Sampling by placeFrontier surveysTh

Sampling by place
Frontier surveys
These are undertaken at points of entry and exit, normally by personal interview after the traveller has passed through border controls. Given that locations are known, the sample is tratified by airports, seaports and land routes, though, as noted earlier, for landlocked countries with open border frontier survey have not proved to be cost-efficient ways of collecting visitor information. By stratifying the sample according to transport mode of entry, the design effect is to reduce sampling error. However it is common not to sample all points of entry and departure at any one time. While countries need to include their main airports, seaports and land routes (if possible) in the sample with certainty, the remainder are often grouped into clusters from which a few will be randomly drawn for sampling. The assumption is that one or two regional airports, for example, will be representative of similar facilities. The design effect of clustering is to increase the sampling error, so it will be appreciated that calculating confidence limits for such large-scale surveys is complex. Once the sample is in the approciate weighting, factors are applied to yield gross totals. These can then be checked against the recorded passenger statistics of transport carriers for accuracy.