A simulation of the insurance industry: The problem of risk model homogeneity
arxiv.org
arxiv.org
Regulators across the globe, practically all of whom are wholly ignorant of what they regulate, demand implementations of their tabular formulas which are often highly specific rather than generalized and so become obsolete as quickly as people find gray spots in the formulations and ways to circumvent the intentions of the regulation. Basic it reads more like law than math and it doesn’t work.
That means all costs go to implementing these sub-par models leaving no room for innovation.
In the end, common people will have to foot the bill and regulations will get even worse. Vicious cycle.
Catastrophe risk models, however, shouldn’t be be lumped together with the types of financial engineering/quant models used in capital markets. The latter are generally very theoretical, based on pretty shallow theory and little data. Academia has been spewing out a lot of quant finance models which have little bearing on real market dynamics.
Cat risk models, however, are generally more data-driven and rich in terms of science and engineering I.e. understanding of underlying hazard mechanisms. But, like any model, they still are only approximations of reality.
In the last 20 years, investors have become a welcome source of risk capital which was driven by (a) the increasing need of insurance coverage relative to existing loss-paying capacity within the system (I.e. Need for extra capital to support the growing market) and (b) the fact that insurers and reinsurers seek to diversify their own insurance coverage, By minimizing credit risk.
The fact is these risk models are quite sophisticated and all make use of pretty much all the science, engineering and data available with respect to the hazards in question. They still are limited in their ability to accurately estimate risk and this is taken into account when pricing transactions.
Furthermore, while all firms have access to the same models, many (who employ scientists and engineers) will apply tweaks based on their evaluation of model weaknesses/limitations. Hence, there is an inherent level of diversification in terms of how the models are used to price risk. This is what makes a market.
The objective of catastrophe risk modeling is to estimate, long-term risk as accurately as possible, while knowing full well that there remains a lot of uncertainty.
The regulators, rating agencies and other market watchers are actually ill-equipped to assess catastrophe model risk.
Advanced epicycles in the guise of macro-economic models.
I did research in that field for about 7 years and at that time the focus was mainly on the paper that was to be published.. it is even a win to have the source available nowadays.
Think of it like an ecology. If there's some firms being on A and some on B, the catastrophe selects for one set. If everyone is the same they all die.
In the extreme case, if all firms use a single model that underestimates the risk of an event then that event occurring poses a threat to the entire industry.
With more models, each making different errors, only parts of the industry are exposed to each error. The more models, the smaller the number of affected firms and with reinsurance between firms the risk to the whole industry is much smaller.
There can be a lot of modeling homogeneity without lots of unanticipated risk (e.g., natural disasters in disparate geographic regions).
Any fans of Fooled by Randomness, Black Swan, etc.?
One of his main ideas, that he has followed up with Mathematics, is fragility to model error in the presence of fat-tails.
If you’re interested in this kind of stuff I’d highly recommend his books.
Anyone with math chops beyond mine checked out his papers? I don’t understand enough of them to endorse or critique.
It's all nice and dandy to say "you have to be anti-fragile, expect the unexpected, yadda yadda yadda...", but putting it into practice is extremely difficult. I don't know of a single institution who did so with an audited track record. Sure, there are (were) many tail event specialized funds, but these will not be providing regular income as one would expect from most funds. Rather, they'll tend to constantly bleed money (basically an insurance premium), and then pay out larger sums when the "midden hits the windmill".
Even looking at something like 2008, it is often cheaper to just accept the occasional drawdown and weather the storm rather than buy protection against it, because of the amount of headwind your investment will face in terms of the insurance premiums.
Books came after
Why exclude when reinsurance can deal with them fairly easily? The insurance company sells bonds that only pay when the massive, correlated event doesn't occur. Financial markets like those, because while they may be 'correlated' in a sense, they are still largely uncorrelated with global growth trends - they're free diversification from their POV. And the "wisdom of the masses" takes care of the modeling - the bond price tells you (and the company) how likely the massive event is expected to be. No added model risk!