The Accounting Choice That Changes What the Loss Does
This chapter focuses on the accounting categories that shape how a U.S. bank experiences an underwater bond: available-for-sale (AFS) and held-to-maturity (HTM). The question is not whether accounting can make a market loss vanish—it cannot—but why the same decline in a bond’s market value can have different immediate effects on earnings, equity, regulatory capital, and future flexibility. That distinction helps explain why a bank might accept a difficult economic position in order to avoid a more immediate and visible accounting event.
AFS and HTM put the same bond into different decision environments
A bank’s carrying value is not always the same as the price at which its security could be sold. Amortized cost starts from the purchase price and adjusts it over time for the gradual accretion or amortization of a premium or discount. Market value is the current trading price. When rates rise, a fixed-rate bond can have a market value below its amortized cost even though the bank still expects to receive its contractual principal and interest.
The AFS and HTM labels determine how that gap enters the bank’s accounts. HTM securities are carried at amortized cost. Rate-driven unrealized gains and losses therefore do not affect net income, assets, equity, or regulatory capital in the same way they would if the bond were continuously exposed to market-value changes in those measures. The market still assigns a price to the bond, and the bond still has duration and extension risk, but the accounting presentation is less directly sensitive to the daily or quarterly market price.
AFS securities are recorded at market value. Unrealized gains and losses do not ordinarily run through net income immediately; instead, they accumulate in a component of equity called accumulated other comprehensive income, or AOCI. That treatment is different from HTM, but it still leaves the market-value change visible in the balance sheet. For many banks, an AOCI filter excludes unrealized AFS gains and losses from regulatory capital. The largest banks are not permitted to use the filter, and some other banks have elected not to use it. As a result, the same underwater AFS position can have different regulatory-capital implications depending on the bank’s status.
The difference becomes sharper when the bank sells. Selling an underwater AFS security transfers the loss from an unrealized market-value position into a realized result for the quarter. It reduces net income, and for banks using the AOCI filter it also reduces regulatory capital. The economic loss was already present before the sale, but the sale changes when and where that loss appears. It can therefore affect reported performance, capital ratios, investor interpretation, and management room at the same moment that it provides cash and shortens duration.
HTM avoids that particular form of immediate market-value volatility, but it is not a free alternative. A bank must have the positive intent and ability to hold the security until maturity. Selling or reclassifying HTM securities can have significant accounting consequences and can “taint” the wider HTM portfolio, meaning the bank may no longer be able to use the classification in the same way. By contrast, AFS securities can generally be reclassified into HTM more freely. The choice is therefore not simply between showing a loss and hiding it. It is a choice between different combinations of present visibility and future freedom.
A simple illustration separates the accounting event from the economic event. Imagine a bank bought a bond for $100 and the bond’s market value later fell to $90 as rates rose. The bank has an economic exposure whose current market value is $90 whether the bond is labeled AFS or HTM. Under AFS treatment, the $10 difference is visible through market value and AOCI, subject to the bank’s capital treatment. If the bank sells, the $10 loss becomes realized and the bank receives $90 of cash. Under HTM treatment, the carrying value may remain based on amortized cost while the bank holds the bond, but the market still values the position at about $90 and the bank still owns the same rate-sensitive cash flows. Reclassification changes the accounting path; it does not turn $90 of market value back into $100.
That distinction also explains why “unrealized” should not be read as “imaginary,” and “realized” should not be read as “newly created.” The loss already reflects the price at which another investor would value the bond. Selling determines when the bank must recognize that price in the relevant measures and gives it cash in exchange. Holding determines that the bank continues to bear the market risk and may eventually receive the bond’s contractual cash flows. The bank’s decision is about timing, flexibility, and who bears the volatility—not about whether the market has noticed the rate shock.
The state changes in the diagram are accounting states, not changes in the bond’s underlying economics. A rate increase can move an AFS bond into an underwater position without changing its contractual claim. A sale can create cash and realize the loss. A reclassification can move the bond into HTM accounting and reduce selected reported volatility, but it does not restore the bond’s market value or shorten its effective duration. If the bond is a callable MBS, its extension risk remains.
Lower visible volatility can mean less room to manage the asset later
The accounting choice matters because it interacts with the other adjustment margins. A bank that sells a risky AFS bond may reduce duration, raise liquidity, and remove a difficult security from its portfolio, but it pays the immediate cost of recognizing the loss. A bank that moves the bond to HTM may reduce the sensitivity of book equity or, for banks without the AOCI filter, regulatory capital, but it also places the bond in a category that is harder to sell and cannot use qualified hedge accounting for interest-rate risk.
This is the central interface between accounting and risk management. The accounting classification affects what a sale costs today; it also affects which actions remain available tomorrow. A bond held in HTM can continue to lose market value as rates move, yet the bank has less freedom to treat it as an ordinary trading or liquidity position. If the bank later needs cash, selling can create additional accounting consequences. If it wants to hedge the interest-rate risk with a qualified hedge, HTM treatment prevents that accounting designation. A choice that stabilizes a reported ratio can narrow the menu of future responses.
The costs are distributed across different audiences and time horizons. Investors may see a quieter book-equity measure while the bank’s fair-value position remains exposed. Regulators may see less capital volatility for a bank using the relevant filter, while managers face a smaller set of securities that can be sold or hedged cleanly. Future liquidity managers inherit the consequences of today’s classification decision: the bond still has market value, but turning it into cash can be more consequential than it would have been under AFS. The accounting label therefore acts as an interface between the bond market and the bank’s internal capacity, not as a detached reporting convention.
This interface also limits the meaning of a capital ratio. A ratio can be stable because the bank has reduced its economic exposure, because it has hedged the exposure, or because a larger share of the exposure sits in a category whose unrealized price changes are not flowing through that ratio. Those explanations have different implications for resilience. The first two can represent risk reduction, although they carry their own costs. The third may represent a change in visibility. Distinguishing them is essential when a portfolio contains callable bonds whose market value and duration can worsen together.
The Philadelphia Fed evidence suggests that banks used this menu selectively. In the 2022–23 period, individual securities with higher interest-rate risk were more likely to be classified as HTM. The relationship was stronger after rates began to rise: within-bank estimates imply that an additional year of duration increased the probability of HTM classification by about 3.6 percentage points in 2022–23, compared with about 1.5 percentage points in 2020–21. Agency MBS were about 21.5 percentage points more likely than Treasuries to be classified as HTM after conditioning on risk and bank-time effects.
The timing is informative. Banks were already more likely to place negatively convex bonds into HTM during 2020–21, before the rate surge. That pattern is consistent with awareness that these bonds would be difficult to manage if rates rose. The evidence does not show the intention behind every individual classification, and a classification relationship is not by itself proof of a deliberate strategy. But it is difficult to reconcile the pattern with a claim that banks discovered the risks only after the market shock. The securities that later became especially problematic were already being treated differently.
The effect on visible risk is the most important result. The study separates gross duration, duration after qualified hedging, and the portion of duration held in AFS portfolios. Gross and net economic duration rose during 2020–23. Yet the share of that duration remaining in AFS—and therefore more directly affecting book equity and, for some banks, regulatory capital—fell to a historically low level by late 2023. In other words, the risk did not move in one direction. The economic exposure increased while the portion most visible in selected accounting measures decreased.
That divergence is easy to misread. A lower AFS share can look like a safer portfolio if the reader is looking only at the market-sensitive slice of book equity. But it can also mean that more risk has been placed in a category with less immediate reporting volatility and less future flexibility. The correct question is not “Did HTM reduce the number?” It is “Which number did it reduce, and which exposure remained?” HTM can reduce a measure of accounting sensitivity without reducing the bond’s sensitivity to market rates, its value if sold, or its contribution to a bank’s fair-value economic position.
The distinction also clarifies why the AOCI filter matters. Banks that exclude AOCI changes from regulatory capital do not see unrealized AFS losses flow into capital in the same direct way. For those banks, the immediate regulatory penalty from keeping an underwater AFS bond is weaker, but realizing the loss through a sale still has consequences. Banks that include AOCI in capital have a stronger reason to limit the capital volatility generated by market values, and the paper finds that the incentive to avoid selling underwater securities is amplified when the filter is in place. The direction of the incentive is not universal across every accounting measure, but the broader point is stable: the capital regime changes the cost of making the loss explicit.
The accounting category also changes the meaning of “waiting.” Waiting with an AFS bond leaves its market value visible and preserves greater selling and hedging flexibility, although it does not prevent further losses. Waiting with an HTM bond suppresses some accounting volatility but makes a later sale more consequential and a qualified hedge unavailable. Neither option is simply passive. Each preserves one form of capacity while sacrificing another.
That is why HTM classification should be understood as a risk-visibility response rather than a complete risk-reduction response. It may be rational for a bank that values stable reported capital and believes it can hold the bonds through maturity. It may be less attractive for a bank that expects deposit outflows, needs to preserve liquidity, or anticipates further rate uncertainty. The choice depends on the bank’s funding structure, capital regime, and ability to use other hedges—not only on whether the bond is economically sound.
The accounting evidence therefore sharpens the original question. Banks did not face a choice between selling a liquid bond and doing nothing. They faced a sequence of state changes. Selling could turn an existing market loss into a current capital and earnings event. Reclassification could reduce selected volatility while narrowing future options. Keeping the bond in AFS preserved more flexibility but left the loss visible. The next issue is behavioral: when banks had AFS securities at different prices relative to book value, did they actually sell the bonds whose market prices made the loss visible, or did the observed portfolio inertia reflect something else?