Deposit Betas and the Ten-Year Bond Temptation
Treasury committee of a bank with USD 5 billion of assets. Funding is 60% demand deposits and 40% term deposits; historical deposit betas are 0.25 for demand and 0.75 for term. On the asset side 70% of loans are prime-linked floating, repricing monthly; 30% are fixed for three years. The policy rate has risen 300 basis points in twelve months and is expected to plateau. LCR is 118%, NSFR 104%, and the bond portfolio shows an unrealised loss of USD 60 million. The treasurer proposes buying USD 400 million of 10-year government bonds to "lock in yields before the cuts come". Two board members want higher near-term NII; the CRO is worried.
1. Estimate the effect of the 300bp rise on net interest income using the stated betas, and state the key assumption risk.
Read the worked answer
A simple repricing-gap approximation:
Asset yield: 70% of loans reprice with the policy rate → 0.70 × 300bp ≈ +210bp on earning assets.
Liability cost:
- Demand: 0.60 × 0.25 × 300bp = 45bp
- Term: 0.40 × 0.75 × 300bp = 90bp
- Total ≈ 135bp
Net: ≈ +75bp of earning assets, i.e. on USD 5 billion roughly USD 375 million of additional annual NII before volume and mix effects.
The bank is strongly asset-sensitive: NII rises when rates rise and falls when rates fall, roughly symmetrically.
The key assumption risk is the beta itself. Betas are estimated from past cycles and are not stable. If competition for deposits pushes the demand beta from 0.25 to 0.45, liability cost rises by an extra 36bp and the NII gain shrinks to about 39bp. Treat the beta as a range, not a point estimate, and stress it.
2. Is buying USD 400 million of 10-year government bonds consistent with this balance sheet?
Read the worked answer
It is a directional bet, not a hedge. The balance sheet has short-duration, rate-sensitive liabilities and already earns more when rates rise. Adding long fixed-rate assets:
- Increases duration on the asset side, widening the duration mismatch rather than closing it.
- Creates large fair-value and OCI volatility: a 10-year bond has duration of roughly 8; a 100bp yield rise produces about an 8% loss, or about USD 32 million on USD 400 million. The bank already carries USD 60 million of unrealised losses, so its capital buffer for further mark-to-market damage is thin.
- Improves LCR only marginally: the LCR is already 118% and the constraint is not HQLA quantity but the funding profile.
The treasurer's "lock in yields before the cuts" thesis only pays if the cuts arrive and if the position is held to maturity without accounting or liquidity pressure forcing a sale. That is a macro forecast, not balance-sheet management.
3. What would you recommend to the committee, with conditions?
Read the worked answer
Recommend against a single USD 400 million 10-year purchase. Instead:
- Size it to capital, not to conviction. Cap the duration extension at a level where a 100bp adverse move is absorbable within the OCI/regulatory capital buffer, and classify according to the actual hold intention and business model.
- Barbell rather than bullet. Shorter-dated HQLA plus a smaller long-dated sleeve gives some yield pick-up with less duration.
- Consider swaps. If the objective is to fix funding cost or hedge NII, an interest rate swap can change the repricing profile without consuming HQLA or creating a large unrealised-loss position.
- Stress the deposit beta and a rate-cut scenario before approving, because the current NII benefit reverses when rates fall.
- Separate the two objectives. Board members wanting higher near-term NII and the treasurer wanting to position for cuts are different trades; approving one on the argument of the other is how duration risk gets under-priced.
Conditions if any purchase proceeds: duration limit, monthly mark-to-market reporting to ALCO, and a pre-agreed trigger for review.