How Banks Hedge Mortgage Rate-Lock Pipelines: Interest-Rate Lock Commitments, Pull-Through, Fallout, TBA Hedges, Basis Risk and Model Validation

Quick answer: when a mortgage lender promises a borrower a rate before the loan closes, it creates an interest-rate lock commitment (IRLC). If market rates rise before closing, the locked loan can become less valuable to an investor; if rates fall, some borrowers may abandon the locked loan and seek a better deal. Banks therefore estimate the probability that each lock will actually close—its pull-through rate—then hedge the expected rate-sensitive pipeline with forward mortgage sales, TBA mortgage-backed securities or other approved instruments. The hedge is continuously resized as locks are added, loans close, borrowers fall out, market rates move and the basis between the loan and hedge changes.

The bank does not hedge every locked dollar as if it will certainly close. It hedges the expected economic exposure—and then keeps correcting that expectation as the pipeline changes.

Why this belongs in mathematics

Mortgage pipeline hedging combines conditional probability, duration, convexity, option behaviour, market-value sensitivity, dynamic rebalancing and model validation. The key difficulty is that the bank is hedging an asset that does not fully exist yet. A rate lock can become a funded mortgage, expire, cancel, change product, extend or fail to close.

The OCC’s Mortgage Banking Comptroller’s Handbook treats pipeline and warehouse interest-rate risk, fallout risk, basis risk, measurement risk and timing risk as distinct mortgage-banking risks. Federal Reserve interagency guidance likewise expects institutions with material mortgage banking activity to measure and control pipeline and warehouse risk and monitor hedge effectiveness.

1. A rate lock creates an option-like commitment

Suppose a borrower locks a 30-year mortgage rate for 45 days. The lender has committed to honour that rate if the borrower satisfies the conditions and closes within the lock period.

If market mortgage rates rise after the lock, the borrower has an attractive below-market rate and is more likely to stay. The lender now holds a commitment to originate a loan whose market value may have fallen.

If market rates fall, the borrower has more incentive to renegotiate, abandon the application or refinance through another lender. The lender’s pipeline can therefore shrink precisely when the original hedge may have gained value.

This asymmetry makes the pipeline option-like: the borrower has meaningful behavioural choice while the bank is trying to predict which locks become assets.

2. Pull-through probability turns locked volume into expected volume

Let L_i be the locked principal of application i and p_i its estimated pull-through probability.

Expected closing exposure = Σ p_iL_i.

If the pipeline contains S$100 million of locks and the weighted average pull-through estimate is 75%, the expected funded exposure is roughly S$75 million before other product and sensitivity adjustments.

The 75% is not a constant of nature. Pull-through can depend on loan-processing stage, purchase versus refinance, borrower characteristics, lock age, current market rates relative to locked rates and operational approval conditions.

3. Fallout is the complement—but not always a simple 1 − p

In a simple binary model, fallout probability is 1 − pull-through probability. A 75% expected pull-through implies 25% expected fallout.

Real pipelines have more states:

• approved and likely to close;
• awaiting appraisal or documentation;
• rate renegotiated;
• lock extended;
• product changed;
• application withdrawn;
• credit declined;
• funded and moved into warehouse.

A richer model estimates transitions among these states rather than reducing every application to one static probability.

4. Pull-through is rate-sensitive

Suppose a borrower locked 6.0%. If current comparable market rates rise to 6.8%, the lock becomes valuable and pull-through can increase. If current rates fall to 5.3%, the borrower may have stronger incentive to seek a lower rate and pull-through can decrease.

This creates wrong-way hedge risk: the amount that needs hedging changes when the same market variable changes the value of the hedge.

A robust model therefore estimates not merely p but p(r, stage, age, borrower, product, channel).

5. From expected principal to rate sensitivity

Two mortgage locks of equal principal can have different interest-rate sensitivity because coupon, expected maturity, product, servicing value and expected prepayment differ.

The bank therefore needs an economic sensitivity measure such as price value per basis point, duration or another approved mortgage-market measure.

A simplified expected pipeline DV01 is:

Pipeline DV01 ≈ Σ p_i × DV01_i.

The hedge then seeks an offsetting DV01, subject to basis, convexity, liquidity and execution constraints.

6. TBA mortgage-backed securities are a natural hedge instrument

Mortgage lenders often use forward sales or TBA mortgage-backed securities because the value of conforming mortgages is closely related to the value of the securities into which many of those loans can ultimately be delivered or hedged.

A simplified hedge ratio is:

Hedge notional ≈ − expected pipeline sensitivity / hedge-instrument sensitivity per unit.

If expected pipeline DV01 is −S$120,000 and one hedge unit contributes +S$4,000 of DV01, a first-order hedge needs about 30 units.

The OCC handbook discusses forward commitments and mortgage securities as common hedge tools but emphasises that hedge selection must reflect the actual pipeline, liquidity and basis risk.

7. Hedge coverage is not always 100% of locked principal

Hedging 100% of gross locked principal can over-hedge the bank if only 70% is expected to close. Hedging only 70% of principal can still be wrong if the closing loans and hedge differ in duration or coupon sensitivity.

The correct object is rate-sensitive expected economic exposure, not gross application volume.

Once a loan funds and enters the warehouse, fallout risk disappears for that loan. The hedge problem changes from uncertain pipeline exposure to funded mortgage inventory risk.

8. Basis risk: the hedge and loan do not move identically

A mortgage loan is not a TBA security. Differences can arise from:

• coupon;
• loan size and product;
• credit characteristics;
• servicing value;
• delivery eligibility;
• prepayment behaviour;
• market liquidity;
• investor demand;
• guarantee and pooling economics.

If mortgage values fall 80 basis points while the chosen hedge gains only the equivalent of 65 basis points, the remaining 15-basis-point difference is basis exposure rather than a failure of simple arithmetic.

9. Product mix changes can create hedge drift

Suppose a pipeline begins as mostly 30-year fixed-rate conventional mortgages. Over the week, more borrowers switch into shorter or different products. Gross locked principal can remain unchanged while the appropriate hedge changes.

The system should therefore track hedge buckets by product, coupon, lock status and investor/delivery route rather than manage the pipeline as one number.

10. Pull-through models need backtesting by rate regime

A pull-through model estimated during stable rates can fail when rates move rapidly. Borrower incentives change, lock-extension behaviour changes and lender processing capacity can change.

Validation should compare predicted and realised closing rates by:

• lock age;
• processing stage;
• purchase/refinance;
• rate movement since lock;
• channel;
• product;
• loan officer or operational segment where governance permits;
• market regime.

A model that is accurate on average can still systematically over-hedge falling-rate periods and under-hedge rising-rate periods.

11. Reverse-price risk can appear when fallout changes

If rates fall sharply, mortgage prices can rise. The short hedge can lose money. At the same time, borrower fallout can rise, causing fewer locked loans to close and leaving less mortgage value to offset the hedge loss.

This is why the OCC identifies fallout and reverse-price risk separately. The hedge loss is not necessarily evidence that the hedge was foolish; it can be evidence that the expected underlying asset disappeared faster than the hedge was resized.

12. Timing risk: a good hedge applied at the wrong moment is still wrong

Pipeline data arrive continuously. Locks are entered, cancellations are recorded, loans fund and investor commitments are updated. If the hedge engine receives those states late, the bank can trade yesterday’s exposure.

A strong system therefore monitors:

• lock-entry latency;
• status-update latency;
• hedge execution time;
• market-data timestamps;
• pipeline-to-hedge reconciliation;
• stale applications and expired locks.

13. Rate-lock extensions create a second option layer

A borrower who cannot close before the original lock expires may request an extension. Depending on the product and terms, extension cost can be borne by borrower, lender or seller. The economic maturity of the commitment changes, which changes hedge duration and expected pull-through.

An extension should therefore be a new state with a new expected closing date, not a cosmetic change to the old lock record.

14. Hedge effectiveness should be measured as a combined portfolio

The question is not whether the hedge itself made money. The question is whether:

pipeline value change + warehouse value change + hedge P&L

remained inside approved tolerances across market moves.

A hedge can lose money when rates fall and still be effective if the mortgage pipeline gains more. Conversely, a profitable hedge can hide a larger loss in the underlying pipeline.

15. Creative-work lens: airport standby seats

An airline can have 100 passengers booked but know from history that some will not appear. Treating all 100 as certain creates one kind of planning error; assuming the historical no-show rate never changes creates another. A mortgage lock pipeline has the same structural uncertainty: locked volume is not identical to eventual closing volume.

The analogy helps with pull-through. Mortgage hedging adds market sensitivity, basis risk and borrower option behaviour that airline seats do not have.

16. The mortgage-pipeline hedging algorithmic pipeline

1. Capture every active rate lock and committed product.
2. Estimate pull-through by application state.
3. Map expected closing date and loan/product sensitivity.
4. Calculate expected pipeline market exposure.
5. Add funded warehouse exposure.
6. Select approved hedge instruments and map their sensitivities.
7. Calculate first-order hedge coverage.
8. Adjust for basis, convexity, liquidity and concentration.
9. Execute and record hedge positions.
10. Recompute after new locks, fallout, funding or product changes.
11. Stress rapid rate rises and falls.
12. Backtest pull-through and hedge effectiveness.
13. Reconcile pipeline, warehouse, hedge and secondary-market delivery states.

17. Failure modes

• Gross-lock hedging. Every locked dollar is treated as certain to close.
• Static pull-through. Rate moves do not change fallout assumptions.
• Principal-only hedge ratio. Sensitivity differences by coupon/product are ignored.
• Basis blindness. TBA and loan values are assumed to move identically.
• Pipeline latency. Expired or cancelled locks remain in hedge calculations.
• Extension blindness. Lock duration changes without hedge duration changing.
• P&L silo. Hedge gains/losses are judged without the underlying pipeline.
• Average-model comfort. Pull-through appears calibrated overall but fails directionally in volatile rate regimes.

18. Diagnostics and falsifiers

• What is predicted versus realised pull-through by lock stage?
• How does pull-through change when market rates move 50 or 100 basis points from lock?
• What is expected pipeline DV01 after pull-through weighting?
• How much of residual P&L is basis rather than rate exposure?
• Which product contributes most to hedge mismatch?
• How quickly are cancellations and funded loans removed from the pipeline?
• Does hedge coverage remain reasonable after a large lock-extension wave?
• Can an independent team reproduce combined pipeline-plus-hedge P&L?

Suppose someone claims, “We hedged 80% of locked principal, so the pipeline was 80% hedged.” A falsifier is a product mix whose expected closing sensitivity is materially different from 80% of gross principal. Hedge percentage by principal is not the same thing as hedge percentage by economic sensitivity.

19. Verification and update triggers

• recalibrate pull-through after large rate-regime changes;
• backtest by channel, product and processing stage;
• independently calculate hedge sensitivities;
• reconcile locks, funded loans and hedge trades daily;
• monitor basis spreads and market liquidity;
• review model changes after new products or delivery outlets;
• stress operational delays and stale pipeline data;
• retain historical pipeline snapshots so hedge decisions remain reconstructable.

Connections across the finance-and-banking algorithms lane

• Interest-rate hedge construction — the general sensitivity-matching framework.
• Mortgage prepayment — another borrower-behaviour option that changes duration.
• Agency MBS cash-flow algorithms — the security-side cash flows used after mortgages enter pools.
• Model validation — pull-through is a model, not an observed future fact.

Research anchors

• OCC — Mortgage Banking Comptroller’s Handbook.
• Federal Reserve — Interagency Advisory on Mortgage Banking.
• OCC — Mortgage Banking / Residential Lending resources.

The deeper lesson

Mortgage pipeline hedging is the mathematics of hedging an uncertain future asset. A rate lock has value before the loan exists. Pull-through converts gross promises into expected closings. Market sensitivity converts expected closings into hedge need. Fallout changes with the same rates that move the hedge. A strong system therefore never asks only, “How much is locked?” It asks, “How much is likely to become a loan, how sensitive will that loan be, and how quickly can the hedge change when the borrower changes the answer?”

Educational note: This article explains public mortgage-banking and hedging mathematics. It is not mortgage advice, trading advice or a hedge recommendation for any institution.