The Cost of Trust: Who Pays to Prove They Are Honest | HL Hunt
The Cost of Trust: Who Pays to Prove They Are Honest
Before any money moves, someone has to establish that you are who you say, that your income is what you claim, and that the account you're using is yours. That work costs real money, it's performed on everyone, and it exists because of a minority who would misrepresent all three. The honest majority pays for it. That's an ordinary and defensible arrangement — every system with a fraud risk works this way. What makes it consequential in consumer finance is a property this desk keeps encountering: the cost is fixed per transaction rather than proportional to it, so the burden of proving your honesty is heaviest for the people transacting in the smallest amounts.
In this report
What verification actually costs
Verification is not one thing. Establishing that a transaction is legitimate involves several separate confirmations, each with its own cost:
| Question | Answered by | Cost driver |
|---|---|---|
| Are you a real person? | Identity checks against records | Data access and matching |
| Are you this person? | Documents, biometrics, knowledge checks | Capture, review, and failure handling |
| Is this account yours? | Account ownership confirmation | Data access, consent flow |
| Is your income what you say? | Documents or source-connected data | Retrieval and review — see our verification guide |
| Are you acting for yourself? | Fraud and coercion screening | Model and review |
| Are you permitted to transact? | Sanctions and regulatory screening | Screening and false-positive review |
Two properties of the total matter more than its size.
Most of it is paid on people who pass. Every applicant is verified; a small share are rejected. So the overwhelming majority of verification expenditure is spent confirming that honest people are honest.
The manual portion dominates. Automated checks are cheap; the cases that fail an automated check and go to human review are expensive, and false positives drive that volume. Which means verification cost is driven less by the fraud rate than by the false positive rate — a system that flags many honest people costs far more than one that catches slightly less fraud with fewer flags.
Almost all verification spending confirms that honest people are honest. The cost is driven by how many of them get flagged, not by how many fraudsters exist.
The fixed-cost property
The structural feature that makes this a distributional question rather than merely an operational one.
Verifying an identity costs approximately the same regardless of the amount at stake. The check is the check. Some elements scale slightly with amount — more scrutiny on larger transactions — but the base is fixed.
Express a stylized $45 verification cost as a share of transaction value:
| Transaction | Verification as a share |
|---|---|
| $200 | 22.5% |
| $1,000 | 4.5% |
| $8,000 | 0.56% |
| $45,000 | 0.10% |
| $320,000 | 0.014% |
A factor of roughly 1,600 between the top and bottom rows, for identical work, driven entirely by transaction size.
This is the same structure our selection analysis identified in origination costs and our barbell analysis identified in institutional overheads. Verification is a third instance of the same mechanism, and the three compound: a small loan bears fixed origination cost, is made by a firm bearing fixed institutional cost, and requires fixed verification cost.
The distributional consequence follows directly and is uncomfortable. The households transacting in the smallest amounts pay the largest proportional cost of guarding against fraud they didn't commit — and those are disproportionately the households with the least margin, per our premium analysis. Nobody designed this and it isn't anyone's policy. It's arithmetic.
Trust substitutes
Verification is expensive, so the system has developed ways to avoid doing it. Each is a trust substitute — something that lets value move without establishing everything from scratch.
- Relationship. An institution that has served you for years doesn't re-verify. Cheapest of all, and the small end's advantage in our barbell analysis.
- Collateral. If the loan is secured, less turns on who you are — which is why secured products are available to people unsecured ones aren't.
- Guarantee. Someone already trusted vouches for you, transferring their standing — the mechanism in our guarantee guide.
- Employment or institutional affiliation. A payroll relationship confirms several facts at once at near-zero cost, which is why the employer-adjacent provision in our shock analysis is structurally cheap.
- Documented history. A record of meeting obligations, observed by others.
The pattern across all five: each converts an expensive question into a cheap one by relying on work somebody already did. That's the entire economic function of trust — not a moral quality but a mechanism for not repeating verification.
A credit file is portable trust
The reframe that makes the credit reporting system legible as infrastructure rather than as a scorekeeping apparatus.
A credit file lets a lender who has never met you rely on the accumulated observations of lenders who have. Its properties are exactly what a trust substitute needs:
- Standardized, so any participant can read it.
- Portable, so it works with a counterparty you have no relationship with.
- Cumulative, so it improves with time at no marginal cost.
- Cheap to consult relative to establishing the same facts directly.
- Costly to fake, which is what makes it worth consulting.
Which reframes several things this desk has treated separately.
The exercise cost in our option analysis is the flip side of this. That report described a credit file as manufactured collateral — valuable, durable, and forfeited on default. The same object viewed from the lender's side is a verification saving. It's collateral to the borrower and a cost reduction to the lender, and both descriptions are of the same mechanism.
Building credit is accumulating a trust substitute, and it explains why the process is slow in a way that isn't arbitrary: the value comes from the observations having been made over time by parties with something at stake. A record that could be assembled quickly would be cheap to fake and therefore worthless.
The bureau system's concentration is a consequence of the function. A trust substitute is more valuable the more counterparties accept it, which is a network property — and network properties concentrate. Whatever one thinks of the resulting structure, the concentration follows from what the thing is for.
What this says about thin files
The most consequential implication, and it reverses the usual framing.
Thin-file applicants are frequently not risky. They're expensive.
The distinction:
- A risky applicant is one whose expected losses exceed what the price supports. Declining them is a credit decision.
- An expensive applicant is one whose assessment costs more than the transaction supports. Declining them is a cost decision.
Someone with no accumulated record has no trust substitute, so a lender must verify directly. On a small loan that cost exceeds what the transaction bears — so the application is declined, and the decline reflects the price of finding out rather than the answer.
Our thin-file analysis documented that these applicants perform better than their treatment implies, and our override analysis found near-cutoff overrides performing at approved rates. The cost framing explains why that gap persists rather than being competed away: a lender who could profitably serve these applicants still can't, because the assessment cost is real and doesn't fall with better underwriting.
And the compounding is worth stating plainly. Someone with no file borrows small, which is where verification is proportionally most expensive, which is why they're declined, which is why they don't build a file. The trap our poverty premium analysis describes has verification cost as one of its mechanisms — and it's the mechanism least addressed by anything aimed at the population, because it isn't about them at all.
The optimal verification level isn't maximum
The same structure as our fraud analysis, applied a step earlier in the process.
More verification catches more fraud and rejects more honest people, in a ratio that worsens as you tighten. So there's an optimum, and it isn't zero fraud:
Tighten only while: fraud prevented × fraud cost > (verification cost + honest applicants lost × their value)
What makes verification worse than fraud screening on this dimension: the cost is incurred on everyone, not just on those rejected. A fraud rule costs you only the false declines. A verification requirement costs you the process on every applicant plus the ones who abandon.
Abandonment is the part that gets missed. An applicant who doesn't complete a verification step is indistinguishable from one who never applied — which makes it the unobserved counterfactual our measurement analysis describes, in Family A. The cost of requiring too much is structurally invisible, so requirements drift upward.
The prediction that follows: verification requirements should be systematically heavier than the economics justify, for exactly the reason fraud screening is. And the remedy is the same one — randomize, by relaxing a requirement for a random share and measuring completion and realized fraud together.
What would lower the cost
Everything on this list works the same way: make a verification reusable rather than repeated.
- Source-connected data. Confirming income from a payroll or bank source rather than from documents removes both the document handling and the review — the mechanism our verification guide describes, and the largest single available reduction.
- Reusable identity credentials. A verification performed once and presentable to many counterparties converts a repeated fixed cost into a one-time one. This is the highest-leverage change available at the small end specifically, since that's where the fixed cost binds.
- Standardized records, which is what the bureau system already provides for credit history and doesn't for identity.
- Employer and institutional channels, where affiliation confirms several facts at once.
- Progressive verification — verifying proportionally to what's at stake, rather than fully at first contact.
- Reducing false positives, since manual review of flagged honest applicants is the expensive part.
The honest trade-off attached to the first three: each concentrates dependence on whoever operates the shared infrastructure. A reusable credential is valuable because many parties accept it, which is the same network property that concentrates the bureau system. Reducing verification cost and concentrating verification infrastructure are the same movement, and anyone advocating the first is advocating the second whether or not they say so.
The strongest objections
"Framing this as a tax on the honest is loaded." Fair, and worth conceding properly. Everyone benefits from a system where transactions can be trusted — including, especially, people who would otherwise be unable to transact with strangers at all. Verification isn't deadweight; it's what makes a market with strangers possible. The distributional point stands independently of the framing: whatever verification is worth, its cost falls unevenly by transaction size, and that unevenness isn't chosen by anyone.
"Much verification is legally required regardless of economics." True and important. Identity and screening obligations exist for reasons unrelated to a firm's cost-benefit calculation, and a firm cannot optimize them away. Our response: the requirement determines that verification happens, not how expensively. The gap between a source-connected check and a document review is entirely within the firm's control, and it's most of the cost.
"You've asserted the cost figures." Correct. The $45 is stylized and actual costs vary by product, channel, and how much manual review a firm's false positive rate generates. The argument depends on the fixed-cost structure rather than the level — at any per-verification cost, the share of a $200 transaction exceeds the share of a $320,000 one by three orders of magnitude, and the distributional result follows.
Testable implications
- Verification cost as a share of transaction value should rise sharply as transaction size falls, and should be a substantial share of the total cost of small-dollar products.
- Thin-file decline rates should exceed what realized performance justifies, with the gap tracking assessment cost rather than loss rates — the sharpest test of the cost-versus-risk claim.
- Source-connected verification should raise completion rates and lower cost simultaneously, since it removes both the document burden and the review.
- Abandonment at verification steps should be substantial and largely unmeasured, and relaxing a requirement for a random share should reveal it.
- Verification requirements should be heavier than the economics justify, for the measurement reason above — testable by randomized relaxation.
- Reusable credentials should expand access at the small end specifically, with little effect on large transactions where the fixed cost is already negligible.
The second is the one worth running and the one most lenders could answer from data they hold. If thin-file applicants who are approved perform close to the approved population — which our own override analysis suggests — then the decline rate is measuring cost rather than risk, and that's a different problem with a different solution. Better underwriting doesn't fix a cost problem; cheaper verification does.
The broader conclusion: trust in financial services is not a sentiment but a cost structure. Its price is fixed, its burden is proportional to nothing, and the people who pay the most for it are the ones with the least to transact.
Frequently asked questions
The cost is largely fixed rather than proportional — the same checks cost the same at $200 or $200,000. As a share, the burden is hundreds of times heavier at the small end.
Anything letting a counterparty extend value without verifying from scratch — a relationship, collateral, a guarantee, or a documented history. Each relies on work somebody already did.
Frequently because they're expensive to assess rather than risky. With no trust substitute, verification costs more than a small transaction supports, so the decline reflects the price of finding out.
Making verification reusable — source-connected data, portable credentials, standardized records. Each also concentrates dependence on whoever runs the shared infrastructure.
Key takeaways
- Almost all verification spending confirms that honest people are honest, and its cost is driven by false positives rather than by fraud volume.
- Verification cost is fixed per transaction, so it consumes 22.5% of a $200 transaction and 0.014% of a $320,000 one.
- Trust substitutes — relationship, collateral, guarantee, history — all work by relying on verification somebody already performed.
- A credit file is portable standardized trust, which is the lender-side view of the same object our option analysis called manufactured collateral.
- Thin-file applicants are frequently expensive rather than risky, so better underwriting doesn't fix it and cheaper verification does.
- Abandonment at verification steps is an unobserved counterfactual, so requirements drift heavier than the economics justify.
This report presents an analytical framework and the authors' interpretation; it is not legal, compliance, or financial advice. Verification cost figures are stylized illustrations and vary substantially by product, channel, and operation. Identity, screening, and customer due diligence obligations are legal requirements whose application is not a matter of cost-benefit judgment — consult qualified counsel about your obligations.