Segmentation by Tourist Expenditure Composition: An Approach with Compositional Data Analysis and Latent Classes

Abstract

In this article, we propose segmenting tourists according to trip budget share (in other words, the proportions of total expenditure devoted to different items, such as transportation, accommodation, and activities undertaken). We put forward a method involving a combination of compositional data analysis (CODA) and mixture models (latent class models), which, by contrast with traditional methods (e.g., cluster analysis), makes it possible to carry out statistical inference with regard to segment characteristics and their relationships with external variables, such as tourist background. The way we use the CODA methodology is intuitively appealing and boils down to building transformed variables that can be tailored to the specific research questions. We analyze the spending profiles of tourists arriving in Spain on low-cost carriers, with relevant and interpretable results.