What Banks Were Protecting—and What They Left Exposed

This final synthesis answers the series’ central question about U.S. bank securities portfolios in the 2022–23 rate surge: why did banks not simply sell liquid bonds when rates rose? The evidence points to a constrained choice, not a single blind spot. Banks were trying to manage liquidity, reported capital, earnings volatility, and economic interest-rate exposure at the same time, but the available actions protected those objectives unevenly. The durable judgment is to ask which objective a bank is protecting, which adjustment margin is actually usable, and whether a lower reported exposure represents a lower economic exposure or only a less visible one.

The direct answer: liquidity did not equal usable adjustment capacity

Banks did not respond to the rate surge by doing nothing. They reduced the volume and duration of new bond purchases, which limited the amount of new risk added to the portfolio. They also classified more securities as HTM, especially higher-duration and more difficult-to-hedge bonds. What they generally did not do was sell large quantities of existing underwater bonds or materially expand qualified interest-rate hedges. The pattern is selective adjustment: the bank could stop worsening its future position more easily than it could repair the position already on its balance sheet.

The phrase “just sell the bonds” misses the feature that made selling different from the other choices. A sale of an underwater AFS bond would have done several economically useful things at once. It could raise cash, shorten asset duration, reduce exposure to further rate increases, and remove a callable position whose risk might extend as rates rose. But the same transaction would also convert a market-value loss into a realized loss. It could reduce quarterly net income and, for banks using the AOCI filter, reduce regulatory capital. A salient sale could also draw attention to losses that had previously been unrealized.

That combination created a powerful institutional friction. The market had already repriced the bond, so holding it did not restore the lost value. Yet selling forced the bank to recognize the price in the measures that governed capital, earnings, and external confidence. The bond was liquid for a buyer who valued it at its market price; it was costly for the bank whose accounting value was higher. The relevant constraint was therefore not the existence of a market. It was the cost of using that market from the bank’s starting accounting and capital position.

Callable MBS made the choice harder still. When rates rose, slower refinancing extended the expected life of fixed-rate mortgage pools. Duration increased and convexity could become negative. The bank was not only holding a bond that had fallen in price; it was holding an asset whose sensitivity to further rate changes had increased. Deposit outflows made cash more valuable at the same time. Selling therefore had a higher liquidity benefit and a higher immediate recognition cost. A bank that did not sell preserved the accounting position for the moment but retained the extension risk and might need to obtain liquidity elsewhere.

The four adjustment margins can be compared directly:

The table is not a ranking. It shows why no single margin could serve every objective. Buying less is a forward-looking response; it cannot quickly fix an existing duration mismatch. Selling is the most direct way to reduce the position, but it makes the loss explicit. A qualified hedge can preserve the asset and offset some exposure, but only if the bank can establish and maintain the relationship, and callable bonds are the most difficult assets to match. HTM can reduce selected reporting volatility but does not shorten the bond’s duration or make its market value recover.

The Philadelphia Fed evidence makes the sale friction observable. In the studied AFS portfolios, the raw quarterly sale propensity was about 1% for underwater securities and about 9% for above-water securities in 2022–23.