The Risk Did Not Disappear—It Became Harder to See

This chapter brings the bank portfolio back to the whole balance sheet. The specific question is whether the choices observed in 2022–23—buying less, selling few underwater bonds, hedging selectively, and using HTM—actually reduced economic exposure or mainly reduced the portion visible in selected accounting measures. The distinction matters because a reported improvement can mean a safer portfolio, a better hedge, a different accounting category, or simply a different denominator. The evidence supports those interpretations only when the measure and its population are kept explicit.

Three views of the same portfolio

The Philadelphia Fed study separates three related quantities. Gross duration measures the interest-rate sensitivity of the securities before qualified hedging. Net duration measures the remaining securities duration after the qualified hedges that the bank has in place. Visible duration is the portion of that exposure held in AFS portfolios, where market-value changes more directly affect book equity and, for some banks, regulatory capital.

These measures answer different questions. Gross duration asks how the securities themselves respond to rates. Net duration asks how much of that response remains after the bank uses qualified derivatives. Visible duration asks how much of the remaining response is exposed to the accounting channels that show market-value changes most directly. A fall in visible duration can therefore occur because the bank sold or hedged risk, or because the bank moved risk from AFS to HTM. Those events have very different implications for economic resilience.

The paper’s pooled results show the divergence clearly. Gross and net securities duration rose significantly from 2020 to 2022. But as more securities moved into HTM, the portion of net duration held in AFS fell to a historically low level by late 2023. The market-sensitive slice of the portfolio looked calmer even as the broader economic exposure was higher. HTM classification had changed the reporting surface, not the bond’s response to a rate move.

The distinction is especially important for callable assets. An agency MBS in HTM still extends when prepayments slow and still loses market value when yields rise. Its duration does not become shorter because the bank records it at amortized cost. Nor does the bank receive a qualified hedge-accounting benefit for interest-rate risk on an HTM security. The category can reduce selected capital volatility while leaving the underlying asset exposed and the future adjustment path narrower.

Whole-bank fair-value evidence points in the same direction. In the Philadelphia Fed sample of publicly traded banks, the mark-to-market equity ratio fell by 3.1 percentage points in 2022–23 relative to the prior period, from a sample average of about 10%. The decline was larger for banks with a high share of securities and especially for banks with a high share of MBS relative to assets. In the key specifications, non-MBS securities did not show the same consistent relationship; MBS holdings did.

This result does not mean that every MBS-heavy bank lacked a natural hedge.