University of Twente Student Theses

Login

Logit, Tobit or Hazard? An analysis of modelling the probability default of a retail mortgage owner.

Verwoerd, M.P. (2015) Logit, Tobit or Hazard? An analysis of modelling the probability default of a retail mortgage owner.

[img] PDF
3MB
Abstract:This thesis is about modeling the 1 year and life time PD of a retail mortgage customer. During this thesis three statistical frameworks (Logit, Tobit & Hazard) are applied in prototypes and the quantitative and qualitative results are compared. During this thesis it turned out that the developed Tobit prototype was not able to sufficiently predict PDs, because of a lack of predictive power. This shortcoming was caused by a bad fit of the latent Tobit variable, which on itself is possibly caused by wrong assumptions. The best performing prototype is based on the Logit framework. The prototype has a good predictive and discriminatory power and therefore meets all the quantitative requirements. Also the Least Squares extrapolation method, used to be able of predicting life time, worked well. The prototype that is based on the Hazard framework did not meet the requirements regarding predictive power. This is probably caused by the bad fit of the hazard rate of Default to the realizations. When this problem will be resolved the predictive- and discriminatory power are expected to grow. The fact that the Hazard framework offers a natural LT framework in which two events could be fitted, makes it a promising prototype.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:83 economics
Programme:Industrial Engineering and Management MSc (60029)
Link to this item:https://purl.utwente.nl/essays/68201
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page