Bankruptcy Prediction of Privately Held SMEs Using Feature Selection Methods
Series
School of Finance Working Paper Series
Type
working paper
Date Issued
2021-08-27
Author(s)
Abstract
In this paper, we test alternative feature selection methods for bankruptcy prediction and illustrate their superiority versus popular models used in the literature. We test these methods using a comprehensive dataset of more than one million financial statements covering the entire universe of privately held Norwegian SMEs in 2006-2017. Our methods can choose among 155 accounting-based input variables derived from prior literature. We find that the input variables chosen by an embedded least absolute shrinkage and selection operator (LASSO) method yield the best in-sample fit and out-of-sample performance. We show in a simulation, which mimics a real-world competitive credit market, that using LASSO to choose bankruptcy predictors improves credit risk pricing and decision making, resulting in significantly higher bank profits. Finally, we show that model performance can be further improved by running feature selection methods on sub-sets of the company universe, such as for example within-industry.
Language
English
HSG Classification
contribution to scientific community
HSG Profile Area
SOF - System-wide Risk in the Financial System
Refereed
Yes
Pages
64
Subject(s)
Contact Email Address
markus.schmid@unisg.ch
Eprints ID
266620
File(s)![Thumbnail Image]()
Name
Bankruptcy Prediction of Privately Held SMEs.pdf
Size
1.05 MB
Format
Adobe PDF
Checksum (MD5)
35160abb688c62e187f24e1007624a6b