6. Asset-Liability Management and Interest Rate Risk

A bank earns most of its income from the difference between the rates it receives on assets and the rates it pays on liabilities, and that difference moves every time market rates move. Asset-liability management (ALM) is the discipline that measures, limits and hedges this interest rate risk in the banking book, and that allocates the resulting margins fairly through funds transfer pricing. For engineers, ALM is a data and simulation problem: cash-flow engines, behavioral models, scenario generators, hedge registers and FTP engines that must be accurate, reproducible and reconciled to the general ledger.

<a id="61-alm-foundations"></a>

6.1 ALM Foundations

This subchapter explains what ALM is trying to achieve and introduces the core lenses and risk types used to describe interest rate risk in the banking book (IRRBB).

Objectives of ALM

Why it matters. Banks perform maturity transformation: they borrow short (deposits that can leave at any time, short-term wholesale funding) and lend long (mortgages, corporate loans). This is economically useful and profitable, because long-term rates are usually higher than short-term rates, but it means assets and liabilities react differently when rates change. Without deliberate management, a rate move that nobody at the bank caused can wipe out a year of earnings or a large part of capital. The US savings and loan crisis of the 1980s and the failure of Silicon Valley Bank in 2023 were both, at their core, ALM failures.

What ALM is. ALM manages the balance sheet as a whole, rather than product by product, to balance profitability against risk within the bank's risk appetite. Its main objectives are:

Objective	What it means in practice
Stabilize earnings	Keep net interest income (NII) within limits across rate scenarios
Protect economic value	Limit the change in the present value of the balance sheet when rates move
Secure liquidity and funding	Ensure the maturity profile can be funded in normal and stressed conditions (see Chapter 5)
Optimize risk-return	Take deliberate, compensated positions (for example, a controlled amount of maturity mismatch)
Comply with regulation	Meet IRRBB, liquidity (LCR, NSFR) and capital requirements
Support business decisions	Inform product pricing, product design and capital allocation through FTP

Scope: banking book vs trading book. ALM covers the banking book: positions held to maturity or for the long term, such as loans, deposits, the investment portfolio and own debt, which are mostly accounted at amortized cost. Positions held for trading are managed as market risk (Chapter 8).

Governance. ALM decisions are taken by the Asset-Liability Committee (ALCO), typically chaired by the CFO, with the Treasurer, CRO, heads of major business lines and the head of ALM. ALCO approves the risk appetite and limits, behavioral model assumptions, hedging strategy and the FTP methodology, and reviews IRRBB and liquidity reports, usually monthly.

                       +----------------------------+
                       |            ALCO            |
                       | limits, models, hedges,FTP |
                       +-------------+--------------+
                                     |
        +----------------------------+----------------------------+
        |                            |                            |
  Treasury / ALM               Risk (2nd line)            Finance / Controlling
  (1st line: measures,         (independent limits,       (NII planning, hedge
   hedges, runs FTP)            model validation)          accounting, reporting)
        |
        v
  +-------------------- Banking book balance sheet --------------------+
  |  ASSETS                              |  LIABILITIES + EQUITY        |
  |  Mortgages (fixed, long)             |  Sight deposits (no maturity)|
  |  Corporate loans (floating)          |  Term deposits (short)       |
  |  Bond portfolio (fixed)              |  Issued bonds (fixed)        |
  |  Central bank reserves (overnight)   |  Equity (no maturity, no cost)|
  +--------------------------------------+------------------------------+

Analogy. ALM is the navigator of a ship. Sales teams are the engines pushing for speed (volume and margin); the navigator watches the weather (rates), the hull (capital) and the fuel (liquidity), and adjusts course so the ship arrives without capsizing.

Example: how maturity transformation creates risk. A simple bank holds EUR 1,000m of 5-year fixed-rate mortgages at 4.0%, funded by EUR 900m of deposits at 1.0% and EUR 100m of equity.

Scenario	Interest income	Interest expense	NII
Today	1,000 x 4.0% = 40.0	900 x 1.0% = 9.0	31.0
Rates +200bp, deposit rates follow one-for-one	40.0 (fixed for 5 years)	900 x 3.0% = 27.0	13.0

NII falls by 58% although nothing went wrong with any customer. ALM exists to decide, in advance, how much of this exposure the bank is willing to keep and how to hedge the rest.

Key takeaways

• ALM manages the banking book as a whole to balance profitability, interest rate risk and liquidity within the risk appetite.
• Maturity transformation is the source of both the bank's margin and its interest rate risk.
• ALCO is the governing body for limits, behavioral assumptions, hedging and FTP.
• ALM covers the banking book; trading positions are managed under market risk.

🧪 Practice

1. Beginner: List three banking book items on each side of the balance sheet and state whether each is typically fixed-rate or floating-rate.
2. Intermediate: In the example above, compute NII if deposit rates move by only 40% of the market move (a "beta" of 0.4) after a +200bp shock. What does this suggest about the importance of modeling deposit behavior?
3. Interview: "Why do regulators require a separate IRRBB framework instead of treating the banking book like the trading book?" Hint: accounting basis (amortized cost vs fair value), long holding periods, behavioral options embedded in customer products, and the role of earnings stability for a deposit-funded institution.

Repricing Gap Analysis

Why it matters. The first question ALM asks is simple: when will the rates on my assets and liabilities change? If more liabilities than assets reprice in the next three months and rates rise, funding costs go up before asset yields do and NII falls. Repricing gap analysis answers this question with a single table, which is why it is still the first report most ALCOs look at.

What it is. Every rate-sensitive position is slotted into a time bucket according to its next repricing date, not its final maturity:

• A fixed-rate loan reprices at maturity (or at each principal payment for amortizing loans).
• A floating-rate loan reprices at its next rate reset, for example in 3 months for a loan indexed to 3M Euribor, even if it matures in 7 years.
• Non-rate-sensitive items (equity, fixed assets, non-interest-bearing balances) go into a separate bucket.

For each bucket:

• RSA: rate-sensitive assets repricing in the bucket.
• RSL: rate-sensitive liabilities repricing in the bucket.
• Periodic gap = RSA - RSL.
• Cumulative gap = sum of periodic gaps up to and including the bucket.

A positive gap means the bank is asset-sensitive (benefits from rising rates); a negative gap means it is liability-sensitive (benefits from falling rates). The approximate change in NII over the next 12 months is:

  dNII ~ sum over buckets within 1 year of:  Gap_i x dr x (1 - t_i)

  where t_i is the bucket midpoint in years, so (1 - t_i) is the fraction
  of the year during which the new rate applies.

Example: gap report (EUR m), +100bp parallel shock.

Bucket	RSA	RSL	Periodic gap	Cumulative gap	Midpoint t	dNII (+100bp)
0-1M	300	500	-200	-200	0.042	-200 x 1% x 0.958 = -1.92
1-3M	400	300	+100	-100	0.167	+100 x 1% x 0.833 = +0.83
3-6M	200	400	-200	-300	0.375	-200 x 1% x 0.625 = -1.25
6-12M	300	200	+100	-200	0.750	+100 x 1% x 0.250 = +0.25
1-5Y	600	300	+300	+100	-	outside 12M horizon
>5Y	400	100	+300	+400	-	outside 12M horizon
Non-sensitive	100	500	-400	0	-	-
Total	2,300	2,300				-2.08

The bank is liability-sensitive within one year: a 100bp rise reduces NII by about EUR 2.1m. Beyond one year it is strongly asset-sensitive, which will show up as an economic value exposure instead.

  Periodic gap (EUR m), one # = 50m
                   negative | positive
  0-1M                 #### |
  1-3M                      | ##
  3-6M                 #### |
  6-12M                     | ##
  1-5Y                      | ######
  >5Y                       | ######
  Non-sensitive    ######## |

Limitations.

Limitation	Consequence
Timing within a bucket is ignored	A position repricing on day 1 and one on day 29 look identical
Parallel shift assumed	Curve shape changes (yield curve risk) are not captured
Single reference rate assumed	Basis risk is invisible
Contractual repricing	Prepayments and deposit behavior (optionality) are missed unless behavioral slotting is applied
Static balance sheet	New business and growth are ignored
Earnings focus	Cash flows beyond the horizon do not affect the result

Analogy. A gap report is like a calendar of contract renewals for a landlord who also rents from someone else. If more of the landlord's own leases come up for renewal this quarter than the leases of their tenants, a rise in market rents hurts before it helps.

Key takeaways

• Positions are bucketed by next repricing date, not by maturity.
• Gap = RSA - RSL; positive gap is asset-sensitive, negative gap is liability-sensitive.
• dNII is approximately gap x rate change x fraction of the year remaining.
• Gap analysis is simple and intuitive but ignores timing within buckets, curve shape, basis risk and optionality.

🧪 Practice

1. Beginner: Slot each item into a bucket: a EUR 50m loan maturing in 6 years with a 6M Euribor reset due in 2 months; a 9-month fixed-rate term deposit; a 2-year fixed-rate bond.
2. Intermediate: Using the gap table above, compute dNII for a 50bp rate cut. Then recompute assuming the 0-1M liabilities are sight deposits whose rate moves by only 30% of the market move.
3. Interview: "How would you build a gap report generator over a heterogeneous set of source systems?" Hint: a common position schema with next repricing date and reset frequency, configurable bucket definitions, behavioral slotting rules versioned separately from data, and reconciliation of bucket totals to balance sheet totals.

Earnings vs Economic Value Perspective

Why it matters. There are two legitimate ways to ask "how much does a rate move hurt us?", and they can give opposite answers. Management and equity analysts care about next year's earnings; regulators and long-term creditors care about whether the balance sheet is worth less than it was. A bank that optimizes only one lens can build up large hidden risk in the other.

The two perspectives.

• Earnings perspective (NII): how much net interest income changes over a short horizon, typically 1 to 3 years, under a rate scenario. It captures the timing of repricing within the horizon.
• Economic value perspective (EVE): how much the present value of all future banking book cash flows changes, over the full life of the positions. It is a long-term, "liquidation value" view of interest rate risk.

Aspect	Earnings (NII)	Economic value (EVE)
Question	How will profits change in the next 1-3 years?	How much is the balance sheet worth today after the shock?
Horizon	Short (typically 12 months, sometimes 3-5 years)	Full remaining life of all positions
Measure	Accrual income, as reported in the P&L	Present value of cash flows
Captures	Repricing timing, deposit pass-through	Long-dated fixed-rate positions, duration mismatch
Blind spot	Cash flows beyond the horizon	Timing of income, new business
Typical users	CFO, investors, budgeting	Regulators, risk management, capital adequacy

The tension in practice: investing equity. Equity has no contractual repricing and no interest cost. In regulatory EVE calculations equity is excluded from liabilities, so the bank must decide in which assets its equity is "invested".

Equity of EUR 100m invested in	NII effect of -200bp	EVE effect of -200bp
Overnight assets	About -2.0m every year (income falls with rates)	About 0 (overnight assets do not change in value)
10-year fixed bonds, modified duration 8.5	About 0 for years (coupons are locked)	About +17m (100 x 8.5 x 2%); and about -17m for +200bp

Stabilizing NII creates EVE sensitivity, and vice versa. Many banks therefore invest equity and stable deposits in a ladder of fixed-rate tranches (for example, evenly spread over 1 to 10 years) that accepts a moderate amount of each risk.

Analogy. NII is your monthly salary; EVE is your net worth. Locking in a 10-year fixed-rate mortgage keeps your monthly budget stable (earnings) but the market value of your debt changes every day (economic value). Neither view is wrong; you need both.

Key takeaways

• NII measures short-term earnings impact; EVE measures the change in present value over the full life of positions.
• The two can move in opposite directions; hedging one can increase the other.
• The treatment of equity and non-maturity deposits is a key driver of the trade-off.
• Regulation (BCBS IRRBB standards) requires banks to measure and limit both.

🧪 Practice

1. Beginner: For each, say whether NII, EVE or both would capture the risk: a 20-year fixed-rate loan funded overnight; a 3-month mismatch between floating loans and floating funding.
2. Intermediate: A bank invests EUR 200m of equity in a ladder with an average modified duration of 3.0. Estimate the EVE change for +200bp and explain qualitatively how NII evolves over the next five years.
3. Interview: "A bank shows a small NII sensitivity but a large EVE sensitivity. What could explain it, and is it a problem?" Hint: long fixed-rate assets funded by liabilities that reprice after the NII horizon, or deposits modeled as long-dated; consider whether the EVE loss would ever be realized and under which conditions (forced sale, deposit flight).

Basis Risk

Why it matters. Even if every asset and liability repriced on exactly the same date, the bank could still lose money, because they may reprice off different reference rates. If the rate the bank receives moves less than the rate it pays, the margin shrinks with no change in repricing dates at all. Basis risk is easy to overlook in gap reports because it is invisible when all rates are assumed to move together.

What it is. Basis risk is the risk from imperfect correlation between the rates on assets and liabilities that reprice at similar times. Common forms:

Type	Example	Typical driver
Reference rate basis	Loans on 3M Euribor, funding on €STR (overnight)	Bank credit and liquidity premium embedded in term IBOR rates
Tenor basis	Assets on 6M Euribor, liabilities on 1M Euribor	Different term premiums by tenor
Administered rate basis	Deposit rates set by the bank vs loans linked to market rates	Competition, commercial decisions
Currency basis	EUR assets funded with USD swapped into EUR	Cross-currency basis in FX swap markets
Central bank basis	Assets at policy rate corridor rates vs market funding	Changes in excess liquidity, corridor width

Example: reference rate basis in stress (EUR m). A bank holds EUR 1,000m of loans paying €STR + 2.00%, funded with EUR 1,000m of notes paying 3M Euribor + 0.30%. Both reprice frequently, so a gap report shows no exposure.

  Margin = (€STR + 2.00%) - (3M Euribor + 0.30%) = 1.70% - (Euribor - €STR spread)

Euribor - €STR spread	Annual margin	NII on 1,000
0.10% (calm market)	1.60%	16.0
0.60% (interbank stress, as seen in 2008 and in a weaker form in 2020)	1.10%	11.0

A purely technical spread widening removes EUR 5m of NII. The bank could hedge it with a basis swap (receive 3M Euribor, pay compounded €STR plus a spread).

Measuring it. Basis risk is measured by shocking the spreads between reference rates separately from the base curve, using historical or hypothetical basis scenarios, and by modeling administered rates through pass-through coefficients rather than assuming they follow market rates one-for-one.

Analogy. Two runners start together, but one is tied to the motorway and the other to the train line. Most days they arrive at the same time; on the day the train stops, the gap between them opens up suddenly.

Key takeaways

• Basis risk arises when assets and liabilities reprice at the same time but off different indices.
• Main forms: reference rate, tenor, administered rate and currency basis.
• It widens in stress, exactly when margins are under pressure for other reasons.
• Measure it with spread scenarios and hedge it with basis swaps or by aligning product indices.

🧪 Practice

1. Beginner: Identify the basis risk: a bank funds 6M Euribor loans with 1M Euribor interbank borrowing.
2. Intermediate: In the example, what basis swap would neutralize the exposure? Compute the bank's NII under both spread scenarios after the hedge, assuming the swap pays €STR + 0.10% against 3M Euribor.
3. Interview: "How would you represent reference rates in a position database so that basis risk can be measured?" Hint: store the index identifier, tenor, spread, reset frequency, lookback and compounding convention per position, not just a "floating" flag; scenarios then shock each index curve separately.

Optionality Risk in Banking Book

Why it matters. Many banking products contain options that the customer, not the bank, decides to exercise. Customers exercise them when it benefits them, which is when it hurts the bank: borrowers refinance when rates fall, depositors move money to higher-paying products when rates rise. This makes the bank's exposure asymmetric and means that simple duration measures understate the risk in large rate moves.

Types of optionality.

Type	Description	Examples
Automatic options	Exercised mechanically when a rate crosses a level, like a financial option	Caps and floors on floating loans; zero floors on Euribor in loan contracts; rate caps on deposits
Behavioral options (assets)	Customer decision influenced by rates and other factors	Mortgage prepayment, early repayment of consumer loans, drawdowns of credit lines
Behavioral options (liabilities)	Customer decision on deposits	Early withdrawal of term deposits, switching from sight to term deposits, withdrawal of savings

Example: the asymmetry of prepayment. A bank holds fixed-rate mortgages at 4.0% funded and hedged on the assumption that they will run to maturity.

  Rates fall to 2%                         Rates rise to 6%
  ------------------                       ------------------
  Borrowers refinance elsewhere            Borrowers keep their 4% loans
  Bank receives cash early                 Loans last longer than expected
  Must reinvest at 2%                      Bank stuck at 4% while funding costs rise
  Hedge (pay-fixed swap) now has no        Hedge too small for the longer exposure
  asset behind it -> loss on the swap
           |                                         |
           +------------- Bank loses in both directions ------------+

In value terms, the mortgage book has negative convexity: its value rises less than a plain bond when rates fall (because prepayments cap the gain) and falls about as much, or more, when rates rise (because the expected life extends).

Illustrative value changes (% of book) for a 5-year plain bond and a mortgage book with similar duration:

Rate move	Plain fixed-rate bond	Mortgage book with prepayment
-200bp	+9.5%	+5.0% (prepayments cap the gain)
-100bp	+4.6%	+3.6%
+100bp	-4.4%	-4.7%
+200bp	-8.5%	-9.8% (expected life extends)

Floors in the negative rate era. Between 2014 and 2022 Euribor was negative. Many loan contracts had a 0% floor on the index, so the bank kept receiving at least the margin; at the same time retail deposit rates could not go below 0% in practice. These two floors had opposite effects, and banks that hedged loans with plain swaps found the hedge no longer matched the floored loan cash flows.

Analogy. Selling an option to a customer is like giving a friend the right to return a borrowed car whenever they like. They will return it early when a better one appears and keep it as long as possible when cars become expensive.

Key takeaways

• Banking book options are mostly held by customers and exercised against the bank.
• Automatic options behave like financial options; behavioral options depend on customer behavior and require models.
• Optionality creates asymmetric, non-linear exposures (negative convexity) that duration alone does not capture.
• Hedging optionality may require options (caps, floors, swaptions) or dynamic hedge rebalancing.

🧪 Practice

1. Beginner: Classify as automatic or behavioral: a 0% floor on Euribor in a corporate loan; a mortgage borrower repaying early after selling the house; a depositor breaking a 2-year term deposit.
2. Intermediate: A bank hedged EUR 500m of 10-year fixed mortgages with pay-fixed swaps. Rates fall 150bp and prepayments double. Describe the impact on the bank's net position and propose how to rebalance the hedge.
3. Interview: "Why does a duration-based EVE estimate understate losses for a mortgage portfolio in a large rate move?" Hint: duration is a first-order (linear) measure; with negative convexity, effective duration itself changes with rates, so revaluation under full scenarios with rate-dependent cash flows is required.

Yield Curve Risk

Why it matters. Rates do not move up and down together. Short rates are driven mostly by central bank policy, long rates by inflation expectations, growth and term premiums. A bank whose assets are linked to long rates and whose liabilities are linked to short rates is exposed to changes in the shape of the curve, even if the average level of rates does not change.

Curve movements.

  Rate change (bp) by tenor        3M     1Y     5Y    10Y
  ---------------------------  -----  -----  -----  -----
  Parallel up                   +100   +100   +100   +100
  Bear steepener                   0    +20    +80   +100
  Bear flattener                +100    +80    +20      0
  Twist (short up, long down)    +50    +25    -25    -50

Movement	Short rates	Long rates	Typical cause
Parallel shift	Up or down	Same amount	Broad repricing of rates
Steepening	Fall or rise less	Rise or fall less	Rate cuts, rising inflation expectations
Flattening	Rise or fall less	Fall or rise less	Rate hikes, recession expectations
Twist / short-end shock	Move strongly	Little change	Central bank surprise

Measuring it: bucketed sensitivities. Instead of one duration number, ALM computes the sensitivity of value to a 1bp move at each key tenor (key rate PV01 or "bucketed DV01", see Chapter 8). The sum of the buckets equals the parallel sensitivity; the distribution shows the curve exposure.

Example (EUR m). A bank holds EUR 1,000m of 5-year fixed-rate assets (modified duration about 4.5) funded by EUR 1,000m of 3-month liabilities (duration about 0.25).

Tenor bucket	Asset PV01	Liability PV01	Net PV01
3M	0	1,000 x 0.25 x 0.0001 = 0.025	-0.025
5Y	1,000 x 4.5 x 0.0001 = 0.450	0	+0.450

Two scenarios with the same "average" move give very different results:

Scenario	EVE impact	12-month NII impact
Bear steepener: 3M +0bp, 5Y +100bp	-0.450 x 100 = -45.0	About 0: funding cost unchanged
Bear flattener: 3M +100bp, 5Y +0bp	+0.025 x 100 = +2.5 (liabilities worth less)	About -7.5: 1,000 x 1% x 0.75 of the year after first reset

The steepener is an economic value problem; the flattener is an earnings problem. A single parallel measure would have missed both.

Analogy. A bridge supported at both ends can survive the whole riverbank rising evenly, but if only one end sinks it twists. Curve risk is the twist.

Key takeaways

• Yield curve risk is exposure to changes in the slope and shape of the curve, not just its level.
• Bucketed (key rate) sensitivities reveal where on the curve the exposure sits.
• Steepeners usually hit EVE of long-asset banks; flatteners and short-rate shocks usually hit NII.
• Regulatory scenarios include steepener, flattener and short-rate shocks for this reason.

🧪 Practice

1. Beginner: Draw the curve before and after a bull flattener (long rates fall more than short rates).
2. Intermediate: Using the bucketed PV01 table, compute the EVE impact of a twist where 3M rises 50bp and 5Y falls 50bp.
3. Interview: "How would you compute key rate sensitivities for a banking book with millions of positions efficiently?" Hint: aggregate projected cash flows onto a common date grid first, then apply bump-and-revalue per key tenor on the grid rather than per position; consider parallelization by currency and portfolio, and caching of discount factors.

<a id="62-irrbb-measurement"></a>

6.2 IRRBB Measurement

This subchapter covers how banks quantify interest rate risk in the banking book: the two core metrics (NII and EVE), the regulatory shock scenarios, the behavioral models that drive the results, and the simulation approaches behind them.

Net Interest Income (NII) Sensitivity

Why it matters. NII is the largest line in most banks' income statements, often 60-80% of revenue for retail and commercial banks. Investors, rating agencies and management want to know how much it will change if rates move. NII sensitivity turns the gap intuition from 6.1 into a full projection that captures exact repricing dates, deposit pass-through and floors.

What it is.

  dNII(scenario) = NII over horizon under scenario - NII over horizon under base case

• Horizon: usually 12 months; many banks also report 2 to 3 years.
• Balance sheet assumption: usually constant balance sheet (maturing positions replaced with like-for-like positions at scenario rates); see the last topic of this subchapter.
• Shock type: instantaneous (rates jump on day one) or ramped (rates move gradually over 12 months). Ramps give smaller, more realistic numbers; regulators focus on instantaneous shocks.
• Behavioral inputs: deposit betas, deposit floors, prepayment speeds, all scenario-dependent.

How it is computed.

  Positions  -->  Cash flow and repricing schedule  -->  Apply scenario rate path
                                                               |
  Behavioral models (betas, floors, prepayments)  -------------+
                                                               v
              Interest income and expense per position per month
                                                               |
                                                               v
                       Sum over horizon  -->  NII(scenario) - NII(base)

Regulatory context. Under the BCBS IRRBB standards (2016) banks must measure NII sensitivity under rate shocks and disclose it. In the EU, the EBA supervisory outlier test (SOT) on NII compares the decline in NII under the parallel up and down shocks with 5% of Tier 1 capital; a breach triggers supervisory dialogue.

Example: 12-month NII sensitivity (EUR m). Assumptions: base 3M rate 3.00%; floating items reprice immediately (for simplicity); sight deposits have a beta of 30% and cannot go below 0%; term deposits reprice in 6 months with a beta of 100%.

Item	Balance	Base rate	Base income (expense)	+200bp effect	-200bp effect
Floating loans (3M + 2.00%)	500	5.00%	25.0	+10.0	-10.0
Fixed mortgages	400	3.50%	14.0	0.0	0.0
Fixed bonds	100	3.00%	3.0	0.0	0.0
Sight deposits (beta 30%, floor 0%)	500	0.50%	(2.5)	-3.0	+2.5 (floor: only -0.50%)
Term deposits (reprice in 6M)	300	2.50%	(7.5)	-3.0	+3.0
Wholesale (3M + 0.50%)	100	3.50%	(3.5)	-2.0	+2.0
Equity	100	-	-	-	-
NII			28.5	+2.0 (+7.0%)	-2.5 (-8.8%)

The result is asymmetric: the deposit floor means the bank cannot pass on the full rate cut to depositors, so the down shock hurts more than the up shock helps. In reality the down scenario would be worse still, because mortgage prepayments would accelerate.

Analogy. NII sensitivity is a household asking: "If rates go up 2% tomorrow, how much will my income from savings rise and my mortgage payments rise over the next year, given which contracts reset when?"

Key takeaways

• NII sensitivity is the change in projected NII over a horizon (usually 12 months) between a scenario and the base case.
• Results depend heavily on behavioral assumptions: deposit betas, floors and prepayments.
• Floors and betas make NII sensitivity asymmetric between up and down shocks.
• Instantaneous shocks are standard for regulation; ramps are common for internal planning.

🧪 Practice

1. Beginner: Recompute the +200bp result if the sight deposit beta is 50% instead of 30%.
2. Intermediate: Recompute both scenarios assuming floating loans and wholesale funding reprice on average after 1.5 months rather than immediately (that is, the new rate applies for 10.5 months).
3. Interview: "Your NII sensitivity jumped by 40% month on month with no major balance sheet change. How do you investigate?" Hint: decompose the change into data (volumes, new positions, missing feeds), market data (base curve), model parameters (betas, prepayment), and code or configuration changes; keep run inputs versioned so two runs can be diffed.

Economic Value of Equity (EVE)

Why it matters. NII looks only at the next year or so. A bank holding 30-year fixed-rate mortgages funded overnight may show a stable NII for a year while its economic value has collapsed. EVE captures the full-life impact of rate moves on the balance sheet, which is why regulators use it as the primary outlier metric.

What it is.

  EVE = PV(banking book asset cash flows) - PV(banking book liability cash flows)

  dEVE(scenario) = EVE(scenario curve) - EVE(base curve)

Regulatory conventions (BCBS 2016, EBA guidelines):

• Run-off balance sheet: existing positions only, no new business.
• Equity excluded from liabilities, so EVE shows the value available to equity holders.
• Risk-free discounting (for example, the €STR or swap curve); commercial margins may be included or excluded from cash flows, but consistently.
• Behavioral cash flows for non-maturity deposits and prepayable loans.
• Average repricing maturity of non-maturity deposits is capped (see the deposit modeling topic).

Duration approximation. For small parallel shocks:

  dEVE ~ -(D_A x PV_A - D_L x PV_L) x dy

  D_A, D_L = modified durations of assets and liabilities

Example: full revaluation from cash flows (EUR m). Net banking book cash flows are discounted with a flat curve at 3% (base) and 5% (after a +200bp parallel shock), annual compounding.

Year	Asset CF	Liability CF	Net CF	DF at 3%	PV at 3%	DF at 5%	PV at 5%
1	300	600	-300	0.9709	-291.26	0.9524	-285.71
2	300	200	+100	0.9426	+94.26	0.9070	+90.70
3	300	100	+200	0.9151	+183.03	0.8638	+172.77
5	200	50	+150	0.8626	+129.39	0.7835	+117.53
EVE					115.42		95.29

dEVE = 95.29 - 115.42 = -20.13, a loss of 17.4% of base EVE. The negative short-term net cash flow (liabilities due in year 1) loses little value; the positive long-dated net cash flows lose much more.

Outlier test. If the bank's Tier 1 capital were EUR 120m, the loss would be 16.8% of Tier 1, above the 15% supervisory outlier threshold for the worst of the six standard scenarios, and the supervisor would expect remediation (hedging, capital, or reduced risk).

Analogy. EVE is the price someone would pay today to take over all the bank's existing contracts. When rates rise, long fixed-rate assets become less attractive to the buyer, so they pay less.

Key takeaways

• EVE is the present value of banking book assets minus liabilities, excluding equity, on a run-off basis.
• dEVE is computed by revaluing all cash flows under shocked curves; duration gives a first-order approximation.
• Long-dated net asset positions drive EVE losses in rising rate scenarios.
• The EVE supervisory outlier test compares the worst dEVE across six scenarios with 15% of Tier 1 capital.

🧪 Practice

1. Beginner: A bank has assets of 1,000 with modified duration 2.5 and liabilities of 900 with modified duration 1.5. Estimate dEVE for +200bp.
2. Intermediate: Using the cash flow table, compute dEVE for a -200bp shock (flat 1%). Is the result symmetric with the +200bp case? Why not?
3. Interview: "Should commercial margins be included in EVE cash flows?" Hint: including them raises PV and changes duration (margins are long-dated fixed flows); excluding them isolates pure interest rate risk; the key requirement is consistency with discounting and with the hedge strategy.

Standard Rate Shock Scenarios

Why it matters. If each bank chose its own scenarios, supervisors could not compare results, and banks could pick mild scenarios. The BCBS IRRBB standards (2016) define six prescribed shock scenarios with currency-specific sizes, which are also used in the EU through EBA regulatory technical standards. Every IRRBB engine must be able to generate them exactly.

The six scenarios.

#	Scenario	Short rates	Long rates
1	Parallel up	+ parallel shock	+ parallel shock
2	Parallel down	- parallel shock	- parallel shock
3	Steepener	down	up
4	Flattener	up	down
5	Short rates up	up, fading with tenor	about unchanged
6	Short rates down	down, fading with tenor	about unchanged

Shock sizes (bp) for selected currencies.

Currency	Parallel	Short	Long
EUR	200	250	100
USD	200	300	150
GBP	250	300	150
CHF	100	150	100
JPY	100	100	100

How the curves are built. Short and long shocks are blended with an exponential decay (x = 4 years):

  Short shock:  dR_short(t) = S_short x exp(-t / 4)
  Long shock:   dR_long(t)  = S_long  x (1 - exp(-t / 4))

  Steepener:    dR(t) = -0.65 x |dR_short(t)| + 0.90 x |dR_long(t)|
  Flattener:    dR(t) = +0.80 x |dR_short(t)| - 0.60 x |dR_long(t)|
  Short up/down: dR(t) = +/- dR_short(t)

Example: EUR shocks (bp) at selected tenors.

Tenor (years)	Parallel up	Steepener	Flattener	Short up
0.25	+200	-147.2	+184.2	+234.9
1	+200	-106.6	+142.5	+194.7
5	+200	+17.7	+14.5	+71.6
10	+200	+69.3	-38.7	+20.5

Down scenarios mirror the up scenarios, subject to a post-shock floor: in the EU the floor starts at -150bp for overnight rates and rises by 3bp per year of maturity, reaching 0% at 50 years. Without a floor, down shocks applied from low starting rates would produce implausibly negative rates.

Aggregation across currencies. dEVE is computed per currency and then aggregated; in the standardized approach, gains in one currency are only partially recognized (weighted at 50%) against losses in others, so currency diversification is not fully credited.

Engineering notes.

• Shock parameters, decay constant and floors are regulatory parameters: store them as versioned, effective-dated configuration, not constants in code.
• Apply shocks to the risk-free zero curve per currency, then derive forwards and discount factors from the shocked curve.
• Log the exact curve used for each scenario run so results can be reproduced for supervisors.

Analogy. Standard scenarios are like a crash test protocol for cars: every manufacturer hits the same wall at the same speed and angle, so the results can be compared, even though real accidents vary.

Key takeaways

• Six prescribed scenarios: parallel up/down, steepener, flattener, short up/down.
• Shock sizes are currency-specific; for EUR they are 200 (parallel), 250 (short) and 100 (long) bp.
• Non-parallel shocks combine short and long shocks with an exponential decay over tenor.
• Post-shock floors and cross-currency aggregation rules must be implemented exactly.

🧪 Practice

1. Beginner: Compute the EUR short-rate-up shock at a tenor of 2 years.
2. Intermediate: Compute the USD steepener shock at 1 year and at 10 years, using short 300bp and long 150bp.
3. Interview: "Design the scenario generation component of an IRRBB engine." Hint: input base curves per currency, versioned parameter tables, deterministic shock functions with floors, outputs as full curves on a standard tenor grid, plus a test suite that reproduces published regulatory examples.

Non-Maturity Deposit Modeling

Why it matters. Current and savings accounts often fund 40-60% of a retail bank's balance sheet. Contractually they can be withdrawn today and their rate can be changed by the bank at any time, so a purely contractual view treats them as overnight. In reality much of the balance stays for years and its rate moves only partially with market rates. How these deposits are modeled is often the single largest driver of a bank's reported IRRBB, and of its hedging strategy.

What is modeled. A non-maturity deposit (NMD) model has three components:

Component	Question	Typical approach
Volume (stability)	How much of the balance will stay, and for how long?	Split into core (stable) and non-core (volatile); runoff or decay profile for the core
Rate (pass-through)	How much of a market rate change is passed to depositors?	Deposit beta, with lags, asymmetry and floors
Repricing profile	When does the bank effectively reprice the core portion?	Slot core balances across time buckets, which determines duration

Regulatory caps (BCBS 2016).

Deposit category	Maximum core share	Maximum average repricing maturity of core
Retail transactional	90%	5 years
Retail non-transactional	70%	4.5 years
Wholesale (non-financial)	50%	4 years

Example: slotting EUR 1,000m of retail savings.

• Statistical analysis of balance history shows 30% of the balance is volatile: slotted overnight.
• The 70% core (EUR 700m) is assumed to amortize linearly over 5 years: EUR 140m per yearly bucket.

  Bucket      O/N    0-1Y   1-2Y   2-3Y   3-4Y   4-5Y
  EUR m       300    140    140    140    140    140
  Average repricing maturity:
    core    = 2.5 years (midpoint of a linear 5-year runoff)
    overall = 0.7 x 2.5 = 1.75 years

Rate model: deposit rate = max(0, margin + beta x market rate). With beta = 0.3 and a +300bp market move, the deposit rate rises by only 90bp. The 210bp the bank does not pass on is the deposit franchise value, the reason why deposits are valuable funding and why EVE treats the core as a long-dated liability.

Why the choice matters.

Modeling choice	Liability duration	Reported risk	Hedge implied
All deposits overnight	Near zero	Large EVE loss on any fixed-rate assets when rates rise; NII loss when rates fall is hidden	Keep assets short or floating; pay-fixed swaps on fixed-rate assets
Core over 5-year runoff	Medium	Balanced	Receive-fixed swaps or fixed-rate assets matching the core profile
Core over 10 years (beyond caps)	Long	Understated rising-rate risk	Too much fixed-rate investment

Lesson from 2022-2023. As rates rose rapidly, depositors moved money from sight accounts into term deposits and money market funds, and betas rose faster than models based on the low-rate decade predicted. Silicon Valley Bank had invested deposit inflows in long fixed-rate securities; when concentrated, uninsured deposits left quickly, the assumed "core" vanished and unrealized losses became real.

Analogy. A café assumes regular customers will keep coming every morning even though they are free to go elsewhere. The core is the regulars; the volatile part is passing trade. A new café opening across the street (higher rates elsewhere) tests the assumption.

Key takeaways

• NMD models split balances into volatile and core, assign the core a behavioral maturity, and model rate pass-through with betas and floors.
• The modeled duration of deposits largely determines EVE sensitivity and the hedging strategy.
• Regulators cap the core share and the average repricing maturity.
• Betas and stability are regime-dependent; models calibrated in low-rate periods can fail in fast hiking cycles.

🧪 Practice

1. Beginner: A deposit pays 0.25% with beta 0.4 and a 0% floor. What is the new rate after a +150bp and after a -150bp market move?
2. Intermediate: Re-slot the EUR 1,000m example with 20% volatile and the core amortizing linearly over 8 years. Compute the average repricing maturity and check it against the BCBS caps for retail transactional deposits.
3. Interview: "How would you estimate the core portion of deposits from historical data?" Hint: minimum balance or envelope methods over a long history, percentile of monthly outflows, segmentation by product and customer type, exclusion of known one-off events, and stress-testing in rising rate regimes.

Prepayment Modeling

Why it matters. Mortgages and many consumer loans can be repaid early. A 20-year mortgage portfolio rarely lives 20 years; prepayments can cut its expected life in half. Because prepayment speeds depend on rates, ignoring them misstates cash flows, duration, NII and hedge sizes, and leaves the bank exposed to the optionality described in 6.1.

Drivers of prepayment.

Driver	Effect
Refinancing incentive	Contract rate minus current market rate; the larger the gap, the faster prepayments
Turnover	Home sales, relocations, divorces; occurs regardless of rates
Seasoning	New loans prepay slowly, speeds rise over the first years
Burnout	Borrowers who did not refinance at past opportunities are less likely to do so later
Seasonality	More house sales in spring and summer
Penalties and regulation	Early repayment fees slow prepayments; some jurisdictions cap them
Curtailment	Partial extra repayments

Measures.

  SMM (single monthly mortality) = prepaid principal in month / balance at start of month
  CPR (conditional prepayment rate, annualized) = 1 - (1 - SMM)^12
  SMM = 1 - (1 - CPR)^(1/12)

  Example: CPR = 10%  ->  SMM = 1 - 0.9^(1/12) = 0.874% per month

Models. In increasing order of sophistication:

1. Constant CPR per product, calibrated on history.
2. Incentive-based S-curve: CPR as a function of the refinancing incentive.
3. Statistical or machine learning models (logistic regression, survival models, gradient boosting) at loan level, using borrower, loan and market features.

  CPR
  35% |                                  ___________
      |                               _-'
  25% |                            _-'
      |                         _-'
  15% |                      _-'
      |                   _-'
   5% |________________.-'
      +-----+-----+-----+-----+-----+-----+-----+---> Incentive
          -200  -100    0   +100  +200  +300        (contract rate - market rate, bp)

  Negative incentive: only turnover (floor around 5%). Positive incentive:
  refinancing accelerates, then saturates (not everyone refinances).

Example: effect on weighted average life (WAL). A 20-year amortizing mortgage pool at 4%, monthly payments:

CPR	WAL (years)
0%	11.4
5%	8.1
10%	6.1
20%	3.7
30%	2.6

If the bank hedged on a 10% CPR assumption (WAL 6.1 years) and rates fall so that CPR rises to 20%, the assets shorten to 3.7 years while the swaps keep running: the bank is over-hedged and pays fixed on swaps that no longer have assets behind them, at a time when those swaps have lost value.

Scenario dependence. IRRBB calculations must use scenario-dependent prepayment speeds. The BCBS standardized framework, for example, scales base prepayment rates by multipliers such as 1.2 in rate-down scenarios and 0.8 in rate-up scenarios.

Analogy. Prepayment is like guests leaving a party early. Some leave whatever happens (turnover); many more leave if a better party starts down the street (refinancing incentive); and those who stayed through the last better offer probably stay to the end (burnout).

Key takeaways

• Prepayment shortens asset life and is driven by refinancing incentive, turnover, seasoning, burnout and penalties.
• SMM and CPR are the standard measures; CPR = 1 - (1 - SMM)^12.
• Prepayment speeds depend on rates, which creates negative convexity and hedge mismatches.
• IRRBB scenarios must use scenario-specific prepayment assumptions.

🧪 Practice

1. Beginner: Convert a CPR of 15% to an SMM, and an SMM of 1.5% to a CPR.
2. Intermediate: A pool of EUR 400m at 4.5% is hedged with EUR 300m of pay-fixed swaps sized on a 6-year WAL. Rates fall 150bp and the WAL drops to 4 years. Describe the hedge mismatch and two ways to manage it.
3. Interview: "Design the data pipeline for a loan-level prepayment model." Hint: monthly loan snapshots with balance, rate, age, LTV and region; label full and partial prepayments separately; join market rates at each snapshot date; avoid look-ahead bias in features; backtest predicted vs actual CPR by cohort.

Static vs Dynamic Balance Sheet Simulation

Why it matters. Every NII and EVE calculation has to answer: what happens to the balance sheet over time? If positions are allowed to mature without replacement, NII naturally falls; if growth plans are assumed, results depend on commercial forecasts. Different questions need different assumptions, and mixing them up produces misleading numbers.

Three balance sheet assumptions.

Assumption	What happens to maturing positions	Typical use
Run-off	Not replaced; balance sheet shrinks	Regulatory EVE; pure exposure of existing book
Constant (static)	Replaced like-for-like (same product, tenor, margin) at scenario rates	Regulatory NII sensitivity, internal limits
Dynamic	Driven by business plan: growth, product mix changes, scenario-dependent customer behavior and pricing	Budgeting, ICAAP, stress testing, strategic decisions

Example (EUR m). A bank holds EUR 1,000m of loans at 4%; EUR 200m mature in 6 months. Rates rise 200bp at time zero, so new loans are written at 6%. Loan interest income over 12 months:

Assumption	Calculation	Income
Run-off	800 x 4% + 200 x 4% x 0.5	36.0
Constant balance sheet	Run-off + 200 x 6% x 0.5 (replacement)	42.0
Dynamic (plan: +100m new loans from month 6)	Constant + 100 x 6% x 0.5	45.0

The same scenario gives three different answers, all correct for their purpose. The key is to label every reported number with its assumption.

Simulation engine architecture.

  +-------------+   +----------------+   +------------------+
  | Positions   |   | Market data    |   | Scenario         |
  | (contracts) |   | (base curves)  |   | generator        |
  +------+------+   +--------+-------+   +---------+--------+
         |                   |                     |
         v                   v                     v
  +-------------------------------------------------------------+
  | Cash flow engine: schedules, resets, amortization           |
  +-------------------------------+-----------------------------+
                                  |
  +-----------------------+       |       +----------------------+
  | Behavioral models     |------>+<------| New business rules   |
  | (NMD, prepayment)     |       |       | (run-off / constant /|
  +-----------------------+       |       |  dynamic plan)       |
                                  v       +----------------------+
  +-------------------------------------------------------------+
  | Aggregation: NII by month, PV by scenario, gap buckets      |
  +-------------------------------+-----------------------------+
                                  v
                    Reports, limits, ALCO packs

Analogy. Static simulation is a snapshot of a river with all inflows switched off; dynamic simulation is a model of the river including the rain forecast. The first tells you how deep the riverbed is; the second tells you whether it will flood.

Key takeaways

• Run-off, constant and dynamic balance sheets answer different questions.
• Regulatory EVE uses run-off; regulatory NII sensitivity uses a constant balance sheet.
• Dynamic simulations depend on business plans and scenario-dependent behavior, so they need strong governance.
• Every reported figure must state its balance sheet assumption.

🧪 Practice

1. Beginner: Which assumption would you use to answer "How much of next year's NII is already locked in by existing contracts?"
2. Intermediate: Extend the example: under a dynamic assumption, competitors reduce new loan margins by 30bp in the +200bp scenario. Recompute the dynamic income.
3. Interview: "How would you make a dynamic balance sheet simulation reproducible and auditable?" Hint: version positions, curves, scenario definitions, behavioral parameters and business plan inputs separately; store run manifests; make new business generation deterministic given its inputs.

<a id="63-hedging-the-banking-book"></a>

6.3 Hedging the Banking Book

This subchapter explains how banks use derivatives and modeling techniques to bring IRRBB within limits, and how hedges are tested and accounted for so that they reduce, rather than create, volatility in reported results.

Hedging Strategies with Swaps

Why it matters. Changing the balance sheet itself to fix an interest rate mismatch is slow and commercially costly: the bank cannot ask customers to switch from fixed to floating mortgages because ALM would prefer it. Interest rate swaps let treasury change the rate profile of the balance sheet synthetically, quickly and cheaply, without touching customer contracts.

What a swap does. In a plain vanilla interest rate swap (IRS) two parties exchange fixed-rate payments for floating-rate payments on a notional amount that is not exchanged.

Hedging need	Swap position	Effect
Fixed-rate assets (mortgages, bonds) funded at floating rates	Pay fixed, receive floating	Converts fixed assets into synthetic floating assets
Fixed-rate liabilities (issued bonds, covered bonds)	Receive fixed, pay floating	Converts fixed funding into synthetic floating funding
Stable deposits and equity treated as long-dated	Receive fixed, pay floating (or buy fixed bonds)	Locks in a long-term yield on the deposit base, stabilizing NII
Expected future fixed-rate lending	Forward-starting pay-fixed swap	Locks the rate before the loans are written
Amortizing loan books	Amortizing swap	Notional follows the expected balance profile
Prepayment or floor optionality	Swaptions, caps, floors	Hedges the non-linear part of the exposure

Example: hedging a fixed-rate mortgage book (EUR m). Treasury funds EUR 500m of 5-year fixed mortgages at 3.50% with floating funding at 3M Euribor + 0.20%. It enters a 5-year pay-fixed swap at 3.00% against 3M Euribor.

                  3.50% fixed                     3.00% fixed
   Mortgage  ------------------->   BANK   ------------------->  Swap
   borrowers                                <-------------------  counterparty
                                             3M Euribor
                                     |
                                     | 3M Euribor + 0.20%
                                     v
                                  Funding

   Net margin = 3.50% - 3.00% + Euribor - (Euribor + 0.20%) = 0.30%, locked for 5 years

Without the swap, and with 3M Euribor at 3.00% today, a 200bp rise in Euribor would turn the margin from +0.30% into -1.70%. With it, the margin is unchanged whatever Euribor does.

Sizing the hedge. Hedges are sized on sensitivity, not just notional:

  Hedge notional = PV01 of exposure / PV01 of swap per unit notional

  Mortgage book PV01 (5Y bullet, 3.5% coupon, 3% yield): about EUR 232k per bp
  5Y par swap PV01 per EUR 100m (annuity factor 4.58):   about EUR 45.8k per bp
  Notional = 232 / 45.8 x 100m  ~  EUR 507m

Analogy. A swap is like agreeing with a neighbor to swap electricity tariffs: you pay their fixed bill and they pay your variable one. Your house and your supplier do not change, but your exposure to price swings does.

Key takeaways

• Pay-fixed swaps hedge fixed-rate assets; receive-fixed swaps hedge fixed-rate liabilities or lock in yield on stable deposits.
• Swaps change the rate profile without changing customer contracts.
• Hedges are sized to match PV01 (or bucketed PV01), not simply notional.
• Optional exposures need option-based instruments or dynamic rebalancing.

🧪 Practice

1. Beginner: A bank issues a EUR 750m 7-year fixed-rate covered bond and wants floating-rate funding. Which swap does it enter?
2. Intermediate: In the mortgage example, compute the bank's annual net interest in EUR m with and without the swap if 3M Euribor averages 5.0%.
3. Interview: "What new risks does a bank take on when it hedges the banking book with swaps?" Hint: counterparty and CVA risk (Chapter 8), collateral and margin liquidity needs, basis between swap index and funding index, hedge accounting ineffectiveness, and model risk when the hedged item's behavior changes.

Macro vs Micro Hedging

Why it matters. A bank with a single large fixed-rate bond issue can hedge it with one matching swap. A bank with 300,000 fixed-rate mortgages cannot sensibly enter 300,000 swaps; it must hedge the net exposure of the portfolio. The choice between hedging individual items and hedging portfolios shapes both the risk management process and the accounting.

Definitions.

• Micro hedging: a specific, identified item (or a small group of near-identical items) is hedged with a specific instrument, often with matching terms. Example: a 10-year fixed bond issued by the bank and a receive-fixed swap with the same notional, dates and coupon.
• Macro (portfolio) hedging: the net exposure of a portfolio of many items, usually by repricing bucket, is hedged with a set of instruments that are rebalanced as the portfolio changes.

Aspect	Micro hedging	Macro hedging
Hedged item	Single, identified transaction	Portfolio or net position by bucket
Instrument match	Terms closely matched	Standard swaps by tenor bucket
Rebalancing	Rarely; until maturity	Frequent (monthly or more) as portfolio and behavior change
Netting	No; gross exposure hedged	Yes; natural offsets between assets and liabilities used
Typical use	Issued bonds, large loans, bond holdings	Mortgages, consumer loans, deposits
Accounting	Simpler hedge accounting	Portfolio hedge accounting (more complex)
Cost efficiency	Lower (no netting)	Higher (only the net is hedged)

Example: macro hedge by bucket (EUR m), hedge ratio 80%.

Bucket	Fixed-rate assets	Fixed-rate liabilities (incl. modeled core deposits)	Net exposure	Pay-fixed swap notional
1Y	300	180	120	96
2Y	250	170	80	64
3Y	270	120	150	120
4Y	160	100	60	48
5Y	190	100	90	72
Total	1,170	670	500	400

Micro hedging all assets would need EUR 1,170m of swaps; the macro approach uses the natural offset of fixed-rate liabilities and needs EUR 400m, leaving 20% of the net exposure open as a deliberate position.

  Monthly macro hedge cycle

  New positions + behavioral model run --> Bucketed net exposure --> Compare with target
         ^                                                                 |
         |                                                                 v
   Next month <-- Update hedge register <-- Execute swaps <-- Proposed hedge trades

Analogy. Micro hedging is buying insurance for each item in your house separately. Macro hedging is buying one home insurance policy sized for the total value, after accounting for what is already protected.

Key takeaways

• Micro hedges link specific items to specific instruments; macro hedges cover the net exposure of a portfolio.
• Macro hedging uses netting and is more cost-efficient, but requires frequent rebalancing and robust behavioral models.
• Retail banks rely mainly on macro hedging; micro hedging dominates for issued debt and large single exposures.
• Hedge ratios below 100% are often a deliberate ALCO decision.

🧪 Practice

1. Beginner: Classify as micro or macro: hedging a EUR 1bn fixed-rate senior bond issue; hedging the 3-year bucket of a credit card and personal loan portfolio.
2. Intermediate: Next month, prepayments reduce the 3Y and 4Y asset buckets by EUR 40m each. Recompute the net exposures and the swap adjustments needed to keep an 80% hedge ratio.
3. Interview: "How would you design a hedge register that supports both micro and macro hedges?" Hint: separate entities for hedged items (or portfolios and bucket layers), hedging instruments, and hedge relationships with designation dates; time-versioned links so relationships can be de-designated and re-designated; links to effectiveness results.

Hedge Effectiveness Testing

Why it matters. A hedge only helps if it actually offsets the risk. Changes in the hedged item and the hedge may diverge because of different fixing dates, discount curves, credit spreads or behavioral changes. Effectiveness testing measures this offset, both to manage risk and because hedge accounting requires proof that the relationship works.

What is tested.

• Prospective test: will the hedge be effective going forward? Performed at designation and periodically.
• Retrospective test / ineffectiveness measurement: how well did it actually offset changes in the period? Any ineffectiveness goes to profit or loss.

Framework	Requirement
IAS 39 (still widely used for macro fair value hedges)	Retrospective and prospective offset within 80-125%
IFRS 9	An economic relationship exists, credit risk does not dominate value changes, and the hedge ratio is consistent with risk management; no bright-line range, but ineffectiveness is measured and booked
US GAAP (ASC 815)	"Highly effective", commonly interpreted as 80-125%

Methods.

1. Dollar offset: ratio of the change in value of the hedge to the change in value of the hedged item attributable to the hedged risk.
2. Regression: regress hedge value changes on hedged item value changes over historical or simulated scenarios; typical criteria are slope between -0.8 and -1.25 and R-squared at least 0.8.
3. Hypothetical derivative method (mainly cash flow hedges): model the hedged item as a "perfect" derivative and compare it with the actual hedge.

Example: dollar offset (EUR m).

Period	Change in hedged item (hedged risk)	Change in swap	Ratio	Result
Q1	+1.20	-1.10	91.7%	Effective
Q2	+0.10	-0.16	160.0%	Fails 80-125%
Cumulative Q1-Q2	+1.30	-1.26	96.9%	Effective

Q2 illustrates the small numbers problem: when rates barely move, tiny absolute differences produce extreme ratios. Using cumulative offset or regression avoids false failures.

Common sources of ineffectiveness.

• Different discount curves (swaps discounted on €STR, hedged item on a different curve).
• Mismatched dates, notionals or day count conventions.
• Floating leg value of the swap, which has no counterpart in a fixed-rate hedged item.
• Counterparty credit (CVA, DVA) on the swap.
• Prepayments or deposit behavior deviating from what was assumed.

Analogy. Effectiveness testing is checking that an umbrella actually keeps you dry. An umbrella that covers only half of you still helps, but you need to know how wet you will get.

Key takeaways

• Effectiveness testing checks how well the hedge offsets changes in the hedged risk, prospectively and retrospectively.
• Dollar offset is intuitive but unstable for small moves; regression and cumulative approaches are more robust.
• IAS 39 and US GAAP use the 80-125% range; IFRS 9 relies on qualitative criteria plus measurement of ineffectiveness.
• Ineffectiveness always flows to profit or loss.

🧪 Practice

1. Beginner: A hedged bond changes by -4.0 due to rates; the swap changes by +3.4. Compute the dollar offset ratio and state whether it passes the 80-125% test.
2. Intermediate: List three reasons why a pay-fixed swap might not fully offset fair value changes of a fixed-rate mortgage portfolio attributable to interest rate risk.
3. Interview: "How would you implement an effectiveness testing service that runs monthly for thousands of hedge relationships?" Hint: batch revaluation of hedged items and instruments for the hedged risk only, cumulative histories per relationship, configurable test methods per relationship type, stored inputs for audit, and alerting on failures before period-end close.

Replicating Portfolios

Why it matters. Non-maturity deposits have no contractual maturity, yet treasury must decide how to invest them, how to hedge them and what transfer price to give the deposit business. The replicating portfolio approach answers all three questions with one construct: it represents the deposits as a portfolio of fixed-rate market instruments that behaves, in rates and volumes, as closely as possible like the deposits themselves.

What it is. A replicating portfolio is a set of rolling ladders of fixed-rate instruments of different tenors. In a rolling 5-year ladder, the balance is split into 60 monthly tranches; each month the oldest tranche matures and is reinvested at the current 5-year rate. The yield of the ladder is therefore a moving average of past rates:

  Yield of a rolling N-month ladder at time t:
     y(t) = (1/N) x sum over k = 0..N-1 of r_N(t - k)

  where r_N(t - k) is the N-month market rate observed k months ago.

  Rolling 5-year ladder (60 tranches)
  Month:     1     2     3    ...   59    60
            [==] [==] [==]   ...  [==] [==]   each tranche = 1/60 of the balance
             ^ matures this month, reinvested at today's 5Y rate

The weights of the ladders (for example, 30% overnight, 20% rolling 1-year, 20% rolling 3-year, 30% rolling 5-year) are chosen by optimization, typically to minimize the volatility of the margin between the portfolio yield and the deposit rate over a long history, subject to constraints such as volume stability and regulatory maturity caps.

Example. After a rapid rate rise, current rates are about 3%, but long ladders still reflect older, lower rates.

Ladder	Weight	Current ladder yield	Contribution
Overnight	30%	3.00%	0.90%
Rolling 1-year	20%	2.80%	0.56%
Rolling 3-year	20%	2.20%	0.44%
Rolling 5-year	30%	1.80%	0.54%
Portfolio	100%		2.44%

With a deposit rate of 0.80%, the deposit margin is 2.44% - 0.80% = 1.64%. Because the portfolio yield moves slowly, like the deposit rate, the margin is stable over the rate cycle.

Uses.

Use	How the replicating portfolio is applied
Investment and hedging	Treasury invests or receives fixed on swaps according to the ladder weights
FTP	The deposit unit is credited with the portfolio yield (see 6.4)
IRRBB	Ladder weights define the behavioral repricing profile of the deposits

Risks. The approach assumes the historical relationship between deposit rates, volumes and market rates holds. If deposits run off faster than the ladders (as in 2023 at some banks), treasury must unwind fixed-rate positions at a loss.

Analogy. A replicating portfolio is like explaining the taste of a soup through a recipe: 30% of this, 20% of that. You cannot see the soup's ingredients directly, but a recipe that tastes the same tells you how to make, price and protect it.

Key takeaways

• A replicating portfolio models non-maturity deposits as a mix of rolling fixed-rate ladders.
• Ladder yields are moving averages of past rates, which mimic the sluggish behavior of deposit rates.
• Weights are optimized for stable margins, subject to volume and regulatory constraints.
• The same portfolio drives investment, hedging, FTP and IRRBB, so its assumptions must be governed by ALCO.

🧪 Practice

1. Beginner: A rolling 12-month ladder has observed 12-month rates of 2.0% for the past 6 months and 3.2% for the 6 months before that. What is its current yield?
2. Intermediate: Recompute the portfolio yield in the example with weights 40% / 20% / 20% / 20%. Explain why a shorter portfolio gives a higher yield right now but a more volatile margin over the cycle.
3. Interview: "How would you calibrate replicating portfolio weights?" Hint: long history of deposit rates, volumes and market rates; objective function on margin volatility or margin level; constraints on weights and average maturity; out-of-sample testing across rate regimes; stability of weights between recalibrations.

Hedge Accounting Overview

Why it matters. Derivatives are measured at fair value through profit or loss, while most banking book items (loans, deposits, issued debt) are measured at amortized cost. A perfectly good economic hedge therefore creates accounting volatility: the swap's value swings through P&L, while the offsetting change in the loan's value is never recognized. Hedge accounting is a set of optional rules that removes this mismatch, provided strict conditions are met.

Accounting mismatch without hedge accounting.

  Rates rise 100bp
                                      Economic view       Accounting view (no hedge accounting)
  Fixed mortgage (amortized cost)     -10.0 in value      not recognized
  Pay-fixed swap (fair value)         +10.0               +10.0 in P&L
  Net                                   0.0               +10.0 P&L (and -10.0 when rates fall)

Types of hedge relationships.

Type	Hedges	Accounting treatment	Typical banking book use
Fair value hedge	Changes in fair value of a fixed-rate item	Hedged item's carrying amount adjusted for the hedged risk through P&L, offsetting the derivative	Fixed-rate mortgages, bonds held, issued fixed-rate debt
Cash flow hedge	Variability of future cash flows of floating-rate items or forecast transactions	Effective portion of the derivative's change goes to OCI (cash flow hedge reserve), recycled to P&L when the hedged cash flows occur	Floating-rate loans or funding, forecast issuance
Net investment hedge	FX exposure on investments in foreign subsidiaries	Effective portion to OCI (translation reserve)	Structural FX (Chapter 8)

Example: fair value hedge (EUR m). Rates rise; the pay-fixed swap gains 10.0; the change in fair value of the mortgages attributable to the benchmark rate is -9.6.

Line	P&L
Swap fair value change	+10.0
Basis adjustment to mortgages	-9.6
Net P&L (ineffectiveness)	+0.4

Example: cash flow hedge (EUR m). A pay-fixed swap hedges floating-rate funding. The swap gains 6.0; the hypothetical derivative representing the hedged cash flows changes by 5.8. The OCI reserve takes the lower of the two cumulative amounts, 5.8; the excess 0.2 (over-hedge) goes to P&L.

Frameworks.

• IFRS 9 introduced a more principles-based hedge accounting model, but allows banks to keep applying IAS 39 hedge accounting, and most European banks do so for macro hedges using the IAS 39 portfolio fair value hedge of interest rate risk, with the EU "carve-out" that permits hedging of core deposits and relaxes some effectiveness rules. The IASB has been developing a dedicated model for dynamic risk management (now called risk mitigation accounting) to replace it.
• US GAAP (ASC 815) offers fair value and cash flow hedges, including the portfolio layer method for closed portfolios of prepayable assets.

Requirements. At inception, formal documentation of the risk management objective, hedged item, hedging instrument, hedged risk, hedge ratio and effectiveness method; ongoing effectiveness assessment; and de-designation rules when the relationship ends.

Analogy. Hedge accounting is like filing both halves of a matched pair of receipts together. Without it, the accountant sees only the expense today and the refund next year, and your monthly statement swings wildly.

Key takeaways

• Hedge accounting removes the mismatch between derivatives at fair value and hedged items at amortized cost.
• Fair value hedges adjust the hedged item's carrying amount; cash flow hedges park the effective portion in OCI.
• Many banks still use IAS 39 portfolio fair value hedging (with the EU carve-out) for macro hedges.
• Formal documentation and effectiveness assessment are mandatory; ineffectiveness always hits P&L.

🧪 Practice

1. Beginner: For each, choose fair value or cash flow hedge: swapping a fixed-rate bond issue to floating; fixing the rate on floating-rate interbank borrowing.
2. Intermediate: In the cash flow hedge example, suppose the swap gains 5.5 and the hypothetical derivative changes by 5.8. How much goes to OCI and how much to P&L?
3. Interview: "What data must a system retain to support hedge accounting audits?" Hint: designation documentation with timestamps, hedge ratios, hedged risk definitions, the exact market data and models used for each valuation, effectiveness test results, basis adjustment amortization schedules, and de-designation events.

<a id="64-funds-transfer-pricing-ftp"></a>

6.4 Funds Transfer Pricing (FTP)

This subchapter explains how banks use internal transfer prices to move interest rate and liquidity risk to treasury, measure the true profitability of products and business lines, and steer commercial behavior.

Purpose of FTP

Why it matters. A bank's NII is earned jointly: the deposit-gathering branch brings in funds, the lending team lends them out, and treasury manages the mismatch. Without an internal price for money, there is no way to say how much of the margin each of them earned, which products are profitable, or who should bear the loss when rates move. Business units would also be left holding interest rate and liquidity risk they are not equipped to manage.

What FTP is. Funds transfer pricing is an internal pricing mechanism in which treasury acts as the bank's internal market for money:

• Treasury buys funds from liability-generating units (deposits) and credits them with a transfer price.
• Treasury sells funds to asset-generating units (loans) and charges them a transfer price.
• Interest rate and liquidity risk are transferred to treasury, which manages them centrally (hedging, funding).

            Deposit rate 1.0%                               Loan rate 5.0%
  Depositors ----------------> Deposit unit           Lending unit <---------------- Borrowers
                                  |  ^                   |  ^
             funds transferred    |  | FTP credit 3.0%   |  | FTP charge 3.0%
                                  v  |                   v  |
                              +-------------------------------+
                              |           TREASURY            |
                              |  manages rate and liquidity   |
                              |  risk, hedges, external funds |
                              +-------------------------------+

Objectives.

Objective	How FTP achieves it
Profitability measurement	Splits NII into margins by product, customer, branch and business line
Risk transfer	Moves interest rate and liquidity risk from business units to treasury
Pricing guidance	Gives the minimum rate a loan must earn (and the maximum a deposit can pay) to be profitable
Behavioral incentives	Rewards stable funding, charges for long-term or contingent liquidity use
Performance management	Business units are judged on what they control: client margins and volumes

Example. Using the diagram, NII on EUR 100m of loans funded by EUR 100m of deposits is 4.0% (5.0% - 1.0%). With FTP at 3.0%:

Unit	Margin	EUR m per year
Lending	5.0% - 3.0% = 2.0%	2.0
Deposits	3.0% - 1.0% = 2.0%	2.0
Treasury	3.0% - 3.0% = 0.0% (no mismatch here)	0.0
Total	4.0%	4.0

Without FTP, the lending unit might claim the full 4.0% (because "it made the loan") and the deposit unit might appear to make a loss (because it only pays interest).

Analogy. FTP is like an internal wholesale price in a retail chain. Stores buy from the central warehouse at a set price and are judged on what they sell above it; the warehouse is responsible for sourcing and inventory risk.

Key takeaways

• FTP is the internal price of money between business units and treasury.
• It splits NII into commercial margins (loans, deposits) and a treasury margin.
• It transfers interest rate and liquidity risk to treasury, where it can be hedged.
• It guides pricing and steers behavior, so its design shapes the bank's balance sheet.

🧪 Practice

1. Beginner: A loan pays 6.0%, the FTP charge is 4.2%. What is the lending margin? A deposit pays 1.5% with an FTP credit of 3.5%. What is the deposit margin?
2. Intermediate: Explain how a single bank-wide FTP rate for all products could lead the bank to over-produce long-term fixed-rate loans.
3. Interview: "Why should business units not bear interest rate risk on their own books?" Hint: netting and economies of scale in hedging, expertise, comparability of performance, and avoiding business decisions driven by rate bets rather than client franchise.

Matched-Maturity FTP

Why it matters. If every product received the same transfer rate, a 10-year fixed loan and a 3-month floating loan would look equally profitable at the same customer rate, even though the long loan exposes the bank to much more interest rate and liquidity risk. Matched-maturity FTP prices each transaction at the market cost of funding with the same repricing and liquidity characteristics, so business units see the true economic margin.

Methods compared.

Method	How it works	Strengths	Weaknesses
Single pool	One rate for all assets and liabilities	Simple	Ignores maturity, distorts incentives
Multiple pool	One rate per maturity or product pool	Better than single pool	Rates change for existing deals; risk not cleanly transferred
Matched-maturity (deal by deal)	Each deal gets the FTP curve rate for its own repricing and liquidity profile, locked at origination	Accurate, transfers risk cleanly, locks margins	Data and system intensive

How matched-maturity works.

• Interest rate component: based on the repricing tenor. A fixed-rate loan locks the rate for its tenor at origination; a floating loan reset quarterly gets the 3-month rate at each reset.
• Liquidity component: based on the liquidity (contractual or behavioral) tenor, locked at origination even for floating loans.

Example 1: fixed vs floating loan, both 5 years. FTP curve at origination: 3M base 3.00%, 5Y base 2.80%, 5Y liquidity premium 0.40%.

Loan	Customer rate	Rate component	Liquidity component	FTP	Margin
5Y fixed	4.50%	2.80% (locked)	0.40% (locked)	3.20%	1.30% locked
5Y floating, 3M reset	3M + 1.60%	3M base (resets)	0.40% (locked)	3M + 0.40%	1.20% locked spread

If rates later rise and the 5Y base becomes 4.20%, the fixed loan's margin stays at 1.30%; the rate risk sits in treasury, which hedged it at origination.

Example 2: amortizing loan. A EUR 300m loan repays EUR 100m at the end of each of 3 years. FTP curve: 1Y 3.00%, 2Y 3.20%, 3Y 3.40%. Each principal repayment is funded to its own maturity (strip funding), and weighted by amount and time outstanding:

  FTP = (100 x 1 x 3.00% + 100 x 2 x 3.20% + 100 x 3 x 3.40%) / (100 x 1 + 100 x 2 + 100 x 3)
      = (3.00 + 6.40 + 10.20) / 600 x 100
      = 19.60 / 600 x 100  =  3.267%

Using the 3-year rate (3.40%) would overcharge; using a simple average (3.20%) would undercharge, because longer tranches are outstanding longer.

Analogy. Matched-maturity FTP is like a car rental company charging by the exact rental period and car type, instead of one flat daily rate for every booking.

Key takeaways

• Matched-maturity FTP prices each deal at the funding cost matching its repricing and liquidity profile.
• The rate component follows repricing; the liquidity component follows (behavioral) maturity.
• Margins on fixed-rate deals are locked at origination; rate risk moves to treasury.
• Amortizing products need cash-flow-weighted or duration-based FTP rates.

🧪 Practice

1. Beginner: A 3-year fixed deposit is raised when the 3Y base rate is 2.9% and the 3Y liquidity premium is 0.3%. What FTP credit does the deposit receive?
2. Intermediate: Compute the cash-flow-weighted FTP rate for a EUR 400m loan repaying EUR 100m per year over 4 years, using 1Y 3.0%, 2Y 3.2%, 3Y 3.4%, 4Y 3.5%.
3. Interview: "A loan is prepaid after 2 years of a 5-year term. How should FTP handle it?" Hint: treasury hedged and funded to 5 years, so the unwind has a cost; options include a prepayment fee passed through to the customer, an FTP adjustment charged to the business unit at prepayment, or an option cost built into the original FTP rate.

Liquidity Premiums

Why it matters. Funding a 10-year loan is not only an interest rate decision; the bank must also secure funding that stays for 10 years, which costs more than rolling overnight money. If FTP charged only the risk-free swap rate, long-term lending would look artificially cheap, the bank would underprice liquidity risk, and short-term funding would be over-used, a pattern behind several failures in 2007-2008.

Components of liquidity cost in FTP.

Component	What it charges or credits	Basis
Term liquidity premium (TLP)	Cost of term funding above the risk-free curve	Bank's own funding spreads by tenor (senior unsecured, covered bonds, term deposits)
Contingent liquidity cost	Cost of holding a buffer for potential outflows	Negative carry of HQLA needed for undrawn commitments and volatile deposits
Liquidity credit for stable funding	Benefit of liabilities that reduce the buffer need or count toward NSFR	Behavioral maturity and LCR/NSFR treatment of the deposit

Example 1: term liquidity premium from a funding mix. The bank funds 5-year assets with 40% covered bonds at swap + 30bp and 60% senior unsecured bonds at swap + 90bp.

  5Y TLP = 0.40 x 30bp + 0.60 x 90bp = 12 + 54 = 66bp

Example 2: contingent liquidity cost of an undrawn facility. A EUR 100m undrawn committed credit line to a non-financial corporate carries an LCR outflow assumption of 10%, so the bank must hold EUR 10m of HQLA against it. HQLA yields swap - 10bp; the bank's funding costs swap + 30bp, a negative carry of 40bp.

  Annual cost = 10m x 0.40% = EUR 40k  ->  4bp on the EUR 100m commitment

The relationship manager should charge at least 4bp in commitment fees, or the facility destroys value.

Example 3: differentiated deposit credits. Two deposits of EUR 100m each:

Deposit	LCR outflow	HQLA needed	Liquidity value to the bank
Stable retail, insured	5%	5m	High: credited with a large share of the term premium
Non-operational corporate	40%	40m	Low: credit reduced by the buffer negative carry

Analogy. A liquidity premium is like paying more for a long-term lease than for a monthly rental. The landlord gives up flexibility, so the long lease costs more, and the tenant who needs certainty should pay for it.

Key takeaways

• Liquidity premiums charge assets and credit liabilities for the cost and value of term funding.
• The term premium is derived from the bank's own funding spreads by tenor.
• Contingent liquidity costs charge products that require liquidity buffers, such as undrawn commitments.
• Stable deposits deserve higher credits than volatile ones, in line with LCR and NSFR treatment.

🧪 Practice

1. Beginner: Compute the 3Y TLP if the bank funds 50% with covered bonds at +25bp and 50% with senior bonds at +75bp.
2. Intermediate: An undrawn liquidity facility to a non-bank financial institution has a 100% LCR outflow. With a 40bp negative carry, what is the annual contingent liquidity charge on EUR 50m, in EUR and bp?
3. Interview: "During a funding crisis the bank's own spreads double overnight. Should the FTP liquidity curve follow immediately?" Hint: consider marginal vs average funding cost, impact on new vs existing deals, smoothing and ALCO overrides, and the signal sent to the business about scarce liquidity.

FTP Curve Construction

Why it matters. Every transfer price in the bank comes from the FTP curve. An error in the curve flows into thousands of deal prices and margins, and because fixed-rate FTP is locked at origination, an error on one day can affect results for years. Building the curve is a precise, governed, daily process.

Building blocks.

  FTP rate(tenor) = Base rate curve(tenor)
                  + Term liquidity premium(tenor)
                  + Adjustments (contingent liquidity, basis, option costs,
                                 regulatory or strategic add-ons)

Block	Source	Notes
Base curve	Risk-free curve: €STR OIS curve, or Euribor swap curve for Euribor-indexed products	One per currency; consistent with the hedging instruments
Liquidity curve	Own issuance spreads, secondary market spreads, CDS, peer spreads, term deposit pricing	Smoothed, often reviewed by ALCO
Adjustments	Behavioral option costs (prepayment), basis spreads, strategic incentives	Must be transparent and documented

Construction steps.

1. Collect market quotes (deposits, futures, swaps) at a fixed cut-off time.
2. Bootstrap zero rates and discount factors for each base curve.
3. Build the liquidity spread curve from funding data at standard tenors.
4. Interpolate to all tenors (for example, linear on zero rates, or monotone convex to avoid unrealistic forward rates).
5. Apply adjustments; validate (no negative liquidity premiums, smooth forwards, day-on-day change checks).
6. Approve and publish with an effective date and version.

Example: EUR FTP curve.

Tenor	Base (swap)	Liquidity premium	FTP rate
1M	3.00%	0.05%	3.05%
3M	3.05%	0.10%	3.15%
1Y	3.10%	0.25%	3.35%
3Y	2.95%	0.55%	3.50%
5Y	2.90%	0.75%	3.65%
10Y	3.00%	1.00%	4.00%

A 4-year deal, interpolated linearly between 3Y and 5Y: 3.50% + (3.65% - 3.50%) x (4 - 3) / (5 - 3) = 3.575%.

Note that the base curve is slightly inverted between 1Y and 5Y while the liquidity premium rises steadily with tenor, so the FTP curve is still upward sloping.

Engineering notes.

• Curves must be immutable once published; corrections are new versions with explicit effective timestamps, never silent overwrites.
• Each deal stores the curve version used, so its FTP can be reproduced years later.
• Intraday curves may be published for front-office pricing; end-of-day curves are used for booking.

Analogy. The FTP curve is the bank's internal price list for money by tenor. Like a supermarket price list, it must be correct, dated, and the same at every checkout.

Key takeaways

• FTP curve = base risk-free curve + term liquidity premium + documented adjustments.
• The base curve must be consistent with the instruments treasury uses to hedge.
• Construction involves bootstrapping, interpolation, validation and approval, daily per currency.
• Published curves are versioned and immutable; deals reference the version used.

🧪 Practice

1. Beginner: Using the table, interpolate the FTP rate for a 2-year deal.
2. Intermediate: Explain why using linear interpolation on par rates can create unrealistic forward rates, and what alternative you would use.
3. Interview: "Market data for the 7-year swap is missing at cut-off. What should the curve service do?" Hint: fallback hierarchy (secondary source, previous close plus shift of neighboring tenors), flagging of the curve as degraded, approval workflow before publication, and logging for audit.

Allocating Margins to Business Lines

Why it matters. FTP is only useful if the margins it produces add up to the bank's actual NII and are accepted by business units as fair. If the sum of business line margins does not reconcile, or if some income is left in an unexplained "residual", management cannot trust the profitability figures and pricing decisions suffer.

The decomposition.

  Bank NII = Sum of commercial margins (assets)       customer rate - FTP charge
           + Sum of commercial margins (liabilities)  FTP credit - customer rate
           + Treasury margin                          mismatch, liquidity and hedge results
           + Equity / non-interest-bearing credit     return on equity allocated

Example: one-year view (EUR m).

Unit	Volume	Customer rate	FTP	Margin	NII contribution
Mortgages	600	4.00%	3.40% (charge)	0.60%	3.6
Corporate loans	400	5.50%	3.60% (charge)	1.90%	7.6
Retail deposits	700	0.80%	2.44% (credit, replicating portfolio)	1.64%	11.48
Corporate deposits	200	2.00%	2.90% (credit)	0.90%	1.8
Treasury					see below

Treasury receives FTP charges of 600 x 3.40% + 400 x 3.60% = 20.4 + 14.4 = 34.8 and pays FTP credits of 700 x 2.44% + 200 x 2.90% = 17.08 + 5.8 = 22.88. Its external funding of the remaining EUR 100m (assets 1,000 vs deposits 900) costs 100 x 3.20% = 3.2. Treasury margin = 34.8 - 22.88 - 3.2 = 8.72.

Reconciliation: 3.6 + 7.6 + 11.48 + 1.8 + 8.72 = 33.2 = customer interest income (600 x 4.00% + 400 x 5.50% = 24.0 + 22.0 = 46.0) minus interest expense (700 x 0.80% + 200 x 2.00% + 3.2 = 5.6 + 4.0 + 3.2 = 12.8). The margins add up exactly to bank NII.

Common issues.

Issue	Description	Mitigation
Unreconciled residual	Sum of margins differs from GL NII	Daily reconciliation, every balance assigned to a unit
Equity benefit	Who earns the return on interest-free equity?	Explicit equity allocation by capital usage, credited at a defined rate
Treasury margin too large	Hidden subsidies or overcharging	ALCO review; separate mismatch, liquidity and hedge results
Disputes over FTP	Business units contest charges	Transparent methodology, documented curves, change governance
Credit risk confusion	Expected loss mixed into FTP	Keep credit risk out of FTP; handle it in risk-adjusted pricing (RAROC)

Analogy. Margin allocation is like splitting a restaurant bill in a large group: every item must be assigned to someone, the shares must add up to the total, and the rules must be clear before the meal, or people argue.

Key takeaways

• Bank NII decomposes into commercial asset margins, commercial liability margins, a treasury margin and an equity credit.
• The sum of allocated margins must reconcile exactly to reported NII.
• Equity, non-interest-bearing balances and credit risk need explicit, documented treatment.
• Transparency and governance are what make FTP results accepted by business units.

🧪 Practice

1. Beginner: In the example, what is the mortgage unit's margin if the customer rate is cut to 3.80%?
2. Intermediate: In the example, rates rise and the bank's external funding cost increases from 3.20% to 4.20%, while all locked FTP rates stay unchanged. Recompute the treasury margin and explain why only treasury is affected.
3. Interview: "Business units complain that FTP charges make their products look unprofitable. How do you respond as the FTP system owner?" Hint: show the curve construction and deal-level calculation, compare with external market funding costs, explain what the business would pay without treasury, separate methodology questions (ALCO) from data or calculation errors (system).

FTP Systems Design

Why it matters. FTP is computed for every loan and deposit in the bank, often millions of accounts, every day. The results feed profitability reporting, pricing tools, management accounts and sometimes regulatory submissions. A wrong rule, a missing curve or a non-reproducible calculation can distort profitability for years, so the FTP system must be accurate, scalable, reproducible and auditable.

Functional architecture.

  +----------------+  +----------------+  +--------------------+
  | Core banking,  |  | Curve service  |  | Rule repository    |
  | loan, deposit  |  | (FTP curves,   |  | (method by product,|
  | systems        |  |  versioned)    |  |  behavioral params)|
  +-------+--------+  +-------+--------+  +---------+----------+
          |                   |                     |
          v                   v                     v
  +----------------------------------------------------------------+
  | FTP engine                                                     |
  |  - assign method per deal (matched maturity, replicating       |
  |    portfolio, pooled)                                          |
  |  - compute rate and liquidity components                       |
  |  - lock rates for fixed deals; reprice floating at resets      |
  |  - compute daily FTP accruals                                  |
  +-------------------------------+--------------------------------+
                                  |
       +--------------------------+---------------------------+
       v                          v                           v
  Profitability / MIS      GL reconciliation           Pricing API
  (by product, branch,     (sum of margins = NII)      (real-time quotes for
   customer, unit)                                      front office)

Key requirements.

Requirement	Why	How
Reproducibility	FTP must be explainable years later	Store curve version, rule version and inputs per deal at origination
Effective dating	Methodologies change; past deals keep their rules	Bitemporal rule and curve storage (valid time and recorded time)
Rate locking	Fixed-rate deals keep their FTP for life	Persist locked components; never recompute from current curves
Scale	Millions of accounts daily	Batch processing partitioned by product and entity; incremental processing of changed deals
Reconciliation	Credibility of results	Daily automated reconciliation of balances and NII to the GL
Low-latency pricing	Front office needs FTP when quoting	Pricing service using the latest intraday curve, same rules as the batch engine
Auditability	Internal audit, supervisors, disputes	Immutable outputs, change logs, approval workflow for rules and overrides

Example: deal-level FTP record.

Field	Example value
deal_id	MTG-2026-0045812
product_code	MORTGAGE_FIXED_10Y
ftp_method	MATCHED_MATURITY_CASHFLOW
origination_date	2026-10-19
curve_version	EUR_FTP_2026-10-19_v2
rule_version	FTP_RULES_2026Q4_v1
rate_component	2.95% (locked)
liquidity_component	0.82% (locked)
option_cost	0.12% (prepayment, locked)
ftp_rate_total	3.89%
customer_rate	4.60%
commercial_margin	0.71%

Example: simplified FTP assignment logic.

  for each deal in deals_changed_or_new(as_of_date):
      rule  = rules.lookup(deal.product_code, valid_at = deal.origination_date)
      curve = curves.get(deal.currency, version = rule.curve_selection(deal))
      if deal.is_new:
          rate_comp = curve.base_rate(tenor = repricing_tenor(deal), method = rule.method)
          liq_comp  = curve.liquidity_premium(tenor = liquidity_tenor(deal, rule))
          lock(deal, rate_comp, liq_comp, curve.version, rule.version)
      elif deal.is_floating and deal.has_reset(as_of_date):
          update_rate_component(deal, curve.base_rate(tenor = deal.reset_tenor))
      accrue_daily_ftp(deal, as_of_date)

Analogy. An FTP system is like a payroll system: it runs every period for everyone, must apply the rules valid at the time, must reconcile to the bank account to the cent, and must be able to explain any individual payslip years later.

Key takeaways

• The FTP system combines deal data, versioned curves and versioned rules in a scalable engine.
• Fixed-rate FTP components are locked at origination and stored with their curve and rule versions.
• Effective-dated (bitemporal) storage, reconciliation to the GL and audit trails are core requirements.
• A real-time pricing API must use the same rules as the batch engine to avoid inconsistent quotes.

🧪 Practice

1. Beginner: Name four fields that must be stored per deal to reproduce its FTP rate in five years' time.
2. Intermediate: ALCO changes the liquidity premium methodology effective from next month. Describe how the system should treat existing fixed-rate deals, existing floating-rate deals and new deals.
3. Interview: "Design an FTP engine that processes 20 million accounts nightly and also serves real-time quotes." Hint: shared rule and curve libraries for batch and online paths, partitioned batch processing with incremental change detection, idempotent reruns, caching of curves in the pricing service, and consistency tests comparing online quotes with batch results.

---

<a id="7-funding-and-capital-management"></a>