The Fairness of Credit Scoring Models
نویسندگان
چکیده
In credit markets, screening algorithms discriminate between good-type and bad-type borrowers. This is their raison d’etre. However, by doing so, they also often individuals sharing a protected attribute (e.g. gender, age, race) the rest of population. this paper, we show how to test (1) whether there exists statistical significant difference in terms rejection rates or interest rates, called lack fairness, unprotected groups (2) only due worthiness. When condition not met, algorithm does comply with fair-lending principle can be qualified as illegal. Our framework provides guidance on algorithmic fairness monitored lenders, controlled regulators, improved for benefit groups.
منابع مشابه
Credit Risk Scoring Models
Credit scoring models play a fundamental role in the risk management practice at most banks. They are used to quantify credit risk at counterparty or transaction level in the different phases of the credit cycle (e.g. application, behavioural, collection models). The credit score empowers users to make quick decisions or even to automate decisions and this is extremely desirable when banks are ...
متن کاملNeural network credit scoring models
This paper investigates the credit scoring accuracy of "ve neural network models: multilayer perceptron, mixture-of-experts, radial basis function, learning vector quantization, and fuzzy adaptive resonance. The neural network credit scoring models are tested using 10-fold crossvalidation with two real world data sets. Results are benchmarked against more traditional methods under consideration...
متن کاملSample selection in credit-scoring models
We examine three models for sample selection that are relevant for modeling credit scoring by commercial banks. A binary choice model is used to examine the decision of whether or not to extend credit. The selectivity aspect enters because such models are based on samples of individuals to whom credit has already been given. A regression model with sample selection is suggested for predicting e...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Social Science Research Network
سال: 2021
ISSN: ['1556-5068']
DOI: https://doi.org/10.2139/ssrn.3785882