[pt] Esta dissertação investiga a presença de informação assimétrica no mercado de planos de saúde do Brasil e os principais efeitos gerados por essa assimetria de informação, conhecidos como seleção adversa e risco moral (moral hazard). A partir de dados do Suplemento Saúde da PNAD 98, são feitas análises descritivas e realizadas inferências estatísticas para avaliação tanto do risco moral quanto da seleção adversa. Na busca por um melhor entendimento do problema de seleção adversa estima-se também um modelo logístico para investigação da relação entre a condição de uma pessoa ter plano de saúde e variáveis explicativas tais como, renda, idade e auto- avaliação do estado de saúde, dentre outras. Em uma etapa subseqüente, são realizadas inferências sobre o risco moral por intermédio de dois procedimentos distintos: no primeiro, a partir da construção de um indicador de risco moral, denominado IRM, analisa-se a diferença entre o número de consultas médicas ambulatoriais realizadas por indivíduos com e sem plano de saúde levando-se em consideração o plano amostral da PNAD 98; o segundo, consiste num procedimento que tem por objetivo verificar a relação entre o fato de uma pessoa ter plano de saúde e o número de consultas médicas realizadas, por intermédio da estimação de um modelo binomial negativo com barreira (hurdle negative binomial model). / [en] The aim of this dissertation is to investigate the presence of asymmetric information in the Brazilian health insurance market and the effects that arises due to this asymmetric information: adverse selection and moral hazard. Using the data set of the Health Supplement of the Brazilian Household Sample Survey (Pesquisa Nacional por Amostra de Domicílios - PNAD) for the year 1998, descriptive analysis are produced and statistical inferences are realized to evaluate the moral hazard and adverse selection. In order to have a better perception of the adverse selection problem it is estimated a logistic model to evaluate the relation among the condition of having a health plan and independent variables such as income, age, and health self- perception. In a next step inferences related to moral hazard are realized by two different procedures: the first one, based on the construction of a moral hazard indicator, named MHI, analyses the differences between the number of physician visits for individuals who have and who have not a health plan, considering the sample design of the Health Supplement of the Brazilian Household Sample Survey. The second procedure aims to verify the relation between having a health plan and the number of physician visits by estimating a hurdle negative binomial model, which allows the elimination of some bias in the average number of physician visits.

Submitted by Ricardo Barros de Aquino Fonseca (ricardobafonseca@gmail.com) on 2017-06-28T06:01:34Z No. of bitstreams: 1 Ricardo_Fonseca_Dissertacao.pdf: 1923906 bytes, checksum: 31859d5e2f131b25276d0fc435835757 (MD5) / Approved for entry into archive by GILSON ROCHA MIRANDA (gilson.miranda@fgv.br) on 2017-06-30T12:38:04Z (GMT) No. of bitstreams: 1 Ricardo_Fonseca_Dissertacao.pdf: 1923906 bytes, checksum: 31859d5e2f131b25276d0fc435835757 (MD5) / Made available in DSpace on 2017-07-04T18:13:40Z (GMT). No. of bitstreams: 1 Ricardo_Fonseca_Dissertacao.pdf: 1923906 bytes, checksum: 31859d5e2f131b25276d0fc435835757 (MD5) Previous issue date: 2017-03-28 / I test for the existence of moral hazard and adverse selection in the Brazilian health insurance market. I use the positive correlation test, proposed by Chiappori and Salanié (2000), to assess the existence of asymmetric information in this market. I estimate the effect of health plan on treatment usage, controlling for many variables (measured blood pressure, age, smoker, etc.), identifying the effect of the plan. I use this coefficients as plan characteristics and estimate their effects on plan purchase using a simple logit regression. The coefficients from this estimation show the relative importance of anticipated moral hazard of different treatments on health plan choice. I find evidence of asymmetric information in the market for every state, in either the Metropolitan Area or not, and for the country as a whole. When performing the positive correlation test using hospitalizations as the usage variable, where moral hazard is assumed to be controlled, I find this evidence only for young women and for the last age group (59 years or more). I find no evidence of asymmetric information for plans paid through work, as expected. Several treatments are used more by people with plans, even when controlling for adverse selection. The analysis suggests that some of those are relevant for plan purchase.