40 COMPOSITIONAL ANALYSIS OF TOURISM-RELATED DATA – Contributions by Berta Ferre
40 COMPOSITIONAL ANALYSIS OF TOURISM-RELATED DATA – Contributions by Berta Ferrer-Rosell
Approaching Compositional Data Analysis in Tourism
Compositional data analysis is the appropriate methodology to employ when dealing with data carrying relative information. Compositional data (CoDa) can be defined as arrays of positive numbers – called components, or parts – whose relative size is of interest to the researchers. In some cases, the components are parts of a whole and their sum is irrelevant or even constant. This is the case in geological and chemical analyses, which use proportions adding up to 1, or in time-use studies, where the sum is up to 24 hours. These two fields of application are the first and most traditional fields using CoDa (Aitchison, 1986).
It is taken for granted that the total time of a day is an uninteresting fact. A common time-use research interest is to know how the distribution of daily time of an individual – the time allocated to commuting, to work, to family and household tasks, to sleeping – may affect one’s health, quality of life and life satisfaction. In other cases, components do not constitute any whole or do not have a constant sum, and the only key issue is that the researchers’ interest and questions lie in the relative importance of components to one another (Egozcue & Pawlowsky-Glahn, 2019). This might be the case of the content uploaded on a tourism product, company or destination website or on a printed brochure. Larger websites or wider brochures may have more content of all types. There are, however, other cases in which both the relative importance and the total volume are of interest. This is the case with tourist expenditure. Destinations and tourism companies may like to know how travellers allocate their trip budget into different expenses – transportation, accommodation and food, activities, and others – as well as how much tourists spend during the trip as a whole.
As can be deduced, many research questions involving tourism-related data to analyse consumer (or company or destination) behaviour, are related to distribution of a whole (e.g., share or allocation) or to relative importance (e.g., prevalence, concentration, dominance). Possible questions tourism researchers may be interested in are, for instance, how does the relative popularity of search terms in Google relate to tourism market share? How do hospitality firms allocate their capacity to their product portfolio? How does time allocated to different types of activities at the destination relate to tourist satisfaction?
CoDa has started to be used in several fields of social science which often face similar research questions, such as education (Batista-Foguet, Ferrer-Rosell, Serlavós, Coenders & Boyatzis, 2015), finance (Carreras-Simó & Coenders, 2020; Linares-Mustarós, Coenders & Vives-Mestres, 2018), marketing (Morais, Thomas-Agnan & Simioni, 2017; Vives-Mestres, Martín‐Fernández & Kenett, 2016), sociology (Hlebec, Kogovšek & Coende