The following article, written by Dr. Paul Russell, Director of Analytics at Experian Decision Analytics EMEAI, was originally published in Mortgage Finance Gazette.

Historical perspective

The use of credit scoring can be traced back to World War 2, when scoring techniques were developed in the US to fill the void left by human decision makers serving in the armed forces.  Since that time, credit scoring has evolved both in the sophistication of the analytical techniques employed and the level of integration into lenders’ decision-making processes.  Early uses of credit scoring focussed on the prediction of the risk of arrears or default.  Confidence in the use of such scoring models was low and, consequently, so was the use of score-based automated decision-making.  However, it became clear that the use of mathematical models allowed for decisions to be made consistently, accurately and, above all, quickly.  With the increasing demand for credit in the post-war era and the requirement to make lending decisions rapidly and cheaply, it was this final point which ensured the acceptance of, and eventual reliance on, credit scoring.

In the UK, scoring techniques started to be used heavily in the mail order sector in the 1970s where there was little customer information available beyond name, address and account behaviour.  During the 1980s, scoring started to be used more widely, particularly in the banking sector, to process new applications.  Towards the end of that decade, behavioural scoring, scores based on the way that a customer uses a particular facility, started to be more widely used.  In the early 1990s the major banks started to recognise the potential of scoring to deliver better decision making, reduced losses and improved efficiency to the extent that, today, lending decisions within branches are extremely rare, with almost all credit decisions being made centrally. 

The rise of the credit bureau

Just as a horse’s form is important in deciding what to back in the 3.30 at Kempton Park, track record plays an important part in making lending decisions.  At the same time that scoring was becoming more widely used, a number of organisations, including the major lenders, realised that sharing information about the way people repaid their debts was a powerful way of establishing an individual’s track record.  Tracing their heritage back to the merchant mutual protection societies formed at the start of the 1900s, these bureaux allow a lender to enrich their own information on any applicant or customer, with details of that individual’s relationships with other lenders.  This gave some comfort to the lenders and made credit more widely available and cheaper to people with a good track record.  The richness of the data held by the bureaux has grown through the years to the extent that in markets, such as the UK and US, it is rare for a significant lending decision to be made without recourse to a credit bureau.

While credit bureau data provides lenders with a more complete picture of the behaviour of potential and existing customers it can be difficult to interpret these complex data sets.  This complexity proved to be a barrier to the drive for fast, consistent and accurate decision making and scoring was the natural answer to this problem through its ability to identify and quantify trends in highly complex data.  The credit bureaux quickly realised the advantage of providing scores to summarise this information and to provide an easily interpreted view of a lender’s customers’ behaviour in the wider world.

Scoring is not the whole story

A scorecard is simply a tool which provides an estimate of risk of something happening in the future, such as the customer not paying.  The gap between assessing the risk and making a decision on a customer is the realm of strategy.  The actions to be taken by a lender in different situations are captured in a decision-making framework which comprises scorecards, policy rules and recommended actions.  The concept of strategy management, in which scorecards are used to drive complex decision logic, implemented in automated rules engines, is at the heart of all major lenders’ systems.  The result of this is an expectation that business decisions and strategy choices can be made almost instantaneously and can be quantified, assessed and improved.

A mortgage perspective

While it is true that most mortgage lenders now use scoring extensively, particularly within the larger institutions, few were among the earliest adopters of scoring technology.  This is understandable because the business model is very different: the lending decision is part of a much longer process involving many parties and this reduces greatly the need for a fast decision.  The security held by the lender mitigates the risk in the customer and makes less pressing the need for a more formal risk assessment; on a recent project in Asia, a bank argued strongly that it did not need scoring because if the customer defaulted, it would simply sell the house.

More recently, benign economic conditions, low interest rates and rising house prices have led to historically low levels of default in the mortgage sector.  While this is good news for organisations running mortgage portfolios, it is not so good for scorecard developers who are looking for bad accounts to include in their scoring model developments.  Basel II and the Capital Requirements Directive are driving banking organisations to improve their risk management capabilities and this is causing many mortgage lenders to implement scoring models as part of an integrated rating system.  In this situation, the limited default experience of many mortgage issuers makes life very difficult if one is trying to estimate default rates and losses when the lender has limited or no default experience.  These challenges are leading the industry to look into alternative techniques for the estimation of rates in so-called low default portfolios. 

In a saturated financial services market, margins are squeezed and lenders are always looking for better returns and the sub-prime sector is seen as an under-served segment which offers the possibility of richer pickings.  The growth of the sub-prime mortgage sector poses some interesting challenges: accurate pricing is vital but this has to be done in an environment where traditionally predictive data items, like the presence of a court judgment, can change their meaning – in the sub-prime sector many people have court judgments so it is not predictive in the same way as in the main stream. 

Recent Trends

With personal debt in the UK having recently passed £1.2 trillion, consumer indebtedness has become a very high profile issue for the government and for the credit industry.  There is an increasing focus within the industry, driven by a concerned and increasingly interventionist government, to ensure that lending decisions are made with full consideration being given to the individual’s ability to repay.  In response to this, the interest in indebtedness scoring has increased, with both lenders and scoring providers shifting their focus to the production of indebtedness measures to supplement traditional risk scores.  Indebtedness scores can act as excellent predictors of future delinquency, even for customers who do not exhibit traditional negative indicators of high credit risk, so enabling lenders to demonstrate responsible lending policies.

The increase in personal indebtedness has been accompanied by a significant rise in the rate of personal insolvency.  The recently enacted Enterprise Act has made it easier for debtors to enter into bankruptcy and Individual Voluntary Arrangements (IVA) and has lessened both the length and severity of the resulting sanctions.  While intended to engender entrepreneurship, it is felt by some that a side effect of the legislation has been to encourage indebted individuals to seek an easy way out of their debts.  In response to this, lenders are seeking ways of identifying potential insolvents before the lending decision is made.  This has seen the development of scores specifically designed to predict the likelihood of bankruptcy separately to traditional default risk. 

The New Basel Capital Accord (Basel II) is driving innovation within risk management practice and is causing banks to refocus risk management from the prediction of simple credit risk to the forecasting of expected and unexpected losses, with loss given default (LGD) and exposure at default (EAD) models joining the more traditional probability of default (PD) scores.  The development of tools to estimate these parameters, to prove that they work and to understand how they behave in an economic downturn is taking vendors and users of scoring technology into new and exciting areas.

A recent technological advance that is starting to deliver significant benefits to lenders is the use of optimisation techniques in areas such as the assignment of credit limits and the pricing of loans.  The challenge is to come up with the ‘best’ strategy for a customer segment, subject to the constraints facing the business at that time: for example, limits on the increase in bad debt that can be tolerated, the capacity of a call centre to handle incoming applications, etc.  The impact of changes in limit or loan pricing on profitability can be modelled using traditional scoring models but choosing the ‘best’ limit or price for an individual to extract maximum profit is not trivial.  This is where the new technique of strategy optimisation has come to the fore.  This is a collection of powerful analytical techniques designed to ensure that information available on customers and applicants is used in the optimal way.  By successfully translating the power of the models gained from scoring into an implementable strategy, optimisation ensures that the full benefits available through scoring are actually realised.

The importance of the credit report

The credit report lies at the heart of many of today’s credit scoring models, at least in those markets, such as the US, UK and most of Europe, where credit bureaux exist.  However, the contents of credit reports vary greatly between different jurisdictions and regulatory frameworks.  Lenders in the UK are fortunate in having access to both positive data and negative data; not all countries are so fortunate, with many allowing the sharing of negative data only and short limits on the length of time even negative data can be stored.  Nevertheless, even in the UK, access to both positive and negative data is not assured.  Because lenders share their data on the basis of ‘reciprocity’, a lender will only be able to see data to the same level that it is willing to share.  If a lender supplies negative data only, it will only see negative data in return, and will be unable to assess a customer’s overall commitments and ability to make repayments if it grants a new credit line.

While most lenders now take access to credit reports for granted, until recently, few of their customers were unaware of their existence.  That situation has changed dramatically in the last decade as a result of education campaigns, mostly by the credit bureaux themselves, and greater coverage of the subject in the media, particularly to encourage consumers to exercise their statutory right to see their credit reports.  Around one million people each year make use of this statutory right to request their credit reports from Experian.  The actual numbers now seeing their reports, however, is much greater, following the introduction of online services from the three credit bureaux to provide online access to their credit reports.  Members of Experian’s online service, which allows unlimited viewings of their credit report, access it, on average, more than twice a month.

Five years ago, research at Experian showed that around 90 per cent of credit report requests from consumers were the result of being turned down for credit.  That figure has fallen to below 40 per cent as people have become more aware of the importance of reviewing their report before making major purchases, particularly when applying for a mortgage.  The advantage for mortgage lenders of their customers reviewing their credit report before making an application is that the consumer can ensure it is accurate and up-to-date and make sure that they haven’t been victim of any fraudulent activity.  In some cases, it might also deter some applicants who recognise that they stand little chance of being granted a mortgage, based on their credit report, thus saving the lender the cost of interviewing and processing the application.

Conclusion

The evolution of strategy within a lending organisation relies on the ability to gather and interpret information from external sources, such as credit reports, and all aspects of a customer’s interactions with a lender.  Scoring techniques have proved very useful in making sense of this complex information and will be at the heart of future development in the mortgage industry.

Dr. Paul Russell - Director of Analytics, Experian Decision Analytics EMEAI

Contact us for further discussions about this article

Top