Fair lending in cash flow underwriting means the rules you apply to bank statement deposits must not treat applicants differently on a prohibited basis, directly or through a proxy. Regulation B bars discounting income because it comes from public assistance, part-time work or a pension. Lenders should test approval outcomes by group and write down why each deposit rule exists.
What fair lending rules reach cash flow underwriting decisions?
The Equal Credit Opportunity Act and its Regulation B govern every credit decision, including one based on bank statement cash flow. They prohibit discrimination on a prohibited basis in any part of a credit transaction, limit how certain income sources may be treated, and require specific reasons when an application is denied.
The current official interpretation is narrower on one point than many compliance teams remember. Regulation B's commentary on section 1002.6 now heads its discussion "Disparate Treatment." It says the Act does not prohibit practices that are neutral on their face, except where neutral criteria function as proxies for protected characteristics, designed or applied with the intent to advantage or disadvantage people on that basis.
For a cash flow lender, that leaves three working duties under federal law:
- Do not treat applicants differently because of race, sex, age, marital status, national origin, religion, receipt of public assistance or the other protected bases.
- Do not build or apply a neutral-looking rule as a stand-in for one of those traits.
- Follow the specific income rules in section 1002.6(b) and the notice rules in section 1002.9.
Plenty of lenders still answer to more than federal ECOA. State laws, investor guidelines and bank partners may set their own standards for outcomes. Ask counsel which apply to your products before you decide how far to take outcome testing.
Deposit rules that act as proxies in cash flow underwriting
A deposit rule acts as a proxy when it sorts applicants in a way that tracks a protected trait closely enough to stand in for it. Cash flow models are full of rules that look neutral: minimum deposit counts, payer-type exclusions, account-age floors and balance thresholds. Most are sound risk measures. A few can be used as substitutes.
The examples below show how an ordinary rule becomes a fair lending question:
- Excluding peer-to-peer transfers. Some communities move rent, wages and side income through payment apps more than others. A blanket exclusion may remove real income for some groups more than for others.
- Excluding cash deposits. Cash-heavy work is common in some trades and neighborhoods. Excluding every cash deposit, rather than asking for a source, can hit those applicants hardest.
- Requiring an account older than a set number of months. Recent arrivals to the country and recently divorced applicants often have new accounts for reasons unrelated to repayment.
- Penalizing deposits from government payers. That rule runs directly into the public assistance protection described in the next section.
The legal question is whether a rule was designed or applied to disadvantage a group. The practical question comes first: can you show a credit reason for the rule that you would defend to an examiner? If the honest answer is "we have always done it," the rule needs review.
Data source matters too. Statements pulled through a consumer-permissioned connection carry different payer labels from a scanned PDF, and the CFPB's section 1033 rule on personal financial data is changing how those feeds work. A rule written for one format can behave differently on the other.
Income types Regulation B protects in cash flow underwriting
Regulation B names income a lender may not discount or exclude because of its source. Under section 1002.6(b), a creditor may not take into account whether income comes from a public assistance program, outside a narrow allowance for judgmental systems, and may not discount income because it comes from part-time work, an annuity, a pension or another retirement benefit.
Section 1002.6 of Regulation B also covers alimony, child support and separate maintenance. When an applicant relies on those payments, the creditor must count them as income to the extent they are likely to be consistently made. The commentary allows a lender to look at how long a payment has been received and how regular it is, judged for the individual rather than by group statistics.
On a bank statement, those protected sources appear as ordinary deposits. Social Security, a pension, a state benefit and a support payment each show up as a line with a payer name. A categorization rule that drops "government" deposits from qualifying income, or that applies a haircut to every pension credit, may break the rule without anyone intending it. The guide to verifying public assistance deposits on bank statements covers how to confirm benefit income without discounting it.
Tax deposits need a separate look. A refund from the IRS is a one-time event and usually not income, while a payment to the IRS is an expense. The guide to reading IRS payments and refunds on bank statements explains how to tell them apart without lumping all government payers together.
The safe pattern is to judge each protected deposit on the same questions you ask of wages. Is it likely to continue? Is it regular? If it passes those tests, it counts.
How do you test cash flow underwriting outcomes for bias?
Testing cash flow underwriting outcomes means comparing approval rates, approved amounts and pricing across groups, then tracing any gap back to the rules that produced it. The goal is to find a rule acting as a proxy before an examiner or plaintiff does, and to confirm the protected income rules are applied as written.
A workable cycle for a small or midsize lender looks like this:
- Pick the decisions. Approvals, declines, counteroffers and the amount granted are the usual four.
- Pick the groups. Where you lack demographic data, lenders often use estimation methods. Have counsel choose the method, because the choice shapes what the results mean.
- Compare outcomes. Look for gaps in approval rate and average amount between groups with similar credit profiles.
- Trace the gap. Rerun the decisions with one rule switched off at a time. The rule whose removal closes most of the gap is the one to study.
- Decide and record. Keep the rule with a written reason, change it, or drop it.
Run the cycle when you add or change a rule, and on a regular schedule otherwise. Many lenders choose quarterly or annually depending on volume.
Mortgage lenders carry an extra layer. Their cash flow analysis feeds an income figure that must also satisfy the ability-to-repay rule for mortgages, so a rule change can affect both the fair lending test and the verification record at once.
Documenting the reason behind each cash flow underwriting rule
Every cash flow underwriting rule should carry a short record: what it does, why it predicts repayment, what data supports that, when it was last tested and who approved it. That record is what lets you defend the rule, and it is also what lets you write accurate decline reasons.
The decline side is set by regulation. Section 1002.9 of Regulation B requires the statement of reasons to be specific and to give the principal reasons for the adverse action. Saying the applicant failed the lender's internal standards is not enough. If you cannot explain a rule to yourself, you cannot explain it to a declined applicant.
The table below gives example records for four common rules. The thresholds are illustrations, not recommendations.
| Example rule | Fair lending question | Record to keep |
|---|---|---|
| At least 8 deposits a month | Does it screen out applicants paid monthly by a pension or benefit? | Loss data by deposit count, and whether protected payers are counted |
| Exclude all payment app credits | Does it remove real income unevenly across groups? | Reason for exclusion, and a review path for documented app income |
| Account open 6 months or more | Does it track national origin or marital status changes? | Repayment data by account age, and accepted alternatives |
| Average daily balance above $1,000 | Does it penalize lower-income applicants beyond what the payment needs? | Link between the threshold and the loan payment size |
Software can carry part of this record. TrueBankFlow lets an underwriter include or exclude deposits, recalculates the averages, and writes the exclusions and notes into the report for the loan file. That gives a reviewer the line-level trail behind each figure.
When a passing fair lending test still leaves cash flow risk
A clean outcome test does not prove a cash flow program is fair. Small samples hide gaps, averages hide individual mistakes, and manual overrides can undo a well-built rule set one file at a time. The test checks the rules; it does not check the people applying them.
Three situations deserve their own review:
- Overrides and exceptions. When an underwriter approves outside policy, record why. A pattern of exceptions granted more often to one group is a disparate treatment risk even if every rule is neutral.
- Low-volume products. A product with a few hundred decisions a year may never show a statistical gap. Review a sample of declines by hand instead.
- Deposit labels you did not set. A vendor or data feed may label a deposit "government" or "transfer" by its own logic. If your rules key off those labels, you inherit the vendor's choices. Ask how they work.
Treat the outcome test as one check among several. File reviews, override logs and the written reason for each rule together give a fuller picture than any single number.