Secondary menu

National Youth Survey Report

SECTION 2: OVERVIEW OF METHODS

2.1 Questionnaire

Before finalizing the questionnaire used in the National Youth Survey, research staff from the consultant worked with Elections Canada to identify key areas of interest. Information from a study entitled Youth Electoral Engagement in CanadaFootnote 2 and a literature reviewFootnote 3 was used to inform the questionnaire design. The final questionnaire included questions about:

2.2 Survey Administration

The survey was timed to coincide with the completion of the 41st general election, held on May 2, 2011.

The questionnaire was designed so that it could be administered using different survey methods (telephone, online and in person) and took on average 14 minutes to complete. Field testing was conducted on May 3, 2011, with 51 respondents. Following the field test, only minor modifications were required, and data from the field test interviews were included in the final data sets. Full survey administration took place between May 5, 2011, and June 13, 2011. The National Youth Survey was administered using multiple methods, including telephone, online and in-person intercept surveys. A draw for an iPad was offered as an incentive to those who completed the survey.

2.3 The Sample

Canadian citizens aged 18 to 34 years as of May 2, 2011, were eligible to participate in the survey. The sample was developed to encompass all regions of Canada. The sampling approach included a random sample stratified by key characteristics of interest (Group A) and a purposive sample designed to represent specific subgroups (Group B).

2.3.1 National Random Sample (Group A)

The sample frame for Group A consisted of 57,634 telephone numbers randomly selected from the ASDEFootnote 4 lists of telephone numbers. The sample was stratified by the following regions:

In total, 1,372 valid completions were obtained, yielding an estimated response rate of 34%,Footnote 5 with an overall margin of error of ±2.6% at the 95% confidence interval.

The distribution of key demographic characteristics was compared with the national random sample (Group A) and the Canadian census of 2006. The slight differences in the distributions were corrected mathematically by post-stratification weighting by age and gender within the regions to reduce any potential bias caused by over- or under-sampling.

Furthermore, response to the telephone survey was found to be biased toward youth with higher educational attainment. Sufficient census data were not available to adjust for these differences through the weighting strategy. As a result of these concerns, and in recognition of the interaction between education and many variables in the survey, the regression analyses controlled for education to ensure that the measured relationships between voting intention and the other variables were not merely an expression of respondents' educational attainment.

Full details of the sample, response rates and weighting are provided in Appendix A.

2.3.2 Purposive Sample (Group B)

Purposive sampling was used to provide data to construct profiles of the voting behaviours of different groups. The consultant used a variety of sampling approaches to better target these subgroups, including:

The numbers of completed surveys by each of the subgroups are shown in Table 2‑1.

Table 2‑1: Number of Valid Completed Surveys by Mode of Completion for Subgroups
Subgroup Random (Group A) Non-random (Group B) Total
Aboriginal 87 196 283
Youth with disabilities 52 101 153
Ethnocultural 196 262 458
Rural 372 180 552
Unemployed 69 172 241
Total 776 911 1,687

2.4 Analysis

At the conclusion of the survey, data were entered and cleaned, open-ended responses were thematically coded and weights were applied to the required survey responses.

The consultant then used the Statistical Package for the Social Sciences (SPSS, statistical analysis software) to produce the final survey results. The results for each question were cross-tabulated by voting behaviour in the general election held on May 2, 2011.

Binary logistic regression modelling was used to examine the association between survey variables and electoral participation in the 2011 general election. A multinomial regression was also used to profile youth based on past voting behaviour (habitual voters, frequent voters, occasional voters or habitual non-voters). Binary and multinomial regression models tested the relationships between voting behaviour and the following factors:

2.5 Considerations

The key strength of the study was the use of both random and purposive sampling. The random sample closely represented the national profile of youth in the 2006 census and therefore provided results that were reasonably nationally representative. The purposive sampling resulted in the inclusion of youth from subgroups who would not have been contacted by telephone sampling alone.

Notwithstanding the strengths of the data obtained from the National Youth Survey, some limitations need to be considered when reading the results:

A full discussion of the strengths and limitations of the study are reported in Appendix A.


Footnote 2 André Blais and Peter Loewen, Youth Electoral Engagement in Canada, Elections Canada Working Paper Series (January 2011).

Footnote 3 Elections Canada Youth Research Action Plan: Literature Review. Draft forthcoming November 2011.

Footnote 4 ASDE lists of numbers are updated regularly from telephone directories across Canada.

Footnote 5 An estimated response rate includes an estimation of the number of refusals who would have been eligible to participate in the survey as most refusals occurred before eligibility could be determined.