(Re)insurers can improve their returns by analysing the essential relationship between their assets and liabilities. Oliver Peterken, Alan Bulbeck and John Hibbert explain.
Insurance and reinsurance companies face a growing number of challenges from increased competition, as well as greater pressure from shareholders. For chief executives and company management to meet these challenges successfully, it is essential that they fully understand their business and the complex relationships between its many elements.
Over recent years investment income has held a predominant position in the overall return for insurance companies and, therefore, their shareholders. For example, the total pre-tax profit for the UK insurance industry for the underwriting years 1995, 1996 and 1997, was approximately £20 billion, of which the underwriting result was -£3 billion and the investment result was £23 billion. For an industry whose primary purpose is to underwrite risks, it may seem surprising that its profitability is so heavily dependent on investment returns.
That investment returns are so significant in a company's final results, and that such investment is made possible only by the acceptance of liabilities and their subsequent premium, emphasises the crucial relationship between an insurer's assets and liabilities. In some ways, this relationship may seem somewhat paradoxical as an insurer's underwriting of liabilities enables him to purchase a variety of assets, the return on which will mitigate any losses resulting from the original underwriting! This highlights the importance of managing simultaneously the assets and liabilities of (re)insurance companies, and selecting the asset allocation strategy which best matches a company's portfolio of liabilities in order to maximise its overall profitability.
However, the assets and liabilities of companies have traditionally been managed separately, with responsibility for the asset side and the liability side delegated to the investment manager and the underwriting manager respectively. Yet growing shareholder pressure for enhanced earnings with reduced volatility raises the issue of the increase in return which insurers and reinsurers may be forfeiting by failing to see the essential relationship between their assets and liabilities and, consequently failing to adopt a concurrent approach to the whole balance sheet.
The recent advances in computer, econometric and liability modelling have revolutionised balance sheet management, as well as increasing recognition of the links between assets and liabilities. These continuing advances are now making it possible to build models of entire (re)insurance operations, facilitating the simultaneous modelling of both sides of the balance sheet and the subsequent delivery of leading edge asset-liability strategy solutions.The economic factors which affect an insurer's cash flow, both in terms of its assets and liabilities, are the starting point for this type of modelling. These drivers include such factors as inflation, GDP and interest rates.
While the diagram in Figure 1 is simplistic, it does illustrate that both sides of the balance sheet are affected by the same economic drivers and, therefore, need to be modelled from a common perspective, using common economic scenarios. By identifying the impact of the various economic factors, on varying mixes of liability and asset types, it is possible to map out the possible joint behaviour of a portfolio of general insurance liabilities and assets over a long time horizon.Let us now look briefly at each of the key components of the diagram.
As has been suggested, economic factors are at the heart of an asset liability model, and are its key drivers, impacting the performance of both asset and liability portfolios.
It is important, therefore, that the modeller uses realistic economic assumptions in order to allow the model to produce logical results. That is for example, the variables selected must be consistent with each other so that the year to year figures produced by the model are feasible. The parameters of the model should, however, allow for a significant change in the economic environment, including sudden shocks, such as an extreme and unexpected rise in interest rates.A simple example such as a recession illustrates the impact of economic change on both sides of the balance sheet. Liability/indemnity claims are likely to rise in a recession as a result of an increase in litigation. In property classes, an increase in thefts and fraud may be expected. While a greater than usual number of claims is adversely affecting the liability side of the balance sheet, the asset portfolio is also likely to be suffering, although the sensitivity of different asset classes to specific economic scenarios varies.
The relationship between GDP and underwriting is effectively illustrated by the graph in Figure 2.
Having recognised that investment returns help to mitigate underwriting losses, insurers face questions such as “What is the optimal asset strategy to adopt in the event of a recession?” and “How would a change in the mix of my investment portfolio from equities to bonds affect my overall results in a recessionary environment or in an inflationary environment?”
Asset liability models tend to use the Monte-Carlo method of simulation and, in order to obtain a meaningful result, apply a range of consistent economic scenarios to both sides of the balance sheet. Often many thousands of scenarios are generated producing results for a period of 5-50 years, depending on the modeller's requirements.
The fundamental task in constructing an asset simulation model is to “mimic” the behaviour of financial markets over a long period of time. An asset liability model must produce computer simulated outcomes for asset returns, yields and interest rate term structure that seem plausible given past experience, economic theory and current market information. There are a number of methods of modelling assets, including historic back testing, scenario analysis and Monte-Carlo simulation, each of which has advantages and disadvantages. As has been suggested, the Monte-Carlo method is perhaps the preferred approach. However, the key thing is to make the model as simple and transparent as possible while capturing the joint behaviour of a wide range of financial and economic variables.
The calibration of the model involves finding a set of model parameters that produce the best fit of the modeller's view of plausible behaviour. A reliance on past experience to determine the parameters may not necessarily be appropriate as the inflation and interest rate environment in the UK, for example, has been dramatically different in the 1970s and the 1990s. It may, therefore, be best to use a combination of past data and current market information and economic thinking to determine the parameters for producing scenarios.
The investment portfolio at the first year of the model is taken and simulated under thousands of different economic scenarios, to produce a probability distribution of outcomes. This will show the insurer the expected rate of return on a particular asset portfolio over a number of years, as well as the likely variability of the return, given extreme and unpredictable events.
Liability and premium scenarios
The linking of liability projections to economic scenarios represents a radical break from the past when it was believed that liabilities could be reasonably forecast by looking at past developments alone.
In addition to the impact which economic factors have on claims experience, the same factors will also affect premium levels. For example, premium levels will generally rise in line with GDP and inflation. This leaves us with a situation where shareholders' returns depend on investment returns, which depend on the amount of premium available for investment, which in turn depends on underlying economic forces.
The liability model must also take account of both the payments for business already written and in run-off, including incurred but not reported (IBNR) claims, and the likely payments for future business as yet not written. The desired outcome is a plausible estimate of ultimate claims and the variability of these payments which is achieved by the use of statistical models called generalised linear models. A liability model, consequently, comprises two parts - past liabilities and loss projections - and includes the effect of a number of variables such as premium levels, expenses and commissions, the level of economic activity and the probability of catastrophic losses in the period.
Putting it all together
The ALM process simulates thousands of different possible scenarios for economic conditions over a specific time period, say five years, together with the development of the liabilities over the same period. This allows an analysis of the impact and likelihood of infrequent economic events that would affect both sides of the balance sheet, as well as particular hits to either side in isolation. From this, the likely net profit of the company and the shareholders' funds under different asset allocation strategies over the next five years can be forecast and optimum asset allocations, reinsurance purchase decisions and integrated risk management strategies may then be determined.
Different levels of reinsurance protection can also be modelled to illustrate the benefits of current or future reinsurance programmes under different economic scenarios.
While the results are only a forecast of the distribution of possible outcomes and not a guarantee, the asset liability modelling process will assist in quantifying and, hence, managing the various risks to earnings and capital to which a general insurer is exposed.
The applications of ALM therefore include:
• Optimising asset allocation strategy.
• Capital allocation by business class.
• Reinsurance strategy.
• Capital return/raising.
• New business growth targets.
Thus, it addresses each of the every day problems facing managers of (re)insurance companies.
Insurance and reinsurance companies are complex entities. They are significantly affected by economic change and the time taken for the effects of this change to become understood. Asset liability models can help management to understand this complexity by linking the effects of different economic variables to the behaviour of both the asset and the liability sides of their company's balance sheet.
Asset liability modelling can help with a wide variety of problems, ranging from capital optimisation to asset strategy, but its real value is in its ability to enhance management's understanding of the business. This allows executives to reach better decisions by providing a tool for exploring the complex inter-relationships that make up a modern insurance company.
Oliver Peterken is managing director, and Alan Bulbeck a principal at Cordis Consulting. John Hibbert is a director of Barrie & Hibbert Ltd. Cordis Consulting is part of the Willis Corroon Group and focuses on helping senior executives of risk carrying businesses enhance earnings and manage risk. Barrie & Hibbert is an independent consultancy based in Edinburgh which services the European savings and investment industry. Tel: +44 (0) 171 488 8867; fax: +44 (0) 171 481 7193; e-mail: firstname.lastname@example.org