e-news

Credit Scoring and Credit Control

The Credit Scoring and Credit Control Conference is the leading international conference in the credit scoring industry.

Held every two years at the University of Edinburgh, the event attracts delegates from around the world. The format comprises a number of parallel themed streams punctuated by three plenary talks from invited speakers on topical issues from the industry.

Even in these difficult times, attendance was just over 300 this year with Experian contributing no fewer than 10 papers, including one of the invited talks. Experian people came from the UK, US, Scandinavia, Bulgaria and Latin America. The conference provides a valuable opportunity to take the temperature of the industry and pick up on developments that Experian can adopt for itself.

Unusually for an academically based conference, around 80% of the participants are industry practitioners, the remainder coming from academic institutions. It is run by the Credit Research Centre (CRC), an affiliation of credit risk researchers from the Universities of Edinburgh and Southampton and Imperial College. The CRC is independent of credit granting organisations and engages in both theoretical and highly applied research of interest to all stakeholders in the industry, including lenders, credit suppliers, credit scoring organisations, borrowers and government agencies.

The dramatic changes to the credit world in the two years since the last conference were reflected by the organisers' choice of papers, both invited and contributed. A senior economist at the Federal Reserve Board, gave an interesting and informative keynote talk on the causes of the financial crisis, the complexities of securitisation and the lessons learned.

Macroeconomics in credit scoring
Predictably, talks on macroeconomics in credit scoring were prominent. William Thomson (Experian Business Strategies) described how important it is understand the economic variables used in credit scoring models and how the use of appropriate segmentation is essential in identifying the groups most at risk. A paper from the University of Edinburgh described the use of Survival Analysis for a large sample of credit card applications from 1999 to 2006. Although this period exhibited far less of the volatility experienced recently, it was found that the economics did make a small but statistically significant contribution to the performance of the models which also incorporated application and behavioural data. Certain advantages of the Survival Analysis approach over logistic regression were described, such as the ability to estimate the probability of default at different points in time rather than for a fixed interval.

Another theme discussed was the development of loss given default (LGD) models, particularly with relation to the Basel capital adequacy regulations. In a talk bridging two themes, a Professor from the University of Edinburgh described the use of macroeconomics in LGD models for retail credit cards. The data consisted of over 55,000 credit card accounts in default from 1999 to 2005. Again, this period included no downturns on the scale of recent events, but there was enough evidence to show that base rates and unemployment rate in particular could make a small, yet statistically significant contribution to the model performance. Interactions between the account variables and economic factors, however, led to slightly worse results, in agreement with other research. Finally, some early results on the use of macroeconomics for scenario generation and stress testing were described.

Also presenting on the LGD theme were Stefan Stoyanov (Experian Decision Analytics) with ''The event agenda covered a vast array of subject matter and there were a number of papers in the Techniques of Scoring stream that were thought-provoking...'' a case study for a leading Central and Eastern European Bank and a speaker from Charles University in Prague, presenting a paper on Modelling Bank Loan LGD of Corporate and SME Segments. They concluded that modelling LGD gives better estimates than the conventional segmenting approach and that the absolute exposure was strongly related to the rate of loss.

Changing economic conditions may be the cause of changing score-odds relationshipsover time and this theme was the subject of talks by presenters from FICO and the University of Southampton. The speaker from the University of Southampton reported on an approach for monitoring this relationship, which may be considered for Decision Analytics' Consumer Delphi Monitor service.

The event agenda covered a vast array of subject matter and there were a number of papers in the Techniques of Scoring stream that were thought-provoking, and a Professor from ICL gave a very insightful talk on the pitfalls of using inappropriate model performance criteria.

'Where did it all go wrong?'
In addition to the talks, there were also two discussion sessions. 'Where did it all go wrong?' featured two of the conference organisers and a speaker from the FSA attempting to sum up the causes and consequences of the Credit Crisis. 'This house believes … that Credit Scoring methods and applications are stuck it the 1980s' debated this stand-point drawing upon expert witnesses from both sides. After hearing the evidence, the audience was asked to vote. Perhaps unsurprisingly, the majority verdict was that credit scoring had moved on appreciably in the last 20 years.

The next two years will bring further developments to the world of credit and it's likely that CSCC XII will feature financial regulation heavily. Equally likely is that Experian will, again, make a significant contribution.

Paul Russell
Head of Analytics
Decision Analytics
Experian

 

Read more about Scoring and Analytics

Visit the website to read more »

 

Contact us for further discussions about this article