Data quality, concepts and methodology: Description of methods

View the most recent version.

Two methods are presently used to collect international travel statistics: the "Frontier counts" and the "Questionnaire surveys". Both of these methods depend greatly on the co-operation of Canada Border Services Agency (CBSA) in the collection of the number of border crossings and the distribution of questionnaires to international travellers.

Frontier counts

All ports of entry across Canada participate in determining the number of travellers by selected categories and by type of transportation, as well as the number of cars, trucks, motorcycles, snowmobiles and bicycles in the case of highway and ferry points.

The forms used by CBSA officials in the enumeration process are described as follows:

E-62 Entry tally. Form used to record the number of daily travellers and vehicles arriving by land and by ferry, at points of entry on the United States-Canada border. Each form indicates the number of automobiles, trucks, motorcycles, bicycles, snowmobiles and their associated travellers cleared for entry by CBSA agents, as well as the travellers coming to Canada by other modes of transport such as bus, train and on foot. Since 1998, the Integrated Primary Inspection Line (IPIL) formerly known as the Primary Automated Lookout System (PALS) has been used to replace the manual E-62 tally. Presently, 37 ports across Canada utilize this automated system to record over 88.0% of Canada's international automobile traffic.

E-63 Commercial and private craft/passenger and crew arrivals. Form used to record travellers entering Canada by private plane or boat. This form is also used to record travellers and crews on commercial freighters, passenger ferries, cruises and some commercial air flights. The CANPASS telephone reporting system also records the number of travellers entering Canada by private plane or boat. The CANPASS system also allows pre-authorized travellers, as well as special permit holders, to cross the border by car without CBSA's agents interaction.

E-311 Customs declaration card. Form used to record on a census basis travellers entering Canada by commercial plane, including schedule and charter arrivals, at the 18 major international airports. The family Customs Declaration Card was introduced in April 1999. The purpose of this card is to facilitate the clearing of international travellers at the point of entry. Up to four family members living at the same address are allowed on the same card compared to only one person on the old individual E-311 card. In order to estimate the number of travellers recorded on the family Customs Declaration Cards, data capture is done on a sample or census basis depending on the traveller type and airport size (see text table 1). For other airports, census based administrative data is obtained from E-63 forms to produce the estimates.

Detailed instructions are provided to aid CBSA officials in the task of collecting data on these documents. A continual liaison function is performed by the International Travel Section with the supplying department, CBSA, to review, discuss and resolve inconsistencies in the reported figures.

A monitoring system has been established to compare incoming data with information available from independent sources, such as airport management reports, toll figures, provincial road counters, etc. The above, as well as reference to weather reports, special events and direct contacts with port authorities, helps the verification and the explanation of irregular fluctuations in reported figures.

The completed forms are submitted continuously to Statistics Canada for processing, analysis and publication. In 2009, the number of documents processed was estimated as follows: 132,700 E-62; 644,100 E-63; and 16,131,800 E-311.

Questionnaire surveys

Continuous questionnaire surveys are used to secure information on the expenditures and other characteristics of international trips and travellers to and from Canada.

Collection methods of mailback questionnaires

Mail-back questionnaires are handed out each quarter to the travel party on entry (non-residents) or re-entry (residents of Canada) by CBSA officials according to pre-arranged schedules. Five different questionnaires are used in the collection process.

Since 1979, a stint distribution system has been used to distribute these questionnaires to travellers at all border ports sampled. A stint consists of a selected period of several days during which questionnaires are to be distributed to eligible travellers. Each port involved in this scheme receives, for each of its stints, a specific quantity of numbered questionnaires and a date on which to start the distribution. On the start date, the officers hand out the questionnaires on a continuous basis to the appropriate travelling population until they have all been distributed. The Liaison staff of the International Travel Section call each port the day before the stint is to start to make sure the questionnaires have arrived and to remind the officers to start the distribution the next day. In this way, the questionnaire return rate of each stint and the length of the stint can all be used as a measure of performance for each stint.

The number of questionnaires distributed in each port for each quarter is set by taking into account the number of travellers who cleared customs through the port in that quarter during the previous year, as estimated by the frontier counts. Respondents are requested to mail back their completed questionnaires in Canada (for Canadian and overseas travellers) or from the United States (for U.S. travellers).

Here is a brief description of the mail-back questionnaires.

Questionnaire (8-2200-356) for United States travellers visiting Canada. Distributed by CBSA officials during sampling periods to travellers residing in the United States at most land, air and some sea ports. However, at land ports, questionnaires are distributed only to United States residents taking an overnight car trip to Canada or entering Canada by a mode of transportation other than car. The questionnaires are completed by the travellers and returned to Statistics Canada for processing.

Questionnaire (8-2200-337) for residents of countries other than the United States visiting Canada. Distributed by CBSA officials during sampling periods to travellers residing in overseas countries at most land, air and some sea ports; the questionnaires are completed by the travellers and returned to Statistics Canada for processing.

Questionnaire (8-2200-336) for Canadian residents returning from trips outside Canada. Distributed by CBSA officials during sampling periods to Canadian residents re-entering Canada, at most land, air and some sea ports. However, at land ports, questionnaires are distributed to Canadian residents who took a trip abroad, except those returning from a same-day car trip to the United States. The questionnaires are completed by the travellers and mailed to Statistics Canada for processing.

Questionnaire (8-2200-338) for Canadian residents returning from a same-day trip to the United States by auto (a visit of less than 24 hours). Distributed by CBSA officials to Canadian residents who return to Canada by car after a same-day trip to the United States (one questionnaire per vehicle). The questionnaires are completed and mailed to Statistics Canada for processing. Since the same-day traffic is considered to be fairly homogeneous, a less frequent sample is taken than for the longer-term traffic.

Questionnaire (8-2200-345) for United States residents entering Canada by auto for a same-day trip (a visit of less than 24 hours). Distributed by CBSA officials to United States residents taking a same-day car trip to Canada (one questionnaire per vehicle). The questionnaires are completed and mailed to Statistics Canada for processing. Since the same-day traffic is considered to be fairly homogeneous, a less frequent sample is taken than for the longer-term traffic.

Collection methods of Air Exit Survey of Overseas Visitors

In the case of overseas travellers returning directly (as opposed to via the United States) to selected overseas countries, additional questionnaires are obtained via the Air Exit Survey of Overseas Visitors (AES).

This survey has been conducted by Statistics Canada since 2000. Its primary objective is to improve the quality of estimates on the characteristics of trips and travellers by air from overseas countries. Personal interviews are conducted in designated airports in Canada. It is an add-on to mail-back questionnaires distributed by CBSA officials to residents of overseas countries visiting Canada.

At the international airports in five cities (Vancouver, Calgary, Toronto, Montreal and Halifax), Statistics Canada's interviewers conduct personal interviews with a sample of overseas travellers as they await their return flights to targeted overseas countries. The countries targeted are those from whom we attract the most travellers. They include principally the United Kingdom, France, Germany and Australia, as well as a number of smaller markets, such as Japan and Mexico. These interviews are conducted each month and the collection period lasts 3 to 7 days.

The number of interviews targeted at each airport for a particular month and a particular overseas country is obtained by selecting a representative sample of flights to the selected country. During the collection period, interviewers arrive at the departure lounges for the selected flights in order to interview departing overseas travellers. All overseas travellers waiting to embark on the selected flights are interviewed until the first boarding call, whether or not they are of the targeted countries. The interviewing team consists of interviewers of different language skills, enabling interviews to be conducted in the native language of the travellers where possible. The questionnaire used (Questionnaire 8-2200-400) is a customized version of questionnaire 8-2200-337 and is available in ten languages.

About 7,600 interviews were completed in 2009. Since 2000, this collection method has achieved a response rate of over 95.0%.

Processing and imputation of questionnaires

All mail-back and AES questionnaires for a given quarter are submitted to a complex control system and the reported data are captured, coded and verified. Only questionnaires accepted by the system are ultimately used for the production of the estimates for the quarter.

A certain amount of data imputation is performed in specific fields of accepted questionnaires. Missing international transportation fares and/or total trip spending are imputed if the remainder of the fields on the questionnaire are valid. The imputed values of the recipient questionnaire are calculated by taking into account the mean of corresponding fields on donor questionnaires that have trip characteristics identical to those of the recipient questionnaire.

Total imputation (i.e. imputation of complete questionnaires) is carried out for all Port Factor groups (PFGs) ¹  or strata that are outside the scope of mail-back questionnaire distribution. There are 120 Canadian and American PFGs for which Statistics Canada never receives questionnaires. These PFGs refer to modes of entry that are not targeted by the questionnaire distribution (train, private plane or boat, motorcycle, bicycle, pedestrians, etc.) or to ports of entry that do not participate in the distribution of questionnaires. For these out-of-scope PFGs, the characteristics of travellers and trips are estimated, using imputed questionnaires. These imputed questionnaires are duplicates of questionnaires that were obtained in 1990 for the same quarter and same PFGs. In 2009, these questionnaires represented only 3.4% of total travellers from the United States to Canada and 1.1% of Canadian residents who travelled outside Canada.

Total imputation is also performed for any in-scope PFG for which we have received an insufficient number of questionnaires for the quarter. In these instances, all the questionnaires from the same quarter of the previous year that belong to the PFG are brought forward and added to the sample of that PFG for the reference quarter.

If necessary, additional total imputation is also performed for U.S. car travellers by state of origin to meet minimum requirements based on the frontier counts.

In 2009, 45,400 questionnaires from non-resident travellers and 50,900 questionnaires completed by resident travellers were used to produce estimates.

Weighting

For estimation purposes, the responses obtained through the questionnaire surveys must be treated as a simple random sample from the total traffic in each stratum (port or group of ports, by type of traffic, by quarter). The data may in fact be subject to some degree of "distribution bias", due to the fact that the questionnaires may not be handed to a random selection of travellers, or to a "non-response bias" due to the fact that the individuals replying may not be representative of the population.

Complex weighting techniques used in the estimation process attempt to reduce the effect of these two types of bias. In order to do so, the questionnaires completed by the international travellers are disaggregated by known characteristics into homogenous groups (PFGs). In the case of questionnaires obtained from the United States and Canadian travellers, those characteristics are port of entry, type of traveller, mode of entry and duration of stay. In the case of questionnaires obtained from overseas travellers, the characteristics used are country of residence and type of entry. For example, two of the main factor groups include questionnaires from American travellers entering and leaving on the same day by car and those from American travellers entering by car and staying one or more nights.

The questionnaires for each PFG are first weighted according to frontier counts. The weights assigned to all questionnaires for a given PFG are such that the estimate of the number of travellers derived from these questionnaires matches the frontier count for that PFG.

Once the initial weights are assigned, bias adjustments are applied to the weights of certain questionnaires.

A second stage of weighting called the bias adjustment is done for questionnaires completed by certain categories of air travellers. These include U.S. travellers, Canadian travellers returning from the U.S. and Canadian travellers returning directly from overseas countries who came through the international airports of Vancouver, Toronto, and the province of Quebec. They also include overseas air travellers originating from the United Kingdom, Germany, France and Japan.

This second stage consists of modifying the initial weights of questionnaires that were obtained from these travellers in order to incorporate more reliable information on trip purpose and duration. This information is derived from a sample of Customs declaration cards (E-311 cards) filled by these travellers. Adjustments to the initial weights are such that estimates on trip purpose and duration compiled from the questionnaires correspond to those obtained from the E-311 cards.

By using more data from the E-311 cards, the procedure for weighting air traveller questionnaires has become more accurate. The use of more precise weights provides better quality estimates and limits the possibility of bias in the questionnaire results obtained for air travellers.

Starting with 2009, another bias adjustment is done for U.S. travellers entering Canada by car. For those travellers who came by car through the largest land ports equipped with the Integrated Primary Inspection Line (IPIL) system, a second stage of weighting is also done. Referred to as the state bias adjustment, it is based on the vehicle state registration information obtained from the IPIL system. The initial weights of these questionnaires are adjusted to better reflect the distribution of travellers from the states that provide the largest volume of travellers. For the top states of origin by PFG by quarter, the statistics obtained from the questionnaires will match those obtained from the frontier counts. The 2008 U.S. traveller statistics have also been revised to incorporate this state bias adjustment.

In addition, a supplementary adjustment, called the regional weighting adjustment, is done to the weights assigned to overseas questionnaires to take into account the region of entry of the travellers. Using the raking-ratio statistical technique, this adjustment ensures that the derived estimates of the number of travellers from the questionnaires closely match the frontier counts at the region of entry level (by type of entry), while maintaining an exact match at the country of residence level. However, the raking-ratio technique does not guarantee an exact match when the country of residence as well as the region of entry or type of entry are considered. The provincial composition of the five regions used in the process is as follows: Atlantic (Newfoundland, Prince Edward Island, Nova Scotia, and New Brunswick), Quebec, Ontario, Prairies (Manitoba, Saskatchewan, Alberta) and British Columbia (including Yukon, Northwest Territories and Nunavut).

In general, this supplementary adjustment allows for the production of more consistent and more reliable estimates on the characteristics of overseas travellers at the regional level.

Production schedule

Each quarter's data on the characteristics of international trips are first released as preliminary estimates, about five months after the end of the quarter. These estimates are based on all questionnaires for the quarter that are obtained before the preliminary cut-off date for the collection, which is about two and a half months after the end of the quarter.

Three months later, the revised estimates for the quarter are released, together with the preliminary estimates for the next quarter. Included in the production of the revised estimates are all questionnaires for the quarter that are received before the final cut-off date for the collection, which is about five and a half months after the end of the quarter.

Data quality and reliability indicators

Questionnaire distribution and response may be biased. The International Travel Survey (ITS) uses several methods to minimize this bias and improve the estimates. Questionnaires are weighted to frontier counts, which are robust statistics obtained from administrative sources. Thus, spending tends to be driven by volumes (frontier counts). Bias adjustments are done for U.S. and Canadian air travellers entering through the three largest international airports, for air travellers coming from four major overseas countries of origin, and for car travellers coming from the United States.

In sample surveys, since inference is made about the entire population covered by the survey on the basis of data obtained from only a part (sample) of the population, the results are likely to be different from the "true" population values. The true population values in this context refer to the values that would have been obtained if the entire population was surveyed under the same conditions. The error arising due to drawing inferences about the population on the basis of information from the sample is termed sampling error.

The sampling error, in addition to the size of the sample, depends on factors such as variability in the population, sampling design and method of estimation.

Additionally, even if the sample design, the size of the sample, and the estimation procedure were the same, different characteristics (on which data have been collected from the sample) would possess different sampling errors, due to the fact that they have different degrees of variability in the population. For instance, the sampling error for expenditure estimates would be different from the sampling error for length of stay, both estimates being based on the same sample. Each of the possible samples would yield somewhat different sets of results. The sampling errors are measures of the variation of all the possible sample estimates around the true values.

The expected value of a variable is the average of the variable taken over all possible samples weighted by the probability of drawing the sample. The difference between the expected value of an estimate and the corresponding true value is called the bias of the estimate. The mean square error of an estimate is defined as the expected value of the square of the deviation of the estimate from the true value. The variance of an estimate is defined as the expected value of the square of the deviation of the estimate from its expected value. Thus, when an estimate is unbiased (i.e., its bias is zero), its variance and mean square error are equal.

The accuracy of estimates from sample surveys is affected by both variance and bias. Under the assumption of simple random sampling within each stratum and with the further assumption of absence of bias, the variance of an estimated characteristic value is a good indicator of its reliability. Since the true variance of the estimate depends, like the estimate itself, on the whole population, it must be estimated from the available sample.

A notable feature of probability sampling is that the quality of the estimates may be estimated from the sample itself. The estimated coefficient of variation is defined as the ratio of the square root of the estimated variance to the estimate itself. Guides to the potential size of sampling errors are provided by the estimated coefficients of variation. The quality of the estimate increases as the corresponding coefficient of variation decreases.

Given the complex nature of the weighting procedures of the ITS, simple calculations of the variance using standard methods tend to underestimate it. Consequently, Statistics Canada uses the "bootstrap" method for estimating the variance. Under this method, the sample units are sub-sampled and re-weighted many times over. From these sub-samples, an estimated variance is obtained to calculate the estimated coefficient of variation.

In all tables in the publication, the estimated coefficients of variation were used to evaluate the reliability of estimates derived from the questionnaire surveys. The reliability of these estimates is identified as follows:

Because Statistics Canada does not control the distribution of the questionnaires by CBSA border service officers, response rates cannot be calculated for questionnaires. It is known that distribution, particularly at high volume ports, can be poor. In 2000, the survey had 54,600 returned and completed questionnaires as a result of CBSA distribution; this number dropped to 27,400 in 2009. As a result, in the corresponding period the number of imputed records has increased from 39,900 in 2000 to 61,800 in 2009. This is less of a problem on the overseas residents' file where the Overseas Air Exit Survey supplies many questionnaires. Imputation is proportionately higher for the United States travellers' file and the Canadian travellers' file. Initial trend analysis on the United States file has shown that much of the imputation is of short duration trips and the trend, in terms of spending, is not changed much by the imputation. Statistics Canada is working with CBSA to improve questionnaire distribution and is looking into developing alternative methods of imputation.

Dissemination

In addition to this publication, international travel statistics can be obtained by accessing or purchasing the following products.

Data from the frontier counts are available in CANSIM, tables 427-0001 to 427-0006. Data on the international travel account are also available in CANSIM, tables 376-0001, 376-0003 and 376-0005.

Annual data on certain characteristics of international trips are available free of charge on Statistics Canada's website.

Statistics Canada Daily, Travel between Canada and other countries, catalogue no. 11-001-X. Publishes preliminary monthly counts of international travellers as soon as they become available – usually six weeks after the reference month. This release covers the four travelling populations, usually at the national level: United States residents visiting Canada, residents from overseas countries visiting Canada, Canadian residents returning from the United States and Canadian residents returning from overseas countries.

Statistics Canada Daily, Characteristics of international overnight travellers, catalogue no. 11-001-X. Approximately five months after the end of the quarter, a brief text about the trends in the main characteristics of international overnight travellers is released in the Daily to announce the availability of preliminary data.

Statistics Canada Daily, International travel account, catalogue no. 11-001-X. Approximately two months after the end of the quarter, an analysis of the trends in the international travel account is released in the Daily to announce the availability of preliminary data.

International Travel, Advance Information, catalogue no. 66-001-PWE. A monthly document listing preliminary figures on international trips is usually available six weeks after the reference month. Table 1 breaks down the number of trips by foreign travellers and trips by Canadian travellers returning to Canada by mode of transport and length of stay. Tables 2 and 3 outline a selected list of categories of foreign overnight travellers to Canada by province of entry. Table 4 gives a detailed list of Canada's major overseas travel markets and Tables 5 and 6 provide a breakdown of overnight trips by Canadians returning by province of re-entry from both the United States and all other countries.

Micro-data files. Available per flow per quarter. Four flows exist: Canadian residents returning from the United States, Canadian residents returning from overseas countries, United States residents visiting Canada and residents from overseas countries visiting Canada.

Next | Previous