The analysis of alternative reinsurance structures is a key component in our DFA analysis. Such analyses are meaningless if not carried out under consistent and proper assumptions. In the particular case of the workers compensation loss model, we subsequently used this model to assist in pricing alternative reinsurance arrangements. If the same loss model is not used to price current and alternative reinsurance structures, then perceived differences in the efficiencies of these structures might be a function of different underlying loss models as opposed to true differences in efficiencies. Inconsistencies in actual reinsurance coverage and re~ated premiums available in the market surely exist. Our focus here, however, is to seek more efficient reinsurance structures, not over/under priced coverage.

Developing reasonable and consistent parameter assumptions for a DFA model is challenging at best, and can be particularly difficult when dealing with reinsurance arrangements. It is important to continually test for the reasonableness of assumptions both in isolation and in tandem with other assumptions.

# Step 4: Running the Model

In order to become comfortable with a particular modeling system for implementing DFA, one must understand the system's underlying methodology and how that particular methodology will impact the results of the analysis. By DFA model methodology we refer to the specific technical implementation of the DFA process.

Whereas the general DFA process has become fairly standardized, there are still a number of different methodologies that are used in the technical implementation of a DFA model. Since the technical implementation of a model can have a significant impact on the results of an analysis, it is imperative that the users of a model sign off on the technical implementations and understand how the model's methodology will impact the analysis. The risk that model results are specific to a particular DFA methodology is referred to as "model risk." This is a difficult risk to evaluate; due to the time, effort and expense of performing DFA, it is often impractical to duplicate the analysis using different DFA modeling systems. As such, users should look for systems that provide a significant amount of flexibility and whose underlying fixed methodologies are consistent with their views of the insurance and financial markets.

At Swiss Re Investors, we developed our Financial Integrated Risk Management (FIRMTM)System as the modeling tool backing our DFA process. The FIRM System, like most DFA systems, uses simulation techniques to model both the assets and liabilities of an insurance company; The projected cash flows are transformed into future balance sheets and income statements that reflect GAAP, statutory, tax and economic viewpoints. The simulations are generated by a series of stochastic differential equations that are designed to allow the model user to reflect a full range of distributions, dynamics and relationships with respect to the underlying stochastic variables. The 'FIRM system is designed to allow a high level of flexibility in describing how the underlying stochastic variables behave in an attempt to minimize model risk. This

78