National Travel Survey (NTS)

Detailed information for third quarter 2023

Status:

Active

Frequency:

Quarterly

Record number:

5232

The National Travel Survey provides statistics on the activities of Canadian residents related to domestic and international tourism. It was developed to measure the volume, the characteristics and the economic impact of tourism.

Data release - February 27, 2024

Description

The National Travel Survey was developed to fully replace the Travel Survey of Residents of Canada (record number 3810) and replace the Canadian resident component of the International Travel Survey (record number 3152). The National Travel Survey collects information about the domestic and international travel of Canadian residents.

The National Travel Survey, sponsored by Statistics Canada, aims to measure the characteristics and the economic impact of the tourism activities of Canadian residents. The objectives of the survey are to provide information about the number of trips and expenditures by Canadian residents by trip origin, destination, duration, type of accommodation used, trip reason, mode of travel, etc.; to provide information on travel incidence and to provide the socio-demographic profile of travellers and non-travellers. From a macroeconomic point of view, the NTS measures the domestic and international tourism demand by Canadian residents.

Reference period: Quarterly (Q1, Q2, Q3 and Q4) and annually from January to December.

Collection period: The month following the reference month.

Subjects

  • Domestic travel
  • International travel
  • Travel and tourism

Data sources and methodology

Target population

The target population is the civilian, non-institutionalized population 18 years of age or older in Canada's ten provinces. Specifically excluded from the survey's coverage are persons living on Indian reserves and persons living in the territories. Also excluded are out-of-scope trips such as routine trips and trips taken by commuters and diplomatic or military personnel.

Instrument design

The content of the National Travel Survey electronic questionnaire was drawn from the Travel Survey of Residents of Canada and the International Travel Survey, which were based on consultation with several tourism provincial organizations/departments. Statistics Canada System of National Accounts participated in the questionnaire design.

The questionnaire underwent 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 a new introduction to the survey and different trip definitions. There were two pilot tests done. The first pilot test done in February-March 2016 was used to evaluate multiple letter-based respondent selection methods. The conventional method of random selection was to select a household and use the application to select a respondent. The first pilot provided information on the ability of household members to interpret and comply with the random selection method described in the letter.

The second pilot test done in August 2017 was used to evaluate the online response application and to estimate the take-up rate. The second survey pilot was also used to evaluate multiple nonresponse follow-up strategies including mail out of letters, follow-up courtesy calls and phone calls to offer to complete the questionnaire over the phone.

Sampling

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

Sampling units:
The NTS has a three-stage sampling design, so there are three sampling units. The primary sampling unit (for the first stage) is the household. The sampling unit for the second stage is the person in the dwelling chosen from the first stage. Finally, the sampling unit for the third stage is the trip from a given dwelling and a selected person.

Stratification method:
The sampling frame is stratified by province, household income level and information obtained from Primary Inspection Kiosks (PIKs). The National Travel Survey is a flexible survey that allows stakeholders to add sample in specific cities or regions.

Sampling and sub-sampling:
For the first stage, the household, the sample is first allocated among the provinces using the cubic root of the number of dwellings in each province. For each province, a small portion of the sample (maximum of 400) is associated with a particular stratum based on information from PIKs. This stratum maximizes the chances of sampling Canadians who have travelled outside the country. Then, the rest of the sample is allocated according to income level in relation to the square root of the sum of the incomes. Within each stratum, by province and income level, dwellings are sorted by postal code. In the strata formed using information from PIKs, a simple random sample is drawn. In all other strata, a systematic sample is drawn. This allows the different regions of each province to be represented in the sample.

For the second stage, one adult per selected dwelling is randomly chosen using a selection method based on the age of household members. This method randomly chooses one adult for dwellings with up to six adults. Depending on the number of adults living in the dwelling, the oldest; the second oldest; the third oldest; the youngest; the second youngest; or the third youngest adult is selected.

For the third stage, the electronic questionnaire asks the respondent for a short description of all trips ending in the reference month. If the respondent reports three trips or less, then all the trips are chosen for the remainder of the questionnaire. Otherwise, an algorithm in the electronic questionnaire selects three trips according to a Poisson sequential sampling plan. Overnight trips abroad and overnight trips to a province other than the province of residence have a greater sampling fraction.

Data sources

Data collection for this reference period: 2023-08-01 to 2023-11-10

Responding to this survey is voluntary.

Data are collected directly from survey respondents.

Selected households receive an invitation letter in the mail. The letter explains who, from the household, is selected to participate in the survey using the age selection method. A household may receive up to two paper reminders, three email reminders, or one SMS reminder. The access code in the letters gives the respondent access to the electronic questionnaire. The electronic questionnaire is offered in the two official languages: French and English. The respondent must provide basic information on all of his or her trips (domestic and international) that ended in the reference month. The respondent then provides details on the trips selected. The average time required to complete the survey is 15 minutes.

In the case of a natural disaster, any dwellings that are located in the impacted areas (as identified by Natural Resources Canada) are removed from the collection strategies for that period of time. The removal of these units is taken into account during the household weighting (see Estimation).

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

Error detection

The electronic questionnaire is the only collection mode; respondents enter their responses to the survey questions directly in the electronic questionnaire. The use of electronic questionnaires reduces data processing time and the costs associated with data entry, transcription errors and data transmission. Responses are sent securely using industry-standard encryption protocols, firewalls and encryption layers.

Some edits are done as the electronic questionnaire is completed by the respondent. When the information is outside the range of expected values (too large or too small), or inconsistent with previous entries, the respondent is prompted, through messages on the screen, to check the information. However, for some questions, the respondent may ignore the edits and skip questions if they do not know the answer or refuse to answer. For this reason, the response data undergo further edit and imputation processes after being received at the Head Office.

Data are sent to Statistics Canada, where the information is processed in stages in preparation for dissemination. Data are checked to identify any inconsistencies. Trip records are validated to ensure that values in mandatory fields are acceptable. For some variables, a range of acceptable values is used. For example, we make sure that the number of nights falls within the logical range, that the type of trip is valid, etc.

For the majority of trip records, the geographic area is coded automatically. For a small number of records, coding is done manually at Statistics Canada's Head Office.

Several consistency edits are carried out on the data to verify the relationship between two or more variables. For example, the number of adults in a household who went on a trip cannot exceed the total number of adults in the household. If a city or other specific geographic location does not correspond to the province or other larger geographic area, only one location will be retained, depending on the question. For expenditure variables, several edit rules are applied to limit these values. If the value does not fall within the predetermined acceptable range, it is imputed later.

Imputation

The number of adults and children in the same household who went on the trip is imputed if the data are missing.

Expenditures are imputed in trip records to ensure that all trips have valid expenditures. Expenditures are imputed by category: commercial transportation, accommodations, restaurants, etc. Travel expenses that failed the edit are imputed using the donor imputation method. Variables correlated with the variable to impute are used to create imputation classes. Before choosing a donor, the Sigma-gap outlier detection method is applied to each variable to impute in order to exclude atypical values from the potential donor pool. Finally, a donor with a ratio between the 5th and 95th percentiles in the class is randomly selected. A ratio is imputed rather than the variable itself, since the variable may be strongly correlated with other variables. For instance, the ratio could represent daily expenses or expenses per person during the trip.

Another step in the imputation process consists of distributing package trip expenditures to specific expenditure categories. If the total amount is missing, it is imputed using a donor. Trips are then combined by class and the distribution of the expenditures observed for non-package trips is applied to package trips.

Another step involves distributing expenditures from the domestic components of international trips and the external components of domestic trips to specific expenditure categories. In the questionnaire, respondents are asked to provide the total amount spent in Canada during international trips, and the total amount spent outside Canada during domestic trips. If the total amount is missing, it is imputed using a donor. Next, a fixed distribution by trip type is used to distribute expenditures by category. This distribution is estimated from historic TSRC data.

The final step consists in distributing the expenditures for each category of a trip to each visit on the trip. This distribution depends on the expenditure category, the trip type and duration, what is included in the trip, the type of visit, etc.

Estimation

NTS estimates are produced using survey data to which weights are applied, so that these data can be inflated to the non-institutionalized Canadian population 18 years and older. The weights calculated to produce these estimates are household weights, person weights, trip weights and person-trip weights.

The household weight is calculated as being the inverse of the probability of selection of a household. Next, some adjustments are done within homogeneous groups to correct the unresolved units, i.e., households that are not definitively in scope or out of scope. Then the weight of out-of-scope households, such as vacant or destroyed homes, is set to 0. Weights are then adjusted within homogeneous response groups to correct the effect of non-response. For both the unresolved and the non-response adjustments, models are first applied to the data to estimate the corresponding probability for each sampled unit which is then used to create the homogenous groups. After adjusting for non-response, the weights are calibrated based on the number of households per province and household size (one person, two people, more than two) to known control totals.

The person weights are derived from the household weights. The inverse of the probability of selecting the person is multiplied by the household weight, giving a first version of the person weight. A simple reweighting is done at the stratum level to adjust the weights for the unresolved persons, such as people with a missing province of residence or missing age. The weight for out-of-scope units, such as Canadian non-residents and residents under 18 years of age, is set to 0. Then another simple reweighting is done at the stratum level to adjust the weights for non-response. It should be noted that most non-response comes from the first degree, i.e., the household, and that is why a decision was made to use a simple adjustment to correct non-response at the person level. Finally, the person weight is calibrated to known control totals (province-age-sex groups and Census Metropolitan Area (CMA)).

The initial person-trip weight is derived from the person weight, which is multiplied by the inverse of the probability of selecting a trip and the number of identical trips plus 1. Next, the weight for the out-of-scope trips is set to 0. The remaining weights are then adjusted to offset the presence of non-response. This adjustment is a post-stratification, where post-strata are created by province of origin, destination, type of trip and the main reason for the trip. Finally, the person-trip weights for selected domestic trips are calibrated to the weighted estimates for reported domestic trips. The person-trip weights associated to international trips are calibrated to counts of adult Canadian travellers estimated by Statistics Canada's Frontier Counts program.

Finally, the trip weight is derived by dividing the person-trip weight by the number of adults (18 and over) in the household who went on the trip. These trip weights are used to estimate expenditures. The Sigma-gap outlier detection method is applied to the weighted expenditures to reduce the impact of highly influential units by restricting the trip weight. After this step, person-trip weights are recreated, and the person-trip calibration steps are repeated on the new weights. The final trip weights are derived by dividing the new person-trip weights by the number of adults in the household who went on the trip.

The NTS uses the bootstrap method, a replicate-based method, for calculating variance. The bootstrap method involves taking subsamples with replacement from the sample and weighting them. Weighting is repeated several times (500 times for the NTS).

Quality evaluation

Data quality is systematically evaluated every quarter. Statistical tables required for analysis are produced and compared with related data sources. A set of indicators is also produced. They are used to determine whether general tourism trends reflect those of the NTS. Furthermore, we work in close cooperation with provincial tourism departments, which provide additional viewpoints and information sources, helping us evaluate data quality at a more refined geographic level.

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.

In order to prevent any data disclosure, confidentiality analysis is done using the Statistics Canada Generalized Disclosure Control System (G-Confid). G-Confid is used for primary suppression (direct disclosure) as well as for secondary suppression (residual disclosure). Direct disclosure occurs when the value in a tabulation cell is composed of or dominated by few enterprises while residual disclosure occurs when confidential information can be derived indirectly by piecing together information from different sources or data series.

Revisions and seasonal adjustment

This methodology does not apply to this survey program.

Data accuracy

Sampling variability is the error in the estimates caused by the fact that the survey is conducted on a sample of respondents rather than on the entire population. Standard error and the related concepts of coefficient of variation (CV) and confidence interval provide an indication of the magnitude of sampling variability. The standard error and coefficient of variation do not measure systematic biases in survey data that might affect estimates. Rather, they are based on the assumption that sampling errors follow a normal curve.

An estimate based on a larger sample size usually has a smaller CV and therefore is more reliable. Moreover, given two estimates based on the same sample size, the one associated with the most evenly distributed characteristic throughout the population has a smaller CV than the other.

The CV is used to identify five levels of data quality:

- 0.00% to 5.00%: excellent
- 5.01% to 15.00%: very good
- 15.01% to 25.00%: good
- 25.01% to 35.00%: acceptable
- Greater than 35.00%: use with caution

Estimates that do not meet an acceptable level of quality are either flagged for caution or suppressed.

Response rates:
The response rates are calculated, for each domain (province or Canada), using the following formula:
Response rate = number of respondent units / (number of contacted units - number of out-of-scope units)
The overall weighted response rate (Canada level) for the NTS for the third quarter of reference year 2023 was 25.9%. Response rates vary from province to province.

Non-sampling error:
There are sources of error other than sampling error, such as non-response bias, recall error, measurement error, etc. For the NTS, corrective measures are only applied when there is non-response bias.

Non-response bias:
Since the household response rate is less than 50%, there is a risk of non-response bias. A logistic regression is used to model the response probability for sampled units. The independent variables used in the regression model are correlated with travel-related expenditures and the probability of responding to the survey. Units with similar response probabilities are combined and the weight of non-respondents is distributed to respondents in these classes. This procedure reduces the potential bias caused by the presence of non-response.

Coverage error:
Approximately 10% of the dwellings in the frame do not have a valid address. These dwellings are not covered by the NTS, since the invitation to participate in the survey is sent by mail. These dwellings can still be sampled and considered unresolved. Therefore, the weight adjustment that redistributes the weights of this type is used to correct coverage error.

Report a problem on this page

Is something not working? Is there information outdated? Can't find what you're looking for?

Please contact us and let us know how we can help you.

Privacy notice

Date modified: