Fair Lending — Data Analysis

March 31, 2022 BY MQMR Blogger

Question: What type of data analysis should a mortgage lender perform when evaluating fair lending?


Answer: The following is a summary of the type of analysis a mortgage lender should perform:


Analyze lending applications and loan parameters for signs of discrepancies in any of the following areas: 

  • Loan Approvals/Denials
  • Loan Pricing (Fees and Interest Rate)
  • Loan Program offerings and guidelines


Data analysis should first be performed for an organization’s portfolio as a whole. Basic data that can be found on a company’s Home Mortgage Disclosure Act (HMDA) Loan Application Register (LAR) Summary sheet can begin to tell a story. For example, you will learn the following:

  • What percentage of applications do you receive from each racial, ethnic, gender, and age group?  Are the percentages what you would expect based on the jurisdictions where you lend, and the population demographics in those jurisdictions?
  • Is there a particular group that performs worse than other groups, in terms of applications received, applications approved or denied, and loan pricing?  If so, are there opportunities that present themselves to increase the applications coming in from that particular group, which could lead to more funded loans?


In addition to analyzing data from an overall portfolio perspective, a company should break down its data to the following levels:

  • Purchase loans vs. refinance transactions
  • Conventional loans vs. Government Loans
  • By Channel (retail vs. wholesale vs. correspondent, as applicable)
  • Geographic area (by region or state)
  • By Branch Office, as applicable
  • By broker, as applicable

If statistically significant disparities are uncovered for one or more protected classes, perform regression analysis to account for credit factors, such as the following:

  • FICO Score
  • Loan-to-Value Ratio (LTV)
  • Debt-to-Income Ratio (DTI)

If the regression analysis does not explain all of the statistically significant disparities that exist, identify matched pairs for comparative file review. A matched pair would be two customers, with similar credit profiles, who applied for loans in the same state, for approximately the same loan amount, where both customers should, based on their credit, be approved. And yet, a majority group customer was approved, and the minority group customer was denied. The goal of the comparative file review would be to explain why the majority group customer was approved, and the minority group customer was denied. If a legitimate, non-discriminatory reason can be found, that should be documented.  If not, feedback should be provided to the appropriate staff, and the company should consider whether changes to policies and procedures may be needed to mitigate fair lending risk.