After Credit Suisse, markets find bail-in less credible

When Credit Suisse failed in March 2023, Swiss authorities, against market expectations, set aside the resolution plan and arranged a UBS takeover backed by public guarantees, sparing the bail-in bondholders who were next in line to absorb losses. This column traces how that choice reshaped investor beliefs about bail-in across Europe. Spreads on bail-in debt narrowed, the funding penalty on weaker banks shrank, and yields moved less on earnings news. Each points the same way: investors now see bail-in as less likely, a sign that its credibility has declined.

When a large bank fails, the losses do not disappear. Someone absorbs them. The reforms of the past decade settled who: shareholders first, and when that does not suffice, a designated class of bondholders, through a mechanism called bail-in. The design was meant to put an end to too-big-to-fail and to protect taxpayers. A promise like this is tested only by a failure. Bail-in had been imposed on smaller banks, in calmer markets. No global systemically important bank had failed under the framework built for it.

Untested does not mean distrusted. If anything, belief in the promise had been strengthening. Berndt et al. (2025) find that the largest US banks had lost much of the funding discount the expectation of rescue once bought them. In Europe, Schäfer et al. (2016) find that credibility rose most where authorities had imposed losses, not merely where they had legislated for them.

How far that credibility extended was the open question; doubts grew with the size of the failure (Goodhart et al. 2021, Brunetti 2023). In March 2023, the test finally came. Credit Suisse, one of the 30 global systemically important banks, was failing. Concerned that a disorderly collapse would cause contagion, Swiss authorities set aside a prepared resolution plan and arranged a takeover by UBS, backed by public guarantees. The bank’s most junior bonds, known as Additional Tier 1 or AT1, were written down in full. The bail-in bonds one rank above them, which the plan would have converted next, were spared.

Markets had not expected this. In the days before the merger, investors had been pricing Credit Suisse’s bail-in bonds for losses. Interviews the Financial Stability Board conducted that summer found the same: large investors had expected resolution, with bail-in bonds bearing losses alongside AT1. Market participants later told the Expert Group reviewing the collapse that they had judged contagion risks less severe than the authorities did (Expert Group 2023). Figure 1 tracks Credit Suisse’s own AT1 and bail-in spreads into the merger. The two climb together through the final weeks as losses look likely in both layers. The announcement splits them: the bail-in spread falls back as those bonds are spared, while the AT1 bonds are written down in full.

Reading the change in the probability of bail-in from prices

The signal reached past Credit Suisse. The framework the Swiss set aside is the international standard for winding down a large bank. Any authority facing a failing global bank would meet the same pressures. If markets read the weekend as a lesson about how authorities act when a global bank fails, then bail-in should have looked less certain across the board. In Di Stefano et al. (2026), we test whether it did, using the prices of other banks’ debt.

Interpreting these prices takes some care. A bank’s creditors are ranked, AT1 first, bail-in next, senior last, and each layer’s spread reflects how likely losses are to reach it. A spread contains the product of two probabilities: the chance a bank gets into trouble at all, times the chance losses reach that layer once it does. So, a narrower bail-in spread might mean a lower chance of trouble, or a lower chance of bail-in once trouble comes. In addition, the layers are entangled: bail-in stands behind AT1, so what happens to bail-in depends on what happens to AT1 first.

That dependence is what we use. Losses reach bail-in bonds only after they have reached AT1, so everything that prices AT1 risk is inside the bail-in spread too. The bail-in spread adds one factor: whether authorities then impose bail-in or step in. When the bail-in spread moves relative to AT1, the change is in that factor. Senior is the check: it would have moved had the odds of outright failure shifted, and it did not.

Figure 1 Credit Suisse bond spreads before the merger

Notes: Weighted-average yield spreads over the ten-year sovereign benchmark for Credit Suisse’s AT1, bail-in, and senior unsecured bonds in the months before the merger, announced on 19 March 2023. The bail-in and senior series are restricted to bonds with at most four years of remaining maturity to keep duration broadly comparable. Trading resumed on 20 March 2023.
Source: Di Stefano et al. (2026).

Bail-in less credible, discipline weaker

We ran the comparison across roughly 2,000 bonds from 94 banks in 22 countries, over the year around the merger. It gave one answer everywhere: the bail-in spread fell relative to AT1. Figure 2 plots the change in each spread by region, as a percentage of its pre-event level. In Switzerland, authorities had written AT1 down, so AT1 spreads rose, yet bail-in spreads fell by about 6%. A rise in AT1 risk should raise bail-in spreads as well: when losses reach AT1 more often, they reach the next layer more often too. For bail-in spreads to fall while AT1 rose, the chance of bail-in itself had to fall. In the euro area and the UK, AT1 spreads also fell, but bail-in spreads fell by more: 13% and nearly 14%. That is a large cut in the premium for bearing bail-in risk: markets marked down the odds that authorities would impose bail-in.

Figure 2 Long-run change in AT1 and bail-in spreads, by region

Notes: Estimated change in each tier’s average spread in the 181–365 days after the 19 March 2023 announcement, relative to the pre-event year, as a percentage of the pre-event level. Whiskers show 95% confidence intervals of the estimated spread change, scaled by the pre-event mean; standard errors clustered at the bank level. Credit Suisse excluded.
Source: Di Stefano et al. (2026).

A fall in the odds of bail-in has two interpretations. It may be that credibility declined, as authorities look less likely to impose bail-in when the choice arrives. Alternatively, it may be that necessity declined: a failing bank can be sold privately instead, and bail-in is needed less often. The Credit Suisse takeover looks like the latter. But it depended on a federal loss guarantee, and the price UBS paid rose with the size of that guarantee. The sale itself broke no new ground. When Spain sold Banco Popular to Santander in 2017, its bail-in creditors bore losses. What was new in 2023 was the sparing: a global bank sold with its bail-in layer left whole, on public support (Coeuré et al. 2024).

The same answer comes from which banks benefited most. Had Credit Suisse made bail-in look less necessary, the smallest banks, which are the easiest to sell, should have seen the largest spread falls. A public rescue points the other way, worth most to the bank closest to needing one. In the data, the lower a bank’s credit rating, the more its bail-in spread fell; size made little difference.

Bail-in was never only about who pays. Creditors with money at risk watch how much risk a bank runs, and charge for it. A likelier rescue dulls both. After Credit Suisse, the charging eased: the extra yield weak banks paid over strong ones narrowed. The watching faded too: a bank’s earnings news moved its bail-in bonds about a third less than before. The episode gave back part of what the reforms had been winning.

What repair would take

The reflex after every failure is more equity. More equity means fewer failures, but not none: no level of it removes the states where a bank stops being viable. Bail-in exists for those states. More equity may make them rarer; it does not change the choice authorities make there, and that choice is what markets repriced after Credit Suisse. The reforms built both: equity against losses, bail-in for the states beyond it. Neither substitutes for the other.

Repairing the credibility of bail-in depends on why the Swiss set the plan aside. Either resolution was not feasible that weekend, or it was feasible and the authorities were not willing to use it. Prices cannot tell these apart. Authorities can address feasibility in advance: arrange the funding for a resolution, settle the cross-border steps, test the plans under stress. They can show willingness only at a failure, by imposing losses as the rules prescribe. Credibility grew for a decade and declined in one episode. The next failure will therefore decide more than one bank’s fate. It will show whether the promise holds when keeping it is hardest.

Source : VOXeu