Sequeira, FernandoCosta, Gonçalo Ribeiro Sá da2024-05-102024-05-1020242022http://hdl.handle.net/10451/64712Trabalho de Projeto de Mestrado, Matemática Aplicada à Economia e Gestão, 2024, Universidade de Lisboa, Faculdade de CiênciasEste projeto surge no âmbito da cadeira de projeto do mestrado de Matemática Aplicada à Economia e Gestão da Faculdade de Ciências da Universidade de Lisboa. Com o estudo aqui apresentado pretende-se obter boas projeções sobre a composição do parlamento ao longo da contagem de votos numa noite de eleições legislativas. Este estudo pode ser aplicado quando ainda só estão contabilizados um pequeno número de votos em cada círculo eleitoral. Procuro captar o padrão de variabilidade nas percentagens de votos nos partidos entre o momento em que se apura 50% do total de votos e o fecho das contagens, em função da dimensão do partido. A variabilidade na percentagem de votos foi modelada por regressão linear e foram criados cenários para a distribuição dos votos finais em cada partido. A aplicação do método de D´Hondt às projeções dos votos obtidos nos diversos cenários mostrou que este devolve uma distribuição de mandatos por partido em cada círculo eleitoral que se mantém bastante estável entre o momento mediano e o momento em que termina a contagem dos votos, o que permite antecipar para pouco depois do fecho das urnas a projeção da configuração do parlamento. Para este trabalho foram analisados os resultados eleitorais das eleições legislativas de 2019. Analisou-se os resultados das 3092 freguesias que compõem os 20 círculos eleitorais do continente e ilhas. Como a contagem de votos dos círculos da Europa e do Resto do Mundo é só efetuada em momento posterior à noite eleitoral, estes não foram explorados.This project arises within the scope of the conclusion of the MSc of Mathematics Applied to Economics and Management at the Faculty of Sciences of the University of Lisbon. The goal of this study is to obtain a good perspective on parliament composition during the vote count legislative election night. This study can be applied when only a small number of votes are scrutinized in each constituency. I try to capture the pattern of variability in the percentages of votes in parties between the moment when 50% of the total votes are counted and the count ends, depending on the size of the party. The variability in the percentage of votes was modeled by linear regression and scenarios were created for the distribution of final votes in each party. The application of the D'Hondt method to the projections of votes obtained in the different scenarios showed that it returns a distribution of mandates by party in each electoral district that remains quite stable between the median moment and the moment when the vote count ends, which allows the projection of the parliament's configuration to be brought forward to shortly after the polls close. For this work, the electoral results of the 2019 legislative elections were analyzed. The results of the 3092 parishes that make up the 20 constituencies of the mainland and islands were analyzed. As the counting of votes from the constituencies in Europe and the Rest of the World is only carried out after the election night, these data were not explored.porPrevisão eleitoralMétodo D'HondtRegressão Linear SimplesReta Resistente de 3 GruposTeses de mestrado - 2024Previsões eleitoraismaster thesis203883241