1. Introduction
The expansion of the education systems is one of the main correlates of the processes of sociocultural modernization and economic development that have taken place in all advanced societies for at least the last 150 years. As is well known, Spanish society underwent a dramatic educational enlargement during the 20th century. In terms of supply, the education system grew following a trend toward continuous expansion; in terms of demand, there were sustained advances in access to formal education for increasing fractions of the population. Four main milestones marked the trajectory of change (Requena and Bernardi, Reference Requena, Bernardi, González and Requena2005): the almost entire eradication of illiteracy, the full schooling of children, the extension of compulsory education, and the unprecedented increase in access to university education in the country. The educational panorama in Spain at the beginning of the 20th century was that of a country with a predominantly rural base, archaic in its economic and productive structure, underdeveloped in terms of its cultural facilities, and fragmented by flagrant social inequalities. At the start of the last century, more than half of the Spanish population was illiterate. The school experience was alien to almost half of the boys, who began their working lives at a very early age, and even more alien to girls, traditionally excluded from access to formal education. Spain was, at that time, a backward country in terms of its educational level compared to other European countries. Thanks to the change, the Spanish population reached an aggregate educational level similar to that of these countries by the end of the century.
The secular process of human capital growth in the Spanish population has been extensively studied by the specialized literature (Núñez, Reference Núñez1992; Reher, Reference Reher1997; Viñao, Reference Viñao2004). Many of its characteristics have been dealt with in depth and are beginning to be well understood. Among others, for example, regional disparities in the rates of growth of literacy and schooling (Beltrán Tapia and De Miguel Salanova, Reference Beltrán Tapia and De Miguel Salanova2021; Reher, Reference Reher2022), the impact of institutions and public policies on the educational expansion process (Viñao, Reference Viñao2012) or the contribution of the increase in human capital to the country’s economic growth (Prados de la Escosura and Rosés, Reference Prados de la Escosura and Rosés2010). The impact of educational expansion on inequality has also deserved the attention of scholars in various fields, such as income or class inequalities (Carabaña, Reference Carabaña1983), gender inequality (Flecha, Reference Flecha2014), educational inequality itself (Bernardi and Requena, Reference Bernardi and Requena2007), or social mobility (Carabaña, Reference Carabaña1999; Beltrán Tapia and De Miguel Salanova, Reference Beltrán Tapia and De Miguel Salanova2021). These analyses of the relationship between educational expansion and inequality have mainly focused on inequalities between (individuals from different) families. Results from these studies indicate that expanding education in Spain led to a reduction in educational and other forms of inter-family inequality (Ballarino et al., Reference Ballarino, Bernardi, Requena and Schadee2009).
Nevertheless, one aspect overlooked in this literature is how educational expansion affected inequalities within families, particularly between siblings. The intra-family dimension of inequalities can be very relevant for understanding people’s socioeconomic position. According to Conley (Reference Conley2004), a significant part of the observed socioeconomic inequality in contemporary societies occurs between siblings from the same families. More generally, abundant evidence has been found in different historical periods and societies that dissimilar cognitive, psychological, behavioral, social, and economic outcomes are conditioned by sibship size and sibling position. Despite this wide evidence, in Spain, very few studies have considered the unequal social and economic outcomes achieved by siblings from the same families;Footnote 1 and even fewer studies have considered, at the same time, the two crucial factors that shape sibling inequality: birth order and number of siblings. Carabaña (Reference Carabaña2004) reported on the extent to which the number of siblings depressed educational levels and showed how this negative association increased over time in 20th-century Spain. In a later work, the same author (Carabaña, Reference Carabaña2013) demonstrated that the number of siblings was a relevant factor in the transition to upper secondary education. Requena (Reference Requena2022) has recently considered jointly how birth order and number of siblings influenced the educational attainment of those cohorts born in the first six decades of the 20th century. Reported findings from this study confirm the negative association of birth order and number of siblings with educational attainment among the observed population: both the number of siblings and birth order have been essential and relatively independent factors reducing the educational attainment of Spaniards and producing inequality between siblings. This latter work, however, does not provide a detailed analysis of the change over time in the associations between family position and educational attainment in families of different sizes.
This paper aims to address that task, filling the corresponding gap in the literature. Its main objective is to analyze whether the observed effect of family size and position on educational attainment in Spain has changed over the 20th century, precisely when the significant expansion of the educational system took place. Did inequality between siblings in Spain grow with the expansion of the education system? Was family position and size more or less determinant of educational outcomes due to the modernization of education? Although the literature on family position and educational outcomes is abundant, there are no systematic studies available that analyze the change over time of these effects. In principle, as formal education becomes more societally relevant because of the expansion of the education system, inequalities within families will be more pronounced. The effects of family position on educational outcomes and intra-family inequalities may become more visible when educational expansion provides access to formal education not only to a small, privileged minority but to most people. On the other hand, the expansion of education systems could have counteracted the impact of birth order and number of siblings on educational outcomes. In a recent study on the Swedish case (Barclay, Reference Barclay2018), it was found that during periods of educational expansion, younger siblings tend to spend longer in the educational system than older siblings.
To answer these questions, data from the Sociodemographic Survey (SDS) will be used (INE, 1993), and the cohort analysis approach will be adopted. The rest of this article is organized as follows. First, the background to the issue under analysis and the potential relevance of change over time are shown. Then, the interest of the Spanish case is presented and defended. Next, the data and methods used for the analysis are detailed. Afterward, the main findings are reported and, finally, the main conclusions are stated.
2. Background
The number of siblings and birth order can have consequences of many different kinds: cognitive, behavioral, psychological, demographic, social, and economic; but the educational effects stand out. There is a large body of evidence that family size and position are good predictors of educational outcomes in different countries and historical periods. At least in developed societies, the number of siblings and birth order are inversely related to educational attainment. The fewer siblings there are and the lower their ordinal position among the siblings, the greater and better the educational outcomes. Siblings from small families are more likely to achieve higher levels of education than siblings from large families. Later-born siblings perform worse academically than their earlier-born siblings (Barclay, Reference Barclay2015; Black et al., Reference Black, Devereux and Salvanes2005; Blake, Reference Blake1989; Conley, Reference Conley2004; DE Haan, Reference DE Haan2010; Hauser and Wong, Reference Hauser and Wong1989; Iacovou, Reference Iacovou2007; Kuo and Hauser, Reference Kuo and Hauser1997). Some of these studies, particularly those using within-family fixed effects models, allow for a causal interpretation of the relationship between birth order and educational outcomes. Regarding the causes, a recent study (Isungset et al., Reference Isungset, Freese, Reassen and Lyngstad2022) has shown that the educational gradient associated with birth order is not dependent on biological differences present at birth. The cognitive advantages of older children—visible at very early ages (Lehmann et al., Reference Lehmann, Nuevo-chiquero and Vidal-fernandez2018)—are due to differences in parental behavior in terms of quality time spent with children (Price, Reference Price2008), disciplinary strategies (Hotz and Pantano, Reference Hotz and Pantano2015), early maternal attitudes, and behaviors (Lehmann et al., Reference Lehmann, Nuevo-chiquero and Vidal-fernandez2018), and parental investments as measured by the quality of the family environment (Pavan, Reference Pavan2016).
Three main theories have tried to explain the negative effect of family size and position on educational outcomes. The sociological theory of the dilution of family resources (Blake, Reference Blake1981, Reference Blake1989) predicts that the allocation to each child of family money, energy, time, attention and care of parents, and other goods (such as space at home) is a negative function of the number of siblings: young children and/or those with more siblings are at a disadvantage compared to older children and/or those with few siblings because competition for family resources is necessarily greater among the former. The confluence theory (Zajonc and Markus, Reference Zajonc and Markus1975) argues that the intellectual development of siblings—and, by implication, their educational outcomes—depends on the family cognitive environment, which, measured by the average mental age of all members of the household, necessarily decreases with the arrival of each new child in the family. The economic theory of human capital (Becker and Tomes, Reference Becker and Tomes1976; Hanushek, Reference Hanushek1992) argues that the quality of children—measured by their educational attainment or future income—depends on the investments their parents make in them. Due to financial constraints, parents will choose between fewer children to invest more or more children to invest less. This trade-off between quality and quantity of children implies that the more children the parents have, the lower their average quality will be.
These theoretical frameworks are neither contradictory nor incompatible, offering complementary perspectives on the educational effects of sibling position and family size. However, all three frameworks may be too simplistic to understand how and why the relationship between family size and sibling outcomes varies over time and space, particularly if sibling outcomes are not solely or primarily dependent on the resources provided by their parents. Another possibility is that the connection between parental resources and child outcomes is mediated by factors that change from one historical or sociocultural context to another. Different scenarios may change the provision of non-familial resources for children. As a result of these processes, economic circumstances, cultural norms, demographic realities, social practices, or institutional settings set the range of variation in which family configuration operates and may induce different outcomes among siblings. The conditional resource dilution model (Downey, Reference Downey2001; Gibbs et al., Reference Gibbs, Workman and Downey2016) explicitly recognizes that these non-family resources can counteract the impact of family resource dilution. The association between children’s educational attainment and family size and position may be conditioned by non-family sources of resources. It may also happen that families with unequal social standing can mobilize resources for their children’s educational outcomes in different ways. Better-off families, for example, may seek to mitigate the negative effects of birth order or number of siblings by implementing compensatory strategies for children who are otherwise more likely to attain lower levels of education. Some studies (Grätz, Reference Grätz2018) suggest that the negative effects of higher birth orders on educational attainment are concentrated in families with low socioeconomic status. Still, evidence to the contrary has also been found (Härkönen, Reference Härkönen2014). Spanish parents in high socioeconomic statuses were able to limit the effects of dilution of resources induced by the number of siblings, but not the dynamics of the dilution associated with the birth order (Requena, Reference Requena2022).
Things are more complicated in the developing world, where the conditioning of educational outcomes by family size and position may operate differently and more complexly than in developed countries. While there are studies that find clear negative educational effects of birth order—for example, in Mexico (Esposito et al., Reference Esposito, Kumar and Villaseñor2020)—in other contexts the impact may be more intricate: for example, in sub-Saharan Africa (Kravdal et al., Reference Kravdal, Kodzi and Sigle-rushton2013; Tenikue and Verheyden, Reference Tenikue and Verheyden2010), Vietnam (Anh et al., Reference Anh, Knodel, Lam and Friedman1998), or Indonesia (Maralani, Reference Maralani2008; Feng, Reference Feng2021). There are other Asian, African, and American cases—India (Kumar, Reference Kumar2016), Kenya (Buchmann, Reference Buchmann2000), Botswana (Chernichovsky, Reference Chernichovsky1985), Brazil (Marteleto and DE Souza, Reference Marteleto and DE Souza2012)—where a positive relationship between a number of siblings and educational attainment has been found. However, even in developed countries, historical and social contexts matter (Bras et al., Reference Bras, Kok and Mandemakers2010). For example, community networks often linked to ethnic/religious groups may mediate the association between sibship size, family position, and educational outcomes. In some ethnic groups, where parental resources are complemented by extended family resources, size and position in sibling groups do not play a negative role in educational achievement (see Gibbs et al., Reference Gibbs, Workman and Downey2016 for different religious groups in the United States; or Shavit and Pierce, Reference Shavit and Pierce1991 for Arabs and Jews in Israel). Political institutions can play their part, as revealed by specific differences between Italy and France (Ferrari and Dalla Zuana, Reference Ferrari and Dalla Zuana2010). Weak or non-existent effects of family size on sibling outcomes in adulthood have been observed in countries with long social democratic traditions, where generous and universalistic social benefits are available (Park, Reference Park2008; Xu, Reference Xu2008).
3. Change over time and the Spanish case
The theoretical framework of conditional resource dilution invites consideration of changes over time in the relationship between a number of siblings, birth order, and educational outcomes. If the impact of family configuration on educational attainment depends on the circumstances in which families live, change over time offers a good opportunity to evaluate this theoretical framework and to pinpoint the contextual factors that may modify the basic relationship between family and education. The historical context of demographic transition, social modernization, and economic development seems particularly attractive in this respect as it implies changes that affect both the structure of families and their shifting societal circumstances. Without being exhaustive, two aspects of modernization processes are particularly relevant here. One is the expansion of education systems and the policies that made it possible. Another is the transformation of families in terms of their size, which followed the demographic transition and other related societal processes. These transformations raise important questions. For this study, the crucial one is how, if at all, the relationship between the number of siblings and ordinal position and education changes in the new context of small family sizes and expanded education systems typical of post-transitional populations.
The evidence accumulated on how the associations between intra-family characteristics and schooling changed over time due to modernization processes is not entirely conclusive. Gibbs et al. (Reference Gibbs, Workman and Downey2016) detected a decline in the association between the number of siblings and educational attainment during the first half of the 20th century in the United States and attributed this waning to the development of welfare policies supporting public education. Increasing public spending on educational programs reduced the constraints on families’ access to education and reduced the extent of the dilution of family resources among American families. China is another interesting case suggesting the same type of contextual influences. Lu and Treiman (Reference Lu and Treiman2008) found significant fluctuations in the effect of the number of siblings on educational attainment depending on educational policies. While egalitarian policies (during early communism and the Cultural Revolution) tended to reduce the effect of a number of siblings, competitive policies (in the pre-communist era and during post-Mao economic reform) increased it.Footnote 2 This argument is consistent with Park’s (Reference Park2008) study on the important role that public policies supporting childcare, universal child benefits, and large public spending packages on education and families play in mitigating the educational disadvantages of large families. To support the thesis, Park offers data on 20 OECD countries with different types of welfare regimes from the Program for International Student Assessment (PISA) database. The case of Hong Kong, previously analyzed by Post and Pong (Reference Post and Pong1998), is relatively similar and supports this thesis.
On the other hand, evidence from Thailand (Knodel et al., Reference Knodel, Havanon and Sittitrai1990; Knodel and Wongsith, Reference Knodel and Wongsith1991) and Taiwan (Hermalin et al., Reference Hermalin, Seltzer and Lin1982; Parrish and Willis Reference Parish and Willis1993) shows increasingly negative relationships between family size and educational attainment throughout socioeconomic development. Marteleto (Reference Marteleto2010) found that in Brazil number of siblings was inversely related to adolescent’s educational attainment and this inverse relationship became stronger for adolescents born after the demographic transition. Siblings from large families who grew up in the post-transitional environment of small families experienced greater relative disadvantage than their counterparts in the pre-transition cohorts. The overall reduction in family size did not contribute to reducing the disadvantage of siblings from large families, but instead increased it. At least in Brazil, the demographic transition seems to have intensified educational differences between siblings. An equally interesting case is Peninsular Malaysia, where Sudha (Reference Sudha1997), using data collected in the mid-1970s and late 1980s, found among younger generations a negative effect of the number of siblings on educational attainment that reversed the positive effect observed among older generations. However, when decomposing the analysis by ethnic groups, the author found that this negative effect for the whole population was only found among the Chinese and Indian subgroups, but not among the Malays, the beneficiaries of the state affirmative action policies for education. In other words, subsidizing the education of ethnic Malay families mitigated the dilution of resources among them.
The difficulties limiting the comparability of these studies stem mainly from the fact that they use different methodological designs and non-equivalent data. One obvious reason for the disparity between the cases, in addition to the large cultural and economic distance between the countries compared, is the different moments of expansion of their respective education systems. Take, for example, the case of the United States and Brazil. While the range of variation of the American cohorts analyzed by Gibbs and co-authors (Reference Gibbs, Workman and Downey2016) goes from 9.8 years of schooling at the beginning of the 20th century to 13.8 for those born in 1970–1979, the two Brazilian cohorts compared by Marteleto (Reference Marteleto2010) increased their years of education from 3.4 (born in 1963) to 4.7 (born in 1983). It is quite clear that the degree of expansion of the Brazilian education system in the 1960s was far from the American educational outcomes at the beginning of the century.
Another different limitation of these studies, also due to the characteristics of the available data, is that they only took into account family size, but not birth order. The two dimensions are closely related, but they are not the same and for both theoretical and empirical reasons need to be disentangled (Rodgers et al., Reference Rodgers, Cleveland, Van Den Oord and Rowe2000; Booth and Kee, Reference Booth and Kee2009; DE Haan, Reference DE Haan2010; Iacovou, Reference Iacovou2007). In a more recent study on the impact of educational expansion on the Swedish population, Barclay (Reference Barclay2018) considered both the number of siblings and birth order and estimated fixed-effect models (within-family comparisons). His paper finds that, although the net effect of birth order on educational attainment is negative, when educational opportunities increase as a result of system growth, this negative impact is cancelled out and even reversed by positive secular trends toward educational expansion. When opportunities for education expand, later-born siblings do better.
The Spanish 20th century provides a good context for assessing how the process of resource dilution is conditioned by changing historical circumstances. Throughout the century, Spanish society underwent a process of social, economic, political, and cultural modernization of considerable consequences. The economic and productive structure changed radically with the transition from an agrarian to an industrial and service economy (Nicolau, Reference Nicolau, Carreras and Tafunell2005). Economic growth (Prados de la Escosura, Reference Prados De La Escosura2017), although discontinuous, substantially raised the material standard of living of the people. Health standards improved, mortality was reduced, and longevity increased (Pérez Moreda et al., Reference Pérez Moreda, Reher and Sanz-gimeno2015). Fertility also declined after a moderate baby boom and the country ended up entering a regime of very low fertility (Reher and Requena, Reference Reher and Requena2015; Requena, Reference Requena1997; Requena and Salazar, Reference Requena and Salazar2014). The health transition and the fertility transition completed the demographic transition.
Two important correlates of the processes of economic development and social modernization in the country were (1) the decrease in family size because of the sharp drop in fertility and (2) the tremendous expansion of the educational system (Table 1). Families lost, on average, 22% of siblings. The increase in the educational attainment of Spanish population was impressive (Núñez, Reference Núñez1992; Viñao, Reference Viñao2004; Bernardi and Requena, Reference Bernardi and Requena2007). According to the SDS data (Table 1), the average number of years of education grew steadily across cohorts since the beginning of the century. Cohorts born in the mid-1960s multiplied their years of education by a factor of 3.Footnote 3 Despite the fact that Spain’s social history during the 20th century involved a great deal of complexity, educational change was dramatic and took place continuously, without setbacks. In terms of intergenerational change, the process of educational expansion did not stop as a consequence of the Civil War or Franco’s dictatorship.
Average years of education, number of siblings, and birth order by birth cohort in Spain
Values represented as means, standard deviations (SD), and coefficients of variation (CV). 1991 Sociodemographic Survey (Instituto Nacional de Estadística).
The tremendous structural changes that the economy, society, politics, and culture underwent in 20th-century Spain offer an ideal context for assessing historical changes in the relations between family size or sibling position and educational achievement. According to Öberg (Reference Öberg2017), this is one of the possible tasks that should be done to advance in the field given the model of conditional resource dilution.
4. Data and methods
Data for this study come from the SDS, a multipurpose survey conducted by the Spanish Statistical Office (Instituto Nacional de Estadística, INE) in 1991 to complement and validate the results of the census of that year (INE (Instituto Nacional De Estadística), 1993; Zárraga, Reference Zárraga2009). The SDS contains a big deal of information on households, relatives (parents, siblings, partners, and children), places of residence and mobility, housing characteristics, education, and economic activity of the Spanish population. Data are collected from individuals, not from whole families, meaning that detailed information on all members of the sibling group is not available; however, for each individual interviewed, it is possible to know how many siblings he/she had and what his/her birth order was. Although the data collection design was cross-sectional, the SDS was a life cycle survey that, through retrospective questions, records a rich information about the demographic, family, residential, educational, and work trajectories of the Spanish population. Besides reporting basic socio-demographic information, informants provided the levels of education they achieved and the number of years they spent in education. They also reported the number of siblings they had and the year of birth of all of them.
The whole SDS sample is made up of 157,100 people over the age of 10 and under the age of 91 who lived in Spain during the last quarter of 1991, when the information was collected. An analytical reduced sample was chosen for this study. This reduced sample is limited to cohorts born between 1901 and 1966 (i.e., population aged 25 to 91 at the time of the survey), under the realistic assumption that by the age of 25 the selected cohorts had completed all their education and that changes in their education levels after that point were not relevant or non-existent. Another reason for excluding cases from the sample is the nature of siblings. Only natural siblings of both parents were considered in the analysis—excluding half-siblings (1.53%), stepsiblings (0.23%), and adopted children (0.02%)—to make the sample more homogeneous and minimize additional confounding factors. The same birth order was assigned to twin siblings. At the end, the operational sample contained 120,779 individuals, with 63,039 women and 57,740 men.
The two main independent variables are birth order and number of siblings.Footnote 4 The dependent variable is the number of years of formal education completed, measured from the informant’s statement about the years he/she studied without counting interruptions.Footnote 5 Now, given the process of educational expansion experienced by the generations observed in this study, the dependent variable (years of education) is a moving target. So it is desirable to have a specification of educational attainment that somehow controls for its steady growth across cohorts (see Table 1). The solution adopted has been to standardize years of education with z-scores normalization by year of birth and sex. The transformed dependent variable will be the number of standard deviations from the mean years of education attained separately for women and men born in the same year.
As control variables to be included in the regression models, the year of birth of the informant, the duration in years of the interval subsequent to his or her birth, if cohabitation with the father ended before the age of 16, the age of father and mother at the time of birth, and the education of the father and mother (seven categories: no education, less than primary, primary, lower secondary, upper secondary, university first stage, university second stage) were considered. The social class of family of origin is measured by the occupation of the father when the informant was 16 years old (six categories: bourgeoisie, white-collar workers, urban petty bourgeoisie, farmers, urban working class, and agricultural workers). Finally, the region of residence (five categories: North, Centre-North, Centre, East and South) is included in the models as a population control variable to consider the territorial heterogeneity of the country and, in particular, the well-known and very persistent disparities in the aggregate education level of the different Spanish regions. Table 2 presents descriptive statistics corresponding to the analytical sample used.
Descriptive statistics. Analytical sample, N = 120,779. Means and proportions
Source: Sociodemographic Survey (Instituto Nacional de Estadística, 1993).
Given the type of dependent variable (years of education), several linear ordinary least square regression models have been estimated with birth order and number of siblings as the main independent variables, in addition to a vector with all the controls considered. However, this type of exercise must address the need to avoid confusion between birth order and number of siblings and estimate their effects separately (Kessler, Reference Kessler1991; Iacovou, Reference Iacovou2007; Booth and Kee, Reference Booth and Kee2009). A first element of confusion arises from the intrinsic relationship between the two variables. Obviously, the probability of occupying a certain order among children depends crucially on the number of siblings. This means that there is never complete independence between these two dimensions. The probabilities of occupying low orders (earlier-born children) are greater in small families, and vice versa, and the linear correlation between the two dimensions oscillates around 0.65 in conditions of low fertility (that is, low average number of siblings and low average birth order). This also means that although the correlation between number of siblings and birth order is not perfect, when these are considered in relation to third variables, compositional effects and spurious associations can occur between the two factors and the result of interest. In addition, due precisely to the functional form of the relationship between the two variables, it can be expected that, ceteris paribus, the cumulative differences in resources that siblings receive from their parents depending on birth order are smaller as the number of siblings is greater and, therefore, that the associations of the number of siblings or birth order with social outcomes are nonlinear (Blake, Reference Blake1981; Downey, Reference Downey1995).
As Iacovu (2007) argued, given the non-linearity of the relations of family size with the resources to be received by children, it is necessary to always analyze the associations of size and family position separately and allow for non-linearities when modelling these associations (Booth and Kee, Reference Booth and Kee2009). To disentangle the effects of number of siblings and birth order, the specification of relative birth order devised by Booth and Kee (Reference Booth and Kee2009) will be used. This birth-order index is defined as B = ϕ /A, i.e. the ratio of each sibling’s absolute birth order (ϕ) to each family’s average birth order, A, calculated as (N + 1)/2, where N is the total number of siblings. As the authors point out, by deflating the absolute birth order (ϕ) by the average birth order in each family (A) the initial correlation between ϕ and N is substantially reduced. By construction, this birth order specification effectively removes the correlation between number of siblings and birth order,Footnote 6 purges family size from birth order, and the results of estimations using it are analogous to fixed, within family effects models.
5. Results
Table 1 presents the data defining the basic context for understanding the change inter-cohorts in the association between family configuration and educational attainment. Cohorts born in the first six decades of the 20th century increased their years of education threefold, while reducing the number of siblings they grew up with by about 30%. However, while the distribution of years of education became increasingly uniform over cohorts, the distribution of the number of siblings maintained or even increased its dispersion across cohorts. As the century progressed, Spaniards studied for more and more years and educational inequality decreased. They also grew up in increasingly smaller families, but these families did not become more similar.
Table 2 summarizes the descriptive statistics of the sample used. The left-hand column presents the distributions of the independent variables; the right-hand column presents the distributions of the dependent variable (years of formal education completed). This right-hand column shows the associations that define the main patterns of educational inequality among these cohorts. Considering all cohorts, men were ahead of women by one year of education, but the female cohorts born in the 1960s balanced the disadvantage and eventually turned it around. Absent fathers (if the co-residence with the father ended before the child’s sixteenth birthday) reduced the total time spent in education by 1.4 years. In terms of social inequalities, the three variables defining the socio-economic status of the family of origin (father’s education, mother’s education and parent’s class of origin) show a clear gradient in the educational outcomes of these cohorts in the expected direction: the higher the socio-economic status of the parents or the family of origin, the more time spent in education and consequently the better the qualifications obtained. Children of farmers accumulated more years of schooling than children of day laborers, but less than children of urban working-class parents, a relatively anomalous but well-known result (Carabaña, Reference Carabaña1999). Moreover, the SDS data reflect well the north-south territorial gradient of educational outcomes in Spanish regions: people in the northern regions of the country had better educational outcomes than those in the south (Beltrán Tapia et al., Reference Beltrán Tapia, Díez-minguela, Martinez-galarraga and Tirado-fabregat2021; Reher, Reference Reher2022).
All these educational inequalities can be considered as inequalities between families. In terms of bivariate relationships, the difference between the educational attainment of children in families with two siblings and children in families with five or more children is almost three years (2.8). With the exception of only children—who performed slightly worse than children with one sibling—the more siblings, the fewer years of education. Special attention should be also paid to differences within families. As for birth order, the pattern is clear and without exception: the higher the birth order, the lower the educational attainment; moreover, if the relationship were perfectly linear, being born after another sibling represented an average loss of half a year of education. Given the high degree of correlation between birth order and number of siblings, it is also important to examine how birth order is associated with educational attainment within families of different sizes (Figure 1). These data show that both birth order and number of siblings were relevant factors in educational attainment in 20th-century Spain.Footnote 7 Only children have poorer educational performance than the eldest of two siblings, the best performing child of all birth orders observed. Birth order is inversely related to years of education in families of two, three, and four siblings. In families with five or more siblings, the elder sibling outperforms their lower-order siblings, but no other significant difference emerges. It must be stressed that these relationships observed in the Spanish case—and, in particular, complexity in families with large sibship sizes—are consistent with other studies conducted in developed countries using comparable data (Booth and Kee, Reference Booth and Kee2009; Iacovou, Reference Iacovou2007; Van Ejick and DE Graaf, Reference Van Ejick and DE Graaf1995).
Average years of education by sibship size and birth order, with 95% CI. Spanish cohort born 1900–1966.
Figure 2 presents these same relationships disaggregated for the six decennial cohorts into which the sample used was split. The figure allows, firstly, to appreciate without difficulty the educational expansion that Spanish society experienced throughout the 20th century. Invariably, each cohort achieved better educational results than the preceding cohort and worse than the subsequent cohort, regardless of the number of siblings or the order of birth. What is important now, however, is that some of the basic associations between number of siblings and birth order are observed in virtually all cohorts. An important caveat is in order here: associations are becoming more distinct and clearer in the younger cohorts. In the 1910–1919 cohort, an inverse relationship between number of siblings and average years of education can be seen, but this negative relationship does not hold for all birth orders. In fact, the inverse birth-order relationship with educational attainment can be only clearly observed from the cohort born in the 1930s onward. In the cohorts born later, the number of years of education attained becomes negatively ordered by number of siblings and birth order. At first glance, while the effects of number of siblings were more or less constant, those of birth order became stronger over the cohorts.
Average years of education by sibship size, birth order, and birth cohort, with 95% CI. Spanish cohort born 1900–1966.
Figure 3 presents the same data, but educational attainment is operationalized using a different metric. Educational achievement is now measured in standard deviation units from the mean years of education of individuals of the same sex born in the same year. By z-normalizing educational attainment, the effect of the constant expansion of the educational system, as well as the different pace of expansion between men and women, is cancelled out and the possible educational impact of the number of siblings and birth order is estimated under the counterfactual, but in this case reasonable, assumption that there was no inter-cohort change in the educational attainment of the Spanish female and male populations. In fact, the differences depicted in the figure oscillate around the value of 0 for all cohorts. The entire modelling exercise presented below uses standardized years of education as the dependent variable (
$\bar x\,$= 0; s = 1), and the results should be interpreted in units of standard deviation.
Average years of education (standardized by year of birth and sex) by sibship size and birth order, with 95% CI. Spanish cohort born 1900–1966.
Table 3 provides the estimates of three regression models, with sequential inclusion of covariates, for the years of education of these cohorts standardized by sex and birth cohort. All three models reveal the significant impact of birth order and family size on educational outcomes. In the model without covariates (1) and in the model with birth cohort only (2), the impact of number of siblings is larger than that of birth order.
OLS regression coefficients for standardized years of education
Source: Sociodemographic Survey (Instituto Nacional de Estadística, 1993).
When controlling for the other covariates, birth order becomes more important than the number of siblings: while each additional sibling subtracted 0.06 standard deviations of years of education, being born after another sibling reduced final educational attainment by 0.08 standard deviations. It should be stressed that the behavior of the covariates tends to be as expected, as can be seen, in particular, in the estimated coefficients for the variables referring to social position (father’s education, mother’s education, and social class of origin), to the different regions or the absent father.Footnote 8
Finally, to assess the change over time in the impact of the number of siblings and birth order on educational attainment, an additional regression model has been estimated that includes interactions of both variables with birth cohort, besides controls for the remaining covariates. Figure 4 shows the results of the model post-estimation, specifically the average marginal effects of number of siblings and birth order for the six cohorts.Footnote 9 The small but significant effect of the number of siblings grew with the change over time from the cohort born before 1920 onward. Among families of children born in the early 1960s, each additional sibling reduced the number of years of education by 0.08 standard deviation units, meaning that between cohorts born between 1960 and 1966 and those born before 1920, the effect had more than doubled. On the other hand, the birth-order effects are minor and not statistically significant among cohorts born in the 1910s, 1920s, and 1930s. However, among those born in the 1940s and, above all, in the 1950s and 1960s, birth-order conditions educational outcomes considerably more strongly than sibship size. In cohorts born in the first half of the 1960s, each change of one position in the sibling age hierarchy decreased years of education by 0.16 standard deviation units; in these cohorts, the birth-order effect was more than five times the effect recorded for cohorts born at the beginning of the century. At the end of the processes of demographic transition and educational expansion, birth order is a more important determinant of educational outcomes than the number of siblings.
Average marginal effects of sibship size and birth order on standardized years of education by birth cohort, with 95% CI. Spanish cohort born 1900–1966.
6. Discussion
This paper assesses the change over time of within-family inequality in the educational outcomes of Spaniards by measuring the effects of birth order on the number of years of formal education attained in families of different sizes. Although very little research has been done so far in Spain on this topic, the literature has established the impact of the number of siblings (Carabaña, Reference Carabaña1999, Reference Carabaña2004) on educational outcomes among the cohorts born in the first six decades of the 20th century, as well as the need to separate the effects of family size and birth order to understand these processes (Requena, Reference Requena2022) fully. Another study has found a significant degree of continuity of these inequalities in educational outcomes associated with birth order in families of different sizes that extends to cohorts born in the 1980s (see Requena, Reference Requena2024). All these findings, although scarce, invite us to wonder about the degree of persistence over time of within-family educational inequality associated with birth order. So far, no one has analyzed in detail the historical evolution of these mechanisms behind within-families inequalities and how the role of family size and birth order in the educational attainment of Spaniards changed over time. This is the main purpose of this study. To undertake it, the approach adopted is that of inter-cohort change analysis, and the data used come from the 1991 Sociodemographic Survey (INE, 1993; Zárraga, Reference Zárraga2009).
The turbulent societal contexts in which these cohorts lived are well known. During the period under analysis, Spain underwent tremendous economic, demographic, social, political, and cultural changes. Economic development, demographic transition, and social and cultural modernization dramatically changed the country. Economic development substantially improved the material standards of living in the country. Demographic transition reduced mortality and fertility to unprecedented levels. Social and cultural modernization greatly expanded the educational system. Because of these changes, the cohorts observed in this study grew up in smaller families than their parents and gradually but steadily increased, without reversals, their endowments of educational human capital. However, the expansion of the education system does not necessarily mean that educational inequalities decrease, and in fact, historical experiences of education growth in different societies have produced mixed results (Bernardi and Requena, Reference Bernardi and Requena2007). Interestingly, Spain is a clear case where the expansion of the education system was compatible with the reduction of educational inequalities (Ballarino et al., Reference Ballarino, Bernardi, Requena and Schadee2009). The level of between-families inequality in educational attainment declined significantly with the expansion of the system in Spain. This can be seen in Table 1 from the inter-cohort change in the coefficients of variation of the average years of education. The dispersion in the distribution of years of education clearly decreased over cohorts.
Did the expansion of the education system in Spain imply a parallel reduction in sibling inequalities within families? Two conclusions can be drawn from the Socio-demographic Survey data. First, educational differences associated with the number of siblings and birth order are easily discernible in all the cohorts observed. Regardless of which phase of the educational expansion process these cohorts experienced in their learning years, educational differentials associated with family size and birth order are visible in the expected direction: the smaller the family size, the more years of education attained; and the higher the birth order, the lower the educational attainment. In a society such as Spain’s, these differentials should be considered a very consistent and persistent regularity, a long-standing pattern capable of manifesting itself in very different historical periods. Second, after z-normalizing educational attainment—measuring it in units of standard deviation concerning the mean number of years of formal education by sex and year of birth to control for the process of educational expansion—the effects on educational attainment of both the number of siblings and birth order tended to increase across cohorts. The increase in the effect of sibship size across cohorts was slight but relatively constant. The impact of birth order was also small and statistically insignificant in cohorts born at the beginning of the century; however, starting with those born in the 1930s, it began to increase rapidly and steadily across cohorts. The change in educational inequalities between siblings associated with birth order can thus be answered conclusively: educational inequality within families dependent on birth order did increase with the expansion of education in Spain.
Spanish results can be interpreted within the model of conditional dilution of family resources associated with more siblings and high birth orders. According to the model (Downey, Reference Downey2001; Gibbs et al., Reference Gibbs, Workman and Downey2016), children’s educational outcomes do not depend exclusively on parental resources because the availability of non-family assets can counteract the dilution processes. Dilution relies not only on the number of siblings but also on other factors such as economic circumstances, cultural norms, demographic realities, social practices, or institutional settings where educational attainment occurs. Although this model is up-and-coming, very few studies have analyzed the change over time of the association between the number of siblings and birth order. A study on China (Lu and Treiman, Reference Lu and Treiman2008) showed how the effects of the number of siblings on educational attainment depended on educational policies. In the case of the United States, secular increases in state-sponsored investments in education reduced the impact of sibship size on educational attainment over time (Gibbs et al., Reference Gibbs, Workman and Downey2016).
The Spanish results point in the opposite direction: in Spain, unlike what is reported for the United States, the secular change indicates an increase, not a decrease, in the role of both the number of siblings and birth order in educational attainment. Results relatively similar to those found in Spain have been reported for urban areas in Indonesia throughout socioeconomic development among cohorts born between 1948 and 1977 (Maralani, Reference Maralani2008). Thailand (Knodel et al., Reference Knodel, Havanon and Sittitrai1990; Knodel and Wongsith, Reference Knodel and Wongsith1991), Taiwan (Hermalin et al., Reference Hermalin, Seltzer and Lin1982; Parrish and Willis Reference Parish and Willis1993), and Brazil (Marteleto, Reference Marteleto2010) are also similar. The disparity between the United States and Spain is remarkable considering that the cohorts analyzed were mostly born in the same years (1900s–1960s). However, some differences complicate the comparison between Spain and the United States. Firstly, the US study did not consider birth order (nor did the aforementioned China study), with all the estimation problems that this implies. Secondly, although the cohorts compared were born in the same years, the phases of expansion of the American and Spanish educational systems in which these cohorts were educated were very different, as well as the speed of expansion: while the cohorts born in the first decade of the century in the United States reached an average of 9.83 years of education, the Spanish cohorts achieved 3.85; for those born in the 1960s, the respective in the United States and 11.70 in Spain. Although the differences between the two countries were significantly reduced, initial levels in Spain were far below those in the United States. Thirdly, the education systems of the two countries and especially access to higher education levels were very different (Dill, Reference Dill, Tavares, Sá, Sin and Amaral2022).
It is beyond the scope of this paper to test the specific mechanisms that could explain the increasingly negative relationship between birth order and educational attainment in the observed cohorts in Spain. The expansion of the education system, which contributed to decreased educational inequalities between families, failed to mitigate inequalities within families because the differentials associated with birth order grew. This means that other factors may have acted in the opposite direction to expanding the education system. As pointed out by the proponents of the conditional family resource dilution model (Gibbs et al., Reference Gibbs, Workman and Downey2016), in addition to public investments in education, community structures, and extended family networks can substantially buffer the dilution of parental resources. Both factors became less prominent during Spanish socioeconomic development in the 20th century due to intense urbanization, massive internal migration (Silvestre, Reference Silvestre2002), and extensive family changes connected with the demographic transition (Reher, Reference Reher1996). A consequential aspect of these transformations is the change in family size. In the Spanish cohorts born in the first two decades of the century, only 11% of families had two siblings; in those born in the 1960s that proportion had risen to 23%. Seventy percent of the families of the generations at the beginning of the century had four or more children, a proportion that had fallen to only 47% of families in the 1960s. As noted above, the process of parental resources dilution is not linear because with the birth of a new sibling, a child who already has several siblings will lose less in terms of paternal inputs than a child with fewer siblings. Suppose small families, in which dilution is more pronounced, tend to become increasingly prevalent. In that case, it is possible that by a simple demographic composition effect, birth order will become increasingly important in educational outcomes.
In short, the results presented in this paper about the Spanish case do not contradict but reinforce the theoretical framework of conditional dilution of family resources. But they also imply an essential caveat. The societal context in which the dilution associated with sibship size and sibling position occurs matters. However, the elements that condition it respond to a complex etiology in which very diverse factors may intervene.
Acknowledgements
None declared.
Open access
