Multi-Level Effects on Behaviour Outcomes in Canadian Children

1.1 Introduction

Researchers have long been interested in delineating the variables that affect children's functioning and development. Our overall research objective was to investigate multi-level effects on childhood behaviour outcomes. In particular, we examined the impact of neighbourhood, family, and individual variables on mother-reported behaviours among 2-11 year old Canadian children. We determined the amount of variation in childhood behaviour outcomes occurring at the neighbourhood, family, and individual level to examine the relative influence of each level on behavioural functioning. We included four behaviour outcomes (i.e., physical aggression, hyperactivity-inattention, anxiety-emotional problems, and prosocial behaviour) to examine possible differences in the amount of variation explained at the three levels. Finally, we examined the association between a number of neighbourhood, family, and individual variables and children's behaviour outcomes.

1.1.1 Individual Effects

Past studies traditionally have focused on characteristics of the child and his or her family. Age and sex are individual-level variables that are frequently included in models of childhood behaviour outcomes. Tremblay et al. (1996) examined the effects of child's age and sex on mother-reported aggression using data from the first collection cycle (1994-1995) of the National Longitudinal Survey of Children and Youth (NLSCY). Results revealed that boys from age 4 through to 11 years had higher physical aggression scores than girls. Age increases were associated with decreases in physical aggression, which the authors attributed to family and environmental socialization factors that generally discourage the use of physical aggression. Girls from age 4 through to 11 years scored higher than boys on indirect aggression (e.g., spreading gossip, excluding someone from a group). Mothers reported an increase in indirect aggression until age 7, followed by relative stability. While the cross-sectional design of Tremblay et al.'s (1996) study limited conclusions about intra-individual change over time, longitudinal studies have also found decreased physical aggression for girls and boys from childhood to adolescence (e.g., Brame, Nagin, & Tremblay, in press; Broidy et al., 1999; Cairns & Cairns, 1994; Tremblay, 2000). While we have some data on the direct effects of age and sex on children's behaviour outcomes, these variables often interact with other individual-, family-, and neighbourhood-level variables to influence various child outcomes.

1.1.2 Family Effects

Research on characteristics within the family that impact on childhood behaviour outcomes is abundant given that children spend much of their time within a family setting and have many of their interactions (e.g., with school) managed by family members. While a review of family characteristics was beyond the scope of this paper, we focused on family socioeconomic status (SES), family poverty, parental mental health, and parenting practices because they are commonly examined variables that were relevant to the present study.

Several studies from the first NLSCY data collection cycle of Canadian children (Boyle & Lipman, 1998; Tremblay et al., 1996; Wade, Pevalin, & Brannigan, 1999) found lower family SES to be associated with greater childhood problems, such as aggression, antisocial behaviour, conduct problems, and hyperactivity. Tremblay et al. (1996) found that 4-11 year old children from the lowest SES levels had the highest physical and indirect aggression scores. Boys had higher physical aggression scores than girls at every SES level, with the greatest sex differences found at the lower SES levels. In contrast, girls had higher indirect aggression scores than boys at every SES level, but the sex differences remained similar across levels. Two-level (individual and family) hierarchical analyses revealed that the absolute variance in physical and indirect aggression explained by family variables was 38% and 43%, respectively. Also, the total physical and indirect aggression variation between families increased as a function of decreasing SES level. These results indicated that in lower SES households, family factors played a greater role in children's use of aggressive behaviour. Boyle and Lipman (1998) also found strong family SES effects using the same NLSCY 4-11 year olds. Specifically, of the absolute variance in childhood behaviour problems associated with family and neighbourhood variables (25%), 18% was attributable to family SES relative to neighbourhood-level variables.

The effects of family SES have been demonstrated not only in cross-sectional designs such as those from the NLSCY but also in longitudinal studies beginning in early childhood (e.g., Dodge, Pettit, & Bates, 1994; Pagani, Boulerice, Tremblay, & Vitaro, 1997). Dodge et al. (1994) followed a sample of children from preschool to middle elementary school and found that SES assessed in preschool significantly predicted teacher-reported externalizing problems and peerrated aggression in kindergarten through to Grade 3. Boys had significantly higher peer-rated aggression scores than girls. Lower SES was also associated with a host of other family variables, including harsher parental discipline, lack of maternal warmth, and exposure to violence.

Family SES appears strongly related to family poverty, which has been linked to problematic childhood behaviours. McLoyd's (1998) literature review found that poor youth had higher rates of behaviour problems compared to peers from middle-class households, that poverty had a more pronounced impact on externalizing rather than internalizing problems, and that poverty duration was positively correlated with youth behaviour problems. The effects of poverty often are mediated by parenting practices that are harsher and more inconsistent as a result of the increased stress exerted on poor parents (Klebanov, Brooks-Gunn, & Duncan, 1994; McLeod & Shanahan, 1993; McLoyd, Ceballo, & Mangelsdorf, 1997). For example, McLeod and Shanahan's (1993) study of 4-8 year olds found that mother-reported externalizing problems were greater in poor families. However, these effects were mediated by parenting behaviours, specifically poor mothers were more physically punitive toward their children and less emotionally responsive to their children's dependency needs. These frequently punished children, in turn, were found to have more externalizing problems compared with less-punished children. Particularly among single parents, factors such as the greater likelihood of poverty, the overwhelming responsibilities of solo parenting, the increased exposure to discrete stressors (e.g., changes in job, residence), and the greater social isolation place them at increased risk for mental health problems.

Poverty appears to be associated not only with parenting practices but also with parental mental health (Klebanov et al., 1994). One frequently studied mental health problem is parental depression. Downey and Coyne's (1990) review paper found that compared with school-age children who did not have depressed mothers, children of depressed mothers exhibited more internalizing and externalizing problems and were at increased risk for affective disorders. These children also had a number of additional problems, such as higher levels of treatment for psychiatric disturbance, greater deficits in social and academic competence, and greater physical health problems. The relationship between maternal depression and children's adjustment appeared to be mediated by mother-child interactions. Specifically, Downey and Coyne (1990) found that depression decreased the amount of effort that mothers put into interacting with their child and increased the amount of hostility and negativity directed toward the child. Depressed mothers also tended to use more coercive and punitive techniques to manage their child's behaviour. An important finding from the literature review was that these parenting difficulties were not depression-specific but rather were commonly found among mothers experiencing distress, either because of personal problems (e.g., physical illness), family difficulties (e.g., marital conflict), or neighbourhood characteristics (e.g., poverty).

Children's aggressive behaviour has also been linked with mothers' own history of childhood aggression and/or withdrawal. Serbin et al. (1998) examined the intergenerational transfer of psychological risk in a sample of women with histories of childhood aggression and/or withdrawal. Results indicated that childhood withdrawal in mothers significantly predicted aggressive behaviour in their children. Mother's childhood aggression also showed a trend in predicting their children's aggressive behaviour. Interestingly, maternal reports of their children's problem behaviours were predicted primarily by level of education, with less educated mothers reporting greater problems. The authors speculated that highly educated women may have a better understanding of child development, fewer psychiatric problems (e.g., anxiety, depression), and fewer financial and social support problems.

1.1.3 Summary

There is much existing research on the impact of families on children's functioning and development. Evidence from cross-sectional and longitudinal studies shows that low family SES is associated with a number of childhood behaviours including aggression and conduct problems. Family SES is also strongly related to poverty which impacts on childhood behaviours, especially externalizing problems. Research suggests that poverty's effects on children are often mediated by parenting practices, which are harsher and more inconsistent as a result of the increased stress exerted on poor parents. The effects of parental mental health problems (e.g., depression) on childhood behaviours are also often mediated by parenting practices that render parents less responsive and more negative toward their children.

1.1.4 Neighbourhood Effects

While most studies of children's behaviours have focused on individual and family characteristics, investigators have recently begun to incorporate variables within the child's environment. However, empirical studies on neighbourhood variables remain limited compared to those focusing on the impact of individual and family variables on behavioural outcomes. Boyle and Lipman (1998) attributed this finding partly to the fact that until recently, there were no appropriate statistical methods for analyzing multi-level (i.e., individual, family, and neighbourhood) data. It may also be that past studies lacked neighbourhood information because no such measures were administered during data collection. We also found that many of the existing neighbourhood studies used adolescent samples. This finding may be related to adolescents' greater interaction in their neighbourhoods compared to children.

Studies that have incorporated neighbourhood variables for child samples have found internalizing and externalizing problems to be linked to such neighbourhood variables as community disadvantage, based on measures of unemployment, poverty, and low education (Boyle & Lipman, 1998; Dubow, Edwards, & Ippolito, 1997; Wade et al., 1999), crowded living conditions (Chase-Lansdale & Gordon, 1996), single-parent households (Boyle & Lipman, 1998; Shumow, Vandell, & Posner, 1998), low-income neighbours (Klebanov et al., 1994; Shumow et al., 1998), exposure to violence (Shahinfar, Fox, & Leavitt, 2000), and greater perceived danger and less perceived social cohesion (Shumow et al., 1998). Reviews of the literature have found that the most consistent evidence of neighbourhood effects occurs for school-age children, that neighbourhood effects are stronger for cognitive and achievement measures than for behavioural and mental health measures, and that among the most commonly studied neighbourhood variables, SES demonstrates the most consistently powerful effects (Duncan & Raudenbush, 1999; Leventhal & Brooks-Gunn, 2000).

In addition to evidence of direct neighbourhood effects (e.g., SES, poverty) on childhood behaviour outcomes, several studies (Dubow et al., 1997; Shumow et al., 1998) have shown how neighbourhood perceptions can mediate this relationship. For example, Shumow et al. (1998) found that maternal perceptions of neighbourhood danger (e.g., drug dealers, gangs, violence) mediated the relationship between neighbourhood demographic characteristics and reports of psychological distress and misconduct for school-age children. Furthermore, children's perceptions of neighbourhood danger mediated self-reported psychological distress. Another neighbourhood variable that has been found to mediate the relationship between neighbourhood demographics and behavioural outcomes is collective efficacy, defined as social cohesion and mutual trust among neighbours combined with their willingness to intervene on behalf of the common good. While data on childhood behaviours are currently unavailable, Sampson, Raudenbush, and Earls' (1997) study of adults living in Chicago neighbourhoods provided an important illustration of the mediating effect of collective efficacy in the relationship between neighbourhood risk (i.e., concentrated disadvantage, immigrant concentration, residential instability) and violent crime. Results indicated that neighbours' perceptions of collective efficacy showed a statistically significant negative relationship with violent crime, even after having controlled for neighbourhood risk.

1.1.5 Summary

Research on the impact of neighbourhood characteristics on children's behaviour outcomes currently is limited. The available data suggest that neighbourhood SES/poverty shows the strongest and most consistent association with behaviour problems. There is also some evidence that individuals' perceptions of neighbourhood danger and problems and their sense of social cohesion and informal social control may mediate the effects of objective neighbourhood characteristics. The influence of neighbourhoods, relative to that of families, generally is small to modest. Boyle and Lipman's (1998) study from the 1994-1995 NLSCY data collection cycle revealed that the absolute variance in 4-11 year olds' behaviour problems explained by a two-level model including neighbourhood and family variables was 25%. However, only 7% of this variance was explained by neighbourhood variables (i.e., unemployment, low income, low education, and single-parent households) while 18% was explained by family variables (i.e., poverty, singleparent status, low SES, and greater number of siblings in the family). Leventhal and Brooks-Gunn's (2000) review paper also concluded that neighbourhood variables account for approximately 5-10% of the variance in child and adolescent outcomes.

1.1.6 National Longitudinal Survey of Children and Youth

Given the importance of understanding the neighbourhood, family, and individual variables that influence child behaviour outcomes, Human Resources Development Canada initiated the National Longitudinal Survey of Children and Youth (NLSCY). The survey follows a representative sample of newborn to 11-year old Canadian children into adulthood, collecting information related to the children every two years. The objective is to develop a national database on childhood characteristics and life experiences from infancy to adulthood in order to obtain information for policy analysis and program development. The present study used data from the first NLSCY cycle (1994-1995) to examine the impact of neighbourhood, family, and individual variables on behaviour outcomes among 2-11 year old Canadian children. Similar to Boyle and Lipman (1998) who employed this data set, we investigated child-, family-, and neighbourhood-level effects on a range of child outcomes, specifically physical aggression, hyperactivity-inattention, and anxiety-emotional problems. We also expanded the scope of our investigation by including healthy prosocial behaviours (i.e., acts of helpfulness, cooperation, and encouragement of others), incorporating data for 2-3 year olds, and using subjective as well as objective neighbourhood measures.

An important difference between our study and that of Boyle and Lipman (1998) was in the operational definition of neighbourhood. Boyle and Lipman (1998) used enumeration areas (EA), which are the smallest geographic units for which census data can be easily retrieved. Each EA has a minimum of 375 households for large urban areas and a minimum of 125 households for rural areas. In contrast, we defined neighbourhoods using Statistics Canada 1996 census tracts, which represent a larger area than EAs. Because of the small area represented by EAs, there were many instances in which Boyle and Lipman (1998) only had data available for one or a few households. Using multiple family and neighbourhood variables to describe these particular EAs and examining interactions among variables creates a great deal of measurement error. Analyses using such small areas confound family- and neighbourhood-level effects, making it difficult to tease apart their separate relative impact.

Contents