Back in 2021, Martin Bertogg, Head of Property, UKI/MEA at SwissRe wrote: “Our current risk models tend to mean we look in the rear-view mirror too much. The risk landscape is evolving very dynamically. Relying heavily on historic data is creating a blurry picture of current risks. We need to be wary of relying only on model results based on multi-decadal averages when more recent history gives us a very different picture.”  

Instead, we need to recognise that in changing times, historical analysis cannot answer all an insurer’s questions. In particular, it does not give a complete picture of today’s extreme scenarios. We need a broader set of tools to run today's reinsurance book, and Exposure Management (EM) is a powerful additional lens of analysis.  

Let’s dig deeper.

Property represents the largest proportion of the reinsurance market: it is the leader among coverage, premiums, and of course losses; with perhaps the most emblematic event being Hurricane Katrina in 2005, costing the industry some $65BN.

It's no surprise, then, that significant investment goes into ensuring that reinsurers and their clients understand the risks that they are underwriting in great detail. Since the 1980s, they have enjoyed increasingly capable and digitised tools (both internal and from external consultancies and tech partners) which have helped them to model the distribution of losses associated with NatCat events. For ever more zones and ever more nuanced definitions of perils, these tools have provided a probabilistic view of what the losses for a property book might look like, based on parameters about the properties themselves (construction, value, location etc.) and vulnerabilities (weather, geography etc.).

This approach has been hugely important – and will remain so – not least because it is the foundation of effective pricing. Profit depends on premiums, premiums are calculated to cover a particular probability of loss, and those probabilities require modelling.

But as our world changes increasingly rapidly, we need to appreciate more than just the probability of a new Hurricane Katrina. With a diverse book, we need to understand the worst-case scenario – the real physical manifestation of a catastrophe based on quantifiable factors: location, property value etc. - and indeed the amplifying effect of any similar coverage in other books.

Most obviously, climate change means that we are seeing a dramatic increase in significant events. The topline cost of Hurricane Ian in September 2022 was also around $65BN. Admittedly Katrina is a larger sum once inflation adjusted, but black swan or tail events are now alarmingly common. And that means that the predictive value of a probabilistic model is reduced.

This is accentuated by the fact that all probabilistic models also include a number of assumptions. These assumptions change between specialist analysts, and a whole cottage industry has sprung up around categories of specialist regional environmental knowledge and their application to insurers. This makes some aspects of the data feeding probabilistic models a “black box” – not interoperable between reinsurers and therefore making cross-cedant analysis by reinsurers practically impossible.  

Each time a new layer of data is added to a model, aiming to refine it and make it more valuable, it comes with new assumptions - and therefore new complexity and the inherent bias of a well-meaning analyst’s perspective. EM has no interpretative layer: it is easy to understand because it just aggregates fact.

Get a balanced view with Exposure Management

We’re not saying that modelling is wrong or irrelevant. Rather, it is subject to the law of decreasing returns. More effort can lead to more complexity and reduced value; and with more unpredictability thanks to climate change, it should not be the only analysis in town.

We feel that the reinsurance market would benefit from focusing, in parallel with probabilistic modelling, on exposure management: the deterministic analysis of total and real-world exposure to a customisable and easily comprehensible list of risks. With modern tools like Allphins, this can easily be generated for any peril and combination of geography and damage factors or for actual events.  

Because it is based 100% on concrete data which is readily available, it can be completed with simple analytics without the need for hypothesis, giving reinsurers a rapid and reliable analysis of any cedant book (exposure per peril and zone, slice-and-dice analysis of TIV splits etc.). More importantly, cross-book analysis and visualisation is simple; because books can be compared using the same parameters.  

Furthermore, because EM comes without assumptions, adding data increases insight without more complexity. For example, a property portfolio could be overlaid with a new parameter - distance from the coast – which adds to perspectives or aids decision-making without increasing the chance of analysis becoming weaker through guesswork.

‍

Ask a broader set of questions

EM also gives reinsurers plenty of opportunities for rich analysis, all available without data manipulation expertise or a maths degree; and all easily understood from the shop floor to the boardroom:

• Critical live event response for your ‘war room’, using market standard event descriptions, in order to calculate exposure in near real-time when events occur

• Insight to inform cover for zones and perils which have not been modelled

• Easy year-on-year and book-on-book comparisons: stress test your exposure on any combination of multiple criteria (peril, geography, construction, age or even custom factors like distance to shore), all without assumptions or “black box” methodologies which make true comparisons impossible

• Simple visualisations of predictive scenarios, thanks to rendering tools like Allphins

• Get further insights not based on probability which are critical for reinsurers, like worst-case scenario (reinsurers carry the can), clash analysis (essential for assessing multiple cedants’ books) or building deterministic scenarios (e.g. “what would another Katrina actually mean?”) with absolute confidence across any parameter or degree of granularity (e.g. city, county, state).

• Specific analysis: a probabilistic model cannot answer a question like “What proportion of Florida properties across our books are over two stories and more than 20 years old?” Exposure Management always gives a granular, real-world answer.

• Finally, EM is applicable for that category of disaster where institutional and reputational cost are at stake, but occurrences are so rare that predictive models just aren’t reliable: man-made events. The names are etched in our consciousness - Bhopal, Fukushima, Grenfell – but they represent unique circumstances. EM analysis offers a meaningful perspective on these events because analysts can replay them against an existing book (or aggregated across cedants’ books) as realistic scenarios; and so gain a powerful insight into total risk.

Exposure Management is not a replacement for probabilistic modelling, and Cat modellers don’t have to worry about their jobs. In fact, it would be a pretty poor replacement: pricing by exposure management would be ineffective and possibly opaquer.  

But EM adds essential perspectives which improve our ability to understand overall outcomes and compare books. It allows us to identify overexposures and question our tolerances for business scenarios based on absolute certainties. It gives reinsurers a very visible way to execute a business strategy: if you have a preference for a particular geography or peril, and/or certain limits in mind, EM constantly surfaces your book’s alignment to that strategy. And it allows us to put data to work with unlimited granularity; browsing, slicing and dicing data at will.  

Tools like Allphins allow for this sort of analysis at increasing depth without any need for training or previous experience. If anything, EM is another tool in the Cat modeller’s armoury – one which will help them to build a vision and communicate it across the business with greater simplicity, balance and authority.

‍