01whole.pdf (3.39 MB)
Download file

Empirical studies in default and insurance risk

Download (3.39 MB)
thesis
posted on 29.03.2022, 00:18 authored by Feng Liu
This PhD thesis evaluates three types of financial risks with analytical methods that improve the accuracy when making predictions for future risk events. It consists of three key chapters based on three research papers.The research paper titled 'Application of the bivariate negative binomial regression model in analysing insurance count data' analyses insurance claim frequency data using the bivariate negative binomial regression (BNBR) model, and claims data from third-party liability and comprehensive motor insurance. It is found that bivariate regression, with its capacity for modelling correlation between the two observed claim counts, provides a superior fit and improved out-of-sample prediction compared to the more common practice of fitting univariate negative binomial regression (UNBR) models separately to each claim type. Noting the complexity of BNBR models and their potential for a large number of parameters, this study explores the use of model shrinkage with the Lasso and ridge regression. Results show that our models estimated using shrinkage methods outperform the ordinary likelihood-based models when being used to make predictions out-of-sample. It can also be shown that the Lasso performs better than ridge regression as a method of shrinkage in the context considered here. The research paper titled 'Assessing Sovereign Risk: A Bottom-Up Approach' assesses sovereign default risk of individual states in the U.S. using information about default risk at the company level. The integrated risk factors of the private sector are linked to the overall sovereign risk of state governments in conjunction with additional financial variables. Using data from Moody's KMV expected default frequencies (EDFs) on corporate default risk, credit risk indicators for different industries are estimated. Building on these measures, state level credit risk indicators are developed encompassing industry compositions to explain the behaviour of credit default swap (CDS) spreads for individual states. It is found that market-based measures of private sector credit risk are strongly associated with subsequent shifts in sovereign credit risk premiums measured by CDS spreads. The credit risk indicators developed here are demonstrated to be highly significant in forecasting sovereign CDS spreads at weekly and monthly sampling frequencies.The research paper titled 'A joint model for longitudinal and time-to-event data in corporate default risk modelling' applies a joint model for longitudinal and time-to-event data to assess corporate default risk. A linear mixed-effects model is used to describe the trajectory of the predictor variable for company default. The output from this analysis is then used in a survived model where time to corporate default in the response variable of interest. The joint model does not assume constant values for the independent variable between observations, and can take advantage of the fully-specified subject-specific longitudinal trajectories. Data collected on U.S. listed companies from 1997 to 2016 is used to test the ability of the joint model to predict corporate default events. Two independent variables, the distance-to-default and the age of a company, are used to assess the company's probability of default over various time horizons. It is found that the joint models outperform the Cox model and the Weibull model in making predictions of default events. The results show that the joint model is more suitable in assessing corporate default risk than the selected standard survival models.

History

Table of Contents

1.Introduction -- 2. Application of the bivariate negative binomial regression model in analysing insurance count data -- 3. Assessing sovereign risk: a bottom-up approach -- 4. A joint model for longitudinal and time-to-event data in corporate default risk modelling -- 5. Conclusion.

Notes

Theoretical thesis. Bibliography: pages 118-130

Awarding Institution

Macquarie University

Degree Type

Thesis PhD

Degree

PhD, Macquarie University, Faculty of Business and Economics, Department of Actuarial Studies and Business Analytics

Department, Centre or School

Department of Actuarial Studies and Business Analytics

Year of Award

2018

Principal Supervisor

David Pitt

Additional Supervisor 1

Stefan Trück

Rights

Copyright Feng Liu 2018 Copyright disclaimer: http://mq.edu.au/library/copyright

Language

English

Extent

1 online resource (xii, 130 pages)

Former Identifiers

mq:70833 http://hdl.handle.net/1959.14/1268183

Usage metrics

Exports