Chapter 2: Methods
This section contains information on how the analysis sample was selected, the measures of supportive father-child and mother-child relationships used in the report, and statistical methods used. It concludes with an outline of sample characteristics.
2.1 Sample selection
The data used were taken from sweep 8 of the first GUS birth cohort, conducted in 2014/15 when children had a median age of 10 years and were in the Primary 6 year group at school. Interviews were conducted with 3151 families (60% of the 5217 families interviewed in the first sweep of data collection).
Children who were not living in couple families were excluded (n=539), together with cases where the children were not living with the biological mother (a further 16 cases). Three further cases were removed where the mother's partner was female. The final analysis sample contained 2593 families (82% of families surveyed at child age 10), all with children living in families headed by two resident parents, one being the biological mother and the other her male partner. In most cases this partner was the child's biological father (n=2411), with a minority of children (n=182) having a non-biological resident father figure. Note: unless otherwise specified, "father" throughout this report refers collectively to biological fathers and resident non-biological father figures.
2.2 Measuring father- and mother-child relationships
At age 10, children supplied information on each parent's supportiveness, using an audio computer-assisted self-completion questionnaire conducted in the child's home. Supportiveness was measured using nine items from the trust and communication subscales of the People in My Life ( PIML) scale, a self-report measure of child attachment designed and validated for use in middle childhood (6-12 years) (Ridenour, Greenberg & Cook, 2006) The validation process used a sample of 10-12 year olds to establish that the overall PIML attachment scale was correlated as expected with other measures of children's behavioural and emotional adjustment, as reported by parents, teachers and children themselves. It gives us confidence that items used for this report are appropriate for the GUS ten year-olds.
Children answered items about either a resident biological father or a resident non-biological father figure. The main carer's report of people resident in the household was used as a proxy to determine whether the child was referring to their biological father or a father figure. In most cases, the mother completed the main carer report. (Note that where there is a father figure, we do not have information on how the same child related to the non-resident biological father.)
Children were invited to indicate their agreement with the following statements: "My Dad listens to what I have to say", "My Dad cares about me", "I can count on my Dad to help me when I have a problem", "My Dad can tell when I'm upset about something", "I talk to my Dad when I am having a problem", "If my Dad knows something is bothering me, he asks me about it", "I share my thoughts and feelings with my Dad", "My Dad pays attention to me", and "My Dad is proud of the things I do". Responses were on a 4-point scale: 1 "never true", 2 "sometimes true", 3 "often true", 4 "always true".
Children were also invited to indicate agreement with a similar set of statements about their mother.
Figure 2-A shows the percentages of children giving the most positive response ("always true") to each statement, in relation to fathers and mothers. For most items, the majority of children gave the most positive response, although children's reports for fathers were less positive than for mothers. The vast majority (over 90%) felt that their Dad and Mum "care about me". For both parents, the two items with the least positive responses related to the ability of the child to confide in a parent ("I talk to Dad/Mum when I am having a problem", and "I share my thoughts and feelings with my Dad/Mum"). Less than half of children thought these were "always true" in relation to their father.
A factor analysis of scores demonstrated that the questionnaire items for children's relationship with fathers all related to the same underlying concept, since items all loaded on to one factor. In other words, it confirmed that it was appropriate to combine responses to these questions to produce a scale or aggregate measure of father-child relationship. The same was true of the items for mothers. Average scores were calculated for each set of questions. Scores had excellent internal reliability (Cronbach alpha for the father-child relationship was 0.90, and for the mother-child relationship was 0.84). The high values of Cronbach alpha (>0.7) show that children responded in a very consistent way across all nine items.
Figure 2-A Children's views on parental support, measured using items from the People In My Life scale
Note: Base sample n=2593 (unweighted).
The distribution of average scores for father-child relationships was negatively skewed (see Figure 2-B). In other words, most children had high scores, with a sizeable number having the maximum score of 4. However, there was a "tail" of children giving less positive responses on average (i.e. scoring 1 or 2). For the purposes of this study, these average scores were then recoded to form three groups, using the cut-off points shown in Figure 2-B:
- Poor relationship: average scores from 1 to less than 3 (i.e. on average, children gave one of the two less positive responses to the items, with the father "never" or "sometimes" supportive)
- Good relationship: scores of 3 or more but less than 4
- Excellent relationship: the maximum score of 4 (children "always" felt supported by the parent, across all items).
Figure 2-B Distribution of average father-child supportive relationship score, and cut-off points applied
Note: Base sample n=2593 (unweighted).
This process was repeated for items relating to the mother/ mother figure. Figure 2-C shows the distribution of average scores for the father-and mother-child relationship on the same graph. Mother-child relationship scores also have a strong negative skew, although the "tail" of low values is not as pronounced as for father-child relationships.
Figure 2-C Distribution of father- and mother-child average relationship scores among couple families
Note: Base sample n=2593 (unweighted).
2.3 Statistical methods
All analyses took account of the stratified, clustered sample design and used survey weights. Use of weights helps to compensate for the effects of differential survey attrition from the first sweep of data collection at child aged 10 months, which is more pronounced for disadvantaged groups.
Predictors of poor father-child relationships were first explored using bivariate associations (i.e. simple associations between pairs of measures). Multivariable models were then used to predict the likelihood of a poor father-child relationship for any one risk factor, after controlling for other factors in the model. The analysis of associations between father-child relationships and other aspects of child socio-emotional wellbeing took a similar approach. It first explored bivariate associations between father-child relationship quality and each measure of low wellbeing. Multivariable models were then used to predict the likelihood of each poor wellbeing outcome being associated with either poor or excellent father-child relationships, using good father-child relationship as the reference group. Further explanation of these models is provided in the Technical Annex to this report.
2.4 Characteristics of the analysis sample
Most children in the analysis sample of couple families (90%) lived with both biological parents, with the remaining 10% living with their biological mother and her male partner (father figure to the child). Note that these percentages, and those described elsewhere in the report are all weighted to compensate for survey attrition, see section 2.3 above.
The average age of children was 10 years and 2 months. The sample contained approximately equal numbers of boys and girls. In 3% of families, at least one parent was from an ethnic minority group. Families with a non-biological father figure contained a higher proportion of younger parents: when children were aged 10, 18% of mothers in families with a father figure were under 30, compared to only 3% of families with a resident biological father. Families with a non-biological father figure were also more likely to be disadvantaged: for example, 42% were in the lowest quintile of household income (<£13,450 p.a., equivalised to take account of household size and composition), compared to 15% of couple families with a biological father. Further details are provided in an Appendix (Table 9-1).
Email: Wendy van der Neut