Applying weights to correct distortions in a non-random sample: An application to Chilean tourism time series data

Statistics that accurately represent the evolution of tourism activity are needed for tourism research and planning. Monthly accommodation statistics provided by the Chilean Statistical Institute show a negative trend despite considerable growth in the number of foreign visitors. This paper proposes a methodology to create sample weights to overcome statistics inaccuracy caused, as in the case of Chilean tourism statistics, by the use of a non-random sample with serious attrition bias. The benefits of the methodology are discussed, highlighting their relevance to a broader range of applications where tourism statistics are questionable or demonstrably inaccurate.