Find It and Fix It

The File Comparison Approach
The fair lending examination procedures provide financial institutions with clear and detailed guidance on the design and structure of a fair lending examination. The instructions include a description of how to determine risk factors and set the scope of the examination. What they don't tell you, however, is how to actually conduct a file comparison analysis.

The techniques you use for file comparison are as important as determining where your greatest risks are and setting the scope of the examination. This article contains guidance for the design and execution of a file analysis.

The goal in this process is to find files that are comparable except for a few factors - which could indicate discrimination or could be explained by other facts. To find the files you want to compare and the issues or areas you need to evaluate further, the first step is to create a database. The database is compiled from a sample of files which you select strategically.

Once you have the files, pull data elements from them and enter them in a database which you can (and will) use to sort and analyze the data - the facts drawn from each loan file. There are two important steps in selecting files and compiling the data. First, you target the file group from which you will pull the sample. Second, you decide what issues or practices you are looking at and select relevant items of data to include in your data base.

Set Your Scopes
You need to know what you are looking at. Do this by focusing your attention on one area - and don't compare what you see to non-comparables. In choosing the criteria we use to set our scopes for the analysis, we use factors that may cause a difference in result without being influenced in any way by discrimination.

Target the products that expose the bank to the most risk. Risk is created in several ways. First, risk is directly proportional to the size of the portfolio - the bigger the portfolio, the more the risk. Second, risk is increased by the newness of the product - look to see whether everyone understands the product and delivers it in the same way. Finally, risk is also closely related to how decisions are made. When individual loan officers have decision authority, there is a higher likelihood that there will be inconsistent decisions.

First, choose the product that you will analyze. Make the selection based on the risk analysis detailed in the examination procedures. When setting the loan product category to analyze, subdivide broad general categories into groups or types of loans that have common features. Also use only specific products within those categories. For example, you could divide mortgages into several broad categories based on the term of the loan. Thirty-year mortgages may have different terms than 15-year loans so control for the term of the loan (number of payments) to be sure that you have a homogeneous group.

Another major category division would be to separate variable rate from fixed rate. Fundamentally, these are not the same product. Don't create misleading results by trying to compare them. Analyze 15-year fixed rate loans in a single category. Then look at 15-year variable rate loans. There are numerous data elements that may arguably confuse the picture. If factors that make a difference in loan terms are not used to separate files, you will use them later to understand any differences that you find.

Collecting & Sorting Data
In your basic approach, you need to collect enough information to either identify a possible problem or a possible explanation. Shortcuts may cost you some serious time later on - when you need extra information to understand the results. It is more prudent to include data than to simplify the data collection process. You don't always know at the beginning where the information will take you.

Fundamental to the data you collect is a concept of how you want to look at it. For example, one of the risk analysis questions the examiners ask relates to decision centers. Who makes decisions, when, and how. Any decision can expose the bank to fair lending risk. Therefore your data analysis should help you look at and compare decision centers. Be sure to collect the decision center where the loan was made. This may be the branch or a more centralized location.

Within a specific decision center, you may also want to compare decision makers, so include the identity of the loan officer. For vehicle loans, you may also find it useful to track the dealer. This information could help you find dealers that may be allowing discriminatory pricing or sales practices.

You will need to know a fair amount about the borrower in order to make valid comparisons. In addition to the obvious - the borrower attributes, such as gender, race, age, and marital status that relate to prohibited bases, you may also want to compile policy-related attributes such as whether the applicant is a bank customer or a previous home-owner.

It may be relevant to know how the customer came to the institution - the loan source. If that is the case, include the identity of the real estate or mortgage broker, the car dealer, or other source. Also note whether an applicant was a previous bank customer.

Always include basic information such as the application and loan date. This enables you to evaluate the loan processing time for each customer. It also enables you to determine what rates were in effect when the loan was made, or whether there was a marketing campaign that offered special terms.

Other information you select should reflect your bank's policies. Choose the attributes that are most important to your lending decisions. This may include information such as the applicant's credit bureau score, and debt ratio.

Include the basic terms of the loan, such as the contract rate of interest, APR, amount financed, and number of payments. These are the loan results that you will study after sorting the information by possible discriminatory factors or actors.

If you are preparing a database using automobile loans, be sure to include factors such as the vehicle type and age, and dealer issues such as a dealer premium. The loan terms may also be affected by factors such as a customer down-payment or trade-in. A customer asking for $8,000 on a vehicle is better qualified than a customer asking for a $16,000 loan on the identical vehicle.

Now What?
Once you have this elaborate spread sheet, what should you do with it? This is where it gets interesting. Use your software to organize the data into a series of reports. If you have concerns about inconsistent decisions between branches, one of your reports should sort loan decisions by branch and then several other key factors that may reflect discrimination, such as APR, customer age, and loan terms.

Get creative. Most software programs will let you prepare reports using any of the data elements you have entered. The time-consuming part is selecting files and entering the data. Generating the reports is easy.

Once you have the reports, study them. Look for patterns that need explanation. Use the reports to identify what files and what concerns you will specifically look at. Now you can actually compare files and find the reason - or lack thereof - for the differences you see in the reports.

Finally, understand what your "findings" really mean. Sometimes a pattern is the result of legitimate bank or customer behavior. In some cases, it may be discrimination. Understand your findings and design compliance program responses that are tailored to the situation. Don't over-react; react constructively.

ACTION STEPS

Study the fair lending examination procedures. Note the risk factors in your bank.
Determine, based on your risk analysis, what loan products you will analyze.
Follow the steps in this article: compile the data, generate reports, and study them.
Investigate any questionable or curious patterns you find. Understand what happened - what the data tell you.
Set a calendar for the entire year. Do this type of analysis at least twice - preferably each quarter.
Talk with lenders and branch managers to understand the commercial or demographic behavior that underlies any patterns. For example, know which dealer sells used cars rather than expensive new models.