70,803 French business failures: where do the losses go?

A business failure does not erase debt. It determines which claims will be paid, guaranteed, recovered or abandoned.

France recorded 70,803 business failures in the twelve months to June 2026, up 5.1% year on year and 19.3% above the 2010-19 average. At the same time, France’s prudential supervisor described the country’s six largest banking groups as highly resilient, with an aggregate CET1 capital ratio of 15.6% at the end of 2025. (Banque de France, 7 August 2026, ACPR, July 2026)

The contrast looks odd. Tens of thousands of firms stop meeting their debts, yet there is no comparable banking crisis. The first explanation is a confusion between three measurements: a count of court proceedings, the quality of a loan portfolio and a bank’s capacity to absorb losses. Each has a different denominator, timeline and blind spot.

The useful question begins there. If the banks do not bear the whole bill, where does it go? The answer runs through collateral, guarantees, suppliers, France’s wage guarantee scheme, public creditors, landlords, business owners and less visible non-bank finance. It also exposes a data gap: France counts insolvency proceedings, but does not publish a comprehensive national ledger of final losses by creditor class.

Three gauges tell three stories

France’s business-failure statistic gives the same weight to each opening of judicial reorganisation or liquidation. A tradesperson and a mid-sized company each count as one case. The indicator measures frequency, not the euro amount of liabilities in the proceedings.

The composition of the total immediately clarifies the gap. Of the 70,803 failures recorded over twelve months, 65,182 involved microenterprises or firms whose size was not determined. That is 92.06%, our calculation from Banque de France data. All other size categories together accounted for 5,621 proceedings. (Banque de France)

The European Central Bank finds the same coexistence across the euro area: bankruptcies have risen while the aggregate quality of bank loans to companies has remained broadly stable. It points to four factors: small firms’ limited weight in loan books, a shift towards equity, bonds and non-bank lenders, active management of distressed loans, and wide variation by borrower size. It finds no systematic sign so far that banks are artificially delaying the recognition of deteriorated loans. Non-performing loan ratios for small and medium-sized firms have nevertheless been rising since 2023. (ECB, Rising bankruptcies, resilient loan books, May 2026)

CET1 adds a third layer. It does not mean a bank incurs no loss. It measures its highest-quality capital against risk-weighted assets. A bank can therefore report more defaults and remain profitable and well capitalised if exposures are modest, already provisioned, partly recovered and absorbable through earnings and capital. In 2025, the six groups monitored by the ACPR generated €39.8 billion in net income, up 10.2%, providing another cushion before losses reach capital. (ACPR)

The size of the cases changes the diagnosis

The overwhelming share of microenterprises explains part of the apparent paradox. It does not make the wave harmless.

Banque de France found that in 2025 failures among SMEs other than microenterprises were 68% above their 2010-19 average, compared with 12% for microenterprises. It concludes that, outside the smallest firms, a post-Covid catch-up of weak companies temporarily preserved by public support no longer explains the rise on its own. Weaker financial structures and greater difficulty preserving activity now matter more. (Banque de France, Business failures: the dynamics at play after six atypical years, July 2026)

The denominator matters too. Banque de France counted 1.2 million business creations over the twelve months to June 2026, up 11.1% year on year. That flow cannot be used directly to calculate a failure rate because creations and failures cover neither the same population nor the same cohorts. It simply reminds us that a larger business population can generate more proceedings without an identical increase in individual risk. (Banque de France)

The missing figure is a consolidated national balance of liabilities: how much did the failing firms owe to banks, suppliers, public bodies, employees and other lenders? Available public statistics do not provide that final breakdown. Banque de France publishes an indicator of failed firms’ weight in outstanding bank credit, but it does not reconstruct all debts or losses after recovery.

Liquidation turns the balance sheet into an order of claims

A statistical failure does not necessarily mean that a French business closes that day. Judicial reorganisation aims to continue operations, preserve employment and restructure liabilities. Liquidation follows when recovery is manifestly impossible. (French public service, judicial reorganisation, French public service, judicial liquidation)

In liquidation, the court-appointed liquidator inventories and sells assets owned by the company, then distributes the proceeds according to applicable rights and ranks. French law is more complex than a single queue. Some assets may be reclaimed by their owner. A retention-of-title clause may allow a supplier to recover unpaid goods. Security gives a creditor a claim over a specific asset. Some wage claims receive super-priority, claims properly incurred after proceedings open may rank highly, and tax, social-security and secured claims follow statutory privileges. Ordinary unsecured creditors share what remains pro rata. (French Commercial Code, Article L643-8, European e-Justice Portal, insolvency in France)

Shareholders sit at the end of this economic logic: they receive residual value only after every creditor has been paid. A business owner may also be exposed personally if they gave a valid guarantee. Neither loss appears in a bank’s non-performing loan ratio or the court’s count of cases.

The bank records what it will not recover

If a company still owes 100 to its bank, that does not automatically become a loss of 100. There is no universal recovery rate. The bank assesses expected payments, the value of security, guarantee coverage and recovery costs. It records provisions when expected credit losses rise, potentially long before liquidation. Its final loss is the part that will be neither paid, recovered nor covered.

European insolvency data assembled by the European Banking Authority confirm the importance of rank and collateral. Secured loans recover more than unsecured loans, large-company loans more than SME loans, and judicial costs weigh proportionally more on small cases. The comparison covers 1.2 million loans with reference dates before December 2018. It documents the structure of the mechanism, not a current French recovery rate. (EBA, 2020)

Banks remain central to French corporate finance. In May 2026, Banque de France measured €2,184.9 billion in financing for non-financial corporations, including €1,429.6 billion in bank credit. SMEs and firms of undetermined size accounted for €551.8 billion. A large share of the risk therefore remains on bank balance sheets. (Banque de France, corporate financing, May 2026)

Aggregate resilience means that the risk currently appears absorbable at the system level. It guarantees neither the absence of losses at a particular bank nor future access to credit for a vulnerable borrower. A bank can preserve its capital by tightening standards, raising margins or demanding more security. The economic cost then returns through the surviving borrowers.

A guarantee changes the final creditor

France’s state-guaranteed loan scheme offers the clearest example of a loss changing address. When a guarantee is called, the bank receives the covered share. The claim does not vanish: the public guarantee mechanism bears that portion of the risk and acquires recovery rights. The bank retains the uncovered part and must follow the scheme’s rules.

The pandemic-era prêts garantis par l’État, or PGE, show the order of magnitude without being attributed to the 2026 wave alone. By 31 October 2025, 39,531 guarantee calls had been recorded among 686,414 recipients. Calls paid through the end of 2024 totalled €4.8 billion. A Banque de France model published in July 2025 estimated gross state losses at €6.7 billion for 2020-28 and net losses after guarantee premiums at €3.8 billion. These amounts span several loan vintages and are not the bill for the 70,803 failures recorded in June 2026. (French Senate, Economy Ministry answer, 12 March 2026)

The mechanism protects the bank on the guaranteed share while transferring some risk to the public finances. It is a concrete reason why a bank may remain strong even though an economic loss has surfaced elsewhere.

Suppliers form a second credit system

A French company does not borrow only from its bank. A supplier that delivers today and accepts payment in 30, 60 or 90 days extends trade credit. If the customer enters liquidation before payment, the invoice becomes a claim in the proceedings.

The supplier may be protected by retention of title, trade credit insurance or security. Without enough protection, it bears some or all of the shortfall. That drains cash, may reduce purchases, delay investment and weaken the supplier’s own payments. Academic work documents this channel of failure propagation through supplier networks. (Banque de France, The amplifier effect of trade credit, Jacobson and von Schedvin, Econometrica, 2015)

This risk is particularly hard to aggregate. France’s Payment Terms Observatory measures delays, not final losses after every liquidation. Its 2024 report found that more than 9% of French companies suffered payment delays exceeding 30 days. The figure indicates cash-flow strain but cannot isolate the share caused by firms already in insolvency. (Payment Terms Observatory, 2024 report)

Wage claims have an employer-funded buffer

When a failed business lacks the necessary cash, the Association pour la gestion du régime de garantie des créances des salariés can advance covered wages, severance and other claims within statutory limits. AGS is financed by an employer contribution. After payment, it is subrogated to employees’ rights and seeks recovery in the insolvency proceedings. Treating every advance as direct French government spending would therefore be inaccurate. (French Labour Ministry)

In 2025, AGS advanced a record €2.233 billion for nearly 250,000 employees at 27,000 companies. This measures a cash advance and social guarantee, not necessarily a final loss, because AGS subsequently recovers part of the money as a creditor. (AGS, 2025 annual report)

Employees remain exposed beyond guaranteed claims. Closure destroys jobs and future income, while some claims may fall outside the scheme or exceed its caps. None of that lost income appears in a bank loan book.

Non-bank finance widens the statistical gap

Bonds, debt funds, private credit, insurers, factoring companies and leasing firms may also hold claims. Their presence does not establish that French credit risk has largely left the banks: the financing figures show otherwise. It does make the map less complete.

Across the euro area, the ECB estimates that private credit funds managed about €100 billion in 2025. More important for national measurement is the cross-border structure: 60% of euro-area investors’ private-credit commitments go to foreign funds, while 70% of the private-credit financing received by euro-area companies comes from funds outside the euro area. National aggregates therefore struggle to identify the final holder. (ECB, The growing role of private credit, May 2026)

The ACPR says the direct and indirect exposure of major French banking groups to private-credit funds remains below 0.5% of their balance sheet. That constrains the identified direct banking risk. It does not measure all private debt owed by French firms or all losses incurred by the ultimate investors. (ACPR)

This is the French ground-level counterpart to our broader analysis of the migration of credit risk out of banks and out of sight. The conclusion here must remain narrow: non-bank finance explains part of the visibility gap, but it is not an estimate of where France’s insolvency losses went.

The indicators that would change the diagnosis

The raw count of cases remains useful for measuring how widely stress is spreading. Six indicators are more informative when looking for a change in regime:

1. The size of failing firms. A sustained acceleration among SMEs, mid-sized companies and large firms increases the jobs, debt and supplier chains affected much faster.
2. Failed firms’ weight in outstanding credit. This connects court cases to the amount actually exposed.
3. Non-performing loans and cost of risk by segment. Deterioration among small firms can remain hidden in an aggregate bank ratio.
4. Calls on public guarantees. They track the conversion of guaranteed risk into a potential public-finance charge.
5. Payment delays and trade-credit insurance losses. They flag propagation towards suppliers.
6. Collateral values and recovery rates. Falling property, machinery or inventory values increase the final loss for the same debt.

The ECB also identifies a vulnerable tail: about 6% of companies in its sample are classified as non-viable, with a similar proportion just below the viability threshold. This is a euro-area firm sample, not a census of French businesses. It identifies where to look, without forecasting a specific failure wave. (ECB)

The final loss remains partly invisible

Available evidence establishes four points. French business failures are high. Their count is dominated by very small firms. French banks still have substantial earnings and capital. A collective proceeding distributes losses across several holders according to assets, contracts, guarantees and law.

It does not establish what national share of the bill was borne by banks, suppliers, public bodies, employees, landlords, insurers or funds. Adding AGS advances, PGE guarantee calls and bank provisions would create double counting: a guarantor that pays may become a creditor, a provision may precede the loss, collateral may be realised later and an invoice may eventually be recovered in part.

Current bank resilience therefore provides important but limited information. It says observed losses have not so far overwhelmed the banking core’s buffers. It does not say the real economy paid nothing. Insolvency distributes value that has become insufficient. To find who ultimately bears the difference, the claim must be followed to its last holder. France does not yet publish that ledger.

Main sources and limitations

• Banque de France, business failures to June 2026: number of proceedings, size breakdown, changes and business creations.
• Banque de France, failure dynamics after six atypical years: post-Covid catch-up and financial weakening outside microenterprises.
• ACPR, major French banking groups at end-2025: capital, profitability, liquidity and exposure to private-credit funds.
• ECB, bankruptcies and bank credit quality: euro-area analysis and differences by firm size.
• French Commercial Code, Article L643-8, French public service and European e-Justice Portal: liquidation, creditor rights and ranking.
• French Senate, PGE losses and guarantee calls and AGS, 2025 annual report: two mechanisms that advance or transfer losses.
• EBA, European recovery comparison and Jacobson and von Schedvin, Econometrica: the effect of collateral and propagation through trade credit.

The figures do not all cover the same date or perimeter. ECB and EBA data are European and are used to explain mechanisms, not presented as a French loss allocation. No historical recovery rate is used as a current estimate. The absence of a national creditor-by-creditor ledger is treated as a documentary limit, not as confirmation that losses sit in any particular category.