Probability of default calibration for low default portfolios: revisiting the Bayesian approach
| dc.contributor.advisor | Conradie, W. J. | en_ZA |
| dc.contributor.author | Venter, Edward Stevens | en_ZA |
| dc.contributor.other | Stellenbosch University. Economic and Management Sciences. Dept. of Statistics and Actuarial Science | en_ZA |
| dc.date.accessioned | 2016-03-09T14:53:48Z | |
| dc.date.available | 2016-03-09T14:53:48Z | |
| dc.date.issued | 2016-03 | |
| dc.description | Thesis (MCom)--Stellenbosch University, 2016 | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT : The Probability of Default is one of the fundamental parameters used in the quantification of credit risk. When estimating the Probability of Default for portfolios with a low default nature the Probability of Default will always be underestimated. Therefore, a need exists for calibrating the Probability of Default for Low Default Portfolios. Various approaches have been considered in the literature review, with the main approaches being the Confidence Based Approach and Bayesian Approach. In this study the Bayesian Approach for calibrating the Probability of Default for portfolios of high grade credit is reconsidered. Two alternative prior distributions that can be used in the Bayesian Approach are proposed; these are an informative, Strict Pareto distribution and a non-informative Jeffreys prior. The performance of these proposals are then compared to existing calibration techniques by using real data. | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING : Die Waarskynlikheid van Wanbetaling is een van die fundamentele parameters in die beraming van kredietrisiko. Wanneer die Waarskynliheid van Wanbetaling beraam word vir ’n portefeulje met lae wanbetaling observasies in die historiese data, vind onderberaming altyd plaas. Dus bestaan daar ’n nood vir kalibrasie tegnieke vir die Waarskynlikhied van Wanbetling vir Lae Wanbetaling Portefeuljes. ’n Verskeidenheid van benaderings word in die literatuur voorgestel, waaronder die Vertroue Gebasseerde Benadering en die Bayesiaanse Benadering die bekendste is. In hierdie studie word die Bayesiaanse Benadering vir die kalibrasie van die Waarskynlikheid van Wanbetaling vir portefeuljes van hoë vlak krediet heroorweeg. Twee alternatiewe apriori verdelings word voorgestel om in die Bayesiaanse Benadering te gebruik. Hierdie apriori verdelings is die streng Pareto verdeling wat ’n inligting-gewende apriori verdeling is en die Jeffreys apriori verdeling wat ’n nie-inligting-gewende apriori verdeling is. Die prestasie van die tegnieke wat voortvloei uit die gebruik van die voorgenoemde twee apriori verdelings word dan vergelyk met bestaande kalibrasie tegnieke deur gebruik te maak van werklike data. | af_ZA |
| dc.format.extent | xiv, 153 pages : illustrations (chiefly colour) | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/10019.1/98723 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject | Probability of default (PD) | en_ZA |
| dc.subject | Low default portfolios (LDP) | en_ZA |
| dc.subject | Credit risk | en_ZA |
| dc.subject | Financial risk management | en_ZA |
| dc.subject | Bayesian statistics | en_ZA |
| dc.subject | UCTD | en_ZA |
| dc.title | Probability of default calibration for low default portfolios: revisiting the Bayesian approach | en_ZA |
| dc.type | Thesis | en_ZA |