Most people agree that children from all walks of life should be able to reach their full potential. This is precisely the argument why all educational systems at some point sort students into different schools or classes based on their ability (varying from as early as age 10 in Germany to age 16 in, e.g., the US and the UK)1,2. Such between- or within-school ability tracks would make it easier to adjust the learning environment to the needs of students of diverse potential as opposed to differentiating the curriculum within the same classroom3. This presupposes that students are always sorted into the track that fits their potential ability, while there are concerns that sorting is also based on family background, especially when the tracking decisions are taken at a relatively young age. The weight of empirical evidence shows indeed that the influence of family background on educational attainment is stronger in educational systems where tracking occurs at a younger age, while average performance is not higher (for reviews of the literature, see refs. 2,4,5). These studies interpret this as evidence that tracking at a young age does not allow everybody to make most of their talents.

However, the way in which studies on inequality of educational opportunity typically test whether delaying tracking to a later age makes educational attainment depend less on family background and more on potential ability is to use parental socioeconomic status (SES) measures to capture family background and measures such as IQ to capture potential ability. This is problematic for several reasons. One issue is that it is virtually impossible to have measures for all the possible ways in which family background impacts educational attainment. In an attempt to get around this problem, researchers have used sibling similarity in educational attainment as an omnibus measure of family background6. But sibling similarity does not only reflect similarity in educational outcomes due to siblings sharing their family background, but also due to their genetic sharing. Sibling similarity thus mixes up the influences of family background and potential ability. Including measures of actual ability to capture potential ability is not a satisfying solution either. Similar to family background, it is difficult, if not impossible, to have measures for all the cognitive (e.g., verbal, information processing speed, memory) and non-cognitive (e.g., self-control, motivation, grit) abilities that are important in school7,8. Moreover, family influence starts at a very early age, possibly prenatally, so every test administered to capture ability will also reflect partly the influence of family background9.

This study makes use of a ‘genetics toolbox’ to address much of these problems. We analyze data from mono- and dizygotic twins to disentangle genetic influences from environmental influences shared by children from the same family and environmental influences specific to the individual. This approach is often referred to as the classical twin design, which allows for the estimation of heritability and the moderation of heritability and environmental impact by a measured variable10,11. It has been argued that heritability, or the variance explained by genetic influences, is a good overall measure for the opportunities that people have to realize their potential and shared environmental influence a good overall measure for the impact of family background12. If tracking at a later age increases the importance of potential ability and reduces the importance of family background, one would thus expect genetic influences to be larger and shared environmental influences to be lower in educational systems in which tracking occurs later.

Note that this hypothesis comes with caveats. First, genetic influence may not only reflect realization of “positive” potential but also of “negative” potential that can hamper education (think, e.g., of disorders), although positive traits seem to carry most weight13,14. Second, because the shared environment captures all environmental influences that make siblings more similar, it should be seen as a broad indicator of all the different ways in which family background matters, including parental resources and behaviors, neighborhood characteristics, and so on. On the one hand, this is a strength because it circumvents the impossible task of having to measure them all. On the other hand, downsides are that the shared environment forms a black box, and that it may capture things one does not want to. For example, without controlling for year of birth, the shared environment would reflect any differences in average educational attainment between age cohorts (for an excellent discussion, see ref. 15).

Although it is important to keep these issues in mind, others have suggested in a similar fashion that genetic influence should be larger and shared environmental influence lower in educational systems that promote equality of opportunity (e.g., see refs. 16,17,18,19,20). Empirical testing of this hypothesis is scarce and it has not been applied to tracking age practices. The first question we ask is therefore: how does the influence of genes and the shared environment on educational attainment depend on the age at which children are tracked? We define educational attainment as the level of secondary education, which is a strong determinant of final educational attainment.

If tracking at a young age generates inequality of opportunity, it is important to understand why this is the case. A key distinction in the literature is that between what has been labeled primary and secondary effects of family background. Family background matters for level of education first through impacting how children perform (in terms of test scores etcetera) (primary effects). On top of these performance differences, so even if children with different backgrounds perform equally, their levels of education tend to diverge because of different educational decisions (secondary effects). In our analyses, we separate how tracking age impacts these primary and secondary effects by distinguishing genetic and shared environmental effects related to performance from those net of performance.

The first reason why early tracking could amplify secondary effects is that it may increase the risk for children with a disadvantageous background to attain lower levels of education than would be expected based on their potential. We argue in this paper that if such hampering by disadvantage is the driving mechanism, a lower tracking age decreases genetic and increases shared environmental influence especially among high-performing students. Second, early tracking could introduce possibilities for advantaged children to attain higher levels than their potential merits. We argue that such compensation by advantage would lead to the opposite prediction: a lower tracking age decreases genetic and increases shared environmental influence especially among low-performing students. Because it is unclear if one of these two types of mechanisms is underlying secondary forms of inequality of opportunity, the second question we ask is: how does the moderation of genetic and shared environmental influence on educational attainment by tracking age depend on the performance level of children?

To answer these questions, we analyze data on N = 8847 twins from the Netherlands Twin Register. The Netherlands provides an excellent case to study the consequences of tracking. The tracking procedure starts at a relatively young age (at the end of primary school around age 12) into five tracks (the three lowest tracks prepare for senior secondary vocational training, a middle stream prepares for tertiary vocational college, and an academically oriented track gives access to university). Secondary schools, however, differ in how definite this tracking is. Some schools put students into one particular track immediately at the start of secondary school based on the track recommendation of the primary school teacher. Other schools do not decide immediately on the definitive track level of a student but first put students in a class that combines two or more levels. After observing students 1–3 years (depending on the school) in such a heterogeneous class, the secondary school teachers decide what track level is appropriate for a student. This variation allows us to study if delaying definite tracking to a later age has consequences for inequality of opportunity.

An advantage of studying tracking within one country is that other aspects of the educational system and country are held constant. It is not necessarily the case, however, that whether children are tracked immediately or later is completely exogenously determined. This means that the association of interest, i.e., between delaying tracking and genetic and environmental influences on educational attainment, could be (partly) spurious or suppressed. Dutch children, together with their parents, choose a secondary school from the pool that offers their track level. There are virtually no restrictions on choice (no catchment areas, low or no tuition, often many schools close by). Research shows that important criteria are distance to school, school quality, student composition, reputation, peer effects, pedagogical approach, and denomination, and some find that the weight of the criteria depends on parental SES and child’s performance level21,22,23. We are not aware of a study that explicitly examined how important a homogeneous versus a heterogeneous class is as a criterium. Although not the same, Herweijer and Vogels24 get somewhat close by asking how important parents find the number of tracks a school offers. They show that this criterium is relatively unimportant and unrelated to parental SES, denomination, political attitudes, and whether children go to a heterogeneous class or not. Although this is somewhat reassuring, attending a homogeneous versus heterogeneous class could still be correlated with other school characteristics that lead to selectivity. Van Elk et al. 25 report for those with a MAVO track recommendation (old label for VMBO-t) that children from higher-SES and more urbanized areas more often attend a heterogeneous class, but that none of the effects are significant in a multivariate analysis. Borghans et al. 26 show for those with a HAVO to VWO recommendation that higher-SES and higher-performing children are less likely to attend a heterogeneous class. All in all, the evidence does not point towards strong selectivity, but nevertheless we check whether two potential confounders, namely family SES and school performance of the child, influence the choice to delay tracking or not.

In the remainder of the introduction, we first provide a short theoretical background on how genes and the environment are thought to influence child development, and then we theorize how the tracking age of an educational system impacts these genetic and environmental influences.

In modern industrialized societies, heritability of cognitive ability as assessed by psychometric IQ tests, is estimated to be approximately 55% around the age of 12, i.e., 55% of the variation in cognitive ability is due to genetic differences between these children27. High heritability does not mean that cognitive development or other forms of development are coded in the genes and just emerge with maturation, leaving little influence of the environment28. Genetic differences between people only come to expression through people’s interaction with their direct environment29. To give an example: a child may have an aptitude for reading, but if there are no books around, this genetic potential will be left unrealized. Higher levels of heritability can thus be expected if the opportunities in the direct environment of children to develop their potential improve.

Transactional models argue that children select and are selected into different environments partly based on their genetic predispositions, and these environments in turn have an impact on their development28,30,31 In case of our reading example, it would mean that children with an aptitude for reading are more likely to pick up a book than children without this aptitude (active gene–environment correlation), and that parents are more likely to buy books for children with an eagerness to read than for children with little appetite for reading (evocative gene–environment correlation). This in turn means that the reading ability of children with a genetic predisposition for reading is stimulated more than that of children with a genetic makeup geared less towards reading. Based on this, it is expected that if the wider environment is such that children have autonomy in shaping their direct environment, genetic differences are allowed to be expressed in terms of developmental differences. Herd et al. 32 show, for example, that the genetic influence on educational attainment increased for women to become more like men’s during the period in which women’s access to education became less restricted in society.

An important societal context that impacts whether children realize their potential or not is the way the educational system is organized. Tracking children into different ability groups could in principle be a way to optimize children’s direct learning environment to their genetic predispositions. However, in practice, ending up in a certain track may not only be the result of genetic predispositions, but also in part of socioeconomic background33. In the exposition below, we hypothesize that the younger the tracking age, the less educational attainment is a result of children realizing their genetic potential and the more it succumbs to primary and secondary effects of family background.

Before doing so, it is good to clarify three things about the relation between genes and family background. First, while Jackson34 stresses that primary effects also include transmission of genetic material from parents to children, Boudon35 originally stressed only sociocultural factors. Parental SES can have a causal effect on children’s performance through these sociocultural factors, but not through the transmission of genes. Parental genes may causally influence parental SES (but not the other way around) and children’s performance (via children’s genes), which would create a spurious association between parental SES and children’s performance. When talking about primary (or secondary) effects of socioeconomic background, it thus makes sense to exclude genetic transmission mechanisms. Second, and related to the first, because parental genes may influence both parental SES and children’s genes, genes and family background can correlate (labeled passive gene-environment correlation: e.g., parents with an aptitude for reading do not only pass on these genes to their children but are also more likely to have books in the household)36. The presence of gene–shared environment correlation makes it more difficult to disentangle genetic and shared environmental influences, but in the methods section we discuss why we think that it is not very problematic in our case (see “Assumptions of the fitted twin models”). Third, based on theoretical ideas related to the ones we apply, it has been argued that children have more opportunities in high- than low-SES contexts to select and evoke their environment based on their genetic proclivities31. Genetic influences on cognition indeed have been found to be larger for children from high SES families in the US, but not in the Netherlands and other Western European countries37. For our measure of educational performance, heritability does not differ across parental SES either38. The potentially complicating matter of interaction between genes and parental SES is thus unlikely to play a role in the context of our study.

In the Netherlands, the tracking procedure at the end of primary school is based partly on a national standardized test score (CITO) and partly on teacher’s track recommendations39. There is a strong association between CITO-scores and socioeconomic background40. Moreover, De Zeeuw et al. 38 show that children with a low genetic propensity for educational attainment still tend to score high on the CITO-test if they are from high-SES background. This supports the idea that performance differences by family background do not only reflect differences in talent, but also the extent to which the home environment is conducive to children reaching their potential34. For example, compared to low-SES parents, high-SES parents have more economic resources to invest in e.g. private tuition and learning materials, and can transmit more cultural capital to their children, which helps them to do well in school41. The younger children are, the more their direct environment is shaped by their parents. When children grow up, they “gain increasingly more autonomy in selecting their peer groups, afterschool activities, academic courses, and other positive learning experiences” (Tucker-Drob et al., 2013, p. 351)31. This should enable these choices to be steered to a greater extent by genetic predispositions. In line with this, it has been shown that variance in cognitive ability due to the shared environment decreases from around 60% in infancy to virtually none in adolescence, while genetic variance increases from <25% to ~70%31,42 (if children’s own choices are not genetically steered, more autonomy would mean that shared environmental effects give way to non-shared environmental instead of genetic effects). A similar pattern could pertain to educational performance, which has considerable (genetic and shared environmental) overlap with cognitive ability in the Netherlands43. This would imply that, the younger the age at which performance differences are used to sort children into different educational levels, the less educational attainment will depend on genes and the more on the shared environment (i.e., primary effects become stronger). Primary effects do not only depend on the influence of family background on performance, but also on the influence of performance on attainment, but we do not have clear arguments why the latter would change with tracking age.

Even after taking educational performance differences into account, children from high socioeconomic background tend to enter higher tracks than those of low socioeconomic background (secondary effects)44,45. This also applies to the Netherlands: children with the same CITO-score receive on average a higher track recommendation and attend a higher track the higher educated their parents are46,47. Educational decisions made at young ages depend more on family background than those made at older ages48,49. At a younger age, other considerations than true potential are more likely to come into play when the teachers, parents, and children choose a track50. This could hamper those with a disadvantaged background and compensate those with an advantaged background.

An example of hampering is that teachers would (unconsciously) underestimate children who do not talk, dress, and behave according to the cultural codes that most teachers are familiar with, which tend to be the codes of the higher classes41. For the same reason, they would overestimate children from the higher classes that do possess this cultural capital (a form of compensation). Of Dutch teachers, 98% indeed indicates student behavior and 43% the home situation to play a role in their recommendation, especially for students whose CITO-scores are at the border of two tracks51. It is difficult to assess whether these considerations are the result of teacher bias, or whether teachers are correct in seeing them as important predictors of the chances for students to be successful in a particular track52. Either way, because children become more autonomous with age, teachers are expected to let the home environment weigh less in their track recommendations.

Moreover, parents from diverse socioeconomic background behave differently when guiding their children in making a track choice. Relative Risk Aversion theory proposes that all parents are concerned with avoiding downward mobility for their children53. Whereas entering one of the lower tracks would mean downward mobility for children of high-SES background, it does not for children of low-SES background. High-SES parents would therefore be more likely than low-SES parents to interfere in the educational careers of their children. For example, it is argued and sometimes shown that high-SES parents more often than low-SES parents put pressure on teachers to give a higher track recommendation51,52,54. If children are younger, they depend more on the involvement of their parents. This would make it more likely for low-SES compared to high-SES children to enter a track below their true potential (hampering), and more likely for high-SES compared to low-SES children to enter a track above their true potential (compensation). Additionally, higher tracks tend to take longer and are therefore (perceived to be) riskier, which applies especially to low-SES parents who are unfamiliar with the higher tracks themselves. This uncertainty enlarges if tracking occurs at a younger age because the timeframe ahead is longer, increasing the risk of myopic decisions for those of lower socioeconomic background55. As students grow older, they and their parents gain more information about the student’s abilities and chances of success, and students can increasingly make their own decisions. This makes it more likely for talented students from low-SES families to choose a high track even if it is unfamiliar to their parents56,57.

These different arguments for secondary effects predict, similar to the primary effects, that if tracking happens at younger age, the entered secondary track level depends more on the shared environment, curbing the expression of genetic differences between children. Because the entered track level is very decisive for the secondary educational diploma that children attain, we expect the same to apply to secondary educational attainment.

H1. The influence of genes on educational attainment is stronger and of the shared environment weaker if tracking is delayed to a later
age.

So far, we have implicitly assumed that early tracking increases both hampering and compensation to the same extent. But of course, it may also be the case that either of these processes is the main driver of secondary effects.

The arguments for hampering, so why disadvantaged children are tracked below their actual level, are more applicable if the educational performance of the children is higher. Teachers may find a lack of cultural capital especially problematic for low-SES children entering the highest levels of education. Furthermore, the track suitable for low-performing children is likely to be familiar to low-SES parents, but the higher the performance level of their children, the more likely it is that low-SES parents are unfamiliar with the track that is suitable. This implies that especially among the high performers, family background (shared environment) matters instead of actual potential (genes). Therefore, if early tracking mostly induces hampering, it is predominantly among the group of high-performing children that early tracking increases the importance of the shared environment and lowers that of genes.

H2a. The moderating effect of age of tracking on genetic and shared environmental influences on educational attainment (net of performance) is stronger if the performance level of children is higher.

The arguments for compensation, so why children with an advantageous background are tracked above their actual level, are more applicable if the performance of the children is lower58. The lower the performance of a high-SES child, the higher the risk of downward mobility, so the more actions high-SES parents are expected to take to prevent this. Contrary, if high-SES children have high performance, there is no need for compensatory actions because they will attain a high track anyway. Thus, in case of compensation, family background (shared environment) makes a difference especially among low performers. Therefore, if early tracking mostly amplifies compensatory actions, one would expect that early tracking increases the importance of the shared environment and lowers the importance of genes mainly among children with low performance.

H2b. The moderating effect of age of tracking on genetic and shared environmental influences on educational attainment (net of performance) is stronger if the performance level of children is lower.

It is not so obvious whether one would expect mostly hampering (H2a) or compensation (H2b) to be increased by immediate compared to delayed tracking in the Netherlands. It could very well be that both are affected to the same extent, in which case the impact of tracking age on genetic and environment influences would not depend on the performance level of children. Johnson et al. 59 show that genetic effects on educational attainment are smaller and shared environmental effects are larger in the US than in Sweden for low intelligence, but not so much for high intelligence. One could interpret these results as support for the compensation of lack of talent hypothesis (H2b) by arguing that Sweden’s school system provides more equal opportunities than the US school system (parallel to our argument that delayed tracking provides more equal opportunities than immediate tracking).

Leave a Reply

Your email address will not be published.