Priya has just joined Kestrel Reinsurance Pte. Ltd. as a cat modelling analyst. Kestrel writes property catastrophe reinsurance across Asia-Pacific, Japan earthquake, typhoon across Japan and the Philippines, Australian cyclone, and growing exposure to flood across Southeast Asia.
On her second day, the chief underwriter puts a one-page summary on her desk. It is Kestrel's current cat exposure position, a table with two columns: RDS scenario and PML estimate. Japan Earthquake: USD 340M. Japan Typhoon: USD 180M. Philippines Typhoon: USD 95M. Australia Cyclone: USD 62M.
"This is the first document I look at every Monday morning," he says. "Before I accept any new risk, I need to know whether it moves one of these numbers in a direction I don't like."
Priya recognises the acronyms from her reading. But she realises she does not yet understand what these numbers actually represent, or why they matter more to a reinsurer's chief underwriter than the full EP curve output she has been trained to produce.
The Probable Maximum Loss is the largest loss a portfolio is likely to sustain from a single catastrophe event. It is not a probability statement in the precise sense of the EP curve, it does not say "there is a 1% chance of exceeding this." It is a risk management concept that asks a simpler, more operational question: what is the worst plausible single-event outcome for this portfolio?
The "probable" in PML is deliberately imprecise. Different organisations and different contexts define it differently. In some frameworks it corresponds to the 1-in-100 year OEP loss. In others it corresponds to the 1-in-250 year level. In others still, it is a judgement-based scenario estimate rather than a probabilistic output at all. The imprecision is a feature, not a bug, PML is designed to be a practical risk management tool, not a mathematically rigorous metric. What matters is that everyone in a conversation is using the same definition.
PML is most useful as a portfolio stress test, not a pricing metric. While AAL drives premium calculations and the EP curve drives reinsurance structuring, PML answers a different question: if the worst plausible single event occurs, can we pay? For a reinsurer like Kestrel, this translates directly into capital adequacy, the PML estimate for each major peril tells the CRO whether Kestrel has enough capital to survive that scenario and continue operating.
When a senior underwriter or CRO quotes a PML figure, always clarify which definition they are using. A Japan earthquake PML of USD 340M means very different things depending on whether it is the 1-in-100 year OEP, the 1-in-250 year OEP, or a judgement-based scenario estimate. The number is only meaningful with its definition attached. In practice, Lloyd's uses the RDS framework to provide a standardised definition that removes this ambiguity.
There are two fundamentally different approaches to estimating a portfolio's catastrophe exposure. Everything covered in Lessons 3.1 and 3.2, EP curves, OEP, AEP, AAL, belongs to the probabilistic approach. The RDS framework introduces a parallel deterministic approach. Understanding both, and why both are used together, is essential for working at Lloyd's or any major reinsurance market.
Define a specific event, a named storm, a defined fault rupture, a historical analogue. Estimate the portfolio's loss from that event. No probability is attached to the scenario itself. The question answered is: "What do we lose if this specific thing happens?"
Strengths: simple, auditable, comparable across portfolios and companies. Everyone can understand the scenario. Weaknesses: does not tell you how likely the scenario is, and is sensitive to the choice of scenario.
Run the full stochastic catalogue across the portfolio. Generate the complete loss distribution. The question answered is: "What is the probability of exceeding any given loss level?"
Strengths: captures the full range of possible outcomes and their likelihoods. Weaknesses: requires complex modelling, is sensitive to catalogue assumptions, and is harder to explain to non-practitioners.
Lloyd's uses both approaches together. The probabilistic output (EP curves and AAL from the CMR submission) feeds the capital model and sets the mathematical basis for regulatory capital. The deterministic output (RDS scenarios) provides a standardised, auditable, cross-market comparison of peak single-event exposure. Neither approach alone is sufficient, together they give a more complete picture of catastrophe risk than either provides individually.
The Realistic Disaster Scenarios are a suite of specific catastrophe events defined by Lloyd's that every syndicate must estimate its exposure to. They are deterministic scenarios, each RDS describes a specific event with defined characteristics, and the syndicate calculates how much it would lose if that event occurred. No probability is attached to the RDS itself. The question is purely: "What does Halcyon lose if the Japan earthquake RDS event occurs?"
The word "realistic" is carefully chosen. The RDS events are designed to be severe enough to constitute a genuine stress test, not the average bad year, but a major catastrophic event that has historically occurred or is physically credible. They are not designed to represent the absolute worst imaginable outcome. Their purpose is to provide a standardised, consistent stress test that allows Lloyd's to aggregate exposures across all syndicates and assess the total market's capacity to absorb a major event.
Lloyd's periodically reviews and updates the RDS suite to reflect current market exposures and emerging risks. The core scenarios include:
A major hurricane making landfall along the US Gulf Coast, the single largest potential source of insured catastrophe loss globally.
A major hurricane making landfall in Florida, a densely insured coastal state with the highest hurricane wind exposure concentration in the US.
A major rupture on the southern San Andreas or nearby faults, the largest potential single earthquake loss in the developed world.
A major subduction zone or inland fault earthquake affecting the Tokyo metropolitan area, Japan's dominant cat peril given its insured asset concentration.
A major extratropical windstorm tracking across Northwest Europe, historically the largest source of insured property losses in the European market.
A major typhoon making landfall near Tokyo, Japan's secondary cat peril and a significant concern for Asia-Pacific reinsurers like Kestrel.
A major aviation event, a mid-air collision or large aircraft loss, producing significant aviation liability and hull losses.
Offshore energy, marine catastrophe, and other specialty scenarios. Lloyd's also publishes emerging risk scenarios covering cyber and pandemic as its market exposures evolve.
For syndicates operating at Lloyd's, the RDS scenarios are not merely a risk management tool, they are a regulatory obligation. Every Lloyd's syndicate must submit its estimated loss for each RDS scenario to Lloyd's on a quarterly basis through the Catastrophe Model Return (CMR). This submission uses approved vendor model outputs in a prescribed format.
The CMR serves two purposes simultaneously. Internally to each syndicate, preparing the CMR forces the cat modelling team to run the full portfolio through approved models on a disciplined quarterly schedule, maintaining up-to-date exposure assessments. Externally, Lloyd's aggregates CMR submissions across all syndicates to assess the total market's exposure to each RDS scenario, understanding whether the Lloyd's market as a whole has adequate capital to pay claims if the Japan earthquake RDS event occurred tomorrow.
The distinction between the CMR and everyday model runs matters. A cat modelling analyst might run dozens of model configurations in a given week, different vendor settings, different exposure cuts, different policy structure tests. The CMR is a specific standardised run: approved vendors, approved model settings, the specific portfolio of in-force business, the prescribed RDS scenarios. It is a regulated deliverable with a hard deadline, not an exploratory analysis.
As a cat modelling analyst at a Lloyd's syndicate, the CMR submission dates will be among the most fixed points in your calendar. Missing or submitting incorrect CMR data has regulatory consequences. The practical workflow, which model versions to use, which settings Lloyd's has approved, how to handle mid-quarter portfolio changes, is something you will learn in your first weeks on the job. What matters at Foundation level is understanding why the CMR exists and what it is measuring: the syndicate's peak scenario exposure, estimated on a standardised basis that allows Lloyd's to aggregate across the market.
The RDS scenarios represent severe but realistic events, credible extremes, not absolute maxima. A more severe earthquake than the Japan RDS is physically possible. A more destructive hurricane than the Gulf of Mexico RDS has occurred. The RDS suite is designed to be a consistent, plausible stress test, not to bound the absolute worst case, which is unbounded in principle.
They may be defined to coincide, some organisations define PML as the 1-in-100 year OEP, but they are conceptually different. PML is a risk management concept that can be defined at any return period or as a scenario-based estimate entirely outside the EP curve framework. Always ask which definition is being used.
The CMR is both a regulatory obligation and a genuine portfolio risk management output. The discipline of running the full in-force portfolio through approved models on a quarterly schedule, against standardised scenarios, is substantively different from, and complementary to, individual risk assessment runs. CMR results often surface portfolio-level accumulation issues that individual risk runs miss.
A low RDS PML means the portfolio is not heavily exposed to the specific scenarios Lloyd's has defined. It says nothing about exposure to perils and geographies not covered by the RDS suite, emerging markets, novel perils, or secondary perils that are not yet represented. A portfolio with low RDS PMLs but heavy exposure to non-modelled risks can be far more vulnerable than its RDS figures suggest.
The largest loss a portfolio is likely to sustain from a single catastrophe event. A risk management concept rather than a precisely defined metric, its meaning depends on the definition used. Always requires its return period or scenario basis to be stated explicitly.
A specific defined catastrophe event used to estimate portfolio loss, without attaching a probability to the scenario itself. Answers "what do we lose if this happens?" rather than "how likely is each loss level?" The RDS framework is deterministic.
A suite of specific catastrophe events defined by Lloyd's that every syndicate must estimate its exposure to. Designed to be severe but realistic stress tests, not absolute maxima. Used for internal risk management and external Lloyd's market-level aggregate monitoring.
The formal quarterly submission every Lloyd's syndicate must make to Lloyd's, reporting cat model loss estimates for prescribed RDS scenarios and return periods using approved vendor models. A regulatory obligation and a discipline that forces regular portfolio-level exposure assessment.
The most extreme loss considered credible for a portfolio under a worst-case but physically plausible scenario. More severe than a typical PML definition. Used in some capital frameworks and engineering assessments as the upper bound of scenario stress testing.
The maximum premium income a Lloyd's syndicate is authorised to write in a given year. PML and RDS figures are frequently expressed as a percentage of stamp capacity, providing a normalised measure of catastrophe exposure relative to the syndicate's size and premium base.
Five questions · 4 of 5 correct (80%) to pass