Introduction

The Wirecard financial scandal shocked the German economy in 2020 when the company had to file for insolvency. The Wirecard scandal is one example of large-scale financial frauds in recent history. The question arises why the fraud was not uncovered earlier by involved external auditors and public oversight bodies. The business operations of companies have grown in complexity, volume, and speed over the last decades enabled by technology and globalisation. This makes it increasingly difficult for those in charge to detect fraud. This study investigates whether the Wirecard scandal could have been predicted with the use of machine learning techniques by using publicly available data and software. It combines machine learning analysis with case study research to illustrate how the analysis techniques can be used, what the output and the implications for involved parties are. The results show that eight out of ten financial statements from Wirecard published between 2009 and 2018 are classified as fraudulent by the best performing machine learning model. Un-folding the predictions using Shapley values explains that the classifications are mainly driven by abnormal relationships between revenues and cash recorded by the company. This study discusses how the fully trained model performance compares to other studies in the field and what the implications for involved parties are for use in the future.

General information

This seminar will be organised in a hybrid setup. If you are interested in joining this seminar, please send an email to the secretariat of Amsterdam Business School at secbs-abs@uva.nl.