Asurity Technologies. LLC (“Asurity”) appreciates the opportunity to offer public comment, primarily on the Equitable Access and Regtech questions on which the FHFA has invited input. Asurity develops and builds RegTech software products aimed at advancing efficiency, fairness, and legal compliance in the home lending industry through its products RiskExec®, Propel™, and RegCheck®. Asurity’s responses herein are very much informed by the decades of experience of its senior team in advancing these critical business and regulatory objectives through the embrace of technology. Below Asurity sets forth each question presented for comment which is relevant to its business expertise, and offers its response in the immediately adjacent column.

Asurity believes these important topics warrant active consideration, ongoing industry engagement, and careful deliberation. Technology has a fundamental role to play in helping to achieve Equitable Access. That being said, to optimize the effectiveness of technology solutions access to consistent and actionable data across the loan origination and servicing spectrum is essential. Only with the benefit of actionable data can technology help to bring greater transparency to these critical stages in the home lending lifecycle, thereby increasing visibility and empowering more informed and goal-driven management. Most certainly, accessible and consistent information will drive better decisioning and over time provide the impetus for improved outcomes. For these reasons, it is imperative that the FHFA focus on establishing consistent data standards and improving access to data, particularly with respect to loan servicing where such data currently is plainly inadequate. With strong baseline data, technology can aid lenders in identifying trends, understanding patterns, analyzing comparisons, and designing efficacious policies and procedures to address any identified concerns.

Asurity has framed each of its responses with these overarching principles in mind. In the event there are any follow-up questions, Asurity’s contact details are included at the end of this submission. Asurity looks forward to continuing to be an active participant in this very important regulatory and public policy dialogue.

QUESTIONS AND RESPONSES SET FORTH BELOW

C. Equitable Access to Mortgage Credit

Question C.1: What new fintech tools and techniques are emerging that could further equitable access to mortgage credit and sustainable homeownership? Which offers the most promise? What risks do the new technologies present?

Asurity Response: Asurity believes there are a plethora of new alternative underwriting models offering great opportunity with some risks described here. In part because of the absence of good data, the development of strong models to improve the treatment of borrowers in the loan servicing process (i.e., related to sustainable home ownership) is further behind.

1. As fintech lenders increasingly offer products that do not follow traditional underwriting and pricing factors, fintech lenders must watch for bias that AI or alternative credit models might unintentionally replicate based on systemic bias in underlying data.
2. The alternative lending models that provide the most promise are the ones which can be independently validated and tested (i.e., not 'black-box', to the extent a result can be replicated for review/audit).
3. To ensure equity in pricing and decisioning, lenders, in general, need independent fairness analytics that are data agnostic, and flexible to assess their own factors, models, and algorithms. As newer underwriting tools make rapid decisions, often without human intervention, real-time assessments which can identify gaps in the fairness of pricing and underwriting decisions will enable lenders to innovate while correcting models if inequity creeps in.

In loan servicing, the same fairness analytics can be applied. Servicing treatment of borrowers - regarding late payments (e.g., fees, fee waivers, customer service treatment), workout steps (notifications, provisions of options, terms of loss mitigation), forbearance and other activities – can result in inequitable results.  However, servicers do not have access to the kind of consistent demographic and other data to enable them to assess the consistency of their treatment of borrowers with the same rigor that current tools allow them to do with respect to underwriting.

Question C.2: What emerging techniques are available to facilitate or evaluate fintech compliance with fair lending laws? What documentation, archiving, and explainability requirements are needed to monitor compliance and to facilitate understanding of algorithmic decision-making?

Asurity Response: Asurity’s response to this question would overlap with its response to the RegTech inquiry below; Asurity will address without duplication.  For compliance software to keep up with the pace of innovation and regulatory change, lenders and servicers will benefit most from software that is regularly updated (typically SaaS software which is frequently updated for all users), and which also scales and flexes to provide meaningful results based on a lender's unique products. Furthermore, embedding testing of aspects of fairness in pricing and underwriting during the marketing and origination phases should improve prevention and reduce reactive / corrective actions which are more time consuming and costly. Similarly, fairness assessments of servicing steps could also improve equity in loan servicing. This response describes compliance technology that is accessible to be integrated with origination technology for this purpose.

1. Techniques: Geocoding and mapping software has matured in the broader software market, and can be used in anticipatory fair lending analysis, as well as retroactively. Marketing outreach can be proxied using BISG to determine/validate data collection regarding race, sex, and ethnicity to ensure fairness in efforts to convert leads into applications. In addition, a ‘second look’ assessment can be utilized in the underwriting stage, helping to identify risks earlier.
2. Documentation/Explainability: Lenders can create specific plans with quantitative goals and measure against them. An institution's specific parameters/goals can be compared against results. For documentation/archiving, beyond any detailed loan application, lenders should store all data points used in pricing and underwriting decisions in a fair lending tool in order to properly assess (hence, software should be open to accepting lender parameters, not force prescribed or possibly antiquated credit decisioning data only). Lenders often model different scenarios before setting credit policy or other decisions. These scenarios should be stored as standard discipline, and also so that explanation of decisions can be shared with auditors and regulators.
3. Monitoring: Automated redlining analysis can easily identify an institution's performance against a custom group of similar lenders across all of their strategic markets. Gaps can be identified and actionable steps can be taken regularly, with analysis being performed either on a quarterly or monthly basis. Fintech lenders that service their loans should also utilize testing to ensure equitable treatment for those customers that end up in loss mitigation. Testing must be performed to ensure fairness across the loss mitigation process, but can also be used for operational compliance.

Question C.3: Are there effective ways to identify and reduce the risk of discrimination, whether during development, validation, revision, and/or use fintech models or algorithms? Please provide examples if available. 

Asurity Response: As described above, inserting compliance testing earlier in the process to keep up with AI or machine-learning based models has a better chance to identify troubling patterns as they emerge.

1. Marketing outreach/leads can be proxied using BISG to determine race, sex, and ethnicity to ensure fairness in efforts to convert leads into applications.
2. Disparity testing can be run across multiple factors such as steering, pricing, decisioning, timeframe to close, and more.
3. Decisioning and pricing factors can be incorporated into regression models to determine if there are outlier minority borrowers who should have been approved but were denied, or who were priced higher when they had similar credit factors.
4. Any outliers identified can be run through a match pair test to see if a control group borrower with similar credit was approved or received better pricing. Lenders might consider a 'second look' model if risk factors are identified (applies to decisioning, AVM, and other steps in process).

E. Regtech 

Question E.1: What are the most promising areas for applying technology to regulatory and compliance functions? Please describe opportunities for “regtech” to simplify or improve compliance with FHFA, Enterprise, or FHLBank requirements. 

Asurity Response: The technology described in sections C.1 - C.3 above has advanced considerably in recent years and can readily be accessed through Software-as-a-Service for rapid implementation. The most cost-effective way to achieve ‘good’ compliance is to proactively apply software to assist with pricing and underwriting review, demographic data assessment, mapping for comparison to census data, comparison to peers in markets, and where applicable, provide statistical and matched pair assessments for independent reviewers to gauge fairness. (As noted above, Asurity’s technology is built with these principles in mind - https://www.asurity.com/riskexec/ - and can be utilized by lenders and agencies when conducting portfolio reviews.)

The larger challenge Asurity sees is the application of fairness concepts to loan servicing. If the FHFA could set formal guidelines for submission of data by servicers, this should greatly improve this discipline. There is no clear equivalent to the HMDA submission process for servicing. As a result, data is very inconsistent in the marketplace, and, in the absence of regulatory guidance, servicers are unsure how much data to assess and where to look.  For instance, inequitable treatment may occur regarding fees and fee waivers, complaint handling, or loss mitigation processes. Data about customer treatment may reside in loan servicing systems, customer relationship management (CRM) platforms, call center systems, or other applications. In the servicing realm, obtaining this information is not a simple LOS extraction.

F. Office of Financial Technology Activities and Stakeholder Engagement 

Question F.2: What are some topics for a housing finance-focused “tech sprint” and how could FHFA encourage participation? 

Asurity Response: Federal regulators have provided clearer guidance concerning origination and underwriting than they have on fair servicing. A tech sprint focused on creating a view for what might constitute a common data file for the FHFA to review fairness in the servicing process would be extremely helpful.