- The proposed rule, if adopted, would apply to mortgage originators and issuers of mortgage-backed securitizations that rely on automated valuation models (“AVMs”) to determine the value of real estate collateral.
- The proposed rule is currently based on a flexible, principles-based approach, but examples of more specific policies and practices or guidance may be included in the final rule to allow institutions to set quality control standards for themselves.
- Although market participants may already be complying with the requirement for a non-discrimination standard, they should take care to address any additional non-discrimination requirements not covered by current practices and guidelines.
What are AVMs?
AVMs are algorithms designed to generate market value estimates for real properties ranging from commercial warehouses to residential homes. The concept of AVMs captured the public’s attention because real estate platform companies, such as Zillow and Redfin, employ AVMs for value estimates of property listings and to streamline real estate transactions.1 Not only are AVMs used for sales of real property, but they have also become integral for institutional lenders who rely on AVM-generated values to originate mortgages. Some examples of AVMs include: Home Value Explorer (provided by the Federal Home Loan Mortgage Corporation); Home Value X (provided by CoreLogic); AVM Insight (provided by Equifax); and Zestimate (provided by Zillow), which is freely available to consumers.2 These AVMs are widely used in connection with making credit decisions and securitization determinations, whereas some activities—such as lenders employing computerized tax assessments to verify valuations, a certified appraiser using AVMs to generate an appraisal, reviewing completed collateral valuation determinations and providing broker opinions of value—are not considered AVM uses subject to the proposed rule.
Proponents of AVMs point to the efficiency and accuracy with which AVMs are able to estimate real estate values. However, because using AVMs may lead to unintended consequences, such as privacy and transparency risks or discriminatory practices, regulators are paying closer attention to the adequacy of current quality control standards with respect to AVM usage.
Background
In June 2023, the Office of the Comptroller of the Currency, the Board of Governors of the Federal Reserve System, the Federal Deposit Insurance Corporation, the National Credit Union Administration, the Consumer Financial Protection Bureau and the Federal Housing Finance Agency (collectively, the “Agencies”) proposed a rule to reinforce quality control standards when AVMs are used by mortgage originators and secondary market issuers to determine the collateral value of a securitized mortgage or to determine whether to waive the appraisal requirement for a mortgage origination.3 Further, the proposed rule covers certain third-party providers, such as servicers who process loan modifications,4 acting on behalf of such originators and issuers if those third-party providers relied on AVMs to “determine the collateral worth,” which covers using AVMs to make credit decisions and confirm collateral valuations.5
If the proposals are accepted, then the Agencies will require certain mortgage originators and secondary market issuers to adopt policies and procedures to:
1. Ensure a high level of confidence in estimates produced by AVMs;
2. Protect against data manipulation;
3. Avoid conflicts-of-interest;
4. Require random sample testing and review; and
5. Comply with applicable non-discrimination laws.6
Response from Market Participants Subject to the Proposed Rule
Entities Within Scope
A significant number of responses to the proposed rule from market participants have focused on the scope of its application and whether and why certain entities should be subject to or exempt from the AVM quality control standards. In its broadest application, the rule would require all entities, including mortgage originators, appraisers, secondary market participants, servicers and securitizers, to adopt policies consistent with the newly proposed standards. The rule would also apply to government sponsored entities, including Fannie Mae and Freddie Mac, with respect to mortgage origination and, in some cases, subsequent securitization if AVMs are used for any revised valuations.7
However, certain commenters, such as the American Bankers Association, national credit unions and smaller market participants, expressed concern with the proposed rule because it would over-burden and discourage AVM use in mortgage origination by imposing additional layers of compliance. Furthermore, these participants claim that it would be inefficient and ineffective to require each mortgage originator to perform quality control reviews for AVMs because, in addition to the inefficiencies of the proposed regime, the AVMs’ technicalities and processes are usually opaque to the originators. Therefore, some market participants have emphasized that that rule should not directly affect AVM users, such as mortgage originators, but rather should apply directly to AVM providers who have a better understanding of the technology.8
Flexible, Principles-Based Approach
The majority of market participants also voiced concerns that a more prescriptive rule would restrict their efforts to establish realistic risk management practices and procedures.9 In addition, because individual guidelines for AVMs are already in place, market participants suggested that the proposed rule should be sufficiently open-ended to be compatible with each participant’s guidelines.10
However, some market participants, especially smaller entities, have expressed the need for additional guidelines for compliance to avoid unnecessary over-compliance and maintain fiscal efficiency. These participants believe that they would benefit from specific minimum standards and consistent reporting elements for quality control factors.
The Non-Discrimination Factor
Although some participants have voiced support regarding the broad application of the non-discrimination factor and commented on the need to articulate baseline standards and examples of compliance, certain trade groups rejected the need to enumerate the factor because all mortgage-related transactions are already subject to long-standing fair lending laws. Therefore, the addition of the non-discrimination quality control standard may be duplicative of laws and rules already in place to uphold non-discriminatory practices.12
As a large number of comments weighed in on the non-discrimination factor, the Agencies have also shown interest in how to best implement quality control measures to eliminate sweeping biases and discriminatory practices. For instance, in September 2023, the Consumer Financial Protection Bureau released the “Consumer Financial Protection Circular 2023-03” in September, which clearly states that creditors may not solely rely on algorithms and models created “using artificial intelligence or complex credit models” because of the potential biases and discriminatory consequences of AVM use.13 Such examples of the Agencies’ interest indicates that non-discrimination is a focal point in their efforts to reinforce quality control standards of AVMs.
Final Rule
The Agencies are tentatively considering a 12-month timeline for implementing the new AVM quality control requirements. However, in light of several comments that highlighted the need for an extended timeline for certain businesses and small entities, the Agencies may consider a longer implementation timeline.14
Conclusion
Given the benefits of using AVMs and the growing reliance on them by real estate platform companies and institutional lenders, we expect that the discourse on AVMs’ long-term ramifications and requisite quality control standards will continue amongst the Agencies and various market participants. While it is widely recognized that AVMs improve accuracy and efficiency in determining collateral values, it is also true that the proliferation of AVM use may lead to a cascade of negative consequences, such as transparency concerns and discriminatory biases. As such, the Agencies will likely continue their efforts in striking the right balance between harnessing the benefits of AVMs but also minimizing detrimental consequences with properly placed quality control standards. We will continue to monitor for any developments on this proposed rule.
1. Sylvia Brown and Alex C. Engler, Governing the Ascendancy of Automated Valuation Models – Regulating AI in Residential Property Valuation (Oct. 5, 2023), Governance Studies at Brookings, https://www.brookings.edu/articles/governing-the-ascendancy-of-automated-valuation-models/.
2. Hayes, Adam, What is an Automated Valuation Model (AVM)? How They Work (Sept. 28, 2023), Investopedia, https://www.investopedia.com/terms/a/automated-valuation-model.asp#:~:text=Leading%20AVM%20providers%20include%20CoreLogic,well%2Dknown%20type%20of%20AVM.
3. Quality Control Standards for Automated Valuation Models, 88 FR 40638 (proposed June 21, 2023) (hereinafter “Quality Control Standards for Automated Valuation Models”).
4. See “Loan modifications and other changes to existing loans” in Quality Control Standards for Automated Valuation Models, pg. 40642.
5. See id.
6. Id.
7. See Quality Control Standards for Automated Valuation Models, pg. 40640.
8. See American Bankers Association, Comment Letter on Quality Control Standards for Automated Valuation Models (Aug. 21, 2023) (hereinafter “American Bankers Association”).
9. See Housing Policy Council, Comment Letter on Quality Control Standards for Automated Valuation Models (Aug. 21, 2023).
10. See Zillow, Comment Letter on Quality Control Standards for Automated Valuation Models (Aug. 21, 2023).
11. See Independent Community Bankers of America, Comment Letter on Quality Control Standards for Automated Valuation Models (August 21, 2023).
12. See American Bankers Association; Real Estate Valuation Advocacy Association, Comment Letter on Quality Control Standards for Automated Valuation Models (Aug. 17, 2023).
13. See Consumer Financial Protection Circular 2023-03, Consumer Financial Protection Bureau (Sept. 19, 2023), https://files.consumerfinance.gov/f/documents/cfpb_adverse_action_
notice_circular_2023-09.pdf.
14. Quality Control Standards for Automated Valuation Models, pg. 40654.
