Learning algorithms, and current developments – banking at the crossroads
| dc.contributor.advisor | Pott, Christiane | |
| dc.contributor.author | Erhardt, Joachim | |
| dc.contributor.referee | Hoffjan, Andreas | |
| dc.date.accepted | 2017 | |
| dc.date.accessioned | 2018-10-10T06:52:17Z | |
| dc.date.available | 2018-10-10T06:52:17Z | |
| dc.date.issued | 2017 | |
| dc.description.abstract | Since the 2008 financial crisis much discourse questioned the relevance of the banking industry to society; in particular given sovereign bailouts of failed banks. Such questions provide the context in which this thesis covers four key topics for wholesale banking in the future. Firstly, in relation to efficiency improvements in the wholesale banking model and promotion of market liquidity, ‘Machine Learning’ techniques were applied to predict the behaviour of wholesale customers in addition to the financial markets they operate in. Central to the analysis is the design of predictive features i.e. the independent variables. In relation to the customer analysis, institutional data was applied to forecast ‘when’ and ‘what’ customers will trade. Accuracies range from 43% to 95% (up to a 17% improvement versus customers’ historic mean). In relation to market prediction, results varied substantially across feature sets and markets. The concept of ‘Genuine Model Robustness’ is proposed by which statistical importance and improvement scores are applied as a measure for significance for the underlying feature mechanic. Secondly, in relation to the mitigation of bank failure, the impact of the new ‘Bail-in mechanism’ introduced by the EU was analysed. In particular the effects of the new regime on bank funding costs and the corresponding interplay between increased covered bond issuance and ‘Asset Encumbrance’ was considered. Thirdly, the ethical dimension of ‘Sustainability’ in investing was reviewed. Via a proposed ‘Green Covered Bond’ framework (with eligibility standards and enhanced credit quality) green projects can benefit from the inherent funding advantages of the covered bond market. Finally, the stimulus effect of finance on the real economy was considered, in particular the influence on ‘Resource Availability’ via international trade from cheaper and reliable financing. Through fixed and pooled effect regression analysis a statistically significant negative relationship was established between availability of base metals and finance costs. | en |
| dc.identifier.uri | http://hdl.handle.net/2003/37145 | |
| dc.identifier.uri | http://dx.doi.org/10.17877/DE290R-19141 | |
| dc.language.iso | en | de |
| dc.subject | Machine learning | en |
| dc.subject | Asset encumbrance | en |
| dc.subject | Sustainability | en |
| dc.subject | Green bonds | en |
| dc.subject | Import of natural resources | en |
| dc.subject.ddc | 330 | |
| dc.subject.rswk | Maschinelles Lernen | de |
| dc.subject.rswk | Nachhaltigkeit | de |
| dc.subject.rswk | Firmenkundengeschäft | de |
| dc.title | Learning algorithms, and current developments – banking at the crossroads | en |
| dc.title.alternative | Essays on Machine learning, Applied data analytics, asset encumbrance, Bail-in, sustainability and resource availability | en |
| dc.type | Text | de |
| dc.type.publicationtype | doctoralThesis | de |
| dcterms.accessRights | open access | |
| eldorado.secondarypublication | false | de |