Common Starting Asset Target (CSAT) Iterative Convergence for VM-20 (March, 2017)
In the Embedded Block, powerful new functionality has been introduced in AXIS version 2017.13.01 to allow iteratively converging the starting assets in the block projections to within a specified tolerance of a user-defined output target such as the reserve. This “Common starting asset target” (CSAT) functionality is particularly beneficial for implementing VM-20 valuations, but it may become valuable in the future for other proposed reserve methodologies as well.
The adoption of VM-20 for valuing life insurance products, with an effective date of January 1, 2017, introduced complex modeling needs in terms of how to establish an appropriate amount of starting assets to use for the deterministic and stochastic reserve model projections. Beginning in version 2017.13.01, AXIS has introduced a simple and elegant solution to this complex problem which will greatly ease the burden of model setup for the user.
The need for this functionality was introduced by the following requirement from VM-20 Section 7.D.2:
“If for all model segments combined, the aggregate annual statement value of starting assets is less than 98% or greater than the larger of NPR or 102% of the final modeled reserve, the company shall provide documentation in the PBR Actuarial Report that provides reasonable assurance that the modeled reserve is not materially understated as a result of the estimate of the amount of starting assets.”
The “modeled reserve” in this statement refers to the greater of the Deterministic Reserve or the Stochastic Reserve, as defined in VM-20. Both the Deterministic and Stochastic Reserve outputs are dependent upon the amount and composition of starting assets used as inputs in the model projections. Due to the nature of its reserve methodology, the Stochastic Reserve can be particularly sensitive to the amount of starting assets. Consequently, the complicating factor is that one of the model’s outputs (reserves) is required to be within 2% of one the model’s inputs (starting assets). When setting up the model, it is impossible to know with certainty in advance if the selected amount of input starting assets will result in an output reserve that is within the allowed tolerance. This problem can only be solved by iteratively adjusting the amount of assets up or down, if necessary, and recalculating reserves until the results converge toward an acceptable answer.
The AXIS implementation of this solution utilizes a new concept within the Embedded Block of a “Common starting asset target” (CSAT), which can be defined at the top of the Block Projection Assumptions screen through a formula table. The user-coded “target” defines the model output that is required to be within the specified tolerance of the starting assets. For VM-20, the target can be coded to the maximum of the Net Premium Reserve, the Deterministic Reserve, and the Stochastic Reserve. The Deterministic and Stochastic Reserve components can be modeled and calculated using the multiple Embedded Block functionality, while the Net Premium Reserve is calculated on a seriatim basis at the liability Cell level. If the resulting aggregate maximum VM-20 reserve is not within the specified tolerance (i.e. 2%), the starting assets used in the blocks will be automatically adjusted up or down, and the deterministic and stochastic projections will be recalculated. The asset adjustment options are to add cash, to scale the inforce assets pro rata, or to scale asset pro rata with limits between 0%-100% (with cash used after the limits are reached). AXIS has embedded solver logic to optimize the efficiency of this asset convergence process.
The advantage of implementing the CSAT functionality within the Embedded Block is that not only can it be used at the original valuation date for valuation purposes, but it can also be used for pricing purposes by revaluing the reserves at future dates along a pricing path. At each future “pivot date” selected for reserve revaluation, the CSAT functionality allows for iteratively adjusting the existing inforce assets at that date until an acceptable reserve is calculated.
The iterative nature of this functionality will necessarily increase model run-times. Please be assured, however, that every effort is being made to implement these iterations in AXIS in an optimal manner. This includes maintaining an efficient convergence solving algorithm and only recalculating the items that need to be recalculated from one iteration to the next.
Please see the VM-20 user guide in Knowledge Base article 2032 for detailed instructions on how to implement this CSAT functionality within the context of a VM-20 model.
This CSAT functionality also has the potential to be useful in implementing proposed changes to the VACARVM (also known as AG-43 or VM-21) reserve methodology.