Universität zu Köln

Advanced Methods for Loss Given Default Estimation

Töws, Eugen (2016) Advanced Methods for Loss Given Default Estimation. PhD thesis, Universität zu Köln.

Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (4055Kb) | Preview


    This thesis consists of three essays on the estimation of the risk parameter LGD. The first essay (Hartmann-Wendels, Miller, and Töws, 2014, Loss given default for leasing: Parametric and nonparametric estimations) fills a gap in LGD related literature by focusing on elementary differences of the examined estimation approaches. We find that finite mixture models are quite capable of reproducing the unusual shape of the LGD distribution. However, out-of-sample the estimation error increases significantly. Model trees produce robust in-sample and out-of-sample estimations. Furthermore, we find that the improvement of advanced models increases with an increasing dataset. The second essay (Töws, 2014, The impact of debtor recovery on loss given default) addresses the economic consideration of the workout process of defaulted contracts. Dependent on the lessor’s workout strategy, defaulted contracts may develop in two distinct ways. Either the default reason can be dissolved and the debtor recovers or the contract must be written off. The study indicates the benefits of establishing the lessor’s expertise in assessing a defaulted contract’s continuation worthiness. If successfully implemented, the resulting workout process should produce lower LGDs than before and thereby strengthen the lenders competitiveness. The third essay (Miller and Töws, 2015, Loss given default-adjusted workout processes for leases) contributes to the LGD estimation literature considering unique features of leasing contracts. We economically account for leasing peculiarities and develop a particularly suited approach to estimate the leasing contracts’ LGD. We separate the LGD into its cash flows of the workout process of defaulted leasing contracts. We find that the estimated values of the parts of the LGD are sturdy indicators for the success of the workout process. Thus, the lessor can benefit from the consideration of both these forecasts for his actions concerning the workout process.

    Item Type: Thesis (PhD thesis)
    Töws, Eugen
    Corporate Creators: Universität zu Köln
    URN: urn:nbn:de:hbz:38-65332
    Subjects: General statistics
    Management and auxiliary services
    Uncontrolled Keywords:
    Loss given default, leasing, workout process, recovery, economic model, forecasting, classification, finite mixture models, regression and model trees, random forest, C5.0, boostingEnglish
    Faculty: Wirtschafts- u. Sozialwissenschaftliche Fakultät
    Divisions: Wirtschafts- u. Sozialwissenschaftliche Fakultät > Seminar für Allg. BWL und Bankbetriebslehre
    Language: English
    Date: 05 February 2016
    Date Type: Publication
    Date of oral exam: 11 January 2016
    Full Text Status: Public
    Date Deposited: 02 Feb 2016 15:35:25
    NameAcademic Title
    Hartmann-Wendels, ThomasUniv.-Prof. Dr.
    Hess, DieterUniv.-Prof. Dr.
    Schradin, HeinrichUniv.-Prof. Dr.
    URI: http://kups.ub.uni-koeln.de/id/eprint/6533

    Actions (login required)

    View Item