Increasing demand for credit requires automated underwriting
As demand for credit increases, the Federal Reserve Bank of NY (NY Fed) reports a high number of credit applications by those with low credit scores is resulting in decreased acceptance rates. Yet these “thin file” consumers are reliably paying rent, utilities, internet and other monthly costs. Lenders are leaving potential good customers’ business on the table. Credit and loan underwriting may be relying on outdated practices.
The NY Fed recently published its October 2021 Survey of Consumer Expectations (SCE) Credit Access Survey (SCE), the most recent of its consumer surveys conducted every four months. In addition to finding that credit card and loan applications for consumers with credit scores below 680 were increasing, they also noted that mortgage refinancing had slowed down in the last quarter.
How much demand is there for credit?
Credit card balances increased by $17 billion each in Q2 and Q3 2021 in the U.S.. Consumers seem to have shifted back from the austerity of COVID-19 spending when they chose to pay down their debt instead, observes National Mortgage Professional. This is evidenced by the total household debt climbing by $286 billion (1.9% since Q2) to $15.24 trillion in the third quarter of 2021.
And it’s not just existing credit that’s being used. The NY Fed found in the October 2021 SCE report that 26.5 percent of respondents applied for credit cards in the preceding 12 months, compared to 15.7 percent in the year ending October 2020. Credit demand in 2021 is back to 2019 rates.
Why credit rejection rates are high
NY Fed data shows applications increased in Q3 2021 among all age groups and credit scores, the largest increase being from applicants aged 40 or younger, 60 or older, and those with lower credit scores.
That’s reflected in a higher rejection rate of 18.6 percent (compared to 15.4 percent in the previous quarter). While only 2.8 percent of applicants with high credit scores were rejected, applications by those with low credit scores were rejected at an alarming 48.2 percent rate. The NY Fed reports that 6.6 percent of consumers were simply too discouraged to even apply for credit.
Rejection rates vary amongst credit instruments:
- Credit cards: 8%
- Mortgages: 3%
- Auto loans: 3%
- Card increases: 5%
- Mortgage refinance: 1%
As credit demand is increasing by potential borrowers with low credit scores, lenders may be missing out on good customers through antiquated loan underwriting.
Inflation and consumer fragility underscore the importance of credit underwriting
While formerly bullish about containing inflation to 2 percent, U.S. Federal Reserve Chair Jerome Powell now warns that with the new variant of COVID-19, “it now appears that factors pushing inflation upward will linger well into next year.” A continued pandemic also may affect productivity, he warned, noting that health-related concerns could “reduce people’s willingness to work in person, which would slow progress in the labor market and intensify supply-chain disruptions.”
Effective credit underwriting must be able to take these disparate factors into account, enabling lenders to look at the applicants’ full profiles. Underwriting is still a largely manual process based on the credit manager’s experience and the use of inflexible, rules-based systems. That is problematic when receiving applications from borrowers who have little credit history, low scores and they need fast decisions. Legacy underwriting solutions are not up to the job of evaluating whether these are high-risk loans for poor credit borrowers, or opportunities for new, reliable customers.
High-risk loans or building cases for eligibility?
Some thin‑file clients – those with little or no credit history – may need help creating the case for eligibility. According to Forrester Analytics Consumer Technographics, banks focus on the preliminary data presented to them when evaluating lending behavior, labelling applicants as high-risk borrowers. It may be, however, that the applicant simply is inexperienced. Forrester’s data shows that 14% of US adults banked online for the first time as a result of COVID-19. Forrester urges lenders to move beyond static underwriting models and conduct loan portfolio diagnostics to assess borrowers’ capacity to pay on an ongoing basis. Forrester’s 2021 report on lending makes the case for assessing stress factors such as supply chain disruption and increasing consumer debt as factors in banks’ lending portfolios to uncover new lending opportunities.
How automated AI underwriting systems meet these challenges
In a traditional lending scenario, the loan officer collects social security numbers, employment status and income and reviews the applicants’ credit scores if they are available.
While an automated AI underwriting system will review these factors, it can also collect regulation-permissible data such as loan and credit card history. It can also include alternative data points such as phone, internet and utility payments and monthly rent. The AI model enables a more robust credit history, creating a full picture of the applicant’s ability to manage regular payments.
AI helps to segment clients into risk stages as they tend to overlap. For instance, at origination there is the potential for delinquency in the first three to six months, so that group is closely monitored.
Unlike traditional credit management tools, advanced AI provides real-time scoring and can predict abnormal behaviors. When the AI model identifies late or minimum payments, it alerts the loan officer who can monitor the credit file or contact the customer. This provides opportunities to help borrowers who may be struggling after a sudden job loss or other circumstances.
Why are lenders hesitant to use AI for loan underwriting?
In 2020, LendIt sent a survey to thousands of lenders to learn how they are investing in technology for credit risk. As reported in Using AI to manage credit risk: lenders report on current AI use and future investments, lenders revealed their biggest concern was explainability. In 2021, explainability remained a high concern, second only to model governance and regulatory requirements (LendIt and Brighterion, AI perspectives: Credit risk and lending, 2021).
Model explainability has become increasingly necessary given the potential for adverse outcomes as AI adoption increases. At Brighterion, we score hundreds of billions of events every year. Our AI models provide a layer of transparency that is used by thousands of banks globally. Not only does Brighterion AI provide predictive scores, it helps lenders understand the reasons behind them. This allows credit risk managers to supervise their portfolios on a one‐to‐one level and develop more personalized strategies to improve their borrowers’ experiences.
Brighterion’s explainability techniques also make it easier for institutions to manage their model governance and meet regulatory requirements.
Automated loan underwriting systems: reliability in uncertain times
Over the past two years, lending money has become much more complex. A global pandemic, inflation, unemployment and supply chain interruptions have all contributed to uncertainty in the industry. Yet lenders have a tool available to them – AI-based loan underwriting – that can create more robust borrower profiles, help them make well-informed lending decisions, and manage the credit lifecycle with confidence. Advanced AI solutions ensure lenders are onboarding new customers who have proven records, including making measured, data-driven decisions on those with lower credit scores or little retail lending experience.
Our Ebook Credit risk and how to manage the customer lifecycle with AI discusses how the models work, using AI for credit origination (including with thin-file clients) and our AI Express implementation methodology.