The last form in this specialization is the most radical, and it answers the course's question with the only answer that points to nobody outside: the risk does not leave the group. A reinsurance captive is an insurance or reinsurance company owned by an industrial or financial group, whose purpose is to carry all or part of its parent's risks rather than cede them entirely to the market. It is neither a disguised optimization scheme nor a curiosity: it is a genuine risk carrier, licensed, capitalized and supervised, which happens to belong to the party whose risks it insures. The whole difficulty of the subject sits in that last clause.
The circuit is worth holding in order, because it has three parties where one would expect two. The group insures with a conventional insurer, the fronting insurer, which issues the policies. The fronting insurer then cedes those risks to the captive, which keeps the retention and in turn buys market reinsurance for the peaks. The captive is therefore never in contact with the policies: it sits at the second tier, reinsurer of the company that insures its own shareholder. And it buys reinsurance itself beyond what it wants to keep, which makes it an ordinary cedant toward the market. One and the same euro of risk can thus cross three balance sheets before settling anywhere.
The fronting insurer is there for one reason, and it is regulatory. Fronting is the arrangement by which an insurer licensed in a given territory issues a policy and formally assumes responsibility for it toward the insured, while ceding most, or even all, of the risk to a reinsurer or captive that lacks the license required in that market. The fronting insurer lends its license and its signature for a fee. It thereby answers two concrete needs: letting a group place its risks in its own captive while meeting local compulsory insurance obligations, and letting a reinsurer reach a market where it is not licensed. A group present in thirty countries needs local policies in thirty countries, and a single captive cannot issue them.
What a group seeks there breaks down into four things, none of them mysterious. Funding predictable losses internally, for which paying a premium to a third party amounts to paying your own claims plus an intermediary's expenses. Smoothing results by accumulating the surpluses of good years. Reaching the reinsurance market directly, often cheaper than direct insurance because it deals with professionals and carries lower acquisition costs. And managing treasury and tax within the applicable rules, the point the caricature remembers and in practice the most tightly framed.
In return, a captive imposes two things that reserve its use to groups of a certain size, and that make it a question of threshold rather than of preference. It requires own funds to be tied up against the commitments retained, capital that leaves the group's industrial uses. And it brings full prudential regulation, with its requirements on governance, reporting, the actuarial function and controls. What the group saves in premium it must therefore hold in capital and administer, and the comparison between the two routes only means something net of the cost of that capital and of those running costs. A premium saving quoted without that correction measures nothing.
There remains the point on which the arrangement lives or dies, and it is neither fiscal nor actuarial, it is credit. The fronting insurer remains legally liable toward the insured. It must therefore pay the claim even if the reinsurer or captive to which it ceded the risk were to fail. Its safety rests not on the retrocession contract but on the security it takes against it: deposits, pledges, letters of credit. That is precisely the point a documented incident laid bare. In the Vesttoo affair, falsified letters of credit were discovered in July 2023, and the company filed for bankruptcy in January 2024. Security is therefore worth only what its verification at source is worth, with the institution supposed to have issued it.
One development is worth flagging to close, and it is prospective rather than established. The AlgoPolis paper on the uninsurability of the agentic model argues that the captive could become a mechanism of last resort for highly automated firms unable to transfer their risk to the conventional market, on condition of tying up massive reserves. The idea should be taken for what it is, a reasoned hypothesis and not a market fact, but it lights up this tool's place. When the market declines a risk, three routes remain: do not take it, have it carried by investors who price it on other terms, which is the subject of the preceding lessons, or carry it yourself and capitalize against it. The captive is that third route, and that is why it belongs to alternative transfer even though, at bottom, it transfers nothing at all.
A global pharmaceutical group sets up a captive in 2026 to carry the first 20 million euros of annual cyber losses. Its fronting insurer issues local policies in every country where the group operates, then immediately cedes the risk to the captive. The captive then buys market cover above 20 million for peak losses. The arrangement cuts the external premium by about 30%. The chief risk officer presents it to the audit committee as "taking our cyber risk out of the market, for 30% less". What must be corrected in that sentence?
The facts are accurate and the arrangement is regular; it is the sentence that is wrong twice, and each error runs in the direction that flatters the deal. The first bears on the word out. The group has not left the market: its captive buys cover above 20 million there, so it remains a buyer for the part of the risk that could actually hurt it, and that is precisely the part whose price hardens when the market turns. What has changed is not the presence in the market but the point at which it begins. The second bears on the 30%, and it is the weightier because it compares two things that do not compare. It is not a saving, it is a shift: the group no longer pays premium on the first 20 million because it carries them itself, so it must now hold in own funds what it no longer pays in premium, tie that capital up outside its industrial uses, and run a company under full prudential regulation. The comparison only means something net of the cost of that capital and of those running costs; it may well stay favorable, but it is not 30%. A third point deserves to reach the committee even though it is not in the sentence: the fronting insurer remains legally liable toward the insureds, so it will require security against the cession to the captive. That security ties up funds in its turn, and verifying it at source with the issuing institution is the exact point a documented incident laid bare in July 2023 in the Vesttoo affair, followed by a bankruptcy filing in January 2024.
- 01A reinsurance captive is a genuine risk carrier, licensed, capitalized and supervised, owned by the party whose risks it insures: the risk does not leave the group.
- 02The circuit has three parties: the fronting insurer issues the policies, cedes them to the captive, and the captive in turn buys market reinsurance for the peaks.
- 03The fronting insurer exists for a regulatory reason: it lends its local license for a fee, which lets a group present in thirty countries hold thirty policies there and a single captive.
- 04What the group saves in premium it must hold in capital and administer: a saving quoted without correcting for the cost of that capital measures nothing.
- 05The point on which the arrangement lives or dies is credit: the fronting insurer stays liable toward the insured and must pay even if the captive fails, so its safety rests on the security it takes.
- 06The Vesttoo affair, falsified letters of credit discovered in July 2023 and a bankruptcy filing in January 2024, showed that security not verified with its issuer secures nothing.