Steven Grey is CEO and chief investment officer of Grey Value Management, an investment firm.
Many investors see banks as the investment equivalent of the famous white, windowless van with “Free Candy” spray-painted on the side. Sure, maybe there’s nothing to worry about, but you’re still reflexively wary of what might be lurking inside.
Take banks’ unrealised losses on investment securities.
As the chart below shows, during the many years that interest rates hovered near zero, banks’ gains and losses on their investment securities were relatively modest. But when the Fed began raising rates in 2022, bond prices plummeted. That generated huge losses for banks that had accumulated them.
The bond market has stabilised since then, but US banks still entered 2025 with unrealised losses on investment securities of about $482.4bn according to the FDIC.
Banks are allowed to keep these losses off their books — holding them at cost rather than fair value — by classifying their bondholdings as “held-to-maturity” (HTM) rather than “available-for-sale” (AFS). But the accounting rules stipulate that they can only classify bonds as HTM if they have both the intent and the ability to hold on to them.
As Silicon Valley Bank’s 2023 implosion demonstrated, it’s the latter that ultimately matters. Generally, if a bank sells a portion of its HTM portfolio, the entire portfolio should be reclassified as “available-for-sale” (AFS) and marked to market. Which means that once a bank saddled with substantial unrealised losses is forced to sell HTM securities, failure is almost unavoidable, even if you appear adequately capitalised.
For example, Silicon Valley Bank’s common equity tier one capital ratio was 12 per cent at the end of 2022 — that was 200 basis points higher than its peer group at the time. It didn’t matter.
To be clear, SVB managed its interest-rate exposure poorly, as Alphaville wrote at the time. By deliberately removing hedges as rates were rising, management amplified its financial risk. It also failed to manage the risks of its liabilities, which in the case of deposits proved much more unstable than anticipated.
While banks love big depositors — they often make lucrative clients — these customers have every reason to run for the door at the first hint of a problem, given how most of their money isn’t covered by FDIC insurance. And as SVB illustrated, it doesn’t take many of them heading for the exit to quickly trigger an existential crisis.
The government has many tools to help arrest bank runs and prevent them from becoming systemic. Because of reforms implemented after 2008, the Federal Reserve can no longer take on some of a bank’s financial risk to facilitate a takeover. But the Fed can still invoke a “systemic risk exception” to raise the FDIC deposit insurance threshold. Which is what it did on March 13, 2023, when the FDIC announced that it would protect all of SVB’s depositors.
Because no two bank crises are alike, interventions are necessarily improvisational. But once panic sets in — contrary to the popular perception — liquidity alone seldom saves the day. As a recent paper from the Yale Program on Financial Stability Survey concluded after studying 22 cases of ad hoc emergency liquidity provision:
Despite AHEL assistance, which the authorities in several cases sized to meet all potential funding outflows from the troubled firms, in no cases did liquidity provision alone prove a “cure” to the run on the institution.
If almost all bank liquidity crises relate to an underlying solvency problem — and the evidence indicates that no amount of liquidity alone can resolve it — then authorities should seize every opportunity to prevent them from arising in the first place. A big part of that is regulatory oversight, but this can only accomplish so much.
Unfortunately, this remains a big issue, as the recent jitters in US regional bank stocks demonstrates. While that was mostly caused by fears over the quality of corporate loan books, as mentioned earlier, US bank balance sheets began 2025 with almost half a trillion dollars in unrealised losses on investment securities, and this could balloon even higher.
As the chart below shows, US banks dramatically increased the average maturity of their securities portfolios in a hunt for yield during the pandemic. This was arguably the more conservative choice in the binary option between reaching for duration or credit risk, but it set them up for a fall when interest rates began to rise again.
Importantly, while the chart indicates that holdings of longer-duration securities have fallen back towards pre-pandemic levels, that decline is probably due to lower valuations, rather than a material reduction in duration risk. As a result, if interest rates begin to rise again — perhaps if inflation refuses to return to below 2 per cent — then losses could rocket once more.
Given this alarming but plausible scenario, here’s the question: What if we could not only help pre-empt the next large-scale bank crisis in the US, but ensure that taxpayers financially profit from doing so?
The way to accomplish this is via what you might call a “bank bail-under.” It’s essentially a variation on a manoeuvre common in a small corner of the corporate debt market known as “busted converts”.
Briefly, companies looking to raise capital at a lower borrowing cost and delay shareholder dilution will often do so by issuing convertible bonds. But if the stock subsequently plummets and makes a conversion vanishingly unlikely, the bond’s price will drop. It is now a busted convert.
If the company has sufficient cash reserves and a savvy CFO, it can offer to repurchase the bonds at a premium to their market value but at a discount to their face value. Imagine an investor who bought a $100 busted convert for $60 being offered $80 by the company that issued it. The investor immediately realises a 33 per cent gain and the company in effect repays its own debt at a 20 per cent discount.
Now imagine that the US Treasury approaches a bank that has substantial holdings of long-duration bonds issued by the US government. These bonds are trading at a steep discount — for example, the 30-year Treasury issued in August 2021 is trading at just 60 cents on the dollar — to the par value that the bank will receive when they mature. The bank balance sheet doesn’t reflect that loss because the bond is parked in their HTM portfolio. But management of course knows what the current price is.
Against that backdrop, the Treasury could offer banks to buy back US government bonds at, say, 90 cents on the dollar, almost wiping out the entirety of their losses and trimming the size of America’s debts at the same time.
Yes, for the government it would often mean issuing expensive debt to retire cheap debt. After all, the 30-year Treasury issued in the summer of 2021 carries a coupon of just 2 per cent, while the most recently issued 30-year UST costs the government 4.75 per cent a year.
But the headline debt reduction should balance this out, and there are other potential upsides. As discussed in a recent paper published by the Federal Reserve Bank of Kansas City, even unrealised losses on bond portfolios can reduce economy-wide lending through four channels:
First, unrealized losses can increase equity costs as investors’ perceptions of financial health deteriorate. Second, deterioration of financial strength combined with increased liquidity needs can increase debt funding costs. These increased equity and debt funding costs are likely to be passed on to borrowers as higher interest rates, potentially reducing loan demand. Third, unrealized losses can also make banks more reluctant to sell securities, creating liquidity demand that could limit future loan supply. Lower loan demand due to higher prices and lower loan supply due to higher funding costs could reduce total loan growth. Fourth, unrealized losses can dampen merger and acquisition (M&A) activity because potential buyers may be reluctant to purchase a bank holding securities in a deep loss position. Reduced M&A activity can result in a less effective banking system to the extent that it allows inefficiently run banks to continue operating. In this way, a slowdown in M&A activity can result in poorly allocated, or reduced, aggregate lending.
How many loans are now not being made because too many banks bought too many “risk-free” US government bonds when interest rates were pushed so low that buying them was almost guaranteed to eventually backfire? It’s impossible to say, but certainly enough to affect the economy.
Selling HTM securities even just a little below their book value would naturally trigger losses for banks. But because the losses would be realised, they could qualify for a capital loss carry-over that would offer valuable tax benefits.
Provided total capital losses for the year exceed capital gains, a portion of the net capital loss could be used to offset capital gains or a limited amount of ordinary income in future tax years. How limited that amount would be, how far into the future it could be carried, etc, would be for the regulators to decide, but the greater the tax benefit the bigger the discount banks would be willing to accept on their HTM holdings.
If this proposal seems more than a little unconventional, sceptics need to consider how ultimately similar it is to the Bank Term Funding Program that the Federal Reserve created in 2023. Under the BTFP scheme, deposit-taking institutions were offered one-year loans that would be collateralised by US Treasuries, agency debt and mortgage-backed securities. The crucial aspect of the BTFP was that this collateral would be valued at par, not the depressed current market price.
Under this plan, rather than just lending against the par value of deep-in-the-red HTM holdings, the US government should buy them back at a slim but still significant discount to par. Financially speaking, doing so is enlightened self-interest. Unlike the BTFP’s short-term collateralised loans, not only would taxpayers directly and immediately profit, but the underlying debt would be eliminated, freeing up capital and encouraging lending activity.
Runs occur because depositors know banks are operating with a cellophane safety net. With well over 40 per cent of all domestic deposits uninsured, the only thing preventing isolated panics from becoming system-wide contagions is regulators’ willingness to aggressively intervene.
If banks can be bailed under at a financial gain, pre-empting much bigger (and potentially more expensive) problems in a manner that actually socialises profits rather than losses, why not consider a more opportunistic, market-oriented approach?
