B.3.2. Visitor surveys at accommodation establishments or tourism sites
4.60. This type of survey offers a valuable means of complementing information obtained from border surveys (see chapt. III, sect. C.2.2.2). Such surveys can also be a useful source of information in cases where observing expenditure at the borders is not a viable option (e.g., when there are significant inbound flows by road and no practical means of stopping visitors near the border). They can be especially useful for obtaining “average daily expenditure”, for example, and can be an important complementary source of information when the goal is a more detailed examination of the characteristics of visitors and their expenditure.
4.61. However, visitor surveys also have a number of limitations (see paras. 3.90-3.91), which need to be carefully evaluated and accounted for as much as possible during the statistical design stage. A given visitor might visit more than one such tourism site or none at all, making his or her probability of selection variable or unknown. The case is the same for visitors surveyed at accommodation establishments, since some visitors might stay at more than one establishment or none at all (if e.g., they stay with family or friends or in their second home).
4.62. Estimating expenditure at the time of the interview, before most visitors have concluded their stay in the country, is even more challenging: they can report what has happened only up to that moment. If information is also being collected on expenditure, there may be significant biases, since people often leave the purchase of souvenirs or other items to take back home, to the very last minute before departure. Also, unexpected events, either in the country of origin or the country visited (such as natural disasters, bad weather or political turmoil), or personal factors might oblige visitors to change their mind about their stay and anticipated expenditures.
4.63. It may be more useful to ask for expenditure only on the previous day (including the value of accommodation for that day) if the interviews are spread over most, or all, of the days of the reference period. This would help minimize the bias associated with obtaining the value of total expenditure to date.
4.64. In addition, it has been observed not only that persons staying with family and friends spend less on accommodation, but also that the whole structure of their expenditure is significantly different from that of persons staying at market accommodation establishments (see Box IV.5): they tend to take meals at home, travel around with relatives, and engage in totally different types of activities during their stay in the country. Consequently, estimating their expenses based on those of persons staying in market accommodation would likely generate biases in both the level and structure of total inbound tourism expenditure.
 An interesting example of the construction on probabilistic of samples is given in Deville, J.C. and Maumy-Bertrand, M. (2006), ‘Extension of the Indirect Sampling Method and its Application to Tourism’, Survey Methodology, volume 32 (2), pp. 177-185 (online) available at: http://www.statcan.gc.ca/pub/12-001-x/2006002/article/9552-eng.pdf (23-03-2015).