At least two developers of mortgage industry computer software have written programs they believe will improve underwriting dramatically. Key to acceptance, however, is the reaction of lenders and investors, including the Federal National Mortgage Association and the Federal Home Loan Mortgage Corporation.
The new programs are intended to make it easier for borrowers to obtain loans while more clearly identifying risks for lenders.
"There are three principal deficiencies in the current process," said one of the software developers, Jack M. Guttentag, also a professor of finance at the University of Pennsylvania's Wharton School.
"First, the underwriter's ability to adjust to a changing market is poor. You see the evidence of that in the behavior of underwriters during the 1988-90 period of problems in the real estate industry when standards, if anything, became more liberal.
"Second, the ability of most underwriters to distinguish among different types of mortgage designs - adjusted rate versus 30-year fixed, for example - is very poor.
"Third, accountability for underwriting decisions is obscured by the current system.
There is no way to distinguish between macroeconomic and microeconomic determinants of risk. The underwriter can't possibly be responsible for the former."
Guttentag said his system essentially limits the underwriter's responsibility to make a judgment about a particular parcel of land.
Macroeconomic projections would be the responsibility of the lender or investor, who would choose different scenarios to insert into the system.
"Within this underwriting system, the individual underwriter does not accept or reject loans," Guttentag explained.
An exception might be categorical rejections based on adverse credit reports, he added.
The underwriter, he said, places the different aspects of the loan into specific categories arid the computer projects all the risk variables, comparing them every month with the limiting values specified by the lender. If none of the limits has been breached, the computer then automatically approves the loans.
If the loan fails, the underwriter can ask the system which variables were violated, and when and what changes could be made in the loan so that it would pass. Recasting, for example, might involve buying down the rate, increasing the down payment or shifting to another type of mortgage, Guttentag said.
Guttentag, who published an article in Fannie Mae's Housing Policy Debate, Volume 3, Issue 1 (see The Mortgage Marketplace, July 6, page 4), said he has had difficulty getting the idea across to potential purchasers but he is still trying.
But another software purveyor was enthusiastic about Guttentag's ideas and said his firm, HNC Inc. of San Diego, is trying to meet the same objectives with its own product.
Nick van der Schalie, director of mortgage products, said his firm has developed 13 models that cover variables on properties for more than 90% of the geographic areas of California.
He acknowledged the role of Fannie Mae and Freddie Mac in gaining acceptance of such a system. HNC has plans to present the programs to both companies, he said.
Officials of the two secondary market agencies were unavailable for comment on the proposals.
Some experts have proposed that artificial intelligence systems be developed for underwriting. Artificial intelligence involves "teaching" a computer to make the same calculations currently done by humans.
That may be acceptable, said D. James Croft, executive director of the Mortgage Asset Research Institute in Reston, Va.
"But if all that is done is teach computers how to perform underwriting the way it is done now, there is hardly any point to it. A system like Guttentag's could move things along. It should be tested by Fannie Mae or Freddie Mac to see if it's the answer."