Section 3: Data and methods

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

• 3.1 Data
• 3.2 Dependent variable: timing of first post-secondary enrolment
• 3.3 Independent variables
• 3.4 Method

3.1 Data

The Youth in Transition Survey (YITS) is a Canadian longitudinal survey designed to examine the patterns of, and influences on, major transitions in young people's lives, particularly with respect to education, training and work. It collects very rich information on education and provides dates of educational enrolment and educational transitions, such as high school graduation, start of a post-secondary program as well as graduation from a post-secondary program. Data were collected from two age groups (cohorts) of youth in the first cycle of the survey in 2000. One began its participation at age 15 (Cohort A) and the other at ages 18 to 20 (Cohort B). In this report, all five cycles of Cohort B are used, providing information every two years from 2000 to 2008. At the time of the last cycle in 2008, respondents were aged 26 to 28. The sample used in all analyses is restricted to youth who had graduated from high school by Cycle 5; thus, about 550 high school leavers and high school continuers (about 5% of the sample) were removed from the analysis. The remaining sample size after removing missing cases is 8,508.

3.2 Dependent variable: timing of first post-secondary enrolment

Cox proportional hazard models are used to examine the impact of pertinent factors on the 'risk' of starting a first post-secondary program prior to age 28 for high school graduates.¹ In each cycle, YITS measures the month and year of high school graduation as well as the month and year of first post-secondary enrolment. As a result, it is possible to update the timing of the end of high school and the start of PSE in each cycle until Cycle 5 when the youth were age 26 to 28. The dependent measure is therefore defined as the probability of starting PSE for the first time, conditional on the fact that respondents have not yet experienced the event and that they were still under observation by Cycle 5. Respondents who had not yet started their first PSE program by December 2007 (the date of the last information on PSE status) were right censored at that time.² Approximately, 10% of this sample of high school graduates is right censored, that is, they had not started PSE by December 2007.

Kaplan-Meier life table estimates presented in Chart 1 show the proportion of high school graduates starting PSE in each month after the month of high school graduation. The cumulative proportion of high school graduates increases very quickly immediately after leaving high school. By month 3, almost 50% of youth had started their first PSE program; by 15 months, 73% had started and by 28 months, 81% had started. Not surprisingly, we observe significant increases in the months which correspond to normative times for starting PSE , that is, in the autumn 3 months after graduating high school, the autumn of the 1^st year out of school (15 months), and the autumn of the 2^nd year out of high school (28 months).

Chart 1 Cumulative proportion of high school graduates who started post-secondary education by age 28 (Weighted Kaplan Meier Life Table Estimates)

3.3 Independent variables

3.3.1 Demographic characteristics and geographic location

In order to take into account some basic and very important background indicators, six measures related to demographic characteristics and geography are included in all models. Demographic characteristics include three dummy variables measuring sex, being Canadian born, being of Aboriginal descent (off-reserve) and mother tongue (English, French, and other). With respect to geographic location, given the complexity of the secondary and post-secondary systems in Canada, it is imperative that a measure for province of high school is included. In the present case, it is sufficient to collapse provinces into four regions based on relative similarity of education system: Atlantic Provinces (Newfoundland, Prince Edward Island, Nova Scotia and New Brunswick), Quebec, Ontario, and Western Canada (Manitoba, Saskatchewan, Alberta and British Columbia).³ Lastly, we include a dummy variable indicating if the respondent lived in a large population centre in Cycle 1 when they were age 18 to 20.

3.3.2 Family factors

As noted, earlier, the family of origin has a tremendous influence on all facets of a young adult's life, which extends into the education system and decisions to finish high school and start PSE. Four factors are used to measure family influence. First, parental education (the highest level of education of either parent), which is split into less than high school, high school only, some PSE and PSE graduate, is included. Second, parental educational aspirations for the child is included, based on a question asked of youth of how important it is to their parents that they 'obtain more than a high school diploma.' The response categories are 'not important at all,' 'slightly important,' 'fairly important,' and 'very important.' This measure (ranging from 1 to 4) is used as a continuous variable in the models. Third, given the importance of parent-child communications, a measure is included that assesses the frequency with which youth and parent talk about future education and career options. This variable is categorized as 'never,' 'less than once a year,' 'a few times a year,' 'a few times a month,' 'a few times a week,' and 'daily.' It is included as a categorical variable in order to measure any nonlinear effects of parent-child communications. Lastly, the effect of siblings is captured via a continuous measure ranging from zero siblings to five or more (values range from 0 to 5).

3.3.3 Academic performance and commitment to the educational system

How engaged a student is in high school and his/her academic performance can both indicate ability as well as aspirations for higher education. To measure academic performance, average academic marks in the last year of high school are included. This measure is split into four broad categories: high, medium high, medium, and low, which correspond to overall grade-point averages of 80% to 100%, 70% to 79%, 60% to 69% and 59% or less, respectively. Issues related to commitment to the education system are measured via three factors. First, frequency of skipping or cutting classes is assessed via the question asking the students the number of times per month that they skipped classes without permission. Respondents were offered five response categories: 'never,' 'less than once a month,' 'once or twice a month,' 'about once a week' and 'more than once a week.' This measure is used as a continuous variable (ranging from 0 to 4) in the models. Second, number of hours spent each week in extracurricular school activities during the last year of high school is included, which is grouped into four categories: (1) none; (2) one to three hours; (3) four to seven hours; (4) eight hours or more. Third, future aspirations and commitment to the education system are taken into account by asking the youth the highest level of education they would like to obtain. This variable is categorized as follows: (1) high school or less; (2) some post-secondary; (3) college, trade or other diploma; (4) bachelor's degree or higher; and (5) undecided or do not know.

3.3.4 Potential barriers to high school completion and post-secondary education attendance

Last, the multivariate models need to take into account potential barriers to the eventual completion of high school as well as to starting PSE. There is a great deal of overlap between the factors from the previous section and this section in that a higher degree of commitment and better academic performance can be seen as non-barriers; however, in this paper, several specific measures are included that can be considered to be distinct barriers. First, working during high school has been linked to a greater risk of dropping out of high school (Sunter 1993) as well as slowing down the transition to PSE (Hango and de Broucker 2007; Tomkowicz and Bushnik 2003). This variable is separated into four categories: no hours, one to less than 10, 10 to less than 20, and more than 20 hours per week. Second, extracurricular activities not organized by the school are included as a potential barrier since it may pull youth away from the education system and leave less time to be spent in school or on school-related activities.⁴ This measure of extracurricular activities is coded as: no hours; one to three hours; four to seven hours; and eight hours or more per week.

The third barrier is measured as a dummy variable signifying whether the respondent (either male or female) had a child prior to age 18. The variable in YITS covers those children for whom the respondent is 'financially responsible and/or has sole or joint legal custody;' thus, the emphasis is placed on responsibility and not simply the occurrence of the birth event. The fourth barrier involves the effect of friends and peer networks via the measure asking respondents 'how many of your close friends were planning to continue with their education beyond high school.' A dummy variable equals 1 if the respondent stated that 'some,' 'very few' or 'none' of their friends were planning on continuing; the reference category is 'most' or 'all.' The fifth barrier is related to potentially negative effects arising from changing high schools; this continuous indicator ranges from 1 to 4 or more.

The final measures related to barriers explicitly ask respondents 'Is there anything standing in your way of going as far in school as you would like to go?' Four dummy variables are included indicating whether finances, average marks, wanting to work or caring for children would be potential barriers to their academic aspirations.⁵

3.4 Method

Cox Proportional Hazard models (Cox 1972) are used to predict the 'risk' or chance of first PSE enrolment prior to age 28. The dependent variable is defined as the instantaneous rate of entry into first PSE and is specified as a function of time constant covariates. Cox's semi-parametric regression model, which does not make any assumptions about the shape of the hazard of starting PSE over time, can be expressed as follows

where r_(t) is the rate at which the respondent experiences the hazard of starting PSE at time t; h_(t) is the baseline hazard rate at t, and X is a vector of time constant covariates. This instantaneous rate is defined as the ratio of the number of respondents entering PSE to the number of those who are still at risk of entering PSE. For all respondents, the observation period begins in the month following high school graduation and ends with the month they start PSE for the first time, or at the date of the last interview for those who had not yet started PSE (right censored cases).

All analyses are appropriately weighted and in addition standard errors are estimated using a replication approach in order to control for the survey clustered design. Thus, all results reflect standard errors derived from re-sampling each model 1,000 times using bootstrap weights (Statistics Canada 2003).⁶ In all analyses the statistical program Stata Version 10.1 is used (StataCorp, 2008).⁷

---

Notes

1. Within survival analysis, the term 'risk' essentially signifies the probability that the event will occur at a given time period.
2. In survival analysis terminology, right censoring is defined as when the failure event, in this case the date of first PSE enrolment, occurs sometime after the respondent is no longer under observation.
3. A handful of respondents who went to high school in the Yukon were included with Western Canada.
4. Alternatively, this measure may also be picking up discrepancies in school resources; perhaps youth who partake in nonschool activities do not have the option of registering in school activities. This type of information however is not available in YITS.
5. Some exploratory analysis revealed that these four barriers were the most detrimental to returning to school, they are also useful because they encompass a broad range of issues from problems with money and family to wanting to work instead of going to school.
6. For more complete information the reader is advised to consult the user guide for these data (Statistics Canada 2003).
7. Specifically, the svy procedure is used in Stata 10. This version of Stata enables researchers to utilize the 1000 bootstrap weights in YITS for the estimation of Cox Proportional Hazard models.