National Cannabis Survey

Detailed information for fourth quarter 2019

The main objective of the National Cannabis Survey is to better understand the frequency of cannabis usage in Canada and to monitor changes in behaviour as a result of the legalization of cannabis for non-medical use.

Data release - February 19, 2020

• Questionnaire(s) and reporting guide(s)
• Description
• Data sources and methodology
• Data accuracy

Description

The survey will be used in conjunction with other data sources to understand how the legalization of cannabis for non-medical use could impact the Canadian economy as well as other health and social services.

Reference period: The reference period in the questionnaire is the previous three months.

Subjects

• Economic accounts
• Health

Data sources and methodology

Target population

The target population for the survey is non-institutionalized persons 15 years of age or older living in Canada's ten provinces.

Instrument design

The content for the National Cannabis Survey electronic questionnaire was drafted in consultation with the System of National Accounts team within Statistics Canada as well as several other federal government departments and agencies, including Health Canada, the Public Health Agency of Canada, Public Safety Canada and the Department of Justice of Canada.

The questionnaire was subject to cognitive testing in the form of in-depth interviews in both of Canada's official languages, conducted by Statistics Canada's Questionnaire Design Resource Centre. The goal of the qualitative study was to test the survey content.

Sampling

This is a sample survey with a cross-sectional design.

Frame:
The frame is the Dwelling Universe File. Institutions, collective dwellings and dwellings on Indigenous reserves were excluded.

Sampling unit:
The NCS sample has a two-stage design: the sampling unit for the first stage is the dwelling, and the sampling unit for the second stage is the person.

Stratification method:
The NCS frame was stratified by province, and a simple random sample of dwellings was selected independently within each province.

Sampling and sub-sampling:
Sufficient sample was allocated to each of the provinces so that the survey could produce provincial level estimates. An initial sample of 12,000 dwellings was selected and sent to collection.

Data sources

Data collection for this reference period: 2019-11-13 to 2019-12-17

Responding to this survey is voluntary.

Data are collected directly from survey respondents.

Data are collected either through an electronic questionnaire or through computer assisted telephone interviewing.

View the Questionnaire(s) and reporting guide(s) .

Error detection

Electronic files containing the daily transmissions of completed respondent survey records were combined to create the "raw" survey file. Before further processing, verification was performed to identify and eliminate potential duplicate records and to drop non-response and out-of-scope records.

In addition, some out-of-scope respondent records were found during the data clean-up stage. All respondent records that were determined to be out-of-scope and those records that contained no data were removed from the data file.

After the verification stage, editing was performed to identify errors and modify affected data at the individual variable level. The first editing step was to identify errors and determine which items from the survey output needed to be kept on the survey master file. Subsequent to this, invalid characters were deleted and the remaining data items were formatted appropriately.

Imputation

This methodology type does not apply to this statistical program.

Estimation

The estimation of population characteristics from a sample survey is based on the premise that each person in the sample represents a certain number of other persons in addition to themselves. This number is referred to as the survey weight. The process of computing survey weights for each survey respondent involves several steps.

1) Each selected dwelling is given an initial weight equal to the inverse of its selection probability from the sampling frame. Dwellings identified as out-of-scope during collection are dropped from the sample.

2) The weights for responding households are adjusted to represent the households that did not respond. Adjustment factors are calculated separately by province and dwelling type (single-detached house / other). In quarter 4, the adjustment factors calculated were broken down further within the dwelling type "other", based on the collection period length, as some of these units had a shorter collection period.

3) The household weights are calibrated so that the sum of the weights match province level household size demographic counts.

4) Person weights are computed by multiplying the household level weights by the inverse of the probability of selecting the person within the household.

5) The person weights are calibrated so that the sum of the weights match demographic population counts at the province by age group by gender level. The weights are also calibrated to demographic counts for large census metropolitan areas.

Variance estimation is based on a resampling method called the bootstrap.

The Generalized Estimation System was used to generate the survey weights and bootstrap weights.

Quality evaluation

While quality assurance mechanisms are applied at all stages of the statistical process, the validation and review of data by statisticians is the final verification of quality prior to release. Validation measures were implemented, they include:

a) verification of estimates through cross-tabulations
b) consultation with stakeholders internal to Statistics Canada
c) consultation with external stakeholders.

Disclosure control

Statistics Canada is prohibited by law from releasing any information it collects that could identify any person, business, or organization, unless consent has been given by the respondent or as permitted by the Statistics Act. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.

Revisions and seasonal adjustment

This methodology type does not apply to this statistical program.

Data accuracy

Survey errors come from a variety of different sources. One dimension of survey error is sampling error. Sampling error is defined as the error that arises because an estimate is based on a sample rather than the entire population. Sampling error can be expressed through a confidence interval or coefficient of variation.

The following are approximate sampling error estimates for Canada level estimates. These are based on average results; these are not results for a specific variable.

- Approximate length of 95% confidence intervals for a proportion of 50% (Canada level): 4.0%
- Approximate length of 95% confidence intervals for a proportion of 10% (Canada level): 2.5%
- Approximate coefficients of variation for a proportion of 10% (Canada level): 6%

Response rates
The response rate for Quarter 4 of the 2019 National Cannabis Survey was 47.24%.

Non-sampling error
The first type of errors treated were errors in questionnaire flow. For skips based on answered questions, all skipped questions were set to "Valid skip" (6, 96, 996, etc.). For skips based on "Non-response", all skipped questions were set to "Not stated" (9, 99, 999, etc.). The remaining empty items were filled with a numeric value (9, 99, 999, etc., depending on variable length). These codes are reserved for processing purposes and mean that the item was "Not stated".

Non-response bias
The survey estimates are adjusted to account for non-response through the survey weights. To the extent that the non-responding households and persons differ from the rest of the sample, the results may be biased.

Coverage error
Coverage errors arise when there are differences between the target population and the observed population. The observed population is persons living in dwellings with mailable addresses on the frame. Approximately 95% of the dwellings on the frame had mailable addresses. To the extent that the excluded population differs from the rest of the target population, the results may be biased.