In property insurance, a loss leaves behind a record the insured did not have to manufacture: a fire brigade report, a police statement, a joint expert appraisal. Political risk almost never offers that. The facts that decide a file are refusals, silences, repeated delays and pressure applied without a written word, and a state that intends to be able to deny does precisely not write. Evidence here is therefore not gathered afterward, it is provoked during: it exists only if the insured wrote first. That is the heaviest difference in method between this line and every other, and it is the one discovered latest.
Provoking the record rests on a simple mechanism that is poorly practiced. A written request, filed and time-stamped, necessarily produces one of two useful results: either an answer, which is evidence, or a silence, which becomes datable as soon as the request was dated. A meeting in which a ministerial chief of staff orally announces the withdrawal of a license proves nothing and can be produced before nobody. A letter sent the next day summarizing that meeting, naming those present, and left undenied for three weeks proves a great deal. The practical rule fits in one sentence: every oral event must be answered by a writing that fixes it, sent quickly enough that a denial would have been the natural response.
This method runs into a conflict that is not negligence, and that must be named to be handled. The natural author of those writings is local management, and it is local management that has the strongest reasons not to sign them. Its executives live in the country, their families reside there, their residence permits and sometimes their liberty depend on the very administrations to which a dated protest would have to be addressed. Their reluctance is rational. It is resolved by moving the pen rather than by insisting: the writings that build the file are issued by the parent company or its counsel, from abroad, and local management is no longer put in the position of signing what would expose it.
A further difficulty catches most groups too late: they are in fact building two files at once, and those files do not want the same thing. There is the notification to the insurer, and there is the claim against the state, most often an investment arbitration founded on a bilateral treaty. Both demand evidence, and not the same evidence. What the insured asserts to its insurer can be produced against it before the arbitral tribunal, and the reverse is just as true. A notification placing the loss of effective control in March serves the cover and weakens an arbitration claim arguing that the investment retained its value until September. These two files are written together, or they contradict each other.
The notification calendar deserves particular attention, because the policy packs into it two obligations pulling in opposite directions. It requires notification without delay, and it recognizes only a constituted loss. Notifying too early invites the answer that no loss yet exists, and the notification may then be set aside without the deadline being preserved for all that. Notifying too late breaches the notification obligation, with whatever consequences the contract attaches. Many wordings resolve the vise with a distinct instrument, the circumstance notification, which flags a fact liable to engage the cover without asserting that the loss is established. Not using it means condemning oneself to choose between two faults.
Meanwhile, the accounts are writing dates, and they consult nobody. The impairment of a shareholding, the deconsolidation of a subsidiary, the note that justifies them: each of those acts carries a date and a written rationale, settled with the statutory auditor, very often before the insurance file has been built at all. They are dated statements, made by the insured, against its own interest, drafted by people who did not know they were writing a piece of evidence. They will be read, because they are exactly what a diligent insurer asks for first.
And yet the two dates can legitimately diverge, which is the point almost nobody knows how to defend. The accounting test and the contractual test do not ask the same question. Deconsolidation looks at who in fact holds the power to direct the entity's activities. The cover looks at the act of the state depriving the insured of its asset or of its control. An accounting loss of control in March and an expropriation triggering event in September can perfectly coexist, on one condition: that the divergence was explained at the time, in a written note when both dates were settled. A divergence explained the same day is an argument. Reconstructed after the claim is notified, it can no longer be explained, it is merely endured.
On September 8, 2025, a decree appoints a public administrator to head the subsidiary of a French group. On December 31, 2025, at the statutory auditor's insistence, the group fully impairs its shareholding and deconsolidates the subsidiary; the note to the accounts states that loss of control has been recognized since March 2025, the date on which the group ceased to be able to appoint the management. In April 2026, the group notifies an expropriation claim relying on September 8, 2025 as the triggering event, a date on which the policy had just been renewed on broader terms. The insurer asks for the audit file. The finance director proposes having the note corrected. What is the answer?
The correction must be refused, and said plainly: it is the one move in the file that is both useless and gravely dangerous. Amending after the fact a note settled with the statutory auditor does not make the original version disappear, since it sits in the audit file the insurer has just asked for. The maneuver would turn a defensible divergence into a documented attempt to align the evidence, which contaminates the rest of the file and additionally exposes the group on the accuracy of its accounts. The right answer is to hold both dates and explain why they differ. The two tests do not ask the same question: deconsolidation looks at who in fact holds the power to direct, which points to March, when the group lost the appointment of management; the cover looks at the state act depriving the insured of its control, which points to the September decree. The difficulty is that this explanation was not written at the time, and that it arrives after a notification whose date benefits the group, under a policy renewed on better terms. The burden is therefore heavier than it should have been, and it is carried with the documents of the period, not with a note drafted in April 2026.
- 01Political risk evidence is not gathered, it is provoked: without a written, dated request, a state's silence cannot be dated.
- 02The writings that build the file come from the parent company: local management has the most to lose by signing them.
- 03The circumstance notification exists to escape the vise between notifying without delay and notifying only a constituted loss.
- 04The date recorded in the accounts is a dated statement by the insured against its own interest, and it will be read first.
- 05A divergence between the accounting date and the triggering event is defensible if it was written at the time; correcting the accounts afterward turns an argument into a suspicion.