| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 962.7 KB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Este estudo tem como objetivo modelar a probabilidade de incumprimento de uma carteira de crédito americana. Para tal, foi utilizado o modelo logit para apurar a probabilidade de incumprimento refêrencia para aceitação de crédito que supõe maior rendimento à Instituição Financeira. A Base de Dados recolhe informação de 338.909 empréstimos terminados, concedidos a particulares, destinados ao crédito ao consumo, e 139 variáveis, posteriormente reduzidas de acordo com o seu nível de significância, apurado nos testes estatítsticos realizados com o programa SPSS.
Pelo principio da parsimónia foram considerados dois modelos e o modelo escolhido foi o que apresenta comparativamente maior rendimento à Instituição Financiera.
This study intends to model the Probability of Default (PD) of an American credit database. The logit model was used in order to determine the PD threshold that presents higher profits. The database contemplated 338.909 terminated loans lent to private individuals, consumer oriented, and 139 variables. From the rough data we reduced the number of variables to the most significant, according to the statistical tests computed in SPSS. By the principle of parsimony two models were considered and the chosen model was the one that presented higher profits for the Financial Institution.
This study intends to model the Probability of Default (PD) of an American credit database. The logit model was used in order to determine the PD threshold that presents higher profits. The database contemplated 338.909 terminated loans lent to private individuals, consumer oriented, and 139 variables. From the rough data we reduced the number of variables to the most significant, according to the statistical tests computed in SPSS. By the principle of parsimony two models were considered and the chosen model was the one that presented higher profits for the Financial Institution.
Description
Mestrado em Finanças
Keywords
Risco de Crédito Modelo Linear Generalizado Logit Scorecard Discriminação Calibração Probabilidade de Incumprimento Gestão do Risco de Crédito Credit Risk Generalized Linear Models Discriminant Analysis Calibration Probability of Default Credit Risk Management
Pedagogical Context
Citation
Argomaniz, Laura de Lacerda (2019). "On credit score models". Dissertação de Mestrado. Universidade de Lisboa. Instituto Superior de Economia e Gestão.
Publisher
Instituto Superior de Economia e Gestão
