3.63. As compared with the aim of phase one, which is dedicated to the observation of global flows, the aim in phase two is to obtain more details about the characteristics of the travellers making the trips. What first needs to be established is, whether the traveller is a non‑resident (but only if the information generated in phase one is insufficient for that purpose). Additional information is then needed from non‑residents to determine whether their travel constitutes a tourism trip. If so, then further information is collected on the characteristics of the visitor, his or her travel party (if travelling in one) and the conditions of the trip (e.g., duration, purpose, modes of international transport, main type of accommodation used and arrangements). This is all the information necessary to characterize visitors and their trips and, in a subsequent stage, to generate statistics on their tourism expenditure.
3.64. The information to be requested should be useful, for policy and analytical purposes, to both national authorities (mainly NTAs, NSOs and central banks) and other stakeholders in the tourism sector. Whether persons travel alone or in parties, for instance, affects not only average expenditure but also the type of accommodation needed (single, double or multiple‑occupancy rooms). The reason of persons for travel, whether business or recreation, will have an impact on their activities, including the times at which they will be available for touring or other recreational or cultural activities.
3.65. With very few exceptions (e.g., when information can be collected from E/D cards completed by visitors upon departure), tourism authorities will need to use surveys to establish the characteristics of inbound visitors. Though flows of inbound tourism are measured on arrival, those surveys will usually need to be conducted upon departure, for two basic reasons:
- To avert prolongation of the border transit process for arriving visitors and immediately assailing them with a multitude of questions
- Because it is more accurate to establish the actual characteristics of the visitor, particularly with respect to expenditure, length of stay and activities undertaken while in the country. Only the expectations of visitors about such matters can be recorded upon their arrival and those expectations can be highly inaccurate, particularly in relation to expenditure.
C.2.1. Issues specific to the observation of the characteristics of inbound visitors
3.66. While, in most tourism observation procedures, the characteristics of inbound trips and inbound visitors are established as visitors leave the country, in fact, this information is then assigned to arrival figures estimated for the period of reference.
3.67. Figure III.2 below provides compilers with the hidden assumptions behind the most frequently used procedure which is to assign all tourism activity to the period of reference corresponding to the arrival of the visitor. As displayed, this practice is strictly appropriate in any one of only four possible types of circumstances.
Flows of visitors and period of reference
Type 1. A person arrives at the place visited before, and leaves during, the period of reference (t). In such a case, the person will be registered as arriving in period t-1, even if part of the person’s tourism activity takes place during the period of reference. Consequently, measurements of all of the person’s activity during the trip, including expenditure, will be assigned to period t-1.
Type 2. A person arrives before and leaves after the period of reference (t). As for type 1, the person’s tourism activity during period t will be assigned to period t-1, although it occurs party in period in t-1, partly in period t and partly in t+1.
Type 3. A person both arrives and leaves during the period of reference (t). In this “ideal” case all of the person’s tourism activity occurs during period (t) and enters into the calculation.
Type 4. A person arrives on a tourism trip during and leaves after the period of reference (t). That person will be registered as arriving during period t although part of his or her activity will take place also during period t+1. Consequently, measurements of all the person’s activity, including expenditure, will be assigned to period t, although part of it occurred in period t+1.
3.68. In order that tourism activity may be measured with precision during a given period of time, ideally, each of the four circumstances presented above should be identified separately and the visitors’ activities allocated proportionally (or otherwise) to the periods of time in which their stay and the period of reference overlap. Obviously, the applicability of this ideal method depends on whether entry and departure dates have been registered for each traveller and on how the information captured on E/D cards has been stored.
3.69. The implementation of this ideal method would entail a higher cost, as it requires systematic re‑estimations of tourism statistics. Countries should document which method they applied when counting travellers. If a country opts to count and survey inbound tourism at arrival, it should make clear that some of the information supplied is not to be taken as certain, e.g., by labelling the information as “expected length of stay” or “expected expenditure”. On the other hand, the country counts and surveys inbound visitors on departure, “actual length of stay” or “actual expenditure” would be measured, as would also be the case for outbound tourism.
3.70. In practice, however, the distinctions between types 1, 2 and 4 are often ignored, it is being assumed implicitly that the mismatches will eventually cancel each other out. In the cases concerned, however, short‑term analyses of arrival and expenditure figures could be distorted (see Box III.12). In publishing the data, it should be made clear that the numbers relate to “arrivals” during the reference period. The longer the period of reference and the shorter the duration of the trip, the smaller the problem of determining when to count a visitor.
3.71. The present section pays special attention to the assignment of tourism activity to the correct period of reference. Related to this issue are possible distortions in the estimated number of non‑resident international arrivals when the information about country of residence is collected upon departure of visitors and is then applied to the arrivals figures. Residents and non‑residents may have opposite behaviours that can lead to wrong conclusions. Similar problems can arise in specific periods if data from passengers entering a country and data from those living the country are combined, e.g., the former may come for a short visit in September (and both arrive and depart in that month) while the latter may leave after a long holiday period ending in September.
C.2.2. Statistical sources
3.72. In an increasing number of countries, the border control system is neither complete nor totally reliable. A clear example is that of the European countries that are parties to the Schengen Agreement, where controls are limited to the zone’s outside borders, with limited control not only over internal movements of residents of non‑Schengen countries (once they have entered the Schengen Area) but also over the internal movements of residents of Schengen countries onece they have returned to the Schengen Area. In these cases, it is necessary to look for other sources of information, whether of an administrative nature or, principally, statistical surveys, in order to measure the flow of non‑resident travellers arriving in the country. There are four classes of possible statistical sources:
- Border surveys
- Surveys of visitors (at market accommodation establishments or popular visitor sites)
- Household surveys in originating markets
- Other (e.g., administrative or, mirror data)
If different types of surveys are available, their questionnaires should be mutually consistent so that they can complement each other: definitions and classifications should be shared and questions formulated similarly.
C.2.2.1. Border surveys
3.73. It should be emphasized at the outset that, before border surveys are considered, the first priority should be to obtain data from administrative sources, particularly the passport control processes. Border surveys should be used to collect data supplementary to the available administrative data, for example, data on expenditure, on activities while in the country, on places visited, etc., as discussed in section C.1.3.1.
3.74. Countries carrying out such surveys usually draw their sample from a universe determined on the basis of administrative data from official sources, or data provided, e.g., by carriers (airlines, railway companies, shipping lines, scheduled bus lines) or though vehicle counts at road border crossings.
3.75. As mentioned above (see para. 3.65), these surveys are carried out upon the visitor’s departure from the country. Often, they also include some type of measurement of expenditure, based on a single question, but preferably on an expenditure module, i.e., a set of interconnected questions serving to elaborate on certain characteristics of the visitor’s behaviour (see chap. IV, sect. B.3.1).
3.76. The statistical design of border surveys must be such that the sampling and characterization of tourism trips and visitors can be considered representative of the entire population of international inbound trips. Designing such surveys requires an analysis of the country information obtained in phase one (see sect. C.1) in order to evaluate its suitability as the population framework of reference from which the sample is to be selected and data are to be collected and grossed up.
3.77. Once this information has been analysed, and depending on the country and the type of border crossing concerned, there will be one of two outcomes:
(a) Comprehensive and reliable data on arrivals of international travellers will be available
In such situations, the variables linking the information in the survey and the universe of reference will be the relevant characteristics measured in the arrival statistics (such as country of residence and/or nationality). These variables also need to be reliably recorded in the sample, being representative of those characteristics in the universe.
In this case, the information obtained from the survey will serve to improve the classification and characterization of traveller flows based on the records of international traveller arrivals.
It is frequently the case, however, that data are collected on nationality only, and not on country of residence. In some countries, border officials are instructed to ask travellers for their country of residence, but often, in practice, this information is not reliable or is not properly collected.
(b) The data available on arrivals of international travellers will not be totally reliable
The survey will need to be designed so as to provide the information required (i) for distinguishing, within the universe of inbound traveller flows (e.g., by railway or at land border crossings), those travellers who are nationals of the country (versus foreign nationals) and those who are residents (versus non‑residents), and (ii) for identifying the characteristics of non‑resident visitors and of their tourism trips.
3.78. The information yielded by the survey, together with the information on border flows, should make it possible to segment the population frame or statistical universe by tourism‑relevant variables.
3.79. The design of a statistical sample is usually extremely complex. There are often official and complementary data sources available for use in structuring the statistical universe or sample frame. The design must account for the particularities of trips: that travellers often travel in groups or parties, that they include children, and that questionnaire respondents need to be selected in such a way as to reflect those particularities. Expanding the data gathered to represent the entire universe of trips presents a real challenge to statisticians.
3.80. Further, through their access to administrative information on the arrival and departure flows of non‑resident travellers at a country’s ports of entry, statisticians can be provided with and advance indication of which flows are most representative at each point in the year, because of their volume or the greater heterogeneity of travellers. Their representative character can also change over time, as was the case for Iceland in 2010 when the ash plume from the Eyjafjallajökull volcanic eruption disrupted air travel in Northern Europe for several weeks, forcing visitors to Ireland to shift to smaller local airports--a change that the existing statistical design was slow to properly assimilate.
3.81. These designs require statistical expertise. In countries whose NTA lacks that expertise, it is necessary to secure the active involvement of the NSO in survey design and the establishment of minimum size and selection of the sample – including its distribution by type of port of entry, by type and characteristics of modes of transport used, by flight (charter/scheduled; low‑cost/conventional), time (day or night), day of the week (working day, holiday, weekend day), and period of the year, as well as other characteristics of the tourism trip or of the visitor considered relevant.
3.82. In determining sample size, consideration should be given to the minimum number of interviews required to ensure that the entire population of travellers arriving at each port of entry is represented, as well as to the distribution of travellers over time. Also, consideration should be given to the need for a sample size sufficient to allow for the production of cross‑tabulations between crucial variables (e.g., country of residence by purpose of visit) with a degree of statistical reliability.
3.83. The cooperation of the NSO should also extend, where possible, to subsequent processing of the data collected, namely, their entry, filtering, validation and tabulation, all processes commonly carried out during the undertaking of surveys and often requiring the expertise mainly or exclusively possessed by an NSO.
3.84. Unlike other statistically observed phenomena, where a given stratification and sampling characteristics might remain virtually stable over the years, travel flows can be volatile, in response to a variety of factors; hence, the initial sample design should be updated regularly.
3.85. UNWTO has developed a set of questions (see Annex I) that could serve as a starting point for the design of a questionnaire and should be adapted by each country’s own particular tourism reality and resource availability. Here are some examples of particularities that should be considered:
- In many countries, the number of inbound same‑day visitors or excursionists is not significant or, if it is, the cost of surveying such visitors would be disproportionate to the benefit of including them (in terms of their total expenditure)
- It might be reasonable to use a simplified version of the questionnaire at border crossings, where there is little time for interviews. More time is available at airports than at road crossings
- Questions related to the breakdown of transport could be broken down into further levels of detail, as indicated in IRTS 2008 (figure. 3.2)
- For questions on “Types of accommodation”, it is recommended that each country separate paid and market accommodation from non‑paid and non‑market accommodation and, on this basis, use the nationally established classification, that is, the relevant classification for analysing supply and expenditure
3.86. While the proposed set of questions focuses on information requested during a border survey, it could be also used as a reference for other types of surveys. It is divided into five blocks:Core module
- Module 1: Mode of transport
- Module 2: Accommodation
- Module 3: Activities while in the country
- Module 4: Expenditure
3.87. A few words of caution regarding the use of these questions:
- The set of question should be adapted to the particularities of tourism in each country and discussed with key stakeholders in the tourism sector as well as officials involved with the Balance‑of‑Payments and the National Accounts.
- Since border surveys are expensive, the entire process should be tested (including the tabulation of results) by means of a pilot exercise so as to ensure that the final version is as efficient as initially envisaged.
- The tables of expected results to be produced alongside the questionnaire should be designed and their content tested using the data collected through the pilot exercise, since the final objective of a questionnaire is to produce data that cross‑classify the main variables observed; the usefulness of the tables for tourism analysis should also be reviewed.
- Since border surveys are technically challenging, it should be established that the necessary resources and technical expertise have been committed to them and that their funding will be stable over time.
- Guidelines for field personnel should be drafted and a serious commitment made to training before the launching of the survey.
Implementing the survey
3.88. Procedural factors to bear in mind in implementing border surveys include the following:
(a) Survey vehicle used: while any method considered suitable may be used to collect information, the most desirable methods are the computer‑assisted interview (which generates highly reliable information, albeit at higher cost) and the printed questionnaire, to be completed by the respondents (which is less reliable, with a higher incidence of non‑responses, but also much less costly). All means of access may not be equally relevant for identifying inbound tourism (e.g., some small and out‑of‑the‑way land border posts may not be used much, if at all, by visitors); those considered less so could be excluded (at least in the first stages of observation, in line with the principle of gradualism);
(b) Recording and grouping possible answers:, questions regarding activities deployed during the stay for instance, should be adapted to local conditions;
(c) Subjects targeted: in the case of air travel, countries often use a multistage selection process, which starts with the selection of a flight from an entire flight programme. Some countries then select all passengers on board; others select only certain passengers (possibly based on the seats occupied). Information on the number of passengers on a flight provides the basis for grossing up the information and should be collected and stored for future review and follow‑up with the source. After the required passengers have been interviewed, the number of questionnaires can be compared with the number of passengers, making it easier to strictly control the country of nationality/country of residence of all passengers on the selected flight and gross up the findings to encompass the entire universe of flights. At other types of borders, as mentioned, a similar statistical design is desirable, in which all passengers in, e.g., a land vehicle or train carriage, are treated as clusters. The case of passengers who are in international transit but do not leave the international zone of an airport should be given appropriate consideration (e.g., are they significant enough in terms of numbers and expenditure, to count (see chapt 4, sect. C.4));
(d) Place of sampling: at airports, data collection on inbound visits, should occur, preferably prior to flight departure in the waiting area at the boarding gate. However, the increasing use of airline lounges for first class, business travellers and frequent fliers makes an increasing number of passengers, with specific characteristics, inaccessible for interviewers. A specific strategy needs to be designed to deal with this issue (e.g., conducting the interviews at the check‑in stage). At other types of border, particularly land border posts, the strategy needs to be carefully planned, because time is short, and traveller flows cannot be easily stopped. It might be necessary to look for alternatives, such as selecting respondents at toll booths or rest areas situated near the border post (although this approach could bias the procedure);
(e) Personnel conducting the sample: the sampling should be carried out preferably by bilingual personnel who are specially trained for this purpose by the NTA and, so as to ensure that the interviews are conducted as much as possible in the passengers’ first language, inasmuch as interviews conducted in another language often lead to misunderstandings and thus to inaccurate data;
(f) Treatment of non‑responses and outliers: non‑responses and outliers directly affect the quality of the information collected and impair that of the resulting estimates. They inevitably have a negative impact on the sample size initially established and create biases in the sample which are difficult to control. The distribution of non‑responses among the population of international travellers needs to be determined in order to define strategies for reducing their numbers and develop formulae for correcting the resulting biases;
(g) Additional considerations: various factors conducive to high response rates also need to be considered; including where the questionnaires are distributed or the interviews are conducted, the method used to collect the information, the survey team’s level of experience, the language used to address travellers, the subject’s country of residence (persons from some countries are more reluctant than others to participate in interviews) and legal requirements (whether survey responses should be compulsory). Finally, imputation techniques should be applied (that is, similar, reliable and complete data sets should be utilized) to supplement or substitute for missing or unreliable data.
3.89. For a comprehensive overview of border survey implementation, see UNWTO, “Tourism as an international traded service”, sect. 3.D.
C.2.2.2. Surveys of visitors (at market accommodation establishments or tourism sites)
3.90. Guest registers could form the basis for surveys at market accommodation establishments, and are, in fact used in much of Europe, where border surveys are less feasible. They do not, however, permit visitors to be distinguished from other travellers, a key objective in compiling inbound tourism statistics in countries where the difference is relevant. If used, those register should therefore be evaluated, with particular attention to:
- The updating mechanism for such registers (particularly in the case of repeat customers, whose particular characteristics hoteliers might fail to update) even if they have changed over the years
- Relevance of same‑day travellers
- Visitors who stay in the homes of friends or relatives, in private homes or in other forms of accommodation not officially registered as market establishments
3.91. Exogenous information needs to be used to correct for the underestimation and bias that results from considering only visitors who use such forms of accommodation. If the flow of visitors who are not staying at market accommodation establishments is considered relevant (as usually, it would be), their activity can be measured by including “tourism modules" in local household surveys (to estimate the number of resident households that have received that visits from non‑resident relatives and friends or rent rooms or apartments that they own to visitors).
3.92. Once the limitations mentioned directly above have been overcome, surveys of visitors staying at market accommodation establishments might be used to characterize visitors and their tourism trips doing so more precisely and possibly more completely than border surveys alone (see sect. C.2.2.1), as the time constraint might be less relevant.
3.93. Estimating expenditure at the time of the interview, before visitors have concluded their stays in the country, is more challenging, since they can report only what has happened up to that point in time. If information is also being collected on expenditure, there may be significant biases, since people often leave the purchase of souvenirs and other items to be taken back home to the very last minute before departure purchasing them, (e.g., in the duty‑free shops at the airport). Also, unexpected events, occurring either in the country of origin or in the country visited (such as natural disasters, bad weather or political turmoil) or personal reasons might oblige visitors to change their minds about anticipated expenditures.
3.94. An alternative to surveying travellers at market accommodation establishments might be to survey visitors at popular tourism sites. However, this kind of survey poses similar difficulties as surveys at accommodation establishments: a traveller might visit more than one site, while others might not visit any, particularly if the main purpose of the trip is other than recreation, e.g. visits to family and relatives or for business. It is not always feasible, moreover, to use statistically designed samples for such surveys, making the process of grossing up the findings to represent the entire population of inbound travellers a particular challenge, probably resulting in biased or misleading data. However, if implemented consistently over time, surveys at tourism sites can be useful in measuring changes or trends in visitors’ activities or characteristics over time and also to estimate average daily expenditure. Compilers should bear in mind that surveys at tourism sites may not be a reliable indicator of total visitor numbers.
C.2.2.3. Household surveys in originating markets
3.95. Surveys designed to collect information on the tourism behaviour of individuals in their usual environment, use the sum of residents as the population frame or statistical universe. Consequently, one way of estimating the number of arrivals and the expenditure of non‑resident travellers from country X in the country of reference (Y), and of identifying the characteristics of those travellers, could be to draw from the information reported by such travellers in response to household surveys in their country of residence X. If countries are willing to share this information (especially if, like most European countries, they have harmonized their surveys), then estimates can be made of the number of arrivals and expenditures by non‑resident visitors in country Y for all countries concerned.
C.2.2.4. Mirror statistics
3.96. Since most non‑resident visitors arriving in any European Union (EU) country come from other EU countries, Eurostat has repeatedly emphasized the usefulness of mirror statistics, which enable EU countries to estimate the number and characteristics of most inbound visitors by using outbound tourism data provided by other countries in the subregion. While mirror statistics, seem to be simple and attractive, the countries using them may not have paid sufficient attention to the challenges such statistics pose (see Box III.20 and Box III.21). For example, if a country uses mirror statistics to estimate numbers of inbound visitors from the numbers outbound visitors of several source countries, then there is a need for consistency in definitions (and, preferably, collection methodologies) among, those countries. Otherwise, it will not be possible to aggregate the data, or to compare the characteristics of visitors from one source country with those of another source country. Also, in cases where an outbound trip involves a number of destination countries, only the main destination country (however defined) may be recorded by the source country, which will result in the underestimation of the number of inbound visitors from those source countries in the countries that are only secondary destinations.