Publication - Research and analysis

Growing up in Scotland: the impact of children's early activities on cognitive development

Published: 18 Mar 2009
Directorate:
Children and Families Directorate
Part of:
Children and families, Education
ISBN:
9780755919680

This report uses data from the first three waves of the Growing Up in Scotland study (GUS) to explore children’s cognitive ability.

92 page PDF

1.8 MB

92 page PDF

1.8 MB

Contents
Growing up in Scotland: the impact of children's early activities on cognitive development
Chapter 5 The Relative Importance of Children's Activities and Socio-Demographic Factors

92 page PDF

1.8 MB

Chapter 5 The Relative Importance of Children's Activities and Socio-Demographic Factors

5.1 Key findings

  • The two questions addressed in this report are: do children's early activities have an influence on cognitive development in addition to socio-demographic factors?; and, do children's early activities moderate the effect of socio-demographic factors on cognitive development?
  • To answer these questions multivariate analysis was carried out to explore whether any of the activity measures were independently associated with cognitive ability scores once a range of socio-demographic factors were controlled for.
  • The analysis considered each of the cognitive ability assessments separately and looked at the influence of:
  • socio-demographic factors alone,
  • activity measures and socio-demographic factors together, and
  • activity measures and socio-demographic factors in a sub-set of less advantaged children.
  • For the naming vocabulary assessment area deprivation and family composition/employment type were no longer significant when all socio-demographic factors were considered together. Three activity measures were independently associated with ability when all factors were considered: being read to every day at age 10 months, being in the most active group at age 22 months for daily activities, and visiting a wide range of events/places at age 22 months. The last two activity measures were still significant when the analysis focused on less advantaged children.
  • For the picture similarities assessment mother's education was no longer significant when all socio-demographic factors were considered together, while age of mother at birth and family composition/employment type were no longer significant once all factors were considered. Two activity measures were independently associated with ability when all factors were considered: being in the most active group at age 22 months for daily activities, and visiting a wide range of events/places at age 22 months.
    The daily activity measure was still significant when the analysis focused on less advantaged children.
  • Collectively these finding suggest that activities do have an influence on children's cognitive development and that they can moderate the effect of socio-demographic disadvantage. The individual factors associated with ability were different for the two assessments.
  • The overall amount of variation in children's scores explained by the analysis was relatively low, but typical for analysis of a social survey, other factors that haven't been explored here are also likely to be important, including genetic factors.

5.2 Introduction

This chapter brings together the analyses presented in the two previous chapters and attempts to establish whether children's cognitive ability is still influenced by the activities they experience and their parents' views of those activities when socio-demographic factors are also taken into consideration. The method used to carry out this analysis is described and the results are presented in summary form.

5.3 Approach to the analysis

5.3.1 Overview and limitations

The results in the previous chapter demonstrated the relationships between a range of socio-demographic measures, children's activities and cognitive ability at age 34 months. However, it is also clear from Chapter 3 that activities are themselves highly socially patterned. It is possible, therefore, that the significant associations between activities and cognitive ability are simply a reflection of the socio-demographic composition of the families whose children experience a wide range of activities, rather than the activities themselves exerting an independent influence. It is possible to carry out a form of analysis to disentangle these relationships and help establish whether activities alone can help explain differences in cognitive ability levels, over and above socio-demographic factors. The analysis used for this task was multiple linear regression, a multivariate technique that is able to assess the relationship that one factor has with an outcome measure ( e.g. cognitive ability) while simultaneously controlling for the effects of all other factors.

It is important to note that while multivariate techniques can help to identify which factors are independently associated with cognitive ability scores, establishing that these factors cause any differences in cognitive ability is slightly trickier. Some studies struggle to establish causation because they only investigate relationships between factors at one point in time and the sequence of events is not always clear. For example, low levels of physical activity are associated with poor cardiovascular health; however, people with poor health are also more likely to have reduced activity levels as a result of their condition. Whether low physical activity causes poor cardiovascular health can only be answered by tracking people over time with a longitudinal study and comparing the health outcomes of people with different activity levels earlier in life. GUS has the advantage of being a longitudinal study so it will be possible to draw some conclusions about the direction of some relationships; for example rather than looking at activities at the same point in time that the cognitive ability tests were administered this analysis looks at children's activities one or two years before the assessments. However, all analyses, whether longitudinal or based on one point in time, share the limitation that any relationships that are identified may simply mask other underlying associations that have not been considered, either because there is no data about them or because they have not been included in the analysis. So, although the temporal limitations faced by some data analysis are reduced with a study like GUS, the potential for key explanatory factors to be missing means caution must be drawn when drawing conclusions based on its results.

5.3.2 Regression model stages

The regression analysis of cognitive ability was done in four stages. These are described below.

Stage one

The first stage in the analysis looked at socio-demographic measures to identify which had an independent association with naming vocabulary scores. Not only are socio-demographic factors and children's activities highly correlated, but also many socio-demographic measures are themselves correlated. This stage helps to clarify which of the relationships shown in the simple two-way tables presented in previous chapters are meaningful in their own right and which ones are simply replicating other underlying patterns in the data.

Stage two

The second stage added a longitudinal dimension to the socio-demographic analysis to investigate the impact of persistent disadvantage. Whereas the first stage looked at household income, area deprivation and household type at the point when the cognitive assessments were carried out, this second stage looked at the cumulative impact of a child being in the most disadvantaged category for each of these factors for all of the first three years of life. The three persistent disadvantage measures were based on: the bottom household income quintile, the 15% most deprived areas, and households containing no adult working >16 hours a week. The classification was as follows (using income as an example):

  • All three years in bottom income quintile (persistent disadvantage)
  • Two out of three years in bottom income quintile
  • One out of three years in bottom income quintile
  • Never lived in household in bottom income quintile

Stage three

Having established which of the demographic factors have an independent association (based on the results of the previous two stages) the third stage added the suite of child activities measures into the analysis. This stage of the regression was then able to identify which of all these factors were independently associated with cognitive ability scores; this could then be used to answer the first question addressed in this report:

Do children's early activities have an influence on cognitive development in addition to socio-demographic factors?

Stage four

The third stage was able to identify whether activities have an influence on cognitive ability independent of socio-demographics, but it is not possible to conclude from this whether activities could have the potential to lessen some of socio-demographic differentials evident in children's outcomes. For example, it could simply be that any added contribution conferred by activities exerts its influence among affluent children rather than all children, and therefore exacerbates rather than ameliorates relative disadvantage. To address this, the final stage removed from the analysis three groups of children who have been shown to have a particular advantage in terms of their socio-demographic background and participation in activities. Children falling into any of the following three groups were removed:

  • Mothers with degree-level education,
  • Households in the 20% least deprived of areas, and
  • Households with incomes in the highest quintile.

By removing the children from more advantaged backgrounds from this final stage of the analysis it is possible to investigate whether any independent association found between activities and cognitive ability still exists among relatively less advantaged children. This then helps to answer the second question addressed in this report:

Do children's early activities moderate the effect of socio-demographic factors on cognitive development?

5.4 Results of multivariate analysis

Table 5.1 and Table 5.2 show the results for each of the four stages of analysis for the naming vocabulary and picture similarity scores, respectively. To keep the interpretation simple the tables only show information about the significance of the association between the overall factors being examined and cognitive ability; they do not go into further detail about the patterns of association between the categories within each factor as well (this is presented in the Appendix). The analysis carries out a test of the statistical significance of the association between each factor and the cognitive ability score of interest, while holding all other factors in the analysis constant. Statistical significance is normally reported using two thresholds:

  • the 95% level, which means that there is a one in twenty risk that the association with cognitive ability has occurred by chance rather than being a genuine relationship, and
  • the 99% level, where there is a one in a hundred chance that the association is not genuine.

Chance findings are a consequence of having selected a sample of children rather than the entire population as there is always a risk attached that the sample is atypical of the population (though robust sampling techniques and sufficient numbers reduce this risk considerably). The results of the regression analysis of naming vocabulary and picture similarity scores are presented in turn below.

5.4.1 Naming vocabulary

Full significance levels are presented in the Appendix, here the table simply sets out which factors had a significant association with naming vocabulary scores at the 95% and 99% levels, and which factors were not significant. Two asterisks (**) indicate the strongest associations, at the 99% level, one asterisk is used for the 95% level, and n.s. means the association was not significant. Blank cells in the table are used to indicate factors that were not part of that stage of the analysis ( e.g. the activities measures are missing from stages one and two).

Table 5.1 Naming vocabulary multiple linear regression - summary results

Naming vocabulary

Regression stages

Stage one

Stage two

Stage three

Stage four

Independent variables

Mother's highest qualification

**

**

**

n.s.

Area deprivation

n.s.

Area deprivation - persistent

-

**

n.s.

n.s.

Equivalised household income quintile

**

Household income - persistent

**

**

**

Family and employment type

n.s.

Family and employment type - persistent

n.s.

Age of mother at child's birth

**

**

*

n.s.

Developmental concern

**

**

**

**

No. of children in family

**

**

**

**

Gender of child

**

**

**

**

Birth weight

**

**

**

*

At 10 months:

Reading with parent

*

n.s.

Library visits

n.s.

At 22 months:

Daily activities

**

**

Educational games

n.s.

Number of annual visits/events

**

*

Number of activities rated very important

n.s.

Importance of educational games

n.s.

Satisfaction with activities

n.s.

R squared

15.6

15.7

19.5

18.2

Constant

35.7

36.5

31.9

31.2

Significance

.000

.000

.000

.000

Sample size

3560

3534

3484

1985

Notes: **=significant at 99%, *=significant at 95%, n.s.=not significant, blanks cells indicate factors not included in the model.

The first stage of the analysis just looked at the socio-demographic factors explored previously in Table 4.1-Table 4.4. It reveals that there is no significant independent association between area level deprivation and family/employment type (a measure that captures the employment status of the household as well as the number of resident parents) and naming vocabulary score, once the other factors are considered at the same time. All other socio-demographic factors are, however, highly significant.

The next stage looks at the influence of socio-demographic factors through a slightly different lens. This time the measures of income, area deprivation and family/employment type are not simply concerned with their impact at one point in time (when the cognitive assessments were carried out) but instead look at the association between experiencing persistent disadvantage and cognitive performance. The findings for income and family/employment type are fairly straightforward: being in a low income household is associated with cognitive ability both persistently and immediately. Conversely, once everything else is accounted for, family/employment type is not significant in either scenario. Area deprivation follows a different pattern. Stage one suggested that living in the 15% most deprived of areas is not significantly associated with cognitive ability, but stage two suggests that doing so persistently for the first three years of life is.

The third stage integrates the socio-demographic and activity factors. This shows that three of the activity measures retain a significant influence once all factors are considered together: being read to daily at 10 months, engaging in lots of activities on a daily basis at 22 months, and visiting a wide range of places/events at 22 months. All but one of the socio-demographic factors (area deprivation) remain significant. The slightly curious performance of area deprivation in these analyses suggests that the interaction between area, household and individual level factors is more complex than this type of analysis can reveal. Other techniques, such as multi-level modelling, that can explore relationships across many levels might be able to provide further insights. Furthermore, the area deprivation persistence measure might also be capturing something about geographic mobility in a way that the other variables might not be. A family can remain in the same housing for three years but see their income and employment status change; for area deprivation to change in this short period of time a change of address would also be necessary. It might therefore be worth conducting further analysis of the impact of changes in family circumstances on developmental outcomes, especially changes that could affect a child's potential to carry out activities, such as periods of maternal ill-health, partnership breakdown or moving house. However, in terms of addressing the critical first question of interest here - whether activities have an influence independent of socio-demographic factors - this analysis is sufficient. While the influence of socio-demographic factors on naming vocabulary scores is very strong, there is evidence to suggest that children who have the benefit of engaging in a wide range of activities are also at an advantage when it comes to their cognitive development. The crucial point to note from this analysis is that the important issue for children is the extent and range of activities they do, rather than specific pursuits, though being read to at an early age appears to confer an advantage in this respect as well (though note its significance level is lower than for the other two activity factors).

As noted above in the outline of the analysis stages, a final step needs to be taken before conclusions can be drawn regarding the second question, that is whether activities can moderate the influence of socio-demographic factors, or whether they instead simply enhance the outcomes of already advantaged children. The fourth stage in the analysis looked at the factors found to be significant at the third stage and explored their association with cognitive ability using a sub-set of children who have not experienced relatively high levels of advantage in terms of their household's income, their area's deprivation level or their mothers' education level. In this analysis mother's education is no longer significant, which is unsurprising as the main effect of this factor appears to be linked to degree level education, and these cases have been excluded from the analysis. Two other factors also become non-significant: mother's age at birth and being read to daily at 10 months (as mother's age at birth is strongly linked to education levels it's also not surprising to see it disappear at this final stage). The other two activity measures of the extent and range of activities children experience do remain significant. We can therefore conclude with greater confidence than is possible on the basis of the stage three results alone that activities do appear to moderate the effect of socio-demographics on cognitive ability. This means that activities appear to have a positive association with cognitive ability even among children who are not currently growing up in circumstances of relatively high affluence and advantage.

5.4.2 Picture similarities

We now look at our second measure of cognitive ability, picture similarities. Table 5.2 follows the same format as Table 5.1 and simply presents indications of the significance levels for each factor. The full results are in the Appendix. It is worth highlighting the figures in the "R squared" row and comparing these with the corresponding figures in the previous table. R squared is a measure of the amount of variation in the variable of interest that has been explained by all the different factors in the analysis. It can be expressed as the percentage of variation explained. The first point to note is that none of the analyses explain a particularly high amount of the variation (19.5% for the naming vocabulary third stage and 12.3% for the picture similarities third stage). This is typical of most social survey data. There will always be a large amount of underlying differences that are either missing from the analysis ( i.e. they were measured in the study but have not been selected for consideration in the analysis) or have been omitted from the study completely ( e.g. the kinds of genetic and biological information that is known to be associated with cognitive ability, or other aspects of child development that are hard to capture via a survey). The more important point to note, however, is the fact that the percentage of variation explained in each of the picture similarities analyses is lower than it is in each of the corresponding stages of the naming vocabulary. This suggests that the factors being used to explore cognitive ability in this report explain less of the variation in children's ability in the picture similarities assessment than is the case with the naming vocabulary test.

Table 5.2 Picture similarities regression models - summary results

Picture similarities

Regression stages

Stage one

Stage two

Stage three

Stage four

Independent variables

Mother's highest qualification

n.s.

n.s.

Area deprivation

**

Area deprivation - persistent

**

**

*

Equivalised household income quintile

**

Household income - persistent

*

*

n.s.

Family and employment type

**

Family and employment type - persistent

*

*

n.s.

Age of mother at child's birth

*

n.s.

n.s.

n.s.

Developmental concern

**

**

**

*

No. of children in family

**

**

**

**

Gender of child

**

**

**

**

Birth weight

**

**

**

**

At 10 months:

Reading with parent

n.s.

n.s.

Library visits

n.s.

At 22 months:

Daily activities

**

**

Educational games

n.s.

Number of annual visits/events

**

n.s.

Number of activities rated very important

n.s.

Importance of educational games

n.s.

Satisfaction with activities

n.s.

R squared

9.9

10.0

12.3

11.8

Constant

36.6

37.4

36.0

34.5

Significance

.000

.000

.000

.000

Sample size

3764

3736

3687

1990

Notes: **=significant at 99%, *=significant at 95%, n.s.=not significant, blanks cells indicate factors not included in the model.

The stage one analysis of socio-demographic factors shows that only mother's education level is no longer significant once other factors are also considered. This contrasts with the naming vocabulary analysis where education was significant at all stages. Area deprivation and family/employment type are also significant, neither of which was the case in relation to naming vocabulary.

When the three persistence measures are included (at stage two) instead of just the sweep 3 measures, area deprivation retains its significant association, as do the other two factors, albeit with lower levels of significance. While it isn't possible to conclude from this that persistence experiences of disadvantage are more or less influential than shorter-term ones, it does signal that persistent experiences of disadvantage should be an additional matter of concern for policy makers and analysts. Certainly as the children age, and further rounds of GUS are carried out, the role that persistent disadvantage plays in determining outcomes will accrue greater significance.

All of the socio-demographic factors that were significant at stage two retain their significance in the third stage of the analysis. In addition, two of the activities measures found to be important influences of naming vocabulary performance (engaging in lots of activities on a daily basis at 22 months and visiting a wide range of places/events at 22 months) were also significantly associated with picture similarities scores. This reinforces the conclusion drawn above that activities have an effect independent of socio-demographics, though it is clear that the precise dynamics of these relationships differ for the two assessments that were carried out.

As with the naming vocabulary analysis, the final stage then removed children from the most advantaged groups within the population and explored the impact of activities on the group remaining. As before, some of the socio-demographic factors lose their significance at this point: household income, family and employment type, and age of mother at birth. In this analysis only one of the activities measures retains a significant association - the extent of daily activities - while annual visits/events no longer has an independent effect. However, the fact remains that some activities appear to be associated with cognitive ability, once socio-demographic factors have been controlled for and once the analysis is restricted to children who have not experienced high levels of relative advantage.