Multi-Level Effects on Behaviour Outcomes in Canadian Children

1.3 Results

Table 1.1 presents the fixed and random effects variances for the four childhood behaviour outcomes estimated by the three-level base model controlling for child's age and sex. Given that there often are important behavioural differences in children of different ages and that most same-aged children do not belong to the same family, it was important to include child's age in the base model. Otherwise, between-family variance would mostly reflect differences between children of different ages. Similarly, the sex of a child is often associated with behavioural differences in children. Therefore, this variable was included in the base model to avoid between-family variance mostly reflecting between-sex differences.

All random effects variances achieved statistical significance, although they became increasingly smaller from individual to family to census tract level. 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. Hyperactivity-inattention had the highest percentage of variation at the individual level (76.31%), physical aggression had the highest percentage of variation at the family level (29.92%), and prosocial behaviour had the highest percentage of variation at the census tract level (9.04%).

1.3.1 Childhood Behaviour and Individual Variables

Table 1.2 presents the three-level model of fixed and random effects for the four childhood behaviour outcomes. The fixed effects estimates indicate the change in childhood behaviour that occurs from a change in child, family, or census tract variables. The random effects estimates indicate the amount of variation in childhood behaviour occurring at the individual, family, and census tract levels.

Findings for individual variables revealed that older children exhibited less hyperactivity-inattention but greater anxiety-emotional problems and prosocial behaviours. Compared with girls, boys had more mother-reported physical aggression and hyperactivity-inattention but less prosocial behaviour. Greater than average exposure to positive child-mother interactions was associated with fewer anxiety-emotional problems. Children who experienced maternal hostility at levels that were higher than the average amount for all children in the family exhibited more physical aggression, hyperactivity-inattention, and anxiety-emotional problems and less prosocial behaviour. Finally, greater than average exposure to punitive parenting was associated with more hyperactivity-inattention and fewer anxiety-emotional problems and prosocial behaviours.

Table 1.1
Multilevel Base (Age and Sex Included) Modelof Random Effects Variances for Childhood Behaviour Outcomes
	Physical Aggression	Hyperactivity-Inattention	Anxiety-Emotional Problems	Prosocial Behaviour
Fixed Effects
Intercept	0.078 (0.075)^a	0.170(0.076)	-0.550(0.075)	-0.554(0.073)
Age	-0.030(0.009)	-0.047(0.009)	0.067(0.009)	0.087(0.008)
Sex	0.196(0.036)	0.324(0.036)	0.029(0.036)	-0.370(0.033)
Random Effects^b
Level 1 (Individual)
Variance	0.668(0.029)	0.770(0.032)	0.686(0.029)	0.627(0.027)
Percent of Total	66.40%	76.31	70.58%	65.18%
Variance
Level 2 (Family)
Variance	0.301(0.031)	0.186(0.030)	0.234(0.030)	0.248(0.028)
Percent of Total	29.92%	18.43%	24.07%	25.78%
Variance
Level 3 (Census Tract)
Variance	0.037(0.012)	0.053(0.013)	0.052(0.013)	0.087(0.018)
Percent of Total	3.68%	5.26%	5.35%	9.04%
Variance
Notes: a=Standard errors are in parentheses; b=All random effects are significant at p < .05.

1.3.2 Childhood Behaviour and Family Variables

Table 1.2 indicates that older mothers reported less childhood physical aggression and that children from single-parent families had greater hyperactive-inattentive and anxious-emotional problems than those from two-parent households. Children from higher SES families had less hyperactivity-inattention, and mothers with greater depressed mood reported higher levels for all four behaviour outcomes. Greater family dysfunction was associated with less prosocial behaviour while greater family social support was associated with more prosocial behaviour. Mothers with higher levels of positive interaction and consistent parenting within the family reported more prosocial behaviour in their children. Consistent parenting was also associated with fewer hyperactive-inattentive problems. In contrast, greater maternal hostility was associated with more physical aggression, hyperactivity-inattention, and anxiety-emotional problems. As well, the greater the amount of punitive parenting used by mothers within the family, the higher the level of children´s physically aggressive and hyperactive-inattentive behaviours and the lower the level of prosocial behaviour.

Table 1.2
Multilevel Model of Fixed and Random Effects for Childhood Behaviour Outcomes
	Physical Aggression		Hyperactivity-Inattention		Anxiety-Emotional Problems		Prosocial Behaviour
	Estimate	T-Ratio^a	Estimate	T-Ratio	Estimate	T-Ratio	Estimate	T-Ratio
Fixed Effects
Intercept	-0.091 (0.504)^b	-0.18	-0.172 (0.567)	-0.30	-0.913 (0.562)	-1.62	-2.256 (0.670)	-3.37
Child Variables
Age	-0.015 (0.009)	-1.69	-0.031 (0.009)	-3.45^*	0.067 (0.009)	7.22^*	0.110 (0.009)	12.23^*
Sex^c	0.100 (0.032)	3.17^*	0.229 (0.032)	7.10^*	-0.029 (0.033)	-0.88	-0.330 (0.032)	-10.33^*
Positive Interaction	0.013 (0.011)	1.16	0.020 (0.012)	1.71	-0.031 (0.012)	-2.60^*	0.004 (0.011)	0.38
Hostility	0.164 (0.001)	16.92^*	0.142 (0.010)	13.72^*	0.101 (0.010)	9.62^*	-0.055 (0.010)	-5.66^*
Consistency	-0.008 (0.012)	-0.62	-0.016 (0.013)	-1.24	0.018 (0.013)	1.34	0.018 (0.012)	1.46
Punitive Parenting	-0.007 (0.019)	-0.36	0.048 (0.020)	2.33^*	-0.055 (0.021)	-2.67^*	-0.052 (0.019)	-2.70^*
Family Variables
Mother Age	-0.009 (0.004)	-2.53^*	-0.006 (0.004)	-1.75	-0.001 (0.004)	-0.28	-0.002 (0.004)	-0.62
Mother Educational Level	-0.005 (0.017)	-0.29	-0.006 (0.017)	-0.36	0.005 (0.017)	0.28	0.007 (0.018)	0.43
Family Status^d
One biological parent	0.064 (0.050)	1.29	0.096 (0.049)	1.96^*	0.110 (0.051)	2.15^*	-0.017 (0.051)	-0.34
No biological parent	0.031 (0.142)	0.22	0.254 (0.141)	1.79	0.092 (0.147)	0.63	-0.175 (0.144)	-1.21
SES	-0.014 (0.034)	-0.41	-0.095 (0.034)	-2.85^*	-0.004 (0.035)	-0.12	-0.028 (0.035)	-0.78
Mother Depressed Mood	0.021 (0.004)	5.44^*	0.016 (0.004)	4.33^*	0.026 (0.004)	6.49^*	0.015 (0.004)	3.76^*
Family Dysfunction	0.001 (0.005)	0.31	0.001 (0.005)	0.12	0.004 (0.005)	0.94	-0.016 (0.005)	-3.39^*
Social Support	-0.008 (0.008)	-1.03	-0.010 (0.008)	-1.26	-0.003 (0.008)	-0.30	0.031 (0.008)	3.67^*
Mother Positive Interaction	0.002 (0.007)	0.27	0.002 (0.007)	0.24	0.008 (0.007)	1.07	0.061 (0.007)	8.28^*
Mother Hostility	0.107 (0.007)	15.91^*	0.093 (0.007)	14.02^*	0.096 (0.007)	13.84^*	0.006 (0.007)	0.81
Mother Consistency	-0.001 (0.006)	-0.12	-0.019 (0.006)	-3.12^*	0.006 (0.006)	0.94	0.029 (0.006)	4.61^*
Mother Punitive Parenting	0.023 (0.011)	2.05^*	0.023 (0.011)	2.12^*	-0.009 (0.011)	-0.82	-0.058 (0.011)	-5.03^*

Table 1.2 (continued)
	Physical Aggression		Hyperactivity-Inattention		Anxiety-Emotional Problems		Prosocial Behaviour
	Estimate	T-Ratio^a	Estimate	T-Ratio	Estimate	T-Ratio	Estimate	T-Ratio
Census Tract Variables
Area Size^e
100,000 - 499,999	0.152 (0.087)	1.73	-0.004 (0.107)	-0.04	-0.087 (0.103)	-0.85	0.123 (0.134)	0.92
30,000 - 99,999	-0.007 (0.088)	-0.07	-0.050 (0.106)	-0.47	-0.094 (0.102)	-0.91	0.056 (0.129)	0.44
15,000 - 29,999	0.024 (0.100)	0.24	-0.130 (0.117)	-1.11	-0.160 (0.114)	-1.40	0.123 (0.141)	0.87
Urban with < 15,000	-0.073 (0.095)	-0.77	-0.079 (0.111)	-0.71	-0.112 (0.109)	-1.03	0.069 (0.134)	0.51
Rural Area	-0.007 (0.078)	-0.09	-0.092 (0.094)	-0.98	-0.206 (0.091)	-2.25^*	0.070 (0.116)	0.61
Poverty	-0.445 (0.224)	-1.98^*	-0.109 (0.238)	-0.46	-0.206 (0.242)	-0.85	-0.324 (0.261)	-1.24
Problems	0.124 (0.061)	2.05^*	0.018 (0.073)	0.25	0.076 (0.071)	1.07	-0.050 (0.090)	-0.55
Collective Efficacy	-0.006 (0.024)	-0.26	0.008 (0.030)	0.26	0.031 (0.028)	1.09	-0.023 (0.037)	-0.63
Random Effects^f
Level 1 (Individual)
Variance	0.503 (0.022)		0.571 (0.024)		0.589 (0.025)		0.503 (0.022)
Percent of Total Variation	69.48%		77.37%		75.03%		63.43%
Level 2 (Family)
Variance	0.209 (0.023)		0.129 (0.022)		0.167 (0.024)		0.216 (0.023)
Percent of Total Variation	28.87%		17.48%		21.27%		27.24%
Level 3 (Census Tract)
Variance	0.012 (0.006)		0.038 (0.010)		0.029 (0.009)		0.074 (0.015)
Percent of Total Variation	1.65%		5.15%		3.70%		9.33%
Notes: a=T-Ratio determined by dividing estimate by its standard error. Values may vary slightly because they were rounded to three decimal places; b=Standard errors are in parentheses; c=Reference group is girls; d=Reference group is a family with both biological parents; e=Reference group is urban. f= All random effects are significant at p<.05.

1.3.3 Childhood Behaviour and Neighbourhood Variables

Compared with living in an urban area with a population of 500,000 or more, children living in rural areas were reported to have fewer anxious-emotional problems. In neighbourhoods with a greater proportion of poor families, mothers reported less physical aggression in their children. In addition, mothers who perceived greater problems in their neighbourhood reported more childhood physical aggression (see Table 1.2).

1.3.4 Multi-Level Effects on Childhood Behaviour

Table 1.2 showed all random effects to be statistically significant. The greatest amount of variation for all four behaviour outcomes occurred between individuals, a moderate amount of variation occurred between families, and a small amount occurred between census tracts. Hyperactivity-inattention had the highest percentage of between-individual variation (77.37%), physical aggression had the highest percentage of between-family variation (28.87%), and prosocial behaviour had the highest percentage of variation occurring at the census tract level (9.33%).

To examine multi-level variation in childhood behaviour after having incorporated individual, family, and census tract variables, random effects variances in Table 1.1 were compared with those in Table 1.2. These results are presented in Table 1.3.

The total variation in physical aggression explained by the three-level model was 28.03% ([(0.668 + 0.301 + 0.037) - (0.503 + 0.209 + 0.012)] / [0.668 + 0.301 + 0.037]). The amount of explained variation in physical aggression increased from the individual to family to census tract levels. Thus, our findings were best able to explain the variation in physical aggression between census tracts and least able to explain the variation in physical aggression between individuals. For hyperactivity-inattention, the three-level model explained 26.86% of the total variation. Our results were best able to explain why families varied in hyperactivity-inattention and least able to explain individual-level variation. For anxiety-emotional problems, the three-level model explained 19.24% of the total variation. Our results best explained the variation in anxiety-emotional problems between census tracts and least explained the variation between individuals. For prosocial behaviour, the three-level model explained 17.57% of the total variation. Our findings were best able to explain the variation between individuals and least able to explain the variation between families.

Table 1.3
Variation in Childhood Behaviour Outcomes Associated With Individual, Family, and Census Tract
Childhood Behaviour	Explained Variation (%)
Physical Aggression
Three-Level Model	28.03
Individual Level (66.40)^a	24.70
Family Level (29.92)	30.56
Census Tract Level (3.68)	67.57
Hyperactivity-Inattention
Three-Level Model	26.86
Individual Level (76.31)	25.84
Family Level (18.43)	30.65
Census Tract Level (5.26)	28.30
Anxiety-Emotional Problems
Three-Level Model	19.24
Individual Level (70.58)	14.14
Family Level (24.07)	28.63
Census Tract Level (5.35)	44.23
Prosocial Behaviour
Three-Level Model	17.57
Individual Level (65.18)	19.78
Family Level (25.78)	12.90
Census Tract Level (9.04)	14.94
a=The numbers in parentheses represent the percentage of total variation from the three-level base model controlling for child's age and sex (see Table 1.1).

1.3.5 Summary

Overall results indicated that children's physical aggression was associated with characteristics of the neighbourhoods in which they lived, specifically neighbourhood poverty and perceived neighbourhood problems. Children's anxiety-emotional problems were also associated with the size of the neighbourhoods in which they lived. Despite these findings, the influence of neighbourhoods was minimal compared with that of family and individual variables.

Characteristics of children's families, namely maternal age and depressed mood, family status, family SES, family dysfunction, family social support, and parenting practices, played a bigger role in influencing children's behaviour outcomes. The age and sex of children also impacted on their behaviours. Most of the variation in children's physical aggression, hyperactivity-inattention, anxiety-emotional problems, and prosocial behaviour was at the individual level followed by the family level. Only a very small amount of the variation in childhood behaviours occurred at the neighbourhood level. Our three-level model explained the most variation in children's physical aggression and the least variation in children's prosocial behaviour.