Reject inference in consumer credit scoring with nonignorable missing data

Bücker, Michael; Kampen, Maarten van; Krämer, Walter

doi:10.1016/j.jbankfin.2012.11.002

Cited by 35 publications

(15 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can accommodate binary or ordered dependent variables and generate the probability of default under the Basel Accord. Hence, Bücker et al (2013) used an improved logit model to enhance the performance of consumer credit scoring. Zeng and Zhao (2014) presented a twophase augmentation model integrated with weighted logistic regression.…”

Section: Methodsmentioning

confidence: 99%

“…Hence, Bücker et al . () used an improved logit model to enhance the performance of consumer credit scoring. Zeng and Zhao () presented a two‐phase augmentation model integrated with weighted logistic regression.…”

Section: Literature Reviewmentioning

confidence: 99%

“…For obvious reasons, the behaviours of the rejected applicants are not observed, which is regarded as a missing information problem in modelling. Many organizations such as credit bureaus, commercial banks, lending companies and other financial houses encounter this in their modelling processes (Bücker et al, 2013). The missing data may lead to sample bias, which means that the credit models are not actually built on the through-the-door population (Hand and Adams, 2014).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Inferring the Outcomes of Rejected Loans: An Application of Semisupervised Clustering

et al. 2019

Journal of the Royal Statistical Society Series A: Statistics in Society

View full text Add to dashboard Cite

Summary Rejection inference aims to reduce sample bias and to improve model performance in credit scoring. We propose a semisupervised clustering approach as a new rejection inference technique. K‐prototype clustering can deal with mixed types of numeric and categorical characteristics, which are common in consumer credit data. We identify homogeneous acceptances and rejections and assign labels to part of the rejections according to the label of acceptances. We test the performance of various rejection inference methods in logit, support vector machine and random‐forests models based on data sets of real consumer loans. The predictions of clustering rejection inference show advantages over other traditional rejection inference methods. Inferring the label of the rejection from semisupervised clustering is found to help to mitigate the sample bias problem and to improve the predictive accuracy.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Literature Reviewmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Inferring the Outcomes of Rejected Loans: An Application of Semisupervised Clustering

et al. 2019

Journal of the Royal Statistical Society Series A: Statistics in Society

View full text Add to dashboard Cite

show abstract

“…For example if a given application has a probability of being rejected of 0.80, then all similar applications would be weighted up 1/(1 − 0.8) = 5 times [2]. None of the empirical research using augmentation shows significant improvements in either correcting the selection bias or improving model performance, see [2,4,5,7,10,14,49]. The augmentation technique assumes that the default probability is independent of whether the loan is accepted or rejected [3].…”

Section: Related Workmentioning

confidence: 99%

Deep generative models for reject inference in credit scoring

Mancisidor

Kampffmeyer

Aas

et al. 2020

Knowledge-Based Systems

View full text Add to dashboard Cite

Credit scoring models based on accepted applications may be biased and their consequences can have a statistical and economic impact. Reject inference is the process of attempting to infer the creditworthiness status of the rejected applications. In this research, we use deep generative models to develop two new semi-supervised Bayesian models for reject inference in credit scoring, in which we model the data generating process to be dependent on a Gaussian mixture. The goal is to improve the classification accuracy in credit scoring models by adding reject applications. Our proposed models infer the unknown creditworthiness of the rejected applications by exact enumeration of the two possible outcomes of the loan (default or non-default). The efficient stochastic gradient optimization technique used in deep generative models makes our models suitable for large data sets. Finally, the experiments in this research show that our proposed models perform better than classical and alternative machine learning models for reject inference in credit scoring.

show abstract

“…However, developing a scorecard that ignored "rejects" is not applicable to the total population, and forecasting for all applicants will not be accurate and realistic (Siddiqi, 2012). While a stream of literature focuses on "reject inference techniques" that address this problem (Jacobson and Roszbach, 2003;Crook and Banasik, 2004;Bücker et al, 2013), including certain types of loans that should not be in the development sample, also deteriorates the model, especially for behavior scorecards. Including loans for special customers such as staff and VIPs or lost/stolen cards, deceased customers and restructured payment plans in the observed data set will change the true characteristics that predict the target (Siddiqi, 2012).…”

Section: Details Of Scoring Typementioning

confidence: 99%

A review of credit scoring research in the age of Big Data

Onay¹,

Öztürk

2018

JFRC

View full text Add to dashboard Cite

Purpose-This paper aims to survey the credit scoring literature in the past 41 years (1976-2017) and presents a research agenda that addresses the challenges and opportunities Big Data bring to credit scoring. Design/methodology/approach-Content analysis methodology is used to analyze 258 peer-reviewed academic papers from 147 journals from two comprehensive academic research databases to identify their research themes and detect trends and changes in the credit scoring literature according to content characteristics. Findings-The authors find that credit scoring is going through a quantitative transformation, where datacentric underwriting approaches, usage of non-traditional data sources in credit scoring and their regulatory aspects are the up-coming avenues for further research. Practical implications-The paper's findings highlight the perils and benefits of using Big Data in credit scoring algorithms for corporates, governments and non-profit actors who develop and use new technologies in credit scoring. Originality/value-This paper presents greater insight on how Big Data challenges traditional credit scoring models and addresses the need to develop new credit models that identify new and secure data sources and convert them to useful insights that are in compliance with regulations.

show abstract

Reject inference in consumer credit scoring with nonignorable missing data

Cited by 35 publications

References 17 publications

Inferring the Outcomes of Rejected Loans: An Application of Semisupervised Clustering

Inferring the Outcomes of Rejected Loans: An Application of Semisupervised Clustering

Deep generative models for reject inference in credit scoring

A review of credit scoring research in the age of Big Data

Contact Info

Product

Resources

About