beta

You're viewing our new website - find out more

Publication -

Growing up in Scotland: changes in child cognitive ability in the pre-school years

Published: 6 Jun 2011
Directorate:
Children and Families Directorate
Part of:
Children and families, Education
ISBN:
9781780452005

This report examines whether the gap in cognitive ability between children from different social backgrounds changes between ages three and five and which factors influence improvement in cognitive ability.

95 page PDF

4.8 MB

95 page PDF

4.8 MB

Contents
Growing up in Scotland: changes in child cognitive ability in the pre-school years
CHAPTER 3: FACTORS WHICH HELP OR HINDER IMPROVEMENT

95 page PDF

4.8 MB

CHAPTER 3: FACTORS WHICH HELP OR HINDER IMPROVEMENT

3.1 Introduction

Children who differ according to the level of their parents' education also differ in a range of other ways such as their family characteristics, living circumstances and their experience of parenting behaviours. Many of the dimensions along which these families differ are also known to impact on children's early cognitive ability (Washbrook and Waldfogel, 2010). It is not possible to assume, therefore, that improving educational qualifications amongst parents alone would close all or most of the education-related gaps in cognitive ability because some of those gaps are created by the different experiences of children whose parents have different levels of education. For example, better vocabulary ability amongst children with degree-educated parents is known to be, at least in part, a function of higher levels of parent-child reading amongst these parents. Thus, improving vocabulary ability could require increasing parent-child reading as well as improving parental educational qualifications.

This chapter examines the extent to which other factors that exist in children's lives contribute to improvement in cognitive ability and may help explain the developmental gap between those with poorly educated parents and those with highly educated parents. In so doing the analysis will permit some suggestion of where policy interventions designed to maximise children's cognitive ability during the pre-school period could be focused.

Existing research on child cognitive ability identifies a range of factors experienced by young children that impact upon their cognitive development. Given that these factors are related to level of cognitive ability at a single time point, we are anticipating that they may also be related to how children's cognitive ability changes over time - particularly during the pre-school period. The factors can be summarised across a range of 'domains':

  • Demographic
  • Family composition
  • Parenting styles
  • Experience of childcare and pre-school education
  • Child health and development
  • Parenting support
  • Maternal physical and mental health
  • Economic and material circumstances

These domains have been selected for three main reasons: first, other research has shown that the characteristics, circumstances and experiences they represent are associated, in different ways, with early cognitive ability; secondly, they cover a large part of the important experiences of children's early lives; and thirdly, GUS has collected data suitable for exploring them. In the following sections, the evidence which links measures within each of these domains to children's cognitive development in their early years will be cited.

A particular aim of this report is to identify factors which may lead to a narrowing of the gap in children's cognitive ability at age 3 during the pre-school period - factors which are associated with a relative improvement in ability from age 3 to age 5. Thus the analysis is not looking for factors necessarily associated with higher or lower cognitive ability - but those which are associated with change in ability, particularly a positive change from earlier scores. Whilst a certain characteristic may predict higher ability at age 3 or at age 5, it will not necessarily be associated with a change in scores in the pre-school period. However, we are assuming that the same factors are associated with both level of ability and change in ability to some degree.

The effect of factors in each domain was explored using multivariate analysis 12 . The results of this analysis allow us to determine which characteristics, circumstances and experiences of children's lives were independently associated with a relative improvement or decline in cognitive ability in the pre-school period after controlling for level of parental education. In addition, by looking at whether these factors weaken the strength of the relationship between parental level of education and cognitive ability at age 5, it is possible to measure whether variations in the additional factors are behind some of the education-related differences. That is, for example, to see whether some of the difference in ability by parents' education actually occurs because children whose parents have different levels of education have different experiences, circumstances or relationships.

3.2 Key findings

  • Compared with children whose parents are degree-educated, those whose parents have no qualifications are more likely, amongst other things, to have younger mothers, live in lone parent families, experience lower levels of home learning activities and household rules, to have had a low birth weight, poorer general health, and a mother who smokes.
  • Some of these differences in circumstances and experiences of children from different educational backgrounds explain part of the education-related gaps in their cognitive ability and there are a number of factors which appear to impact on change in cognitive ability over and above parental level of education. Indeed, only degree-level education continues to have an independent positive effect on change in either ability after the full range of other factors is taken into account. This suggests that much of the difference in ability amongst children from lower educational backgrounds is explained by differences in the home and external environments, and parenting experiences of the children in these groups.
  • Changes in vocabulary ability during the pre-school period are more strongly related to aspects of the child's home environment and the choices and behaviours of parents than external influencing factors such as pre-school education. After controlling for parental education, greater consistency of parenting, stronger parent-child attachment, attendance at ante-natal classes and breastfeeding were each independently associated with a relative improvement in vocabulary ability in the pre-school years.
  • Early language development is also important - those children who display better communicative skills at an earlier stage are those who are more likely to see their skills improving during the pre-school period. It would appear generally beneficial therefore, to seek to improve children's communication ability from the very earliest stages and establish better skills earlier in order to ensure continued positive language development.
  • Parenting and the home environment were also associated with change in problem solving ability; a higher frequency of home learning activities and being breastfed were each independently associated with a relative improvement in problem solving scores. External factors were also related to changes in this ability. Attending a private nursery school for pre-school education and some experience of primary school 13 were both associated with positive development whereas not attending pre-school and living in an area in the most deprived quintile were associated with a relative decline in ability.
  • The significant variables present a complex picture of the numerous elements of children's lives which, taken together, can influence their cognitive development. Influencing just one factor is unlikely to generate any change in children's ability.

3.3 Domains of influence on cognitive development

First we examine how demographic characteristics vary by levels of education, then we will look at whether they are independently associated with change in cognitive ability and whether they explain any of the education-related differences.

3.3.1 Demographic characteristics

A range of evidence exists demonstrating the importance of demographic characteristics - including gender, parental ethnicity and maternal age - in predicting pre-school cognitive test scores. Previous analysis of GUS data, for example, has shown that at age 3, on average girls scored significantly higher than boys in both vocabulary and problem solving assessments (Bromley, 2009). In the same analysis, children with younger mothers - particularly mothers who are aged under 20 at the child's birth - are shown to have lower scores than are those with older mothers. Similar patterns are observed at age 3 in other comparable UK data from the Millenium Cohort Study which demonstrates that such demographic differences persist at age 5 and age 7 (Hansen, 2008; Hansen et al, 2010). The MCS data, and a considerable range of other research (Magnuson and Duncan, 2006; Keels, 2008) also show difference in early test scores between children of different ethnic backgrounds, particularly in the US.

Table 3.1 displays the demographic characteristics of children in each education group.

  • There are few differences in the gender split of children from different educational backgrounds.
  • There is a greater concentration of children with a non-white parent amongst the lower educational groups with fewer represented in the higher education groups. Eight per cent of families with no qualifications have at least one parent who is non-white compared with 3% of those with Higher Grades or upper level vocational qualifications.
  • Maternal age at the child's birth shows the most stark differences by level of education. Younger mothers were considerably more likely than older mothers to be among those with lower qualifications. Just 6% of mothers with degree level qualifications were aged under 25 compared with 33% of those with no qualifications.

Table 3.1 Selected demographic characteristics by parental highest level of education

Parental highest level of education

Demographic characteristic

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

Child's gender

Male

52%

51%

55%

49%

52%

52%

Female

48%

49%

45%

51%

48%

48%

Parental ethnicity

Both parents white

92%

89%

97%

97%

94%

95%

At least one parent non-white

8%

11%

3%

3%

6%

5%

Maternal age at child's birth

25 or older

67%

53%

57%

73%

94%

75%

Under 25

33%

47%

43%

27%

6%

25%

Bases

Weighted

222

201

769

1170

1222

3589

Unweighted

145

153

660

1202

1431

3596

A summary of the demographic domain regression is included in Table 3.2 14 . The results suggest that demographic factors are more important for change in vocabulary ability than for problem solving. After accounting for parental education, having non-white parents and a younger mother at birth are both associated with a decrease in vocabulary ability during the pre-school period.

Table 3.2 Demographic domain linear regression - summary results

Demographic characteristic

Cognitive ability

Knowledge of vocabulary

Problem solving

Sig.

Direction of change*

Sig.

Direction of change*

Child's gender (ref: male)

Female

NS

NS

Parental ethnicity (ref: both parents white)

At least one parent non-white

0.01

-

NS

Maternal age at child's birth (ref: 25 or older)

Under 25

0.03

-

NS

Parental level of education (ref: no qualifications)

Lower SGs or VQs or 'Other' quals

NS

NS

Upper level SGs or intermediate VQs

<0.01

+

0.02

+

Higher Grades or upper level VQs

<0.01

+

<0.01

+

Degree level academic or VQs

<0.01

+

<0.01

+

Weighted base

3324

3336

*A plus sign (+) indicates relative improvement in ability score and a minus sign (-) indicates relative decline in ability score for children in the various sub-groups as compared those in the reference sub-group. The reference sub-group is indicated in brackets. Where the variable is not significant, direction of change has not been included.

The effect of education persists after the domain variables have been added to the model - this is illustrated in Figure 3-A and Figure 3-B. The graphs display the values of the standardised regression co efficient (which measures the strength of the association) between parental level of education and cognitive ability at age 5. The solid line displays the values in a model with just education and ability score at age 3 as explanatory variables. The dotted line displays the values after the domain (in this case, demographic) measures have been added. If the dotted line falls below the solid line, this indicates a weakening of the association between education and change in ability after the domain measures have been added suggesting that differences in the domain measures by level of education help explain some of the association between education and change in ability. In contrast, if the dotted and solid lines show little separation, this indicates that the domain variables explain little or none of the association between education and change in ability.

The results in Figure 3-A and Figure 3-B show that, for each ability, both the solid and dotted lines are very close together indicating that there is little change in the association between parental education and cognitive ability after the demographic characteristics have been added, and thus that these factors explain very little of the education differences.

Figure 3-A Associations between parental education and change in vocabulary ability before and after taking account of demographic characteristics

Figure 3-A Associations between parental education and change in vocabulary ability before and after taking account of demographic characteristics

Note: reference category is 'no qualifications'.

Figure 3-B Associations between parental education and change in problem solving ability before and after taking account of demographic characteristics

Figure 3-B Associations between parental education and change in problem solving ability before and after taking account of demographic characteristics

Note: reference category is 'no qualifications'.

3.3.2 Family composition

The particular make-up of a family - for example, the number of parents and siblings a child lives with, and their birth order - is known to be related to children's performance on cognitive assessments. Analysis of GUS and MCS data shows that children in lone parent families, and those with two or more siblings tend to have a lower score than children in couple families and singleton children or those with just one sibling (Bromley 2009; Hansen 2008). Table 3.3 shows how these characteristics vary by parental education.

Table 3.3 Selected family composition characteristics by parental highest level of education

Parental highest level of education

Family composition characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or Intermediate VQs

Higher grades and upper level VQs

Degree level academic & vocational qualifications

All

Family type transitions,

10mths to 34mths 15

Stable couple

38%

57%

58%

78%

94%

75%

Couple who separated

6%

7%

7%

5%

3%

5%

Stable lone parent

47%

28%

26%

10%

2%

14%

Lone parent who re-partnered

9%

9%

8%

7%

1%

6%

Number of children in household at 10mths

One

29%

32%

35%

36%

31%

34%

Two

30%

45%

39%

45%

50%

44%

Three or more

42%

23%

26%

18%

19%

22%

Cohort child's birth order

First born in household

41%

44%

48%

51%

54%

50%

Not first born

59%

56%

52%

49%

46%

50%

Bases

Weighted

230

206

779

1172

1228

3621

Unweighted

150

157

668

1204

1437

3621

  • Almost all (94%) families with a degree-educated parent are headed by a stable couple, compared with around two-fifths (38%) of families where parents have no qualifications. Families in the lower educational groups are significantly more likely to be stable lone parents.
  • The key variation in number of children by education is on the proportion of households with three or more children This is significantly higher in the no qualifications group than in all other groups and is lowest for the degree-educated group. There is very little difference in the proportion of singleton households by education.
  • Among the lower qualified groups, the child was significantly more likely to be the oldest in the household - 59% were the first born in the no qualifications group compared with 46% in the degree group. This reflects, at least in part, the differences in maternal age amongst the two groups seen in Table 3.1 above.

When entered into the multivariate model alongside score at age 3 and parental level of education, none of the family composition variables were found to be significantly associated with a change in cognitive ability - on either assessment 16 . Neither was there any notable change in the strength of the relationship between education and ability. Thus, these factors are neither independently related to change in cognitive ability during the pre-school years nor do they explain any of the difference in ability by education.

3.3.3 Parenting factors

Considerable attention has been focused in recent years on the relationship between parenting activities and children's development - including their cognitive development. In particular, numerous studies have shown a significant relationship between aspects of parenting - such as the nature of the parent-child relationship and the pursuit of home learning activities 17 - and children's early language skills (Bromley, 2009; Waldfogel and Washbrook, 2008; Waldfogel and Washbrook 2010; Foster et al, 2005; National Evaluation of Sure Start, 2008; Sylva et al, 2003).

Five aspects of parenting are considered in this section:

  • Frequency of home learning activities.
  • The existence of rules.
  • Use of harsh discipline.
  • Early infant-maternal attachment.
  • Parental problems with reading or writing.

Table 3.4 provides information on how these factors vary according to parental level of education.

Table 3.4 Selected parenting characteristics by parental highest level of education

Parental highest level of education

Parenting characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or Intermediate VQs

Higher grades and upper level VQs

Degree level academic & vocational qualifications

All

Mean score on home learning activities scale, age 2 and 3 years

38.6

40.9

42.6

45.9

47.6

45.2

Banded level of rules in household at 58 mths

Low

61%

51%

41%

34%

27%

36%

Medium

35%

41%

51%

54%

63%

55%

High

4%

8%

8%

11%

10%

10%

Use of harsh discipline at 22 mths and/or 46 mths

Never used smacking

52%

55%

48%

49%

54%

51%

Used smacking before age 4

48%

45%

52%

51%

46%

49%

Mean score on infant-maternal attachment scale at 10 mths

3.7

3.8

3.8

3.9

3.8

3.8

Parental problems with reading/writing at 46 mths

Does not have any problems

67%

80%

84%

90%

93%

88%

Does have some problems

33%

20%

16%

10%

7%

12%

Bases

Weighted

192

167

699

1096

1151

3309

Unweighted

127

130

603

1134

1354

3353

  • Children in higher educated households are more frequently involved in home learning activities such as reading, painting and games involving shapes and numbers than are children from lower educated households.
  • In addition, parents with higher qualifications are more likely to set rules for the child than are parents with lower qualifications. Almost two-thirds (63%) of degree educated parents were categorised in the 'medium rules' group, around twice the proportion amongst parents with no qualifications.
  • Figures on use of smacking are similar across the education groups - the differences which exist are not statistically significant.
  • In contrast, whilst the differences in mean attachment scores are small, they are statistically significant (p <0.001). Parents with no qualifications have slightly weaker attachment than parents in the other education groups.
  • As may be expected, there is a close relationship between qualifications obtained and literacy problems. Parents with no and lower qualifications were considerably more likely to report having at least some difficulties with reading and/or writing.

Children who experienced a higher frequency of activities, those living in households with greater rule-setting and those with a stronger early parent-child attachment were each more likely to see their vocabulary scores improve in the pre-school period, irrespective of parental education level and age 3 score (Table 3.5) 18 . More frequent home learning activities were also associated with improved problem solving scores but other parenting measures were not significantly related to changes in that ability.

Table 3.5 Parenting domain linear regression - summary results

Cognitive ability

Knowledge of vocabulary

Problem solving

Parenting characteristic

Sig.

Direction of change*

Sig.

Direction of change*

Mean score on home learning activities scale

<0.001

+

<0.001

+

Banded level of rules in household (ref: low rules)

Medium

NS

NS

High

0.016

+

NS

Harsh discipline (ref: never smacked)

Used smacking before age 4

NS

NS

Mean score on infant-maternal attachment scale at 10 mths

< .001

+

NS

Parental problems with reading/writing when child aged 4 (ref: does not have any problems)

Does have some problems

NS

NS

Parental level of education (ref: no qualifications)

Lower SGs or VQs or 'Other' quals

NS

NS

Upper level SGs or Intmed VQs

0.004

+

NS

Higher Grades or Upper level VQs

0.011

+

0.05

+

Degree level academic or VQs

<0.001

+

<0.001

+

Weighted base

3098

3190

*A plus sign (+) indicates relative improvement in ability score and a minus sign (-) indicates relative decline in ability score for children in the various sub-groups as compared those in the reference sub-group. The reference sub-group is indicated in brackets. Where the variable is not significant, direction of change has not been included.

Figure 3-C and Figure 3-D show, the addition of parenting variables to the model weakened the relationship between education and ability on both assessments, suggesting that these approaches to parenting help to explain some of the differences observed by education.

Figure 3-C Associations between education and change in vocabulary ability before and after taking account of parenting characteristics

Figure 3-C Associations between education and change in vocabulary ability before and after taking account of parenting characteristics

Note: reference category is 'no qualifications'.

Figure 3-D Associations between education and change in problem solving ability before and after taking account of parenting characteristics

Figure 3-D Associations between education and change in problem solving ability before and after taking account of parenting characteristics

Note: reference category is 'no qualifications'.

3.3.4 Experience of childcare and pre-school

Improvement of child outcomes via the provision of childcare and pre-school based interventions and programmes have been a focus of much recent early years policy.

This follows a range of research and evaluations demonstrating the significant impact such experiences and intervention can have on a broad range of child outcomes (see Burger, 2010 for a review). Evidence shows that as compared to no experience of centre-based care or pre-school education, children with any experience tend to have improved language and cognitive skills (Sylva, 2009; Butt et al, 2007; Magnuson et al, 2010). In terms of the statutory pre-school provision, to which all children in the UK are entitled at ages 3 and 4, the EPPE study (Sylva et al, 2009) found that quality of provision is the key factor associated with making the greatest impact on on intellectual and cognitive development - a finding echoed in US-based research (Butt et al, 2007). However other factors such as duration of attendance (in months and years rather than the number of hours per day) have also been shown to be important (Butt et al, 2007).

In this section we consider a range of factors which seek to capture children's experience of childcare, pre-school and primary school in the period between birth and the assessment at age 5. They include: any experience of formal childcare before age 3, any attendance at pre-school and type of provision attended, weekly duration of pre-school, months of pre-school attended, child's perceived readiness for pre-school, and whether the child had started primary school. Differences across these measures are shown in Table 3.6.

  • Children from higher educated backgrounds were considerably more likely than those in all other groups to have experienced some formal childcare before the age of 3; 72% had done so compared with 35% of children whose parents had no qualifications.
  • Data from age 3 onwards shows that the type of pre-school varied slightly amongst the different groups. Whilst children in all groups were most likely to have attended a nursery class attached to a state or independent school, reflecting the dominant type of provision offered in Scotland, those in the higher education groups were more likely than those in the lower groups to have attended a private nursery school and those with no qualifications a local authority nursery school.
  • There were no significant differences in the number of hours attended per week nor in the duration of attendance prior to the assessment at age 5 (note that this measure was approximately derived - further details in appendix 1), neither were there any significant differences in primary school attendance - as may be expected given that eligibility is driven by date of birth.
  • Differences are evident in perceived school readiness. Parents with higher qualifications were significantly more likely to consider the child 'ready' for pre-school than were parents with lower qualifications which appears justified given the higher ability scores amongst children in the former group.

Table 3.6 Selected childcare and pre-school characteristics by parental highest level of education

Parental highest level of education

Childcare and pre-school characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

Experience of formal childcare before age 3

No

65%

61%

54%

43%

28%

42%

Yes

35%

39%

46%

57%

72%

58%

Type of pre-school attended

No pre-school

2%

4%

2%

2%

2%

2%

Nursery class attached to state or independent primary school

63%

74%

70%

63%

63%

65%

Local authority nursery school

25%

15%

17%

15%

11%

14%

Private nursery school

7%

5%

9%

16%

20%

15%

Other

3%

2%

2%

4%

4%

4%

Weekly duration of pre-school

No pre-school

2%

4%

2%

2%

2%

2%

Less than 12 hours

7%

6%

7%

8%

9%

8%

Between 12 and 12.5 hours

62%

68%

66%

62%

55%

61%

Between 12.5 and 15 hours

17%

11%

11%

11%

13%

12%

15 hours or more

12%

11%

13%

18%

21%

17%

Mean duration of pre-school experienced (months)

18.0

17.1

17.4

17.9

18.0

17.8

Parent's perception of child's readiness for pre-school at age 3

Average or above readiness score

42%

52%

49%

57%

60%

55%

Below average readiness score

58%

48%

51%

43%

40%

45%

Whether child started school at 58 mths

No

72%

73%

67%

67%

67%

68%

Yes

28%

27%

33%

33%

33%

32%

Bases

Weighted

185

151

639

1015

1072

3066

Unweighted

123

118

559

1053

1262

3119

Few of the childcare and pre-school factors remained significant in the regression model. Those which did were slightly different for each type of ability 19 . For naming vocabulary, only perceived readiness was significant - children with above average perceived readiness were more likely to improve their relative scores during the pre-school period. For problem solving pre-school type was significant. The results suggest that, as compared to children attending a pre-school class attached to a primary school, those who attended a private nursery school were more likely to see their relative problem solving ability improve. Attending primary school (P1) was also associated with a relative improvement in problem solving ability despite those children who had started school having only spent a small amount of time there. 20

3.3.5 Child health and early development

For children who suffer poorer health and early developmental difficulties the effects on outcomes can be persistent. A range of evidence indicates that early health problems and developmental delays continue to impact on developmental outcomes, including cognitive outcomes, in later life. For example, Bradshaw (2010) found that children who were reported by their parents to have delays in motor development and language development at age 3 were more likely to display difficulties with their social, emotional and behavioural development at school entry.

This section examines the association between a number of indicators of child health and early development and change in cognitive ability. The indicators considered include: child's general health from 10 months to age 3, low birth weight, total score on the Infant/Toddler Checklist of the Communication and Symbolic Behaviour Scales at age 2 (22 months) 21 , and child's level of physical activity at age 3. Descriptive information on how these characteristics vary by parental level of education is included in Table 3.7.

Table 3.7 Selected child health and early development characteristics by parental highest level of education

Parental highest level of education

Child health and early development characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

General health from
10 mths to 34 mths

Always good or very good

84%

78%

82%

87%

90%

86%

Temporarily or always fair, bad or very bad

16%

22%

18%

13%

10%

14%

Low birth weight

Not low

88%

93%

91%

95%

95%

94%

Low

12% 7% 9% 5% 5% 6%

Mean total score on CSBS infant/toddler checklist at age 2 (22 mths)

48.2

48.2

48.6

49.9

50.0

49.5

Level of physical activity at age 3 (34 mths)

Low

56%

41%

41%

29%

26%

32%

Medium

27%

30%

31%

36%

37%

35%

High

17%

30%

28%

35%

37%

33%

Bases

Weighted

191

167

699

1096

1151

3308

Unweighted

126

130

603

1134

1354

3352

  • Those children with better educated parents tended to have persistently better general health than those with lower educated parents. Nine in ten (90%) children with degree-educated parents were reported to have consistently good or very good health between the ages of 10 months and 3 years compared with around eight in ten children from the lowest education groups.
  • 12% of children whose parents had no qualifications were born with a low birth weight, a proportion twice as high as that amongst children whose parents are degree educated (5%).
  • Whilst the differences are small, children from lower educational backgrounds scored statistically significantly lower on average on their assessment of communication and and language development at age 22 months.
  • Variations in levels of child physical activity are quite stark; children in the lowest education group were around twice as likely as those in upper two groups to be classed in the 'low activity' category.

Few of the variables considered remain significant in the regression model, as summarised in Table 3.8 22 . For vocabulary, early indications of good progress in language and communicative development are associated with improvement in vocabulary ability in the pre-school years. In other words, children given a good start in language development show better improvement in these skills in the later pre-school period. Early communication issues do not have the same association with change in problem solving ability although this factor is only just non-significant. General health appears to be more important for problem solving ability. Those children who report even temporarily poorer general health are more likely than those with consistently good general health to see their relative problem solving ability decline between age 3 and 5.

Table 3.8 Child health and early development domain linear regression - summary results

Child health or early development characteristic

Cognitive ability

Knowledge of vocabulary

Problem solving

Sig.

Direction of change

Sig.

Direction of change

General health from birth to age 3
(ref: always good or very good)

Temporarily or always fair, bad or very bad

NS

<.05

-

Birth weight (ref: not low)

Low

NS

NS

Total score on CSBS Infant/Toddler checklist

<.001

+

NS (.06)

+

Level of physical activity

NS

NS

Parental level of education
(ref: no qualifications)

Lower SGs or VQs or 'Other' quals

NS

NS

Upper level SGs or Intmed VQs

<.01

+

NS

Higher Grades or Upper level VQs

<.01

+

NS

Degree level academic or VQs

<.001

+

<.01

+

Weighted base

2998

3006

*A plus sign (+) indicates relative improvement in ability score and a minus sign (-) indicates relative decline in ability score for children in the various sub-groups as compared those in the reference sub-group. The reference sub-group is indicated in brackets. Where the variable is not significant, direction of change has not been included.

Aspects of a child's health and early development do appear to offer some part of the explanation of differences in ability by parental education. As shown in Figure 3-E and Figure 3-F, there is a noticeable drop in the association between education and ability when the health variables are added to the regression model. This effect is slightly larger for vocabulary than for problem solving.

Figure 3-E Associations between education and change in vocabulary ability before and after taking account of child health and early development characteristics

Figure 3-E Associations between education and change in vocabulary ability before and after taking account of child health and early development characteristics

Note: reference category is 'no qualifications'.

Figure 3-F Associations between education and change in problem solving ability before and after taking account of child health and early development characteristics

Figure 3-F Associations between education and change in problem solving ability before and after taking account of child health and early development characteristics

Note: The reference category is 'no qualifications'.

3.3.6 Parenting support

This section examines the impact of aspects of support for parents on change in cognitive ability. Indicators of this domain included levels of service use at 10 months, access to informal support networks (family and friendship groups), and attendance at ante-natal and other parenting classes.

For parents of young children, having access to the services and support necessary to assist them in their parenting role can be important for improving child outcomes across a range of developmental domains. Evidence from the Sure Start Impact evaluation (National Evaluation of Sure Start, 2008) indicated that the improvement of services in areas of high deprivation through the Sure Start programme led to better social development amongst children in those areas, along with beneficial effects on parenting, higher rates of child immunisation and lower rates of accidental injury (although caution was advised on interpreting this finding due to issues related to the timing of measurement). Evaluations of a range of early childhood interventions demonstrate the significant improvements which can be gained in cognitive and academic achievement via engaging at-risk parents and providing the necessary support and services they, and their children, require (Geddes et al, 2010). Although getting those parents engaged with such services can be particularly challenging (Mabelis and Marryat, 2011).

The data in Table 3.9 illustrate how aspects of service use and support vary by parental education. 23

  • More highly educated parents were significantly more likely to draw on a wider range of services when the child was aged 10 months. Degree-educated parents reported accessing an average of around five different health, care and parenting services compared with an average of around three for parents with no qualifications.
  • More highly educated parents also reported higher levels of informal social support. A little over 60% of parents in the two upper education groups had both a satisfactory friendship and family network compared with 48% of those in the lowest group.
  • Attendance at ante-natal classes was much higher among the higher educated groups. Sixty-one percent of degree-educated parents attended some ante-natal classes, around three times the proportion of those in lower educated groups who attended.
  • Attendance at other parenting classes was generally low and there were no significant differences in attendance by education level.

Table 3.9 Selected parenting support characteristics by parental highest level of education

Parental highest level of education

Parenting support characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

Mean score on service use scale at 10 months

3.3

4.0

4.3

4.7

5.4

4.8

Existence of social networks at 22 mths

Has both satisfactory family and friendship networks

48%

55%

58%

65%

62%

61%

Only has satisfactory friendship network

8%

9%

12%

12%

19%

14%

Only has satisfactory family network

26%

19%

21%

16%

12%

16%

Has neither satisfactory social network

18%

17%

9%

7%

7%

8%

Attended any parenting classes from birth to age 3

No

91%

93%

92%

93%

91%

92%

Yes

9%

7%

8%

7%

9%

8%

Ante-natal classes

Went to all or most

12%

16%

25%

39%

50%

37%

Went to some

4%

10%

9%

13%

11%

11%

Did not go to any

84%

73%

66%

49%

39%

52%

Bases

Weighted

192

167

699

1095

1150

3309

Unweighted

127

130

603

1134

1354

3353

None of the parenting support factors remained associated with change in vocabulary ability and only attendance at all ante-natal classes remained associated with problem solving ability, having a positive effect on ability during the pre-school period 24 . Parenting support factors do slightly affect the relationship between education and ability as shown in Figure 3-H and Figure 3-H. The coefficients for education are weakened more in relation to vocabulary than problem solving ability. This suggests that variations in parenting support contribute more to explaining education differences in vocabulary ability than explaining education differences in problem solving. This overall effect, without the individual variables being significant, suggests there may be some correlation between the various explanatory variable measures.

Figure 3-G Associations between education and change in vocabulary ability before and after taking account of parenting support characteristics

Figure 3-G Associations between education and change in vocabulary ability before and after taking account of parenting support characteristics

Note: reference category is 'no qualifications'.

Figure 3-H Associations between education and change in problem solving ability before and after taking account of parenting support characteristics

Figure 3-H Associations between education and change in problem solving ability before and after taking account of parenting support characteristics

Note: The reference category is 'no qualifications'.

3.3.7 Maternal health and health behaviours

The next domain considered the impact of a number of indicators of maternal health and health behaviours on change in cognitive ability. Good maternal health and wellbeing promotes better development and outcomes for children whereas poorer physical and mental health can be detrimental to child development.

Marryat and Martin's analysis of GUS data (2010) showed that, at age 4, children whose mother's had reported poor mental health had, on average, lower cognitive ability and higher social, emotional and behavioural difficulties. In addition, Bromley (2010) illustrated the strong associations between maternal health and health behaviours (such as smoking) and a range of child outcomes including cognitive ability. Bromley's findings indicated, for example, that children whose mothers did not smoke had far fewer negative health outcomes than children whose mothers did smoke.

A range of studies have demonstrated higher average cognitive ability, at several ages, amongst children who were ever breastfed compared with those who were formula fed (Quigley et al, 2009; Iacovou and Sevilla-Sanz, 2010; see also Anderson et al (1999) for a meta-analysis). One interpretation of this association is that breastmilk has nutritional value of benefit to child cognitive development. However, the decision to breastfeed may also reflect a desire to adopt a particular parenting approach and positive health behaviours which may involve other parenting and childcare practices associated with better cognitive ability. Thus the effect of breastfeeding on cognitive ability may be transmitted via these additional parenting behaviours rather than solely linked to the nutritional content of breastmilk itself.

The indicators selected for inclusion were: maternal smoking between the child's birth and age 3, whether the child was ever breastfed and whether the mother had reported poor mental or poor general health between the child's birth and age 3. These factors all showed significant differences by parental level of education as shown in Table 3.10. Some of the largest differences relate to smoking and breastfeeding.

  • In more highly educated households, the child was significantly more likely to have been breastfed and the mother less likely to smoke than in lower educated households.
  • Mothers in the two lower education groups were around 6 times more likely to have smoked during the child's first three years than were mothers in the highest group.
  • Those in the lower groups were also more likely to have reported poorer mental and general health. A little over two-fifths (44%) of mothers in households where parents had no qualifications reported poor mental health compared with around one-fifth (19%) in households where a parent was degree-educated.

Table 3.10 Selected maternal health characteristics by parental highest level of education

Parental highest level of education

Maternal health characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

Ever smoked between child's birth and age 3

No

34%

43%

47%

72%

89%

69%

Yes

66%

57%

53%

28%

11%

31%

Was child ever breastfed?

No

76%

67%

57%

42%

16%

39%

Yes

24%

33%

43%

58%

84%

61%

Reported poor mental health between child's birth and age 3

No

56%

71%

71%

79%

81%

76%

Yes

44%

29%

29%

21%

19%

24%

Reported less than 'good' health between child's birth and age 3

No

61%

77%

76%

82%

87%

81%

Yes

39%

23%

24%

18%

13%

19%

Bases

Weighted

192

167

698

1096

1151

3308

Unweighted

127

130

602

1134

1354

3352

Breastfeeding emerged as significant in both multivariate models 25 . Children who were breastfed were more likely than those who were not to show a relative improvement in their cognitive ability. For problem solving ability, maternal mental health was also significant. The results indicate that children whose mothers suffered poorer mental health saw a relative deterioration in their abilities during the pre-school period. The maternal health variables made only a small reduction to the education co efficients, as shown in Figure 3-I and Figure 3-J. This means they explain only a small amount of the education-related difference in cognitive ability. The effect was of a similar magnitude for both vocabulary and problem solving.

Figure 3-I Associations between education and change in vocabulary ability before and after taking account of maternal health characteristics

Figure 3-I Associations between education and change in vocabulary ability before and after taking account of maternal health characteristics

Note: reference category is 'no qualifications'.

Figure 3-J Associations between education and change in problem solving ability before and after taking account of maternal health characteristics

Figure 3-J Associations between education and change in problem solving ability before and after taking account of maternal health characteristics

Note: the reference category is 'no qualifications'

3.3.8 Material and economic circumstances

A considerable amount of evidence, including that discussed in Chapter 2 of this report, illustrates the differences in child cognitive ability by level of household income. However, other research has also demonstrated that additional, related, material and economic circumstances - such as maternal employment and area deprivation - can impact on child outcomes above and beyond the effect of income itself. For example, McCulloch and Joshi (2001), in analysis of data from the British National Child Development Study, found a small but statistically significant relationship between neighbourhood poverty and lower test scores in children aged 4-5. Barnes et al (2010) found associations between indicators of poverty such as car ownership and housing tenure and a range of child health and development outcomes.

The association between several measures of family material and economic circumstances and change in cognitive ability was explored. The measures included housing tenure, level of material deprivation, maternal employment, area deprivation, and the extent to which the parent reported the family to be coping on their current income. Given these measures are associated with income, and that there is a close relationship between income and education it is unsurprising to also find considerable variations in these measures by parental education (Table 3.12).

  • Children with degree educated parents were significantly more likely than those whose parents had lower qualifications to live in an owner-occupied home, experience lower material deprivation, have higher rates of maternal employment and to live in an area in the lowest two deprivation quintiles.
  • In contrast, children whose parents had lower qualifications, and particularly those in the lowest education group, were more likely to live in social rented housing, experience higher material deprivation than children whose parents had higher qualifications and live in an area with higher deprivation.

Table 3.11 Selected material and economic characteristics by parental highest level of education

Parental highest level of education

Material and economic characteristics

No qualifications

Lower Standard Grades or VQs or Other

Upper level SGs or intermediate VQs

Higher Grades & upper level VQs

Degree level academic & vocational qualifications

All

Housing tenure at age 3 (34 mths)

Owner occupied

17%

30%

39%

69%

89%

65%

Social rented

74% 61% 47% 23% 6% 27%

Private rented

8% 7% 10% 5% 3% 6%

Other

1% 3% 4% 3% 2% 3%

Mean score on material deprivation index at age 4 (46 mths)

22.1

11.0

12.6

7.1

4.0

7.9

Did mother work between child's birth and age 3

No

65%

51%

40%

21%

16%

28%

Yes

35% 49% 60% 79% 84% 72%

Area deprivation - quintiles - age 3 (34 mths)

Least deprived

2%

8%

8%

17%

33%

19%

2

6%

7%

14%

20%

28%

20%

3

10%

20%

20%

22%

19%

20%

4

26%

25%

22%

19%

11%

18%

Most deprived

56%

40%

35%

22%

9%

23%

Reported difficulty coping on present income between birth and age 3

No

39%

63%

62%

73%

81%

71%

Yes

61%

37%

38%

27%

19%

29%

Bases

Weighted

192

167

698

1096

1151

3308

Unweighted

127

130

602

1134

1354

3352

The relative ability of children who experienced higher levels of material deprivation tended to deteriorate in the pre-school period. In addition, living in an area in the highest deprivation quintile was detrimental for problem solving ability. No other factors were significant in the regression models. Despite this, the addition of material and economic circumstances to the models did make a noticeable reduction to the education coefficients (Figures 3-K and 3-L) with slightly larger effects in relation to vocabulary, meaning that these factors explained some of the variation in cognitive ability by parental education.

Figure 3-K Associations between education and change in vocabulary ability before and after taking account of material and economic characteristics

Figure 3-K Associations between education and change in vocabulary ability before and after taking account of material and economic characteristics

Note: reference category is 'no qualifications'.

Figure 3-L Associations between education and change in problem solving ability before and after taking account of material and economic characteristics

Figure 3-L Associations between education and change in problem solving ability before and after taking account of material and economic characteristics

Note: reference category is 'no qualifications'.

3.4 Summary of single domain effects

Thus far the analysis has considered each of the various domains of influence in isolation. This analysis has demonstrated that a range of factors, measuring different aspects of children's circumstances and experiences, are associated with their change in cognitive ability during the pre-school period. However, across all the domain analysis, parental level of education has remained significantly associated with change in cognitive ability although some factors have weakened this association. The factors which have emerged from the various domains as being important differ according to each ability and are summarised in Table 3.12.

Table 3.12 Factors significantly associated with change in cognitive ability from the single domain analysis

Cognitive ability

Knowledge of vocabulary

Problem solving

  • Mother aged under 25 at child's birth
  • Having a non-white parent
  • Frequency of home learning activities at age 2-3 years
  • Level of rule-setting in household at age 5
  • Level of infant-maternal attachment at 10 months
  • Parent's perception of child's readiness for pre-school at age 3
  • Language and communicative development at age 1
  • Attendance at ante-natal classes
  • Breastfeeding
  • Maternal mental health
  • Experience of material deprivation
  • Frequency of home learning activities at 2-3 years
  • Type of pre-school attended
  • Whether child had started primary school
  • Child's general health between 10 months and 2 years
  • Breastfeeding
  • Area deprivation

Only two factors are shared across the two types of ability at this stage - frequency of home learning activities and breastfeeding. The variation across the remaining factors indicates that progress in each ability is influenced by quite different characteristics, circumstances, environments and experiences. This is perhaps unsurprising; whilst both abilities form part of the core assessment in the British Ability Scales Early Years Battery and contribute to the measurement of General Conceptual Ability via that battery, they are clearly measuring quite different cognitive concepts and thus are likely to be affected in different ways by different factors present in the child's life. Indeed, although scores on both assessments are statistically significantly correlated (so that children who score better on one assessment also tend to score better on the other), the strength of the association is only moderate (at around 0.4 at each time point) indicating that consistently high or low scores on both measures for any single child are not inevitable.

3.5 Combined domain effects

To what extent do the various factors in each domain, when taken together, explain differences in change in ability by education level? The next stage of analysis involved entering the significant domain factors together into a single regression model for each ability alongside parental level of education. In so doing, this analysis explores the extent to which each factor remains independently associated with change in ability and an examination of the combined effect of all factors on the relationship between education and change in ability.

The results of the regression analysis are summarised in Table 3.13 for vocabulary and in Table 3.14 for problem solving with the remaining significant variables for both models included in Table 3.15 26 .

In each model, factors across a range of domains remain independently associated with change in the respective skill. However, the particular pattern of variables and domains is different.

Changes in vocabulary (Table 3.13) are more related to aspects of the child's home environment and the choices and behaviours of parents. After controlling for education, greater consistency of parenting, stronger parent-child attachment, attendance at ante-natal classes and breastfeeding were each independently associated with an improvement in vocabulary ability in the pre-school years.

Alongside these home experiences and parenting behaviours, early language development is also key. A higher score on the CSBS infant/toddler checklist at 22 months remained significantly associated with improvement in expressive language skills in the pre-school period. This indicates that those children who display better communication and language skills at an earlier stage are those who are more likely to see those skills improving during the pre-school period. It would appear generally beneficial therefore, to seek to improve children's communication ability from the very earliest stages in order to ensure their continued positive language development. Indeed, it is possible, and perhaps likely, that those factors associated with relative improvement in language ability during the pre-school period are also associated with very early communicative development such as that measured by the CSBS scale.

Table 3.13 Knowledge of vocabulary cross-domain linear regression - summary results

Sig.

Direction of change

Standardised ability score at age 3

0.001

+

Maternal age at child's birth (ref: 25 or older)

Under 25

NS

Parental ethnicity (ref: both parents white)

At least one parent non-white

NS

Mean score on home learning activities scale

NS

Banded level of rules in household (ref: low rules)

Medium

NS

+

High

0.01

Infant-maternal attachment score

0.001

+

Perceived readiness for school scale

NS

Total score on CSBS Infant/Toddler checklist

0.02

+

Attendance at ante-natal classes (ref: did not attend)

Went to some classes

NS

+

Went to all classes

<0.001

Was child ever breastfed? (ref: no)

Yes

0.03

+

Poor mental health since child's birth (ref: no)

Yes

NS

-

Score on material deprivation index

NS

Parental level of education (ref: no qualifications)

Lower SGs or VQs or 'Other' quals

NS

Upper level SGs or Intmed VQs

NS

Higher Grades or Upper level VQs

NS

Degree level academic or VQs

0.015

+

Weighted base

2689

Table 3.14 Problem solving cross-domain linear regression - summary results

Sig.

Direction of change

Standardised ability score at age 3

<0.001

+

Mean score on home learning activities scale

0.002

+

Pre-school type (ref: nursery class attached to school)

No pre-school

0.03

-

Local Authority nursery school

NS

Private nursery school

<0.001

+

Other provider

NS

Has child started primary school? (ref: no)

Yes

0.03

+

General health from birth to age 3 (ref: always good or very good)

Temporarily or always fair, bad or very bad

NS

Was child ever breastfed? (ref: no)

Yes

0.04

+

Area deprivation (ref: least deprived)

2

NS

3

NS

4

NS

5 Most deprived

0.02

-

Parental level of education (ref: no qualifications)

Lower SGs or VQs or 'Other' quals

NS

Upper level SGs or Intmed VQs

NS

Higher Grades or Upper level VQs

NS

Degree level academic or VQs

0.04

+

Weighted base

3344

Change in problem solving ability appears much more susceptible and responsive to external influences, although the home environment and parenting behaviours - through home learning activities and breastfeeding - continue to have an impact.

The experience of pre-school education itself had an impact. Those children who did not attend any pre-school education were more likely to show a deterioration in problem solving ability. In contrast, when delivered via a private nursery school, the pre-school experience itself remained independently associated with an improvement in problem solving skills. Early primary school experience, of just a few months, also had a positive impact with those children showing an improvement in problem solving ability although we may expect this effect to disappear when all children enter school at this stage. That is, all children will mutually benefit in terms of their problem solving ability from their early experience at primary school when they come to attend and it cannot be presumed that sending children to primary school early will necessarily additionally benefit their cognitive development.

Area deprivation was a further external influence, but a negative one. Compared to those living in the least deprived areas, the relative problem solving ability of children living in the most disadvantaged areas declined.

3.5.1 Summary of combined domain effects

The factors which have emerged as being statistically significantly associated with change in each ability in the combined domain analysis are summarised in Table 3.15.

Table 3.15 Factors significantly associated with change in cognitive ability from the cross-domain analysis

Cognitive ability

Knowledge of vocabulary

Problem solving

  • Level of rule-setting in household at age 5
  • Level of infant-maternal attachment at 10 months
  • Language and communicative development at 22 months
  • Attendance at ante-natal classes
  • Breastfeeding
  • Frequency of home learning activities at 2-3 years
  • Type of pre-school attended
  • Whether child had started primary school
  • Breastfeeding
  • Area deprivation

Only a single factor, breastfeeding, is shared across the two types of ability in this analysis. The variation across the remaining factors further confirms that progress in
each ability is influenced by different characteristics, circumstances, environments and experiences.

3.5.2 Explaining the effect of education on gaps in ability

Throughout this section we have been considering how much of the difference (or variance) in ability scores at age 5 between children from the various educational groups can be explained by differences in measures in each of the domains. When differences in ability at age 5 are considered across all children, the analysis shows that level of parental education and score at age 3 explain 30% of those differences in relation to vocabulary and 11% in relation to problem solving. When the additional significant measures are added, no further variance is explained in relation to vocabulary, and only an additional 2% is explained in relation to problem solving indicating that the domain measures largely form part of the differences already seen and measured by parental education. It is therefore unsurprising to see that the strength of the association between parental level of education and change in ability is visibly reduced in these final models as shown in Figure 3-M and Figure 3-N. As indicated by the position of the dotted lines in the graph, the association between level of parental education and ability at age 5 is weaker (that is, the coefficient has a lower value) for both models after the additional explanatory measures have been added.

Figure 3-M Associations between education and change in vocabulary ability before and after taking account of combined domain characteristics

Figure 3-M Associations between education and change in vocabulary ability before and after taking account of combined domain characteristics

Note: reference category is 'no qualifications'.

Figure 3-N Associations between education and change in problem solving ability before and after taking account of combined domain characteristics

Figure 3-N Associations between education and change in problem solving ability before and after taking account of combined domain characteristics

Note: reference category is 'no qualifications'.

In addition, it is notable that of the various levels of qualifications included in the education measure, only degree-level education continues to have an independent positive effect on change in either ability. This suggests that much of the difference seen earlier across the lower qualifications is explained by differences in the home and external environments, and parenting experiences of the children in these lower groups.

The evidence in this chapter has shown that there are a number of factors which appear to impact on change in cognitive ability over and above parental level of education. Moreover, the direct effect of education level on change in ability is reduced by the addition of these factors. Thus, whilst these additional factors do not explain any more of the variance in ability than education itself, we have confirmed some of the processes through which the effects of parental education are transferred to children's cognitive development. As such, these findings indicate that improving educational qualifications amongst parents alone would not close all or most of the education-related gaps in cognitive ability because some of those gaps are created by the different experiences of children whose parents have different levels of education. Raising education levels would require accompanying change in the circumstances outlined here in order to move towards a more narrow cognitive ability gap.