Methodology

View the most recent version.

Data source
Analytical techniques
Odds ratio
Description of variables

Data source

The data source for this study is the International Youth Survey (IYS) conducted by Statistics Canada in 2006. The main objectives of this survey are to measure the prevalence and incidence of various types of self-reported delinquent behaviours among youth in industrialized countries, to review the variability and correlates of self-reported youth delinquency and to provide detailed data in support of national policy development.

The target population for this survey was youth in grades 7, 8 and 9 from 175 schools in the Toronto census subdivision. Schools in the Toronto Public School Board and private schools participated in the survey.

Schools were first stratified by grade and geographic area, and a sample of schools was selected in each stratum systematically with probability proportional to size, where size was measured as the number of students in the chosen grades. This strategy was designed to provide an adequate representation of the various Toronto neighbourhoods including the city's demographic and socioeconomic characteristics. Selected schools were then visited in order to obtain the principal's consent. Once this was granted, interviewers randomly selected classes in the desired grades. Students in the selected classes were asked to obtain parental consent in order to participate in the survey. The survey was administered to students in April and May 2006. The sample size was 3,290 students across 177 classes; after weighting, this represented 60,900 students. Greater detail on the survey design and methodology are available in Savoie (2007).

Analytical techniques

A combination of bivariate and multivariate statistical techniques was used in this study. As we are interested in the relative impact of three different groups of factors in explaining youth delinquency (i.e., school variables, victimization, and peer and family relationships), each of these groups of factors are presented in separate logistic regression models that control only for the factor of interest and demographic variables. After we present these factors separately, we then present them all together in a full model to assess whether or not they retain their unique effects on property-related and violent delinquency. Cross tabular analyses were also conducted to compare various groups (e.g., by generational status and age) along various risk factors such as victimization and time spent with family. Respondents who were missing responses for any of the variables used in the logistic regression analysis were excluded¹ . This reduced the sample size by 16 % from approximately 3,300 youth to 2,778².

As some variables appeared to be highly correlated (e.g., the school variables), variance inflation factors and tolerance were calculated to check for multicollinearity in the models. The results were found to be within acceptable ranges.

Cross-sectional weights accounted for unequal probabilities of sample selection. To account for the complex sample design, the bootstrap technique was used to estimate coefficients of variation, confidence intervals and to test for statistical significance of differences (Rao, Wu and Yue, 1992; Rust and Rao, 1996).

Odds ratio

When an outcome variable for a regression model is dichotomous (e.g., committing a delinquent act versus not committing a delinquent act), researchers are interested in determining the probability of the occurrence of that event under a particular set of circumstances (e.g., having low income, being female, or living in a single-parent family). In this case, logistic regression is the most appropriate technique to use.

An odds ratio is a statistic generated by a logistic regression and, in this study, was used to assess whether, other things being equal, youth with specific characteristics are more or less likely to engage in delinquent behaviour than those in another (reference) group. For example, consider the risk of delinquency for youth from a low income background compared to those from an average income background, the reference category: An odds ratio near 1.0 implies that there is no difference in delinquency between the two groups; an odds ratio that is less than 1.0 implies that those in the group being considered (i.e., youth from a low income background) are less likely to be delinquent than those in the reference group; and an odds ratio that is greater than 1.0 implies that those in the group being considered are more likely to be delinquent than those in the reference category.

When an explanatory variable is continuous (e.g., age measured in years), the odds ratio indicates how much the ratio P/(1-P) is greater or less than a one unit increase of this variable (e.g., for an individual who is one year older than another individual). For example, an odds ratio of 2.0 indicates that the odds of delinquent behaviour are twice as high for a 12-year-old as they are for an 11-year-old.

Description of variables

Dependent variables

Property delinquency

The property delinquency variable was derived from the following seven questions: In the past 12 months have you, damaged something on purpose, such as a bus shelter, window, car or seat on a bus or train? Stolen something from a store? Broken into a building with the purpose of stealing something? Stolen a bicycle? Stolen a motorbike or car? Stolen something out of or from a car? Set fire on purpose to a mailbox, garbage can, building or car (not your own property)?

The property delinquency variable contained two categories, where 1 indicated that youth reported at least one property-related delinquent act within the last 12 months, and 0 indicated that youth did not report such an act within the last 12 months.

Violent delinquency

Violent delinquency was measured using five questions: In the past 12 months have you, snatched a purse, bag or something else from a person? Carried a weapon, such as a stick, chain or knife (not a pocket knife)? Threatened somebody with a weapon or threatened to beat them up to get money or other things from them? Participated in a group fight on a school playground, a football stadium, in a street, or in any other public place? Intentionally beaten up someone, or hurt them with a stick or knife, so badly that they had to see a doctor?

The violent delinquency variable contained two categories, where 1 indicated that youth reported at least one violent delinquent act within the last 12 months, and 0 indicated that youth did not report such an act within the last 12 months.

Independent variables

Generational status

Generational status is a concept derived from the youth's place of birth, his or her parents' place of birth, and, if applicable, the age at which the youth immigrated to Canada. This is in accordance with previous research which shows that educational attachment and achievement vary for youth depending on these factors (Aldous, 2006; Dinovitzer, Hagan and Parker, 2003; Kao and Tienda, 1995). For this analysis, youth were divided into four groups: native-born (youth and parents were born in Canada); second generation (youth was born in Canada; at least one parent is foreign-born); earlier immigrant, or youth who immigrated before the age of 5 (both parents are foreign-born, youth is foreign born and the youth immigrated before the age of 5); and recent immigrant, or youth who immigrated after the age of 5 (both parents are foreign-born, youth is foreign-born and immigrated after the age of 5).

Family composition

Youth were coded to one of four family-types according to their responses to a series of questions about with whom they lived. Options included an intact family (birth mother and father), single-parent family (either birth mother or birth father), step/blended family (either birth mother and stepfather or birth father and stepmother) or another family situation (grandparents, aunt, brother/sister, foster family or someone else).

Other demographic characteristics

The logistic regressions in these models also controlled for sex and age which ranged from 12 to 17. Since there were few youth aged 16 and 17, these two age categories were collapsed into one (16 years). Descriptive statistics for the variables used in this study are presented in the Appendix.

School-related variables

Negative attitudes towards school

Youth were asked a series of questions about their feelings towards or observations about school. Three of these capture positive feelings towards school. On a 4-point scale where 1 corresponded to the most positive feelings and 4 corresponded to the least positive feelings, youth were asked: 'Do you usually like school'; 'If I had to move I would miss my school'; and 'I like my school'. Responses were added together to form a scale which ranged from 0 (youth who feel very positively towards school) to 9 (youth who disagree with positive statements about school). The reliability coefficient (Cronbach's alpha) for the negative attitudes towards school scale was =.73³.

Perception of school safety

Youth were asked to rate, on a scale of 1 to 4 (where 1 corresponds to fully agree and 4 corresponds to fully disagree), how much they agreed with these statements: 'There is a lot of stealing in my school'; 'There is a lot of fighting in my school'; 'Many things are broken or vandalized in my school'; and 'There is a lot of drug use in my school'. These items were reversed and added together to form a scale which ranged from 0 (youth who perceived their school as safe) to 12 (youth who perceived their school as unsafe). This scale had a Cronbach's coefficient alpha of .76³.

University aspirations

Youth were asked what level of education they aspired to. A dichotomous (or two-category) variable was derived where 1 corresponded to graduating from high school and attending university and 0 corresponded to educational aspirations at the college or trade school level, high school, or non-completion of high school⁴.

Skipping school

Youth were asked if and how often they had skipped school in the last 12 months. This variable took the value of 0 (not having skipped), 1 (skipped 1 or 2 times), or 2 (skipped 3 or more times).

Victimization variables

Victimization

Youth reported whether or not they had experienced four different forms of victimization. Specifically they were asked whether, in the past 12 months, 'Someone wanted you to give him/her money or something else (watch, shoes, cell phone) and threatened you if you did not do it?'; 'Someone hit you violently or hurt you so much that you needed to see a doctor?'; 'Something was stolen from you (such as a book, money, cell phone, sport equipment, bicycle...)?'; or 'You were bullied at school (other students humiliated you or made fun of you, hit or kicked you, or excluded you from their group)?'.

These variables were included separately in the logistic regression models. For each type of victimization, 1 corresponded to having experienced the behaviour one or more times in the last 12 months, while 0 corresponded to not having experienced the behaviour.

Discrimination

In addition, youth were asked about their experiences of discrimination, or specifically, 'Have people ever treated you badly because of your religion or the language you speak, or the colour of your skin?' Youth who replied once, sometimes or often were coded a value of 1, while youth who responded never were coded a value of 0.

Peer and family variables

Positive mother-youth relationship

Youth were asked 'how do you usually get along with the woman you live with (your mother or stepmother)'. Response categories included 'I get along just fine', 'I get along rather well', 'I don't get along so well', 'I don't get along at all' and 'There is no mother or other woman in the house'. A dichotomous variable was created that gave a value of 1 to youth who replied either the first or second categories as a measure of a positive relationship. Youth who reported that there was no woman in the house were coded a value of 0, along with youth who did not report having a positive relationship with their mother.

Positive father-youth relationship

Youth were asked 'how do you usually get along with the man you live with (your father or stepfather)'. Response categories included 'I get along just fine', 'I get along rather well', 'I don't get along so well', 'I don't get along at all' and 'There is no father or other man in the house'. A dichotomous variable was created that gave a value of 1 to youth who replied either the first or second categories as a measure of a positive relationship. Youth who reported that there was no man in the house were coded a value of 0, along with youth who did not report having a positive relationship with their father.

Goes out at night

Youth indicated whether or not they went out at night without their parents and responses were coded a value of 1 if youth reported going out at night without their parents and 0 if youth reported that they did not go out at night.

Time spent with family

Youth were asked with whom they spent most of their free time. They were coded into three groups: youth who reported spending most of their free time alone; those who reported spending most of their free time with family; or those who reported spending most of their free time with friends. Youth who reported spending the majority of their free time with family were the reference category.

Illegal activities acceptable to peer group

Youth were asked if committing illegal acts was accepted by or okay for their peer group. Youth who reported yes were coded a value of 1, while youth who replied no or that they did not have a group of friends were coded a value of 0.

---

Notes

1. Cases in which the youth reported all of the independent variables and at least one delinquent act were included even if the youth did not respond to all of the delinquency questions. However, the majority of youth replied to most of the delinquency questions: 98% of youth responded to more than one-half of both the property-related and violent delinquency questions.


2. Missing data analysis showed that non-respondents were roughly similar to respondents in terms of age and sex, but were slightly less likely to come from an intact family and slightly more likely to be born outside of Canada.


3. Cronbach's alpha is a measure of internal consistency, based on the average correlation between items. It is assumed that items are positively correlated with each other because they are attempting to measure a common construct; therefore, a Cronbach's alpha close to 1 indicates a perfect consistency between items. A suggested level of reliability is typically .80 or greater; however, this may vary by the type of data.


4. The effect of all postsecondary aspirations where 1 corresponded to aspiring to any type of postsecondary education and 0 corresponded to no postsecondary aspirations was not significant.