Best estimates for Solvency II

Best-estimate technical provisions in Life and P&C

In Solvency II, the quantitative requirements for insurance companies are based on an economic balance sheet approach. This approach requires liabilities to be valued at the amount for which they could be transferred, or settled, between knowledgeable and willing parties. In particular, the market-consistent value of technical provisions is determined by adding a risk margin to the company’s "best estimate" of the value of their insurance and reinsurance obligations. Realistic assumptions and appropriate actuarial techniques are essential for calculations of company-specific best estimates. A best-estimate view will also be required for the next quantitative impact study, QIS4.

A great deal of technical terminology is used by practitioners in this area. This newsletter seeks to summarise – in an understandable way – some aspects of deriving best-estimate assumptions for claims, as future claims cash flows are an important element in the valuation of insurance obligations.

We focus on how best-estimate technical provisions can be derived for P&C insurance and important factors relevant for deriving best estimate mortality and morbidity rates in life insurance.

next page »

Best estimates in P&C insurance

Technical provisions

Non-life insurance companies deliver a service. For a defined premium an insurance company undertakes to pay an (unknown) amount, if an event occurs. During the term of a policy, as premiums are earned, a loss might or might not occur. If a loss does occur, it is reported to the insurer, which pays (or declines) the claim.

The stages in this process are spread over a period of time but the insurance company needs to answer the following questions as soon as the contract is signed:

• —
  How many unknown claims should be accounted for?
• —
  Does the company have sufficient financial resources to pay known claims?

These potential future payments are accounted for by the "incurred but not reported" (IBNR) reserves, which, together with the individual claims reserves (i.e. reserves that are posted for each single claim), form the "technical provisions" or "best-estimate reserves" for claims. In Solvency II, the term "technical provision" and in IFRS the term "provision" are used to denote the traditional "reserves". We will use "technical provisions" to be consistent with Solvency II.

Economic view of technical provisions

There are a number of reasons why insurance companies estimate technical provisions on an economic basis. For example, it is useful for measuring profitability and supports the pricing process. It is also necessary for the production of their financial statements in accordance with accounting principles and supervisory regulations. While accounting regulations frequently require prudence in technical provisions, both IFRS and Solvency II favour an economic view in the balance sheet. Thus, assets and liabilities are accounted for at their current market value using prices that can be monitored or other appropriate methods.

Estimating technical provisions for (incurred) claims, in particular the IBNR ("Incurred but not reported") reserves, is a difficult task. The time horizons in which claims will be paid are unknown and can be very long (e.g. in casualty business). Therefore, companies usually use statistical methods and actuarial expertise to project past experience in order to estimate the ultimate costs generated by an expired insurance period. Although practitioners use a variety of methods, most involve similar contracts being analysed at an aggregated level using so-called claims triangulation. However, the most common method of triangular projection is the Chain-Ladder Method, which is explained in the following example.

Let us suppose that business in a particular line was written from 2006 and the payments for claims arising in the 2006 accident year (AY) were €50. In 2007, €25 was paid out in respect of claims originating in 2006 and €70 for claims originating in 2007.

Thus, the cumulative payments for the year 2006, after one development year (DY), were €75. If we assume that there will be no further payments relating to 2006 after two years, then the experience of the 2006 accident year can be used to estimate the cumulative payment for the 2007 accident year. On the assumption that the ratio between the first and the second year is constant,
cumulative payments after 2 DYs for AY 2007/cumulative payments after 1 DY for AY 2007 = 75/50
cumulative payments after 2 DYs for AY 2007 = 70 * (75/50) = 70 x 1.5 = 105.

AJ/EJ	1	2
2006	50	75
2007	70
AJ/EJ	1	2
2006	–	1,50
2007	–
CL-Factor	–	1,50
AJ/EJ	1	2
2006	50	75
2007	70	105

Figure 1: IBNR estimation using aggregate claims triangles

Actuarial expertise

The assumption that "there will be no further payments after two years" (in actuarial terms "no tail") is a critical one. In long-tail businesses, where claims development may evolve for as long as 30 years, the available claims triangles often do not provide information on the complete loss development horizon. Similarly, data availability for older accident years is frequently not representative of more recent accident years. In such cases, actuarial techniques are used to complete the missing time horizon in the model.

The example in Figure 1 shows, in a very simplistic way, the general principle used to estimate technical provisions. In reality, the transfer of past experience to the future is much more complex due to uncertain influences of various economic factors on the payments that actually become due. For example, in Europe many countries have rules to prevent accident victims falling into poverty and compensate people for what they might have become if the accident had not happened. It is, of course, difficult and requires a lot of experience to forecast such future developments and match them with investments. If business is written in another currency, the exchange rate can change. The most important issues that influence the development of claims are:

• —
  the time horizon (tail)
• —
  inflation
• —
  currency effects
• —
  legislative changes in the line of business (e.g. Workers Compensation)
• —
  cultural changes (e.g. attitude towards litigation)

These are factors outside a company which affect the timing and size of claims. However, it is also important to understand the internal procedures in claims departments.

• —
  How are individual technical provisions set?
• —
  How long does it take to set the technical provision after the reporting of a claim?
• —
  When are technical provisions changed?
• —
  When and why do they review technical provisions?
• —
  Have there been any changes in claims settlement practices?

The (individual) technical provisions are set by the claims department on the basis of subjective appraisals. However, using incurred claims, i.e. the total amount of claims (instead of paid claims only), to produce an IBNR-projection, might provide a more reliable estimation of the technical provisions required. The "Munich Chain Ladder"¹ projection method uses both information sources to add credibility to the estimates.

Discounting "best estimates" in Solvency II

In addition to the procedures explained above, in Solvency II (and IFRS) the "tech-nical provisions" need to be adjusted for the time value of money, i.e. the present value must be calculated. This adds a new factor to the equation which requires expertise and a good data base. In the current QIS4 exercise, the European commission will permit the use of proxies² (i.e. simplifications of best-estimate calculations) to lower the barriers for participation. However, companies are ex-pected to build up the necessary expertise in time for the introduction of the Solvency II regime.

---

¹ http://www.aktuar.de/download/MunichChain.pdf (PDF)
² http://www.ceiops.eu/media/docman/public_files/publications/submissionstotheec/CEIOPS-DOC-23-07%20QIS4%20-%20Technical%20Specifications%20%20Rev.pdf (PDF)

« previous page  |  next page »

Best-estimate rates in Life

Company data

Homogenous risk groups

Crude best estimate rates

Best estimate rates

Figure2: Process for deriving bestestimate rates in Life

Company-specific best-estimate mortality and morbidity rates form the basis for projecting realistic claims cash flows. Apart from Solvency II requirements, such best estimate rates are used for several other tasks within a life company, e.g. Embedded Value, IFRS, monitoring, pricing, reserving and profit testing. Most of these tasks are closely linked to the "actuarial control cycle", which is used by life insurers to monitor and control their best-estimate rates and hence the expected return on and the embedded value of the portfolio.

Best-estimate rates in life insurance are documented in company-specific mortality and morbidity tables. These tables are derived statistically from historical company data or by combining company experience and market data. Obviously, for this purpose a company’s portfolio should be subdivided into homogenous risk groups. The statistical analysis of these homogenous risk groups produces crude best-estimate rates which are usually quite volatile, depending on the volume of data being used. Hence, actuarial methods (e.g. smoothing techniques and comparison to reference tables) are used for deriving the final best-estimate rates so as to better represent reality.

A number of aspects should be considered when subdividing a portfolio into homogenous risk groups and when deriving or working with best estimate rates:

Risk factors

A risk factor is a characteristic statistically associated with the risk of mortality or morbidity, e.g. age, sex, smoker status, lifestyle, etc. There are many other risk factors. Such risk factors can significantly influence the best-estimate rates and should be taken into account for company-specific best-estimate analyses if the required data is available. Examples are:

Age and sex

Major risk factors for mortality and morbidity are age and sex. For example, it may be observed that mortality in women is lower than in men and often has a different shape. This is illustrated in the figure below.

Mortality table 2002/2004 Germany, age and sex

Source: Statistisches Bundesamt, Wiesbaden, 2005

Smoker status

The level of mortality and morbidity varies with the smoker status. Smoking increases the probability of developing severe diseases. For a best estimate, smokers and non-smokers should be considered separately.

Marital status

It may be observed that married people have lower mortality than single people. The reason may be that marriage encourages healthy types of behaviour and discourages risky ones like drinking, smoking, drug use, etc. Usually marriage also increases family income, which may result in lower best-estimate mortality rates.

Socioeconomic status

Socioeconomic status measures the correlation of income, education and wealth with mortality levels. In a best-estimate analysis it makes sense to consider the following factors (if the appropriate database is available within the company):

• —
  occupation
• —
  income
• —
  sum assured
• —
  postcode

For example, the postcode could be a proxy for socioeconomic status as the population in a particular postcode area is likely to have a similar socioeconomic status.

Method and quality of underwriting

Medical and financial underwriting is an instrument to avoid anti-selection in life insurance. Recently underwritten and accepted lives have a lower level of biometric risk than older business. The length and degree of that selection effect varies with the quality of the underwriting.

For example, policies with high sums assured have to fulfil thorough underwriting standards. This leads to a lower level of best-estimate rates for policies with higher sums assured.

Sales channel

The level of best-estimate rates may vary with the sales channel. For example, it may be observed that business transacted through independent financial advisors has a different level of mortality than that transacted via internet.

Types of product

Best-estimate rates depend on the individual characteristics of a product. Hence, when analysing multiple products the product characteristics are important for arriving at homogenous risk groups. The grouping of products is, of course, subject to the availability of sufficient data. Some product characteristics of interest are type of underwriting, sales channel, policy wording and policy duration.

Data quality and observation period

Obviously the quality of the historical data used to determine the best-estimate rates is essential for their significance. Hence, everything depends on a proper database. Deriving best estimate rates for morbidity is usually more challenging then deriving mortality best estimates as the database for morbidity is often smaller.

The period of data used for the analysis should be long enough for the data to be credible, but not too long, as the results may then be distorted by changes in processes or systems. With regard to mortality, an observation/investigation period of 3 to 5 years would usually be suitable. The end date of the observation period may also influence the results. It should be far enough in the past to ensure that all claims have been reported and recent enough to ensure that the best estimates are applicable.

Trends should be considered when determining best-estimate rates for long-term business. Future best-estimate rates may be different from today’s best estimate rates, e.g. due to better medical care. However, for a mortality trend analysis a minimum of 10 to 20 years of experience is usually required.

« previous page  |  next page »

Summary

There are many factors to consider when deriving best estimates. Both in life and in P&C insurance the crucial issue is the availability and quality of data together with actuarial expertise and professional tools.

In P&C insurance the major challenges are the estimation of the amount, time horizon and inflation of claims payments. On the life side it is essential to cluster the portfolio in homogenous risk groups. However, the more sophisticated a best-estimate analysis in life is intended to be, the less likely it is that adequate data will be available.

Hence, in the context of Solvency II, companies will have to focus on the most important factors and techniques when determining their best estimates – also taking cost and benefit considerations into account.

Services provided by Munich Re

Munich Re assists its clients in all areas of Solvency II. In particular, the PillarOne project combines risk management with the added value of an open source platform for collaboration. The project plan is based on the needs of the participating community, which identified the calculation of best estimates as a Solvency II-driven need. Information on recent developments in the project can be found on the project website at www.pillarone.org.

For life insurance companies, Munich Re provides services in the field of experience studies, assisting its clients in better understanding their portfolio and deriving best-estimate rates. By using this kind of service, life insurance companies can directly benefit from Munich Re’s expertise in risk management and can make an important step forward in preparing for Solvency II.

« previous page