Journal of Pediatric Psychology Advance Access published October 15, 2014

Province-Level Income Inequality and Health Outcomes in Canadian Adolescents Elizabeth C. Quon,1 PHD, and Jennifer J. McGrath,2 PHD, MPH 1

Community Mental Health, IWK Health Centre and 2Department of Psychology, Concordia University

All correspondence concerning this article should be addressed to Jennifer J. McGrath, PHD, MPH, Pediatric Public Health Psychology Laboratory, Department of Psychology, Concordia University, 7141 Sherbrooke St. W., Montreal, Quebec H4B 1R6, Canada. E-mail: [email protected] Received February 17, 2014; revisions received September 9, 2014; accepted September 17, 2014

Key words

adolescents; disparities; health behavior; mental health; public health.

Countries with greater disparity between the rich and the poor—or greater income inequality—have been shown to have worse population health (see Wilkinson & Pickett, 2006 for a summary). These findings have formed the basis of the income inequality hypothesis: A more unequal distribution of income in society, over and above societal average income, has an adverse effect on the health of the individuals in that society (Wilkinson & Pickett, 2007). To test the hypothesis that income inequality has a contextual effect on health, Subramanian and Kawachi (2004) have argued that multilevel consideration of individual income and societal/community income inequality, and their effects on individual health, is essential. In a meta-analysis of 28 multilevel studies, Kondo et al. (2009) found that income inequality had a ‘‘modest’’ adverse effect on adult self-rated health and mortality. Adolescence is a period of transition from childhood to adulthood. During this time, one’s socioeconomic status

(SES) also shifts from being determined by one’s parents or family toward being primarily self-determined. Existing evidence suggests that graded associations between SES and health (or socioeconomic gradients in health), which are well-established in adulthood and childhood (Braveman, Cubbin, Egerter, Williams, & Pamuk, 2010), may be present inconsistently during adolescence (Chen, Martin, & Matthews, 2006; Goodman, 1999; West, 1997). Similarly, associations between income inequality and health may be different in adolescence compared with adulthood. Two main mechanisms have been proposed to explain the link between income inequality and health, both of which may differentially affect adolescent versus adult health. The social cohesion pathway suggests that income inequality leads to low social capital and stressful social comparison, which affect health through psychological processes and associated physiological changes (Wilkinson, 1997a, b; Wilkinson & Pickett, 2009). Social comparison

Journal of Pediatric Psychology pp. 1–11, 2014 doi:10.1093/jpepsy/jsu089 Journal of Pediatric Psychology ß The Author 2014. Published by Oxford University Press on behalf of the Society of Pediatric Psychology. All rights reserved. For permissions, please e-mail: [email protected]

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Objective To examine the effects of provincial income inequality (disparity between rich and poor), independent of provincial income and family socioeconomic status, on multiple adolescent health outcomes. Methods Participants (aged 12–17 years; N ¼ 11,899) were from the Canadian National Longitudinal Survey of Children and Youth. Parental education, household income, province income inequality, and province mean income were measured. Health outcomes were measured across a number of domains, including self-rated health, mental health, health behaviors, substance use behaviors, and physical health. Results Income inequality was associated with injuries, general physical symptoms, and limiting conditions, but not associated with most adolescent health outcomes and behaviors. Income inequality had a moderating effect on family socioeconomic status for limiting conditions, hyperactivity/inattention, and conduct problems, but not for other outcomes. Conclusions Province-level income inequality was associated with some physical and mental health outcomes in adolescents, which has research and policy implications for this age-group.

2

Quon and McGrath

for country/state mean income, which limits their interpretability, as this is a potential confounder of the influence of country/state income inequality. It also remains unclear whether the geographical scale (e.g., country, state, city) of income inequality comparison matters for adolescent health. Existing evidence in adults, as reported in the meta-analysis by Kondo et al. (2009), suggests that stronger associations between income inequality and self-rated health exist for between-country versus within-country comparisons. Moreover, meta-analytic findings suggest that within-country associations may emerge in highly unequal societies only. Ross et al. (2005) found that within-country city-level income inequality was linked to mortality in highly unequal countries (United States, United Kingdom) but not in more equal countries (Canada, Sweden, Australia). To date, within-country adolescent comparisons are limited to U.S. states (Crosby et al., 2003; Singh et al., 2008, 2009) and Brazilian municipalities (Celeste et al., 2009). There is a need for more within-country comparisons outside of the United States, particularly in more equal countries, like Canada. In terms of income inequality, Canada is more equal than the United States, United Kingdom, Italy, Australia, and Japan, and less equal than Switzerland, Ireland, France, Sweden, Denmark, and other peer countries (Conference Board of Canada, 2013). Canada is made up of 10 provinces and 3 territories. Each province is responsible for its funding and delivery of health, social, and educational services. Canadian provinces also vary widely in terms of level of taxation. For instance, the highest provincial taxation rate in Alberta is 10%, while it is 21% in Nova Scotia. As such, Canadian provinces differ in terms of level of income inequality as well as in the provision of programs and services. Therefore, examination of the scale of income inequality at the province level was thought to be the most appropriate within-country comparison for the Canadian context. To our knowledge, no previous studies have tested for the effects of income inequality on adolescent health within Canada. Further understanding of how geographical scale and the inequality level of the country affects within-country effects may help to elucidate the mechanisms by which income inequality may influence health. The aim of the current study was to test the effects of provincial income inequality, independent of province mean income and family SES, across a number of health outcomes in Canadian adolescents. Therefore, using a within-country design, we tested a contextual main effect of province-level income inequality on individual health outcomes in adolescents, while controlling for province

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

and social cohesion may be particularly relevant to health during adolescence, owing to the importance of peer relations during this time (West, 1997). Psychosocial processes are also critical during this developmental period, and mental disorders are the most common health problems (Gore et al., 2011). The policy pathway suggests that the adverse influence of income inequality may operate through social and health policies, such as health care, welfare spending, child care, tax policy, and unemployment compensation (Subramanian & Kawachi, 2004). Policies and spending related to education and mental health care may be particularly important during adolescence. To date, only a handful of studies have examined associations between income inequality and adolescent health outcomes with multilevel study designs. Using data from the Health Behaviour in School-aged Children study, greater income inequality at the country level has been shown to be related to poorer adolescent self-related health (although results did not control for country mean income; Torsheim, Currie, Boyce, & Samdal, 2006), drinking alcohol in young adolescents (with no effect in older adolescents; Elgar, Roberts, Parry-Langdon, & Boyce, 2005), and a steeper within-country socioeconomic gradient in adolescent life satisfaction (but no main effect of income inequality on life satisfaction; Levin et al., 2011). In the United States, greater state-level income inequality was linked to higher adolescent obesity prevalence (Singh, Kogan, & van Dyck, 2008) and lower physical activity levels (Singh, Kogan, Siahpush, & van Dyck, 2009), although these findings did not control for state mean income. State-level income inequality was inversely correlated with birth-control usage but was not significant in multivariate analyses (Crosby, Holtgrave, DiClemente, Wingood, & Gayle, 2003). Finally, higher municipal-level income inequality was associated with worse oral health in Brazilian adolescents (Celeste, Nadanovsky, Ponce de Leon, & Fritzell, 2009). Results from these studies suggest that associations between income inequality and adolescent health may vary by health outcomes, such that income inequality may have a stronger effect on certain health outcomes. Moreover, associations between SES and adolescent health have been shown to vary by health outcome (Goodman, 1999). Examination of multiple health outcomes within a single sample would help to elucidate how income inequality may be differentially linked to health outcomes. To date, no studies have examined associations between income inequality and mental health outcomes, a critical domain of adolescent health. Moreover, some of the previous findings did not statistically control

Province-Level Income Inequality and Health Outcomes

mean income, household income, and parental education. We expected that greater provincial income inequality would be associated with worse adolescent health outcomes. We also tested the interaction between provincelevel income inequality and family SES on adolescent health. We expected stronger associations between family SES and adolescent health (or steeper socioeconomic gradients in health) in more unequal provinces.

Method Sample

Individual/Family SES Characteristics Family SES information was collected by phone interviews with the parent and their spouse. Household income (before taxes and transfers) from all sources of income for all family members during the previous 12 months was derived from open-ended interview questions. Parental education (years)

was derived from questions about the highest level of education attained for parent and spouse. Mean years of education between the two parents was calculated (except in cases where there was no spouse).

Province Income and Income Inequality Income measures for each Canadian province were drawn from the Canadian Socio-economic Information Management System database from the Income Statistics Division of Statistics Canada. Income inequality was measured using the Gini index, a measure of inequality that ranges from 0 (perfect equality) to 1 (perfect inequality), based on household income after taxes and transfers, adjusted for household size (Statistics Canada, 2013a). Mean income was measured as the average household income after taxes and transfers, adjusted for household size (Statistics Canada, 2013b). Data from 2000 and 2006 were extracted to match the years of NLSCY data collection. Thus, we included information from the 10 Canadian provinces from two different time points. Gini indices by province and year are presented in Table I.

Health Outcomes Health outcomes were broadly categorized into five categories: self-rated health, mental health, health behaviors, substance use behaviors, and physical health. All health outcomes were coded such that higher scores indicate worse health. Self-Rated Health Adolescents rated their health status as ‘‘excellent,’’ ‘‘very good,’’ ‘‘good,’’ ‘‘fair,’’ or ‘‘poor’’ via self-completed questionnaires (12–15 years) or telephone interviews (16–17 years).

Table I. Gini Index by Province and Year Gini index Province

2000

2006

Alberta

.312

.314

British Columbia

.312

.319

Manitoba

.290

.304

New Brunswick

.291

.293

Newfoundland

.302

.299

Nova Scotia

.295

.295

Ontario

.325

.320

Prince Edward Island Quebec

.285 .294

.265 .291

Saskatchewan

.295

.323

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Participants were from the National Longitudinal Survey of Children and Youth (NLSCY), a population-based longitudinal survey of Canadian children and adolescents conducted by Statistics Canada and Human Resources Development Canada. The NLSCY sample is representative of children aged 0–11 years who were living in any Canadian province in 1994–1995, when survey weights are applied. A full description of the NLSCY and its sampling design is available elsewhere (Human Resources Development Canada & Statistics Canada, 1995). Data were accessed with permission from the Social Sciences and Humanities Research Council of Canada. The current study used data from the original longitudinal cohort of the NLSCY, a sample that was 0–11 years old at initial recruitment in 1994–1995. Data collection occurred every 2 years, with eight collection cycles. Using a cross-sectional design, data were included from Cycle 4 (2000–2001) and Cycle 7 (2006–2007) to measure all participants from the original cohort during adolescence (between 12 and 17 years old). In Cycle 4, we included 5,580 adolescents who were 6–11 years old at initial recruitment in 1994. In Cycle 7, we included 6,319 adolescents who were 0–5 years old at initial recruitment in 1994. Data collection for the NLSCY was completed via computer-assisted telephone interviews and paper-and-pencil questionnaires. Data collection methods for SES and health variables are noted in the sections below. The ‘‘person most knowledgeable’’ was the youth’s biological mother (90%) or biological father (8%) and will hereafter be referred to as ‘‘parent.’’

3

4

Quon and McGrath

Health Behaviors Health behaviors were assessed via self-completed questionnaires for adolescents aged 12–15 years and via telephone interviews for adolescents aged 16–17 years. Television watching was derived from adolescent report of average number of hours spent watching TV or videos or playing video games per day. Response categories were recoded to create a continuous variable using the median value, where applicable. Physical activity was derived from adolescents’ responses to two questions about frequency of playing sports or doing physical activities during the week, with or without a coach or instructor. Responses were summed to create a total score. Breakfast eating was derived from adolescent report of frequency of eating breakfast from Monday to Friday. Substance Use Behaviors Substance use behaviors were assessed via self-completed paper questionnaires for all adolescents. Alcohol use was measured by adolescent report of their experience with alcohol, ranging from ‘‘I have never had a drink of alcohol’’ to ‘‘About 6–7 days a week.’’ Cigarette use was measured by adolescent report of their experience with smoking cigarettes from ‘‘I have never smoked’’ to ‘‘About 6–7 days a week.’’ Physical Health Limiting conditions, injuries, and chronic conditions were assessed via parent telephone interviews for adolescents aged 12–15 years and via adolescent telephone interviews for adolescents aged 16–17 years. Limiting condition was measured by report of a physical or mental condition or health problem that reduces the amount or kind of activity

the adolescent can do (‘‘yes’’ or ‘‘no’’). Responses were summed across three domains: home, school, and leisure activities. Injuries were measured by report of an injury requiring medical attention in the past 12 months (‘‘yes’’ or ‘‘no’’). Chronic conditions were measured by report of a health professional diagnosis of the following long-term conditions: asthma, bronchitis, food allergies, respiratory allergies, other allergies, heart condition, kidney condition, epilepsy, cerebral palsy, mental handicap, learning disability, attention deficit/hyperactivity disorder, psychological disorder, or other (‘‘yes’’ or ‘‘no’’). Responses were summed to create a total score. General symptoms and sleep difficulties were assessed via self-completed paper questionnaires for adolescents aged 12–15 years and via telephone interviews for adolescents aged 16–17 years. General symptoms were derived from adolescent report of frequency of occurrence of headaches, stomachaches, and backaches in the past 6 months from ‘‘seldom or never’’ to ‘‘most days.’’ Responses were summed to create a total score. Sleep difficulties were measured by adolescent report of how often they had difficulties in getting to sleep in the past 6 months from ‘‘seldom or never’’ to ‘‘most days.’’ Body mass index (kg/m2) was derived from self-reported height and weight via a paper questionnaire for all adolescents.

Missing Data Longitudinal response rate for the NLSCY was 68% in Cycle 4 and 57% in Cycle 7. We were unable to include data for adolescents who did not participate in these cycles. Multiple imputation (five data sets) was performed using SAS (version 9.2) to impute missing information for partial nonresponse data. Multiple imputation is preferable to listwise deletion of data, as it retains information to increase power and reduce bias (Enders, 2011). To impute health outcomes, we included all other health outcomes along with age, sex, cycle, and province in the imputation model. To impute household income and parental education, we included these variables along with parental employment status, family size, single parent status, number of bedrooms in the home, and type of dwelling. Results were largely identical when analyses were run on the original versus imputed data set; therefore, only results based on the imputed dataset are presented. The characteristics of the current study sample are provided in Table II.

Analytical Strategy Multilevel modeling techniques (Bryk & Raudenbush, 1987) were used to fit regression models to the nested data. A two-level model was specified in which participants (Level 1) were nested within province year (Level 2). The

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Mental Health Mental health was assessed via adolescent self-completed paper questionnaires (12–15 years); items were aggregated to form indices. For older adolescents (16–17 years), aggregated indices were derived from the previous cycle of data collection (i.e., when they were 14–15 years old). Selfesteem was measured by four items taken from the General Self Scale of the Marsh Self-Description Questionnaire (Cronbach’s a ¼ .73). Adolescents completed the ‘‘Behaviour Checklist,’’ which was factor analyzed by Statistics Canada to identify six factors: indirect aggression (5 items; a ¼ .73–.74), physical aggression (6 items; a ¼ .74–.82), emotional disorder (7 items; a ¼ .76–.79), hyperactivity/inattention (7 items; a ¼ .75–.79), prosocial behavior (10 items; a ¼ .77–.88), and property offenses (6 items; a ¼ .67–.77). These Cronbach’s a ranges are presented for Cycles 4 and 7 of the NLSCY.

Province-Level Income Inequality and Health Outcomes Table II. Sample Characteristics Characteristic

Age (in years)

Mean (SD)

N (%)

14.33 (1.71)

12 13

2,229 (18.7) 2,321 (19.5)

14

1,927 (16.5)

15

1,857 (15.6)

16

1,855 (15.6)

17

1,710 (14.4)

Sex Male

5,983 (50.3) 5,916 (49.7) 13.10 (2.14)

Household income ($Canadian)

77,024 (55,433)

Cycle 4 (2000/2001)

5,580 (46.9)

7 (2006/2007)

6,319 (53.1)

Province Alberta

1,253 (10.5)

British Columbia Manitoba

988 (8.3) 912 (7.7)

New Brunswick

699 (5.8)

Newfoundland and Labrador

646 (5.5)

Nova Scotia

843 (7.1)

Ontario

2,993 (25.1)

Quebec

2,267 (19.1)

Prince Edward Island

349 (3.0)

Saskatchewan

949 (8.0)

Level 1 model describes the effect of individual socioeconomic variables, and the Level 2 model describes the effect of province socioeconomic variables. Multilevel models were specified using HLM 6.2 software. To test the effect of income inequality, we entered province income inequality as a Level 2 predictor while controlling for Level 2 province mean income and Level 1 household income and parental education. To examine whether provincial income inequality moderated the effect of family SES on health, we tested cross-level interactions of Gini index and household income, and Gini index and parental education, while controlling for provincial mean income.

Results

Discussion

Gini coefficients by province and year are presented in Table I. The lowest Gini coefficient was .265 (Prince Edward Island in 2006), which is similar to the level of income inequality in Finland in 2000 (.269) or Belarus in 2011 (.265). The highest Gini coefficient was .325 (Ontario in 2000), which is similar to the level of income

Using a within-country design in Canadian adolescents, the aim of the current study was to examine the independent effects of income inequality, after accounting for mean income and family SES, on multiple domains of adolescent health. We tested for a main effect of provincial income inequality on adolescent health and for a moderating effect

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Female Parental education (years)

inequality in France in 2008 (.327) or Bangladesh in 2010 (.321). International Gini coefficients were obtained from the World Bank (2014). Using the categories of low (.244–.284), medium (.290–.354), and high (.355–.456) from Elgar et al. (2005), most Canadian provinces fall within a ‘‘medium’’ level of income inequality, as does the Canadian mean level of inequality. Descriptive sample characteristics for the 11,899 adolescents included in the study are presented in Table II. Cohort effects were not observed; thus, sample characteristics are presented for the entire sample. Overall, the sample was evenly divided across age and sex categories. Mean parental education was about 13 years, which corresponds to completion of secondary education or 1 year of postsecondary education, depending on the Canadian province. Mean pretax household income before taxes was about $77,000, which is similar to national averages. Descriptive statistics for health outcomes are presented in Table III. We hypothesized that greater provincial income inequality would be associated with worse health outcomes. Results (presented in Table IV) indicated that greater income inequality (higher Gini index) was associated with more injuries requiring medical attention, more general physical symptoms, and more life domains affected by limiting conditions, after controlling for provincial mean income, household income, and parental education. We also hypothesized that associations between family SES and health would be stronger in provinces with greater income inequality. Cross-level interactions of income inequality with household income and parental education are presented in Table V. Results indicated that greater income inequality was associated with stronger associations for household income with limiting conditions, and for parental education with limiting conditions, hyperactivity/inattention, and property offenses. In contrast, greater income inequality was associated with a weaker association for household income with cigarette use. Selected significant interaction effects are illustrated in Figure 1; income inequality tertiles were created for interpretation.

5

6

Quon and McGrath Table III. Descriptive Statistics Health outcome

Self-rated health (1–5)

Mean (SD)

N (%)

1.93 (0.80)

Excellent (1) Very good (2)

3,717 (31.2) 5,798 (48.7)

Good (3)

1,970 (16.6)

Fair (4)

361 (3.0)

Poor (5)

53 (0.4)

Injury (past 12 months; 0–1) No (0)

9,675 (81.3)

Yes (1) Chronic conditions (number; 0–14) Sleep difficulties (1–5)

2,224 (18.7) 0.63 (0.97) 2.19 (1.22) 4,426 (37.2)

Once per month (2)

3,474 (29.2)

Once per week (3)

2,152 (18.1)

Two or more time per week (4)

1,010 (8.5)

Most days (5)

837 (7.0)

General symptoms score (3–15)

5.79 (2.32)

Body mass index Limiting condition

21.51 (3.62) 0.18 (0.62)

(number of domains: 0–3) 0

10,751 (90.4)

1

507 (4.3)

2

271 (2.3)

3

370 (3.1)

Physical activity score (2–8)

4.76 (1.64)

Television watching (hr/day) Breakfast eating (1–4)

2.49 (1.66) 1.88 (1.04)

Every day (1)

5,950 (50.0)

3–4 days per week (2)

2,839 (23.9)

1–2 days per week (3) Never (4)

1,739 (14.6) 1,371 (11.5)

Cigarette use score (1–8)

2.09 (1.90)

Alcohol use score (1–9)

3.28 (2.06)

Self-esteem score (0–16) Indirect aggression score (0–10)

4.17 (2.50) 1.35 (1.55)

Emotional problems score (0–16)

3.45 (2.70)

Physical aggression score (0–12)

1.10 (1.64)

Hyperactivity/inattention score (0–16)

4.00 (2.68)

Prosocial behavior score (0–20)

8.76 (3.76)

Property offenses score (0–12)

1.02 (1.36)

Note. For all health behaviors and conditions, a lower score indicates better health.

of provincial income inequality on associations between family SES and adolescent health. The first hypothesis was that greater provincial income inequality would be associated with worse health outcomes in adolescents. A significant association between income inequality and poor health was observed for certain physical health outcomes, which provides limited support for this hypothesis. Greater income inequality was related to

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Never (1)

more injuries requiring medical attention, more frequent physical symptoms like headaches, stomachaches, and backaches, and more limitations at home, school, and leisure due to a physical or emotional condition, after controlling for family SES and mean province income. However, provincial income inequality was not associated with self-rated health, health behaviors like diet, physical activity, or substance use, or on mental health problems such as hyperactivity/inattention, emotional problems, or aggression. Previous research on the effects of income inequality on health in adolescents has shown mixed results across studies and outcomes. Greater country income inequality was associated with poorer self-rated health in adolescents (Torsheim et al., 2006), and greater state income inequality was associated with higher obesity prevalence and lower physical activity levels (Singh et al., 2008, 2009). In contrast, the current study did not observe significant associations between province income inequality and self-rated health, body mass index, or physical activity. Of note, the previous studies did not adequately control for average income levels, which may be an important confound of the effects of income inequality, while we included mean province income as a covariate, along with household income and parental education. The lack of significant findings in the current study may be due in part to the limited range of income inequality among Canadian provinces, while previous studies have examined across countries and states with more variation in level of income inequality. Other factors that may contribute to the differences in findings are the scale of the study (between country vs. within country), overall level of income inequality in the country (high inequality in the United States vs. medium inequality in Canada), and measurement differences (self-report vs. measured, change in methods of assessment). Based on current and previous findings, independent effects of income inequality on adolescent health are not consistently observed. However, when significant associations are observed, they indicate that greater income inequality is associated with poorer health in adolescents. The second hypothesis was that greater provincial income inequality would be associated with steeper socioeconomic gradients in health. A significant cross-level interaction between income inequality and family SES was observed for limiting conditions, hyperactivity/inattention, and property offenses in the expected direction, which provides partial support for this hypothesis. Levin et al. (2011) also observed a significant interaction between Gini index and individual SES (as measured by the Family Affluence Scale), which indicated that as country

Province-Level Income Inequality and Health Outcomes

7

Table IV. Main Effects of Province and Family SES on Health Outcomes Province (Level 2) Gini index

Health outcome

b

Individual/family (Level1) Mean income

b

95% CI

95% CI

Household income

b

95% CI

b

95% CI

0.09, 0.03

.062

0.08, 0.04

Self-rated health

.007

0.04, 0.06

.018

0.01, 0.05

.061

Injuries

.049

0.02, 0.08

.024

0.06, 0.02

.002

Chronic conditions

.013

0.05, 0.08

.008

0.05, 0.07

Sleep difficulties

.042

0.09, 0.01

General symptoms Body mass index

.048 .035

0.002, 0.09 0.08, 0.01

Limiting conditions Low physical activity

0.02, 0.02

.027

0.01, 0.05

.036

0.06, 0.02

.001

0.02, 0.02

.095

0.05, 0.14

.026

0.05, 0.01

.037

.017 .005

0.06, 0.02 0.03, 0.04

.015 .028

0.03, 0.004 0.05, 0.01

.012 .071

0.02, 0.06 0.03, 0.01 0.09, 0.05

.047

0.01, 0.09

.024

0.01, 0.06

.033

0.05, 0.01

.042

0.06, 0.02

.015

0.07, 0.03

.002

0.05, 0.05

.074

0.09, 0.06

.100

0.12, 0.08

.042

0.01, 0.09

.096

.012

0.07, 0.09

.001

0.15, 0.04

.034

0.05, 0.01

.112

0.13, 0.09

0.08, 0.08

.053

0.07, 0.03

.098

0.12, 0.08

.038

0.06, 0.02

.085

0.10, 0.06

0.01, 0.04

.028

0.04, 0.01

0.08, 0.04 0.02, 0.01

.031 .029

0.05, 0.01 0.05, 0.01

Cigarette use

.048

0.03, 0.13

.102

0.18, 0.02

Alcohol use

.038

0.03, 0.10

.025

0.09, 0.04

.022

.020 .022

0.06, 0.02 0.01, 0.05

.040 .012

0.01, 0.07 0.05, 0.03

.058 .002

Emotional problems

.028

0.02, 0.07

.013

0.03, 0.05

.041

0.06, 0.02

.001

0.02, 0.02

Physical aggression

.001

0.06, 0.06

.046

0.01, 0.11

.032

0.05, 0.01

.081

0.10, 0.06 0.07, 0.03

Hyperactive/inattention

.005

0.05, 0.05

.043

0.01, 0.09

.028

0.05, 0.01

.051

Prosocial behavior

.021

0.06, 0.02

.018

0.02, 0.06

.037

0.06, 0.02

.038

0.06, 0.02

Property offenses

.013

0.06, 0.04

.040

.0.01, 0.09

.031

0.05, 0.01

.031

0.05, 0.01

Expected direction

Positive

Negative

Negative

Negative

Note. Standardized b coefficients and 95% confidence intervals are presented. Bold values are statistically significant at p < .05. All models include age, sex, parental education, household income, Gini index, and mean income.

Table V. Cross-Level Interaction of Gini Index With Household Income and Parental Education Gini  Household income

Health outcome b

Self-rated health

95% CI

Gini  Parental education b

95% CI

.010

0.02, 0.04

.011

0.03, 0.01

.005

0.03, 0.02

.001

0.02, 0.02

.006

0.01, 0.03

.009

0.01, 0.03

Sleep difficulties

.005

0.03, 0.02

.008

0.03, 0.01

General symptoms

.006

0.03, 0.02

.006

0.03, 0.01

.011

0.01, 0.03

.011

Injuries Chronic conditions

Body mass index Limiting conditions

.022

0.04, 0.004

.037

0.01, 0.03 0.06, 0.02

Low physical activity Television hours

.020 .005

0.01, 0.05 0.02, 0.03

.000 .003

0.02, 0.02 0.02, 0.02 0.02, 0.02

Low breakfast eating

.012

0.01, 0.03

.001

Cigarette use

.029

0.01, 0.05

.019

0.00, 0.04

Alcohol use

.002

0.02, 0.02

.002

0.02, 0.02

.002

0.03, 0.03

.006

0.03, 0.01

Indirect aggression

Low self-esteem

.008

0.03, 0.01

.013

0.03, 0.01

Emotional problems

.004

0.02, 0.02

.017

0.04, 0.002

Physical aggression Hyperactivity/inattention

.003 .003

0.03, 0.03 0.02, 0.02

.023 .024

0.04. 0.001 .0.04, 0.004

Prosocial behavior

.013

0.04, 0.02

.009

0.03, 0.01

Property offenses

.003

0.02, 0.02

.023

0.04, 0.004

Expected direction

Negative

Negative

Note. For interaction terms, standardized b coefficients and 95% confidence intervals are presented. Bold values are statistically significant at p < .05. Gini  Income models control for age, sex, mean province income, Gini index, and household income. Gini  Education models control for age, sex, mean income, Gini index, and parental education.

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Television hours Low breakfast eating

Low self-esteem Indirect aggression

Parent education

8

Quon and McGrath

income inequality increases, the socioeconomic gradient in life satisfaction increased. In the present study, income inequality displayed a main effect on limiting conditions, as well as a moderating effect on the family socioeconomic gradients for this outcome. Moreover, steeper gradients were observed in more unequal provinces for two ‘‘externalizing’’ mental health issues: hyperactivity and property offenses. In other words, low family SES was most strongly linked to externalizing problems in more unequal provinces. Previous research has linked income inequality to juvenile homicide and bullying (Wilkinson & Pickett, 2007). For cigarette use, we observed that individual socioeconomic gradients decreased as income inequality increased. This finding may be linked to regional variations in youth cigarette use across Canada (Reid & Hammond, 2009), which may confound the associations. For instance, youth smoking rates are much higher in Quebec compared with other provinces; thus, a socioeconomic gradient may be less apparent in this province. The current study did not evaluate pathways directly; however, these results may have implications for the two main proposed pathways between income inequality and health. The social cohesion pathway considers the importance of psychological processes of social trust and social comparisons. Thus, this pathway may be more directly linked to mental health outcomes. In the current study, we found that steeper family SES gradients existed for inattention/hyperactivity and property offenses in more unequal provinces. Crime and violence may be uniquely linked to income inequality through lower social trust, increased importance on status, and increased sensitivity of shame and humiliation (Wilkinson & Pickett, 2009). Associations between income inequality and internalizing mental health issues, such as self-esteem and emotional problems, were not observed, which suggests that stressful

social comparisons may not explain associations between provincial income inequality and adolescent health in Canada. The policy pathway considers the importance of social, education, and health policies. For example, differential provision of special education services across Canadian provinces may influence the degree of limitation a child with a disability faces. Funding of and access to primary health care, which vary across Canadian provinces, may affect general physical symptoms like stomachaches or headaches. Similarly, differences in bicycle helmet legislation across Canadian provinces have been linked to injuries (MacPherson et al., 2002). Further research is required to investigate these associations. Although not the primary aim of the current study, it is important to note that higher family SES was associated with better adolescent health for most of the health outcomes and health behaviors that were measured. Even after accounting for provincial income and income inequality, strong and consistent socioeconomic gradients were observed for Canadian adolescents. This suggests that more proximal socioeconomic indicators (household income and parent education) had stronger and more consistent associations with adolescent health than distal socioeconomic indicators (provincial income, provincial income inequality), which has both theoretical and policy-related implications. This article adds to the literature that has used a multilevel design to examine associations between income inequality and health during adolescence. One of the strengths of the current study was our ability to examine the independent effects of income inequality, while statistically controlling for mean income, and parent-reported household income and parental education. This study tested within-country associations between income inequality and adolescent health in Canada, a country with

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Figure 1. Effects of parental education on (a) hyperactivity/inattention and (b) limiting conditions are presented by income inequality tertiles (low, medium, and high).

Province-Level Income Inequality and Health Outcomes

of these associations. Moreover, longitudinal data are required to test the mediational pathways between income inequality and adolescent health. For instance, social comparison or social cohesion may be measured as a mediating pathway between income inequality and health outcomes, particularly limiting conditions, injuries, general symptoms, and mental health. Additionally, specific policies or programs that are offered differentially across provinces or countries, such as day-care policies, special education services, access to primary health care, or access to mental health services, may be examined as a mediating pathway between income inequality and adolescent health outcomes that have shown significant associations in the current study or previous studies. In conclusion, this study provided limited evidence for independent associations between provincial income inequality and health in Canadian adolescents. We observed a main effect of income inequality for some adolescent physical health outcomes, and a moderating effect on associations between parental education and adolescent externalizing mental health. Using a multilevel, withincountry design in Canada, we found that provincial income inequality was not related to most adolescent health outcomes, including self-rated health, health and substance use behaviors, and internalizing mental health problems. Further understanding of the effects of income inequality on health in childhood and adolescence, as well as adulthood, will help us to promote interventions to reduce inequality or its impact on health and well-being.

Acknowledgments This analysis was based on the Statistics Canada master files NLSCY Cycles 4 and 7, which contain anonymized data collected from 1994 to 2007. The responsibility for the use and interpretation of these data is solely that of the authors. The opinions expressed in this article are those of the authors and do not represent the view of Statistics Canada. The authors extend their sincere thanks to members of the International Network for Research in Inequalities in Child Health (INRICH).

Funding This work was supported by funding from the Canadian Institutes of Health Research (J. McGrath OCO-79897, MOP-89886, MSH-95353, MOP-123533; E. Quon CGM89256) and Quebec Inter-University Centre for Social Statistics. Conflicts of interest: None declared.

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

medium-level income inequality, which was important given the evidence that within-country effects of income inequality may exist only in highly unequal societies. Finally, we were able to examine the effects of income inequality in a broad range of health outcomes and health behaviors, which allowed for a more thorough investigation of these associations in adolescence. There are several methodological limitations of the current study. First, the amount of variability in income inequality between Canadian provinces is limited compared with previous cross-country analyses, which may contribute to the lack of significant associations in the current study. A previous meta-analysis of adult findings noted weaker associations in within-country comparisons compared with between-country comparisons (Kondo et al., 2009). Second, although we examined associations in both 2000 and 2006 to increase our statistical power, we used a cross-sectional design and are not able to determine the direction of the observed associations. Third, the NLSCY relies on adolescent and parent reports of health behaviors and health outcomes, which are subject to differences in response styles and are a potential source of bias. Moreover, the method of assessing certain variables changed at age 16 years (parent report to self-report, paper questionnaire to telephone interview), which introduces additional measurement error and potential for bias. Fourth, the current study aimed to provide an overview of the associations between income inequality and a range of adolescent health outcomes. Thus, many of the health outcomes are not examined in great depth or detail, and statistical bias is possible given that multiple health outcomes were tested. Replication of these results, focusing on specific health outcomes, is recommended. Fifth, although we used after-tax income to derive the Gini coefficient, in line with previous studies (Torsheim et al., 2006), the NLSCY data set included before-tax household income only. Finally, although the original sample was representative of the Canadian population at initial recruitment, significant attrition occurred over time in the NLSCY. To maximize available data, we used multiple imputation to examine associations in all remaining participants. Future research in this area may address some of the limitations of the current study, as well as further the understanding of the association between income inequality and health. Studies that link to health records may help to reduce bias associated with self-reported health variables. In addition, further examination of age and developmental influences on associations between income inequality and health during adolescence will further conceptual understanding. Further, longitudinal study designs that document changes in income equality and subsequent changes in health will help to determine the directionality

9

10

Quon and McGrath

References

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

Braveman, P. A., Cubbin, C., Egerter, S., Williams, D. R., & Pamuk, E. (2010). Socioeconomic disparities in health in the United States: What the patterns tell us. American Journal of Public Health, 100, S186–S196. Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to assessing change. Psychological Bulletin, 101, 147–158. Celeste, R. K., Nadanovsky, P., Ponce de Leon, A., & Fritzell, J. (2009). The individual and contextual pathways between oral health and income inequality in Brazilian adolescents and adults. Social Science and Medicine, 69, 1468–1475. Chen, E., Martin, A. D., & Matthews, K. A. (2006). Socioeconomic status and health: Do gradients differ within childhood and adolescence? Social Science and Medicine, 62, 2161–2170. Conference Board of Canada. (2013). How Canada performs: Income inequality, Retrieved from http://www. conferenceboard.ca/hcp/details/society/incomeinequality.aspx Crosby, R. A., Holtgrave, D. R., DiClemente, R. J., Wingood, G. M., & Gayle, J. A. (2003). Social capital as a predictor of adolescents’ sexual risk behavior: A state-level exploratory study. AIDS and Behavior, 7, 245–252. Elgar, F. J., Roberts, C., Parry-Langdon, N., & Boyce, W. (2005). Income inequality and alcohol use: A multilevel analysis of drinking and drunkenness in adolescents in 34 countries. European Journal of Public Health, 15, 245–250. Enders, C. K. (2011). Applied missing data analysis. New York: Guilford Press. Goodman, E. (1999). The role of socioeconomic status gradients in explaining differences in US adolescents’ health. American Journal of Public Health, 89, 1522–1528. Gore, F. M., Bloem, P. J. N., Patton, G. C., Ferguson, J., Joseph, V., Coffey, C., . . . Mathers, C. D. (2011). Global burden of disease in young people aged 10-24 years: A systematic analysis. Lancet, 377, 2093–2102. Human Resources Development Canada, & Statistics Canada. (1995). National Longitudinal Survey of Children: Overview of Survey Instruments for 1994-95 Data Collection, Catalogue no. 89F0078XIE. Ottawa, ON: Statistics Canada. Kondo, N., Sembajwe, G., Kawachi, I., van Dam, R. M., Subramanian, S. V., & Yamagata, Z. (2009). Income

inequality, mortality, and self rated health: Meta-analysis of multilevel studies. British Medical Journal, 339, b4471. Levin, K. A., Torsheim, T., Vollebergh, W., Richter, M., Davies, C. A., Schnohr, C. W., . . . Currie, C. (2011). National income and income inequality, family affluence and life satisfaction among 13 year old boys and girls: A multilevel study in 35 countries. Social Indicators Research, 104, 179–194. Macpherson, A., To, T. M., Macarthur, C., Chipman, M. L., Wright, J. G., & Parkin, P. C. (2002). Impact of mandatory helmet legislation on bicycle-related head injuries in children: A population-based study. Pediatrics, 110, e60. Reid, J. L., & Hammond, D. (2009). Tobacco use in Canada: Patterns and trends (Vol. 2). Waterloo, ON: Propel Centre for Population Health Impact. Ross, N. A., Dorling, D., Dunn, J. R., Henriksson, G., Glover, J., Lynch, J., & Weitoft, G. R. (2005). Metropolitan income inequality and working-age mortality: A cross-sectional analysis using comparable data from five countries. Journal of Urban Health, 82, 101–110. Singh, G. K., Kogan, M. D., Siahpush, M., & van Dyck, P. C. (2009). Prevalence and correlates of state and regional disparities in vigorous physical activity levels among US children and adolescents. Journal of Physical Activity and Health, 6, 73–87. Singh, G. K., Kogan, M. D., & van Dyck, P. C. (2008). A multilevel analysis of state and regional disparities in childhood and adolescent obesity in the United States. Journal of Community Health, 33, 90–102. Statistics Canada. (2013a). Gini coefficients of market, total and after-tax income of individuals, where each individual is represented by their adjusted household income, by economic family type, annual (number), CANSIM (database). Table 202–0709. Ottawa, ON: Statistics Canada. Statistics Canada. (2013b). Market, total and after-tax income of individuals, where each individual is represented by their adjusted household income, by economic family type and adjusted after-tax income quintiles, 2011 constant dollars, annual, CANSIM (database). Table 202–0707. Ottawa, ON: Statistics Canada. Subramanian, S. V., & Kawachi, I. (2004). Income inequality and health: What have we learned so far? Epidemiologic Reviews, 26, 78–91. Torsheim, T., Currie, C., Boyce, W., & Samdal, O. (2006). Country material distribution and

Province-Level Income Inequality and Health Outcomes

Wilkinson, R. G., & Pickett, K. E. (2006). Income inequality and population health: A review and explanation of the evidence. Social Science and Medicine, 62, 1768–1784. Wilkinson, R. G., & Pickett, K. E. (2007). The problems of relative deprivation: Why some societies do better than others. Social Science and Medicine, 65, 1965–1978. Wilkinson, R., & Pickett, K. (2009). The spirit level: Why greater equality make societies stronger. New York: Bloomsbury Press. World Bank. (2014). Gini index. Retrieved from http:// data.worldbank.org/indicator/SI.POV.GINI

Downloaded from http://jpepsy.oxfordjournals.org/ at University of Birmingham on October 31, 2014

adolescents’ perceived health: Multilevel study of adolescents in 27 countries. Journal of Epidemiology and Community Health, 60, 156–161. West, P. (1997). Health inequalities in the early years: Is there equalisation in youth? Social Science and Medicine, 44, 833–858. Wilkinson, R. G. (1997a). Comment: Income, inequality, and social cohesion. American Journal of Public Health, 87, 1504–1506. Wilkinson, R. G. (1997b). Health inequalities: Relative or absolute material standards? British Medical Journal, 314, 591–595.

11