In the second of a series of articles on the systemic risk of modelling, we look at the results of the work so far.
Oxford academic Anders Sandberg has just been through a two-year crash course in underwriting and insurance.
He’s a respected neuroscientist who is part of a project looking at the systemic risk of modelling – in other words what kind of threat is posed by the growing use of automated computer models by the insurance industry?
Harsh lessons were learnt from the banking crisis of 2008. It was one of the reasons MS Amlin helped set up the Systemic Risk of Modelling (SRoM) working party to look at the potential impact on the insurance industry of this increasing reliance.
One of the most interesting findings was that the danger of systemic risk increases when everyone uses the same models. It was one of the significant factors in the banking crisis of 2008.
Back then, the value of property-based derivatives that underpinned the collapse were miscalculated by the models that the financial institutions were using. The lack of diversity in modelling meant that everyone had the same exposure based on flawed assumptions and calculations.
The insurance industry relies heavily on a handful of main players in the modelling sphere. But if the market is to continue to be vibrant and successful, the working party thinks perhaps there ought to be less reliance on this small number of players.
As Anders, senior fellow of the Oxford Martin School (OMS) at Oxford University points out: “Shared risk models can act as a conduit of systemic risk by making different companies behave similarly; model diversity reduces the risk of a large fraction of the market becoming insolvent when an unusually bad disaster occurs.
“It’s a challenge for insurers. On the one hand models are an essential part of the business, and indeed the regulators demand that insurers use models to calculate risk.
“So models are essential, but there is a lack of diversity. Having more models to choose from would seem to be a good way to make sure everyone doesn’t make the same mistake.”
Another eye-catching finding from the research is that very competitive markets have more company and systemic risk, since the incentives push towards higher risk appetite and smaller margins.
While competition in markets is generally considered to be healthy, and good for customers, there is a downside for the continuing health of the insurance and reinsurance markets.
Anders said: “Very competitive markets mean that companies are willing to take bigger risks.
“And if that leads to a lot of companies going bankrupt at the same time, from a social standpoint, this isn’t good for anybody.”
Crucial to the research is how the underwriters interact with the models which they use to help make their decisions.
And this formed a significant part of the work Anders did with the working party.
Anders added: “I spent a lot of time talking to underwriters, and when you do this you discover that moving numbers around spreadsheets is only part of the job. Building relationships, applying knowledge and experience, creating long-term value; these are all a fundamental part of the job.”
While talking to the underwriters Anders was exploring cognitive bias, one of his specialities. Cognitive bias refers to errors in thinking and reasoning that can adversely affect decision-making.
And it’s an area we’ll be looking at in more detail in our next article.