Quick answer: a bank does not price a loan properly by taking its funding rate and adding an arbitrary spread. A risk-adjusted pricing engine separates several jobs: the funding and liquidity cost of supporting the loan, expected credit loss, operating and servicing cost, option or commitment costs, the return required for capital exposed to unexpected loss, and the commercial margin needed to make the transaction worthwhile. The quoted borrower rate is then constrained by competition, law, customer relationship value, collateral, covenants and product strategy. A loan can have a high interest rate and still be badly priced if the bank is being paid for the wrong risks.

The price of a loan should answer two questions at once: what does this loan cost the bank, and what risk remains after those costs are paid?

Why this belongs in mathematics

Loan pricing combines discounted cash flow, probability, expected loss, capital allocation, optimisation and sensitivity analysis. It is also an incentive system. If the pricing model understates one component—say term funding, prepayment or economic capital—the bank can systematically originate loans that look profitable at the desk while destroying value at the portfolio level.

The European Banking Authority’s loan-origination guidance gives a useful public decomposition: loan pricing should consider capital cost, funding cost matched to the loan’s characteristics and behavioural life, operating costs, credit-risk costs, other real costs and market competition. It also calls for risk-adjusted performance measures when assessing pricing and profitability. See the EBA Guidelines on Loan Origination and Monitoring.

1. Start with the all-in economics, not the coupon

For a simplified fixed-rate loan, define:

r_L = customer loan rate;
c_F = matched funding/FTP cost;
EL = expected credit loss as a rate;
c_O = operating and servicing cost;
c_X = option, commitment, liquidity or other product cost;
K = economic or allocated capital;
h = target return on that capital.

A simplified break-even pricing floor can be written:

Minimum economic rate ≈ c_F + EL + c_O + c_X + required capital return.

The exact implementation differs by bank and product, but the decomposition prevents one cost from being hidden inside another.

2. Funds transfer pricing isolates the funding job

A five-year loan should generally carry the economic funding and liquidity cost associated with its expected life rather than being compared mechanically with a short-term deposit rate. That is the role of funds transfer pricing (FTP).

Suppose a five-year floating-rate loan reprices every three months. Its interest-rate component may behave like a short reset, but the bank can still be committed to funding the asset for years. Federal Reserve FTP guidance explicitly separates those dimensions: the interest-rate component can map to the repricing horizon while the liquidity component maps to the longer holding horizon.

See How Banks Use Funds Transfer Pricing Algorithms.

3. Expected loss is the average credit cost—not the capital charge

A simple credit-cost identity is:

EL = PD × LGD × EAD.

If a S$1 million exposure has a one-year PD of 1.5%, LGD of 35% and EAD of S$1 million, simplified expected loss is S$5,250, or 0.525% of exposure.

Expected loss should not be confused with the return required for unexpected loss. A portfolio can have a manageable average expected loss but still require substantial capital because actual losses can vary around that mean.

Federal Reserve research published in 2025 finds that expected credit risk is an important component of observed bank loan pricing and that higher-risk loans tend to carry higher interest rates, although pricing depends on more than credit risk alone. See Examining the Relationship Between Loan Pricing and Credit Risk.

4. Economic capital prices the uncertainty around expected loss

Suppose two loan portfolios each have expected annual loss of 0.5%. Portfolio A contains many weakly correlated borrowers. Portfolio B contains a few concentrated borrowers exposed to one property market. Their average losses can match while the tail loss of Portfolio B is much larger.

Economic capital is an internal attempt to measure the capital needed for unexpected loss at a chosen confidence standard and horizon. The exact definition is institution-specific; it is not automatically the same as regulatory capital.

If a loan consumes S$100,000 of allocated economic capital and the bank targets a 12% pre-tax return on that capital, the loan needs about S$12,000 per year of economic profit contribution after expected loss and other allocated costs merely to meet that target.

5. RAROC turns profit into a capital-efficiency ratio

A simplified Risk-Adjusted Return on Capital (RAROC) can be written:

RAROC = risk-adjusted income / economic capital.

Risk-adjusted income might begin with interest and fee income, then subtract FTP/funding cost, expected loss, operating cost and other attributable costs. If the resulting annual risk-adjusted income is S$18,000 and allocated economic capital is S$120,000, RAROC is 15%.

If the bank’s hurdle is 12%, the transaction appears to create value under those assumptions. But every term matters. Understate PD, overstate collateral recovery or use a cheap FTP curve and the same loan can cross the hurdle without becoming economically better.

6. A worked all-in example

Imagine a S$2 million commercial loan priced at 6.30%.

Component	Rate-equivalent	Annual amount
Customer interest	6.30%	S$126,000
FTP/funding and liquidity	−3.70%	−S$74,000
Expected credit loss	−0.60%	−S$12,000
Operating/admin cost	−0.35%	−S$7,000
Commitment/option cost	−0.15%	−S$3,000
Risk-adjusted income	1.50%	S$30,000

If the transaction requires S$250,000 of economic capital, simplified RAROC is 30,000 / 250,000 = 12%.

Now increase expected loss by 0.30 percentage points because the borrower’s sector deteriorates. Risk-adjusted income falls by S$6,000 and RAROC falls to 9.6%. The loan’s contractual rate has not changed. Its economic attractiveness has.

7. Pricing floors prevent relationship pressure from erasing economics invisibly

A pricing engine can calculate a minimum economic rate or minimum required spread. The business may still choose to price below it for a documented strategic reason—relationship value, cross-sell economics, promotion, public-policy programme or competitive entry—but the exception should be visible.

The danger is not every exception. The danger is systematic override without measurement. If relationship managers repeatedly win approval below the risk-adjusted floor, the bank can accumulate a low-return portfolio while each individual exception appears small.

8. Fees must be converted into the same economic language

An upfront fee can make a loan appear more profitable, but its value depends on expected life. A S$20,000 fee on a five-year loan is not economically equivalent to S$20,000 per year.

The pricing model should amortise or discount fees and costs across the expected cash-flow profile. If the borrower is likely to prepay after one year, an upfront fee can be more valuable per year than if the loan remains outstanding for ten years—while the loss of future interest income also changes.

9. Prepayment and undrawn commitments alter the price

A borrower may have the right to prepay, redraw or use an undrawn commitment. Those options change the bank’s expected cash flows and contingent funding need.

A revolving facility with a low current draw can still consume liquidity capacity and capital. A fixed-rate mortgage can prepay precisely when rates fall and the asset becomes valuable to the bank. Good pricing needs to charge for those behavioural options rather than discovering them later inside Treasury or liquidity risk.

10. Collateral lowers some loss estimates but can create false comfort

Strong collateral can reduce LGD and therefore expected credit cost. But collateral value can fall at the same time the borrower becomes weak. A commercial-property borrower and its property collateral are not independent if both depend on the same property cycle.

The pricing model should therefore stress collateral, liquidation time and recovery costs rather than converting current appraised value into a permanent LGD assumption.

11. Covenants change the payoff distribution without changing the coupon

Two loans with the same interest rate, maturity and borrower PD can have different risk if one gives the bank earlier information and intervention rights through covenants while the other does not.

A covenant is therefore not simply legal text around the price. It can alter expected recovery, loss timing and the bank’s ability to stop further exposure growth. Pricing and terms belong in one economic package.

12. Competitive price and economic price can disagree

If competitors offer 5.2% while the bank’s risk-adjusted model needs 5.8%, the pricing engine has not “failed.” It has identified a strategic choice:

decline the loan;
change collateral or covenants;
reduce size or maturity;
seek fees or ancillary value;
accept a documented below-hurdle return;
challenge whether the model is too conservative.

The current OCC Lending and Loan Portfolio Risk Management handbook, issued in June 2026, treats pricing and loan terms as part of disciplined lending and portfolio risk management rather than an isolated sales decision.

13. Selection effects can make high rates create worse borrowers

Increasing price can compensate for risk only up to a point. A very high rate can change which borrowers accept the loan, increase debt-service burden and make risky projects more attractive to borrowers who have less to lose.

This is the classic adverse-selection and moral-hazard problem in credit markets: price is not merely compensation for a fixed risk distribution; price can change the distribution itself.

A strong pricing model therefore cannot assume that raising the coupon always restores RAROC mechanically.

14. The pricing algorithmic pipeline

Define product cash flows and expected life.
Assign matched FTP/funding and liquidity cost.
Estimate PD, LGD and EAD or another defensible expected-loss model.
Allocate operating and servicing cost.
Price embedded options and contingent liquidity where material.
Estimate regulatory and/or economic capital usage for the decision framework.
Calculate risk-adjusted income and RAROC or equivalent measure.
Derive an economic pricing floor or hurdle.
Compare with market and relationship considerations.
Optimise non-price terms such as size, tenor, collateral and covenants.
Record any override and its rationale.
Monitor realised margin, default, recovery, prepayment and utilisation.
Backtest whether loans priced above the hurdle actually outperform.
Update the pricing engine when funding, capital or borrower behaviour changes.

15. Failure modes

Coupon-only thinking. High borrower rate is treated as proof of high profitability.
Cheap-FTP subsidy. Long assets inherit a short funding cost.
Expected-loss/capital double count. The same risk is charged twice under inconsistent definitions.
Collateral permanence. Today’s collateral value becomes a fixed recovery assumption.
Fee illusion. Upfront fees are counted immediately without expected-life effects.
Override normalisation. Below-floor exceptions become routine without portfolio-level visibility.
Adverse-selection blindness. Higher price is assumed to compensate risk without changing borrower behaviour.
Relationship-value fiction. Cross-sell value is claimed but never measured.

16. Diagnostics and falsifiers

Which component contributes most to the pricing floor?
How far does RAROC move if PD or LGD is stressed?
What funding tenor is used, and does it match expected life?
How much capital does the transaction consume under the chosen internal framework?
Does collateral still protect the bank in the same scenario that weakens the borrower?
How many loans were approved below the floor last quarter?
Do below-floor loans actually generate compensating relationship income?
Does raising price reduce application quality or increase subsequent default?

Suppose someone claims, “The loan is profitable because its 7% coupon is well above the bank’s 3% deposit cost.” A falsifier is a full pricing decomposition showing that matched funding, expected loss, operating cost, options and required capital return total more than the apparent 4% spread. A nominal spread is not a risk-adjusted return.

17. Verification and update triggers

compare predicted expected loss with realised defaults and recoveries;
reconcile FTP with actual marginal funding economics;
compare projected utilisation with drawn exposure;
measure profitability after prepayment and restructuring;
backtest pricing exceptions separately from standard approvals;
recalibrate capital and hurdle assumptions when portfolio concentration changes;
update operating-cost allocations when processes automate;
challenge any model that consistently approves loans that later earn below the stated hurdle.

Connections across the finance-and-banking algorithms lane

Funds transfer pricing — isolates funding and liquidity cost.
Expected credit loss — the credit-cost model used in pricing and accounting with different purposes.
Credit concentration — portfolio context can make one apparently attractive loan costly at the margin.
Bank capital models — regulatory and economic-capital concepts must not be conflated but both can affect pricing decisions.

Research anchors

The deeper lesson

Loan pricing is the mathematics of making different costs visible before the bank commits capital and funding. FTP prices the balance-sheet resource. Expected loss prices average credit deterioration. Economic capital prices uncertainty around that average. Operating costs price the process. Options price behavioural choices. The quoted rate is only the final surface. A strong pricing model can explain what lies underneath—and can show exactly which assumption would make the apparent profit disappear.

Educational note: This article explains public banking mathematics and risk-management concepts. It is not lending advice, consumer financial advice, credit advice or institution-specific pricing guidance.

Small Group Tutorials

How Banks Price Loans for Risk-Adjusted Return: Expected Loss, Funds Transfer Pricing, Economic Capital, RAROC and Pricing Floors