By Dr Kevin Somerville
Ah, rating tables – convenient, elegant and simple. They’ve been used for decades and even a trainee underwriter can use them. So much easier to understand than those calculators and so much easier to compare the offerings in the various underwriting manuals and complain that ratings are too high or low… Besides there’s the current mantra in the market: simplicity. Tables have the virtue of being as simple as we want them to be.
Yet there’s a problem that refuses to go away: actuarial risk pricing. Actuaries price for the risk to multiple decimal places on a product-by-product, market-by-market basis. By contrast, underwriting manuals (outside of North America and its preferred lives model for life insurance) have tables based upon +25% rating steps that can be used everywhere and anywhere. Some allowance might be made for European/Asian build and some other differences, but that’s about it.
Tables can’t do it for us anymore and it could be argued that they are inherently unfair. Consider the different pricing margins, disease patterns and treatment outcomes from country to country and from region to region. A +50 rating for life cover applied for myocardial infarction in India produces a different absolute increase in event rates to that in the UK because of the differential in pricing and disease patterns and outcomes between those countries. The application of the same rating tables in different markets will produce absolute differences that can be substantial and not evidence-based. If the margins are tight, as in the UK, then the drive is to risk-rate everything and anything that moves; not so in regions where there’s some leeway.
For critical illness the issue is more marked with the number and type of diseases differing from product to product, particularly since partial payments and stand-alone cancer products have become more common. In addition, table ratings can’t deal with proportionate influences, eg a 20% treatment improvement, as tables deal in absolutes and are crude with large steps in ratings which are accentuated into old age. We are being admonished to treat our customers fairly but for many applicants tables won’t do this. Tables do not cope well with the variety and complexity of individual differences.
So what’s to be done? Calculators offer at least a partial and perhaps complete solution. They have a better ability to adjust for individual and regional differences than rating tables. Not transparent enough? Just what is transparent about tables except that the structure can be easily seen? Does any underwriter know how those +50s etc are derived? And does transparency of process really matter that much? Big data is characterised by black box-derived associations and risk calculations, and it seems that the risk assessment world is pinning its hopes on and moving in this direction. Data analytics with its confidential company algorithms will never be entirely transparent for any pricing system. Furthermore, by contrast with tables, calculators can deal with the interactions between risk factors and impairments and calculate the risks using market-specific event rates, background data and relative risks and return an integer rating without the 25 band limit restriction for each input. At the least, partial transparency can be achieved by showing how the inputs influence the outputs (ratings).
Is there an alternative to calculators? Consider flat extras as are used for most cancer life ratings. These are independent of the base rate although they assume that the excess death or event rate from an impairment is much the same globally. However, flat extras are only applicable in certain situations.
Rating tables have had their day, useful while they lasted but no longer fit for purpose. RIP rating tables?
Dr Kevin Somerville is an independent medical consultant with many years’ experience working with major reinsurers.