With credit scoring now in wider use, customizing the technology to mortgage companies' needs has become a top priority in the industry.
Reviewing credit scores can seem like a "house of mirrors," acknowledged David Shellenberger, product manager at Fair, Isaac & Co., at a forum sponsored by the firm this week.
Interest in a borrower's credit risk can come from a number of sources, including brokers, insurance companies, banks and thrifts, rating agencies, servicers, and investors. As a result, borrowers are often scored many times and from many perspectives, Mr. Shellenberger said.
"A lot of the mortgage industry still doesn't understand how scores relate to performance," said Dan Feshback, president and chief executive officer of Mortgage Information Corp., San Francisco.
While consumer lenders have used credit scoring for some time, the mortgage world started to incorporate credit reports in 1995 when Fannie Mae and Freddie Mac began requiring that Fair, Isaac credit bureau risk scores, or Fico scores, be used to evaluate the loans the two companies bought from lenders.
A credit score is a snapshot of an individual's credit profile, Mr. Shellenberger said. It incorporates a summary of all the data on a credit report into an index that tells the propensity of a borrower to pay bills.
Mortgage companies are moving toward "a systematic approach to assessing the credit that they do," said Ken A. Palla, mortgage industry initiative manager at Fair, Isaac. Credit risk management is coming into the picture as companies are increasingly hiring credit risk employees, he said.
Credit bureau risk scores rank borrowers based on the likelihood they will repay all their credit obligations, but more specialized scores are available to rank borrowers by the likelihood of repaying mortgage obligations.
Lenders may use pooled-data scoring models, which predict behavior based on data provided by numerous lenders, or custom scoring models based on their own data.
The latter models have more predictive power, but some lenders do not have enough data on their own borrowers to create them. A minimum of 1,000 defaulted loans is required to develop the model.
Mr. Palla said use of models is becoming more prevalent as lenders venture into high-loan-to-value mortgages and other new products.
Richard A. Craine, assistant vice president of support services at United Companies Funding Inc., said his Bloomington, Minn., company hopes to use scoring technology to identify "a segment of the population that will perform strongly for us."
Marco A. Del Hierro Caraza, risk manager for mortgage and auto lending at Banco Nacional de Mexico, came to the conference to improve his bank's behavioral scoring model.
"We are allocating a lot of resources into behavior scoring since our biggest concern is still the collections," as opposed to the volume of new loans, he said.
But Mr. Del Hierro Caraza said credit scoring lets the bank's credit analysts decide whether to grant loans and is "a way to maximize our resources to keep the costs down.
"We don't have to have an army of credit analysts looking through every application," he said.