UK: Report for the ABI on Key Correlation Assumptions in ICA for Life Offices - Executive Summary

This report has been commissioned by the Association of British Insurers (ABI) to assist its life assurance member firms in their work to prepare their company’s Individual Capital Assessment (ICA). The report considers how firms may demonstrate the benefit of the diversification effect between key risks and take credit for this in their capital calculation. The terms of reference were agreed between ABI and Deloitte & Touche LLP.

Our report considers only the diversification effects on UK life insurers in general. It does not consider the circumstances of any particular life company or the impact on any particular life company’s individual capital assessment.

In preparing our report, we have relied upon a number of published sources for our analysis. These are detailed in the references section at the back of the report. This is a very significant business issue since the diversification effect may reduce the ICA by between 25% and 50% over what might otherwise prevail.

At its simplest, the effect of diversification is to reduce the total amount of capital an insurer will need because, using an agreed benchmark of probability, we can demonstrate that not all risk events that could occur will in fact occur at the same time. This means that for an insurer managing a diverse range of risks, with appropriate risk management policies in place, the amount of capital needed to support the entire business is less than the figure suggested by the simple summation of the capital amounts required to support each of the risks individually.

Accordingly, the extent to which key risks are correlated is highly significant. Only when two events, or outcomes, always occur simultaneously will the correlation be perfect (or 100%). In most cases there will be a lower level of correlation and in some instances there will be a negative correlation so an occurrence of one type of event will typically be offset to some extent by another type of event.

For insurers, the nature of such correlations is highly significant in determining the level of capital they need to hold. In this report we explore the relationships between the following key risks:

• Market risk and lapse & surrenders.

• Market risk and mortality.

• Annuitant mortality and assured lives mortality.

• Intra-market correlations.

These risks were chosen for their significance to life insurers. They represent areas in which a comparatively greater amount of robust primary data exists and other research has already been undertaken. Where available this work has been drawn upon in the report. Clearly, diversification effects are not limited to these correlations, however for this report we agreed to confine the scope of research to a manageable range of relationships and correlations. We hope that this has allowed us to make a timely contribution to firms’ work in preparing their ICAs.

The report identifies common themes and quotes observed ranges of correlations along with argumentation to support these ranges. For individual firms suitable correlation values will vary and must be determined with reference to the composition of their individual businesses including asset/liability management and the risk profile of the company.

In considering our findings it is important to differentiate between the correlation of risk parameters, such as interest rates and lapse rates, and the correlation between the associated risks, i.e. the actual loss amounts which may not move in the same way. We set out our key findings below.

This is a challenging area to analyse since the factors driving the relationship between market risk and lapse & surrender risks will tend to be product and company specific. Our main findings are that :

• For lapse & surrender rates one might expect to see a strong dependency on product type and design, especially around guarantees.

• Guarantee related effects appear to be limited by the extent of consumer understanding of the value of guarantees. FSA research¹ suggests a poor level of consumer understanding of guarantees and that affordability and product performance are key factors affecting lapse & surrender rates.

• Lapse & surrender rates will reflect the individual characteristics of life insurance firms, for instance the distribution channel, asset allocation, approach to surplus distribution, surrender penalties and MVAs.

• Empirical studies are inconclusive.

• Macroeconomic analysis shows a possible link between lapse & surrender rates and the consumer confidence index; however, there is no direct relationship to either earnings or unemployment rates.

The tables below show the key contributing factors:

Given that many companies have significant blocks of immediate annuity business and are at risk to the effects of increasing longevity this is a key area of study. As part of our investigation we have made the following observations:

• Little explicit evidence of direct correlation.

• Correlation may exist for certain insured populations or arise in extreme conditions.

• Clear link between wealth measured by occupational class and longevity (the link between longevity and capital markets may exist if wealth is directly related to capital markets).

• Strong correlation may arise in extreme conditions such as in the case of a pandemic or a terrorist attack or, in the long term in the case of a lengthy economic recession. This may result in a negative correlation between mortality rates and equity markets, and a positive correlation between interest rates and mortality rates.

• A link between population age structure, longevity and economic trends can be demonstrated, particularly in relation to supply and demand of fixed interest borrowing and property prices. However, these are long term effects which are not directly measurable over a short time horizon.

• Severity and duration of the correlation effects is difficult to predict. An indicative range for this correlation is -10% to +10% but typically taken as 0%. The tables below show the key contributing factors:

For many insurers this is an important area of focus. The effect on capital requirements of the resulting risk correlation will be highly dependant on the relative sizes of the annuitant and assured life blocks of business.

• Annuitant and assured lives mortality rates at the same ages are likely to be highly correlated, but generally annuitants and assured lives belong to different age groups.

• Patterns of mortality improvement may differ between age groups and thus correlation between annuitant and assured lives may be lower due to age profile of the populations.

• The effect of high correlation is to reduce, rather than to increase, overall capital requirement, since insurance risk affects the capital required to support assurance business and annuity business in opposing ways.

• Impact of selective lapse and re-entry of assurance business must be considered as annuity terms are typically fixed and policies are not currently transferable.

An indicative range for this correlation is -75% to -20%.

Key correlations in most ICA calculations are:

Equities with bonds, interest rate term structure, property with equities & bonds, credit spreads with interest rates & equity markets.

The following considerations are important:

• Difference between longitudinal & cross sectional correlations.

• Economic logic requires that yields on bonds with term maturity dates close to one another are closely correlated.

• Yield spreads on corporate bonds and equity markets are expected to be strongly negatively correlated, as wider spreads imply increased likelihood of bond defaults.

• Similarly, there is positive correlation between capital values of corporate bonds and shares issued by the same company.

• Negative correlation between the level of interest rates and equity markets based on discounted cash flow valuations is often proposed but difficult to justify in practice due to the complex interactions between interest rates and dividend estimates, and second order effects relating to the movement of funds between bond and equity markets.

• Property prices may be strongly linked to interest rates. Contributing factors include: consumer behaviour, relationship between commercial property prices and lending, and highly leveraged residential property.

Indicative ranges for these intra-market correlation factors are:

This paper presents a number of strong arguments to support ranges of correlation factors for given combinations of risk. The values chosen by an individual company for its ICA calculation may differ significantly from these ranges owing to the risk profile and general nature of their business. However, it is clear that there are significant diversification benefits which should be explicitly recognised in the ICA calculation.

Further work by individual firms and by the industry as a whole may develop this argumentation further and deliver more specific correlation ranges based on individual company experience.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.