While credit scoring is the dominant assessment tool the market uses, not much is known about how it works. MPA went in search of answers
Apparently there are as many ways to skin a cat, as there are to credit score a loan application.
But the one most commonly used is known as a FICO, because of its origins with a US company called Fair Isaac and Company. They were the first to come up with idea of a credit scoring tool, nearly fifty years ago.
A credit score is a statistical model used to assess the likelihood a credit application has of going into default. It does so by making a mathematical assessment of a customer's willingness to repay a debt. "This, rather than assessing their servicing capacity or security is the key element of a credit score," says Nelson Yiannakou, NAB's head of mortgage credit intelligence.
Think of a credit score like a grade at school; calculated from test scores, homework, participation and anything else appropriate. Weight each category according to its importance and roll all of them into a single number. The score.
Yiannakou says credit scoring's relevance in the third party channel rest on brokers' reliance on lending products from the major banks. "In the end all credit applications are subject to an assessment process. In addition, some brokers source their own funds and have their own credit scoring tools," he says.
Peter Hall, country executive at Genworth Financial agrees. So much so that Genworth has recently implemented a scoring tool to assess loan applications.
Genworth bought the license for the software technology to build the model and developed the score in-house, using international best practice standards learned in the US and Canadian markets.
It has all the key attributes of an unsecured loan score, but is welded into a mortgage framework. "It has a fraud score included, which makes it more robust," he says.
But, Hall adds, it is the quality of the data that runs through the model that makes the score either meaningless, or accurate. He runs a combination of Genworth's own, and data from the credit bureau, through the score.
Yiannakou agrees. Predictive models use data from the past to predict the future, so scoring models are only as good as the comparable data sets that populate them. "One of the problems that lead to the sub prime market crunch is that many of the approved credit scores used data drawn from an earlier, benign, environment," he says.
To the extent that mortgage lending differs from unsecured lending, the standard 'approved/ declined/ refer' decisions common to credit card or personal loan application assessments are often replaced for mortgage applications, by a risk ranking score.
This risk ranking score is used to set policy rules and security requirements, or to support risk-based pricing or risk-based servicing decisions.
Hall says implementing the credit scoring capability at Genworth is a natural progression of wanting to improve operational efficiency, and, it has.
Turnaround times have been reduced from hours - even days in some cases - to seconds, where the system generates immediate responses (provided all the relevant boxes are ticked).
Genworth's credit score process is straight-forward. No loan is declined. It's either approved or 'recommended refer'. The 'recommended refer' cases are routed to an underwriter for further scrutiny. "However we've found that 60-70% get approved automatically," says Hall.
It's a new system and this approval rate is improving all the time, he says, as the new system beds itself down. 'We cut through the red tape to provide tangible benefits for our clients,' he says.
And scoring, Yiannakou adds, is all about slashing excessive administrative procedure. It is the most efficient way of assessing credit since scoring produces a consistent result as the same decision is reached every time. "And it does so in a quicker time," he says.
Also, it does all of this while providing a very granular picture of risk to lenders. This allows them to set detailed risk lending strategies, and provides for a certain amount of procedural flexibility. Particularly in areas where higher risk thresholds allow lifting of risk layers. "Lenders are able to make well informed risk reward decisions," Yiannakou says.
The 5 C's
The willingness of a borrower to repay is often determined by 5 criteria:
1. Capacity
2. Capital
3. Character
4. Collateral
5. Credit history
Nelson Yiannakou, NAB's head of mortgage credit intelligence on the benefits of credit scoring
1. It produces a consistent result as the same decision is reached every time.
2. It is efficient. There is a quicker time to 'yes'.
3. It enables direct channels and leverages broker channels.
4. It reduces bad debt, for those who've converted from manual to automated channels.
5. It facilitates the automatic capture of data for management information systems (MIS) purposes; for both products and processes. Without MIS, lenders might not know what they didn't know.
Nelson Yiannakou on the limitations of credit scoring
1. It is not always correct. It uses a pooled portfolio data in making its assessment. In a risk group where there is a 10% chance of default we know one in ten will go bad, but not which one.
2. Predictive models use data from the past to predict the future. The models are only as good as the comparable data sets they are using. One of the problems in the sub prime market crunch is that many of the approved credit scores used data drawn from an earlier, benign, environment.
3. It requires ongoing monitoring and adjustment. Setting and forgetting is a mistake.
An approximate breakdown of how credit scores are determined:
- 35% based on your payment history.
- 30% based on outstanding debt.
- 15% based on the length of time you've had credit.
- 10% based on the number of inquiries on your report.
- 10% based on the types of credit you currently have.
