The total cost of misclassification in credit scoring: A comparison of generalized linear models and generalized additive models

• Author: Christian Lohmann,Thorsten Ohliger
• Date: 01 August 2019
• Published date: 01 August 2019
• DOI: http://doi.org/10.1002/for.2545

RESEARCH ARTICLE

The total cost of misclassification in credit scoring: A

comparison of generalized linear models and generalized

additive models

Christian Lohmann

1

| Thorsten Ohliger

2

1

Schumpeter School of Business and

Economics, University of Wuppertal,

Wuppertal, Germany

2

parcIT GmbH, Cologne, Germany

Correspondence

Christian Lohmann, Schumpeter School

of Business and Economics, University of

Wuppertal, Gaußstraße 20, 42119

Wuppertal, Germany.

Email: lohmann@wiwi.uni‐wuppertal.de

Abstract

This study examines whether the evaluation of a bankruptcy prediction model

should take into account the total cost of misclassification. For this purpose, we

introduce and apply a validity measure in credit scoring that is based on the

total cost of misclassification. Specifically, we use comprehensive data from

the annual financial statements of a sample of German companies and analyze

the total cost of misclassification by comparing a generalized linear model and

a generalized additive model with regard to their ability to predict a company's

probability of default. On the basis of these data, the validity measure we intro-

duce shows that, compared to generalized linear models, generalized additive

models can reduce substantially the extent of misclassification and the total

cost that this entails. The validity measure we introduce is informative and

justifies the argument that generalized additive models should be preferred,

although such models are more complex than generalized linear models. We

conclude that to balance a model's validity and complexity, it is necessary to

take into account the total cost of misclassification.

KEYWORDS

bankruptcy prediction, cost of misclassification, generalized additivemodel, generalized linear model

1|INTRODUCTION

All creditors need to perform credit scoring before they

grant credit to potential debtors. Credit scoring involves

quantifying the probability of the debtor going bankrupt

and the costs that are associated with the decision to lend

that debtor credit. If the credit is granted, the lender faces

potential costs if the debtor defaults on the credit; such

costs may arise from defaulted repayments and interest.

If the credit is not granted, the potential costs are usually

opportunity costs that arise from foregone business

transactions. To determine reliably how much equity is

required on the basis of economic targets and regulatory

standards and to calculate the appropriate credit interest

rates, the creditor has to be able to quantify accurately

the debtor's risk of bankruptcy and the cost that might

arise as a result of the decision to lend that debtor credit.

When credit interest rates are relatively high, they can

balance the expenses that may arise from a credit default.

However, to avoid rejecting low‐risk debtors because

of adverse selection, credit interest rates must not be

disproportionately high in relation to the debtor's

bankruptcy risk. In this context, economists recognized

as far back as the 1930s that it is crucial to evaluate

accurately a debtor's bankruptcy risk (Fitzpatrick, 1932a,

1932b, 1932c).

The debtor's bankruptcy risk is typically evaluated by

means of empirical models that estimate the probability

of default. Every type of empirical model involves a

trade‐off between validity and complexity. Specifically, a

Received: 10 April 2018 Revised: 10 July 2018 Accepted: 3 August 2018

DOI: 10.1002/for.2545

Journal of Forecasting. 2019;38:375–389. © 2018 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/for 375