WorkflowConsumer Loans & Cards

Fraud procedure · for verification, fraud and underwriting leaders

Loan application fraud: a lender's procedure for a falsified pay stub

When a pay stub or bank statement in a loan application looks falsified, the lender's exposure lies less in the forgery than in what it does next: whether the file is held, who decides, what the applicant is told and whether a report is due. This guide sets out that procedure for verification, fraud and underwriting leaders, and how Income and Document Verification supports it.

Updated · 13 min read

See one case, screen by screen →
Watch with transcript →

A lender's procedure for loan application fraud involving a falsified pay stub or bank statement has six parts: hold the file without a decision, preserve the document exactly as received, have a fraud analyst decide whether the signs amount to a finding, decide the credit on what is proved or decline with a specific reason a second person co-signs, assess suspicious activity reporting on its own clock, and charge off any loan already funded within the period the classification rules allow. The law behind it is severe (a false statement to a federally insured lender carries up to $1,000,000 and 30 years under 18 U.S.C. § 1014), which is exactly why the procedure has to be careful: a tamper signal establishes that a figure is inconsistent, not that anyone meant to deceive. Botminds' Income and Document Verification raises each signal with the figures behind it, holds the file, and leaves every consequence to a named person.

The files, applicants and figures in the examples are fictional sample data from the solution; the statutes and rules are real and linked where they appear.

One sample lender's verification desk since the agents took over the checks

1,148

files verified

23

altered documents confirmed by a fraud analyst

14min

median from a complete file to a decision, against 2.6 days by hand

Fictional sample lender in Income and Document Verification. 'Confirmed' means a fraud analyst made a finding; signals that were cleared with a written reason are counted separately.

What the law makes of a falsified income document

Two federal statutes reach it, and both turn on knowledge. Section 1014 covers whoever "knowingly makes any false statement or report ... for the purpose of influencing in any way the action of" a listed federal or federally insured institution "upon any application ... loan, or insurance agreement", with a maximum of a $1,000,000 fine, 30 years, or both. Bank fraud under 18 U.S.C. § 1344 covers whoever "knowingly executes, or attempts to execute, a scheme or artifice" to defraud a financial institution or obtain its money by false pretences, with the same ceiling. State fraud and forgery laws apply alongside them.

The elements of § 1014, and what a document check can establish about each

ElementWhat it meansWhat a tamper check establishes
KnowinglyThe person knew the statement was falseNothing: arithmetic cannot tell a forgery from a payroll correction the applicant never read
False statement or reportThe document or assertion is untrueA great deal: a net figure that is not gross less deductions is wrong as arithmetic
For the purpose of influencingIt was made to move a credit decisionPartly: a stub sent with an application was sent to support it
The institutionA listed federal or federally insured lenderNothing: it is a fact about the lender
Our reading of 18 U.S.C. § 1014. A check is very good at the second element and silent on the first, which is why a signal is a question for a fraud analyst, not a finding.Source: 18 U.S.C. § 1014 (Cornell LII)

For a lender the statutes matter less as a threat to the applicant than as a discipline on its own staff. An accusation of a federal felony should not rest on a font. The evidential standard for a decline is the lender's own and lower than a criminal one, and nobody writing in a file, a letter or a call note should confuse the two.

Why a tamper signal holds the file instead of declining it

Because every integrity test checks consistency, and inconsistency has innocent causes. Gross to net, statutory tax rates, year-to-date against the pay calendar, balances carried from page to page and a re-save by an editing tool each test whether figures agree. A payroll correction, a retroactive adjustment, a reprinted stub, a partly scanned page or a photographed document can each fail one of them.

So a signal in Income and Document Verification does one thing: it stops the file where it stands and sends it to the fraud analyst with the evidence. No decision is issued and nothing goes to the applicant until fraud decides. Clearing a signal takes a written reason a second person can read, and the fraud analyst reviews every clearance. The checks themselves and the arithmetic each runs are set out in how to identify fake pay stubs.

The path a held document takes, and who decides at each step

  1. Agent

    Check

    Runs the integrity checks on every page and raises a signal naming the figure or date each sign rests on.

    Never: Declines a file or calls it fraud.

  2. Agent

    Hold

    Stops the file. No decision is issued and no communication goes to the applicant.

  3. A person decides

    Fraud analyst

    Reads every sign against the pages behind it and decides: finding, or clearance with a written reason.

  4. A person decides

    Verification analyst

    Decides the credit on what is proved, or declines on a specific stated reason.

  5. A person decides

    Underwriting team lead

    Co-signs every decline for altered documents, and any decision above the analyst's limit.

  6. A person decides

    BSA team

    Decides separately whether suspicious activity reporting thresholds are met, on the reporting clock.

As the fictional sample lender runs it in Income and Document Verification. Only the first two steps are automatic, and the second does nothing but stop.

Hold, refer or clear: what each outcome must leave behind

A held file has three futures, and each must produce its own evidence. A hold that produces nothing is the state to avoid, because it cannot be told apart from a file nobody opened.

The three outcomes for a held file

OutcomeWhat it requiresWhat the applicant experiencesWhat stays on the file
ClearedA written reason, reviewed by the fraud analystA short delay, then the file proceedsThe reason, the reviewer and the date, so clearances can be counted
Declined on the documentsA fraud finding and the team lead's co-signatureA decline with a specific principal reasonThe finding, both signatures, the evidence and the reason given
Decided on the merits insteadEnough verified information to decide without the disputed pageAn ordinary decision at verified figuresWhy the disputed document was not relied on
Our table, as the fictional sample lender runs it. The third row is often the cleanest: where deposits, the payroll response and tax forms already settle income, no decision needs to rest on the disputed page.

Clearances deserve counting for the same reason policy exceptions do. The OCC's retail lending guidance expects exceptions to be "controlled, with appropriate reporting to identify and track exceptions and appropriate analysis to determine causes, impact, and any need for corrective action" (OCC Comptroller's Handbook, Retail Lending). A rising clearance rate on one product or one analyst's files is the pattern that reporting exists to show.

One flagged file, five signs, referred with the evidence

On the sample lender's second personal loan file, a biweekly net of $2,410.00 lands in savings on five of seven pay dates, but the two August stubs have no deposit behind them: $4,820.00 of pay on the stubs, $0.00 in the bank. The signal names five signs. No deposit in August; the late-August stub shows year-to-date gross of $58,500.00, the same as the September stub; the payroll response records unpaid leave for most of August; advice numbers repeat from July and September; and both August stubs were re-saved by an online PDF editor in October, late at night.

Three of the five come from outside the document, which is what makes the combination actionable where any one sign alone would be a question for the applicant. The analyst referred the file to fraud; it was held, and no notice went to the applicant until fraud decided. Why the deposit behind each stub is the hardest thing in a file to fake is covered in pay stub verification against bank deposits.

Five signs in Income and Document Verification, each naming the figure or date it rests on

Tamper signals for a flagged sample file: two August pay stubs with each sign marked on the page: no deposit in August, identical year-to-date gross of $58,500.00 on the late-August and September stubs, a payroll response recording unpaid leave in August, advice numbers repeated from July and September, and both stubs re-saved by an online PDF editor in October, with two actions: refer to fraud, or clear with a written reason.
Fictional sample file. The two actions offered are the only ones the analyst has; neither is a decline.

How should the decline be worded?

With a specific principal reason, and never with anything that hints at a report. Regulation B requires that a statement of reasons for adverse action "must be specific and indicate the principal reason(s) for the adverse action", and that statements resting on "the creditor's internal standards or policies" are insufficient (12 CFR § 1002.9(b)(2)). A decline for a document the lender could not verify should say what could not be verified, in terms the applicant can act on if there is an innocent explanation.

The stated reason is also the only thing the applicant will ever be given, so it has to stand on its own. That is a further argument for deciding on what is proved where possible: a decline that rests on verified income below the policy minimum is easier to state, defend and explain than one that rests on a judgement that a page was altered.

When does loan application fraud require a suspicious activity report?

When the bank's thresholds under the SAR rule are met, and the decision runs on its own clock. For a national bank the OCC rule requires a report for a known or suspected federal criminal violation aggregating "$5,000 or more where a suspect can be identified", or "$25,000 or more regardless of potential suspects", filed "no later than 30 calendar days after the date of the initial detection of facts that may constitute a basis for filing a SAR"; where no suspect was identified, filing may be delayed 30 more days, and "in no case shall reporting be delayed more than 60 calendar days" (12 CFR § 21.11). The other banking agencies keep parallel rules for the institutions they supervise.

Two clocks that start the same morning and never meet

The credit decisionthe applicant is told, specifically
Hold and fraud review
Decision, co-signed
Notice with a specific principal reason
  • Hold and fraud review
  • Decision, co-signed
  • Notice with a specific principal reason
The report, if thresholds are metthe applicant is never told
Initial detection
30 calendar days to file
30 more only if no suspect was identified
  • Initial detection
  • 30 calendar days to file
  • 30 more only if no suspect was identified
Reporting periods from 12 CFR § 21.11(d). The credit lane is illustrative; the notification period itself comes from Regulation B. Lane widths show elapsed days.Source: 12 CFR § 21.11

Confidentiality binds everyone who touches the file: "No national bank, and no director, officer, employee, or agent of a national bank, shall disclose a SAR or any information that would reveal the existence of a SAR." Management must "promptly notify its board of directors, or a committee of directors or executive officers designated by the board", and the bank must keep a copy of the SAR and the supporting documentation "for a period of five years from the date of the filing". For a verification desk the practical rule is short: front-line staff never mention a report in a call, an email or a decline letter, and the original document is preserved as supporting documentation from the moment the file is held.

What if the loan has already funded?

Then the finance function has a clock too. The OCC's retail classification guidance is that "fraudulent loans should be classified loss and charged off no later than 90 days from discovery or within the time frames specified in the classification policy, whichever is shorter." A falsified stub found in a post-funding quality review, or after an early payment default, starts that clock on discovery, alongside the SAR clock.

Post-funding discovery is also the case for checking every file before decision rather than sampling after. A lender that finds altered documents only when a loan defaults finds them after the money has gone, with a charge-off deadline and a report to file, instead of before, with a hold and a question.

Six controls a lender should be able to show for falsified documents

  1. The original document preserved as received

    Not a printed copy or a re-saved file; re-saving overwrites the history that raised the question.

  2. Every sign recorded with the figure behind it

    "Year-to-date gross identical on two stubs a month apart" is evidence; "tamper suspected" is not.

  3. No decline by a check

    A signal holds the file; a named fraud analyst decides; a second person co-signs a decline for altered documents.

  4. Clearances with written reasons, counted

    By product and by analyst, so a drift in clearing is visible.

  5. A specific adverse action reason that stands alone

    Never resting on internal standards, never hinting at a report.

  6. SAR assessment on its own clock

    Thresholds, 30 or 60 days, board notification and five-year retention, decided by the BSA team.

Drawn from 18 U.S.C. §§ 1014 and 1344, 12 CFR § 21.11, Regulation B § 1002.9 and the OCC's retail lending guidance; each is a step or setting in Income and Document Verification for the fictional sample lender.

Who signs, and why the decline needs a second signature

At the sample lender the underwriting team lead co-signs above the analyst's $35,000 limit and on every decline for altered documents, whatever the amount. The two conditions have different reasons. A large approval is co-signed because the money at risk is large; a decline for altered documents is co-signed because the consequence to a person is severe and the evidence is interpretive.

The error that costs most in this work is not the forgery that gets through, a credit loss the lender can size and price, but the honest applicant refused on a misread document, a harm the lender cannot see and the applicant cannot easily appeal. A second reader is the cheapest control against it and the only one that works on a judgement rather than a number. Every signature in the solution records its meaning with the signer's name, which is what lets a reviewer months later tell the person who decided from the person who processed.

Signing the decision, with the meaning recorded

The signing screen: the analyst signs a decision to approve at verified income with one condition before funding, with the meaning recorded beside her name as 'I reviewed the verified figures and decide this application', and a note that a second signature is required above her approval limit.
Fictional sample file in Income and Document Verification. The same screen requires the team lead's co-signature on any decline for altered documents.

Where Income and Document Verification fits

The solution's agents line up every document in a file, tie each pay stub to its deposit, check the arithmetic and file history of each page, match identity and employer across sources, and raise a signal with the evidence when something does not add up. They hold the file and route it; they never decline, call anything fraud or contact the applicant. Approval limits, the co-signature rule, how tamper signals are handled and the time to answer a request are settings the lender owns. How often the checks flag documents that turn out to be genuine is measured in automated income verification.

What changes for a fraud or verification leader is the starting point and the evidence. Every held file arrives with its signs, the figures behind them and the pages they came from, and every outcome, cleared, declined or decided on the merits, leaves its reason and its signatures. To see the procedure on your own files, start from Income and Document Verification.

Questions people ask

What should a lender do first when it suspects a falsified pay stub?

Hold the file without a decision and preserve the document exactly as received. Then route it to a fraud analyst with every sign and the figure behind it; nothing goes to the applicant until fraud decides.

Can a lender decline for a suspected fake document?

Yes, on a fraud finding and with a specific principal reason, as Regulation B requires. Often the cleaner route is to decide on what is proved, such as verified income from deposits and a payroll response, so the decision does not rest on the disputed page.

What are the SAR thresholds for loan application fraud?

For a national bank, $5,000 or more where a suspect can be identified and $25,000 or more regardless of suspects, filed within 30 calendar days of initial detection, or up to 60 where no suspect was identified. The other banking agencies keep parallel rules.

Can a lender tell the applicant it filed a SAR?

No. The rule prohibits the bank and its staff from disclosing a SAR or any information that would reveal its existence. The applicant receives only the credit decision and its specific reason.

What happens to a funded loan later found to be fraudulent?

The OCC's guidance is to classify it loss and charge it off no later than 90 days from discovery, or sooner if the bank's classification policy says so. The SAR assessment runs alongside on its own clock.

Why count cleared signals?

Because clearances are exceptions to a control, and the OCC expects exceptions to be tracked and analysed for causes. A rising clearance rate on one product or one analyst's files is the signal that the checks, the training or the judgement need attention.

Income and Document Verification.
See it on your own files.

An agentic solution from Botminds. Every pay stub, bank statement and ID checked against each other in minutes, with what does not add up shown on the page. Book a demo, or explore how it works.