What a Price Has to Cover: Decomposing the Cost of a Loan

The phrase "risk-based pricing" suggests that what a borrower pays is mostly a function of how likely they are to repay. It isn't. A credit price has to cover six distinct things, and only one of them is about the borrower's risk. The other five are properties of the lender's operation and of the loan's size and term. On small loans, the expected loss component is frequently not the largest — fixed origination and servicing costs spread over a small balance can exceed it substantially. Which means a great deal of the public argument about credit pricing is conducted about the wrong component, and several proposed remedies address the part that isn't dominant.

The six components

Component	What it is	Scales with
Cost of funds	What the lender pays for the money	Amount and term
Expected loss	Losses across the pool	Amount, and borrower risk
Origination	Acquisition, verification, decisioning, disbursement	Nothing — fixed per loan
Servicing	Statements, payments, support, collections	Largely fixed per month
Cost of capital	Return on equity held against the asset	Amount and risk weight
Margin	Profit	Whatever the market allows

Read the third column. Two of the six don't scale with the loan at all — and those two are the ones that determine what happens to small loans.

Only the second row is about the borrower. Which means that for two otherwise identical borrowers, the difference in price attributable to risk is one component of six, and for small amounts it's a minority of the total.

Only one of the six components is about the borrower. Two of them don't scale with the loan at all.

Two loans, worked

Stylized figures to show the structure. These are illustrative rather than any lender's actual costs.

Assume: funds at 5%, origination at $180 per loan, servicing at $7 per month, capital charge at 2% of balance annually, and expected loss of 9% of the amount.

Loan A: $20,000 over 36 months.

Component	Total cost	As annual % of loan
Funds	$1,650	2.8%
Expected loss	$1,800	3.0%
Origination	$180	0.3%
Servicing	$252	0.4%
Capital	$660	1.1%
Before margin	$4,542	7.6%

Loan B: $500 over 4 months.

Component	Total cost	As % of amount
Funds	$5	1.0%
Expected loss	$45	9.0%
Origination	$180	36.0%
Servicing	$28	5.6%
Capital	$3	0.7%
Before margin	$261	52.2%

Origination alone is four times the expected loss on Loan B. The borrower could be certain to repay — expected loss zero — and the loan would still need to recover $216 on $500 over four months.

The inversion

The finding this report exists for, and it reverses the usual framing.

The public argument about high-cost small credit is conducted almost entirely in terms of risk. Defenders say these borrowers are risky, so the price reflects the risk. Critics say the price exceeds what the risk justifies, so it's extraction. Both sides assume the expected loss component is the dominant term.

On small short-term loans it frequently isn't. Which means:

• The defence is weaker than it sounds. "These borrowers are risky" doesn't explain a price whose largest line is a fixed cost that would be there for anyone.
• The critique is aimed partly at the wrong thing. Arguing that the price exceeds the risk is correct and doesn't establish extraction, because most of the price isn't risk.
• The remedies both sides propose miss. Better underwriting addresses the loss component; rate caps bind on the total. Neither touches origination cost, which is what's actually large.

This is the structural answer to the question our time preference analysis left open. That report established that an extraction mechanism exists and couldn't size it against genuine cost. This decomposition doesn't size it either — but it establishes that genuine cost on a small loan is a much larger number than the risk framing implies, which narrows the space the extraction argument has to work in.

And it confirms, from the cost side, what our verification analysis and selection analysis found from the access side: thin-file and small-balance borrowers are expensive rather than risky, and the expense is mostly in getting the loan made rather than in getting it repaid.

What actually moves a price

The decomposition's practical use: name which component a proposed change acts on, and check whether that component is large.

Intervention	Acts on	Effect on a small loan
Better underwriting	Expected loss	Modest
Cheaper funding	Cost of funds	Negligible
Rate cap	The total	Rations, per our rationing analysis
Automated decisioning	Origination	Large
Source-connected verification	Origination	Large
Payroll-integrated collection	Servicing and loss	Large
Embedded distribution	Acquisition	Large
Longer term	Spreads origination	Large, with a caveat below

Every large-effect row is a fixed-cost row. Which is the same conclusion our distribution analysis reached by a different route — that genuine access expansion has come from cost reduction rather than from product design — now with an arithmetic reason attached.

And it explains why our employer analysis found such large price differences for economically identical advances. The payroll channel eliminates acquisition, verification, and collection simultaneously — three fixed costs at once — which is why it can price at a fraction of an equivalent standalone product without being better at predicting anything.

Why term matters as much as size

The second dimension, and it produces a genuine tension.

A fixed origination cost is recovered over the life of the loan, so a longer term spreads it thinner. The same $180 is 36% of a four-month $500 loan and a much smaller annual burden over two years.

Which suggests longer terms make small loans cheaper — and they do, in annual rate terms. The caveat is that total cost paid moves the other way. A longer term means more months of funding cost, more months of servicing, and more total interest, so:

• The annual rate falls while the total dollars paid rise.
• Which of those matters depends on the borrower's situation — and per our disclosure analysis, presenting one without the other misleads in a predictable direction.
• Longer terms on short-lived needs create the term mismatch our structural analysis describes.

So "extend the term to lower the rate" is a real mechanism and a partial one, and it's the mechanism behind a good deal of product design that presents as consumer-friendly. The honest version tells the borrower both numbers.

The component nobody discusses

Cost of capital rarely appears in public argument and it shapes what gets offered.

A lender must hold equity against assets, and equity expects a return. That return is a real cost, it varies with the risk weight assigned to the asset, and it's invisible to borrowers entirely.

Two consequences worth naming:

It creates pressure toward assets with favourable treatment, independent of their economics for borrowers. Per our category analysis, where a product sits in a regulatory taxonomy determines the capital held against it — so two economically identical arrangements can carry different capital costs and therefore different prices, for reasons having nothing to do with either borrower.

It explains part of the barbell our institutional analysis describes. Institutions with cheaper equity can price the same asset lower, which compounds with the fixed-cost advantage of scale.

The practical note for anyone reading a lender's pricing: if a product's price seems disconnected from its apparent risk and cost, capital treatment is frequently the missing term — and it's the one that never appears in a disclosure or a defence.

What the decomposition can't tell you

Being clear about the limit, because the framework is easy to over-claim.

This tells you what a price would have to cover. It does not tell you what any lender's costs actually are. Without real cost data you cannot determine:

• How much of a given price is cost recovery versus margin.
• Whether a lender's origination cost is $180 or $40 — and the difference changes the whole picture.
• Whether expected loss assumptions are honest or conservative.
• What return on capital is being targeted.

Both cost recovery and captured margin produce high prices on small loans, and the decomposition is consistent with either. Anyone asserting the split from the price alone — in either direction — is asserting rather than measuring.

What it does establish is narrower and still useful: a high price on a small short-term loan is not evidence of extraction, because the cost floor is genuinely high. And correspondingly, "these borrowers are risky" is not an adequate defence, because risk isn't the dominant term. The framework removes two lazy arguments and leaves the real question open.

The strongest objections

"Your cost figures are invented." Correct, and stated in the text. The figures are illustrative and real costs vary enormously by lender, channel, and product — an automated digital lender's origination cost is a fraction of a branch-based one's. The structural claim survives any plausible figures: origination and servicing are fixed per loan, so their share rises as the loan shrinks. The magnitudes are illustrative; the direction isn't.

"Lenders choose their cost structure." A good point. A lender with $180 origination costs has made choices — about channel, about verification method, about automation — and a competitor with lower costs could undercut them. Which is an argument that the cost floor is lower than any particular lender's costs, and it's the argument for why cost reduction is the productive lever. It doesn't make fixed costs zero.

"This is a defence of high-cost lending." Half of it functions that way and half doesn't, which is the point. It weakens the risk-based defence at least as much as it weakens the extraction critique, and its main conclusion — that the productive intervention is reducing cost per loan — is a demand on lenders rather than an excuse for them.

Testable implications

1. Price should vary more with loan size than with borrower risk, within any lender's book — the direct test, answerable from a single lender's own pricing.
2. Lenders with automated origination should price small loans materially below branch-based ones at equivalent risk.
3. Expected loss should be a minority of price on small short-term loans and a majority on large ones, across lenders.
4. Channel integration should reduce price more than underwriting improvement does, per the employer comparison.
5. Regulatory category changes should shift prices via capital treatment, independently of any change in borrower risk.
6. Term extension should lower annual rates and raise total cost paid, which is arithmetic and worth confirming in marketed products.

The first is the one that would settle the framing and is available to any lender internally. If price varies more with size than with risk across your own book, your pricing is a fixed-cost recovery structure wearing a risk-based label — and that's worth knowing before defending it as the latter.

The conclusion we'd hold: credit pricing is described in terms of risk and driven substantially by things that aren't risk. The component that dominates small-loan pricing is the cost of making the loan at all, which is the same finding this desk has now reached from access, verification, distribution, and collections — arriving here at the price itself.

Frequently asked questions

---

This report presents an analytical framework and the authors' interpretation; it is not financial, legal, or policy advice. All cost figures are stylized illustrations chosen to demonstrate structure and are not estimates of any lender's actual costs, which vary substantially by institution, channel, and product. The report does not and cannot determine the share of any price attributable to cost recovery rather than margin; the implications identified as testable are hypotheses.