When Rates Rise, Some Bonds Get Longer
This chapter focuses on the mechanism that made U.S. bank securities portfolios unusually difficult to manage in 2022–23: why a rise in interest rates can make a callable mortgage bond lose value and become more sensitive to further rate increases at the same time. That mechanism matters for judgment because it links three events that are often treated separately—market-value losses, deposit pressure, and the bank’s need to rebalance its assets. Once those links are clear, it becomes easier to distinguish an immediate cash problem from a problem of economic exposure and from a problem of accounting visibility.
Duration is only the first approximation
A fixed-rate bond promises cash flows that do not change simply because market yields move. When yields rise, the present value of those cash flows falls. Duration is a compact way to describe that sensitivity: roughly, it estimates the percentage change in a bond’s value for a 100-basis-point change in yield. A longer-duration bond loses more value for the same rate move than a shorter-duration bond, all else equal.
But duration is not a permanent property stamped onto a security. It depends on when investors expect to receive the bond’s cash flows, and that timing can change. Convexity describes the curvature of the price–yield relationship—the way duration itself changes as yields move. For a conventional fixed-coupon bond, convexity is generally positive. As yields rise, duration tends to decline, so the bond’s price becomes somewhat less sensitive to additional increases. That does not prevent a loss, but it moderates the next increment of loss.
Callable bonds reverse part of that intuition. A callable bond gives the issuer an option to repay the bond before its contractual maturity. Agency mortgage-backed securities have a related feature: the underlying borrowers can prepay their mortgages, often when refinancing becomes attractive. In ordinary conditions, that prepayment option can shorten the life of the mortgage pool. For the investor, receiving principal earlier can be valuable because it limits the time spent holding a low-coupon asset when market opportunities change.
When rates rise, refinancing becomes less attractive. Borrowers are less likely to replace an older, low-rate mortgage with a new, more expensive one, so prepayments slow. The expected cash flows of the MBS move farther into the future. The security has effectively extended. Its duration rises precisely as yields are rising, and its convexity can be negative. The price decline from the rate move is therefore not the whole problem: the bond also becomes more exposed to whatever rate movement comes next.
This is why the word “callable” needs care. A corporate or municipal bond may be called by its issuer, while an MBS is typically prepaid by the underlying mortgage borrowers. The economic feature is similar for this discussion: the timing of the investor’s cash flows depends on an option whose exercise changes when rates change. In the Philadelphia Fed study, agency MBS were the dominant example, and callable bonds of all types represented about 60% of the sample’s securities holdings. Agency MBS made up more than half of the portfolio.
The result is a nonlinear portfolio shock. Suppose two banks begin with securities whose initial duration looks similar. One owns mostly conventional Treasury securities; the other owns a large share of fixed-rate agency MBS. After rates rise, the Treasury portfolio may lose value but become somewhat shorter in duration. The MBS portfolio may lose value and extend. The two banks have not merely experienced different-sized losses. They now face different future sensitivities and different incentives to rebalance.
The asymmetry can be stated in terms of cash-flow timing. A conventional bond’s maturity schedule is relatively stable, so a manager can estimate how a rate move changes its price and duration. An MBS manager is estimating a distribution of prepayments. If rates fall, refinancing may accelerate and principal may return sooner; if rates rise, the same option may remain unexercised and principal may arrive later. The bank is therefore managing a security whose risk changes with the very market variable it is trying to hedge. Negative convexity is not simply a more technical label for a larger loss. It means the direction of the rate move changes the future sensitivity of the position.
That distinction matters even without a credit default. A high-quality agency MBS can lose value because its expected cash flows have been repriced, not because borrowers have suddenly become unable to pay. The bank may still receive the promised principal and interest over time, but the market value and interest-rate exposure of those cash flows have changed. A strategy that waits for contractual maturity can therefore avoid a realized market loss while accepting a longer period of exposure and less flexibility if funding needs arrive first.
The 2022–23 evidence follows that logic. The Federal Reserve raised the federal funds rate from near zero in March 2022 to 5.25–5.5% by July 2023, while longer-term yields also rose sharply. By mid-2022, nearly all securities in the Philadelphia Fed sample were trading below amortized cost, and about one-third of holdings by value had unrealized losses greater than 10%. The callable and noncallable parts of the portfolio then moved in opposite directions: noncallable duration fell, while callable duration rose sharply in 2022 and stayed high through the end of 2023.
That divergence also explains why what a bank held before the shock mattered so much. The paper finds that the increase in portfolio duration was closely tied to the share of callable bonds held in 2019, before rates began their rapid climb. A bank did not need to buy a new risky security in 2022 to become more exposed. Its existing MBS could extend as borrower behavior changed. The portfolio’s starting composition acted as a built-in sensitivity to the path of rates.
The diagram’s sequence is important. The bank does not first choose to take on a longer bond and then observe the result. For callable assets, the market environment changes the expected cash-flow timing of what it already owns. That creates a moving target for risk management: a hedge sized for yesterday’s duration may no longer offset tomorrow’s exposure, and a sale that would shorten duration may also crystallize yesterday’s price decline.
Why deposit pressure raises the stakes without making every sale attractive
The asset-side shock arrived alongside weaker deposit growth. In the Philadelphia Fed sample and the broader banking data discussed in the paper, deposit growth slowed and the median bank’s year-over-year deposit growth was negative by mid-2023. Deposits are not a static source of funding during a rate-hiking cycle. Their pricing can change, and some balances can leave. A bank facing outflows must decide how much to raise through more expensive wholesale funding, how much liquidity to preserve, and which assets can be adjusted.
A securities portfolio appears attractive for this purpose because bonds can be sold more readily than many other bank assets. But “can be sold” does not answer “should be sold now.” A bank that sells an underwater MBS receives cash and may shorten its asset duration, yet it also recognizes the difference between the security’s market price and its carrying value. If the bank does not sell, it preserves the accounting position for the moment, but the MBS remains exposed to extension risk and the bank may have to find liquidity elsewhere. The rise in the value of liquidity therefore increases the benefit of a sale while the rise in unrealized losses increases its immediate cost.
This creates a timing problem, not just a valuation problem. A deposit outflow is an immediate funding event. Extension risk is a sensitivity to future rate and cash-flow paths. A realized loss is an immediate accounting and capital event. The bank may be forced to choose between paying more for funding now, selling an asset at a loss now, or retaining an asset whose market value and duration remain unfavorable. None of those choices makes the underlying trade-off disappear; they distribute its cost across time and across different measures of bank condition.
This is the central interaction between funding and market risk. Deposit outflows make a liquid asset more useful as a source of cash. Rising rates make the same asset more dangerous to sell at its current market value. A portfolio can therefore be liquid in a market sense while being costly to mobilize for the bank that owns it. That is not a contradiction in the bond market; it is a consequence of applying a market price to an institution with capital and reporting constraints.
The broader balance sheet matters as well. Economic value of equity, or EVE, is a fair-value measure of assets minus liabilities. It asks how the value of the bank’s net position changes under a rate scenario, including the expected cash flows of existing positions. The Philadelphia Fed paper reports that banks became materially more exposed to an upward rate shock in 2022–23. In the OCC’s supervisory framework, EVE is a long-term measure, while net interest income, or NII, focuses on projected interest income minus interest expense over a shorter horizon. Neither is the same as a cash loss caused by a bond sale.
That distinction prevents three common errors. First, a fall in a bond’s market value is not automatically a realized loss: the loss becomes realized when the institution sells or otherwise recognizes it under the applicable accounting treatment. Second, a modeled EVE decline is not a prediction that the bank will pay out that percentage in cash; it is a sensitivity of modeled economic value to a specified rate shock. Third, stable short-term NII does not prove that the bank has low long-term duration risk. The two measures examine different horizons and cash-flow assumptions.
The OCC’s figures also illustrate why a stress percentage needs a denominator and a time horizon. Its tables report changes in EVE and NII under parallel shocks, not a single measure called “interest-rate risk.” The EVE calculation captures modeled values of existing assets, liabilities, and off-balance-sheet items; the NII calculation projects income and expense. The result depends on assumptions about repricing, deposit behavior, and the timing of cash flows. Those assumptions are necessary for the exercise, but they mean that a reported EVE sensitivity should be read as a scenario-dependent exposure rather than as an observed loss waiting to happen.
The sources also caution against assuming that deposits automatically hedge callable assets. Deposit betas can rise as rates increase, shortening the effective duration of liabilities, while MBS duration can lengthen on the asset side. The combination can create a larger asset–liability mismatch unless the bank actively adjusts. The Philadelphia Fed analysis finds that banks with larger MBS holdings experienced larger declines in mark-to-market equity during 2022–23, and adding deposit-beta or deposit-rate controls did little to remove that relationship. This does not mean no bank had a natural hedge. It means the evidence does not support a broad, sufficient offset for the MBS-related exposure in the studied sample.
The practical implication is that a rate surge changes both the value of an asset and the urgency of managing it. Yet the urgency does not determine which action is feasible. The Philadelphia Fed study finds that banks sharply reduced the volume and duration of new purchases, but did not respond with a comparable increase in outright sales. That behavior is consistent with a bank recognizing that it can stop adding exposure more easily than it can repair an existing position. It can also be consistent with the fact that selling, hedging, and classification carry different institutional costs.
Those costs are not yet an accounting detail added after the market mechanism. They are part of the mechanism’s consequence. A callable MBS that extends in duration creates a larger economic exposure. A deposit outflow makes cash more valuable. The price at which the MBS can be sold makes the loss visible. EVE may deteriorate even if the bank’s selected book-value measures appear calmer. To understand why the bank may choose one response over another, the next question is how its accounting category determines what the loss does immediately—and what flexibility remains afterward.