In times when the search for comfort in an uncertain environment can lead people to rush for the 'certainty' of mathematics, only to find a naked emperor, TGP investigates the practice of Judgement Rating.
We find that it is:

• essential for modern-day risk assessment;

• a valuable route to profitability;

Insurance Pricing and Data

In these days when we are swamped with information about every aspect of many of the risks that are insured it is maybe as well to remember that it was not always thus. Consider the invention of the motor car and how insurers, with no data to refer to, looked at pricing the risks that this innovation then represented. Borrowing from property and liability insurance practice, underwriters took a rate on value to produce the cost of the own damage risk and a flat price based on engine size to reflect the third party risk. Even as late as 1964 matters had not progressed much further. Then came vehicle grouping, to take account of relative repair costs and weight to brake horsepower ratios; refinements for rating districts (before postcodes) and driver ages were also added.

Nowadays motor pricing may seem like little more than an actuarial exercise, but in the 1960's the idea of involving actuaries in work outside life and pensions was considered very novel. And, incidentally, in those days the business was generally profitable.

Don't Blame the Actuaries

The marginal (at best) profitability of motor lines today is not, of course, the fault of actuaries but of competition. Comforted by the thought of plenty of data and fairly low entry barriers, a relative immunity from catastrophe risks, limited accumulation exposure and a simple reinsurance structure, potential financiers may look kindly at funding a new player. But more money was made out of writing catastrophe risks in the days of scant capacity, even though relevant data was scarce and of questionable value.

Comfort in Information

Data often inspires confidence when it should not. Much information is held on areas of geological liability but it cannot be harnessed to forecast when and where the next earthquake is going to occur or how much damage it will cause. And advanced mathematics may look impressive, especially when written down as lengthy formulae, but can easily create a false sense of security (remember LTCM). Flooding? - just look at the UK in the summer of 2007. Models may have their uses but it is dangerous to place too much faith in them.

Unpredictability and Profitability

While risks that appear to be highly unpredictable may seem the least suited to insurance, they are often the most profitable. Indeed, judgement rating, by its nature, may be applied in cases where the very absence of conventional rating information has given rise to a distress market and the high level of rating that such conditions permit. War risks is a good example (although it needs to be remembered that its profitability can be due to the premiums being effectively tax funded, because of the wish of governments to keep commerce flowing in war zones). And although distress pricing obviously holds considerable profit potential, it needs to be borne in mind that the calculation of supporting capital is most important. If the risks really are very volatile then the consequent need to service a high level of risk capital will demand strong pricing in any event.

Judgement Rating

So what happens when data is either unavailable or so constituted as to be of no real practical value in conventional rate-making scenarios? This is the time to apply judgement rating. Sometimes this is referred to as 'seat of the pants underwriting', suggesting that there is no real science to it at all, but in practice nothing is ever so random. The original marine underwriters in Lloyd's coffee house were also venturers. They would have known which trade routes were more hazardous than others, the reputation and experience of individual captains and the seaworthiness of their ships. These features would have been reflected in their pricing - and they must have been right more often than they were wrong since the market is still with us.

Amenability of Risk to Statistical 'Taming'

Some types of risk may not, by their very nature, be capable of being brought within acceptable confidence limits of predictability. It may be safe to say, for example, that the risk of civil war in Zimbabwe is significantly higher than in France, but how much higher in quantitative terms? And if civil war hits either country what form will the damage take, whom will it affect most and what will be the ultimate outcome? We may be asked to quote for loss of business arising out of insurrection, terrorist damage to certain property, murder or sequestration of particular individuals. It is here that judgement pricing comes into its own.

Never Really Absent

Of course the concept of judgement rating is never really absent. A common example is the insistence on a minimum rate on line. Saying that 'nothing will be written at less than one on line' is effectively saying that nothing will be judged as having less than a one in one hundred return period. There are many other examples to be found in minimum premiums and minimum levels of deductible.

How to Proceed

There are some who hold that judgement rating is too unscientific to be made subject to any type of formal discipline. We would disagree. The very fact that it is acknowledged that someone is making a judgement indicates that the result will not be simply taken off the wall. The key is in having a framework whereby relative judgments may be juxtaposed and validated. This is a different discipline from analysing a ton of data, but it is no less important.

For more information on this fascinating subject contact: John Murray, john@tgpadvisers.com

Newsletter information

This newsletter is issued approximately bi-monthly by TGP Management Advisers llp.
Copyright (c) TGP Management Advisers llp 2008. All rights reserved.

To subscribe follow the link below

www.tgpadvisers.com/newsletter.htm

All comments, feedback and other queries to feedback@tgpadvisers.com