Subrogation is the most ordinary mechanism in insurance: the insurer that has paid steps into its insured's place and pursues whoever owes. Against a private debtor it works, and it is in fact what makes pricing possible. Against a state, the whole arrangement runs into immunity, and one must see immediately that the difficulty is not one procedural obstacle among others: it is structural, it was known at placement, and it explains a large part of the price of political risk covers.
The paradox states itself in one sentence. International law has long applied a restrictive immunity: a state is not immune for its commercial acts, those in which it behaves as an ordinary operator, and it is immune for acts of sovereign authority. Yet the facts that trigger a political risk cover are by definition acts of sovereign authority: expropriating, closing the FX market, forbidding a state-owned company to pay a foreign supplier. The act that makes the loss covered is exactly the one immunity protects. The cover and the recourse run into the same fact, which is why bringing the matter before the ordinary courts of the insurer's own country rarely leads anywhere.
One must then distinguish two immunities, because people remember the first while the second decides everything. Immunity from jurisdiction bars bringing the state before a court. Immunity from execution bars seizing its assets, and it survives the first: a creditor can perfectly well obtain a judgment and have nothing to enforce it against. Central bank assets enjoy specific protections in most financial centers, property assigned to a diplomatic mission cannot be seized, and only assets allocated to commercial activity are in principle attachable, something states have long organized so that few exist and they sit in inconvenient places. An award obtained is not a sum recovered, and confusing the two distorts the entire assessment of a file.
The route that remains is investment arbitration, where the state consented in advance to being brought before a tribunal, in a bilateral investment treaty. That consent changes everything, but it opens a question few people ask before they need it: the right to bring the claim belongs to the investor, in its capacity as a national of a state party to the treaty. Can the insurer that paid, and that is not that investor, exercise it? The answer is not in the policy, it is in the treaty. Many treaties contain a subrogation clause expressly recognizing the rights of an insurer that has indemnified, and that clause exists precisely because the problem is real. Where it is absent, the recourse can be closed even though the case is strong on the merits. It is checked before the loss, at placement; afterward, it cannot be added.
One must then look at where the money actually comes from, because it is almost never from a seizure. It comes from a negotiated settlement, and the leverage that produces it is not legal but financial. A state preparing a return to the bond markets, negotiating a program with a multilateral lender, or seeking to attract direct investment again needs to clear its arrears, because pending claims raise the cost of its debt or block its access. The value of a recourse therefore tracks the state's need to refinance far more than the strength of the file. That carries a methodological consequence which disconcerts lawyers: the recovery window opens and closes for market reasons, at moments unrelated to the maturity of the proceeding, and it shuts as soon as the issue has been placed.
Time does the rest. Between indemnification and an outcome, five to twelve years commonly pass, and recoveries are shared pro rata: the insurer for the share it paid, the insured for the share it kept. Both therefore hold the same claim, heavily discounted, on a distant horizon, which looks like perfect alignment and is not.
Because the two do not want the same thing when the time comes to settle, and that is the last mechanism to know. The policy almost always gives the conduct of the recourse to the insurer and places a duty of cooperation on the insured, breach of which can be sanctioned since it prejudices subrogation. But an insurer that has reserved its loss and wants to close a line often prefers to bank a discounted settlement now, while an insured still carrying its uninsured share may want to go all the way to an award. A global settlement concluded by the insurer can extinguish the claim without the consent of the party holding the remainder. That conflict is not resolved on the day the offer arrives: it is settled at placement, in the clause saying who decides to settle, on what conditions, and what becomes of the uninsured share.
In 2021, the host state withdraws a French group's mining concession. The insurer pays 36 million euros, that is 90% of a 40 million loss; the group retains 4 million uninsured. The applicable bilateral treaty contains a subrogation clause. In 2024, an arbitral tribunal upholds its jurisdiction, and the case is considered strong on the merits. In 2026, the state, preparing an international bond issue, offers to clear all investor claims at 35% of face value, payable over four years. The insurer, which has conduct of the recourse, wants to accept. The group refuses and wants to go to an award. Who is right?
The group's reflex is that of a good lawyer and it is probably wrong here. It reasons on the merits: the case is strong, jurisdiction is settled, an award would be worth face value plus interest, so settling at 35% means giving up two thirds of a claim that will be won. The decisive objection is that an award is not a payment. Once rendered, it meets immunity from execution, this state's attachable assets are few and inconveniently placed, and the group would hold an excellent and unenforceable title. The 35% offer does not exist because the state acknowledges fault; it exists because it needs to clear arrears in order to issue: that is the only window in which this file turns into money, and it shuts the day the issue is placed. Contractually, the conduct clause gives the insurer the hand, and the duty of cooperation exposes the group if it obstructs. But its position is not empty, and this is what must be worked: a global settlement concluded by the insurer risks extinguishing the claim over the 4 million uninsured as well, without its consent and without separate consideration. That is the question that should have been settled in 2021, in the settlement clause, rather than discovered in 2026 in front of an offer that expires.
- 01The act that makes the loss covered is an act of sovereign authority, and it is exactly the one immunity protects: cover and recourse run into the same fact.
- 02Immunity from execution decides more than immunity from jurisdiction: an award obtained is not a sum recovered.
- 03A subrogated insurer's recourse depends on a subrogation clause in the applicable treaty, checked at placement and never added afterward.
- 04The value of a recourse tracks the state's need to refinance, and the window shuts once its bond issue is placed.
- 05The conduct-of-recourse clause and the insured's uninsured share can want two different settlements; that is drafted before the loss.