Retrospective Report on the 42nd General Election of October 19, 2015
Appendix 1: Overview of the Public Opinion Research Studies
This report draws on data from several surveys that are outlined below. While surveys are highly valuable research and feedback tools, they are also subject to inherent limitations, among which are the following:Footnote 26
- Self-selection bias. It is increasingly difficult to get people to respond to surveys, and those who do take the time to respond may have different profiles than those who do not. They may simply be more interested in the subject, and thus contribute to skew the findings with certain types of behaviours or attitudinal characteristics.
- Social desirability bias. Respondents can, consciously or not, give the answer that they think the surveyor wants to hear or that they think will put them in the best light. Voting, for example, is a behaviour that is typically over-reported in surveys.
- Memory effect. Post-election surveys sometimes take place days, if not a few weeks, after an election. This may affect the quality of the respondents' answers and amount of details they are able to remember.
- Small groups and low prevalence. Surveys are very useful and relatively accurate for identifying mainstream trends from large samples. However, estimating patterns within smaller subsets of the population or low-prevalence behaviours is much less accurate. For example, an observation based on a sample of 2,500 respondents yields a typical ±1.96 percentage-point margin of error. The same observation based on a sample of only 25 respondents would yield a ±19.6 percentage-point margin of error.
Elections Canada commissioned R.A. Malatest & Associates Ltd. to conduct a survey of electors to measure their experiences, attitudes and opinions regarding various aspects of the 42nd general election. The firm completed 3,516 telephone interviews between late October 2015 and mid-January 2016. The response rate was 19.5 percent, and the margin of error was ±1.8 percentage points, 19 times out of 20 at the national level. The research design also included over-sampling of youth aged 18 to 24, Aboriginal people, immigrants and electors with disabilities to ensure sufficient representation among these groups of interest.
Elections Canada commissioned EKOS Research Associates Inc. to survey the 1,792 candidates to assess their experiences and measure various attitudes, such as their satisfaction with the administration of the 42nd general election. A total of 916 candidates completed the survey, for a response rate of 54 percent (based on the 1,635 candidates with valid contact information). Since the firm attempted to obtain participation from all candidates, just like for a census, there is no applicable margin of error.
Elections Canada commissioned Forum Research Inc. to conduct a survey of election officers to gather their impressions on the overall administration of the vote. The survey firm conducted telephone interviews with 3,503 election officers randomly selected from a list of 201,602 officers. The sample was stratified to ensure representation from deputy returning officers, central poll supervisors, information officers, poll clerks and registration officers as well as to ensure good coverage of ordinary polling stations on election day, advance polls, mobile polls, senior residence or long-term care polls, Aboriginal reserve polls and student campus polls. The overall response rate was 51 percent, and the margin of error was ±1.7 percentage points, 19 times out of 20.
For the second time since 2011, Elections Canada collaborated with Statistics Canada to include questions on voter participation in the Labour Force Survey following a general election. The Labour Force Survey is administered by Statistics Canada on a monthly basis to collect labour market information for approximately 100,000 individuals based on a sample of approximately 56,000 households.Footnote 27 Participation in the survey is mandatory, but answering Elections Canada's questions in the November 2015 survey was voluntary. The response rate to Elections Canada's filter question on citizenship was 96.7 percent. The fact that almost all survey participants volunteered to answer Elections Canada's questions significantly reduces the selection bias usually observed in traditional political or electoral surveys. Compared to other surveys—where the sample size is usually between 1,500 and 3,500 respondents—the Labour Force Survey, with its very large sample, allows for more meaningful statistical analysis, particularly for studying smaller subsets or more marginal aspects of a subject.
Elections Canada commissioned Nielsen Consumer Insights to conduct a series of surveys and focus groups with electors to assess the reach and effectiveness of its Electoral Reminder Program. The firm conducted a total of 6,005 online surveys (using a proprietary online panel) in five phases, as each phase of the Electoral Reminder Program was rolled out. Nielsen also held 17 in-person and online qualitative focus groups in five cities across the country (Halifax, Montréal, Mississauga, Winnipeg and Vancouver). The research design included representation from electors among the general population and from target groups of interest, including youth aged 18 to 24, students, ethnocultural groups, Aboriginal electors and electors with disabilities. The evaluation report is based on the online surveys and focus groups; therefore, no margin of error can be calculated.
Return to source of Footnote 26 For example, see W. Lawrence Neuman, Social Research Methods: Qualitative and Quantitative Approaches, 3rd ed. (Boston: Allyn & Bacon, 1997).
Return to source of Footnote 27 In each dwelling, information about all household members is usually obtained from one household member.