Car insurance in France: non-fault claims and renewal

A driver can be blameless and still become a customer their insurer no longer wants. Rising claims costs help explain that decision. They do not justify every non-renewal, or establish that the insurance industry is in financial distress.

Research current to 29 September 2026. Annual financial results refer to 2025; 2026 figures and hypothetical examples are identified separately. The legal discussion concerns France.

A parked car is hit. Then it happens again. Its owner is responsible for neither accident. Imagine that, when the policy comes up for renewal, the insurer declines to keep the business. To the driver, the logic is baffling: they bought protection against bad luck, only to discover that bad luck has made them unwelcome.

On 29 July 2026, France’s insurance supervisor, the ACPR, published a survey of 14 insurers. It questioned the use of non-fault claims, including notifications resulting in no payment, in termination decisions. Source 1

Does that mean insurers fear for their survival? The available accounts do not support that conclusion. Understanding what is happening requires separating two calculations rarely explained together in a customer letter: the cost of the last accident and the expected future cost of keeping the policyholder.

Renewal depends on an estimate of future risk

Responsibility establishes who caused the damage. Pricing asks a different question: how much might this policy cost during its next period of cover?

Under France’s regulated bonus-malus system, an entirely non-fault accident does not trigger the surcharge associated with an at-fault claim. But that coefficient is only one component of the price. It does not freeze the underlying premium or bind the insurer to offer renewal indefinitely. Source 2

Consider the parked car again. Two collisions might reveal an unusually exposed parking location. They might also be unrelated misfortunes with little lasting predictive value. Assessing their predictive value requires data on comparable risks.

The insurers surveyed cite greater future exposure. Their individual models cannot be validated from this publication. Source 1

A correlation can help predict a loss without establishing fault. But it still needs to hold up on new cases and add useful information. Where parking location is already reflected in the premium, for example, the treatment of associated accidents needs to avoid counting the same underlying exposure twice. The public data examined here do not quantify this possible overlap.

French law also distinguishes declining renewal at the annual expiry date from cancelling mid-term following a claim. For an individual’s motor policy, annual non-renewal requires, among other things, reasons and notification at least two months before expiry. Claim-triggered mid-term cancellation in motor insurance is subject to specific conditions, including alcohol, drugs or certain driving offences leading to licence suspension for at least one month or cancellation. Two non-fault accidents do not, by themselves, satisfy those conditions. Other legal grounds for ending a policy may exist. Source 3 Source 4 Source 5

The bill depends on the cost of each claim

The starting point of insurance economics is a multiplication: how often losses happen, and how much the average loss costs. Improvement in the first can be overwhelmed by deterioration in the second.

According to industry association France Assureurs, the frequency of motor property-damage claims fell 2% in 2025, while their average cost rose 5.3%. That nominal average was 39% higher than in 2020. This is not a constant-quality repair-price index: changes in the mix of vehicles and damage can also move the average. Source 6

Apply the 2025 changes to an unchanged, hypothetical portfolio. One hundred claims costing €2,000 each produce a €200,000 bill. Ninety-eight claims at €2,106 produce €206,388. Costs rise by roughly 3.2% despite fewer claims. This illustrates the arithmetic; it is not an estimate of the actual increase in claims expenditure across the entire market.

l0g calculation: 0.98 × 1.053 − 1 = 3.194%. Bars share a zero-based scale. Changes in motor property-damage claim frequency and average cost in 2025: France Assureurs. The starting portfolio is fictional; the calculated increase does not represent the market total.

Even that multiplication does not finish the insurer’s calculation. After compensating its customer, an insurer can, subject to the applicable legal conditions, pursue the party responsible to recover the money paid. Source 24

Assessing a policy’s economics should therefore involve expected costs after those recoveries, operating expenses and reinsurance, which transfers some losses to another insurer. If one company reimburses another, the charge moves between balance sheets; the repair bill does not disappear from the insurance sector. A count of reported incidents cannot capture all of this.

If expected costs rise, an insurer has several possible responses. It can increase premiums, change deductibles or coverage, reduce expenses, work to reduce damage, or decline certain renewals. More restrictive underwriting is therefore an economically plausible response to cost pressure. That does not establish an industry-wide instruction, or explain any particular customer’s case.

Customer selection in a financially sound sector

France’s motor insurance market recorded a net-of-reinsurance accounting combined ratio of 101.2% in 2025, according to France Assureurs’ study published on 20 July 2026. In simplified terms, claims and expenses included in that measure amounted to €101.20 for every €100 of premiums: an underwriting loss, before considering investment income among other items. Source 7

The position differs across business lines. For the household property-damage branch, which is broader than standard multi-risk home insurance alone, the corresponding ratio was 97%, improving from 98.4% in 2024. Both figures come from the industry’s representative body: a primary source with a stake in the discussion. Source 8

The supervisor supplies a broader view. Across its prudential reporting perimeter for non-life activities, the net combined ratio was 95.3% in 2025. Non-life insurers excluding bancassurers had eligible own funds equivalent to 299% of their solvency capital requirement, up from 278% at the end of 2024. That measures regulatory capital coverage, not available cash. These reporting populations do not exactly match the preceding industry statistics. Source 9

The figures do not describe general financial weakness. Nor do they establish that every company, region or category of policy is profitable.

A financially sound insurer may still conclude that a particular group of customers would weaken future returns. It might keep them at a higher price or decide not to offer renewal. Strong aggregate finances and selective underwriting are entirely compatible. Establishing that non-renewals caused the improvement in profits, however, would require a decomposition that these aggregates do not provide.

For the customer losing cover, the distinction matters. A portfolio-management decision can have serious consequences without the insurer being anywhere near insolvency.

Climate pressure and the scope of Cat Nat

A natural event can damage many insured properties at once. It concentrates losses in a way that simply spreading policies across thousands of customers may not adequately offset. France’s Cat Nat system provides a specific pooling arrangement involving private insurers and state-guaranteed reinsurance offered by CCR. It is distinct from the full range of insurance covering weather-related damage. Source 10

On 1 January 2025, the Cat Nat surcharge rate on the relevant property-damage premiums, excluding motor insurance, increased from 12% to 20%. That does not mean the entire insurance bill rose by 20%. On an unchanged hypothetical €100 base, excluding other taxes, the component calculated this way rises from €112 to €120, or about 7.1%. Source 11

Risk did not stop developing when the 2025 accounts closed. On 4 September 2026, CCR announced a €500 million provision for drought-related clay shrinkage and swelling, ground movements that can damage buildings, for losses occurring in the first half of 2026. Its half-year net result remained positive at €143 million. The provision is an accounting estimate of claims expense, not cash already paid to policyholders or an estimate of France’s full-year drought bill. Source 12

This is an identifiable, current pressure. It does not automatically connect a home insurance questionnaire, or two non-fault motor claims, to a climate-driven withdrawal strategy. Establishing that connection requires evidence linking the particular risk, policy and decision.

Reinsurance coverage depends on its threshold

It would be equally inaccurate to suggest that insurers, in September 2026, have no opportunity to transfer risk. Aon describes a global reinsurance market with record capital and favourable conditions for buyers. Gallagher Re also reports capital growth in the first half. Their assessments point in the same direction, although both firms are brokers with a commercial interest in the market and their measurement perimeters differ. Source 13 Source 23

The price of protection and the losses it covers remain separate questions. At the January 2026 renewals, SCOR described a more competitive pricing environment and broadly stable terms for its portfolio, including the thresholds at which coverage begins. These observations concern SCOR’s own portfolio. Source 14

Imagine a policy that covers losses above €10 million per event, with no additional annual aggregate protection. Three separate €8 million events trigger no reinsurance payment. The insurer retains €24 million of losses, even if the price of purchasing its protection has fallen.

l0g diagram and calculation: three distinct events with no annual aggregate protection. Bars share the same zero-based scale. Framework: ACPR, glossary p. 22. Private excess-of-loss reinsurance example, separate from Cat Nat treaties.

This example illustrates private excess-of-loss coverage. It does not describe the particular reinsurance treaties used in France’s Cat Nat system. Its purpose is to explain why cheaper reinsurance does not mechanically remove the cost of more ordinary losses. Conversely, easing conditions can improve margins or make broader cover affordable. The current softening therefore belongs in the analysis alongside claims-cost pressures.

Updating a home policy can change the eventual payout

A home insurance questionnaire updates the description of the insured risk. French law requires policyholders to report new circumstances that aggravate or create risks and make their earlier answers inaccurate or outdated. It does not require every domestic change to be reported indiscriminately. Source 15

The implications extend beyond the premium. Where an unintentional misstatement is discovered after a loss, the payout can be reduced by the ratio of the premium actually paid to the premium that should have been charged. Paying €300 when €400 was required would, in a simplified example, reduce an otherwise payable €10,000 claim to €7,500. Source 18

l0g calculation under Article L. 113-9 of the French Insurance Code. Unintentional misstatement discovered after the loss. The bar represents the claim before reduction, with no other contractual limit in this example.

An honest error is therefore not necessarily consequence-free. Nor should it automatically be treated as deliberate deception, which is governed by different legal rules. Updating risk information also works in the other direction: a reduction in risk entitles the policyholder to a lower premium under the conditions established by law. Source 16 Source 17

A critical assessment needs to examine the questions actually asked, how clearly the information requirements are explained, and what the insurer does with the answers. Repeated reminders alone establish neither a trap nor a coordinated exercise linked to a motor non-renewal.

A non-renewal can follow the customer elsewhere

The effects of a decision can extend beyond the first contract. The motor termination database operated by AGIRA allows member insurers to check an applicant’s history, including claims characteristics and the reason a previous policy ended. Source 19

It also contains contracts ended at the policyholder’s initiative. It is not an insurance blacklist imposing a ban on coverage. AGIRA states that records are retained for five years for terminations following a claim, compared with two years generally, subject to particular rules for unpaid premiums. Source 20

There is nevertheless a possible circularity. One insurer might decline renewal because of two accidents. The next might then treat both the accidents and the non-renewal as adverse signals. But if the second fact simply results from the first, the two signals may convey overlapping information.

That possibility deserves testing. The public sources examined here do not establish how often it occurs or whether underwriting models fail to account for it. Sharing information may instead improve assessment by distinguishing circumstances more accurately. The outcome depends on the information recorded and how it is interpreted.

A route to compulsory cover exists, subject to procedural conditions, through the Bureau central de tarification. In motor insurance, its remit concerns mandatory third-party liability. It does not guarantee restoration of comprehensive cover equivalent to the policy a driver has lost. Source 21

Access is harder to measure than solvency

The ACPR found relative stability in claim-related terminations over 2021–2023, not an established trend for 2026. Source 1

CCR’s insurability observatory, released on 15 June 2026, found no complete insurer withdrawal within its study perimeter, centred on detached houses and three natural hazards. Local presence, however, guarantees neither acceptance of every household nor an affordable quote. Reporting lags also matter: the next update, planned for early 2027, will use 2024 data. Source 22

Measuring a current tightening would require tracking refusals and non-renewals by initiator and reason, followed by what happens to the affected customers: replacement cover, exclusions, deductibles and premiums on a like-for-like basis. Profitable accounts do not answer those questions. Individual accounts of difficulty, however consistent, cannot supply a national rate either.

Selection can have an economic justification. An insurer that systematically ignores differences in risk may underprice some policies and shift the resulting costs onto other customers. More granular segmentation, however, can also move the burden towards households already exposed to harm, without reducing the probability of an accident or property damage.

The documented picture is of a motor insurance business under cost pressure and contested selection practices within a sector that retains substantial financial buffers overall. The nationwide scale of non-renewals after two non-fault accidents remains unknown. The question worth following is more concrete: when an insurer improves its portfolio, has the underlying risk fallen, or has that risk simply become harder for the customer to insure?

Sources and documents

Method and limitations

Sources consulted on 29 September 2026. Annual accounts cover 2025; 2026 information is dated. The ACPR termination survey describes 2021-2023. Figures from France Assureurs, CCR, Aon, Gallagher Re and SCOR are attributed to those organisations. The three infographics and the Cat Nat calculation use explicitly labelled fictional examples. Possible double counting of risks remains a hypothesis to test. The cover is an AI-generated editorial illustration.

Car insurance: how insurers decide which customers to keep