developmental outcomes
8.1 Data and methods
the method for this section was to use a series of regression models to estimate the association between a lack of parental employment and children’s developmental outcomes and then to sequentially add controls for other factors that may be related both to a lack of parental employment and child wellbeing. the purpose was to assess (if applicable) to what extent these factors provide potential explanations for lower levels of child wellbeing in jobless and short part-time hours families.
the measures of child developmental outcomes were selected to cover cognitive outcomes as well as social–emotional outcomes. Only measures that were available across at least two waves of the study were used. all the measures applied to children aged at least 4–5 years, so data from the B cohort at waves 1 and 2 were not used in this section. the measures used are outlined in table 21, and the means and standard deviations for these measures by cohort and wave are shown in appendix table a5.
Table 21: Measures of child outcomes
Domain Outcome measure Values Notes
Cognitive receptive
vocabulary scaled scorehigher score = better outcome; range 28–101 Mean across the pooled data = 71.1; sd = 7.8
Measured by the Peabody Picture Vocabulary test (dunn & dunn 1997).
available for children from age 4–5 years to 8–9 years
non-verbal
intelligence standardised scorehigher score = better outcome; range 1–19 Mean across the pooled data =10.6; sd = 3.0
Measured by the matrix reasoning test
not designed to measure how children change over time, as test score is a relative measure or rank comparing children to their peers
available for children 6–7 years, 8–9 years and 10–11 years
social–
emotional total difficulties scale
Possible range of 0 (no) to 40 (most difficulties) Mean across the pooled data = 8.2; sd = 5.2
derived as the sum of parent-reported (primary carer) scores of strengths and difficulties Questionnaire (sdQ) subscales of hyperactivity, emotional problems, peer problems, and conduct problems subscales (Goodman 2001).
available for children from age 4–5 years to 10–11 years Prosocial
behaviour Possible range 0–10Mean across the pooled data = 8.1; sd = 1.8
Measured by the sdQ prosocial subscale to capture more positive social–emotional behaviours (Goodman 2001).
available for children from age 4–5 years to 10–11 years
the basic empirical approach taken was to estimate a series of regression models for each of the measures of child developmental outcome. as in the parental mental health section, two model types are estimated: random effects models and fixed effects models. (see subsection 7.4 for a description of, rationale for, and use of, these different types of models.)
for the random-effects models, four models were estimated for each measure of child developmental outcome. the first model estimated the relationship between parental employment and child wellbeing, controlling for family type (couple-parent versus single-parent) but not controlling for any other difference except for the child’s age and gender. the second set of variables included the measures of financial wellbeing in order to begin to assess how much of the relationship between parental employment and child wellbeing can be explained by the effect that joblessness has on financial circumstances, which in turn affects child developmental outcomes. the third model added to the second model the social capital variables, along with the range of socioeconomic, demographic and local characteristics of the families in which children were growing up.22 the fourth model added to the third model measures of parental wellbeing and parenting, to gain some understanding of how the effects of joblessness on child developmental outcomes are mediated by these aspects of family wellbeing.23
the explanatory variables used here are the same as those included in previous sections of the report. in addition, measures of parental mental health (the mental health of the primary carer) and parenting style are included. the measures of parenting style included in the statistical models are those of the primary carer and measure three dimensions of parenting. these dimensions are warm parenting, consistent parenting, and angry parenting (each measured on a scale of 1 to 5). information about these measures is provided in appendix a. appendix table B6 provides information on the extent to which the measures of parenting vary according to parental employment. Only small differences were apparent at this aggregate level for warm parenting and angry parenting, although primary carers in jobless families appeared to be less consistent in their parenting compared to primary carers in full-time/long part-time hours families.
CHILD WELLBEING AND DEVELOPMENTAL OUTCOMES
the fixed effects models estimated the relationship between a lack of parental employment and child
developmental outcomes, holding constant all differences between children and their families that are constant over time, irrespective of whether the characteristic was measured in the dataset. stated slightly differently, the fixed effects models allowed both observable and unobservable differences between children and their families that are constant (fixed) over time to be controlled for. the set of explanatory variables used in these analyses was more restricted than those used in the random effects models, as only those variables that had the potential to change over time could be included.
detailed results from these analyses are presented in appendix B (table B8 to B11). the tables presented below omit the coefficients for the sociodemographic variables of families and children that were included in the analyses as control variables.