Spain’s Persistent Negative Educational Gradient in Fertility

Abstract

A trend towards a reversal of the negative educational gradient in fertility has been detected in some advanced societies, although the end of this inverse relationship is far from being generalized across developed countries. Previous analyses have shown that, for most of the twentieth century, Spain exhibited the steady negative educational gradients in fertility that characterize the low-fertility demographic regimes in Southern Europe. This report presents data collected on the period fertility of Spanish women between 2007 and 2017 and tests for the persistence of educational gradients in their recent reproductive behavior. Using data taken from Spanish administrative registers and standard period indicators of fertility, evidence is offered that proves that a negative educational gradient in fertility continues to exist in the country in accordance with the historical experience of cohorts born during the first decades of the twentieth century. This negative gradient implies significant differences in fertility between educational categories and affects both quantum and tempo. Moreover, the educational differentials in fertility not only persisted during the observed years but probably widened.

Background

The diffusion of fertility control involved in the demographic transitions has been associated with the expansion of mass education and, more specifically, with the growth in women’s participation in these expanded education systems (Basu, 2002; Cleland, 2002; Caldwell, 1980; Rindfuss et al., 1980). As a result, a negative relationship between educational attainment and women’s fertility—the so-called negative educational gradient in fertility—emerged and became a steady feature of post-transitional populations: the more educated than women are, the less children they are expected to have (James et al., 2012).

Some of the key mechanisms that have been proposed to explain fertility decline in modern societies (Van Bavel et al., 2018) place education at the center of the stage. First, changes in the lifecycles of women deriving from their involvement in formal education have generally led to the postponement of childbearing (Gustafsson & Kalwij, 2006; Mills et al., 2011; Ní Bhrolcháin & Beaujouan, 2012; Spéder, 2006). This delay in maternity can lead to a drop in completed fertility, following the old demographic adage that fertility postponed tends to be fertility forgone.1 Second, as stated by the New Home Economics school (Becker, 1987), for women higher educational levels imply higher incomes, greater economic independence, and reduced gains from marriage. Wherever most children are born in wedlock, marriage is a proximate determinant of fertility (Bongaarts, 1982), which then crucially depends on the share of women of reproductive age who are married or have a partner. Third, and in the same vein of economic theories of the family, additional years invested in formal training are supposed to produce higher returns on education and, therefore, higher opportunity costs associated with fertility, as childbearing curbs female participation in the labor market. Fourth, cultural explanations for the reduction in fertility emphasize the growing diffusion of rationality, secularism, and individualism promoted by cultural modernization and, in particular, by large-scale educational processes (Lesthaeghe, 1983). Finally, from a more instrumental point of view, higher educational achievement also implies the more effective use of modern means of contraception, which are the main tools to reduce fertility in advanced societies.

The operation of all these mechanisms arouses the expectation of a negative educational gradient in fertility in post-transitional societies. For the most part, the reality lived up to this expectation. There is no need to quote here the many studies linking low fertility with education and providing empirical evidence for the link (for a recent assessment, see Sobotka et al., 2017). Suffice it to say that this negative educational gradient was supposed to affect both the timing and the quantum of fertility (Rindfuss et al., 1996). More educated women were more likely to remain childless, postpone childbearing, and have fewer children than less educated ones. This negative educational gradient in fertility seemed remarkably pervasive, despite specific variations across countries or regions, institutional contexts, and periods.

Also remarkable was the emergence, about two decades ago, of a new stream of literature suggesting a weakening, or even a reversal, of the negative relationship between education and fertility in advanced societies (Kravdal, 2001; Kravdal & Rindfuss, 2008). In some countries, as educational systems have continued to expand, women with higher educational credentials have come to represent an increasing proportion of the female population and gender equality has reached unprecedented levels, the fertility differentials between educational categories have narrowed or inverted. According to this recent literature, fuelled by a quiet revolution that has transformed women’s employment, education, and family lives (Goldin, 2006; see also Goldscheider et al., 2015), the educational gradient in fertility and other family behaviors (e.g., divorce and union stability) is being upended (Esping-Andersen & Billari, 2015). In institutional contexts that render female work and family duties more compatible, women with higher educational attainment have higher overall fertility levels.

Nevertheless, this trend towards fertility convergence among educational categories is far from being generalized across post-transitional populations. Again, there is a sizeable variation in results observed across countries, much of which is attributable to the different approaches used, the methods and measures applied, and the specific dimensions of fertility that are analyzed. Against this background, the clearest cases of the inversion of previous patterns are the Nordic countries, where women’s negative educational gradients in fertility started to vanish among cohorts born after 1940 (Hoem et al., 2006; Jalovara et al. 2018; Kravdal & Rindfusss, 2008). The inversion is less clear in other European countries. A comparison of the complete fertility of female cohorts born between 1940 and 1970 in three European countries plus South Korea, based on the same indicator, showed educational confluence around moderate fertility levels in Finland, stable differences at low levels in Spain, and a widening gap in Serbia, with South Korea converging at low levels (Sobotka et al., 2017). Another comparative study of 25 European countries (Merz & Liefbroer, 2017) found that negative educational gradients in complete fertility were weakest in Nordic and post-Soviet countries and strongest in Mediterranean and post-communist countries; still, in all these settings the gradient appears to have been reduced among later cohorts.

Turning to English-speaking countries, the panorama is equally complex. A convergence between more and less educated women in childlessness and complete fertility has been observed in the US and is now considered a distinctive feature of post-1960 fertility (Bailey et al., 2014), although other studies (Hazan & Zoabi, 2011) have found the relationship between fertility and female education in the US to be U-shaped. A more recent study (Zang, 2019) reported only a minor reduction in educational gradient in fertility among the cohorts of women born in 1961–80. In Australia, evidence has been presented (Gray & Evans, 2019) that fertility among tertiary-educated women born between 1952 and 1976 remained lower than that of other educational groups, and there was no sign of an inversion of fertility educational differentials. However, an analysis of the changes in period fertility indicators between 2006 and 2011 (McDonald & Moyle, 2019) revealed that the gap in fertility rates between highly educated and low educated women was closing. Finally, an analysis of women born between 1940 and 1969 in the UK (Berrington et al., 2015) showed that educational differentials in fertility had not changed over cohorts, partly because in Britain the postponement of motherhood has not increased the complete fertility of more highly educated women as a result of a late recuperation. Cohort fertility analysis (Reher & Requena, 2019; Requena & Salazar, 2014) has shown that, for most of the twentieth century, Spain exhibited the steady negative educational gradients in fertility that characterize the low-fertility demographic regimes in Southern Europe.

This research note presents data collected on the period fertility of Spanish women between 2007 and 2017 and tests for the persistence of educational gradients in their reproductive behavior. In recent decades female educational attainment and female employment substantially increased in Spain, whereas fertility rates steadily dwindled (Davia & Legazpe, 2014). Today, Spain is a low-fertility country characterized by a Mediterranean welfare state regime where family-friendly policies are not particularly generous and gender inequalities in the distribution of domestic work are widespread. Over the years observed in this study, while women’s employment first contracted and then expanded along the two phases of business cycle, female education continued to grow during the entire period. Since migratory flows declined between 2007 and 2013 but grew between 2013 and 2018, the share of migrant women of reproductive ages hardy changed over these years. Moreover, in 2007 a new, one-time, sizeable, and universal child bonus of 2500 € —the so-called ‘baby check’ in Spain— was introduced in the country, being in force until the end of 2010 (Gonzalez, 2011). Have the enduring negative differentials in fertility observed in Spain shifted recently as a consequence of these changes?

Data and Methods

This study made use of the customary tools of the trade—period fertility rates—computed from two main data sources. First, numerators for calculating fertility rates (i.e., numbers of births by age of mother, educational attainment, and birth order) were taken from Spanish administrative registers. Vital statistics, which are called Estadísticas del Movimiento Natural de la Población (MNP) in Spain, come from individual birth certificates and contain basic information regarding several characteristics of births, newborn babies, and their parents. All data were obtained from the Spanish national statistical office (Instituto Nacional de Estadística, INE). Annual original datasets are available from the INE webpage and fully accessible to the public. For the purpose of this paper, the yearly individual microdata of all births registered in Spain from 2007 to 2017 was aggregated into a new database that constituted a complete representation of the universe of births (N = 4,996,744) that occurred in the country in the period among women born between 1952 (55 years old in 2007) and 2005 (12 years old in 2017).

From 2007 onwards, Spanish data on births include the parents’ educational attainment. In this respect, a source of concern is the recently increasing number of observations with the mother’s education unknown in MNP birth statistics.2 To minimize potential biases induced by this omission, a multiple imputation procedure was followed based on a multivariate model that predicted the education level for those cases with missing values from the information available in the dataset on age, province of residence, size of municipality, marital status, children ever born and migratory status of the mothers with known education levels. Although education is registered in some detail on birth certificates, for the sake of parsimony and comparability in this work educational attainment was recoded into three categories: low education (lower secondary or less, ISCED 0–2), medium education (upper secondary and post-secondary non-tertiary education, ISCED 3–4) and high education (tertiary education, ISCED 5–8).

Second, denominators for estimating fertility rates—i.e., female population by single age from 16 to 55 years and educational attainment—were taken from the Spanish Labor Survey (SLS), a continuous Spanish household sample survey carried out quarterly by the INE in accordance with strict, internationally established methodological standards. As in other countries, this survey is the canonical source for examining the labor market. It also provides quite accurate and detailed quarterly estimates of the Spanish population and its educational level. Its sampling frame is the population living in Spain in family dwellings and its geographical scope encompasses the entire country. Every quarter the survey interviews a sample of around 3822 census tracts, 65,000 households, and 180,000 individuals. For this estimate, average annual data from 2007 to 2017 were used. For each year, the distributions of the female subpopulation by single age and educational attainment over reproductive ages were estimated with SLS data and subsequently used as denominators for calculating age-specific fertility rates.

The variation over time of the educational gradient in fertility was measured by comparing annual period total fertility rates (TFR) for different educational categories. The standard routine for calculating TFRs was followed, which is to simply sum the age-specific fertility rates. Using the SLS female subpopulation as denominators produced TFR annual estimates that differed from the official estimates of INE by no more than 0.02 children per woman during the observed period.

As fertility rates were calculated for synthetic cohorts, the estimates of TFRs refer to hypothetical female populations subject to the schedule of age-specific fertility rates existing in each given calendar year. Data for birth order-specific fertility rates and mean ages at birth by educational attainment are also presented and were analyzed to better assess the changes over time in the educational gradient associated with fertility. To specifically minimize timing distortions inherent to period indicators, tempo-adjusted total fertility rates (TFR*) are also presented, calculated as the sum of age-specific fertility rates by birth order corrected by the respective changes in the mean age at birth (Bongaarts & Feeney, 1998).

Results

Between 2007 and 2017, as the process of educational expansion among the Spanish female population continued to grow, fertility fell in Spain from 1.39 children per woman to 1.32 (Table 1). However, this decline (− 5.5%) was not evenly distributed across levels of educational attainment. Women with low educational credentials increased their fertility by 9%, but those with high levels of education reduced it by almost the same measure (− 10%); while the average TFR among low-educated women rose by 0.15 children per woman, it fell by 0.10 among high-educated women. As a result, the fertility ratio of low to high educational levels rose from 1.79 to 2.15, the fertility differential between educational categories increased, and the educational gradient of fertility expanded among Spanish women. In other words, the fertility of low-educated women not only remained above that of high-educated women over these years, but also the distance between the educational categories widened. Despite the characteristic volatility of TFR*s, the change in the TFR*s (Fig. 1) also demonstrates this widening gap between educational categories.

Table 1.

Female population in reproductive ages, births and TFR by education

	Women 12–54			Births			TFR
	Low educ	Med educ	High educ	Low educ	Med educ	High educ	Total	Low educ	Med educ	High educ	Ratio L/H
2007	6,376,361	196,891	3,779,236	202,104	146,613	143,735	1.39	1.78	1.58	1.00	1.79
2008	6,367,963	3,263,398	3,900,074	212,588	152,916	154,253	1.46	1.90	1.63	1.03	1.85
2009	6,325,107	3,272,275	3,978,133	196,869	143,180	154,880	1.40	1.76	1.54	1.02	1.72
2010	6,140,062	3,245,737	4,134,010	188,329	138,065	160,153	1.38	1.78	1.53	1.02	1.75
2011	5,908,406	3,323,959	4,232,972	173,723	134,184	164,046	1.36	1.73	1.44	1.04	1.66
2012	5,821,031	3,245,566	4,310,993	165,960	128,378	160,281	1.34	1.70	1.48	0.99	1.72
2013	5,669,907	3,218,129	4,304,779	151,398	118,925	155,288	1.29	1.61	1.41	0.97	1.66
2014	5,512,655	3,174,279	4,539,984	151,693	119,662	156,177	1.32	1.72	1.52	0.94	1.84
2015	5,333,933	3,192,560	4,596,523	152,362	116,841	150,986	1.33	1.83	1.51	0.92	1.99
2016	5,171,761	3,199,710	4,657,417	153,262	114,015	142,960	1.34	1.95	1.52	0.93	2.10
2017	5,170,552	3,166,490	4,630,877	149,035	108,858	134,869	1.32	1.93	1.53	0.90	2.15

Spain, 2007–17

Fig. 1.

Change in non-adjusted and tempo-adjusted total fertility rates in Spain, 2007–17

Splitting the fertility changes into two sub-periods is instructive in this respect. As most often happen (Sobotka et al., 2011), times of economic crisis are not particularly propitious for having children. The TFR steadily fell between 2007 and 2013 (− 0.10 children per woman) and then started to rise slightly until 2016 (+ 0.05), but again the reproductive behavior of the educational categories was dissimilar. Among high-educated mothers, change over the two sub-periods (− 0.03 and − 0.04, respectively) was small and smooth. By contrast, among low-educated mothers there was a relatively intense fertility decline (− 0.18) during the hardest years of recession and an even stronger recovery (+ 0.34) during the subsequent period of economic growth. From this, it appears that, in Spain, the fertility of low-educated segments is much more sensitive to economic circumstances than that of high-educated people.

This differential variation in total fertility rates is accounted for by changes in TFRs for birth orders (Fig. 2) and respective mean ages (Fig. 3). Although the decomposition of TFR by birth order presented in Fig. 2 is based on rates of the second kind (incidence rates) and not indicative of parity progressions, it is clear that the higher period fertility of low-educated women is based on their superiority in second and third birth orders. Average ratios for the entire period of low- to high-educated women TFRs oscillated around 1.56–2.01, and 2.7–5.4, for second and third and subsequent birth orders, respectively. This means that the higher the birth order, the larger the TFR differential between low- and high-educated women. It is also important to note that these ratios increased during the period for all birth orders.

Fig. 2.

TFRs of different birth orders by mother’s education. All births occurred in Spain, 2007–17

Fig. 3.

Mean ages at births by birth order and education. All births occurred in Spain, 2007–17

Childbearing postponement and late fertility define the demographic regimes of South European countries. Overall, mean age at births of any order increased by 1.7 years in Spain during the period. Nevertheless, educational attainment made a clear difference to fertility schedules. Figure 3 shows the mean ages at birth by birth order and education. As expected, low-educated women had children of the same order at a younger age than high-educated women. Low-educated mothers had their first babies an average of 5.6 years before high-educated mothers, with the equivalent figures being 4.4 years for second births and 3.9 years for third and subsequent children. Low levels of education imply earlier reproductive schedules, and the negative educational gradient in fertility is not only visible in the quantum but also in the timing of fertility.

Discussion

The existence in Spain of a significant negative educational gradient in fertility has been consistently proven by cohort analysis (Requena & Salazar, 2014; Reher & Requena, 2019; Sobotka et al., 2017; van Bavel et al., 2018). Different from Nordic countries and South Korea, and possibly from Australia and the US, but similar to the UK and other European countries, low-educated Spanish women have shown higher levels of fertility than high-educated ones. As a matter of fact, this negative educational gradient seems to characterise the low-fertility regimes of Southern European countries and Mediterranean welfare states (Meerz & Liefbroer, 2017). Two additional features of the gradient in Spain are worthy of mention. First, the negative gradient describes both the quantum and the timing of fertility: the less education, the more children; the more education, the later the childbearing age. Second, the gradient has persisted over time, from the cohorts born in the first decades of the twentieth century until, at least, the cohorts born in the late 1960s and early 1970s.

Have these educational differentials persisted in the most recent reproductive behavior of Spanish women? Based on period fertility indicators, this paper offers evidence from data on births in Spain between 2007 and 2017 to mothers born between 1952 and 2005. This evidence, which has not been previously compiled, demonstrates three important findings. First, a negative educational gradient in fertility continues to exist in the country in accordance with the historical experience of cohorts born during the first decades of the twentieth century. Persistence in the inverse association of reproductive behaviors and education dominates the fertility scenario in Spain. Second, this negative gradient implies significant differences in fertility between educational categories. During the whole period observed here, women with low education levels had on average 0.8 children more than women with high education levels. The magnitude of such a difference is well gauged by the fact that it is higher than the change in total fertility experienced in Spain during this decennium. Third, the educational differentials in fertility not only persisted during the observed years but also widened. As a result, the negative educational gradient could have become steeper over recent years.

This study is subject to a number of limitations. Using standard period indicators entails the risk of timing biases or distortions. To avoid that risk, tempo-adjusted total fertility rates based on rates of the first kind (Bongaarts & Feeney, 1998) were estimated and presented. These tempo-adjusted rates, however imperfect or volatile they may be, confirm the findings derived from non-adjusted rates. Unfortunately, as parity distributions of Spanish women were not available for the entire period, other alternative solutions to tempo distortions are difficult to find. There is no way to mechanically translating period data into cohort data, but the period evidence presented in this paper can be complemented with cohort data based on completed fertility. For example, for discarding the specific tempo distortion arising from the possibility that today's highly educated young women are not having children at younger ages (< 30), but might eventually show fertility recuperation at later ages (above 30 or even above 35), the 2018 Spanish Fertility Survey (SFS) data, available at www.ine.es, can be used. SFS data show basic stability in educational gradients, around a ratio of 1.2 between low/high educational categories, for two birth cohorts of women born in 1963–68 [50–55-year-old at the moment of the survey] and in 1968–73 [45–49]. Admittedly, the comparison is far from perfect because the approaches and populations are different, but at least it confirms the recent persistence of negative educational gradients in fertility among women born in the sixties. On the other hand, these period indicators do not entirely capture the significance of childlessness for fertility levels. To surmount this obstacle, a preliminary vetting of SFS corroborates the negative educational gradient in childlessness among women aged 18–55. Finally, there is always the possibility that women might increase their educational achievement after having reached the parity that can be observed using period data. It can be assumed, however, that the end of education for most low-educated women comes before fertility (Stange, 2011). Education tends to have causal precedence over fertility. Considering the data gathered in this paper, it seems unlikely that significant numbers of women with only lower secondary education and modal ages at fertility close to thirty will go on to improve their education levels after motherhood.

Another different source of concern comes from the contribution to fertility of migrant women residing in the country. Given that in Spain the fertility level of the immigrants has been higher than that of the native population (Del Rey & Grande, 2015; Gonzalez Ferrer et al., 2017), changes in migration over time can affect the observed educational gradient in fertility if migrant women become an increasingly selected segment of the female population with high fertility and low educational attainment. On the one hand, according to 2011 Census data (not shown here, but available upon request), the educational gradient of migrant women is steeper than that of natives. Therefore, a considerable part of the educational gradient in fertility is due to the reproductive behavior of migrants, although it also characterizes, to a lesser extent, the fertility of women born in Spain. On the other, SLS data show that a substantial proportion (around 80%) of these women with low educational attainment are Spanish-born. From this point of view, low-educated women have not become more selected by migration groups. Interestingly, between 2012 and 2017, when the fertility dissimilarity between low and high educational levels grew the most, the migratory component of female population remained stable.

A final point concerning the explanation of the persistence of an educational gradient in a country like Spain must be briefly addressed. As stated above, several mechanisms explaining fertility decline in modern societies have been proposed to explain the negative association between education and fertility. In one way or another, all of these mechanisms point to education increasing the opportunity costs of female reproduction. Observers of the reversal of educational gradients in fertility in Nordic countries have advanced two main arguments to account for it. Some policies and institutional devices intended to reduce incompatibilities and conflicts between work and childbearing –from subsidized childcare to paid maternity leave—have reduced the wage penalty of motherhood and eased maternal employment (Kravdal & Rindfuss, 2008). A recent work arguing that the educational gradient in Australia has reversed (McDonald & Moyle, 2019) attributes the fertility decline among low-educated mothers to the relatively scant and decreasing government support for these families. Furthermore, the (second half of the) gender revolution (Goldscheider et al., 2015) has fostered the increasing involvement of men in domestic chores, which has enhanced the lure for women of partnering and family formation (Esping-Andersen, 2016; Esping-Andersen & Billari, 2015).

Spain is very far from being one of those countries where an educational gradient reversal is occurring. Among the 37 OECD countries in 2015, Spain ranked 32nd in public expenditure on family benefits (cash, services, and tax breaks for families) as a percentage of GDP. Therefore, it might well be that an improvement in family policies and greater support to working mothers of childbearing age would help to raise fertility among high-educated women. A word of caution is in order here, however: in Spain, the new path taken away from the family/kinship model and towards a dual-earner family system (Naldini & Jurado, 2013) has not resulted in substantial increases in fertility. Between 2001 and 2009, when the public spending on family benefits multiplied by 1.9, fertility only grew by a factor of 1.1 (or + 0.14 children per woman). The short experience (2007–2010) of universal child bonus brought about a positive jump in the number of conceptions after the policy announcement date and a small drop in abortions (Gonzalez, 2011), but the baby checks only increased the fertility rates in 2008 inducing a very modest, almost negligible decrease in the educational gradient between 2007 and 2010. As to the other supposed factor, gender inequalities in the distribution of domestic work are known to be greater among families of low-educated women, but the much greater gender equality within more educated families (González & Jurado, 2009) has not yet promoted their fertility.

Funding

This work was supported by the Spanish Ministerio de Ciencia, Innovación y Universidades (RTI2018-098455-B-C21) and by Programa de actividades de I + D en Ciencias Sociales y Humanidades, Comunidad de Madrid (H2019/HUM-5802).