Growing Up in Scotland: changes in language ability over the primary school years

4 Factors Associated with Improvement

4.1. Introduction

This chapter looks at factors present in children's lives over the primary school years which may help or hinder their language development. It seeks to identify characteristics and circumstances that might help to improve children's language ability relative to their peers.

A previous GUS report (Bradshaw, 2011) demonstrated that children's vocabulary ability differs according to social background upon entry to primary school. In a similar vein, the previous chapter showed clear differences in vocabulary ability according to a number of social background characteristics towards the end of primary school (in Primary 6). However, the analysis undertaken thus far does not allow us to explore whether social background is associated with a relative change in ability over the period - in other words, whether children from less advantaged backgrounds improve at a faster, similar or slower rate than children from more advantaged backgrounds. It also does not allow us to determine whether each characteristic is associated with language ability independently of the other characteristics. For example, it is unclear whether the differences seen by income are driven by other differences among children from different income groups such as the parents' level of education, their parenting practices, and/or the child's experience at school.

The emphasis in this chapter is on factors associated with this relative change in language ability over the primary school period, rather than with ability at a single time point. After outlining the key factors considered in the analysis, the chapter briefly considers the relationship between expressive vocabulary ability at the two time points considered in the analysis, namely around the time the GUS children started school and again when they were in Primary 6. Then, drawing on univariate linear regression models fitted for several characteristics, circumstances and experiences (outlined in Table 4-1 below), it explores which (if any) are associated with a relative change in language ability between the two time points - that is, over the primary school years. Next, it draws on multivariable regression models to explore which of the factors found to be associated with a relative change in expressive language in the initial analysis remain associated with a relative improvement or decline in ability once other known differences between the children are taken into account. Finally, the chapter explores whether any associations found vary according to the children's social background - more specifically, according to their parents' level of education. This is done through the fitting of interaction effects to the final multivariable regression model^[7].

4.2. Factors considered in the analysis

The analysis considers a range of different factors which are known or expected to be associated with children's language development. These are listed in Table 4-1 below; further details are provided in Appendix A^[8].

The factors explored in the analysis have been selected for the following combination of reasons: existing research has shown associations with children's language ability; they are considered likely to play an important role in children's lives over the primary school period and thus be (directly or indirectly) important for their language development and GUS has collected data suitable for exploring them.

Table 4‑1 Characteristics, circumstances and experiences considered in the analysis

Variable	Age/stage of child when measured	Source (parent/child/administrative data)
Child's gender	Age 10 mths	Parent
Social background and location
Highest level of parent education	Age 5/P1	Parent
Annual household income	Age 5/P1	Parent
Level of area deprivation (SIMD)	Age 5/P1	Parent
Urban vs small town or rural location	Age 5/P1	Parent
Other household factors
Whether languages other than English spoken in the household	Age 5/P1	Parent
Whether parent reported any literacy issues	Age 4	Parent
Parent mental wellbeing	Age 5/P1 and P6	Parent
Whether parent has limiting long-term health problem	Age 5/P1 and P6	Parent
Child health and development
Whether child has a limiting long-term health problem	Age 5/P1 and P6	Parent
Whether child has above average levels of social, emotional and behavioural difficulties	Age 5/P1 and P6	Parent
Significant life event occurred
Whether child experienced parental separation or re-partnering	Age5/P1, Age 6/P2, Age 8/P4 and P6	Parent
Whether child changed school	Age5/P1, Age 6/P2, Age 8/P4 and P6	Parent
Whether child experienced adverse life event (death of a parent or sibling; a parent in prison; being in care; or a parent losing their job)	Age5/P1, Age 6/P2, Age 8/P4 and P6	Parent
School
Child's feelings about school	Age 8/P4	Child
Size of school	Age 5/P1	Administrative data
Whether denominational school	Age 5/P1	Administrative data
Proportion of children registered for free school meals	Age 5/P1	Administrative data
Parenting and parent-child relationship
Warmth of parent-child relationship	Age 8/P4	Parent
Parental interactions with child's school	Age 8/P4	Parent
How often parent helps child look for school-related information	Age 8/P4	Parent
Child home reading in last week	Age 8/P4	Parent
Parent belief that they can influence child's achievements at school	Age 8/P4	Parent

Including a measure of the child's gender in the analysis allows us to examine whether there are differences in the level at which boys' and girls' vocabulary improves (or declines), relative to their peers', during the first five years after they start school.

As in chapter 3, household income, area deprivation and parental education are considered as measures of social background. To ensure sufficient base sizes for the type of analysis undertaken, rather than break parents' level of education into four groups as was done in chapter 3, the measure of parental education used in this chapter simply identifies whether or not at least one parent or carer in the household was educated to degree level (or above).

A measure of whether the child resided in an urban, small town or rural area provides a means to explore whether geographical location appears to be associated with children's language development. Any such association may arise through differences in general lifestyle, but may also, for example, arise through differences in school experiences in urban, small town or rural communities, as the size, resources and ethos of schools based in these different areas are likely to differ (e.g. Commission on the Delivery of Rural Education, 2013). To support the analysis, areas in the Scottish Government's six-fold urban-rural classification^[9] were grouped to form a binary variable comparing children living in large and other urban areas with those living in small town or rural areas (see Appendix A).

Bi- and poly-lingual children can have slightly delayed language development, but then catch up with their peers as they grow older (Cattani et al., 2014). Exploring whether languages other than English are spoken in the household at the time the child enters primary school allows us to gauge whether children in bi- or poly-lingual homes appear to be more or less likely to see an improvement in their language skills during the first years of primary school, relative to their mono-lingual peers. The analysis also considers parental literacy, as a parent experiencing literacy issues may impact negatively on their child's language development. For example, parents who have literacy issues may be less inclined to read with their children or engage in other educational activities which may help improve their child's language skills.

Other household factors such as the parent's mental wellbeing, and whether they had a limiting long-term health problem are also explored. Parental mental health and wellbeing has been shown to be strongly associated with children's cognitive development at an early stage (e.g. Marryat and Martin, 2010; Barnes et al, 2010). Compared with parents who have higher levels of mental wellbeing, someone who suffers from poor mental health may have less energy and/or capacity to engage in activities with the child known to improve language development. Parents or carers who have a long-term limiting health problem may be similarly inhibited in their parenting activities and/or there may be further impacts on the home environment such as financial constraints resulting from lower earnings and/or additional costs associated with their health problem.

Children's health and social development are also considered following other evidence linking these to cognitive outcomes (e.g. Gregg and Washbrook, 2011). Having a limiting health problem or social development difficulty may severely affect a child's language development in several ways, depending on the issue in question.

Experiencing significant changes or events can have a substantial effect on children and may, directly or indirectly, influence their language development. For example, changing school often carries with it not only changes in surroundings and staff but also in peer relationships, and may cause a general sense of upheaval which is potentially detrimental to children's learning. Changes in their home environment caused by parents separating and/or step-parents moving in has also been shown to be associated not just with children's wellbeing, but also with their cognitive outcomes (Chanfreau et al., 2011). Furthermore, experiencing an event such as the death of a parent or sibling; a parent being in prison; spending time in care; or a parent losing their job is likely to cause upset and distress which could have indirect effects on children's learning, for example through being off school for a prolonged period or simply being unable to concentrate. The analysis considers three measures related to significant changes in the child's life over the primary school period: whether they experienced parental separation or re-partnering; whether they changed school; and whether they experienced a significant adverse life event such as the death of a parent or sibling, a parent being in prison, spending time in care or a parent losing their job.

In this report we are particularly interested in factors which are potentially modifiable in the short to medium term through relatively discrete initiatives. That is, factors which can be influenced through dedicated changes to policy or practice either at a national level or through targeted interventions and initiatives. As such, aspects of the child's schooling and parenting practices are of particular interest here.

As noted in section 1.3, current education policies in Scotland highlight the importance of children's mental wellbeing for their academic achievement. How a child feels about school is likely to impact on their learning experience and the way in which they engage in school activities - and, ultimately, on their learning outcomes. Conversely, a child's skills and abilities are also likely to influence how they feel about school. Either way, it is useful to understand how, if at all, children's feelings about school may be associated with aspects of their cognitive development - including their expressive language ability - as they progress through school.

Understanding more about which (if any) aspects of a child's school experience are associated with higher levels of improvement is something which is of obvious interest to policy makers and may help focus both policy making processes and further research efforts. The analysis considers several school-related measures taken from both survey and administrative data:

• A measure of the cohort child's enjoyment of school around the time they were in Primary 4 (aged just under 8 years), based on the child's own report.
• Information about the child's Primary 1 school, obtained through linkage with administrative records. Specifically:
  • the size of the school;
  • whether the school was denominational or not;
  • the proportion of children in the school who were registered for free school meals (25% or more compared with less than 25%)^[10].

Children attending large primary schools are likely to have very different experiences than those in the smallest schools, with potential benefits and drawbacks of each. The experience of attending a denominational school may also differ to that of attending a non-denominational school. Furthermore, in statistics of school performance, denominational schools have been found to perform particularly well (see e.g. Andrews and Johnes, 2016; Hinchliffe and Bradshaw, 2015). As for the proportion registered for free school meals, this is a commonly used indicator of the level of poverty among the pupils attending the school. In Scotland, eligibility for free school meals has now been extended to all children in Primary 1 to Primary 3, however, at the time the children in BC1 were in Primary 1, eligibility for free school meals was still determined based on need, primarily though the family's eligibility for and receipt of certain benefits. These measures are obviously not a comprehensive set of indicators of the child's school experience and for a significant minority of cases no information was available on measures of school size, denomination or the proportion registered for free school meals. Nonetheless, they do provide measures of selected aspects which may, if nothing else, suggest possible fruitful directions for future research.

The relationship between parenting and children's development has received much attention in recent years and existing research has shown numerous links between a range of parenting and home learning activities and children's cognitive development (e.g. Waldfogel and Washbrook, 2010; Bromley, 2009; Melhuish, 2010). Among policy makers there has been a particular interest in measuring and encouraging 'parental engagement' in their child's school and education. These efforts are targeted at parents, encouraging them to engage in educational activities with their children at home, as well as at schools and teachers to ensure they are maximising the opportunities for parents to be meaningfully involved in their child's schooling (e.g. Scottish Government 2017b).

The analysis considers the following measures of parenting and home learning activities and characteristics - all are based on data collected from the cohort child's parent around the time the child was aged just under 8 years old, i.e. when most children were in Primary 4^[11]:

• Warmth of the parent-child relationship
• Parental engagement in child's schooling, including:
  • parent interactions with the child's school (those with 7 to 10 different types of different interaction compared with those with 6 or less);
  • how often a parent helped the child look for information about what he/she was learning at school, for example at the library or on the internet;
  • a measure of the extent to which the parent believes they can influence their child's achievements at school (those holding very positive beliefs compared with those holding less positive, neutral or negative beliefs).
• Home learning activities: how many days in the last week the child read or looked at books at home (children doing so at least 6 days per week compared with those doing so less often).

The measure of parent interactions with the school is a count of the number of different activities the parent reported to have engaged in in the two years before the interview^[12].

4.3. Vocabulary ability upon entry to primary school and in Primary 6

Before considering which factors may help or hinder a relative improvement in language ability, it is worth exploring the relationship between the two measures of vocabulary ability used in the analysis - that is, the measure obtained around the time the children started primary school, and then when they were in Primary 6.

As shown in Table 4-2, we see a strong relationship between the two standardised scores, with around 17% of the variation in standardised vocabulary scores in Primary 6 explained by the variation in scores at the start of primary school. In other words, a substantial proportion of the differences in children's expressive language ability at the time they are in Primary 6 appears to be explained by their ability around the time they started school. This also indicates that children's language ability at primary school entry is closely related to their ability towards the end of primary school. Nonetheless, the proportion of variation explained is not as large as has been found in some other studies (see e.g. Goodman, Gregg and Washbrook, 2011), and a rather large proportion of the variation in language skills at the time children were in Primary 6 does not appear to be explained by their earlier ability - at least the way it is measured here. On this point, it is worth bearing in mind that the analysis uses two different measures of expressive vocabulary, something which (despite the use of 'standardised' scores, as outlined in section 2.2) is likely to have introduced higher levels of uncertainty in the analysis than if the exact same measures had been used at both time points. Even with this caveat, however, the results suggest that although children's expressive language ability around the time they start primary school appears to play an important role in explaining their level of ability towards the end of primary school, other factors are also important.

Table 4‑2 Standardised expressive vocabulary score in Primary 6, by standardised expressive vocabulary score in Primary 1

	Standardised expressive vocabulary score (Primary 6)
	p-value*	Regression coefficient**	95% Confidence interval***
			Lower	Upper
Standardised expressive vocabulary score (Primary 1)	.000	0.431	0.385	0.476
R squared	0.172
Weighted base	2698
Unweighted base	2726

* All figures quoted in this report have a margin of error because they are estimates based on a sample of children, rather than all children. The p-value is an estimation of how likely it is that we would find a relationship in our sample of children if there was no actual relationship in the population (i.e., broadly speaking, among children in Scotland who are the same age as the GUS children but who are not part of GUS). Thus, the smaller the p-value (p<0.05), the more confident we can be that our results are likely to apply to children in Scotland more widely.

** The regression 'coefficient' illustrates the relative level of change (positive or negative) in language ability score at P6 if score at P1 is increased by 1 unit. A significant (p<0.05) positive coefficient denotes a relative improvement in ability score and a significant negative coefficient denotes a relative decline in ability score for every one-unit increase in P1 score.

*** The 95% confidence interval is an indication of the level of uncertainty in the coefficient estimate.

4.4. Univariate regression analysis

The following sections explore the extent to which a range of characteristics, circumstances and experiences are associated with a relative change in vocabulary ability over the primary school period.

The first stage of the analysis considers the relationship between each factor - that is, each of the characteristics, circumstances and experiences set out in Table 4-1 - and a relative change in language ability between the start of primary school and Primary 6. This is done by fitting separate linear regression models for each factor with the standardised Primary 6 vocabulary score as the dependent (outcome) variable and the Primary 1 standardised vocabulary score as an additional independent variable (co-variate). This allows us to assess the relationship between each factor of interest and a relative change in language ability between the two time points^[13].

Table 4-3 lists the factors which the analysis showed to be associated with a relative change in language ability over the period (when other differences between the children are not controlled for). Only factors where the relationship with a relative change in vocabulary ability is statistically significant at the 10% level are included in the table^[14]. A '+' in the 'Direction of change' column indicates a positive relationship between having the characteristic in question and a relative improvement in expressive language ability; conversely, a '-' indicates a negative relationship and a relative decline in language ability. For example, the '+' for 'Higher household income' indicates that children in higher income households improved at a higher rate than children in low income households (the reference category).

Table 4‑3 Factors individually associated with a relative change in expressive vocabulary ability (not controlling for other differences)

Factor	Direction of change
Social background and geographical location
Higher income household (vs low income)	+
In less deprived area (vs most deprived)	+
Parent/carer educated to degree level (vs not degree educated)	+
Live in small town or rural area (vs urban)	+
Child development
Child has above average social, emotional or behavioural difficulties (vs average or below)	-
School
Child's primary 1 school has high % of children eligible for free school meals (25% or more vs less than 25%)	-
Parenting
Parent reads with child at least 6 days per week (vs less often)	+
High number of parent interactions with school (7-10 vs 6 or less)	+
Parent holds strong positive belief they can influence child's achievement at school (vs less positive, neutral or negative beliefs)	+

Table 4-3 shows that over the course of the primary school period, children in more advantaged circumstances improved at a higher rate than their less advantaged peers. In relation to income, the analysis shows a clear positive relationship with children in the higher income groups improving at a higher rate than those in the lowest income group. Meanwhile, children living in less deprived areas were more likely to improve than those living in the most deprived areas. Those with degree-educated parent(s) improved at a higher rate than those whose parent(s) did not have a degree and children living in small town or rural areas showed higher levels of improvement compared with their peers living in urban areas.

Looking at the child's social development, those who were reported by their parent as having above average levels of social, emotional and behavioural difficulties saw lower levels of improvement in their vocabulary ability, compared with those who had only average or below average levels of difficulties. None of the other measures of household factors nor the child's health or significant changes in the child's life were found to be associated with a relative change in language ability over the period. Neither were the child's feelings about school, the school size, nor school denomination. However, children who attended a school with more than 25% of pupils registered for free school meals were more likely to see a relative decline in their expressive language skills compared with those attending a school with a lower proportion registered for free school meals.

Three measures of parenting were individually associated with a relative improvement in vocabulary. First, children whose parents read with them at least 6 days a week around the time they were in Primary 4 saw higher levels of improvement compared with children whose parents read with them less often. Second, children whose parents reported a high number of interactions with the child's school improved at a higher rate than children whose parents reported fewer interactions with the school. Third, children whose parents held strong positive beliefs in their ability to influence the child's achievement at school improved more than children whose parents held less positive beliefs.

As already noted, for each of the factors explored, the analysis carried out here did not take into account other differences between the children. Thus, the relationships outlined above may have arisen because of other differences between the children particularly given how these factors tend to vary by social background. This question is explored as part of the multivariable analysis outlined below.

4.5. Multivariable regression analysis

The next stage of the analysis involved entering the statistically significant factors listed in Table 4-3 above, as well as the child's gender, into a single multivariable regression model. As for the regression models outlined above, standardised Primary 6 vocabulary score is included as the dependent (outcome) variable and Primary 1 standardised vocabulary score as an independent variable (co-variate). This approach allows us to explore the extent to which each factor remains associated with a relative change in ability once these other known differences are controlled for.

This analysis showed^[15] that only four factors remained associated with a relative change in vocabulary ability once other factors had been controlled for:

• Parental education - children with parent(s) educated to degree level or above saw higher levels of improvement than those whose parent(s) had lower levels of education.
• Urban or small town/rural location - children living in small town or rural areas improved more than those in urban areas.
• Child's level of social, emotional and behavioural difficulties - children with above average levels of difficulties were more likely to see a relative decline in ability than their peers with average or below average levels of difficulties.
• Home reading - children who read or looked at books at home at least six days per week improved at a higher rate than those who looked at books less frequently.

A final model was fitted with only those factors which were statistically significant in the multivariable analysis described above, as well as the child's gender. The results are outlined in Table 4-4 below.

4.6. Variation by parental education

The final stage in the analysis explored whether any of the associations found vary according to the parents' level of education. This was done through fitting so-called 'interaction effects' between parental education and each of the remaining variables to the final regression model outlined in Table 4-4^[16]. None of the interaction effects were statistically significant. This suggests that the relationship between a relative change in ability and each of the factors identified at the earlier stages of the analysis and outlined above - i.e. urban or small town/rural location; the child's level of social, emotional and behavioural difficulties; and home reading - does not vary systematically according to parents' level of education.

Table 4‑4 Factors associated with a relative change in vocabulary ability (controlling for other differences) - final model

	Change in expressive vocabulary ability
	P-value	Coeff*	95% confidence interval
			Upper	Lower
Child's gender (ref=Girl)
Boy	.119	.063	-.017	.142
Highest parental level of education (ref=Below degree)
Degree	.001	.143	.061	.225
Location (ref=Urban)
Small town or rural	.013	.124	.027	.220
Child level of social, emotional, behavioural difficulties (ref=Average or below)
Above average level of difficulties at one or both time points	.002	-.179	-.290	-.068
Home reading in last week (ref=5 days or less)
6-7 days	.012	.118	.027	.209
Unweighted base	2726
Weighted base	2698

* In this table the regression 'coefficient' illustrates the relative level of difference (positive or negative) in language ability for each sub-group as compared with the reference sub-group. A significant (p<0.05) positive coefficient denotes a greater improvement in ability score and a significant negative coefficient denotes a lower change in ability score when compared with the reference sub-group. The reference sub-group is indicated in brackets.

4.7. Summary

This chapter has shown that, on average, when considered on an individual basis - that is, when not taking into account any other known differences - children living in higher income households, those living in areas with lower levels of deprivation, and those with parent(s) educated to at least degree level saw higher levels of improvement in their expressive language ability relative to their peers than those in the lowest income households, those in the most deprived areas, and those whose parent(s) do not have a degree.

Furthermore, it identified four experiences and circumstances which were associated with a relative improvement in children's expressive vocabulary over the primary school years even after a range of other known differences were controlled for:

• Having parent(s) educated to degree level or above
• Living in a small town or rural area
• Not having above average levels of social, emotional or behavioural difficulties
• Reading or looking at books at home at least 6 days a week when aged 8/in Primary 4

The analysis showed no indications that these relationships vary systematically according to parental education, suggesting that these factors are associated with improvement for all children irrespective of whether their parent or parents are educated to degree level or not.