Corporate solvency study for the second quarter of 2026, covering hundreds of issuers across the European and global credit universe, each measured on the same criteria and compared with its own position a year earlier.
The average of a credit book is the number that reassures most and informs least. This study opens it up.
FGA Research is an independent research firm and a tied agent of Miraltabank. We do one thing in fixed income: measure the credit health of a portfolio’s issuers every quarter on our own consistent, comparable criteria, and deliver an actionable read before the deterioration becomes consensus.
A solvency score per issuer built on six fundamental dimensions and three risk lenses with published literature behind them. The same scale for every name and every quarter.
Every quarter, across the client’s universe: bands, ranking, prioritised watchlist and change alerts between cuts. Not a snapshot: the trajectory.
Report and a reading session with your team. Where the relative risk sits, which names are turning and which specific dimension is driving it.
| What a client receives each quarter | What it solves |
|---|---|
| Solvency map | Score, band and ranking for every issuer on a single comparable scale |
| Prioritised watchlist | Deteriorations and turns, with the flags of the three lenses and the reason for each |
| Change alerts | The names starting to turn between cuts, before it becomes obvious |
| Sector and macro read | Where your relative risk sits and where the alpha is, in cycle context |
| Report and session | Our own document plus a reading session with the investment or risk team |
A solvency aggregate that does not move is an invitation not to look. It is the most expensive signal in credit, because the deterioration does not stop: it accumulates in specific names while the average nets them out.
“The average solvency level of the book is flat. Nothing has happened.”
“Half the issuers deteriorated, and the falls weigh 1.5 times the rises.”
Both statements describe the same universe over the same period. Neither is an opinion: they are two ways of looking at the same data, and only one of them is actionable.
A real case from the FGA panel, identity withheld. A small-cap utility issuer enters surveillance in the upper part of the intermediate band and crosses two full bands in five quarters. Its sector median stayed stable throughout: the sector average never showed anything.
A single fallen angel nobody saw coming can offset years of carry across a whole portfolio: once investment grade is lost the bond stops being eligible for part of the market, the sale comes forced and at the worst moment, and the impact is not spread — it concentrates in the position and in the quarter.
The probability is low; the loss is not. That asymmetry turns surveillance into cheap insurance: a single position avoided in time justifies monitoring all the others.
We tested the system against the history of agency ratings: several hundred issuers over eight years, one agency per issuer so scales are never mixed, and the test adjusted for publication date — it only sees up to the quarter before the rating action. Without that adjustment any credit back-test produces spectacular results and is worth nothing.
The downgrades the system did not anticipate are documented one by one with their verified cause: a leveraged buyout, two large debt-funded acquisitions, the seizure of a subsidiary by a foreign government and the 2020 shock. No analysis of financial statements anticipates a corporate transaction, because the leverage jump is not in the previous quarter’s balance sheet. We publish it because it defines the perimeter of the tool.
Gross sensitivity answers “of all the downgrades that happened, how many had we flagged?”; organic sensitivity asks the same of downgrades whose origin was on the balance sheet. Specificity is the one almost nobody publishes and the one that prevents cheating: of every ten issuers the agency left untouched, the system stayed silent on almost nine.
Hundreds of issuers measured across 24 sectors with sufficient sample. Over twelve months, 44% improve and 50% deteriorate. A split that looks neutral in number and is anything but in magnitude: the typical fall weighs 1.5 times the typical rise, and 17% of the universe suffers a severe setback against 7% posting an equivalent improvement.
One of the strongest levels in the universe and the worst trajectory of the 24 sectors: 83% of its issuers deteriorate. This is not distress, it is deterioration from a high level — and at that breadth, it is not noise.
Also 83% falling, and the highest density of system flags: accounting-quality signals and factor convergence in the same share of the sector.
The largest block and the lowest average level, with 44% in the alert band, but a flat trajectory and the lowest internal dispersion in the book: a level problem, not a trend problem.
At the other end, mining and steel and oil and gas lead with three of every four issuers improving, and the highest absolute level in the universe sits in aerospace and defence.
A back-test measures consistency over data that has already happened. It is not a promise about what comes next. We publish it because the publication adjustment makes it reproducible and because the misses come with a named cause, not because it guarantees anything. Past results do not imply future results and, in this case, that sentence is not legal boilerplate: it is the conclusion of the exercise.
FGA Research & Advisory is an independent firm doing one thing on three fronts: reading complex systems ahead of consensus and saying what we see, with the data in front of us. FGA is a tied agent of Miraltabank.
The audit background is not CV decoration: two of the system’s three risk lenses come from it. Telling profit apart from cash is a craft.
Francisco Salvador has spent more than twenty-five years reading market cycles and has worked with artificial intelligence since 2001, when he founded Tailored Market Monitor, one of the first Spanish research firms built on AI. Before FGA: Arthur Andersen, Santander, Ahorro Corporación, Venture Finanzas, Mirabaud, M&G Valores and Rentamarkets. MBA from IESE (exchange at Kellogg). Ranked by Extel and StarMine.
Seventeen pages: the method, the back-test with its four layers and explicit denominators, the full matrix, the misses with their cause, the 24-sector map entirely in percentages and the annex of declared limits. It identifies no issuer and no portfolio.
There is no direct download: we email it to that address within minutes.
You have just requested the method applied to our own analysis book. For yours there is no PDF worth having: there is a surveillance mandate, quarter after quarter, calibrated by hand against your sector.
A single position avoided in time pays for years of surveillance. That is the whole arithmetic, and it is measured on page 4 of the study you are about to receive.
The specific names are not in the study. They are in the mandate, over your universe and with the system in front of us.
We take a limited number of mandates each quarter, because every calibration is done by hand. The next cut-off is Q3 2026.
Book thirty minutesOr reply to the email you are about to receive telling us how many issuers your universe holds: we will come back with scope and fees on a single page, with no call needed first.
Solvency surveillance · FGA Research, tied agent of Miraltabank