Estudo da sinistralidade no mercado securitário de veículos: uma abordagem multivariada

• Ano de defesa: 2014
• Autor(a) principal: Costa, Priscila Amorim da [UNESP]
• Orientador(a): Salgado, Manoel Henrique [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Apolices de seguro; Seguro de acidentes; Automoveis; Administração de risco; Insurance policies
• Link de acesso: http://hdl.handle.net/11449/122190

The insurer that has an effective control of the risks involved by the insureds can avoid issues such as bankruptcy and loss of profitability. The purpose of the research was to identify the varieties that define the risk associated with future buyers, enabling classify them into one of two groups, the claimed and the unclaimed, based in probabilities defined by a multivariate model. Therefore, were questions considered (variables) existing on the risk evaluation of the insurers questionnaire and others nominated by the expert brokers of the insurers. To achieve the objective, was used in the analyses, the multivariate statistical technique, known as Discriminant Analysis, in order to segregate the individual into one of two groups. A discriminant function was constructed from the independent variables, associated to the risks, and from the dependent variable, which covers the two groups. Other results as estimation of the classification rule, evaluation of the quality of the discrimination rule settings, estimation of the overall probability of correct answers and tests related to the assumptions of discriminant analysis were presented. The studied sample was consisted of 2,000 insured served by the broker, divided into two groups: the first composed of individuals without claims; and the second with those who has one or more claims. The model enabled to classify individuals by the two groups, wherein in the development sample and test the classification represented 69% accuracy. The separation was carried out by variables with higher importance degrees. Where elected: car power, time of insurance, bonus, parking lot use, and professional activities defined by the insured occupation. Such variables can be considered highly discriminating, based on the contribution coefficient for the discrimination. This work has presented results that contradict the common sense in the insurance market, in which technicians say that sex and age determine whether