Beyond Ratings: CDS Spreads & Market Signals

In 2023, within a couple of weeks of each other Credit Suisse and Silicon Valley Bank both failed. Both were rated by numerous rating agencies. In both cases these ratings remained investment grade right up to the collapse, giving no indication of the changing risk profile for customers of those institutions.

Credit ratings are a useful tool; they help you benchmark exposures across counterparties, and give you a scale for setting your risk appetite. They will not tell you when one of your counterparties is about to default.

We look at some of the other indicators below.

Three prices, three different questions

For Treasurers considering bank exposures, there are three sources of tradeable price data which can give valuable insights into the financial health of a counterparty. But they tell different things, and it is important to distinguish between them.

The three tradeable prices, and the question each one answers.

Share Price. Is the franchise viable?

Equity is the residual claim on the value of the assets after all debt has been repaid. It absorbs the first loss and is the most sensitive instrument to bad news. That sensitivity is also why it is the noisiest. Bank shares can halve on an earnings miss or a sector de-rating, a change in interest rates, or new regulation, none of which might impact the risk of default. Market capitalisation is more valuable than share price. Better yet, look at price-to-book ratio, which looks at whether a bank is valued at a premium or a discount to the accounting values of its assets. The IMF found the European banks worst affected in March 2023 were those trading "at significant discounts to their book values."

Bond Yield. How risky is the debt?

First of all, cut out the noise. Base interest rates are the biggest driver of bond yields. Take away the relevant risk-free rate, and what is left is the spread that compensates for expected loss, plus liquidity and a variety of risk premia. Banks can issue thousands of bonds, with different risk, features and terms. The critical point is that expected default loss is generally a small portion of the spread for investment-grade names. Research (Amato and Remolona) has found an average BBB three-to-five-year spread of 171bp against expected loss of 20bp. So a spread doubling from 80bp to 160bp is not a doubling of default probability.

Credit Default Swap (CDS) Spreads. What does it cost, right now, to insure against this name defaulting?

This is the only one of the three prices that directly answers the question corporate treasurers are asking. It is the only one of the three that is a direct traded price of protection.

Table 1 — What each signal is actually telling you

Why CDS is the cleanest read

CDS is the best signal available because it is standardised. Since the 2009 reforms, single-name CDS trade on fixed standard coupons settled with an upfront payment, on standard quarterly roll dates, under the 2014 ISDA definitions, with a single auction setting recovery.

Comparing two banks' bond spreads means normalising for coupon, maturity, currency, covenants, call features and — critically for banks since resolution reform — whether the issuer was the holding company or the operating company. CDS requires none of that.

Five-year senior CDS on Bank A and on Bank B are genuinely comparable.

Curve shape

The highest-value, lowest-effort check is the shape of the curve, not its level. An upward slope is normal: more time, more risk. A flat or inverted curve says the market thinks the near term is the dangerous part. Credit Suisse's curve inverted on 3 October 2022, five months before it failed, and was deeply inverted again by mid-March 2023, when the one-year closed around 836bp against a five-year of 574bp (eighteen times UBS's one-year). Inversion needs no peer group, no regression and no benchmark, which makes it the check most worth building into a routine.

Junior instruments move first. Subordinated CDS trades wider than senior, and is more sensitive to changes in risk sentiment. Contingent Capital bonds like AT1s reprice earliest and hardest because they are the first to bear loss. On 3 October 2022 a Credit Suisse AT1 fell twelve cents in a day, to 58 cents on the dollar. Watching AT1 prices costs nothing and moves before senior anything.

Limitations of market-based indications

Availability

• If a company has no listing, there is no share price to watch.
• If they have no publicly issued debt, there are no bond yields to track.
• Even then, only the largest issuers have active, liquid regularly traded CDS references.

Unfortunately, the biggest drawback of market-based factors is that your counterparty needs to be in the market to be visible.

Unfortunately, of the c. 5,000 banks in Europe, fewer than 20% have one or more of the indicators we discuss above.

Liquidity

Unfortunately, it gets worse. A price is only information if someone traded at it, and in equity, debt and credit markets very often nobody does. And this problem gets worse as the potential informational value gets better.

Equity markets. A company can have a limited free float on an obscure index, small limited issuance or illiquid CDS, and there may simply not be enough trading activity or price action in their securities to give any informational value about a change in circumstances.

Looking at Argentex's example here, no public debt, no CDS, a limited listing on AIM. Share price had fallen off its peak, but any analysis suggested this was a result of it falling behind some of its peers on technology and expansion. Its share price actually rose over 50% in the first quarter of 2025, while they were teetering on the brink behind the scenes.

Credit. A 2025 review by IOSCO (the global body of securities & markets regulators) found that between 2018 and 2023 around 1,169 CDS reference entities traded at least once, but only 496 traded in every quarter — and those accounted for roughly 80% of all activity. Fewer than 3% of names averaged more than ten trades a day. The ESRB's work is starker: across all EU global systemically important banks combined, single-name CDS averages around sixty trades a day, with only thirteen dealers active daily. Against that, commercial datasets publish composites on over 2,200 entities. A daily level exists for a couple of thousand names; genuine two-way trading for perhaps five hundred; deep liquidity for a few dozen. For most mid-market banks, the spread you are reading is a dealer's model output, not a traded price.

Bonds are no better: FINRA found that less actively traded corporate bonds turn over their issue size roughly once every 600 days, and that much of the price impact of a block trade reverses within five subsequent trades. Most bond trades convey inventory movement, not information. So before acting on a spread move, ask whether the instrument trades at all. If your counterparty's only bond is small, old and rarely traded, a large apparent move may be one odd-lot trade with no information content whatsoever.

No indicator gives a precise probability of default

CDS spreads are the closest, but the price level is not a probability (it includes an estimate of recovery). Use CDS for direction, relative standing and curve shape. Converting it into a default probability only gives false confidence.

CDS references an obligation, not an institution

CDS spreads share this shortcoming with credit ratings. They reference a specific obligation: senior secured debt, subordinated debt or something else. Not the institution itself.

When FINMA wrote roughly CHF 16 billion of Credit Suisse AT1 to zero while shareholders received CHF 3 billion (inverting the normal hierarchy) the ISDA Determinations Committee (who determine what is considered a default) ruled this was not a credit event for subordinated CDS, because the reference obligation was a bond issued in 2000 that ranked senior to the written-down AT1s. Protection buyers got nothing.

Idiosyncratic vs market-wide moves

A fifty-basis point widening means nothing without a frame of reference, and there's no point moving your exposure to another bank only to find all the banks moving in the same direction. Here's how to confirm a move is unique:

Four checks that separate a name-specific move from the whole sector moving together.

For example, in the five days to 10 March 2023, iTraxx Europe Main widened six basis points and large US banks moved 14bp to 28bp. Credit Suisse was already around 350bp and heading for 574bp. Nothing more sophisticated than the ratio test would have shown that divergence from October 2022 onwards.

Cost

Now for the good news. Access to financial markets data is becoming cheaper to access and easier to parse.

There are numerous free to access financial data sites offering share price, including history, and trading volumes.

Access to CDS spreads used to require a Bloomberg terminal or a Markit subscription, both running at over €20k per annum.

ICE Clear Credit publishes daily settlement prices on CDS indices and five-year single names free to the public. The cleared universe is just over 1,000 references across 600 names, but includes most UK & European banks with a tradeable CDS.

Bonds are a little trickier given the size of the universe, but it is possible to patch together the necessary data at a relatively low cost, particularly if you avoid investing in corporate debt.

When there is no listing and no public debt

Unfortunately, this is the situation for many banks and most fintechs and brokers. The honest answer: it takes a bit more effort.

You replace market prices with public disclosures, registers and your own dealings with the firm.

The signals that are still there when there is no price to read.

Banks: regulatory disclosure is the richest free layer

The EBA's Pillar 3 Data Hub went live on 28 January 2026 and makes prudential disclosures from all EEA institutions available in one harmonised, machine-readable, free platform, replacing both the chore of hunting individual Pillar 3 PDFs and the EU-wide transparency exercise, discontinued from June 2025. Alongside it: the EBA's biennial stress test (the 2025 edition covered 64 banks and 75% of EU banking assets; next in 2027); the ECB's SREP results and individual Pillar 2 Requirements, which are supervisory judgement made visible; and quarterly ECB Supervisory Banking Statistics. For US counterparties, the FFIEC Call Reports and Uniform Bank Performance Report are the most granular free bank data anywhere, and the UBPR pre-computes the peer comparison for you.

The metrics that move first are not the headline capital ratio: watch CET1 headroom over the combined requirement rather than the minimum, the trend in LCR and NSFR rather than the level, loan-to-deposit ratio, deposit concentration and the uninsured share, and unrealised securities losses against CET1.

Non-bank financials: registers and filings

Payment institutions, e-money firms and FX brokers are where treasurers have the least protection and the least data. Four free checks should be mandatory at onboarding and repeated periodically:

• the regulator's register - checking the permission rather than merely the presence, since an authorised payment institution is not a bank, and checking for restrictions imposed on it;
• filed accounts at Companies House or the CRO, including audit qualifications and late filings;
• the register of charges, where a newly registered floating charge is a funding-stress signal that costs nothing to see;
• and director and adviser changes.

Argentex is the case to study. An FCA-authorised payment institution with no public debt, and a small closely held listing.

There were some signals if you knew where to look:

Parent proxies, consensus data, and your own experience

For an unlisted subsidiary of a listed or bonded group, the parent's market data is often the only real-time signal — but subsidiarisation cuts both ways. SVB UK was separately capitalised and separately resolvable, which is exactly why the Bank of England could sell it to HSBC for £1 with deposits protected in full while the US parent's bondholders entered administration. The question is not "is the parent strong?" but "is this entity separately capitalised, is there a guarantee, and which resolution authority has jurisdiction?"

Word of mouth

An unreliable source of information sure, but ask any of the founders who got their $42bn out of SVB if they are happy they acted on suspicions from the WhatsApp group chat. This may have created a vicious cycle once rumours get out of control and people start acting on them, but banks aren't charities, and your stakeholders won't thank you if you had information about a potential bank run and took a principled stance or chose not to act.

The signal stack, and what each layer costs

Table 3 — The signal stack, with coverage and cost

How much are news alerts actually worth?

Less than people hope, for two reasons.

Velocity. By the time a story is a story, the market has moved. Social media has made it possible to find timely information on breaking issues before they are reported in the press but the sources are rarely the same consistently, and many sources are now fragmented across numerous social media platforms.

Noise. A simple starting point for any credit analysis is setting up news alerts for your counterparts. For smaller names out of the limelight, this can be a valuable source of information. For larger names, it can quickly overwhelm as the volume of news coverage drowns out relevant indicators of credit stress.

If you are looking for signals, look for specific signals. Enforced senior departures; auditor resignations (Wirecard's shares fell 60% in a day when EY refused to sign off its accounts); delayed reporting; material weakness disclosures (Credit Suisse disclosed material weaknesses in its financial reporting and confirmed that CHF 110 billion of outflows had not reversed on 14 March 2023, five days before it ceased to exist); going-concern language; regulator enforcement, or defensive capital raises.

You have the signals, now what?

The purpose of observing market signals isn't to panic at the first dip in price. They allow you to observe a range of factors, over a period of time and decide a graduated response.

First, you need to have the plan.

Five stages, decided in advance, so the bad week is execution rather than debate.

For example:

The same job at three sizes

Start-up. Much of what we have discussed is beyond the capacity of a start-up finance function. … read more show less

Internal assessments and complex peer analysis are too time-consuming. If your main bank is in the headlines, look at its market capitalisation and its CDS. If both deteriorate significantly vs peers, you may have a reason to raise the second-bank plan for discussion.

Established mid-market. A quarterly review of market pricing and the structural metrics for every name on the counterparty list. … read more show less

An hour or two every quarter, and a refresh on bad news flow.

Large corporate. Live dashboards, paid market feeds, ratio and curve-slope alerts. … read more show less

A risk register listing exposures to each counterparty, and a plan of action for each.

Lessons for corporate treasurers