ABSTRACT
The purpose of this study was to compare the conceptual and methodological features of five classification algorithms used to produce socioeconomic stratifications of the Brazilian society, and to measure misclassification tradeoffs between them. We applied these five algorithms to classify the 55,970 households covered in the survey of household budgets (Pesquisa de Orçamentos Familiares [POF]) conducted by Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística [IBGE]). Our results allow us to conclude that maximum-likelihood classifier had the best performance in explaining Brazilian families' consumption patterns across the socioeconomic strata, followed by the adaptive Bayesian, simplified ABEP, old ABEP, and SAE algorithms. The first three classifiers rely on the concept of permanent income, while the first two incorporate an important innovation: classifying a household by considering its geographic location and family composition, the latter defined by the number of adults and minors living in the household. These new classifiers allow researchers and marketers to segment and study markets based on a valid, unbiased, and reliable criterion of socioeconomic stratification.
KEYWORDS:
Market segmentation; socioeconomic class; socioeconomic status; socioeconomic classifier; emerging markets
