Hemant Shah and Peter Nakada take stock of the use of dynamic financial analysis and argue that analysing risk adjusted return on capital is likely to be more immediately useable by the insurance industry.

Throughout the world, insurance and reinsurance shareholders are increasingly demanding that property/casualty business activities generate sufficient risk-adjusted returns to justify the capital required for their support. Excess capital provides an added measure of safety for policyholders, but it also indicates lax capital management and likely destroys shareholder value.

More sophisticated measures of risk, capital, and value are now supplementing traditional measures of profit and loss. Today, leading (re)insurers are increasingly seeking to understand better the links between risk and capital.

Key questions include:
• How much capital is needed to support the diversity of risks on the balance sheet, and should this capital be attributed to the various activities?
• Which activities are creating shareholder value and which are destroying it?
• What are the links between investment risk and underwriting capacity?
• What is an appropriate level of catastrophic risk?
• How should I create incentives that reflect not only profit and loss, but capital utilisation and shareholder value creation?
• What is an optimal strategy for reinsurance or for risk finance?
• How should enterprise-wide risk be communicated to rating agencies? How do I manage the economics of my business within regulatory and rating-agency capital constraints?
This article summarises how insurers have been attacking these issues, highlights some problems with the typical approach and suggests a new method that can provide valuable answers in a short period of time.

DFA in the spotlight
Over the past several years, property/casualty actuaries have responded to these capital management issues by developing capabilities to measure enterprise risk and link these risk measures to management decisions. This effort has been galvanised around the term “dynamic financial analysis” (DFA).

The Casualty Actuarial Society (CAS) broadly defines DFA as “the process for analysing the financial condition of an insurance enterprise. Financial condition refers to the ability of the company's capital and surplus to support adequately the company's future operations”, through a currently unknown future environment. In particular, “the DFA concept broadens the scope of the actuary's analysis to encompass the entire balance sheet as well as the company's business plans over some future horizon.”1

While the CAS's definition of DFA is quite broad, and does not specify the technique for analysis, in practice, DFA has come to mean “multi-period simulation-based stochastic analysis of statutory and GAAP financials.” That is, most DFA implementations take accounting models of the firm, and simulate them forward in time, based on risk characteristics of assets and liabilities and assumptions about the changing environment and management actions.

The intended uses of DFA include all of the capital management issues outlined in the introduction to this article. However, because DFA also incorporates accounting and multi-period effects, it is intended to be used for other applications, such as tax optimisation, insurance-cycle based product strategy, reserve management and investment strategy planning.

The real world
Despite the promise of DFA, many (re)insurers have grown frustrated with it. While there have been some success stories, extensive interviews with the senior management of leading (re)insurance businesses reveal a landscape littered with the wreckage of DFA projects that have simply failed to live up to their expectations. Representative include: “It is easy to get lost in a DFA project . . .”; “. . . the least successful DFA projects I have seen try do to everything . . .”; “We worked on DFA in its early stages, and we are just as frustrated with this as we were four to five years ago . . .”; “. . . the progress in terms of probabilistic DFA is slow and cumbersome.”; “The biggest challenge we have had using DFA is looking at it on a corporate-wide perspective.”

There are several common themes driving these frustrations.

Objectives and applications are too vague. Again and again we have seen DFA projects focus on the techniques, with the key applications following almost as an afterthought. As a result, DFA teams often get bogged down in conflicting opinions with management on what exactly DFA does, what its outputs should be and how these outputs can be used to support decisions. As a result of this lack of focus, many projects end up becoming open-ended research platforms or counting exercises and fail to provide tangible decision support for senior management. After 12-18 months, they are often shelved.

Improper treatment of asset and operating risk. Frequently DFA models are excellent at modelling liability risk, but fall short on modelling asset risk and operating risk. The key problem here is that the diversification benefits offsetting standalone insurance risk are highly dependent on the amount and type of the other risks of the organisation. In other words, it is necessary to get all of the risks right in order to get any of them right.

Asset risk is difficult to model because it is so complex. For example credit risk is a low-frequency, high-severity risk that is difficult to model based on historical data; mortgage-backed securities have complex interest rate options, and equity risk is difficult to integrate with interest rate and credit risk.

Some DFA modellers take an overly simple view of asset risk, and end up mis-estimating the risk by, for example, ignoring embedded options. Other modellers take a highly complex approach to measuring asset risk, modelling the cashflow of each highly complex security, and end up drowning in complexity.

Operating risk is difficult to model, because it is difficult to pin down. This risk type is a “catch-all” that includes volatility in volumes and margins, as well as event risk from fraud, errors and systems interruptions.

Modelled results are difficult to test and debug. DFA models are largely simulation-based, rather than closed form. That is, risk is measured by characterising assets and liability cashflows as functions of risk factors, then generating many different risk factor values to create distributions of potential future cashflows. While simulation models have the advantage of being easy to understand, the results tend to be difficult to test and debug.

Testing individual modules of the model (for example, each risk type), or exploring the sensitivity of results to key assumptions, involves rerunning the simulation repeatedly. Furthermore, when risks are combined, the correlations are “hidden” inside the simulation. This makes it difficult both to validate the diversification benefits, as well as to calculate marginal risk contributions. In addition to the technical concerns, this lack of transparency can undermine a DFA team's ability to communicate the results to management.

Models are unnecessarily complex. It seems that many of the modellers have not made an appropriate trade-off between accuracy and simplicity. Basing the models on statutory or GAAP financials, rather than on cashflows and economic value, introduces significant complexity associated with accounting and reserving rules. Building a complex interest rate term structure model into the DFA model may be overkill if the nature of the interest rate exposures do not warrant it. Some risks are small enough in comparison with others that they can be effectively ignored.

An alternative approach
There is an alternative approach. Over the past decade, the banking industry has successfully addressed a similar set of enterprise risk and capital management issues. The techniques used to develop the necessary metrics and business insights, unlike their current DFA counterparts in the property/casualty insurance industry, were largely embraced by these financial institutions and are now an integral part of how financial institutions measure and manage risk, capital and value. Banks not only use these tools internally, but also communicate capital allocation and return on capital by business line or activity in their annual reports.The output from these techniques are measures of capital required to support risk - economic capital, and measures of return on that capital - risk-adjusted return on capital (RAROC). These techniques have achieved success in large part by focusing only on those applications central to measuring capital and value by activity. In other words, RAROC methods and capital management projects are simply less ambitious than DFA, but are much better suited than DFA in addressing key issues of concern at a senior management level.

The RAROC approach is built on a highly practical, efficient and decision-oriented methodology focused on issues of capital protection, capital adequacy and capital productivity. More specifically, these are:
1. How much risk am I taking? How much capital do I need to hold (overall and by business segment)? Answer: calculate and attribute “economic capital” based on risk.
2. How is my business performing (overall and by business segment) on a risk-adjusted basis? Which businesses create shareholder value and how much? Answer: measure RAROC overall and by business segment. Calculate hurdle rate required by shareholders and compare business RAROCs to these.
3. What strategic decisions do I need to make to maximise shareholder value? Answer: compare RAROCs of alternative value propositions.
Working together, Risk Management Solutions (RMS) and Oliver, Wyman & Company have adapted RAROC methods from banking to the property/casualty insurance industry. This RAROC framework measures and integrates all risks (asset, liability, operating), calculates the economic capital required to support those risks, attributes economic capital by activity and measures the risk-adjusted return on capital for each activity.

Most modules of property/casualty RAROC estimate risk using “closed form” equations. That is, the distributions are measured formulaically rather than by simulation. The correlations within different risk types are explicitly addressed. The advantages of the closed form or mixed close form/simulation over the full simulation approach are:
• Models are quicker to build and run.
• Impact on value of the various risk types is fully transparent.
• Risk contributions across types and business segments are calculated with relative ease so that required capital can be calculated at any level of granularity.
• Closed form solutions can measure the sensitivity of results with respect to parameters, especially the ones which are difficult to estimate such as correlations across certain risk types.

Figure 4 illustrates the distinction between a “typical” DFA model discussed in literature published to date, and property/casualty RAROC in terms of financials used and calculation procedures. As seen in the chart, RAROC reflects true economics of risk, as it is based on market-value of assets and liabilities and is more efficient as most of its modules are in analytically closed form.

The way forward
Enterprise risk management strategies predicated on shareholder value considerations are increasingly becoming an integral component of an overall approach to strategic capital management and risk finance. If the experiences of the banking industry over the past decade provide any indication, we can expect to see RAROC-based performance metrics emerge as the common yardstick used to evaluate property/casualty performance on a consistent basis for different (re)insurance activities. In addition to yielding invaluable internal benefits, this would also improve communication among insurers, reinsurers, risk-transfer counterparties, shareholders, policyholders, rating agencies and regulators.

Although DFA methodologies, properly implemented, can provide useful insights into the relationships between risk, capital and value, the track record to date is quite spotty. While many of the obstacles to a successful DFA application can be overcome with sufficient planning and determination, it is important to clearly understand the potential pitfalls, and to realise that alternative approaches, such as property/casualty RAROC are now available.

Given the capabilities and limitations of the various alternatives, we believe that a best practices approach in the insurance industry would be to apply both RAROC and DFA judiciously, taking care to align the tools and the objectives properly. RAROC should be viewed as an approach that takes risk and return information and calculates economic capital and RAROC by activity. This provides a quick (two months to implement) economic snapshot of the firm that answers the most important capital management questions. DFA should be viewed more as a research platform for evaluating the impact of various strategies on the financial statements of the firm over time.Hemant Shah is a founder, and senior vice president of Risk Management Solutions (RMS), a provider of products and services for the quantification and management of property catastrophe, natural hazard and complex business risks. Tel: +1 650 617 6552; e-mail: hermants@riskinc.com.Peter Nakada is a director of Oliver, Wyman and Company, which is a strategy consulting firm dedicated to the financial services industry, with major practice areas in risk management, wholesale financial services, retail financial services. Tel: +1 212 541 8100; e-mail: pnakada@owc.com1. (The Actuarial Review, 1997).