A carregar...
Publicação
Modelo de previsão de comissões de crédito para particulares
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 525.29 KB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Dado que as comissões bancárias são uma das principais fontes de rentabilidade de qualquer instituição financeira em Portugal, a tentativa de previsão destas componentes acaba por se tornar fulcral no desenvolvimento de campanhas de promoção a certos produtos de crédito. Tendo em conta a estrutura da carteira de créditos concedidos da Caixa Geral de Depósitos, uma vez que é a instituição financeira com mais crédito concedido a particulares em especial o crédito habitação, o estudo da evolução das comissões de créditos concedidos a particulares acaba por ser um trabalho de enorme relevância para o sistema financeiro português, dada a grande representatividade da amostra em questão.
Inicialmente neste trabalho, são realizados testes de estacionariedade das variáveis a serem estudadas, bem como testes de cointegração e de causalidade sendo estes requisitos para a aplicação dos modelos econométricos de séries temporais em estudo.
Depois de preparadas as variáveis, iniciou-se a modelação das comissões usando os métodos de Alisamento Exponencial e os modelos ARIMA, ARIMAX, VAR e VEC e foram comparados os erros de previsão de cada um deles para assim concluir qual o modelo que melhor se ajusta à comissão em questão. Adicionalmente foram realizados testes de avaliação do diagnóstico a cada um dos modelos estimados como ausência de autocorrelação dos resíduos, normalidade dos resíduos, multicolinariedade e significância dos parâmetros. Tendo em conta os resultados obtidos foram escolhidos os modelos que apresentavam os melhores resultados de indicadores de performance e que passavam na maioria dos testes realizados.
Os resultados obtidos apontaram para um melhor ajustamento da maioria das comissões nos modelos ARIMA e ARIMAX, no entanto, existiram duas comissões onde os métodos de Alisamento Exponencial e o modelo VAR foram os que apresentaram melhores resultados. Por fim, foi realizada a previsão out-sample que verificou um crescimento mensal em algumas das comissões estudadas e a manutenção da média dos valores registados nos dois anos seguintes à amostra disponível.
Given that bank commissions are one of the main sources of profitability for any financial institution in Portugal, trying to predict these components becomes crucial when developing campaigns to promote certain credit products. Bearing in mind the structure of Caixa Geral de Depósitos' loan portfolio, since it is the financial institution with the largest number of loans to private individuals, especially housing loans, studying the evolution of commissions on loans to private individuals turns out to be an extremely important study for the Portuguese financial system, given the large representativeness of the sample in question. Initially in this work, stationarity tests are carried out on the variables to be studied, as well as cointegration and causality tests, which are requirements for the application of the time series econometric models under study. Once the variables had been prepared, the commissions were modelled using the Exponential Smoothing methods and the ARIMA, ARIMAX, VAR and VEC models, and the forecast errors of each were compared to conclude which model best fits the commission in question. In addition, diagnostic tests were carried out on each of the estimated models, such as the absence of autocorrelation of the residuals, normality of the residuals, multicollinearity, and significance of the parameters. Considering the results obtained, the models with the best performance indicator results and which passed most of the tests were chosen. The results showed that the ARIMA and ARIMAX models were the best fit for most of the commissions, but there were two commissions where the Exponential Smoothing method and the VAR model were the best fit. Finally, an out-sample forecast was carried out which found monthly growth in some of the commissions studied and the average values recorded in the two years following the sample available were maintained.
Given that bank commissions are one of the main sources of profitability for any financial institution in Portugal, trying to predict these components becomes crucial when developing campaigns to promote certain credit products. Bearing in mind the structure of Caixa Geral de Depósitos' loan portfolio, since it is the financial institution with the largest number of loans to private individuals, especially housing loans, studying the evolution of commissions on loans to private individuals turns out to be an extremely important study for the Portuguese financial system, given the large representativeness of the sample in question. Initially in this work, stationarity tests are carried out on the variables to be studied, as well as cointegration and causality tests, which are requirements for the application of the time series econometric models under study. Once the variables had been prepared, the commissions were modelled using the Exponential Smoothing methods and the ARIMA, ARIMAX, VAR and VEC models, and the forecast errors of each were compared to conclude which model best fits the commission in question. In addition, diagnostic tests were carried out on each of the estimated models, such as the absence of autocorrelation of the residuals, normality of the residuals, multicollinearity, and significance of the parameters. Considering the results obtained, the models with the best performance indicator results and which passed most of the tests were chosen. The results showed that the ARIMA and ARIMAX models were the best fit for most of the commissions, but there were two commissions where the Exponential Smoothing method and the VAR model were the best fit. Finally, an out-sample forecast was carried out which found monthly growth in some of the commissions studied and the average values recorded in the two years following the sample available were maintained.
Descrição
Mestrado Bolonha em Métodos Quantitativos para a Decisão Económica e Empresarial
Palavras-chave
Comissões de crédito ARIMA ARIMAX VAR VEC Alisamento Exponencial. Credit commissions, Exponential Smoothing.
Contexto Educativo
Citação
Ferreira, Miguel Ângelo Lopes (2023). “Modelo de previsão de comissões de crédito para particulares”. Dissertação de Mestrado. Universidade de Lisboa. Instituto Superior de Economia e Gestão
Editora
Instituto Superior de Economia e Gestão