Home Mortgage Disclosure Act Virtual Tech Sprint

March 22-26, 2021

More than 100 participants from community groups, financial institutions, compliance software developers, academia, government, and the tech sector formed 17 teams and worked throughout the sprint week on two tracks. One track focused on furthering development of HMDA data publication capabilities and data products, and the other track focused on improving the process of submitting HMDA data to the Bureau.

HMDA Data Publication Solutions

During the tech sprint, several teams worked to develop consumer-facing tools using HMDA data. The teams proposed a variety of solutions, such as tools that would:

• Allow consumers to enter, for example, zip code, credit score, loan purpose, and loan type to compare lenders on by interest rate, fees, and other measures from available HMDA data.
• Enable consumers to compare mortgage rate quotes they received with quotes other consumers had received and sent to a central database, and when combined with HMDA data, could identify potential “lending deserts.”
• Track customers’ feedback through a survey and use a scoring algorithm to identify the aggregate consumer sentiment.

Other teams worked on solutions to improve the user experience with HMDA data. These teams proposed solutions such as a:

• HMDA data analytics and visualization dashboard.
• Revamped HMDA front-end that included implementing plain-language revisions and documentation, closing gaps between tools and application programming interface (API) usage, and highlighting starting points for typical HMDA use cases.
• A single line of code to combine specific HMDA parameters, including states, years, borrower race, action taken, and lending institution to increase the accessibility of HMDA data and make it easier for users to download and analyze.
• Variety of disclosure options for credit scores that are currently not disclosed to the public, including state-level aggregations for certain loan types and applicants, disclosure in bins or ranges, and aggregate disclosure by geography using graphs or maps.

Some teams focused on HMDA data analytics to solve issues and barriers to learning through HMDA data. There ideas included a:

• Platform that used artificial intelligence/machine learning to identify the potential disparities in the action taken on an application by predicting the outcome of submissions using a peer group.
• A Rural Reporting Reliability Index to capture the scale of the HMDA data unreported for rural areas as a result of current HMDA coverage parameters by linking to other databases.
• A tool intended to help public sector officials to direct public investment or subsidies in a way to stimulate private sector lending and investment in neighborhoods in need of credit and capital. They did this by combining, at the census tract level, HMDA data with data on two HUD programs: the Home Investments program and the Community Development Block Grant program.

HMDA Data Submission Solutions

Teams presented several ideas to improve the submission of HMDA data to the Bureau. These ideas included:

• Exploring methods to speed up processing time of submitted data, especially for lenders reporting large loan/application registers (LARs).
• Techniques to ensure the integrity of the data submitted and to avoid data errors.
• Methods to create a customized submission experience for lenders.
• A method to check HMDA data collected by the lender against credit bureau data before the LAR is submitted.
• Exploring the use of LARs data in a blockchain involving the Bureau.

Other teams focused on the use or improvement of APIs in the submission process. These improvements included:

• A machine-learning based API dashboard proposal that assists consumers in shopping for a mortgage.
• Exploring an API developer’s portal with searching, archiving, management, and reporting functionality to enhance the existing CFPB APIs.