Over the past year, the Federal Housing Finance Agency (FHFA) and Government-Sponsored Entities (GSEs) Fannie Mae and Freddie Mac have sought innovative ways to safely and reliably expand access to home ownership. I’ve been working on it.
This charge could not have been better timed.
The racial disparity in homeownership rates between blacks and whites in America is wider than it was in 1960, driven by a combination of rising mortgage rates, rising home prices, and outdated models for assessing consumers’ ability to repay. , the problem is getting worse. As mortgage professionals, it is our duty to support homeownership and provide sustainable mortgage opportunities in the communities we serve. Fortunately, progress has been made in assessing consumers’ ability to repay, allowing lenders to adopt more responsible and inclusive lending practices.
The easiest way to understand a consumer’s ability to pay their mortgage is to analyze the money going in and out of the bank each month. A long-standing model for determining repayment ability, the consumer credit score only considers past credit to determine credit risk. It leaves out the countless Americans with no credit and poor credit, and could miscalculate the repayment ability of consumers with low credit scores.
A holistic underwriting process that incorporates financial data from multiple vectors (disposable income, discretionary income, credit score) provides a clearer picture of a borrower’s true ability to pay, while providing relevant insights into loan origination and production. You can better protect against the risks that FormFree calls ATP.
ATP gives lenders the best of both worlds. In addition to providing lenders with powerful tools to responsibly expand housing finance opportunities in underserved communities, ATP has been rejected for less sophisticated underwriting models. It enables lenders to increase their volume by capturing the business of likely creditworthy loan applicants. ATP helps lenders turn rejected loans into closed-hit loans, helping more consumers unlock the wealth-building power of a generation of homeowners.
Understanding the language of cash flow data
Analyzing consumer transaction data can provide a more detailed understanding of a consumer’s ATP, but manually reviewing bank statements can be a daunting task for busy mortgage professionals. Luckily, the latest advances in data intelligence are giving lenders a way to re-engineer that process without the hassle.
A powerful innovation of the 21st century, natural language processing incorporates aspects of linguistics, computer science, and artificial intelligence (AI) to identify patterns between digital data and human language. For mortgage lending purposes, natural language processing can be used to extract consumer cash flow data from bank statements, giving lenders a better understanding of a prospect’s disposable and discretionary income. increase.
While this extensive data is useful on its own, it is even more useful when considered in conjunction with traditional models of credit risk assessment. Lenders can corroborate the data collected through natural language processing with other external third-party data sources, such as credit bureaus, providing an additional layer of security when determining a homebuyer’s credit eligibility. is provided.
Natural language processing can also prevent unintended biases that can arise from alternative forms of data intelligence. While other AI or machine learning-based technologies are known to programmatically learn and apply human biases, natural language processing solutions use rule-based algorithms that only calculate income and cash flow. To do. These algorithms intentionally omit information such as the applicant’s ethnicity in order to maintain objectivity. As such, natural language processing is designed to find the truth based on real banking data, rather than supporting long-standing human biases that limit homeownership for minority borrowers.
Democratizing the credit decision process
This year, FHFA Director Sandra Thompson testified before Congress that rent payment history is just as important to a consumer’s credit rating as mortgage payments, and she made a great point. Rent is often the most important payment a consumer makes each month, but not all landlords submit a consumer’s rent payment history to credit bureaus. Smaller landlords with less sophisticated systems often fail to automatically report a positive rental history, so a renter’s positive payment history won’t be reflected in her credit score. . This dynamic, in which tenants must rely on their relationship with landlords, is unfair to consumers and time consuming for lenders.
By requiring landlords to report tenant payment history at the credit decision stage, lenders and borrowers become dependent on third parties for financial data. Not only does this slow down the underwriting process, it adds another layer of risk as the data requested by the landlord may not be reliable. When a rent verification request is submitted, a busy or uninvolved landlord may simply sign it and verify that the consumer paid the rent on time, without the rigorous due diligence required by the lender. I have.
Additionally, natural language processing reduces credit risk by analyzing cash flow data that lenders typically don’t consider. For example, if a person’s income is sufficient to qualify for a loan, but they pay thousands of dollars each month for their children’s daycare or private education, that cost will add up to what they can afford each month for a mortgage. adversely affect. With natural language processing mining consumer banking data, lenders can factor in major monthly expenses such as childcare when determining credit eligibility. This allows lenders to serve more homebuyers and keep the pipeline flowing in a safer and more inclusive manner.
Using technology to support traditional credit risk assessment methods gives consumers more power in the credit decision process. Not all consumers can share their financial data to show lenders why they are reliable borrowers, but using natural language processing, as a typical monthly income and expenses average he ATP can be calculated. Ultimately, this deep insight into consumer financial habits will enable homebuyers to prove their credit risk not just in their credit scores, but in their bank statement transactions, thereby helping to ensure equal affordability. expands housing opportunities for
Protect the economy by prioritizing truth over trust
The housing finance industry is a central player in our economy, and a market of such enormous size and influence essentially affects everything else around it. To maintain, lenders, investors, regulators, GSEs, and fintechs pay special attention to the policies, procedures, and data they use to determine borrower risk. But a commitment to safe and sustainable homeownership doesn’t stop the industry from evolving its processes. That’s why the FHFA and the GSE, two of the biggest players in the housing finance industry, are working to find new and innovative ways to use consumer data to expand access to sustainable credit.
With greater insight into direct-source consumer-authorized banking data, lenders can gain more insight into the financial health of future borrowers, from how bills are paid, to various sources of income, assets, and even employment status. You can judge many things. Natural language processing analyzes these disparate data points in a consistent and structured way and uses rule-based algorithms to quickly and easily identify common monthly expenses.
It’s important to note that natural language processing technology for the mortgage industry didn’t appear overnight. It takes computer science to build reliable algorithms that allow GSE’s automated underwriting system to receive consumer data, identify transactions such as rent payments, and create messages to lenders informing the consumer of her ATP. It took several years. These systems are carefully tested and re-tested before being released to a wide audience of lenders. It has strong regulatory protections and is guaranteed to be free of unknown forms of bias.
Consumers’ own financial data is the most reliable in identifying consumer risk, and is more reliable than being assessed by different types of risk models through multiple hands. Using natural language processing to look at actual financial transactions, the lender will be able to see the “truthful data”, i.e. transaction history from the consumer’s account and more trust-based data such as third-party reports and credit scores. can be considered at the underwriting stage.
what’s next?
Effective natural language processing techniques can extract deeper meaning from unstructured data and make a difference in the lives of millions of home buyers who were blinded by credit or unable to access an affordable mortgage. bring. By using this technology to compare cash flows with other credit risk assessment models, lenders can better manage their risks appropriately and holistically.
The mortgage industry has made great strides in how it leverages direct-source data. Industry leaders are investing in research and development to expand the realm of financial data that consumers can give lenders permission to view and analyze. The FHFA’s support enables the mortgage industry to anticipate policy changes that support more nuanced risk assessments for better lenders, consumers and the overall housing economy.