Family Relationships and Children's School Achievement: Data from the National Longitudinal Survey of Children and Youth - October 1998

4.1 Sample

The records of 4,302 (2,134 girls and 2,168 boys) Canadian children included in the first data collection cycle of the National Longitudinal Survey of Children and Youth comprised the sample for this study. While the survey is intended to provide representative information on Canadian children between birth and 11 years of age, only those school-age children from 6 to 11 years were included in this study. Because of the sampling design of the survey, the children and families in the study reflect a broad range of social classes, ethnic origins (excluding Natives), and geographic locations in all the provinces and territories of Canada.

4.2 Measures

Achievement was measured by a single item from the NLSCY teacher questionnaire. Teachers rated each child on the question, "How would you rate this student's current achievement across all areas [reading, mathematics, written work]? Teachers rated student achievement on a 5-point scale from, "near the top of the class" to "near the bottom of the class."

An Academic Focus Scale was developed by combining scores on six items from the teacher questionnaire. Children were rated by their teachers on a variety of academic skills. Sample items from the scale are, "Listens attentively", "Follows directions", "Works independently." Higher scores indicate better levels of academic focus. Cronbach's alpha for the scale was .91.

The Hyperactivity-Inattention Scale, consisting of 8 items from the parent questionnaire, provided a measure of the children's level of hyperactivity/inattention. Sample items are, "Can't sit still, is restless or hyperactive" and "Can't concentrate, can't pay attention for long". Higher scores indicate greater numbers of hyperactive-inattention behaviours. The alpha for this scale was .84.

The Positive Interactions Between Parents and Child Scale, consisting of 5 items from the parent questionnaire, provided a measure of positive, supportive interactions between parents and children. Higher scores indicate more positive interactions. The alpha value was .81. Sample items are, "How often do you praise (name) by saying something like 'Good for you!' or 'That's good going!" and "How often you and he/she talk or play with each other, focusing attention on each other for five minutes or more, just for fun."

The Hostile-Ineffective Parenting Scale, consisting of 7 items from the parent questionnaire, was selected as a contrast to the positive parenting measure. Higher scores indicate more hostile-ineffective parent-child exchanges. The measure of hostile parenting contained items such as, "How often to you get angry when you punish (name)?" and "How often do you get annoyed with (name) for saying or doing something he/she is not supposed to do?" The alpha for this scale is .71.

The Family Functioning Scale, consisting of 11 items from the parent questionnaire, provided a measure of the level of overall dysfunction in the family with higher scores indicating increased amounts of dysfunction. Sample items are, "In times of crisis we can turn to each other for support," "We express our feelings to each other," and "Making decisions is a problem for our family." The alpha for this scale is .88.

Two measures of the parent characteristics were drawn from the parent questionnaire. Technically speaking, the respondents assessed in each case are called the "person most knowledgeable" (PMK) in the survey documents. Females make up 91.8% of the PMKs which means that the data overwhelmingly reflect the responses of mothers or female guardians. Of the female PMKs, 98.0% are biological mothers. The first of the two measures available for this level of the model is the Parental Depression Scale which consists of 12 items with an alpha of .82. Higher scores indicate increased levels of depression. Sample items from the scale are, "How often have you felt or behaved this way in the last week: I felt lonely; I had crying spells; and I felt hopeful about the future." The second measure is the Social Support Scale on which the respondents rate their feelings of security and sense that they have useful social supports around them. It consists of 6 items. Higher scores indicate greater social support. The alpha for the scale is .82. Sample items are, "I have family and friends who help me feel safe, secure and happy" and "There are people I can count on in an emergency."

Socio-Economic Status (SES) was determined for the NLSCY by standardizing the measures of education levels for the PMK and spouse, the prestige of occupation of the PMK and spouse, and the household income. Then, a mean of the five standardized measures was calculated to yield the SES variable. Higher scores indicate higher levels of socioeconomic status.

4.3 Data Analysis Procedures

The data were analysed using linear structural equation modelling (Jöreskog & Sörbom, 1989) largely because of the technique's capacity to examine simultaneously a large number of direct and indirect associations between variables. Such analyses are often used to examine underlying causal processes although cross sectional data, such as those obtained in the first cycle of the NLSCY, make interpretations of causality difficult. In such situations, strong theory is needed to make strong causal claims. Nevertheless, this technique is highly suited to analyses of large samples with multiple variables that are understood to be interacting in highly dynamic ways.

In all of the analyses reported, the observed models were all derived from covariance matrices. Because single indicator latent variables were used in the analysis, variable error variance was estimated by multiplying the observed variance for each variable by 1 minus alpha (Hayduk, 1987). To begin, we determined maximum likelihood estimates for each predictive relation in just-identified models. Models were then trimmed using statistical significance at the .01 level and then by eliminating all paths with coefficient values less than .1. These two very stringent criteria were used because of the very large sample sizes in the models. For large samples, even relationships of very small effect size can be statistically significant. Because the goal of the present study was to identify highly robust and meaningful relationships between variables, small effects or weak relationships were discarded.

One of the original objectives of the study was to investigate the possibility of developmental change in the relationship between family processes and children's school achievement (Research Question 2). Accordingly, the total sample of 4,302 children was divided into males and females and then into three age groupings of roughly equal size within each sex: 6/7 year-olds, 8/9 year-olds, and 10/11 year-olds. Results showed that the core models supported by the data were essentially the same across all age groups with only weak and unstable differences appearing in the different age groups. We concluded that within the age range covered by the survey data (6 years to 11 years) no meaningful developmental differences were observable. In future data collection cycles, when data can be compared between children in primary school and others in junior or senior high school, age differences may indeed emerge. As a consequence of these preliminary analyses, all the age groups were collapsed. The final models report analyses for boys and girls separately as well as a single model with both sexes combined.

One final general point concerns the placement of the social support variable in the model. Because we used the family-school relationships model to establish the expected relationships between variables, the usual placement of social support would have been at the same level as parental depression. There are logical reasons, however, for considering that social support is more distant from the child's achievement than is parental depression. In a sense, the social support variable, while indicating something of the parent's state of mind, also describes the parent's perceived context and could be properly considered to belong to Level 6 which is formally concerned with contextual circumstances. From this perspective, it is possible that an environment that is not supportive could lead to increased levels of parental depression. For these reasons, we put social support to the left of parental depression in the models and another step further removed from the achievement outcome variable.