Índices preço-lucro baseados no Ibovespa para análises do mercado de ações brasileiro

• Ano de defesa: 2020
• Autor(a) principal: Daniel Penido de Lima Amorim
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; FACE - FACULDADE DE CIENCIAS ECONOMICAS; Programa de Pós-Graduação em Administração; UFMG
• Palavras-chave em Português: Finanças Comportamentais; Índice preço-lucro; Reversão à média; Sobrevalorização; Modelo Fed; Behavioral finance; Price-earnings ratio; Mean reversion; Overvaluation; Fed model; Índices de mercado de ações; Bolsa de valores – Brasil; Finanças – Brasil
• Link de acesso: http://hdl.handle.net/1843/39434; https://orcid.org/0000-0002-2844-3079

Shiller (1989, 2005) provided two price-earnings ratios (P/E) for the U.S. stock market: the P/E1, calculated by the ratio of a stock price index to a corporate earnings index contemporaneously; and the P/E10, also called Cyclically Adjusted Price Earnings Ratio (CAPE), which has as a divisor a ten-year moving average of the earnings index. However, indicators such as these were not found in the context of the Brazilian stock market. This study presents a method for the construction of market P/E ratios based on the Ibovespa, which is adapted to the particularities of the Brazilian stock market. In addition to constructing the P/E ratios, it aimed i) to test the mean reversion in P/E1, ii) to identify moments of overvaluation and undervaluation in the stock market based on P/E1 and iii) to analyze the long-run relationships among both P/E1 and P/E10 and the interest rates corresponding to the returns of treasury bonds. In general, the period considered in this study was from December 2004 to June 2018. Through the unit root tests, it was evidenced that P/E1 has a stationary time series and, therefore, presents mean reversion, which contradicts the Efficient Markets Hypothesis. The non-linearity of P/E1 was evidenced in the Enders-Lee unit root test. Thus, based on regime probabilities recovered from a Markov Switching model estimated with the P/E1, five moments of undervalued stock market and four moments of overvalued stock market were identified. Using Autoregressive Distributed Lags Models (ARDL), which can be represented as conditional Error Correction Models (ECM), it was evidenced significant long-run relationships among both P/E1 and P/E10 and the referred interest rates, indicating that the Fed Model is effective in the context of the Brazilian financial market.