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

Previous reviews of the literature concerned with family process influences on children's school success (see, for example, Christenson, Rounds, & Gorney, 1992; Hess & Holloway, 1984; Kellaghan, Sloane, Alvarez, & Bloom , 1993; Ryan & Adams, 1995; Scott-Jones, 1984) have shown that a very large number of family relationship processes variables and family member characteristics are significantly related to school success or failure. Almost all of the studies in this field, however, are limited by two significant problems. First, they typically employ smaller, convenience samples that usually unduly restrict the range of values associated with each variable studied. For example, when researchers have to rely only on those parents who respond positively to letters inviting them to participate in studies of social process, agreement is typically forthcoming from reasonably well functioning, educated, middle-class families. In these circumstances, variable distributions are highly likely to be truncated which has an impact on the capacity of the researchers to uncover significant relationships among the variables. Many distorted finding likely result.

Second, the number of constructs usually measured in these studies is small due to the financial restraints faced by social science researchers combined with the unwillingness of most families to respond to a large number of measures. The result is that relatively few variables are studied together even though the literature suggests that large numbers of family processes and family member characteristics act simultaneously and jointly to produce schooling outcomes in children.

The NLSCY data set offers an opportunity to avoid these two significant problems in the study of family effects on schooling outcomes. The sample is large and serious attempts were made in the data collection to ensure that it is representative of Canadian children in general (with the exception of Native children who were not included in the survey). This ensures that a wide range of families are included in the study and that the scores on each of the measures are more likely to mirror the distribution of the sampled characteristics in the population. In addition, the survey includes a large number of important constructs; in fact, more measures are included in the survey than can readily be dealt with in any single analysis. Nevertheless, the data offer an excellent opportunity to examine the joint effects of a substantial number of variables. Dealing more effectively with these two problem areas makes the present study particularly important in the field.

The selection of constructs and measures from the survey was guided by the structure of the Family-School Relationships Model (Ryan & Adams, 1995). The constructs included in the first cycle of data collection were examined to determine which variables could be used at each of the seven levels of the model. Constructs from each of the levels in the model were identified with the exception of level two which is concerned with school-focused parent-child interactions. Aside from some of the self-report data from the 10-11 year-old subjects (these self reports were not used in this analysis), the survey does not contain level two items. The operational definitions for each of these constructs are given below in the methods section of the report.

Level 0 - Achievement Outcomes. The survey includes information on overall child achievement as rated by parents and teachers. The data derived from teacher rating of achievement were used in the present study. These data were used instead of the rating of achievement by parents because teachers' knowledge of how the child stands in relations to his or her classmates is likely to be more accurate and informative of school success.²

Level 1 - Child Characteristics. Constructs concerning two child characteristics were selected for study. First, the survey provided data on academic focus which is a general construct encompassing a variety of skills and attitudes that promote learning. The academic focus construct is similar in character to what Gesten (1976) identified as a, 'good student' factor and what Ketsetzis, Ryan, and Adams (1998) referred to as 'intellectual effectiveness'. Prior evidence shows that such characteristics are closely linked with achievement.

Second, a measure of hyperactivity and attention deficit was included. The association between attention difficulties and school achievement has long been noted (Barkley, 1990; Hinshaw, 1992; Maguin, Loeber, & LeMahieu, 1993) although the precise nature of the causal linkage between the condition and achievement is still in question (Coie & Dodge, 1998).

Level 3 - Parent-Child Interactions. From the extensive number of measures of parent-child interactions available in the survey, two were judged representative of the general areas parent-child research suggests are central. First, a measure of positive parent-child interactions was selected. While there is no clear evidence that simply behaving positively (as distinct from, for example, authoritatively) with a child will lead to better achievement, a host of other positive developments for children do appear strongly associated with positive parent-child interactions (e.g., Bar-Tal, Nadler, & Blechman, 1980; Fabes, Eisenberg, Karbon, Troyer, & Switzer, 1994).

Second, the measure of hostile-ineffective parenting was selected largely as a contrast to the positive parenting measure. Patterson's work (e.g., Patterson, Reid, & Dishion, 1992) has extensively documented the effects of coercive parenting practices and their very negative outcomes for children as well as whole families.

Level 4 - General Family Functioning. The NLSCY included only one construct focused on overall family functioning that was applicable to all school-aged subjects: family dysfunction. Previous studies of family process and schooling outcomes (Conger, Conger, Elder, Lorenz, Simons, & Whitbeck, 1992; Forehand, Thomas, Wierson, Brody, & Fauber, 1990; Grolnick & Slowiaczek, 1994) have shown that the general nature of family relationships can have a significant impact on both academic and social success in school.

Level 5 - Parent Characteristics. Two constructs from this level of the model were taken from the survey. The first, parental depression, has been shown in other research to have negative consequences for schooling outcomes (Forehand, McCombs, & Brody, 1987; Roseby & Deutch, 1985; Thomas & Forehand, 1991).

The second construct from the survey at this level was perceived social support which was taken as a measure of a the parent's sense of security. Such support or the perception of such support has been shown to provide parents with a buffer against a wide variety of negative forces that operate on them in difficult contexts (Garbarino, 1992). In general, the social support literature shows that well supported parents are less likely to suffer emotional difficulties and are less likely to be in families of greater dysfunction.

Level 6 - Exogenous Social Variables. The single level 6 variable used in the analysis was the survey's measure of socio-economic status (SES). Perhaps no other single variable in the social sciences over the last 5 decades has been so consistently shown to have an impact on children's school success (Booth & Dunn, 1996). One of the problems with small, convenience samples in the literature focused on families and schools is that the range of SES levels in the samples is often rather narrow and usually tilted toward the middle and higher-middle classes (these families are usually more ready to participate in research studies). As a consequence, SES frequently fails to contribute strongly to effects on schools and when it does its effects are largely mediated by family processes. The representativeness of the NLSCY sample makes the use of the SES measure especially valuable in this context.

• ²A different and more objective outcome variable for the study might have been the math test scores included in the survey. Measurement difficulties, however, arose in the test forms used in the first cycle of data for the survey. Because two form of the test were used, there were marked ceiling effects for the children in grades 3 and 5 which sharply reduces the usefulness of the measure in any analyses encompassing these two grade levels. Another outcome variable possibility might have been the Peabody Picture Vocabulary Test (PPVT) score although, for the purposes of this report, it is probably best regarded as a measure of the children's ability and placed at Level 1 in the model. Moreover, the PPVT was administered only to the 4 and 5 year-old group.