Rating methodologies

Question 1

Summary (14)

Answer 1

• An appropriate pricing model is critical to the long term profitability and success of a general insurance company. There are many possible ways of calculating premiums, ranging from a simple approach to sophisticated rating models dealing with many different parameters and variables for each rating group
• The methodology actually used will depend on:
  o The class of business being priced
  o The availability of relevant data
  o The market in which the company is operating
• The process may start with a calculation of the pure risk premium (ie the expected future claim amount), before loadings are added to give the office premium. This is the cost plus approach.
• The pure risk premium can be expressed as a nominal amount, but usually as a rate per unit of exposure.
• To price policies appropriately the company needs data that is reliable and relevant. Policy data is needed to calculate exposure and identify characteristics of each risk group. Claims data is required to estimate the ultimate cost of claims. In particular, data by type of peril is required to cost policy options and make further adjustments (eg to allow for atypical events or changes in cover). Some data items are required (eg claim reference, loss date, description and amount) and others are useful.
• If the experience underlying the data includes anomalous events or untypical experience, we should remove claims arising from such sources. Alternatively, we can choose a more typical base period, collect more years’ data or apply an adjustment factor.
• Internal data, where available, is generally more appropriate than external data. However, market claims data, publicly available information (eg areas prone to flood) and competitors’ rates are also invaluable. Reinsurers may help provide this.
• The base period needs to be chosen to balance the conflicting requirements of relevance and credibility. Data from the most recent years may need adjusting for IBNR and unsettled claims.
• An insurer’s own data is likely to give the most accurate indication of future claims experience. If the insurer has enough data (which can be assessed using credibility theory) then they can use their data alone to estimate future claims costs. However, if the insurer does not have enough data, there is a high chance that the mean of the data will be distorted by random variation. In this case, the insurer can look to external data to complement their own data. The weight they place on the external data will depend on relevance of the data to their own experience.
• Data should be split into homogeneous groups for analysis, usually considering claim frequency and average cost separately, over time. In general, the sub-division should be in as much detail as possible, subject to there being sufficient data in each risk cell to allow credible analysis. Small (attritional) and large claims may be considered separately.
• The proposed premium rates must allow for the conditions appropriate to the dates of cover. In particular, unusual features in the base period data, trends that are expected to continue, changes in risk (eg mix, cover, underwriting and reinsurance), claims inflation and environmental changes must be allowed for.
• A loading may be applied for catastrophe or large loss claims. This can be estimated using own data, external data or a sophisticated catastrophe model.
• Further loadings would be applied for:
  o The cost of reinsurance, including trends in this cost
  o Expenses (including commission), eg by applying a simple overall loading or a more detailed approach allowing for the different expense types (such as fixed or variable)
  o The cost of capital (or profit loading), reflecting the underlying variability of the class of business (net of reinsurance) and shareholder risk appetite
• The premium may also be reduced to allow for investment income on the premiums for the period until claims and expenses are paid out. This is more important for long-tailed business.