How Mortgage Lenders Are Using Automated Credit Scoring

The United States has the largest and most developed housing finance system in the world, but automated credit scoring used in mortgage underwriting is a relatively recent development. It has become an area of increasing attention and uncertainty in this rapidly changing industry. With credit scoring here to stay, many wonder how it will affect the flow of capital to our nation's homebuyers.

Mortgage Fundamentals

Mortgage lending programs begin with a vision statement that describes the results the program would like to attain and a strategic plan to explain the program's objectives. Planning identifies the primary success factors that will enable the lender to meet and successfully respond to housing market realities. Because good mortgage lending programs and decisions rely on loan policy, the lender builds the organization's policy next.

Credit scoring is one of several automated devices that assist the lender in verifying or qualifying credit information about a customer. It is the first step taken by the lender after receiving an application for credit.

When the credit analysis is complete, evaluation and decision-making follow. The concern of lenders is that good mortgage credit decisions are complex and require hard work. Evaluating and ascertaining what risks are present and how to safely manage those risks are at the core for the lender. Loan policy guides the lender in the consistent use of information from many sources, including mortgage scores, when making a loan decision.

Market Realities

There are different types of mortgage credit scoring methods, but all credit scoring systems attempt to forecast loan performance; manage credit risk; predict the customer's ability and willingness to repay as scheduled; and forecast delinquency, default and bankruptcy rates.

Lenders must satisfy the loan volume, quality and profitability benchmarks contained in the lending program objectives. This includes a consistent flow of mortgage credit applications through the program that can bring the rate of return to the lender as set forth in the business plan.

Consumer demands for fair, quick and affordable mortgage credit parallel lender demands for safety and profitability. Lenders are using credit scoring systems to satisfy the consumer's demand for extraordinary customer services, including speed, accuracy and price. It now is possible to apply for a loan over the telephone, at a kiosk in a mall or on the Internet and receive a response measured in minutes, hours or a few days. Consumers want fast credit decisions, and, without the help of automation, decisions are time-consuming and expensive.

Looking for Patterns

Mortgage credit scoring systems attempt to find patterns in the credit histories of groups of individuals as reflected in credit bureau records. All data used in automated mortgage credit scoring are historical.

Mortgage credit scores use statistics to predict how the majority of loans with common characteristics in a broad group of the population will perform in the future. Performance in this case means loan defaults. Automated data can easily show trends in group behavior and how behavior may change over time. Much of the research conducted on mortgage credit scoring indicates that credit scores accurately predict the performance of the loans in the future and that scoring increases the accuracy of risk assessment.

Credit scores do not predict how individual borrowers may behave in the future but rather that the majority of persons with similar characteristics usually behave in predictable patterns. It may seem like splitting hairs, but there is a difference between the predictable performance of a group of loans with common credit bureau characteristics and an individual's behavior and risk. For example, most people in a movie theater will exit at the end of the movie while only a few will stay for the movie credits, or the majority of people stop at red lights but a few run them consistently. Credit scoring is useful because the greater the ability to predict aggregate or group problem credit behavior, the better chance there is to manage risk.

Issues of Credit Scoring Related to Mortgage Lending Programs

In all walks of life, there will always be exceptions to the statistical norm, and this fact has implications for mortgage lending programs and mortgage scoring usage. Research conducted at the Board of Governors of the Federal Reserve System raises several concerns about credit scoring. The research recognizes the benefits credit scoring systems may offer lenders and borrowers. However, the study states, "Its use raises economic and policy issues, including the ability of credit scoring to accurately quantify an individual's credit risk and the effects of scoring models on credit flows to lower-income and minority neighborhoods."

For individual borrowers, scoring does not include unobservable influences and could have incomplete information. These may include payment history for rent and utilities and participation in informal credit markets. For example, utility and rent payment history can be a substitute when the borrower lacks a consumer credit history.

Credit scoring may not allow for those conditions or variables (for example, a change in an individual's behavior through education, regional economic conditions or natural disasters) that may affect the individual customer. As one of the five "Cs" of credit, conditions may be altered that can affect the standard predictability of an individual's credit behavior. There may have been a general economic downturn and increasing unemployment rates in the region.

Credit scoring can be fair only to the extent that the information is accurate and the model is applied appropriately. A score may be accurate only to the extent that credit bureau information is accurate and complete. A scoring model developed for one product type, population group and loan term cannot be applied to a group or product with different characteristics and maintain its reliability.

Anecdotal evidence may indicate that credit scoring is being used as the only tool for credit evaluation when making a decision on a mortgage application. According to Fair, Isaac and Co., this is a misuse of the scoring system and may result in unsafe lending. Lenders are counseled by Fannie Mae and Freddie Mac that credit scores alone are not to be used in approving or denying an application for mortgage credit. The decision-making system must use a complete credit profile and not just the credit score.

"Our advice on using credit scores is that they can be used to quickly process the vast majority of borrowers, freeing up resources and time for lenders to focus on the more difficult loan files," said Henry Cassidy, senior vice president of single-family risk management for Freddie Mac.

Benefits

Technology can be a tremendous help when the number of variables that affect a decision increases. Automated scoring serves as an initial and primary screen for applications. In theory, automated scoring is used to screen large quantities of credit information for applications that may be approved with the least amount of credit analysis and decision-making resources. Such a screen could remove human subjectivity, especially bias, prejudice or overt discrimination. Applications scoring lower require further credit analysis, probably using manual or traditional underwriting.

The benefits of an automated screen of applicants can include lower acquisition costs, more objective and consistent decisions, and increased speed of decisions for the customer with less paperwork. Scoring also may contribute to the ability of a financial institution to make more loans that are profitable by decreasing the likelihood of default and collection costs. This could increase the amount and flow of funds to more borrowers. Fewer defaults could mean more money available for borrowers and mortgages.

Subscribe

Keep up with what’s new and noteworthy at the St. Louis Fed. Sign up now to have this free monthly e-newsletter emailed to you.

2015 Federal Reserve System
Community Development
Research Conference