A Clustering Method for Categorical Data in Tourism Market Segmentation Research

Abstract

One challenge in tourism market segmentation research is finding a statistical clustering method that can use data from the commonly used qualitative (categorical scale) survey instrument. Current proven methods require the use of quantitative (ratio or interval scale) data. However, quantitative survey instruments are seldom used. Many quantitative clustering methods severely restrict the number of attributes measured despite the fact that segmentation analysis works best when it measures all the multistate attributes that visitors identify as influencing their tourist experience. This study demonstrated that multistate categorical survey data could be successfully used. Using data from a bed-and-breakfast survey (229 guests), a two-stage analysis method was employed. First, multiple correspondence analysis was used to spatially map each of the attributes, and then cluster analysis was used to identify market segments. It is believed this method can be more practical in the field of applied tourism research.

An Extension of Multiple Correspondence Analysis for Identifying Heterogeneous Subgroups of Respondents

“…The use of categorical variables in surveys is becoming increasingly popular for various reasons (see Arimond & Elfessi, 2001;Dolničar & Leisch, 2001). In particular, they offer respondents faster and less tedious response formats in comparison to Likerttype items.…”

“…A convenient practice for accommodating cluster-level respondent heterogeneity in MCA is to adopt a two-step sequential, tandem approach (Arabie & Hubert, 1994): In the first step a low-dimensional representation of the categorical variables is obtained via MCA; in the second step some variety of cluster analysis is used to identify a set of relatively homogenous respondent groups on the basis of the low-dimensional data (e.g., Arimond & Elfessi, 2001;Green & Krieger, 1998;Green, Schaffer, & Patterson, 1988;Lebart, 1994). In addition to the ease in which the two-step sequential approach can be implemented, there can be substantive reasons for adopting this approach as well (see Green & Krieger, 1995).…”

Segmenting the Cruise Market: An Application of Multiple Correspondence Analysis

“…For this study five simple (yes/no/not sure) questions are used in the analysis, but the twostep method is also valuable for segmentation of various types of categorical data (Arimond & elfessi, 2001;Gonzales & Molina, 2009). Likert-type scales can often be tedious for respondents, and surveys used in many market surveys by tourism offices or businesses are often qualitative in nature.…”

“…MCA uses alternating least squares with optimal scaling for estimation (Arimond & elfessi, 2001). MCA quantifies the relationship between categories of each variable as well as between variables, and plots respondents who choose the same categories close to each other.…”

“…this study contributes to the understanding of potential repeat visitors by segmenting cruise tourists based upon their intent to cruise in the future. this study uses a two-step method, Multiple Correspondence Analysis (MCA) combined with a k-means cluster analysis, (Arimond & elfessi, 2001;Pina & Delfa, 2005), to segment the sample based on responses to five questions about future cruising intentions. the purpose of this research note is to reflect on the segmentation method applied.…”