Multi-Level Effects on Behaviour Outcomes in Canadian Children - May 2001

Executive Summary

Previous

Contents

Next

The investigation of childhood behaviours has recently broadened to include the impact of contextual variables such as the neighbourhoods in which children live. With his influential book "The Truly Disadvantaged" William J. Wilson (1987) refocused social science attention on the role of neighbourhoods in shaping children's lives. He argued that increases in concentrated poverty have resulted in a "new urban poverty" in which today's poor children are increasingly exposed to environments characterized by economic hopelessness, community disorder, and violence - realities with far-reaching consequences for children's development (Wilson, 1997). Methodological advances (e.g., Bryk & Raudenbush, 1992; Goldstein, 1995) and improved sources of data have enabled a new generation of researchers to study neighbourhood effects on children. The emerging consensus is that neighbourhoods indeed "matter" for children yet how much they matter and the specific reasons (i.e., "how" they matter) remain somewhat unclear (for a recent review see Gephart, 1997).

While the majority of studies to date have been limited to the United States, research shows that concentrated neighbourhood poverty is also on the rise in Canadian cities (Hajnal, 1995; Hatfield, 1997). One of the problems with these U.S. studies is that many are based on restricted range samples of a single metropolitan area or disadvantaged population. The National Longitudinal Survey of Children and Youth (NLSCY) offers a unique source of data with which to examine neighbourhood effects on children across Canada. The use of the 1996 Census data to describe neighbourhood demographic and economic structures, in combination with parent and interviewer ratings of neighbourhood conditions, enabled us to assess specific characteristics of neighbourhoods assumed to influence children's behaviours. We were also able to assess the relative impact of neighbourhood, family, and individual characteristics and to identify family characteristics that best predicted childhood aggression. The spatial definition of neighbourhoods for the first study was Statistics Canada 1996 census tracts. In order to maintain adequate reliability for the neighbourhood problems and collective efficacy measures, we eliminated census tracts with less than 15 households. Our sample consisted of 96 census tracts including 1,982 families and 2,745 children aged 2 to 11 years. The second study used both census tracts and enumeration areas that have been cluster analysed to yield eight types of neighbourhoods reflecting various combinations of social class, immigration status, and family type composition.

The report is presented in two parts. The first part was a broad examination of neighbourhood and family effects on children's physical aggression, hyperactivity-inattention, anxiety-emotional problems, and prosocial behaviours. We believed that it was important to consider a wide range of behaviour outcomes as previous studies have found differing magnitudes of neighbourhood effects for emotional versus behavioural and antisocial versus prosocial outcomes. We first estimated the amount of variation in these behaviour outcomes occurring within and between neighbourhoods, families, and individuals. We then examined the relative influences of neighbourhood, family, and individual variables on this variation. The essential finding from this analysis was that across the four behaviour outcomes, the greatest amount of variation occurred between individuals, a moderate amount of variation occurred between families, and a small amount occurred between census tracts. This finding replicates the general trend of results from similar studies (see Brooks-Gunn, Duncan, & Aber, 1997a,b). The second part also focused on aggression but distinguished between physical forms of aggression, which are more common among males, and indirect forms of aggression (e.g., gossip, social exclusion), which are more typically exhibited by females. Further, we tested for direct and mediated effects of neighbourhood objective and subjective characteristics on these forms of aggression.

Neighbourhood Effects

What do these two separate studies reveal about the role of neighbourhoods and families on Canadian children's behaviour outcomes? Using slightly different approaches, both studies found that neighbourhood characteristics had much less of an impact on children's behaviour than individual and family characteristics. One consistent finding was that children living in neighbourhoods reported to have higher levels of problems (e.g., crime, disorder) were significantly more likely to be characterized by parents as being aggressive, both physically and indirectly. Another common finding was that poor neighbourhoods in Canada were not necessarily more likely to produce aggressive children. The first study found that the percentage of families living below the poverty line was negatively associated with physical aggression, controlling for family socioeconomic status (SES) and other family and neighbourhood variables. Using a clustering (as opposed to a continuous variable) approach, the second study found that both high SES neighbourhoods and low SES neighbourhoods had lower rates of physical aggression, in comparison to middle-class neighbourhoods. In the case of indirect aggression, the second study found highest rates in enumeration areas with higher percentages of single parents and immigrants. Thus, the effect of neighbourhood SES on children's aggression is complex and may vary depending on the type of aggression measured.

Neither population size nor perceived neighbourhood cohesion was significantly related to physical aggression. However, it seems premature to conclude that neighbourhood cohesion is unrelated to children's behaviours. The cross-sectional design of the studies could have led to this insignificant result, specifically social cohesion at a single point in time may have represented a response by neighbours to existing violence and disorder, which would have cancelled out its expected protective effect. It is also possible that social cohesion is only significant when considered in combination with other neighbourhood characteristics such as SES or the presence of violence and disorder. Further, the protective effect of social cohesion may be more important for older youth that are in more direct contact with the neighbourhood. Data from subsequent NLSCY collection cycles will permit a more definitive answer to these issues.

Turning to childhood behaviour outcomes other than aggression (i.e., hyperactivity-inattention, anxiety-emotional problems, prosocial behaviour), the only neighbourhood effect reported in the first study involved population size. In particular, children living in rural neighbourhoods were less likely to exhibit anxiety-emotional problems compared to children in large cities. However, several issues should be kept in mind prior to concluding that neighbourhoods have little influence on children's behaviours. First, concerns about the study's cross-sectional nature apply here as well. Second, random effects analyses indicated more neighbourhood variation in these outcomes than for physical aggression. Almost 10% of the random variation in prosocial behaviour occurred between neighbourhoods as opposed to only 4% for physical aggression. While aggression and prosocial behaviour are sometimes correlated, they may also be theoretically independent phenomena that require different explanatory neighbourhood variables. Research on neighbourhood effects will also need to separate empirical effects from selection processes. Selection effects may arise as families exert choice within constraints in determining where they live. If the unobserved factors that affect residential location also affect developmental outcomes of children, then the failure to include those unobserved factors in the models may lead to biased estimates, either in the form of the overestimation or underestimation of neighbourhood effects on children's outcomes.

Individual and Family Effects

Results revealed that a number of individual and family variables were strongly associated with children's behaviours. As expected, boys were reported to be more physically aggressive and hyperactive-inattentive and less prosocial than girls. Behaviour outcomes also varied by age, specifically older children were reported to exhibit greater anxiety-emotional problems and prosocial behaviours and less hyperactivity-inattention.

Family structure played a role in behaviour outcomes in that children from single-parent families were more physically aggressive, hyperactive-inattentive, and anxious-emotional than children living with both biological parents. Indicators of family SES had few effects on children's behaviours once other family, neighbourhood, and individual characteristics were controlled. At first glance this finding might seem at odds with our previous research demonstrating a strong SES gradient for physical aggression (Tremblay et al., 1996). We believe the insignificance of family SES is partly explained by the very strong and consistent effects of parent and family social process variables, which are significantly related to family SES. For example, mothers' own psychological well-being had a strong effect on their reports of children's behaviours. Positive and hostile parent-child interactions also had an important impact on children's behaviours, with positive interactions associated with prosocial behaviour and hostile interactions associated with all negative outcomes. Likewise, parenting strategies were consistently significant covariates, with punitive parenting having detrimental effects and consistent parenting having protective effects. Thus, from a proximal perspective, parent and family characteristics likely mediated the relationship between family SES and children's behaviours and from an intergenerational perspective, parent characteristics likely led to both family SES and family characteristics (Nagin & Tremblay, 2001; Zoccolillo, 2000). Of course, intergenerational data are needed to test these hypotheses. Our limited measure of family SES at a single point in time is also an issue as it did not capture issues of onset, duration, and chronicity or their interplay with children's age.

Policy/Service Implications and Directions for Future Research

Policy makers and service providers responsible for behaviour problem prevention policies and services should consider the fact that the present study confirmed results from other studies indicating that, before age 12 years, individual and family characteristics are more strongly related to children's behaviour problems than neighbourhood characteristics. From our analyses of the first wave of the NLSCY data, children who appear most at risk of behaviour problems are young males living in a dysfunctional family with young depressed mothers who do not live with the father. However, policy makers and service providers need to keep in mind that this study is cross-sectional, and correlates are not causes. To better understand the mechanisms that are involved in the development of these costly problems we need data over many more NLSCY waves. In most cases the causes of behaviour problems are complex and appear to build up over long periods of time. Longitudinal and experimental data are needed to understand these mechanisms. With time the NLSCY data collection cycles will allow us to more rigorously test mechanisms. Multiple data points are particularly important for modelling more complex relationships such as mediated and interactive effects. Longitudinal data will also further enable us to more properly conceptualize childhood behaviour adjustment as a developmental process that changes over time.

Because there is good evidence of intergenerational transmission of behaviour problems from the present study, and from many other longitudinal studies, the best advice to policy makers and service providers for the prevention of behaviour problems is for them to take a long-term perspective. To prevent behaviour problems one probably needs to make long-term investments in early child development through support to adolescents and young adults who are and will be the next generation of parents of young children. From this perspective, although males are those with the highest levels of problems, females with problems, apparently less serious, could be a better investment in the long run, since they are the ones most involved in providing the early environment (pre- and post-natal), which appears to be of crucial importance for the development of a brain which will be in control of an individual's behaviour.

To what extent are neighbourhood factors important? Very few studies have investigated the interactions between individual, family, and neighbourhood characteristics using appropriate methodologies. Our second study highlighted a number of interesting "ecological" relationships between neighbourhood and individual characteristics. We believe this is an important area for further investigation. For example, previous studies have suggested that females are less likely to be influenced by neighbourhoods than males. Consider the implications of such a mechanism if it were also true that mothers play the crucial role in early child development. It may be possible to address some of these issues with data from the on-going NLSCY collection cycles. However, there will always be important limits to the use of the NLSCY for the study of neighbourhood effects because there are generally too few families per neighbourhood. This problem is likely to increase with time, since many families will be moving to different neighbourhoods.

Previous

Contents

Next