Lenders Champion the Way Towards Income Inclusivity

Low income car buyers face many challenges when applying for auto loans. And while that isn’t disputed, the root cause of the friction is often debated. The hope is that we protect consumers from predatory lending while making it easy and acceptable to obtain credit – without making people jump through hoops.

Our data shows that low income consumers more often mis-state their income, both over and under. Often this is due to a lack of clarity in how lenders account for variable pay.

Low income car buyers tend to mis-state their income, resulting in higher payments or default likelihood

Borrower with Understated Income: Joe states that he makes $45K/year. He has a well-established occupation as a driver and works for a well-known employer. His understated income will result in him paying over $1,000 in interest over the course of his loan.

Below are examples of paystubs showing types of variable earnings that lenders must decide whether/how to consider. These redacted images are a fascinating visualization of the problem. Because, it’s not merely getting data off of documents – simple OCR can do that. But, using patent-pending intellectual property, Informed provides real intelligence that classifies variable earnings and knows which count as real income.

Joe’s overtime makes it difficult for him to accurately state income

Informed is the only company dedicated to fairness when it comes to instantly calculating applicant income.
OCR tools read data off of documents, but they aren’t an end-to-end solution helping lenders comply with EOCA and Fair Lending. We handle complexities such as variable earnings, duplicate documents, multiple employers, and W-2 wage earners with other forms of fixed income requiring gross ups.

Because there is turnover at lenders, new staffers unknowingly make errors and introduce bias as they manually review line items. Sometimes they count forms of overtime, shift differential, per diem pay, and in other cases, they don’t. This results in disparate treatment and subjects the lender to additional liability of fair lending violations.

Working multiple shifts adds uncertainty and results in over/understatements

Applicants who work certain professions aren’t at a disadvantage when Informed’s AI is deployed.
Informed is committed to solve inequities faced by:

  • Drivers
  • Construction Laborers
  • Nurses
  • Machine Operators
  • Construction Supervisors
  • Mechanics
  • Carpenters
  • Electricians
  • Equipment Operators
  • and more

Consumer-permissioned sources can introduce additional barriers for low income earners since many lenders underwrite based on net income rather than gross income. Every American has a different tax situation leading to unpredictable withholdings, tax deductions, garnishments, 401k contributions, etc. 

So by underwriting on net income a lender can reduce the borrower’s payment-to-income ratio forcing them into a higher interest rate. For low income borrowers these services also introduce friction, require remembering account username and password, and don’t offer full support. They also don’t count important variable pay types including: per diem, shift differential, overtime, per mile, etc. 

Consumer-permissioned data is valuable when consumers need to show sources of income other than W2 paystubs. But, it’s critical to look at stipulation and credit policies holistically so resolving one issue doesn’t introduce bias or fair lending concerns. 

The Bureau of Labor Statistics (BLS) struggles with standardizing consumers’ income and provides education for consumers to learn how to classify their income.  When BLS  surveys Americans, they specify which line items on paystubs count as income and which don’t.

List of possible line item paystub earnings from the BLS

Please EXCLUDE when reporting wages:

  • Attendance bonus
  • Back pay
  • Clothing allowance
  • Discount
  • Draw
  • Holiday bonus
  • Holiday premium pay
  • Jury duty pay
  • Meal and lodging payments
  • Merchandise discount
  • Non-production bonus
  • On-call pay
  • Overtime pay
  • Perquisites
  • Profit-sharing payment
  • Relocation allowance
  • Severance pay
  • Shift differential
  • Stock bonuses
  • Tool/equipment allowance
  • Tuition repayment
  • Uniform allowance
  • Weekend premium pay
  • Year-end bonus

What is included when reporting wages? Please INCLUDE when reporting wages:

  • Base rate
  • Commission
  • Cost-of-living allowance
  • Deadheading pay
  • Guaranteed pay
  • Hazard pay
  • Incentive pay
  • Longevity pay
  • Over-the-road pay (mileage)
  • Piece rate
  • Portal-to-portal rate
  • Production bonus
  • Tips

Source: https://www.bls.gov/respondents/oes/faqs.htm#16

For the average American, and even funding analysts, calculating income in a standard form accurately is challenging.

Mike, the Coca-Cola Machine Worker from Chattanooga, TN, also struggled

Car buyers don’t know how lenders will consider their variable pay.
Auto lenders have incredibly complex formulas to calculate applicant incomes that vary throughout the year and can’t be easily explained. This has left many car buyers, and the dealers serving them, in the dark and concerned about overstating income that could then result in an unwanted repossession.

A Lender Surveys Their Team on Income Calculations

One of Informed’s clients ran an experiment. They provided applicant income documents to several staffers, ranging in titles from executives to funding managers. Participants were asked to calculate income using their credit policy handbook as their guide. (All funding analysts are equipped with the handbook) The results of the experiment showed only two people were accurate based on a threshold of calculated income. No two participants calculated the same income! 

This experiment demonstrates that it’s much more than an OCR challenge. It’s not simply reading text off of documents. The real problem is classifying the text according to the earning type and then assessing whether the earning type fits the lenders’ definition of income.

At times lenders don’t even realize there is an issue as sampling the calculation for accuracy is often a quality control function. But that function doesn’t extend to re-calculating and validating income. The data then contributes to analysis on credit worthiness, ability to pay, and creating a model and attributes for the underlying credit scorecard. This “dirty data” is a problem that results in potential unfair treatment of borrowers.

Better Than OCR

Informed is the only lending solution focused on fairness. Google Cloud Platform and AWS Textract have OCR solutions to read data off of paystubs, but they’re only pulling text from the documents. Their mission isn’t to lower the cost of credit using AI for financial inclusivity, real-time transparency, and improved compliance like Informed.

AWS Textract and GCP Lending AI are focused on mortgage – super prime borrowers who don’t have these challenges. But they encounter other challenges such as a higher synthetic fraud rates and charge offs 18-24 months from booking. Informed focuses on fairness, enabling our clients to better comply with ECOA and Fair Lending laws. 

We’re not an OCR company. We’re a fair lending solution that provides a platform for multiple use cases, portfolios and has a lender pool across the full spectrum of lending. We work tirelessly to ensure we are focused on this strategic and product differentiation as many Americans are counting on fair treatment when applying for credit.

The “clean data” showcases Informed’s ability to move the needle for lenders by instantly, and accurately calculating applicant income. Our customers and investors see our continued commitment to the market, our communities and underrepresented minorities across income brackets. We’re solving for fair income calculation regardless of any other variables. General purpose OCR vendors like AWS Textract and Google Cloud Platform aren’t able to dedicate the resources we can. We’re committed to solving the challenges of accessing credit for borrowers with variable pay.

We’ve invested resources to close the income inequality gap. As our team grows, we hire those aligned with this mission. Our passionate subject matter experts push the limits of machine learning and computer vision to facilitate fairer lending. We’re delivering on the promise we make to lenders around more accurate income calculations.

Our lenders want to do right by their customers. They’ve chartered us with solving this difficult problem and providing a unique approach to aggregating the best applicant income sources. We do this through partnerships with consumer-permissioned data sources and our dedicated fraud team. Our third-party assessor performs disparate impact analysis on an ongoing basis as part of our k-fold cross-validation work. We lead discussions in successful customer implementation calls with benchmark data and a focus on our lenders’ long term strategic goals.

Just Launched: stAI informed Podcast

X