充分利用AXIS，第17卷

IN THIS ISSUEExecutive CornerSuper Datasets forconsolidated model runsIn the SpotlightReinventing annuityreinsurance modelingTips & TricksWhat's New in AXIS? © Oliver WymanExecutiveCornerSUPER DATASETS FOR CONSOLIDATEDMODEL RUNSConsolidating various models, particularly in the context of ALM runs, has been a common challenge forAXIS users. Super Datasets can provide some relief through specialized functionality aimed at producingconsolidated output from a network of underlying AXIS liability models, without the need to merge into asingle Dataset. Additionally, Moody’s continues to enhance this functionality and increase the scopetowhich it can apply, but the existing functionality is already worth exploring. In this article, we willhighlightkey features of the functionality, potential benefits, and some considerations forimplementation.CORE FUNCTIONALITY OF SUPER DATASETSSuper Datasets is defined within EnterpriseLink as a group of version-controlled Datasets. New functionalityenables a Fund in one of the Datasets (designated “Lead Dataset”) to utilize Subfunds from the other Datasets(designated “Member Datasets”). When the Super Dataset is run, EnterpriseLink automatically coordinates runsettings and shared objects across Datasets to produce consolidated results from the Fund, thereby allowingthe Member Datasets to remainindependent.Exhibit 1: Super Dataset processSuper DatasetDataset A(Lead Dataset)Dataset C(Member Dataset)Common DatasetVersion Control ProjectTo ensure consistency across Datasets,the Shared Object Library is created byEnterpriseLink to store shared objects.Dataset B(Member Dataset)Liability SubfundLiability SubfundCOMMON Dataset(Shared Object Library)ScenariosReinvestmentStrategiesReinvestmentsAssetsAll objects required to run the Batch must becontained within the Lead Dataset, with theexception of liability Subfunds from MemberDatasets and their linked objects.Liability Subfunds from Member Datasetsare linked to Funds in the Lead Dataset, andtherefore get processed when the SuperDataset is run. © Oliver WymanSince model components can remain in distinct Datasets, the model development cycle can continue inseparate Version Control Datasets concurrently across underlying liability models. Updates to MemberDatasets are reflected in the Super Dataset by linking to the appropriate model version, facilitatingquickre-runs or refreshes of the consolidated modelresults.With this structure, corporate components of the model can be separated from liability components whichresults in a clearer delineation of model ownership. Furthermore, EnterpriseLink can automatically check thatthe versioning of the shared objects is consistent across all Lead and Member Datasets, allowing modelers tofocus on collaborative aspects of managing the shared objects.Model ownership can be extended to Super Dataset runs as well, since EnterpriseLink can assemble post-runDatasets in a Run Area folder, allowing for efficient analysis and modelreview.CONSIDERATIONS IN SUPER DATASET IMPLEMENTATIONThe benefits provided with the use of Super Datasets can be significant, however, thefollowingconsiderations should be top of mind when determining if your organization can fully leveragethisfunctionality.For entities with a centralized modeling function, the coordination required for the initial implementationof Super Datasets will be easier to manage and likely result in less disruption to current modeling practices.Conversely, those with a de-centralized team may run into additional hurdles setting appropriate permissionsand ensuring that coordinated items across the Member Datasets, such as the AXIS Version and sharedobjects, are properly defined and managed.Those with recurring model merge needs, such as the transfer of Asset Cells to liability models for ALM runs,are likely to identify significant process improvements using Super Datasets. However, those with well-established consolidated ALM models may be reluctant to go back to distinct AXIS liability models, whichis needed to enjoy the core benefits of Super Datasets. The model development efficiency and enhancedanalytical capabilities achieved through the use of Super Datasets may warrant a model “un-merge” if thebusiness case is compellingenough.Lastly, Super Dataset functionality remains under beta release in EnterpriseLink and AXIS, with additionalenhancements being released regularly. The main limitation of Super Datasets is related to the supportedBatch calculations. At the current time, Super Datasets can be run using Fund Batch Testing or Portfolio RateCaptures on single testing targets, with preliminary support for Embedded Blocks. Additional Batch types,such as IFRS17 and GAAP LDTI reporting Batches, are not currentlysupported. 3 © Oliver WymanIn the SpotlightREINVENTING ANNUITYREINSURANCE MODELINGInsurance companies have traditionally viewed reinsurance as a liability risk transfer mechanism utilized mostlyby life insurers. More recently, however, annuity wr