open Secondary menu

Public Opinion Survey Following the November 25, 2013 By-elections

Section 1: Introduction and Methodology

R.A. Malatest & Associates Ltd. (herein "the Consultant") administered and analyzed the results from the Public Opinion Survey following the November 25, 2013 By-elections for Elections Canada. This section places this survey within the broader context of Elections Canada's ongoing post-election evaluation work, describes the methodology implemented by the Consultant, and provides notes for readers to facilitate their understanding of survey findings.

1.1 Context

It is the mission of Elections Canada to ensure Canadians can exercise their democratic rights, both as electors and candidates. As such, it is mandated to be prepared at all times to conduct a federal general election, by-election, or referendum.

Elections Canada conducts post-election evaluations, including public opinion surveys of electors, following each general election, by-election, or referendum. These surveys are carried out by independent evaluators using a standardized survey instrument developed by Elections Canada. This report presents the findings of the survey of electors conducted following by-elections held on November 25, 2013 for four federal ridings:

  • Bourassa, Quebec;
  • Brandon–Souris, Manitoba;
  • Provencher, Manitoba;
  • Toronto Centre, Ontario;

Post-elections surveys provide an opportunity to garner feedback from a broad spectrum of electors. The sample represents both English and French communities, urban and rural regions, and significant populations of new Canadians and students.

1.2 Methodology

The survey was administered between December 19, 2013, and January 24, 2014.

Prior to launching data collection, Elections Canada prepared a fifteen minute questionnaire suitable for telephone administration, which aimed to evaluate electors' opinions and knowledge of Elections Canada's services and to measure various aspects of electors' experiences with the electoral process.

A minimum of 400 individuals from the general population were surveyed in each of the four ridings. In total 1,631 valid completions were obtained, with all respondents verified as living within one of the four ridings during the by-election. Data was then weighted by age and gender to improve the representativeness of the sample based on Statistics Canada's riding profiles (from the 2011 Census). Footnote 1 The margins of error for results reported are ±4.9 percentage points for riding-level data, and ±2.4 percentage points for overall data. The confidence level is 95 percent (19 times out of 20).

Refer to Appendix B, Study Methodology, for more information.

1.3 Limitations

A limitation with surveys of this nature is that they tend to over-represent respondents that indicate that they voted. Among survey respondents, survey data suggests that self-reported turnout ranged from a high of 68 percent in Brandon–Souris to a low of 52 percent in Bourassa. In actuality, the turnout rate was much lower, ranging from a high of 45 percent in Brandon–Souris to a low of 27 percent in Bourassa (see Section 3.2 for further details). Previous post-election research has demonstrated a similar effect.

Two factors may be responsible for these overestimated turnout rates. For one, people who vote are more likely than non-voters to participate in studies about voting. Second, some respondents may have demonstrated social desirability bias, in that they did not in fact vote but reported doing so in order to present themselves in a more positive light. The net result of these factors is two fold. For one, groups generally less likely to vote, such as youth, are under represented in the survey sample than others. Additionally, non-voters' views and experiences are underreported.

The Consultant made all efforts possible to garner the views of non-voters, including setting sub-regional quotas within each riding to ensure a broad diversity of communities were surveyed, coaching telephone interviewers throughout the project to improve response rates, and weighting the final database by age and gender to improve the representativeness of the sample. Weighting in particular improves representation of non-voters as younger electors are both less likely to vote and less likely to participate in post-election evaluations.

1.4 Note to Readers

Throughout this report the terms "electors" and "respondents" are used to denote all survey participants. The term "voters" denotes survey participants who reported having voted in the by-elections. Additionally the ridings of Brandon–Souris and Provencher are often referred to as "the two Manitoba ridings" or "the two rural ridings" and Bourassa and Toronto Centre as "the two urban ridings." This was done for editorial reasons as many trends observed cleave between these two pairs.

Figures illustrating results for questions which were only asked of select respondents are followed by a "base" definition (e.g., "Base: Electors who received a voter information card"); results for questions asked of all respondents have no such note.

For each question the overall finding is presented, followed by results by riding. Then, where possible, statistically significant trends related to socio-demographic groups are discussed (i.e. those deemed significant at the 95 percent confidence interval, and where differences between categories are outside the margin of error). The following groups are analyzed:

  • Age: "Youth" (18–34), "Middle–aged" (35-54) and "Older" (55+) electors;
  • Gender: Male and female electors;
  • Employment status: Employed electors (including full-time, part-time and self-employed), unemployed electors (including those looking for work, and those who stay at home full-time), students, and retirees. Other situations (e.g., disability pension, extended leave due to illness, full-time volunteers) were excluded from analyses due to their relatively low frequency;
  • Household income (before taxes): "Lower-income" (Less than $40,000/year), "Middle-income" ($40,000 to less than $80,000), and "Higher-income" ($80,000/year or more); and
  • Voting behaviour: Voted or did not.

Other dimensions, including disability status, type of dwelling, technology use, and previous voting behavior (i.e., in the 2011 general election) are used to compare results where relevant.


Footnote 1 Note: these riding profiles include ineligible electors such as non-Canadian citizens.