The Credit Research Research Centre (CRC) is one of the UK’s leading centres for research into all aspects of credit to both consumers and customers. Members of the CRC have research interests ranging broadly across the economics of credit, credit scoring, customer scoring and SME risk modelling. Credit scoring involves the application of models to predict the probability that an applicant for credit will default on the loan. It has been described as the most successful application of Operational Research techniques ever. Members of the Centre have written the world’s most authoritative text on the subject as well as having published in the world’s best journals in the area.
Following is a sample of the range of opportunities for doctoral study which currently exist within the Centre.
Ethical Artificial Intelligence models
The adoption of Artificial Intelligence (AI) technologies in the Banking industry has boosted over the last half a decade. A debate has emerged on different ethical concerns and consequences of the use of AI, not only in banking, but also other areas such as justice, law enforcement and medicine. This project will explore the development of AI-based methodologies that can overcome or mitigate these issues, focusing on applications to credit scoring and financial risk management.
Required qualifications: A good first degree in Business Analytics, Operational Research, Data Science, Computer Science, or Engineering.
Supervisor: Dr Belen Martin-Barragan
Stress testing consumer credit portfolios
Regulators and financial, institutions have been ‘stress testing’ their loan portfolios to examine the amount of capital that an institution may expect to maintain to cover for unexpected losses. However there are many different types of stress test and many methodological difficulties in carrying a stress test out. In this project the student will consider how discrete survival models can be used to enable stress tests to be carried out in particular how the macroeconomic variables can be ‘combined’ in some sense to yield a valid ‘test’. The stability of expected loss distributions will be considered in detail.
Modelling loss given default and recoveries for corporate loans
Each bank that is regulated under the Basel III Accord has to maintain an amount of capital in case of unexpected losses. The Accord indicates how the minimum amount of capital should be computed. The amount depends on the amount of risk weighted assets that a bank has and an important component of this is the forecast levels of Loss Given Default for each bucket of a loan portfolio. In this project the LDG for each of a large sample of corporate loans will be modelled. Various algorithms and data transformations will be considered as will the incorporation of macroeconomic variables. Stress testing will be carried out.
The amount of capital a bank is required to hold in case of unexpected losses is governed by the Basel II Accord. The relevant computation of the risk weighted assets makes some ‘assumptions’ about the correlation between the asset values. The aim of this project is to explore methods of measuring the degree of concentration of risk in a credit portfolio. The project will consider ad hoc and model based measures and practical approaches to its empirical measurement.
There is a large amount of literature on reasons for person bankruptcy. In this project the sociodemographic characteristics of those who declare bankruptcy will be modelled using mixed fixed and random effect models. The fixed effects will include states of the macroeconomy. Initially publicly available data from surveys will be used. Later we will be to seek data from a financial institution.
Exploring affordability in retail credit
Traditional credit scoring used by the high street lenders creates two classes those who will be given credit and those who will not. Individuals are therefore excluded from credit and often seek alternative and more expensive forms of credit.
More recently risk-based pricing has been introduced by high street lenders to provide credit to individuals who under traditional models would not have received credit but who will be charged a rate associated with the their risk profile. This means individuals with low risk ratings will be required to higher interest rates. This raises the question of whether individuals will be able to afford the credit and the need for the lender to be able to assess the individual’s ability to pay. The research will explore both of these related aspects. It will explore the available data on individual economic resources arising from Income and Expenditure Surveys, and other published sources. It will also explore how this information can be used to enhance the models currently used to assess affordability.
Comparisons of household credit constraints
Credit constraints have profound implications for the applicability of the Permanent Income Hypothesis, for the effects of certain government fiscal and credit policies. However the empirical literature has not definitively identified which households are credit constrained and how the characteristics of those constrained has changed before and after the recent financial crisis. The aim of this project is to identify the characteristics of households who are credit constrained over the last decade in a variety of countries using different data sources. These include self-reported credit constraints, implied credit constraints of ten used in the excess sensitivity of consumption literature, demand and supply of debt equations and administrative data.
Required qualifications: a good first degree in Economics and preferably an MSc in Economics.
If you have any questions about these topics, please contact Professor Jonathan Crook.