents in detail the classification of occupational categories applied in the analysis and the distribution of women according to their occupational category in the FFS samples.

Another important information regarding the woman's labour market situation is whether the woman is unemployed and for employed women, the number of hours worked per week (part-time vs. full time employment). Women who did any work for pay but are currently jobless, looking for jobs or not, are considered unemployed in the analysis. Moreover, a woman is in part-time employment when she reports less than 26 hours of work per week (EU Labour Force Sample Definition). The FFS supplies also full histories of educational enrolment that includes dates of attainment for each level of education. The educational enrolment variable reflects whether the woman is in or outside of education. The woman's level of education is classified in three levels: primary and lower secondary education (ISCED levels 0–2), upper secondary education (level 3), and university education (levels 5a/5b–6) (EURYDICE 2008).² Finally, a number of demographic and contextual control variables are included to interpret these variables related to the labour market. All the piecewise linear models applied include age as the baseline. Moreover, we include four birth cohorts: 1945–1954, 1955–1959, 1960–1964, and 1965–1977. Models are also controlled by the number of siblings, the place of residence up to age of 15 and the woman's partnership status.

Models 1, 2 and 3 are aimed at empirically analyzing the four hypotheses included in the theoretical section. Findings are presented as follows. First, in Model 1 we see the effect of the woman's employment status on the risk of having the first, second and third birth, with a special focus on unemployment and on whether the woman has a part-time or full-time job (Table 2). Secondly, we investigate in Model 2 whether the impact of women's labour force status differs according to type of employment, following the traditional Erikson and Goldthorpe's classification based on occupational status (Table 3). Model 3 presents the results when the differentiation between ‘health and teaching professionals’ and ‘others’ within the ‘service class’ category is included (Table 4). All risks in the models are relative.

Results corroborate the sociological view that young women who participate in education are not at risk of childbearing. There is a strong negative significant effect of being enrolled in education when explaining the birth of the first child (0.31*** (Model 1); 0.37*** (Models 2 and 3)). Furthermore, Models 2 and 3 show that being out of the educational system increases the probability of having the first birth with respect to women who are still enrolled, but these women do not become mothers unless they are out of the educational system at least for 2 years or more. These results are consistent with the hypothesized prediction that, in a context of unstable work conditions with high unemployment rates and high temporary contracts, Spanish women try to consolidate their careers before even thinking of forming a family. Additionally, most women conclude their studies before the first child is born so the effect of being enrolled in education is somewhat smaller and not significant in the analyses of the second and third birth (0.90/0.89 in Models 2 and 3, respectively). Our data also suggest that the lower the level of education of a woman, the higher the probability of childbearing, in all models and for all birth orders, although the effects are not significant for university studies and the difference between middle and better-educated women does not imply a strong negative monotonic shape of the hazard (with the exception of the third birth).

Secondly, as predicted, labour force participation seems to result in later and lower fertility in Spain (human capital hypothesis (H₁)). Generally speaking, the ‘male breadwinner’ situation in which the woman stays at home seems to encourage fertility according to the results in Table 2. Taking ‘housewife status’ as the reference group, all the other woman's labour force categories show a lower propensity towards motherhood and smaller families. For instance, unemployment has a negative not significant effect in Spain for first (0.97), second (0.89) and third birth (0.90) so it is not seen as an opportunity to have children. As put forward above, unemployment seems to hamper rather than facilitate family formation in Spain and this result corroborates the uncertainty hypothesis (H₂). Data demonstrate that the real distinction is between housewives and women who are in the workforce, because those out of the labour market with previous experience (unemployed women), have a higher probability of childbearing in comparison to full-time and part-time workers (although not significant), but this propensity is still smaller than that of women out of the labour market without any experience (reference group).

Regarding the number of hours at work, we see that the relative risk of having a first child is lower in case of part-time vs. out of employment in comparison to full-time employment vs. out of employment for the first (0.63***/0.68***), the second (0.77*/0.78***) and the third births (0.65/0.80**, respectively, in Model 1). These results seem to indicate that part-time employment does not really matter in Spain and therefore the predicted reconciliation hypothesis (H₃) does not apply. However, the low number of individuals involved in calculating this covariate and the fact that part-time employment is rarely available and is associated with temporary and even precarious jobs in Spain means exercising caution when interpreting this particular result. Previous studies have already shown that there is a very strong and significant effect of temporary jobs on fertility in Spain and that working part time does not have a significant impact (Baizán 2005: 17).

As presented in the theoretical section, the results shown above with regard to the impact of a woman's labour force status should be complemented with the inclusion of the variable ‘woman's type of employment’. Being employed reduces the probability of having a/another child in comparison to women who continue to be defined entirely by their family role but the inclusion of women's type of employment qualifies this result (Model 2, Table 3). In the analysis of the first birth, skilled female workers are the only group with a positive not significant effect of entering into motherhood with respect to women who are out of the workforce (relative risk of 1.46). These women are more likely to have the first child, show a lower probability of having the second birth (0.42) but then, if born, they proceed rapidly to the third child (1.31). Data suggest that they have a high preference for children and they may remain in the labour market for economic reasons.

The probability of childbearing is low for employed women in the ‘petty bourgeoisie’ category. There is a negative significant effect for first, second and third birth (0.64***, 0.73*** and 0.68***, respectively). These women seem to face a strong trade-off between family and work due, partly, to the characteristics of the Spanish labour market over the past decades with respect to small proprietors with/without employees in terms of parental leave and family policy (‘autónomos’). The relative risks of becoming mothers and of having the second and third birth for routine non-manual female workers, the most numerous group, are 0.67***, 0.82*** and 0.77**, respectively. These routine non-manual works are also overwhelmingly female in composition in occupations such as secretarial and clerical work, retail sales, and private domestic employment. However, they do not seem to offer favourable possibilities for reducing children's costs at the workplace. The result, therefore, is a low propensity towards motherhood.

The most interesting findings pertain to the ‘service class’ category. For first births, employed women in the service class have a low probability of becoming mothers, although somewhat higher than for women in other categories (0.78**). Once the distinction is made, service class women in the category ‘health and teaching professionals’ show a higher propensity towards motherhood (0.86) with respect to women in ‘others’ (0.67**) (Model 3, Table 4). The variation is not high and the effect is not significant for the first group of women despite we expected that these women would have had diverse childbearing preferences and that working in environments with different gender distributions would have impacted more positively on their motherhood propensity. However, this distinction is higher for second and third birth (although only significant for health and teaching professionals in the higher birth order). The relative risk of being employed as a health or teaching professional on the second birth is 0.99 (0.41 for ‘others’). In contrast, there exists a positive significant effect for health and teaching professionals when having the third birth (1.98* vs. 0.74 for those in ‘others’) in comparison to women out of the labour market. A plausible explanation is that the most common way to access this type of jobs (often pursued in the public sector) is through various tests and examinations (oposición), and this fact means that it takes longer for women to get established in the labour market and have their careers well underway. Later in life, most women in the ‘health and teaching professionals’ category occupy the so-called ‘women–friendly jobs’, which offer more flexible hours and better opportunities to combine family and work (Martín-García and Baizán 2006: 265). These women, therefore, are in a better position to afford the opportunity costs of children.

Results in Model 3 seem to tell us that this group of employed women postpones motherhood but then they have a higher propensity of having a second and especially a third child. In other words, this specific group of employed women may postpone the entry into motherhood but then they ‘catch up’ and end up having the desired number of children. This confirms our expectations that women can show a particular orientation towards family life and motherhood, irrespective of their employment career, and this is captured by her specific type of employment (mother-friendly occupation hypothesis (H₄)). Women in the ‘health and teaching professionals’ category who already have one/two child/children are probably a selected group with higher preferences for children than other employed women. In addition, the stability and protection of these types of employment play a decisive role because work and family are more compatible here than in other employment areas. More flexible work hours and better conciliation options reinforce the positive effect.

Finally, the control variables have the expected effect. For all the birth orders, there is a strong negative impact on fertility for younger generations. Data demonstrate that for the 1965–1977 birth cohort, the higher the birth order, the stronger this negative effect (negative significant relative risks of 0.47*** and 0.49*** for second and third birth, respectively, in comparison to 0.75*** for first birth in Model 3). Results also show that growing up in a family with a relatively high number of siblings positively influences a woman's entry into motherhood and the number of children in her own family. An only child woman is shown to have a lower probability of having a second and a third child once she has become a mother (significant relative risks of 0.69*** and 0.57*** for second and third birth, respectively, in Model 3). Moreover, women in urban areas have a lower probability of becoming mothers (0.87*** in Model 3), which is consistent with the theory that a rural place of residence may impose higher obstacles in accessing higher education. However, this effect vanishes and losses significance for higher birth orders. Finally, our data show that by far the most important effect on the entry into motherhood is whether the woman is in a union (14.83*** in Model 3). Women in a union have also a higher propensity towards bigger families (relative risk of 1.66*** for third birth in Model 3).

Applying event history models to data from the Spanish Family and Fertility Survey, this paper answers the specific research question of whether all women forgo and/or delay motherhood as a result of their increasing participation in the labour market, or whether there are intra-women differences with regard to fertility according to the occupational category they occupy. The aim of the paper was to investigate how gender specific distribution across occupational categories in the labour market leads to women's differences in the transition to first, second and third birth in Spain. The results can be summarized as follows. First, they confirm the human capital approach that predicts that women's participation in the labour market postpones and reduces motherhood. Due to the scarcity of family policy and the gender-division of labour within the household in Spain, the time of the mother still accounts for the majority of the total costs. Moreover, part-time employment does not seem to really matter in Spain. There might rather be a problem of temporary contracts or unemployment. However, these models that exclusively focus on women's participation in the labour market do not fully capture their returns in terms of family attitudes, (income) and flexibility to harmonize work and family.

The findings in this paper demonstrate that women's type of employment serves as a better indicator of their reproductive behaviour than their mere participation in the labour market because it reflects individual preferences concerning motherhood, and also indicate different possibilities for resuming and maintaining employment after birth. Therefore, in order to get a complete picture on the connection between female labour force participation and fertility outcomes, studies need to take into account the different occupational categories. Once we distinguish these occupational choices, we see that women who participate in the labour market do not always have the lowest probability of becoming mothers and of having a higher number of children in Spain. Health and teaching professionals show a higher propensity of having a higher number of children with respect to other employed women and even to women out of the labour market. In fact, this group of women may be the most advantageous group of employed women in Spain concerning the compatibility of family and work. Hence, this article shows that the conflict between family and working roles does not occur with equal intensity across employed women. Our results alerts that there exists a risk of accumulating advantages and polarization among employed women and this is particularly true for a country such as Spain, where policies aimed at helping parents to combine paid work and unpaid work are scarce.

That said, we have to bear in mind that empirical results shown here should be taken with caution due to data limitations. Small samples, particularly in the analyses of the second and third birth, are reflected in large standard errors and hence insignificant effects. Larger data sets and longitudinal data with more detailed and updated measures are needed for the estimation of the causal effects between female participation in the labour market, occupational choices and fertility outcomes (data on public employment are a must). Additionally, in the present analysis of occupational variations in career intentions and children, we are lacking a decisive issue, that of men and the domestic division of labour.

This research is part of a project financed by the National Plan for Scientific Research, Development and Technological Innovation of the Ministry of Education and Science (Ref. SEJ2006-03485/SOCI). I would also like to thank the anonymous referees for their useful suggestions and comments. Finally, I gratefully acknowledge the Advisory Group of the FFS programme of comparative research for its permission to use the FFS data on Spain.

Teresa Martín García (PhD, EUI, Florence, and Doctor-Member of the Juan March Institute, Madrid). Research Fellow at the Institute of Economics, Geography and Demography, Depart. Population, Spanish Council for Scientific Research (CSIC), Madrid.