ICRFS™ 2018 substantially extends the flexibility of the Reinsurance module of ICRFS™ 2016. Reinsurance not only handles typical Adverse Development Cover on an entire book of business (whether prospective or retrospective), but also can handle a myriad of additional structures of LPTs, Adverse Development Covers, or a combination thereof.
Reinsurance contracts can now also be monitored over time as new diagonals are added by setting the first calendar year in the contract to refer to historical calendar periods. In this way, gross data can be modeled and Reinsurance arrangements (ADCs) can be evaluated for capital efficiency and cost benefit analysis.
- Groups of Accident/Underwriting periods can be split into separate contracts
- Periods do not have to be contiguous
- each contract can specify:
- number of layers
- proportions taken by each layer; quota share contracts (potentially varying by layer)
- calendar periods the contract applies for
- applicable discount rates
- calendar periods can now be set to begin in historical periods (ie preexisting Reinsurance contracts if modeling gross data)
- reinsurance settings can be saved into the ICRFS™ database associated with forecasts
- Calendar year liability streams Net of Reinsurance along with the total distributions Net of Reinsurance are calculated
Reserve positions Net of Reinsurance
Accident years with no Reinsurance cover can still be included in the analysis so the reserve position Net of Reinsurance studies can be calculated. When constructing the contract, simply set the Insurer to take 100% from each layer.
Distributions by calendar year and total, including V@Rs and T-V@Rs, are supplied (along with simulated values for further post processing if required).
Monitoring existing Reinsurance contracts
As Reinsurance contracts now allow the user to set which calendar year the contract begins in - including historical periods - this enables transparent monitoring of contracts. As the data are updated for a new diagonal, the Reinsurance module automatically allocates the losses into the correct layer based on the 'from inception' contract.
Examples of multiple Reinsurance contracts
Below shows a list of three contracts for a particular insurance company; note not all contracts are contiguous. Cover for 2012 and 2014 is retrospective and was purchased in 2017. The 2015~2017 accident years are for adverse development cover arranged in 2016 (retrospectively for 2015, 2016 and prospectively for 2017). The 2018 year is prospective.
Example of quota share in layers
Proportions can be allocated between any of the three layers (insurer, reinsurer or retrocessionaire). The allocations cannot exceed 100% at any layer but there is no requirement that they must sum to 100%.
Output from the Reinsurance module is summarised in total (Aggregate) and for each contract.
As with the ICRFS™ 2016 Reinsurance module all summary statistics (Quantiles, V@Rs, T-V@Rs) and simulated values are available. ICRFS™ 2018 calculates these metrics for the Aggregate of all contracts and for each contract. For prospective adverse development cover - distributions can be compared with the PALD module to show the reduction in risk capital required by the insurer should the adverse development cover be purchased.
The insurer's loss distributions can be compared between the total loss distribution (below right), and the new loss distributions (below left).
What input is required into the Reinsurance module?
A single composite MPTF model is identified from the data, for multiple LoBs, that measures the volatility (and trend) structure in each LOB and the volatility correlations between the LoBs.
In essence, on a log scale, a normal distribution is fitted to every cell in the loss development arrays and the means of these distributions are related in the identified trend structure.
In respect of projections for future years a lognormal distribution is forecast for every cell using explicit, auditable assumptions for each LoB.
Given that there is no closed analytical distribution for the sum of lognormals, to find the distribution of the aggregate of lognormals we simulate from each lognormal also giving consideration to their correlations.
The simulations from the lognormal distributions are the input into the Reinsurance module.
An example of the forecast table with the forecast lognormal distributions (mean and standard deviation) is shown below.