CHARACTERISTICS OF JEWELRY PURCHASING HOUSEHOLDS: A MACHINE LEARNING PERSPECTIVE

Abstract

With an ever-growing market, sized over $300 billion annually, jewelry consumption has a significant place in luxury consumption. Jewelry marketing and consumption are usually studied from a causal perspective, as the most frequent question can be defined as “why do people buy jewelry?” in literature. This study offers a different perspective as the research question of this paper is “what are the characteristics of the households who purchase jewelry in Turkey?”. Three years of household spending data was collected from the Turkish Statistical Institute (TÜİK) and inflation adjusted. Then data is classified using cluster analysis to detect major groups in data, and then the CHAID algorithm is used for conducting decision trees. To the best of the authors’ knowledge, the methodology is unique in jewelry marketing literature. Research findings show that lower and middle-income households show a tree form, and sub-branches are found and the most and the least likely household types are identified: lower and medium-income households with younger and higher educated house heads are more likely to purchase jewelry, and the least likely groups are those with house heads who are less educated and aged over 60, whereas higher-income households look homogenous in terms of jewelry buying behavior.