Smaller slices of PI

After years of bleeding red ink, professional indemnity insurers may at last be returning healthy profits - but that does not mean they are making the most of their opportunities.

Hard markets tend to disguise flaws, only for them to be exposed at a later stage. To make the most of the good times and remain in the black when the cycle eventually turns, underwriters need adequate systems to rate and monitor the business they write. But experience suggests that too many PI insurers have not understood the dynamics of their portfolios.

As a result, they are unwittingly subsidising their weakest market segments and failing to take full advantage of their strongest.

One-way not the right way?

The problem from the rating viewpoint is that analysis is not performed in sufficient detail, usually using too few rating factors and without consideration of how these factors interact with each other. This should concern not just direct insurers, but also those using coverholder or managing general agent arrangements, where there has traditionally been a scarcity of timely data. It is no coincidence that Lloyd's continues to issue strict guidelines for the control and monitoring of MGAs' arrangements.

Rating is normally performed by modelling the claims using 'one-way' tables. This is where claim numbers and amounts are summarised against exposure for each rating factor individually (see Fig 1). But this approach can be misleading. Different mixes of business and interaction between rating factors can distort the results. For example, a larger company may have a higher retention: so are differences in claim frequency the result of their size, or their retentions?

The inappropriate use of one-way tables, and the associated failure to subdivide the rate book into the appropriate segments as defined by rating factors, has many adverse consequences. Some segments are priced too cheaply and are, therefore, unprofitable; other segments are priced too highly and the insurer risks losing business.

This makes it possible for insureds and competitors alike to take advantage of inconsistencies in PI insurers' rate books. Furthermore, their underwriters can only guess at the statistical variability between the key factors that influence the claims within the business they write.

The solution is to price business at the 'micro' level to expose relative claims experience between different segments, as is already common in personal lines insurance. This is done using generalised linear models, where a regression model is fitted to all rating factors simultaneously.

The effects of each factor can then be isolated. Claims frequency and claims severity are modelled individually to provide greater understanding of the processes underlying the risk premium per policy (see Fig 2). This approach has the additional benefit of simplifying over-complicated rating factors by grouping them and also removing any irrelevant factors. Hence the process will 'clean up' the rating factors already in use. A further advantage is that GLMs provide a more rigorous statistical framework from which to base rates. They also allow the exploration of combinations of different rating factors.

Another point in relation to pricing is that PI claims need to be considered in size bands. The issues surrounding these claims have to be dealt with separately and the risk premium allocated to the appropriate sector of the portfolio. For example, the risk premium from basic claims is attributable to an individual rating factor level, while the large claims cost can be smoothed over a broader range, perhaps involving only a few rating factors. The cost of 'mega-claims' is then spread across the entire portfolio.

Several actuarial software packages are now available to perform such analyses quickly and efficiently. The next step is to identify the profitable and unprofitable segments of the book and determine potential changes, enabling the premium rating structure to be adjusted accordingly. Once again, software packages can assist with such 'impact analysis'.

One advantage of this methodology is that it does not require the insurer to collect vast amounts of additional data from potential insureds, since most of the information for these additional rating factors is collected within the current quotation systems.

Attention to claims

Inadequate monitoring, including claims reserving, can compound the problems stemming from a poor rating structure. Claims are usually not monitored often enough, or not at the correct level of detail. Reserving should be analysed at as detailed a level as the rating process, and conducted on at least a quarterly basis. This will prevent problems within the account and help PI insurers to identify pockets of unprofitability. Conversely, if they fail to do so, business written could be making losses for a sustained period before the truth is finally discovered.

A change in expected claim amounts, and hence risk premium, can arise from many areas, ranging from segment-specific issues to economic factors affecting all segments. These trends can be hidden within data, but detailed monitoring will bring them out. To achieve this, it is imperative that a departure from the expected loss amount is identified early so the rating plan can be adjusted before losses escalate.

Although some of these processes may seem technical, the message is simple.

The greater understanding underwriters have of their portfolios, the greater their competitive advantage. The disciplines outlined above can have a direct impact on your bottom line - whatever the state of the market.