Case Study 2 — The LMX Spiral and the Limits of Passing On Risk

DataField.Dev

Case Study 2 — The LMX Spiral and the Limits of Passing On Risk

A real, public episode in the history of the London insurance market, here told qualitatively to draw out its structural lesson. No precise loss or financial figure is invented; the mechanism is the point. The numerical illustration in the final section is a clearly-labeled constructed example.

Background

In the late 1980s and into the early 1990s, the London reinsurance market — including the Lloyd's market (whose origins Chapter 2 owns) — developed an extensive trade in a particular kind of cover known as excess- of-loss reinsurance on excess-of-loss business, widely abbreviated LMX (London Market Excess of Loss). The idea began soundly enough: a reinsurer that had written catastrophe excess-of-loss cover (§27.4) would, in turn, protect its own book by buying retrocession (§27.7) — excess-of-loss cover on the excess-of-loss risk it had assumed. Spreading catastrophe risk further is, in principle, exactly what the system is supposed to do. The trouble was not the concept; it was what the market built on top of it.

Over time, the London market traded this excess-of-loss-on-excess-of-loss business so actively, and through so many interconnected participants reinsuring one another, that the chain of who-was-reinsuring-whom became genuinely opaque. Participants bought retrocession from counterparties who bought it from others, who — often unknowingly — bought it from the original parties, or from a tangle so interwoven that no one held a clear picture of where the ultimate catastrophe exposure actually rested. Each individual transaction looked like a sensible piece of risk transfer. The system those transactions formed was something no single participant could see.

The insurance / underwriting issue

This is the textbook realization of the spiral described in §27.7 — and it is the perfect complement to Case Study 1, because where Andrew and Katrina show reinsurance working (the market reforming to provide the catastrophe cover the industry needs), the LMX spiral shows the limit of reinsurance: the failure mode that appears when risk is passed on so promiscuously, and so opaquely, that it loops back on the parties who believed they had shed it.

The core issue is the gap between transferring risk and eliminating it. Every participant in the LMX chain had, on paper, reinsured away part of its catastrophe exposure. But the risk did not leave the system — it circulated within it. When a catastrophe (or a cluster of catastrophes) struck, the loss entered the bottom of the chain and began to climb, and because the same loss passed through participants who were reinsuring one another, it was effectively counted and re-assumed multiple times as it cycled around the loop. A modest number of underlying catastrophe events could thereby generate reinsurance losses far larger than the original insured damage, amplified by the very mechanism that was supposed to disperse it. Participants who believed their net exposure to any single catastrophe was small discovered, as the spiral unwound, that they were on the hook for far more than they had ever knowingly assumed — because their retrocessional protection turned out to lead, by a circuitous route, back to themselves.

There is a second, quieter issue underneath the spiral, and it is one this book returns to: adverse selection and information loss along the chain (Chapter 1, §1.4). The further a risk travels from the underwriter who first evaluated it, the less the ultimate holder knows about it. By the time a slice of catastrophe exposure had passed through several layers of retrocession, the party finally holding it often had little insight into the original risks, the original accumulation, or even how many times the same underlying exposure sat in its own book through different counterparties. Opacity is not a side effect of a spiral; it is its precondition.

What it shows

The LMX spiral teaches the single most important limiting principle in this chapter, and it teaches it the hard way.

First, passing on risk is not the same as making it disappear. Reinsurance and retrocession relocate risk; they do not abolish it. The risk must ultimately rest on someone's capital, and if the chain is opaque, that "someone" may turn out to be you, several steps removed, holding magnified exposure you never consciously accepted. An underwriter who treats retrocession as an eraser — "I'll write it and retro it" — has misunderstood the tool exactly as badly as a primary underwriter who thinks reinsurance lets them write anything.

Second, a perfectly-priced individual transaction can sit inside a catastrophically mispriced system. As the §27.7 callout argues, each LMX participant might have priced its own slice competently, using sound methods, and the aggregate could still be a disaster — because the danger lived not in any single transaction but in the connections between them, which no participant's analysis captured. This is the deepest version of the book's model-versus-judgment tension: the limits of pricing individual deals well when the real exposure is systemic, emergent, and invisible to any one party's books.

Third, opacity is itself an underwriting risk. The reinsurance market's most expensive lesson here was about knowing where your risk ultimately resides. A market that cannot trace its own catastrophe exposure has not diversified that exposure; it has hidden and concentrated it. The discipline that prevents a spiral is not a cleverer catastrophe model — it is the underwriting governance to understand and limit aggregate exposure to interconnected risk, to know your counterparties and where the cover ultimately leads, and to decline the seductively profitable retro business whose ultimate exposure you cannot see.

Outcome

The LMX spiral produced severe, well-documented losses for the London market and contributed to a painful and much-studied period of distress at Lloyd's. The market's response, over the following years, reshaped how this business is done: far greater scrutiny of retrocessional programs and the chains they form; a deep wariness of "spiral" exposures and of writing excess-of-loss on excess-of-loss without understanding where it leads; better data and aggregation discipline so that participants could see their true catastrophe accumulation; and a lasting cultural memory in the reinsurance market of what happens when risk is passed around a loop no one can map. The episode became, and remains, the standard cautionary tale taught whenever retrocession is explained — which is precisely why it closes this chapter.

Lesson

The transferable lesson is the natural limit on the power Chapter 27 spends most of its length celebrating. Reinsurance is extraordinarily powerful: it makes large risks writable, catastrophes survivable, and earnings stable. But its power is relocation, not abolition, and relocation has limits. The LMX spiral is what those limits look like when they are ignored. For the working underwriter — primary or reinsurance — the discipline is threefold: know that the risk you cede still exists somewhere and could come back; understand your aggregate exposure to interconnected risk, not just your exposure on each deal; and retain the judgment to refuse business whose ultimate exposure you cannot see, however attractive its price. Some risks are managed not by pricing them more cleverly but by declining to be the last party — or, in a spiral, the unwitting last several parties — holding them.

A constructed illustration of the amplification

To make the mechanism concrete, here is a deliberately simplified, constructed illustration — not a real LMX program, just a toy model of how a loss can be re-counted in a loop:

A TOY SPIRAL — how one loss gets counted many times          [constructed teaching example]

  Suppose three reinsurers A, B, C each write catastrophe XOL, and each
  retrocedes part of its exposure to the next in a ring:
        A ──retro──► B ──retro──► C ──retro──► A   (the loop closes!)

  A single catastrophe causes a loss that pierces A's layer. A recovers from B.
  But B's loss now pierces B's layer, so B recovers from C. C's loss pierces
  C's layer, so C recovers from... A — who is now hit a SECOND time by the same
  underlying event, and the loss begins another lap around the ring.

  The ORIGINAL insured damage is fixed. The REINSURANCE losses generated as the
  loss circulates can be a MULTIPLE of it — and no single party, looking only at
  its own next counterparty, can see that the ring closes back on itself.

The numbers are invented and the structure is stylized, but the lesson is exact: when retrocession forms a loop and no one can see the whole loop, a bounded real loss produces an unbounded-looking reinsurance loss, and the participants who thought they were protected are the ones who pay. That is why §27.7 ends the chapter not on the power of reinsurance but on its limit.

Discussion questions

In your own words, explain the difference between transferring risk and eliminating it, using the LMX spiral as the example. Why does the distinction matter to a reinsurer buying retrocession? (§27.7)
The chapter claims "a perfectly-priced individual transaction can sit inside a catastrophically mispriced system." Explain how that is possible, and why it is a deeper problem than any single mispriced deal. (§27.7)
How does opacity — not knowing where ceded risk ultimately resides — function as an underwriting risk in its own right? What governance discipline pushes back on it? (§27.7)
Connect the LMX spiral to adverse selection and information loss (Chapter 1, §1.4): why does a risk become harder to understand the further it travels from the underwriter who first evaluated it?
Contrast this case with Case Study 1. Andrew and Katrina show reinsurance working; the LMX spiral shows its limit. State, in one sentence each, the lesson each case teaches about catastrophe risk transfer.
The chapter says "some risks are managed not by pricing them better but by declining to be the last one holding them." Give one example of such a risk from this case, and explain why pricing discipline alone would not have protected a participant in the spiral. (§27.7)