|
Writing mortgages can be a risky business, risks which every mortgage grantor
must manage and minimise. Whilst every company aims to reduce, or even eliminate
risk, the mortgage business has its own specific challenges - challenges that
credit scoring is addressing with increasing effectiveness.
The economic recession has certainly made credit grantors acutely aware of the
need to improve risk management. For the mortgage industry, some of the most
telling signs of the recession were highlighted in the increasing number of home
repossessions.
So how can credit scoring help reduce these risks? Scorecards offer a
computer-based solution to aid decision making and provide an accurate and
consistent method of appraising the mortgage applicant's ability to honour
repayments.
But why hasn't mortgage scoring played a more significant role? The answer to
this question lies in the relatively high investment needed to develop,
implement and maintain a bespoke system built around specific requirements.
Furthermore, lack of skill and experience in the relevant technologies meant
that many building societies were content to stay with the methods they had
always used and become accustomed.
To minimise risk and maximise growth opportunities, many mortgage grantors are
under pressure to change the way they currently do business.
Benefits
So what are the benefits to the mortgage grantor in developing and implementing
credit scoring?
One of the key benefits of developing a scorecard is that it enhances the
quality of decisions reached. Equally, adopting scoring technology brings an
improved level of quality and consistency to the underwriting process and
reduces decision turnaround times considerably.
Recent developments show a shift in the underwriting processes. The key to
securing a successful borrower now lies not in the value and security of the
property itself, but on the ability of the borrower to make payment. Scoring
technology has facilitated this shift, ensuring that mortgage grantors can
underwrite individual applicants based on a credit score.
Finally, every building society, indeed every business, aims to keep ahead of
competition, increase profit margins and reduce bad debt. Scoring technology
will help ensure that these objectives can be achieved and measured.
Initial considerations
Mortgage lenders weighing up the benefits of developing a scorecard must
consider many complex angles.
Building societies are faced with a choice: to develop a scoring system in
house; to develop a bespoke system with a specialist credit scoring agency; or
to buy an existing scorecard that can be made to fit their business model.
Few mortgage grantors can develop a system internally to compete with the
cumulative expertise of scoring agencies - but many still do. Scoring agencies
have proven track records with many of the major building societies and whilst
the bespoke system will cost from £50,000, the costs of doing it internally
would probably remain unknown.
Equally, once the scorecard has been developed in-house and substantial time and
resource has been spent in the process, there is no complete guarantee that the
scorecard will fit with the business.
In spite of the advantages of the bespoke solution for many lenders, even those
who place great importance on scoring, the time, effort, and capital expense
associated with the development is simply too high.
Mortgages being written now, however, are a future investment and credit scoring
should also be viewed as a long term investment. The most pragmatic solution
would be to adopt a generic system which has been developed as an alternative to
the tailor-made scorecard.
Whilst a generic system cannot reflect the unique requirements of individual
building societies, it can apply the rules learnt and built-up over the years to
assist risk management
Fundamental to the development of a scorecard is the technology to implement it.
The mortgage grantor may have a suitable computer system in place, otherwise it
will be necessary to budget for such a system.
Also vital to success, is the availability of credit reference data - how this
is accessed by the building society, manually or computer to computer link, will
strongly influence the criteria that can be used and must be considered in the
initial stages.
Scorecard monitoring is a final consideration, and processes must be identified
in the early stages. Monitoring applicant profile and customer performance is
crucial in ensuring the expected levels of discrimination.
The development process
Despite being deceptively similar on the outside, each building society is
unique in terms of the product portfolio they offer to customers. Whilst a
generic system may be appropriate to some lenders, a business wishing to
capitalise fully on the benefits of new scoring systems will opt for a scorecard
which is equally unique.
The structure of the scorecard is determined solely by the nature of this
product portfolio: operational considerations, long term objectives and
availability or scarcity of data.
It is essential that every piece of information is accessed for evaluation.
Accounts that have been deleted from the main billing file must be retrieved
from archive records so that they can be included in the process.
The fundamental concept of scoring is based around the notion that the future
broadly resembles the past. Since the scorecard allows the mortgage grantor to
make future decisions based on past activity of applicants, without an accurate
picture of the past, future decisions become harder to predict.
Masterfile analysis
Having established the elements of the mortgage portfolio, the next step in the
development looks at constructing a detailed study, or masterfile, of that
portfolio. At this stage, the home loans credit grantor supplies a copy of the
'current billing' file to the organisation involved in the scorecard
development.
It is important that the file being analysed contains details of accounts that
are live, as well as those that have been closed and written-off.
Once the mortgage portfolio has been constructed, the stages of analysis begin.
The scorecard developer begins a thorough investigation of the current appraisal
processes, noting any differences between the current approach and previous
methods which could be selected for sampling purposes. It is important at this
stage, that the building society defines what makes up good and bad accounts and
explores the time taken for accounts to mature.
In identifying bad accounts, attention is usually turned immediately to the
repossession file which details customers who are obviously an above average
risk.
But how do the customers who represent potential risk fit into the equation?
Surely, customers who have defaulted on payments on more than one occasion need
to be considered since potential risk can easily become proven risk.
Creating the Database to Support the Scorecard Development
From the masterfile analysis stage, 1,500 to 2,000 accounts are randomly sampled
from each of the good and bad definitions. The mortgage grantor then supplies
all of the information available on the accounts at the original time of
application.
In addition, 1,500 to 2,000 accounts rejected, and applications not taken up are
selected from the same time frame to estimate their behaviour, had they been
accepted.
Once captured, this information forms the basis of the development database and
the foundation on which individual behaviour can be predicted. To consolidate
the scorecard, information is drawn from the credit register files of a credit
referencing bureau and appended to the database.
The ability to incorporate credit data into the scorecard development is
essential in ensuring a more effective scorecard. This file contains some 60
million records consisting of in excess of 200 characteristics, such as:
1. County Court judgments, bad debt accounts and good accounts from the credit
grantors in the finance, banking, mortgage, utility and retail sectors.
2. Data derived from the electoral roll.
3. Pinpoint, a socio-demographic indicator.
4. Indication of a registered telephone at the address.
5. Other characteristics in the form of statistics which are combinations of
other data.
Statistical analysis
The first objective of the statistical analysis is to assess the strength of
each criteria as a predictor of repayment behaviour with weak or suspect
criteria being abandoned at this stage.
Data is then stratified which leads to fewer categories within criteria and
larger sample counts within categories. At this stage, the sample is then
re-weighted to reflect the actual proportions of applicants within each
performance definition.
Powerful statistical algorithms are then used to develop the final scorecards
and statistics to demonstrate their power and effect on key sub-populations are
used.
At this point, it is useful to clarify that only 80 percent of the sample is
used to develop the scoring system - the remaining 20 percent is set aside and
used to independently test or validate the scoring system.
Working together
The majority of the work in developing a tailor-made scorecard is carried out by
the scorecard developer. In order to ensure time scales (typically three to four
months) are met and the final scorecard is an accurate fit with the
organisation's requirements, the building society must be equally committed.
Implementation
Sourcing and establishing a database is a mammoth task but is a means to an end,
and not an end in itself.
Once the database has been set up ensuring that the scorecard is implemented is
crucial but often overlooked. Training and advice from the scorecard developer
on getting the most from the data will be essential in ensuring that future
decisions are made accurately.
In order for the building society of the 1990's to build and establish success,
credit scoring must play an active role.
Without these fundamental technologies in place, organisations will find it
increasingly difficult to manage risk and maximise business opportunities.
Mike Ryan is Associate
Director, Banking & Finance Sector at, Equifax Europe (UK) Limited
Source:
Credit Control Journal
|
