WASHINGTON - A subsidiary of Moody's Investors Service said it will launch today a product that it claims will be the first step toward the securitization of middle-market commercial loans.
The product, RiskCalc Private Model, is a Web-based program that estimates the probability that a private firm will default on a loan. Its developers used a database of 28,000 private firms' financial statements and 1,600 private firms' loan defaults to flag characteristics common to bad loans.
"This large number of defaults allows us to identify risk factors statistically, something that we don't believe has been done before," said Moody's Risk Management Services vice president Eric Falkenstein.
Moody's hopes the ratings produced by the Private Model will become the benchmark for a market that does not yet exist: trade in securitized middle-market loans.
A majority of private companies with $100,000 to $100 million of assets are not covered by private debt-rating firms, so there is no standard measure of their debt's risk level. There is no secondary market for that debt because without such benchmarking, securitization is impossible.
Assuming that it produces consistent, accurate assessments of default probability, the Private Model could create a "sea change" in middle-market lending, said Kevin M. Blakely, executive vice president of KeyCorp and chairman of the lender trade group RMA.
"I don't think a lot of people realize the potential that a consistent and reliable rating system will have on middle-market lending," he said. "I have to think what the scorecard companies did for consumer lending systems, this will do for middle-market credits."
Others took a more wait-and-see approach.
"I have not yet seen a private-firm model that works very well," said Alok K. Sinha, a senior vice president with Bank of America who specializes in risk modeling.
"It's fair to say that the more defaults you have in your database, the better you are going to be able to coordinate the factors leading to default," Mr. Sinha said. "But that by itself is not going to make the product more exciting. You have to see what kinds of correlations they have been able to extract from the data."
KMV Corp. of San Francisco; Loan Pricing Corp. of New York; and Zeta Services of Hoboken, N.J., offer tools to assess private firms' default risk, but Moody's says all three rely largely on the default experience of public companies.
"We're the only model that is really estimated and calibrated to private firms. We're a little more apples-to-apples," Mr. Falkenstein said.
Moody's appears to have another advantage over its competition: It owns the spreadsheet software that 85% of U.S. banks use to assess their borrowers' balance sheets.
According to Mr. Blakely of KeyCorp., a risk rating program that functions as an add-on to that software could dominate the market.
Moody's expects to charge banks an annual subscription fee of about $25,000 to use Private Model in commercial lending operations. The fee will increase on a sliding scale, depending on the size of the loan portfolio.
At first, subscribers will use a password-protected Web site to enter financial information about a potential borrower, and receive an estimated default probability - the percentage chance that the loan will go into default in one to five years.
Mr. Falkenstein said that, while most banks have a system for gauging their commercial loans' risk, "I don't know of any bank in the U.S. that has this much data to validate any score that they have."
Moody's also claims that the program may help banks comply with changes to international capital rules expected this year from the Basel Committee on Banking Supervision. The committee has said that it may allow banks with sophisticated internal risk modeling systems to set their own international capital levels. Mr. Falkenstein said that if it is used in conjunction with the RiskCalc Public Model, which calculates public companies' default probabilities, Private Model could help banks meet that systems criterion.
"Whenever you have internal models, you are going to need a benchmark.