Gaspar, RaquelCosta, António da AscensãoVaz, Sónia Melania Oliveira2015-12-012015-12-012012Vaz, Sónia Melania Oliveira (2012). "How efficient is the Portuguese Stock Market?". Dissertação de Mestrado, Universidade Técnica de Lisboa. Instituto Superior de Economia e Gestão.http://hdl.handle.net/10400.5/10329Mestrado em FinançasEsta Dissertação testa a hipótese de eficiência fraca do mercado aplicada a seis índices de mercado Europeus (França, Alemanha, Reino Unido, Grécia, Portugal e Espanha) no período de Janeiro de 2007 a Janeiro de 2012. Para tal, testámos as correlações, realizámos o teste runs, o teste de raízes unitárias bem como teste de variâncias (variance ratio test). Adicionalmente analisámos se seria possível prever os retornos do PSI-20 recorrendo a data mining e, mais concretamente, aos algoritmos de data mining: k-NN e Redes Neuronais. Os nossos resultados evidenciam que no período de Janeiro de 2007 a Setembro de 2008 os índices de referência de França, Alemanha e Espanha, cumpriram a maioria dos critérios referentes à hipótese de eficiência fraca de mercado. Os nossos resultados evidenciam ainda que esta situação se verifica depois para os índices dos seis países considerados, no período de Setembro de 2008 a Janeiro de 2012. Relativamente à previsão dos retornos do PSI-20, desenhámos uma estratégia baseada nas previsões dadas pelo k-NN e Redes Neuronais e concluímos que, ao implementá-la obteríamos retornos consideravelmente elevados face aos alcançados através de uma simples estratégia de buy-and-hold, comprometendo assim a hipótese de eficiência fraca de mercado.This dissertation reports the results of tests on the weak-form market efficiency applied to six European market indexes (France, Germany, UK, Greece, Portugal and Spain) from January 2007 to January 2012. For this matter we use a serial correlation test, a runs test, an augmented Dickey-Fuller test and the multiple variance ratio test. In addition we also analyze if it would be possible to forecast the PSI-20 returns resorting data mining, more specifically using k-NN and Neural Network. Our findings show that from January 1997 to September 2008 France, Germany and Spain meet most of the criteria for the weak-form market efficiency hypothesis, a situation that occurs afterwards for all six European market indexes from September 2008 to January 2012. Regarding the forecast of PSI-20 returns we designed a strategy based on the forecast of k-NN and Neural Network and concluded that by implementing it we would obtain relevant higher returns than the ones achieved by a buy-and-hold strategy, which compromises the weak-form market efficiency.engEficiência de mercadoteste de runsADFteste de variânciask-NNredes neuronaisMarket efficiencyruns testvariance ratio testneural networkHow efficient is the Portuguese Stock Market?master thesis