Family background and child outcomes

An enduring debate in the sociology of education has centred on how best to capture variation in family resources and their effects on child outcomes. Several studies have pointed to the dominant role played by parental education in shaping child cognitive outcomes, with better reading and numerical test scores found among children whose parents have degree-level or higher qualifications (Hansen and Hawkes, 2009; Hansen and Jones, 2010). In contrast, other researchers, mainly but not solely economists, have focused on the effect of household income on child outcomes (Gregg and Macmillan, 2010). Increasingly, however, researchers have highlighted the value of adopting a multidimensional approach to understanding educational inequality in childhood (Sullivan et al, 2013; Washbrook et al, 2014; Hartas, 2015).

Perhaps more importantly, research has shown that the effects of different dimensions of family background on child outcomes are underpinned by different mechanisms. Parental level of education is an indicator of cultural resources and is likely to be reflected in the intellectual stimulation and learning environment at home as well as enhanced knowledge of the educational system. Family income captures parental financial resources at a given time, regardless of parents’ educational level. This is an important resource as it indicates the ability to pay for goods (more sophisticated toys, child’s own room and so on) and services/activities (private schooling or higher quality early childcare, extracurricular activities). Furthermore, the lack of financial resources might lead to financial stress and family conflict, resulting in poorer educational and socio-emotional outcomes for children (Conger and Conger, 2002).

Two important processes through which parental education and income can influence their children’s development in their early years are: the HLE and early childhood care and education (Melhuish, 2010). Therefore, we test the extent to which the HLE mediates differentiation by parental education in children’s cognitive outcomes and whether this mechanism is stronger for certain outcomes than others. Similarly, we examine the extent to which the type of early childhood care used explains differences by income and, again, whether its explanatory power differs depending on the outcome in question in two different countries with broadly similar institutional settings and highly comparable data: Scotland and Ireland.

Testing these mechanisms from a comparative and longitudinal perspective and across different outcomes has important implications for the nature and timing of family policy and interventions to close the gap in cognitive skills at the start of the educational career. More specifically, it can indicate the extent to which parental practices at home and the take-up of early childcare are responsible for existing inequalities and whether structural inequalities remain over and above these mechanisms. The last two decades of family policy in the UK and, perhaps to a lesser extent, in Ireland have focused on parental practices as the main strategy to address existing inequalities. However, a solid body of literature (Sullivan et al, 2013; Washbrook et al, 2014) has provided empirical evidence that structural inequalities persist beyond such practices and have argued that the focus on what parents ‘do’ can result in a culture of ‘blaming’ the parents rather than focusing on reducing broader inequalities (Hartas, 2014; Gillies et al, 2017). The framework used in this article can help shed more light on these dynamics by disentangling the direct and indirect effects of parental resources, whether the timing of ‘doing’ certain activities matters and whether these processes unfold similarly in two broadly similar contexts.

Home learning environment

A dominant strand of research on inequalities has built on Lareau’s (2003) concept of ‘concerted cultivation’ to look at the role of the HLE in fostering the kinds of skills and dispositions that are valued in the educational system. From this perspective, highly educated parents proactively provide more stimulating educational and cultural activities for their children, such as reading to them in the early years and paying for drama and music lessons when they are older. A good deal of recent research has documented the way in which the quality of this home learning environment (HLE) has not only short-term effects in supporting the transition to primary school but longer-term effects in influencing pathways through secondary education (Van Steensel, 2006; Sylva et al, 2010; Sammons et al, 2015). Using Growing Up in Scotland data, HLE was found to have a strong relationship with measures of vocabulary and picture similarities at three years of age, with a larger effect size than for other variables and completely mediating the effect of mother’s education (Melhuish, 2010).

Early childhood care and education

Children’s learning is also shaped by their access to early childhood care and education (ECE). The findings on the impact of non-parental care have been subject to much debate, with some studies indicating benefits of grandparent care for vocabulary development and of centre-based care for primary school adjustment (Hansen and Hawkes, 2009) while other studies point to non-significant differences across different types of care (for a summary, see Russell et al, 2016). There is a greater consensus on the influence of the quality of ECE on child development, with high-quality early years settings leading to better cognitive and non-cognitive outcomes among children (Sylva et al, 2010), especially for children from more socio-economically disadvantaged families (for example, Del Boca et al, 2017; McGinnity et al, 2017; Blossfeld et al, 2017).

Countries have been found to vary in the provision of free or subsidised ECE, the age group at which it is targeted, rates of take-up and the focus on learning activities (Gambaro et al, 2014; Blossfeld et al, 2017; OECD Family Database). These patterns suggest that the institutional context is likely to impact on the cognitive development of preschool children. However, to date (with the exception of Bradbury et al, 2012, and Doran et al, 2020) there has been little attempt to exploit the rich potential of child cohort studies as a basis for comparative research. Cross-national analyses have indicated that parental practices, such as reading to the child, are associated with cognitive development across almost all national contexts (Araújo and Costa, 2015). However, little attention has been paid to potential cross-cultural variation in such practices or the extent to which they respond to institutional factors (for example, preschool or primary school encouragement of home reading). There is suggestive evidence that, for example, rates of reading every day for three-year-olds are higher in Scotland than in the rest of the UK (Sullivan et al, 2010; see also Araújo and Costa, 2015, on diversity among a wider range of countries) but no obvious explanations of why this may be so.

In sum, this article contributes to the literature in at least two ways. First, it directly compares the effects of parental education and household income on several cognitive outcomes (vocabulary, language, reading and numbers) for five-year-old children and the extent to which these are mediated by the HLE and ECE respectively. Second, it takes a comparative approach, analysing not only the way in which institutions (such as ECE) shape child experiences and outcomes but also the extent to which parental practices to support home learning vary across contexts.

Data

There may be considerable challenges in conducting post hoc harmonisation of existing child cohort studies because of differences in the sampling approach, the timing of data collection over the life course and the measures used (see, for example, Waldfogel, 2013; Doran et al, 2020). The current analyses can overcome many of these challenges by using data from two cohort studies with remarkable similarities in the sampling method (a nationally representative survey of 9-month-old infants and their families) and the time intervals at which data were collected (9 months, 3 years and 5 years) as well as a strong similarity in the measures used.

The data for Ireland come from Cohort ’08 of the GUI longitudinal survey, based on a nationally representative sample of 11,134 children drawn from the Child Benefit register. We analyse data from the first three waves collected at 9 months (2010/11), age 3 (2013/14) and age 5 (2015/6). For Scotland, the data come from the MCS, a longitudinal study following the lives of young people born across England, Scotland, Wales and Northern Ireland in 2000–01 and selected through Child Benefit records. For Scotland, the initial achieved sample at the first wave was 2,336 families and 2,370 children. For the Scottish analysis, we use data from the first three waves: 9 months (2000/1), age 3 (2003/4) and age 5 (2006/2007).

In both surveys, at all three waves, detailed interviews were conducted with the primary caregiver, who was usually the mother. In addition, objective individual assessments of cognitive skills were conducted at ages 3 and 5 and a teacher’s survey was administered at the beginning of primary school, which included teacher reports on children’s skills. The objective assessments in both countries used the same tests (the British Ability Scale) and the teacher assessments used the very same instruments (see later). The measures of underlying mechanisms such as HLE and early years education were very similar in both countries, with only some regrouping of categories needed to achieve full comparability.

One limitation in cross-national equivalence relates to the timing of the surveys; the Scottish data relate to the pre-recession period while the Irish data cover the period from mid-recession to economic recovery. The recession resulted in a dramatic drop in average income levels in Ireland, though the scale of income inequality remained broadly stable (Callan et al, 2014). Nonetheless, we suggest that the timing differences should have little impact on our results since we use a measure of relative income (that is, quintiles) which should yield greater stability. Further, we do not expect differences in the impact of parental education to fluctuate over a decade. To test this assumption, additional sensitivity analyses (see Tables S5 and S6) have controlled for the self-reported impact of the recession on families in Ireland. The results show that the impact of the recession does not have a clear (linear) effect on child outcomes and the relationships between family background factors and child cognitive outcomes remain unchanged. Despite timing differences, the presence of the same outcome measures and very similar explanatory factors in both surveys provides a unique opportunity to look at the processes shaping inequalities in child outcomes in two contexts.

Measurement

The outcomes analysed in this study were measured at the age of five around the period of transition to primary education. The analyses focus on three outcomes reported by the teacher, who was seen as a good source of information on the extent to which children had the skills to engage in the curriculum: (1) language for communication and understanding (such as talking and listening confidently); (2) reading (including understanding story) and (3) numbers (including counting). These represent subscales of the measures used in the English Foundation Stage Profile that were replicated for the MCS teacher questionnaire in Scotland and were used for comparative purposes in the GUI study. The questions asked teachers whether or not the child had achieved specific competencies using a series of statements, scored from zero to nine. Given that the distribution was positively skewed, we dichotomised the scores distinguishing between those who exceeded the expectations for their development stage (that is, those who were assigned the highest scores of 9) and those who did not. In addition, we analysed a direct assessment of children’s vocabulary skills which was collected via the British Ability subscale of ‘Naming vocabulary’; scores were treated as continuous. In this wave of the survey, parents were not asked about their children’s cognitive skills but instead asked about socio-emotional difficulties.

The key concepts measured in our study are family background, based on which we assess differences in the achieved scores, and HLE and type of early years care and education, which both play the role of potential mediators in explaining any differences by parental education and income respectively. We also test the explanatory power of vocabulary at ages 3 and 5, separately, to investigate the extent to which inequalities emerge early (before age 3) or continue to develop during those two years.

First, family background was measured based on maternal education and household income recorded when the child was 9 months old (Wave 1). Maternal education was standardised to the following categories: ‘lower secondary or less’, ‘upper secondary’, ‘post-secondary’, ‘university degree or higher’ and ‘unknown or missing’. Finally, household income was measured using equivalised net income quintiles. For Ireland, missing cases for income are retained in the analyses but assigned to a sixth category (‘income missing’), which comprised 7% of cases. Net income provides a better measure of the economic resources available to families than gross income, especially in the context of two systems with different tax and welfare regimes. Using quintiles has advantages in capturing relative differences between the two countries and in allowing for potential non-linearity in effects within countries.

Second, HLE captured the frequency of different activities at age 3 (how often parents were involved in reading to the child, teaching the alphabet, counting, teaching songs, painting and screen time) and at age 5 (how often parents were involved in reading to the child, how often children were listening to music, painting, playing computer games, watching TV and going to the library). Unlike some earlier research (Melhuish, 2010, for example), the HLE indicators are included as separate items rather than as a composite index for two reasons. First, the internal consistency of the combined scale was relatively weak and the models would have lost explanatory power. Second, assessing the role of each specific HLE component may be more informative in understanding the mechanisms behind inequalities. For example, reading to the child might be expected to have a stronger effect on vocabulary skills than painting with them.

Third, childcare and early years education arrangements at 9 months and at age 3 were measured distinguishing between the following types: parental only, relative, childminder, centre-based and other. In addition, a binary variable captured whether the child had any exposure to non-parental care between ages 3 and 5.³

Several control variables were included in the analysis: gender, age, area density, child disability and school stage, given that some respondents were in the first and others in the second year of primary school.

Between-country differences in home learning environment and childcare arrangements

Income levels in both countries are transformed into quintiles for comparative purposes. However, the descriptive statistics indicate that maternal education is more polarised in Scotland than Ireland, with more at either end of the educational spectrum; 35% of mothers in Scotland had lower secondary education compared with 17% in Ireland and 32% of Scottish mothers had degree-level qualifications compared with 29% in Ireland (Table 2).

To what extent are there differences in the HLE? The descriptive statistics show a complex picture (Table 2). At age 3, the proportions being read to every day are more polarised in Scotland than Ireland, with levels of being read to every day and being read to infrequently both found to be higher. Between-country differences are variable across different aspects of the HLE, with doing the alphabet somewhat more common, and screentime slightly longer, in Ireland and counting, singing and painting more frequent in Scotland. By age 5, reading has become more frequent in Ireland than Scotland as has painting/drawing, playing with toys and playing computer games. However, educational outings (such as cinemas, concerts or art galleries) and library use were somewhat greater in Scotland. This latter pattern may be also due to a higher proportion of children living in urban areas in Scotland, which provide more opportunities for this type of cultural activities. Given our second hypothesis, this could create more room for parents to diverge in their childrearing practices in Scotland, and enable, to a larger extent, middle-class parents to practise concerted cultivation (Lareau, 2003) and embody valuable cultural capital (Bourdieu, 1986) in children’s lives early on. Ultimately, children’s familiarity with formal settings and structured activities may make the transition to primary school smoother.

To what extent are relative differences by maternal education different in the two countries? Figure 1 shows the difference in the percentage of parents reading every day to the child by maternal education. In both countries, three-year-olds whose mothers have degree-level qualifications are much more likely to be read to every day than those whose mothers have lower secondary education or less. In both countries, the social gradient in reading reduces between ages 3 and 5 but as the result of very different trends. In Ireland, this is mostly driven by an increase in involvement in reading for those whose mothers have lower levels of education. In contrast, in Scotland, there is a decrease in parental reading involvement among those with upper secondary and degree-level education. It appears therefore that the transition to primary school results in very different parental home-learning behaviours in the two contexts.⁵ Nevertheless, the age-specific gap in the frequency of reading by maternal education seems to be rather similar in the two countries contrary to our second hypothesis based on differences in the broader HLE.

View Full Size

Percentage of children being read to every day at 3 and 5 years of age by maternal education in Ireland and Scotland

Citation: Longitudinal and Life Course Studies 14, 1; 10.1332/175795921X16591139653249

• Download Figure
• Download figure as PowerPoint slide

Figure 2 shows the difference in attendance at centre-based childcare at 9 months and 3 years by household income levels. In Scotland, just under a third of those in the highest income quintile used centre-based care when the child was 9 months old; the numbers in the other income groups are too small to report. At age 3, there is a social gradient in the use of centre-based care, with the main distinction being between the top quintile and all others. In Ireland, the gradient in centre-based care at 9 months is stark, with a marked difference between the top quintile and all others. Centre-based childcare take-up increases by age 3 across all income groups. However, as in Scotland, the main distinction is between the top income quintile and the others.

View Full Size

Percentage of children in centre-based care at 9 months and 3 years by household income (quintile) in Ireland and Scotland

Citation: Longitudinal and Life Course Studies 14, 1; 10.1332/175795921X16591139653249

• Download Figure
• Download figure as PowerPoint slide

In sum, there are differences between the two countries in the level and social differentiation of both childcare and the HLE with Scotland showing a slightly higher take-up of centre-based childcare, particularly among the top-income households and stronger differentiation, especially at 9 months; these will probably be reflected in different patterns regarding the extent to which they explain inequalities in child outcomes, contrary to our ECE hypothesis based on the similarity of childcare policy characteristics in the two countries.

Differences by parental background in children’s cognitive skills at the beginning of primary school

Figure 3 shows sizeable differences between the top and the bottom categories by both maternal education and household income for almost all outcomes considered in this study. However, the income gap was larger than or at least equivalent to the parental education gap in both countries. The differences found were generally larger in Scotland than in Ireland, particularly for household income. In vocabulary skills, for instance, the score of children from the top income quintile was about nine points higher (almost a full standard deviation) than among those from the bottom quintile in Scotland compared to 5.5 points (that is, less than half a standard deviation) in Ireland.⁶ The exception to this pattern was larger differences in teacher assessment of language in Ireland than in Scotland.

View Full Size

Differences in children’s skills at age 5 by maternal education and income (Scotland and Ireland)

Citation: Longitudinal and Life Course Studies 14, 1; 10.1332/175795921X16591139653249

• Download Figure
• Download figure as PowerPoint slide

The role of childhood home learning environment and childcare arrangements in predicting children’s cognitive skills at the beginning of primary school

Reading to three-year-olds was the strongest predictor among the HLE indicators, particularly in Ireland. For the teacher-assessed outcomes, this was the case even when vocabulary test scores at age 3 were taken into account (Supplementary Material Tables S1 and S2). In contrast, in Scotland reading to three-year-olds showed a significant association only with their vocabulary at age 5 and their reading skills. The influence of other aspects of the HLE was much less clear-cut, showing relatively little consistency between time points (ages 3 and 5), particularly in Scotland. These findings are consistent with earlier analyses of the English MCS sample by Hartas (2011) who found little systematic impact of HLE activities other than reading on teacher-rated skills and competencies. In contrast, in Ireland, several HLE indicators proved to have a positive role on children’s cognitive skills (for example, alphabet, counting at age 3; playing computer games and library visits at age 5). In terms of timing of the HLE, both age 3 and age 5 activities played an independent positive role in Ireland, while in Scotland this was the case only for a couple of activities at age 3 and only one activity at age 5. Nevertheless, overall, in both Ireland and Scotland the findings point to a stronger direct effect of family background compared to the indirect effect mediated by HLE, an issue explored further later in this article.

In both countries, being in centre-based care at an early stage (9 months) was associated with enhanced skills in some domains (language and numeric skills in Scotland; vocabulary and reading in Ireland) (see Tables S3 and S4). This relationship was much stronger in Scotland (with differences of around 20 percentage points compared with around 4 percentage points in Ireland). Even after controlling for subsequent childcare take-up and for vocabulary at age 3, the initial strong associations of early centre-based childcare remained almost unchanged in Scotland, while in Ireland, effects remained only for reading skills. In Scotland, the relationship between being in centre-based childcare at 9 months or at age 3 and enhanced vocabulary appeared to merely reflect difference in family income; nevertheless, the take-up of early centre-based childcare shows an independent strong association when it comes to language and numeric skills. Another interesting cross-country difference was that only in Ireland was childcare provided by relatives or a childminder at 9 months and/or age 3 positively associated with vocabulary skills and teacher-rated language skills. Therefore, centre-based care in early childhood (at 9 months) had a stronger relationship with skill development in Scotland whereas in Ireland it proved to have a smaller impact at both 9 months and 3 years but non-centre-based childcare also mattered for child outcomes.

The role of home learning environment and early years care/education in explaining differences by family background

First, to what extent do home learning practices explain differences by maternal education? Tables 3 and 4 show the overall percentage variation mediated by the HLE in Ireland and Scotland respectively based on the KHB method. There is a clear differentiation in the cross-country patterns with Ireland showing a higher percentage mediated by HLE in the directly assessed vocabulary score at age 5 (for example, 36% versus 29% of gap between the highest and lowest educated) and Scotland showing a stronger mediating percentage in the teacher-assessed outcomes (ranging from 33% (Scotland) versus 19% (Ireland) in number skills to 82% (Scotland) versus 29% (Ireland) in language skills). There are some cross-national similarities in the timing of effects, with activities at age 3, particularly reading to the child, contributing much more to the overall indirect effect via HLE than activities at age 5 in both Scotland and Ireland. However, the overall indirect effects via HLE were statistically significant only in Ireland and only for vocabulary and language skills. In Scotland, none of the indirect effects via HLE was statistically significant. Overall, our results suggest that, contrary to our hypothesis, inequalities by maternal education in children’s vocabulary scores and language skills were more strongly mediated by HLE in Ireland than in Scotland. Thus, home learning appears to foster broader vocabulary development and (perceived) school readiness in Ireland but not to the same degree in Scotland.

In order to understand the extent to which inequalities at age 5 are an echo of earlier inequalities, we have also investigated the role of earlier vocabulary scores at age 3 in mediating inequalities in children’s cognitive skills at age 5. The cross-country pattern in the magnitude of the indirect effects already described also holds for the mediating power of earlier vocabulary with 73% of maternal education being mediated jointly by HLE, control variables and vocabulary scores at age 3 in Ireland versus 57% in Scotland. Yet, the corresponding mediation percentage in teacher’s assessed skills was much higher in Scotland (for example, 84% versus 30% in reading skills; 60% versus 29% in numeric skills). This time, all the indirect effects were statistically significant for all outcomes in Scotland (for the gaps between the highest and lowest educated groups), while in Ireland, the indirect effect of early vocabulary scores was only statistically significant for reading skills but not for the other outcomes. This suggests an interesting timing pattern in how inequalities unfold in the two countries, with earlier inequalities at age 3 being more consequential for inequalities at age 5 in Scotland than in Ireland. This, in turn, implies that the inequalities are more deeply rooted in earlier childhood in Scotland and the degree of cognitive stimulation in early childhood before age 3 is crucial in alleviating inequalities on school entry.

A second key question investigated in our study is to what extent does the type of early childhood care and education received explain differences by family background? Generally, ECE mediated a much higher percentage of household income inequalities in Scotland than in Ireland (Tables 5 and 6). The mediation effect of ECE was strongest for language (50% in Scotland compared with 13% in Ireland), followed by numeric skills (25% compared with 15% in Ireland) and reading (19% compared with 17% in Ireland). The extent to which ECE mediated income differences in vocabulary scores was much lower, below 10% in Ireland and below 5% in Scotland for any of the income quintile gaps included in the table. The indirect effects via ECE were statistically significant for all outcomes in Ireland (for several income gaps) while in Scotland, the indirect effects via ECE were statistically significant only for language and numeric skills and only for the gaps between top and bottom income quintiles. This pattern reflects more polarised practices in Scotland. Therefore, our ECE hypothesis about no country differences in the mediating role of ECE is not confirmed, with ECE a more important mechanism through which income inequalities are reproduced in Scotland than in Ireland. The different patterns shown for actual and teacher-perceived skills suggest that ECE provides higher-income families with some advantages in terms of school readiness (rather than vocabulary development per se). The analyses show important cross-national differences in the relative timing of the impact of ECE. In Scotland, early childcare, particularly centre-based care, has the strongest indirect effect for all of the outcomes analysed. In Ireland, although smaller, the mediating impact of ECE is more cumulative, being more evenly spread across the three different stages considered in our analysis.