Análise comportamental de consumidores brasileiros: fatos estilizados por estratificação social e aplicações em modelos de projeção macro

• Ano de defesa: 2016
• Autor(a) principal: Velho, Vitor Vidal Costa
• Orientador(a): Matos, Silvia Maria
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Confiança; Indicadores desagregados; Incerteza; Componentes principais
• Palavras-chave em Inglês: Confidence index; Disaggregated indicators; Uncertainty; Components analisys
• Link de acesso: https://hdl.handle.net/10438/17834

Survey indicators produced by FGV are economic information published relatively quickly and used as a 'thermometer' of the Brazilian level of activity in the short term. The use of the Consumer Confidence Index (CCI), as well as other indicators within the survey, are taken as a antecedent or coincident in household consumption forecasting models and other official quantitative variables. However, the work seeks to show evidence that information at the aggregate level has not always the best predictive power. The work seeks to analyze the behavior of consumers under a more disaggregated view of the research in order to obtain greater correlation and robustness with the target variables. There is evidence that some consumer groups are able to provide better assessments on one topic than others. Using the Principal Component Analysis (PCA), we reduce the dimensionality of these better indicators to obtain robustness in the forecasting scenarios. To complement these predictions, we will use the indicator of dispersion of responses as an uncertainty proxy, in an attempt to understand subjectivity of survey data. The conclusion of the article is given through the analysis of the stylized facts of the behavior of these groups selected for each case.