1This short paper presents a systematic comparison of Latin American studies on the relationship between women’s employment and fertility. Though the link between the two is often considered to be negative, in urban areas especially, several authors have drawn attention to the diverging results obtained in empirical studies (Davidson, 1977; Garcia and de Oliveira, 1989; Youssef, 1982). We will attempt to pinpoint the reasons for this divergence.
2We have chosen to focus exclusively on Latin America for two reasons. Firstly, the continent is relatively absent from recent analyses, and literature reviews of this question are scarce and generally rather old (Davidson, 1977; Youssef, 1982). Secondly, an analysis of the relationship between women’s employment and fertility in the case of Latin America might provide a starting point for new research in this area, given the particular sociocultural, economic and demographic profile of the region.
3In all Latin American countries, the labour force participation rate of women increased steadily in the second half of the twentieth century, rising from less than 20% in 1960 to more than 40% at the end of the century (Celade, 1999). Before the 1980s, the increase was linked mainly to industrial development and a substantial rise in school enrolment rates, which improved the situation and status of women, primarily in the most advantaged social categories (Espinosa, 1994). Women worked mainly as employees in the newly created industries and in the formal services sector (De La Luz, 1989; Suarez, 1989). From the 1980s onward, with the financial crisis, the introduction of economic restructuring policies and the onset of globalization, the situation began to change. Increasingly, the rise in female employment now reflects the survival strategy adopted by families looking for ways to raise their income. With ever fewer employment opportunities on the labour market, women mainly hold informal, low-productivity jobs in the services sector, often on a self-employed basis with no social protection (Garcia and de Oliveira, 1994). The maquiladora [1] industry is a second major source of female employment. It is based mainly in the formal sector, though working conditions are often appalling (Martinez, 1994; Chant, 1991). So the increase in women’s employment now concerns above all the disadvantaged social categories, without necessarily bringing an improvement in their situation or offering them financial independence.
4Until the late 1960s, Latin America had a regime of natural fertility which, associated with early and practically universal marriage, resulted in very high fertility, i.e. 6 to 7 children per woman (Cosio-Zavala, 1998; Guzman and Rodriguez, 1993). The 1970s marked the start of a spectacular decline in fertility to just above replacement level, with some countries even falling below the replacement threshold. This drop is almost entirely due to the practice of birth control by couples, rather than to a decline in the marriage rate or the postponement of marriage, whose influence remained marginal (Rosero-Bixby, 1996; Moreno and Singh, 1996). The decline in fertility has been accompanied by a lowering of the mean age at childbearing, with most births now occurring before age 35 (ECLAC, 1998). However, this general trend conceals major differences between regions and social groups. In rural regions and disadvantaged social categories, the decline in fertility occurred much later and remains smaller than for the country as a whole. Moreover, it has not been accompanied by an improvement in women’s situation or by a process of emancipation. In the disadvantaged social categories, it tends to follow a Malthusian process involving little normative change (Cosio-Zavala, 1992).
5To study the relationship between women’s employment and fertility in Latin America, we will begin by comparing the results obtained in the literature, according to the methodology and measures used, and making a distinction between cross-sectional and longitudinal studies. We will then examine the influence of the context, notably development level and fertility in the different countries, and changes in the relationship between women’s employment and fertility. Lastly, a discussion of the specific characteristics of this relationship and of the growing body of knowledge on this issue in Latin America will enable us to suggest new ways of approaching the question.
I – Comparison of methods
1 – Cross-sectional studies
6Cross-sectional studies are based on the current or most recent employment status of women and their total or recent fertility. Tables in which the two types of indicator are cross-tabulated can be found in older work or in the preliminary documents of more in-depth studies. This approach reveals the influence of certain variables, such as school enrolment, marriage or social category, on the relationship between women’s employment and fertility. For example, we observe that parity is lower among working women than among non-working women (Rosenberg, 1983; Welti, 1989); but when school enrolment is taken into account, the relationship becomes weak or even disappears altogether [2] (Weller, 1968; Welti, 1989; Smith, 1981b). Rosenberg (1983) shows that it appears only after ten years of union and that it is stronger in legal marriages than in common law marriages. Lastly, Stycos and Weller (1968) show that the relationship is strongly negative at the top and bottom of the social hierarchy and low among the middle classes [3] (Table 1).
Relationship between occupational status and fertility based on studies using cross-tabulation, Latin America
Relationship between occupational status and fertility based on studies using cross-tabulation, Latin America
7The cross-tabulation approach is exploratory in nature, while multi variate models claim to determine the precise relationship between women’s employment and fertility. To this end, most studies take account of different employment categories. Three types of model exist, and we will analyse them separately: those which estimate the direct impact of women’s employment on fertility, those which consider women’s employment and fertility as dependent variables determined by the same factors and thirdly, those which study the influence of women’s employment on contraceptive use.
Type 1 models: influence of women’s employment on fertility
8The first type of model follows the standard approach of predicting fertility from a series of independent variables using the ordinary least squares method or a hierarchical model (Table 2).
Type 1 models: influence women’s employment on fertility
Type 1 models: influence women’s employment on fertility
9The United Nations study (1987) based on the World Fertility Surveys of the 1970s includes eight Latin American countries [4] and estimates the effect of employment in different occupational categories [5] on fertility. The result are markedly different from one country to another. However, women’s employment in the modern and mixed occupational categories nevertheless has a more frequently negative impact on fertility, and a stronger impact than employment in other categories. Employment in transitional occupations also tends to reduce fertility, though the effect is only significant in Paraguay. The impact of employment in traditional occupations is not consistent, sometimes being negative and sometimes positive.
10For the same purpose of comparison and with the same data, Rodriguez and Cleland (1981) perform a hierarchical analysis. They start by analysing the difference between rural and urban fertility levels and then introduce school enrolment, occupational status of the husband and that of the wife. The results show that women’s employment has a negative effect on fertility, though the impact differs from one country to another and is greater for wage earners than for the self-employed.
11It is not easy to compare the two studies. Firstly, the categories and statistical methods used are different and secondly, the measurement of fertility used by Rodriguez and Cleland is based on recent fertility, while the United Nations used the number of liveborn children. Despite these differences, the two studies give fairly similar results. The relationship between women’s employment and fertility is stronger in Mexico, Colombia and Costa Rica, but weak in Peru. Moreover, it is noteworthy that the United Nations found a positive relationship in the Dominican Republic, the only country for which the results are non-significant in Rodriguez and Cleland’s study.
12In her study of Puerto Rico in 1982, Szembik (1982) distinguishes between formal and informal employment [6]. She finds that, on average, the fertility of working women is two-thirds lower than that of non-working women and that the impact of employment on fertility is slightly higher in the informal sector than in the formal sector. Gendell and Maraviglia (1970), for Guatemala (1964), contrast household domestic employment with other types of occupation and find that the impact is greater for the former. They attribute this result to the practices of employers who select women with few children, especially when they live on the employer’s premises. This is also probably the reason why Szembik found a stronger relationship between employment and fertility in the informal sector, where household domestic jobs are strongly represented.
Type 2 models: determinants of fertility and women’s employment
13This second approach is based on the idea that women’s employment and fertility are both governed by household decisions based on the value of the woman’s time spent in paid employment and the household budget. The value of women’s time is measured by estimating their potential wage [7]. The household budget is measured either by the husband’s wage (Smith, 1981a; Wong and Levine, 1992), or by total income (McCabe and Rosenzweig, 1976) (Table 3).
Type 2 models: determents of fertility and women’s employment
Type 2 models: determents of fertility and women’s employment
14Wong and Levine (1992) focus on the role of the household structure in reducing conflict between women’s employment and childcare in Mexico in 1982. The household structure is entered into the model as an explanatory variable and is measured by the number of non-working women and girls aged over six present in the household. The results indicate that household income and women’s potential wage have a positive effect on women’s employment and a negative effect on fertility. The presence of women or older girls increases the probability of working, though contrary to expectation, it does not affect recent fertility.
15McCabe and Rosenzweig (1976) focus not only on household structure, but also on the compatibility of employment with childcare in Puerto Rico in 1970. These variables are governed by household decisions and are introduced into the model as dependent variables determined by the same series of predictors as women’s employment and fertility. The authors reason as follows: if their wage is potentially high and if women want to have children, they will choose a job that is compatible with childcare or they will entrust the care of their children to others. The authors show that in Puerto Rico, wage level has a positive effect on labour force participation and fertility, that employed women more often live in extended households and that they choose a job that can be done in the home and that is compatible with childcare [8].
16In Smith’s (1981a) study of Mexico in 1971, the choice of occupation is also a dependent variable. Smith distinguishes between work in the formal and informal sectors [9]. He finds that the woman’s estimated wage has a negative effect on fertility and is associated with an increased likelihood of working in a modern occupation [10].
17Comparing the three studies, we observe that McCabe and Ronsenzweig find that the woman’s wage has a positive effect on fertility, while in the two other studies the effect is negative. The reversal of the relationship can hardly be attributed to difference in the models used (introduction of household structure by Wong and Levine and of attitude variables by Smith). We have identified three possible explanations. The first is that McCabe and Rosenzweig’s study is based on a survey in Puerto Rico in 1970, a period of high fertility, while the other two were conducted ten years later in urban Mexico, when fertility was already in decline. Consequently, as the context is very different, it is possible that the influence of women’s employment on fertility is different in the two sets of circumstances. In addition to these differences of time and location, McCabe and Ronsenzweig focus exclusively on women aged 35 or over.
18A second possible explanation lies in the way in which wages and fertility are measured. McCabe and Ronsenzweig estimate women’s potential wage on the basis of single women’s wages at the time of the survey, while fertility is measured by the cumulative fertility of women aged 35 or more. Consequently, fertility measurement is based on the behaviour of cohorts much older than those whose wage is estimated, since young women are most probably over-represented among single women. This amounts to an implicit assumption that the determinants of wage level have not changed over time. But such is not necessarily the case, especially under the conditions of rapid social and economic change that characterized the 1970s and 1980s. Moreover, single and married women probably do not have the same characteristics. Non-observed variables may simultaneously influence the decision to work and to marry, along with the choice of an occupation that is compatible with household duties. In short, there is selection bias which implies that the wage estimated on the basis of single women may not accurately reflect that of married women at the end of their reproductive life. The wage estimates proposed by Smith and by Wong and Levine, for their part, are based on all employed women, both single and married. This partly avoids selection bias, but mixes women with and without children so that the assumption of an independent wage variable is not fully respected.
19The link between wage and choice of occupation is not a one-to-one relationship either. McCabe and Rosenzweig find a positive link between wage and the choice of a job compatible with childcare, while Smith finds a negative link. Once again, the explanation for this difference may lie in the wage estimation method (strong selection possible with McCabe and Rosenzweig, problem of independence for Smith).
20A third difference between the studies is the relative effect of household structure. McCabe and Rosenzweig find that the presence of adults in the household varies in the same direction as women’s employment and fertility. Wong and Levine (1992), on the other hand, did not find any influence on fertility. Surprised themselves by this result, they suspect that the explanation lies in the inclusion of children in the household composition: the absence of a significant influence on fertility and the negative sign of the relation may reflect the lower probability for a woman who already has grown-up children of also having very young children (Wong and Levine, 1992, p. 98).
Type 3 models influence of women’s employment on contraceptive use
21Despite the importance attributed to intermediate variables in demography, they are relatively absent from studies on the link between women’s employment and fertility in Latin America. Breastfeeding and fertility are only briefly mentioned in the United Nations study (1987), and contraceptive use is merely skimmed over in a few articles (Table 4).
Relationship between women’s employment and contraceptive use in the studies using different approaches
Relationship between women’s employment and contraceptive use in the studies using different approaches
22By presenting cross-tabulations, Remez (1992) finds a positive relation between women’s employment and contraceptive use at the time of survey in Guatemala in 1987. Welti and Paz (1994) focus on the period between marriage and birth of the first child. They also find a strongly positive relation in all social categories in Puebla, Mexico, in 1992.
23By contrast, the United Nations regression analysis (1987) gives results that vary from one country to another. Only in Ecuador and Peru do women who are employed at the time of survey make significantly higher use of contraception than those who do not work [11]. It may be that contraceptive use is more frequent among employed women, but that the difference between women who work and those who do not is no longer significant once control variables are introduced into the model.
24Regarding the relationship between contraception and level of fertility, the results do not always agree. In the study by Welti and Paz, the more frequent use of contraception by working women does not lengthen the interval between marriage and first birth. Likewise, in the United Nations study, in Peru and Ecuador, the relationship between women’s employment and fertility is weak, while between employment and use of contraception it is strong. In other countries, such as Colombia, Mexico or Costa Rica, the influence of women’s employment on fertility is strong, though it has no effect on contraception. On the basis of this observation, two hypotheses can be put forward. Firstly, the effect of women’s employment on contraception only partly explains its effect on fertility and secondly, this explanation becomes all the more partial as the effect of employment on fertility increases. Moreover, Peru and Ecuador are countries where contraception was still not widespread. This suggests a third hypothesis: the effect of women’s employment on contraception decreases as contraceptive practice develops. It is only when contraception becomes widespread that employment affects fertility. Studies taking account of the different types of contraception, their efficacy and the frequency of discontinuation are needed in order to examine this question in more depth.
The influence of several important confounding variables
Employment before marriage and duration of employment are factors with a strong influence on age at time of marriage (Rosenberg, 1983; Welti, 1989). Moreover, most women who were working at the time of survey or had worked during their married life had already worked prior to marriage. Consequently, a part of the effect of women’s employment on fertility can be attributed to selection of women who were in work prior to marriage and who delayed their union. To determine the net impact of employment on fertility, it is therefore necessary to control for age at time of marriage, and this is not the case in all studies.
Surprisingly, the specific roles of age and duration of marriage have rarely been studied, despite the obvious influence of these variables. Firstly, age at time of survey serves to determine the birth cohort to which the women belong, an important factor since the older cohorts have higher fertility and a lower labour force participation rate. Secondly, age and duration of marriage provide information on the moment reached in the life cycle, with its contrasting influence on employment and fertility. Employment opportunities are more limited at higher ages, though there is less conflict between work and motherhood. Duration of marriage is also an indication of stability which reduces the probability of employment. The study results show the contradictory effects of these variables. Szembik (1990) and Rosenberg (1983) find a negative relation between duration of marriage and women’s employment, though age appears to have a positive influence (Szembik, 1990; Welti, 1989).
2 – Longitudinal approach
25Cross-sectional studies examine the relation between women’s occupational status and their fertility at a given moment. The life cycle approach, on the other hand, follows women over their childbearing and working life and analyses the way in which they fulfil their different roles, either successively or simultaneously. This approach provides a partial response to the difficulties raised by cross-sectional studies, notably in determining the direction of causality between women’s employment and fertility (Ní Bhrolcháin, 1980) and with regard to the effect of selection of the least fertile women by the labour market (Rodriguez and Cleland, 1981).
26One way to avoid these problems, in part at least, is to focus on the relation between employment before marriage and age at first birth or the interval between marriage and the first birth. In the city of Puebla in Mexico in 1992, Welti and Paz (1994) find that age at first birth varies substantially according to occupational status, with women who have worked bearing their first child 2.4 years later, on average, than other women. Regarding the interval between marriage and first birth, neither Welti and Paz (1994), nor Stycos and Weller (1968) find any difference between women who have worked and those who remained economically inactive during this period, except in the most advantaged social categories. So it would appear that the effect of women’s employment on the age at first birth is played out largely through the effect on age at union.
27Another approach involves comparing the relative number of women who work at different times in their reproductive life. Mier y Teran (1996) compares the percentage of women who worked before marriage, during the interval between marriage and first birth and at the time of the DHS survey in 1987 in Mexico. Welti and Paz (1994) make the same comparison on the basis of parity using data from their survey in Puebla (Mexico). Focusing exclusively on certain moments in women’s life history and lacking any control variables, these studies clearly have their limits. They nevertheless give an initial insight into the relation between demographic events and employment over women’s lives: 1) employment is more frequent before marriage than after; 2) few women who did not work before marriage work afterwards; 3) many women continue to work during the interval between marriage and first birth, especially in the most advantaged social categories, though most give up work afterwards. But women with a large number of children (three or more) are often obliged to return to work to meet the growing demands of the household, especially in the most disadvantaged social categories.
28Implementation of the longitudinal approach is optimal when associated with life history analysis, which estimates the probability of an event according to the changing conditions of the different phases of life. Using the data from Puerto Rico in 1982, Kahn and Whittinton (1994) estimate the influence of the value of women’s time [12] on the probability of having a first child at different times [13]. The authors observe that the effect of the value of time on the probability of having a first child is negative for women aged 15-18, non-significant at ages 20-23 and positive at ages 24-26. They attribute this result to the fact that young women with a high value on the labour market postpone births until a later age and then catch up on the delay.
29For Colombia in 1984, Florez (1989) estimates the annual probability of having a child by current occupational status and at a previous moment in time [14]. In an urban environment, the occupational status during the observation period has no effect on the probability of having a child, whereas in the rural environment, being in employment has a positive effect. On the other hand, being in employment before the observation period reduces the probability of having a child, in both urban and rural environments.
30Gurak and Kritz (1982) analyse the mutual relationship between employment and fertility in the Dominican Republic in 1978. They assess the effect of employment, measured according to the number of months worked during the two years preceding each observation year, on the probability of having a child during that year, along with the impact of having a child during an observation year on the number of months worked during the following year [15]. The results indicate an absence of any relationship between employment and fertility: on the one hand, women have the same probability of working during the following year whether or not they have had a child during an observation year; and secondly, the probability of having a child is the same for women who did or did not work in the years prior to the observation period [16]. We nevertheless observe a negative impact of the parity reached at the start of the observation period on the probability of working.
31Both Florez and Gurak and Kritz obtain comparable results, indicating the absence of a short-term relationship. In both studies, this result is verified for all types of employment. But over the longer term, a relationship emerges: Florez finds a negative effect of work experience on fertility, while Gurak and Kritz show that total fertility tends to reduce the probability of working.
II – The influence of context
1 – Level of development and fertility
32In developed countries, the relation between women’s employment and fertility is generally negative and relatively stable. This observation raises the question of whether industrialization in developing countries strengthens the relation between women’s employment and fertility. The United Nations (1987) tested this hypothesis for all developing countries and found that having a modern occupation has a stronger impact on fertility in countries with a high development index value than in countries where the value is low. On the other hand, the effect of other types of employment does not vary according to the development index value.
33We propose to test this hypothesis for the countries of Latin America and study the relation between employment and fertility according to the fertility levels in the different countries. To this end, we divided the eight Latin American countries of the United Nations study (1987) into two groups, according to the human development index values (UNDP, 1990), and the fertility level (CELADE, 1996, 1997). The human development index value is considered to be high above 0.80 and low below this threshold. Likewise, the overall fertility rate is considered to be high above 4.5 children per women, and low otherwise. For each group of countries thus defined, we calculated the mean regression coefficients of fertility on women’s employment taken from the United Nations study (Figure 1).
Effect of type of occupation (modern, transitional, mixed and traditional) on fertility according to the human development index value and fertility level in Latin American countries, based on United Nations study data (1987)
Effect of type of occupation (modern, transitional, mixed and traditional) on fertility according to the human development index value and fertility level in Latin American countries, based on United Nations study data (1987)
34Figure 1 shows that the relationship between employment in a modern occupation and fertility is much more negative in countries with a high human development index value than in countries where it is low. This relationship is also more strongly negative in countries where fertility is low. The reverse is observed for traditional and transitional occupations. The relationship between women’s employment in mixed occupations and fertility is negative but varies little with the development index value and the overall fertility level in the country. Lastly, it is interesting to note that the differences according to the type of employment are larger in the context of a high development index value and low fertility.
35These results are congruent with the hypothesis of the influence of development. It is primarily the effects of employment in a modern occupation on fertility which change with the level of development. However, the strong correlation between development and fertility suggests that the effect of development is perhaps no more than apparent. The countries included in the study are at the start of their demographic transition and the effect of employment on fertility is only visible in countries where women are starting to control their fertility. It is not surprising that women employed in modern occupations are the first to reduce their fertility.
36Though these results based on eight countries cannot be generalized, it is clear that macro-social variables can play a role in the relationship between women’s employment and fertility, and should not therefore be almost systematically excluded from empirical studies.
2 – Changes in the relationship over time
37Mier y Teran (1996) studies the working histories of women using data from the 1976 and 1987 FHS surveys in Mexico. The first date corresponds to a period of prosperity, and the second to a period of crisis. She observes that the percentage of women who work before marriage and of women who have worked at least once in their life changed little over the period. However, the proportion of women who were working at the time of survey increased. Moreover, on the two dates, labour force participation is higher among women with young children than among women at the end of their reproductive life.
38Garcia and de Oliveira (1994) apply a logistic regression model to the same data and find that between 1982 and 1987, the negative relation between fertility and labour force participation grew stronger in the most advantaged social categories and disappeared in the most disadvantaged categories.
39Overall, the relation between women’s employment and fertility has changed over time, as has the influence of certain variables such as age, stage in the life cycle or social status. Mier y Teran’s study points out that the crisis primarily affected the continuity of employment or women’s return to the labour force, especially for those with young children. Consequently, it would appear that labour force participation rates have become increasingly similar for married and single women, and for women at different stages in their life cycle. Garcia and de Oliveira, for their part, have shown that during the crisis of the 1980s, women from advantaged social categories with a relatively large number of children were the first to lose their jobs, resulting in a closer relationship between fertility and female employment in 1987. Women from disadvantaged social categories, for their part, were obliged to increase their labour force participation rate despite — or perhaps even due to — their large number of children, thereby weakening the relationship between employment and fertility.
III – Discussion
40In our review of the literature on the relationship between women’s employment and fertility in Latin America, we have seen that the conclusions vary from one study to another. The relationship is sometimes strong and negative, sometimes weak or non-existent, and sometimes even positive. The aim of this paper is to understand the reasons for this divergence. We began by comparing the methodological approaches, and then the contexts in which the relationship is measured.
41Concerning the methods used, we have distinguished between cross-sectional and longitudinal approaches. The first group includes studies based on cross-tabulation and multiple regression analyses. These analyses make use of three types of model, depending on the type(s) of variable explained: fertility, occupational status and fertility together, and contraceptive use. The most important reasons for the differences observed between the studies appears to be the statistical method used, the omission of important control variables, the measurement of variables and the variability of the phenomena studied.
42The longitudinal approaches include working history studies and life history analysis. These studies offer the advantage of avoiding, in part at least, ambiguity in causality between phenomena due to confusion in the time sequence of variables and due to selection bias in cross-sectional analyses. The longitudinal approach also distinguishes between short- and long-term effects.
43In the second part, we examined how the study context — a component almost totally ignored in the literature — affected results. We showed that macro variables such as development level, fertility level and period are liable to strongly influence analysis results.
44If we compare Latin American studies with those concerning Northern countries, notably the USA and Europe, we observe that their approaches are very similar. Until very recently, the vast majority focused exclusively on individual characteristics, without explicitly taking account of the broader constitutional context (Spitze, 1988; Bernhardt, 1993; Brewster and Rindfuss, 2000). At present, as is the case for the social sciences in general, in Europe especially, this purely micro vision is giving way to a more institutional vision. During recent decades, the relationship between women’s employment and fertility in several Northern countries has become positive, indicating that a large proportion of women are no longer obliged to make a choice between family and work (Kögel, 2004; Adsera, 2004), and this observation has probably accelerated this paradigm change. Research is shifting towards the analysis of broader societal factors, the most widely studied being childcare (increased availability and lower relative cost of day care, parental leave, etc.) and labour market characteristics (flexibility, technological changes, etc.).
45In the most recent Latin American sociodemographic analyses, we find a similar shift in research paradigms towards a more institutional vision. Though the link between women’s employment and fertility is not always the central issue, these new studies provide three useful keys to understanding this relationship, firstly through the focus on women’s employment and reproduction in studies on poverty and family vulnerability (Retamoso, 2002; Tuiran, 1993; Wong and Levine, 1992), secondly through studies of the increased autonomy and decision-making powers within the family of women employed outside the home (Casique, 2003; Garcia and de Oliveira, 2004); and thirdly through the study of women’s experience in different social situations and the meaning attached to their work and their motherhood (Garcia and de Oliveira, 1997). Family and gender are central to much of this recent research (Acosta, 2003).
46Despite a relatively similar methodological development in Latin America and in the Northern countries, our literature review shows that the study results present certain major differences. Firstly, the relation between women’s employment and fertility is much less stable in Latin American countries than in industrialized countries. This generates a much larger variation in the influence of the type of employment on fertility, whatever the employment classification used. This is mainly because the labour market is more diversified in Latin America, ranging from high-level salaried employment to casual labour in the home, such as several hours per day of childcare.
47A second general difference in results concerns the changes that take place over time. In the North, the weakening or even the reversal of the relationship between employment and fertility are linked to improvements in institutional conditions that make it easier to reconcile family and working life. It is among the most advantaged social categories that these changes are most visible. In Latin America, on the other hand, the weakening of the relationship between employment and fertility is due mainly to a reduction in the share of wage employment and other types of modern occupation, combined with an increase in the number of women working in the informal sector or who are self-employed. These trends occurred in conjunction with the economic crises of the 1980s and 1990s. The weakening of the relationship between women’s employment and fertility is thus linked above all to the behaviour of women in the most disadvantaged social categories who are obliged, for economic reasons, to cope with the difficulties of reconciling work and childcare.
Conclusion
48This panorama of the relationship between women’s employment and fertility in Latin America shows that the many available studies are based on statistical measures and methods that are difficult to compare. Faced with this apparent impasse, the standard response has been to look for increasingly sophisticated statistical methods. The divergences between studies are therefore partly the result of methodological advances which have certainly contributed to a better understanding of the relationship between women’s employment and fertility. However, this technical quest has not produced a consensus. The supporters of cross-sectional methods and those who favour a longitudinal approach each vigorously defend their own choice and highlight the numerous drawbacks of the rival method (Cramer, 1980; Smith-Lovin and Tickamyer, 1982).
49This situation has led to a change of paradigm: from a purely micro vision, there has been a shift towards institutional analyses taking account of the context and of the specific experience of players. Though the effects of this change on the study of fertility determinants are not clear, we believe that these new analyses may lead to a better understanding of the influence of women’s employment on fertility. An institutional analysis could shed light on the social interaction process which underlies the relationship between the two phenomena. Generalization — without losing sight of contextual particularities — thus becomes possible. Even though the relationships between the phenomena are different and are observed in different contexts, the underlying social process may be the same. An excellent example of such a methodology is provided by Skockpol (1979) who, in a comparison of the French, Russian and Chinese revolutions, suggests a set of common mechanisms operating in all three cases which explain their very different triggers and outcomes.
50In short, it could be said that until now, empirical studies of the relation between women’s employment and fertility have sought to obtain generalizations of the following type: “work of type x leads to fewer liveborn children than work of type y, controlling for z”. In our view, we should move towards generalizations of the following type: “work of type x and fertility evolve in opposite directions (or in parallel) via the mechanism y which can be understood by the social relations z”. We would thus obtain a theoretical generalization via an institutional analysis which sheds light on the relationship between the phenomena in question.
Acknowledgements
This work was funded by an Alpha Pop II grant from the European Commission.Notes
-
[*]
Institut de Démographie, Université Catholique de Louvain-la-Neuve, Belgium.
Translated from the French by Catriona Dutreuilh. -
[1]
The term maquiladora is used to describe companies that process or assemble imported components that are then re-exported.
-
[2]
The relation is weaker among married women and disappears among ever-married women (Cf. Box).
-
[3]
Stycos and Weller did not give their definition of the different social categories.
-
[4]
Colombia, Costa Rica, Dominican Republic, Ecuador, Mexico, Panama, Paraguay and Peru.
-
[5]
The four occupational categories are: modern (professional, clerical); transitional (domestic household employee, service); mixed (sales, skilled and unskilled); and traditional (agricultural).
-
[6]
The formal sector includes wage employment; the informal sector includes non-wage employment and domestic household employment.
-
[7]
Estimations based on a logistic regression model applied to the wage of a sub-group of working women, classified according to schooling, rural or urban origin, migratory status and length of stay in Mexico City in Smith’s study, and according to schooling, age and birth inside/outside Mexico City for Wong and Levine.
-
[8]
The choice of occupation is based on the distinction between work inside and outside the home and on a compatibility indicator which is a function of the number of liveborn children for 35 different types of employment. The household size is measured by the number of servants and family members living in the household.
-
[9]
The distinction between formal and informal work is based on a consensus of different researchers. If the working hours and place of work are flexible, if relations with the employer and with customers are informal and if the tasks do not require constant attention, then the job is classified in the informal sector. If the opposite is the case, it is classified in the formal sector.
-
[10]
Unlike the two previous studies, Smith does not take account of household structure, though he measures the influence of traditional attitudes through questions on the roles of the spouses, patriarchal dominance and an indicator of religious devotion. None of these measures has a significant influence on the dependent variables.
-
[11]
Analysis is limited to married women, not pregnant at the time of survey.
-
[12]
The value of women’s time is estimated using a scale of occupations considered as a proxy of wage differences.
-
[13]
Religious affiliation, area of residence, schooling and husband’s occupational status were introduced as control variables.
-
[14]
The control variables are level of education, age at marriage, migration and contraceptive use.
-
[15]
The control variables are number of previous liveborn children, age, work experience, schooling of the woman and her husband, and contraceptive use.
-
[16]
Smith-Lovin and Tickamyer (1982) point out that Gurak and Kritz’s conclusion — that a recent birth does not affect the probability of working — contradicts the results of analyses conducted in Northern countries. They attribute this difference to a problem of correlation between fertility before the observation period and during the observation period.
