EconSouth (Third Quarter 1999)


s productivity innovations go, credit scoring has been as important to the banking industry in the 1990s as automated teller machines were in the 1980s. Just as output per bank worker — the most common definition of productivity — surely increased in the banking industry after ATMs began handling routine banking transactions, automated underwriting has allowed banks to write more loans more quickly and at less cost than ever before.

But some critics of the banking industry say that credit scoring leads to discrimination. They charge that the industry's reliance on a nameless, faceless and colorblind set of computer programs allows banks to satisfy the law while still discriminating against borrowers in low- and moderate-income (LMI) neighborhoods.

Recent research by Michael Padhi, Lynn Woosley and Aruna Srinivasan of the Federal Reserve Bank of Atlanta suggests, however, that credit scoring applied to small business lending not only does not increase discrimination against business in LMI neighborhoods, but it also seems to increase the level of small business funding available in these areas. The research by Padhi, Woosley and Srinivasan may have serious implications for banks as they endeavor to comply with the Community Reinvestment Act (CRA).

The Community Reinvestment Act
Twenty-two years after Wisconsin Senator William Proxmire authored the anti-redlining provision that would become the CRA (see below), its goal remains to ensure that banks serve LMI areas. In 1999 CRA compliance for large banks — defined as having $250 million or more in assets or being owned by a holding company with $1 billion or more in assets — is determined largely by three separate tests. A lending test examines banks' mortgage, small business and community development lending; an investment test examines the availability of grants to and equity participation in community development activities; and a service test examines branch locations and the provision of banking services to low- and moderate-income groups.

While the tests for CRA compliance are notably more specific than they were two decades ago, scoring the tests — assessing the results — remains a somewhat subjective exercise. Bankers argue that regulators use the threat of noncompliance to force them to make loans that don't make economic sense. Bank critics, meanwhile, argue that regulators are too easily swayed by bankers' reliance on credit scoring and other objective lending criteria. Moreover, bank critics argue that as lenders have increased their reliance on credit scoring, lending to low- and moderate-income areas has declined. It's this last charge that the Atlanta Fed researchers address.

Proxmire's Progeny

n June 7, 1977, after two days of debate, the U.S. Senate approved H.R. 6655, the Housing and Community Development Act of 1977, by a margin of 79 to 7. It was a mostly unexceptional bill, reauthorizing the popular Community Development Block Grant program for an additional three years.

But the Senate version of H.R. 6655 included a provision that had not been included in the House-approved version. At the behest of Chairman William Proxmire of Wisconsin, the Senate Banking Committee added Title VIII, an anti-redlining provision that directed banking regulators to consider an institution's record of meeting the credit needs of "its entire community, including low- and moderate-income neighborhoods" when evaluating the institution's application to open a new branch.

The report that accompanied the bill explained that Title VIII had been added because the committee was "aware of amply documented cases of redlining in which local lenders export savings despite sound local lending opportunities." And in arguing against an amendment by North Carolina Senator Robert Morgan to cut Title VIII from the bill, Proxmire said it was needed "to reaffirm that banks and thrift institutions are indeed chartered to serve the convenience and needs of their communities . . . and needs does not just mean drive-in teller windows and Christmas Club accounts. It means loans."

Ultimately, of course, Title VIII was retained in the legislation signed into law by President Jimmy Carter. But while the bill's main provision — the Community Development Block Grant program — would become a casualty of Washington's 1980s budget wars, Proxmire's anti-redlining provision would continue as the Community Reinvestment Act.

Adding it up
To better understand the results of this research, it's worth considering how credit scoring works and how it's said to discriminate. Credit scoring derives a single quantitative measure — the score — from a vast statistical sampling of past borrowers in order to predict the future payment performance of an individual loan applicant. For banks, credit scoring increases profitability by reducing the labor costs associated with reviewing a loan applicant's credit reports and financial statements. This labor savings not only compresses the approval time of loans, but it also frees personnel to perform other functions, including marketing more new loans.

But while credit scoring has been used for more than a decade to underwrite credit cards, auto loans and home equity loans, its use in small business loan underwriting is a much more recent development. The common factor linking these different types of loans is personal credit history. In the case of the small business owner, her personal loan payment history is a very strong predictor of her business loan payment prospects — especially if the loan is for less than $100,000 — and her consumer credit scorecard can be adapted with little difficulty.

Charges that credit scoring reduces lending in low- and moderate-income areas arise from concerns that credit scorecards and the databases they're built on can contain bias. These charges may seem ironic. For bankers, one of the real virtues of a credit scorecard is its objectivity: it allows the same lending criteria to be applied to all applicants without regard to their membership in a protected class. Moreover, income, race, gender, length of employment and home rental-ownership are not included in the leading credit scorecard for small business lending produced by Fair, Isaac & Co., a firm at the forefront of developing credit scoring models. Nevertheless, critics argue that credit scoring models unlawfully discriminate by assessing LMI applicants against a pool of more affluent non-LMI borrowers. Critics also charge that in small business lending and other areas in which the need for business-lender relationships is thought to be important, banks may use credit scorecards as a surrogate for close relationships.

What the Atlanta Fed found
Padhi, Woosley and Srinivasan tested these arguments of credit scoring critics. They examined the small business lending practices of 99 of the 190 largest U.S. banks lending in metropolitan areas in the Federal Reserve Bank of Atlanta's district.

The data for their research was built on a previous study by Srinivasan, Woosley and Scott Frame of the U.S. Treasury Department, which found that 63 percent of banks surveyed used credit scoring to underwrite small business loans. That study found that for loans of less than $100,000, all of the credit-scoring banks used scorecards; for loans of less than $250,000, 73 percent of those banks did. In addition, Srinivasan, Woosley and Frame found that just 42 percent of scoring banks used scorecards to automatically approve or reject loans, but 98 percent employed them as part of the loan decision process. Both studies considered banks that used scoring for part or all of the loan decision process to be credit scorers.

   What about

s large banks increasingly turn to automation in their dealings with customers, smaller banks may find a lucrative opening for old-fashioned character loans. Last year, in a speech before the Georgia Bankers Association, Federal Reserve Bank of Atlanta President Jack Guynn likened community banks to AAA, the American Automobile Association.

In that speech Guynn said, "You know how AAA works: you're in Hawkinsville, Ga., and you want to drive to Staunton, Va. There are probably a 1,000 ways to get there, but only two or three are any good. AAA will send you a map, highlight the best routes, and note all the trouble spots along the way: construction in Greenville, traffic in Raleigh and so forth.

"The financial system of the 21st century will be as big, as efficient and occasionally as encumbered as our transportation system. It will be a massive, interlocking network of technology and financial resources. And for the average customer, it could be easy to get lost. Small and community banks know their passengers, they know their destination, and they know how to get there. They provide an invaluable service and one that I think will be increasingly in demand."

By comparing the small business lending activities of banks that use credit scoring with banks that do not, Padhi, Woosley and Srinivasan were able to assess the impact of scorecards on small business lending practices inside and outside of low- and moderate-income areas.

They found that while nonscoring banks were likely to lend significantly less in LMI areas than in higher-income areas, the difference in loan levels for scoring banks was not statistically significant. In other words, the dollar value of small business loans made by credit scorers was about the same in LMI areas as in non-LMI areas when neighborhood and bank characteristics were statistically accounted for, whereas nonscorers lent significantly less in LMI communities.

In addition, the researchers found that credit scorers actually originate more loans — number of loans, not dollar value — in low-income tracts than in moderate- or high-income tracts. Nonscoring banks originated significantly fewer total loans in LMI neighborhoods. Finally, the researchers found that among credit scorers, the presence or absence of a bank branch in an LMI area has no significant effect on the level of small business credit available. For nonscoring banks, the location of a branch in the community is a significant positive for small business lending in LMI neighborhoods.

While Padhi, Woosley and Srinivasan have not attempted to assess any underlying causes with their research, their results suggest at the very least that scoring alone does not result in small business lending discrimination in LMI neighborhoods. Moreover, considering that the presence of a branch in an LMI tract does not have any impact on scorers' small business lending activities, credit scoring may actually serve as prima facie evidence of compliance with the CRA's lending test. As for the role of character or personal loans in small business lending, the Atlanta Fed research suggests that credit scoring might discourage discrimination against protected groups and communities.

What's next?
The Atlanta Fed research is the first such study on an issue with serious social, economic and political repercussions. While much more remains to be done on this issue — especially at the national level — the initial results suggest that the use of credit scoring in small business lending can help ensure fair treatment of small business borrowers.

Return to Index  |  Next