There exists a wide and divergent body of literature analyzing the effects of family structure on the educational outcomes of children. These studies attempt to determine which elements of the family environment are meaningful in shaping the educational success and cognitive development of children. Though the work on this topic is extensive, it has produced few consistent conclusions as to the type of family composition that yields the best educational outcomes. Additionally, while few would contest that the family is pivotal in determining the future potential of a child, there is no single, prevailing explanation of the mechanism through which family composition translates to different experiences educationally. As work in this area continues, various different explanatory variables have attracted interest as the driving force behind this observed relationship between family structure and education. Most earlier authors first investigated the effect of absolute family size on education, establishing a rather firm negative relationship between number of children and educational achievement. Subsequent work has drilled deeper into the family structure, considering the effects of birth order and sex composition in turn. Though several studies attempt to capture these particular nuances of the family, one noticeable void becomes apparent when surveying the literature. There is a dearth of research concerning the effect of birth spacing on educational outcomes. Further investigation in this vein would enable researchers to more thoroughly understand one additional layer of complexity of this enigmatic relationship.
Despite the extent of the literature on this topic, no single theory has gained prominence as a compelling and unshakeable explanation. However, two theories have gained prominence in authors’ attempts to explain these observed effects. The confluence theory holds that the addition of every subsequent child harms the other children in a family by depressing the average maturity level of the household. (Powell and Steelman) Because a child’s intellectual development is a function of the unweighted average of the intellectual levels of the household, additional closely spaced children will detract from this average. Accordingly, first borns and only children should benefit from time spent in households consisting of only two adults, as the intellectual climate is more stimulating. This effect does not entirely play out in the data. In fact, Black et al find an only child penalty when considering the effect of family size on number of years of education. Children in two-child families receive 0.26 additional years of schooling on average than only children according to this study. (Black et al) Many authors also often employ the resource dilution theory to explain the observed effects of family structure on children’s educational outcomes. According to this theory, the family acts as a conduit to transfer resources to the next generation. The level of this transfer per child depends both on the absolute level of resources in a family and the number of siblings with whom they must share this endowment. Not only does the addition of another sibling dilute a family’s economic resources but also cultural, social and intellectual capital. (Steelman et al) Thus, larger families are detrimental, as parents are able to allocate fewer resources to each child; and according to some studies, even those, which are made available, are less beneficial in large families than in smaller ones.
Scholars in this area have winnowed down the wide range of potentially relevant family characteristics and focused primarily on the size, density and sex composition of sibships and the ordinal position of the child in that sibship. A sibship is defined as the group of children all born to the same set of parents or a group of full siblings. Additionally, the density of a sibship refers to the number of children born within a given time frame or the number of siblings born within a given distance from a reference child. Accordingly higher density sibships imply closer spacing between births. The research most heavily concentrates on family size as an explanatory variable; however, more recent studies have expanded the scope of this focus to include the aforementioned variables as well. Existing work establishes a relatively firm inverse relationship between family size and educational outcomes for children. This observed negative correlation lends support to both the confluence and resource dilution theories, as resources are spread more thinly and the average intellectual level is lower in these larger families. A number of studies have found higher birth orders and higher density births to be associated with poorer outcomes as well, though these relationships are not as well established. Applications of both the confluence and resource dilution theories can again be employed to provide plausible stories for the effects of both birth order and density on educational attainment. Conversely, the effect of sex composition remains indeterminate. While the other elements of family composition fit neatly in the framework proposed by the field’s two prevailing theories, sex composition does not. However, this does not suggest that it should be discounted as a potentially important explanatory variable.
While extensive work has been done in this area, there is a noticeable lack of insight into how the spacing of births affects the outcomes of children. Though birth density serves as imprecise lens into the relationship between birth spacing and education, it is unable to capture the exact nature of this relationship. Density captures whether or not more closely spaced births are detrimental, but is unable to provide insight into impact of exact birth intervals. With this paper I hope to achieve this additional level of granularity. I would like to look more closely at the effect of the number of years between subsequent births on the educational outcomes of children. This seems like a highly relevant variable as it fits within both the resource dilution and confluence frameworks. Very closely spaced children detract from the intellectual climate of a household, as younger children are less able to benefit from the higher intellectual levels of siblings that are a substantial amount older than them. Also, closely spaced births act as a drain on household income as parents have to make financial outlays for their children all at once. Not only does this family arrangement dilute economic resources, but also closely spaced siblings reduce the amount of attention dedicated to each child, especially when children are younger. I plan to examine the effect of this variable on the level of educational attainment for children in the family. Family composition not only plays a role in determining the length of time an individual remains in school, but also shapes the cognitive development of a child. The latter necessarily partly informs the former, however, I do not plan to make this the primary focus of my study. Previous research shows that family composition has the stronger influence on educational attainment, measured as years of schooling, than on cognitive development, captured by variables, such as test scores.
Interest in the effect of the family environment and dynamic on the future outcomes of children has persisted for over a century, as scholars attempt to disentangle exactly what determines the ultimate success of a child. Given the length of time that this subject has garnered the attention of scholars, there is an impressive body of research aimed at determining which characteristics of a family are most pivotal in shaping educational outcomes of children. A large volume of this literature centers around the effect of sibship size, though other variables such as birth order, birth density and sex composition also make frequent appearances. Most studies analyzing the latter list of variables generally accept the existence of a negative relationship between family size and educational outcomes and tend to use this as a starting point for their own papers. Though there are several studies that refute this relationship, it remains widely accepted. Caceres-Delpiano (2006) finds that and exogenous increase in family size reduces the probability that a child attends private school by 1.2 percentage points. Like many other studies, this paper employs twin births as an instrumental variable to capture an exogenous increase in family size. The author’s primary focus is to determine how the addition of another child affects the level of investment in each child in a family, rather than focusing on the outcomes that result from these investment decisions. His primary findings indicate that an additional child both lowers the probability of private school attendance and reduces the mother’s labor force participation. (Caceres-Delpiano) While this workforce exit may translate to a lower level of economic resources, the author argues that this decline in income is mitigated by the additional time the mother is able to spend with the children. Though this paper demonstrates that additional children reduce the level of economic resources available to each child, parents may exit the workforce to compensate for the monetary loss. However, because the scope of this article excludes actual outcomes, it is indeterminate whether the substitution between market and home-based ameliorates the negative effects on the children.
Downey (1995) produces similar results as above, finding that children in larger households demonstrate lower levels of academic achievement, as the available financial and interpersonal resources decline with the addition of each sibling. Interaction terms for sibship size and level of parental resources suggest that even if the same level of resources are available to children of large families, they reap fewer benefits from these resources than children in smaller families.(Downey) This produces a doubly negative effect for children of large families, as they tend to both live in households with fewer resources and benefit less from any given level of resources compared to children of smaller families. Family size bears significance in the consideration of this topic, and this paper in particular, not only in its own right as a major element of family composition, but also in its implications for birth spacing. Though larger families are not necessarily more closely spaced, one would anticipate that this is generalization is fairly accurate due to the limited age range in which most mothers give birth. In order to have more children within this restricted time frame, the mother would presumably give birth in more frequent intervals.
Many studies look beyond the effects of family size in order to determine if this variable is capturing other nuances of family structure as well. Birth order is another variable that has been heavily studied. Black et al (2005) find that when controlling for birth order, the effect of sibship size becomes insignificant, replaced instead by large birth order effects. They conclude that higher birth orders result in poorer outcomes and that the last child suffers disproportionately to the rest of the children in the family. Though not a focus of this particular study, the authors do make cursory mention of birth spacing, stating that first children whose two immediate siblings are closely spaced fare worse as a result. (Black et al) Another study on the effects of birth order reaches a contrary conclusion, finding that higher birth order children fare better than their older siblings. According to Ejnaes and Portner (2004) parents make fertility decisions based on the genetic endowments of their children, and are most likely to stop having children once they have a child of high genetic ability. The authors find highly statistically significant support for this theory, estimating that the increase of one ordinal position in absolute birth order increases educational attainment by approximately 0.245 years. This would suggest that the last child tends to benefit from a better genetic endowment than his older siblings and thus outperform them academically. Also later birth order children tend to have older parents, who might in turn have more financial resources to allocate to these younger children. (Ejnaes and Portner)
Birth density and sex composition also make appearances in the literature though they are more seldom than the two variables discussed above. Birth density refers to the number of siblings born within a specified time frame from the reference child. Though this concept is similar to the idea of birth spacing, it fails to capture the exact distance of each child from his other siblings and provides a more imprecise measure of the effect of birth spacing. Generally, studies tend to find that closely spaced siblings fare worse than those spaced more widely. Theory suggests that wider spacing provides “breathing room” for parents, as they have time to financially recoup from the birth of the previous child before the next is born. In addition to financial recovery, a longer interval between subsequent births allows parents to recover from the mental and emotional strain of having an infant. However, there is an argument to the contrsry in so far that children spaced more closely are able to share material goods and therefore reduce the financial pressure on the family. Powell and Steelman (1990) find that having siblings within two years of each other significantly reduces both math and verbal scores, as well as GPA. Also children with closely spaced siblings recall being read to as children less frequently than those in more widely spaced families. This evidence lends credence to the resource dilution theory, as parents have less time to dedicate to each child, especially when they belong to a dense sibship. (Powell and Steelman)
Data and Model
In order to examine the effects of age spacing on the educational attainment of children, I will use data from the National Longitudnal Survey of Youth. This survey is comprised of questions pertaining to multiple dimensions of an individual’s family life, educational outcomes and personal characteristics. It was administered to a nationally representative sample of approximately 9000 individuals, aged twelve to sixteen in 1997. The available data covers the period from 1997 through 2008 and follows the original sample over time. Though variables for sibship size or birth spacing are not present in the raw data, these can both be constructed from other existing survey questions. In addition to collecting information regarding the respondent to the survey, the interviewer obtains basic information, including birth date and relationship to respondent, on every individual linked to a particular household. From this basic information I will be able to derive a measure of number of siblings, birth order and age spacing in a given family. Additionally, because the survey targets all children in a given household that fall within the specified age range, it will be possible to account for household fixed effects in these cases. The sample contains 3855 instances in which multiple siblings responded to the survey. Of these, 3134 were occurrences of two sibling respondents and the remainder included cases of three, four or five siblings. In these cases, I will be able to identify the effects of age spacing on the educational outcomes of siblings within the same family environment. Unfortunately, for those siblings that are not the immediate focus of the survey, educational data is not available. Therefore I will primarily be focused on the effects of age spacing across households, rather than within them. Ideally I would like the focus of the paper to be on within household variation, to better capture the influence of household fixed effects on educational outcomes, yet the available data limits this possibility.
While the NLS has several shortcomings, one advantage of this data is the quality of the educational data. The survey contains questions that ascertain not only the level of the respondent’s educational attainment, but also includes questions that better gauge the individual’s cognitive ability. Variables of particular interest include the respondent’s highest grade completed, whether or not the individual skipped or repeated a grade, and the respondent’s math and verbal SAT scores. Though I am primarily concerned with the effects of birth spacing on educational attainment, these additional variables to some extent control for the individual’s inherent cognitive ability. This in turn will sharpen the analysis of the effect of birth spacing on educational attainment alone, aside from its effect on the cognitive development of a child. An array of other control variables can also be found in the survey, including family income, race, sex and parents’ age and education. These variables constituted the standard range of controls in the existing literature on family structure.
A large volume of the existing work on this subject employs an instrumental variable approach in order to avoid potential omitted variable bias. Endogeneity problems may complicate the analysis, as an unobservable characteristic of the parents may both influence the family composition itself as well as the educational attainment of their children. Parents may easily manipulate certain elements of family composition in order to produce a family of desired size and spacing. The unobserved quality that induces certain parents to carefully plan the number and timing of births in their family may also influence the educational success of the children within that family. Though most individuals would likely choose to exercise some degree of control over their fertility outcomes, it is possible that those who are capable of doing so successfully better exhibit this unobserved trait. Therefore family size or widely spaced birth intervals may be capturing the effect of this unobserved quality rather than acting on their own to influence a child’s educational achievement. Because the timing and occurrence of a birth is never truly exogenous, the potential for bias is almost unavoidable without an instrumental variable.
Within the existing literature twin births and sex of existing children are the variables chosen most often for use in this approach. In the case of the former, the birth of the second child is a truly exogenous event, as parents are unable to anticipate whether or not a pregnancy will result in twins ex ante. The theory behind sex composition is somewhat weaker, but some authors argue that families with children all of the same sex are more likely to have an additional child. Though I would prefer to use twin births as an exogenous shock to family size, the data limits this possibility. Conversely, the sex composition of a given family is fairly easily discerned from the available information, though this instrumental variable may be relatively weaker than the alternative. While these are capable of predicting an exogenous change in family size, they do not provide an adequate replacement for birth spacing variables. It will be more difficult to identify a variable that influences the actual timing of births within a family. Due to the difficulty associated with identifying a strong instrument variable for spacing, I will likely attempt to mitigate the problem of endogeneity by controlling for various parental characteristics. The NLSY includes information concerning both parents’ levels of education, employment status, income, and the age of the mother at first birth. This wide array of parental traits can potentially capture the effect of unobservable characteristics that influence the timing of fertility decisions, as well as the educational attainment of children born to a given set of parents.
I intend to use a standard ordinary least squares regression to model the effects of birth spacing on a child’s educational attainment. The primary left-hand variable of interest will be highest grade completed, though I anticipate running the regression on test scores and whether or not the child repeated a grade as well. In addition to birth spacing variables, the regression will also control for sibship size and birth order in attempt to more fully capture the nuances of the family structure. The existing literature has also established family size and birth order as important determinants in the outcomes of children, further supporting their inclusion in the model. The model will also control for various characteristics of the individual child, as well as for parental traits and total household income. Depending on the final definition of family size, I will also include dummy variables to indicate if a particular child is adopted, or a step or half sibling. However, these children may be excluded from the final sample altogether in order to avoid the introduction of potential bias from the effect of differential treatment by parents or other siblings.
Initially I intend to break the data into various groups depending on family structure of the survey respondent. The primary group will consist of children with at least one younger and one older sibling, with additional groupings that will include children with only older or only younger siblings. This distinction will make it possible to determine not only the effect of the length of the birth interval between siblings, but will also capture the additional nuance of whether the sibling in question is older or younger than the survey child. This is of interest, as the length of the birth interval may have differential effects depending on whether the spacing variable is capturing the distance to a younger sibling versus that to an older one. Within these various groups the regression would control also for family size and birth order effects, in order to ensure that only pure spacing effects are represented by the coefficients.
Black et al. “The More the Merrier? The Effect of Family Size and Birth Order in Children’s Education.” The Quarterly Journal of Economics 120 (May 2005): Print.
Caceres-Delpiano, Julio. “The Impacts of Family Size on Investment in Child Quality.” The Journal of Human Resources 41.4 (Fall 2006): Print.
Downey, Douglas. “When Bigger is not Better: Family Size, Parental Resources, and Children’s Educational Performance.” American Sociological Review 60.4 (Oct. 1995): Print.
Ejnaes, Mette and Portner, Claus. “Birth Order and the Intrahousehold Allocation of Time and Education.” The Review of Economics and Statistics 86.4 (Nov. 2004): Print.
Powell, Brian and Steelman, Lala. “Beyond Sibship Size: Sibling Density, Sex Composition, and Educational Outcomes.” Social Forces 69.1 (Sept. 1990): Print.
Steelman et al. “Reconsidering the Effects of Sibling Configuration: Recent Advances and Challenges.” Annual Review of Sociology 28 (2002): Print.