Bureaucracy’s love of silos and task-churning is often pernicious. Clean the data, upload to the model, crank the handle, put the results into tables, populate the Solvency 2 capital model, job done; and repeat next quarter.
There is an epistemic question which should be asked but often isn’t: how close to the truth are cat models’ predictions? Some silos – notably those running cat models and broader capital models – tend to believe the results. Others – notably those in commercial decision-making functions like underwriting – don’t. The silos don’t compare notes. The former does its thing, the latter shrugs its shoulders and gets on with doing business.
I have seen situations where the modelled AAL for one natural peril is higher than the total premium for property insurance in that region. This is striking: it suggests the model and the market differ in their view of average cat losses by a factor of more than three.
Markets can be inefficient and wrong, but surely not that wrong.
I can think of reasons for this error – notably around the observation that most attention is paid to the tail and not the mean, especially for the stakeholders like regulators and reinsurers whose existence depends on not underestimating such downside scenarios.
I’ve been told by underwriters that some cat models for some perils are heavy (as I observe above) and others are reasonable or even light. This is hearsay, I confess, and in their defence let’s observe the massive effort the cat modelling firms put into their work and refining it over time.
But the main point is this: do not take cat models’ AALs as gospel. Always understand and challenge
Cat models may not always be good predictors of the true average cat loss but it does not mean they have no value.
The intellectual property behind the cat model is full of insight. They have thousands of events, each one defining where it affects and to what degree: meteorologists and seismologists have put in thousands of person-years of work. The damage it would cause to certain buildings of certain construction types has been analysed, with civil engineers and construction experts.
In my opinion it is likely that where a model is out of sync with the market, or common sense, it is driven by the probability assumption. As I’ve explained, the vast majority of these events are hypothetical, they haven’t actually occurred. Even for an event that has occurred it is almost perverse to ask what its probability was. In any event, the judgment of likelihood as indicated by this probability assumption is speculative.
So it is a simple matter of recalibrating the results, which is possible with cat modelling output, to deliver a truer reflection of your own judgment of a fair AAL. This has value when you might consider the effects of reinsurance, or of geographically rebalancing your book of business.
I approach dangerous ground. But I am confident that what I say here is both true and practised by smart insurers.
Firstly let me say that Technical Price (TP) must always include the full AAL. Where I’m going in this section is about the strategy of handling Actual Price (AP) for cats.
A large enough cat event will always cause a dislocation in the market. In the following years there is both greater demand for cat cover (as customers have been reminded of how severe events can be) and lower supply of insurance capital (as some insurers may go out of business, and others reassess their appetite nursing big losses).
So: lower supply for greater demand. The perfect recipe for prices to rise.
And indeed they do. It is very clear both anecdotally and from published academic papers (and indeed reinsurance brokers’ published rate indices) that events large enough always cause premiums to rise. For several years the market supports richer pricing, such that on average a large event is “paid back” within 2-4 years. This is true in both commercial lines and personal lines business, wherever I have seen analysis of the effects of cat events.
What this means for the strategy of pricing for cat risk is that it is reasonable to accept AP:TP ratios a little below 1 in normal times. If there is not a cat, the allowance for cat in the AP (albeit below the AAL) will still generate realised profits. If there is a cat, you take a material loss this year (this is the point of insurance!) but you stay in the market and for several years thereafter you demand AP:TP ratios above 1.
Over the cycle, you should ensure you cover the AAL on average. But holding a pricing standard that says all business must cover the full AAL each year will mean you write far less than you might otherwise do and thus make far lower profit over the long run.
While some sophisticated insurers do this, and some others also do it less explicitly, I do think that it is a widespread fallacy that full annual cat loads should be covered each year.
And back to that observation that some insurers exit the market after a large event: I cannot think of a more irrational strategic play. “RATES ARE ABOUT TO RISE”, I want to scream at them.
This doesn’t look like a fallacy at all. This is what reinsurance is for! And it is certainly true that year-on-year the loss absorbing nature of reinsurance will take the big hits.
But in the long run all reinsurance does is smooth your results. The effect I discussed in Fallacy 3 above applies in a more “out of the money” way to reinsurance. A truly big event will lead to recoveries for you but the reinsurers will subsequently put their rates up. So while you would benefit from rising retail prices, you will also be paying more to cede your exposures.
Reinsurance is a smoothing mechanism. It is of value for the stability of returns. And the ability to continue trading rather than run out of money is crucial! But in the long run reinsurers will not lose money, and so in the long run smoothing is all that reinsurance buys you.
There has been a lot of speculation recently around the use of AI in insurance and how this signals the death knell for Underwriters. Having mopped up the coffee that I spat out and checked my life insurance provisions, this got me to thinking. Surely not all underwriters? But if not, then who?
The ineffable question of the “effs”. What should a business aim for – and more specifically, for us, the underwriting function in an insurance business – efficient operations or effective decision making?
We have all heard of the boiling frog syndrome which says: if a frog is put in boiling water, it will jump out to protect itself. However, if you place a frog in cold water and slowly turn up the heat, it won’t notice the subtle increase in temperature and be boiled alive.
Game theory aims to shed light on outcomes where multiple actors operate. Dry academe often assumes “ceteris paribus” even in social sciences, but in the real world when we are trying to anticipate the outcomes of decisions one actor makes it is essential to consider the possible and likely responses of other actors…
In various parts of our lives, we are examined and tested to grade our ability and knowledge, be that driving tests, ‘A’ levels, or piano exams.
And so it is with the insurance industry. Boards and regulators are increasingly pushing for demonstrable oversight. This is especially true in the field of Delegated Authority (DA)…
Managing General Agents (MGAs) provide insurers a broad opportunity for profitable growth. Operating in niche sectors with lean efficient models and providing customers with products that they often find hard to come by in the mainstream. All of this is wrapped up with the relevant expertise and providing insurers with access to markets that…
As 2022 was drawing to a close, insurers across the UK were starting to look at their results, keeping their fingers, toes and everything else that they could, crossed in the hope that December would fizzle out in a whimper.
But alas it was not to be as temperatures plummeted and the industry suffered…
What use statisticians and actuaries when the world is being transformed by modern data-analysis techniques? What hope for clunking humans in the era of Big Data (Facebook sells data which influences US presidential elections), of Artificial Intelligence (China can identify disliked racial groups through automatic facial recognition technology)…
In part 1 of this triptych on insurance pricing we looked at Technical Price (TP): what must be covered over the long run to make reasonable profits. I finished that note by considering TP in the context of the market. Here, I take this theme further and explore more how to think strategically about pricing.
This note covers the fundamentals: the three building blocks of a technical price and a note on the role technical price plays. It is followed with a deeper look at pricing strategies. Finally we consider a knotty topic, cat(astrophe) pricing, in particular the use of cat models to inform the price.
On 26th September 1983, at the height of the Cold War, Soviet missile monitoring picked up apparent signs of five nuclear missiles launched from the US. Retaliation was the default planned reaction.
Charles Dickens opened A Tale of Two Cities with these lines, contrasting calm enlightenment London with turbulent revolutionary Paris: “It was the best of times, it was the worst of times…
Calibrant was established to bridge a gap. The gap is that between MGAs and their carriers. And it finds many forms.