Helix Catastrophe Modeling Operating Standards

June – September 2020

1. Introduction and Executive Summary

At the time of writing, various initiatives across the insurance/reinsurance and catastrophe modeling communities are investing in establishing common operating standards, prioritizing open common data standards and catastrophe model interoperability. Aligned with these efforts, a group of industry practitioners and model and service providers convened in November 2019 to address key issues impacting use of catastrophe models, and initiate work to explore and advance work in these areas.

The resulting Catastrophe Modeling Operating Standards (CMOS) initiative comprises over 40 companies across the insurance and catastrophe modeling industries. In early 2020 and working under the guidance of The Institutes, the CMOS initiative determined to engage an exploratory survey and proof of concept project addressing implementing an open common exposure data standard and assessing model interoperability issues.

This document presents the proceedings, results, and recommendations from that project, which took place from June through September 2020.

1.1 Executive Summary

In Spring 2020 and under the guidance and support of The Institutes, the CMOS initiative participants (representing insurers, reinsurers, intermediaries and model/data/service providers) agreed to embark on a limited project to explore and survey issues regarding implementing a common standard in the marketplace, and advance multi-model interoperability. The Project Team comprised volunteers from AIR Worldwide, Aon, Chubb, Guy Carpenter, KatRisk, Renaissance Re, and Xceedance, all of whom represented the various constituencies concerned.

The Project Team adopted a project charter, agreed criteria for identifying a common standard, and selected the Open Exposure Database (OED is the Open Exposure Data standard, developed by Nasdaq and currently supported by the Oasis Loss Modeling Framework. More information is available at https://github.com/OasisLMF/OpenDataStandards. ) as the standard to use for the project. Key activities undertaken included:

Engaging Nasdaq and Oasis to educate the team regarding OED design, origins, and operations, and coordinating with the IDF ITWG regarding their work in developing technology tools to convert data to and from OED
Assessing OED implementation and operating issues at a high-level, leveraging the differing perspectives of the participating companies
Surveying OED capabilities by directly converting individual policy and portfolio level data to OED and back again, identifying challenging areas in describe exposures and contracts across standards while maintaining fidelity
Identifying interoperability opportunities and issues as respects converting between OED and other standards

Based on this survey work and observations, the Project Team is making five coordinated recommendations to progress this work as follows:

The community should unify the current diverse investments and initiatives in the marketplace regarding common standards and interoperability by forming a single consortium or collaborative to oversee carrying the work forward
The community should start using OED in market areas where it can support business activity today
The community should use the organizational model recommended above to address OED design, development, curation, and evolution by leveraging experience from OED use in market
The community should encourage/assist/support model providers in consuming OED directly, and in the absence of that, and to address exceptional situations, continue the ITWG work to develop conversion and mapping assets with model providers owning mapping definitions/interpretations to ensure fidelity of treatment
The community should promote choice and innovation by helping the community of model developers and service/data providers continue to advance risk analytics capability by protecting their unique IP

2. Project Proceedings

2.1 Background

In November 2019, Chubb and SpatialKey hosted a Forum for Change in New York City to address issues affecting catastrophe modeling practices and operations across the insurance and reinsurance industry. Over 40 companies were represented, including insurers, reinsurers, intermediaries, model providers, and industry and risk analytics service providers. The Forum topics and speakers addressed multiple topic areas including data collection and establishing standards, modeling costs/transparency/interoperability, efficiency challenges, and industry collaboration.

Building on the discussions at the Forum, a call to action proposed forming two related Workgroups to address a) exposure and modeled loss data standards and solution interoperability, and b) risk modeling complexity, validation, industry data, new capabilities. Representatives of over 40 companies volunteered to participate in these Workgroups, and the collective group determined to collaborate with The Institutes to ensure appropriate compliance, governance, and to provide executional guidance where needed.

The Workgroup volunteers met in March 2020 to assess options for defining and progressing work addressing the issues raised at the Forum. As part of a supporting survey process, the Workgroup agreed that the top priorities to begin the work included identifying and implementing open common exposure data standards, and addressing multi-model interoperability issues.

The Workgroup reconvened in May 2020, adopting the acronym CMOS (for Catastrophe Modeling Operating Standards) to identify the initiative, and determined to proceed with a time-bound Proof of Concept project to explore and define issues and challenges in adopting a common data standard and enhancing model interoperability, using a small team of industry cat modeling practitioners and model providers. Working under the guidance of The Institutes, volunteers from AIR, Aon, Chubb, Guy Carpenter, KatRisk, Renaissance Re, and Xceedance agreed to pursue the project, facilitated by Mark Cravens as independent consultant (see Appendix).

2.2 Project Proceedings

Project Initiation
The Project Team convened on 9-June-2020. In the initial meeting the team adopted a proposed charter, agreed criteria for selecting a common standard, and designated OED as the standard to use in the project based on those criteria (See Appendix for Project Charter and information on OED selection). The original intent included assessing CEDE (CEDE is an open exposure data format designed and implemented by AIR Worldwide, and on which parts of the OED design and code set are based. More information is available at https://www.air-worldwide.com/data- standards/cede-7-0-database-schema/) to OED mappings and conversion, and “field testing” any available conversion tools in a production environment (per below, these plans changed to avoid duplicating other efforts already underway).

Education
The next two weeks focused on educating the team on OED and work underway by the Insurance Development Forum’s Interoperability Technical Working Group (ITWG) on building technology to convert CEDE format data to OED and back again. The Project Team arranged educational sessions with Oasis and Nasdaq as respects OED, and separate sessions with the ITWG team as respects their conversion work. In the course of that, it was clear the ITWG was already addressing one of the PoC goals in building the CEDE-OED technology, and the ITWG invited the PoC team to participate in a technical review of that work, initially scheduled for early July.

Change in Direction: OED Usage Profiling and OED Coding
The PoC team determined to change course so as not to duplicate the ITWG work, and addressed assessing OED operational readiness in two stages. Initially the team performed use case assessment work in which the team profiled OED usage through different operational use cases such as geocoding, vulnerability description, financial modeling, and data development, administration, and sharing.

Subsequently, the ITWG rescheduled the technical review to mid-August, and the PoC Team determined to continue survey work on OED by coding actual market data into that format. While the original intent was to execute this work in a short timeframe, the team needed to extend timing on the project to include the ITWG technical review, and also due to interruptions from tropical storm activity requiring immediate attention.

ITWG Technical Review
The ITWG technical review on 13-August-2020 provided the PoC team insight into the design and progress attained in prototyping a model-agnostic exposure mapping framework. The framework design uses independently defined mapping files defining conversion rules between a given individual standard and OED (avoiding choice constraints), and is built using commonly available technology for flexibility in deployment and development. Further information is available at https://github.com/OasisLMF/OasisDataConverter.

The meeting was an important step in cross-team collaboration, and surfaced strategic considerations that also had arisen in the usage profiling work regarding market adoption and collaborative governance going forward.

Ongoing OED Coding Work
The OED coding work focused on independent efforts by Aon and Guy Carpenter working with Chubb data already in each intermediary’s possession, thereby reducing the need for any data sharing agreements (as the data is already part of existing business relationships). The Guy Carpenter team focused on converting several individual accounts identified by Chubb to explore account and contract-specific aspects. The Aon team addressed conversion at the portfolio level, converting an agreed Chubb portfolio into OED, and then reconverting it back to its original format to test for data fidelity.

Consolidation of Project Outcomes
Alongside the OED coding work, the Project Team assembled assets for collecting observations and recommendations for this report from the individual teams/participants, with the intent to collate feedback and priorities using the voices of each company. This feedback collection addressed four dimensions:

OED Design/Content/Operations: design and operating considerations for the format as it exists today
Interoperability: recommendations from the teams regarding prioritizing expanding interoperability capabilities in the market
Market Adoption: Considerations in socializing, publishing, implementing, and supporting OED and relevant tools across the market
Organization and Governance: Perspectives on how best to organize around this work going forward

In the last weeks of September, the Project Team consolidated mutual observations from the combined education, profiling, and OED coding work, using those to formulate the recommendations presented in this document.

3. Accomplishments and Observations

3.1 Solution Precepts

The Project Team approached the work at hand using critical solution design precepts established in the November 2019 Forum for Change: Choice, Innovation, Transparency, and Efficiency.

Choice includes choice amongst catastrophe models, modeling approaches, providers, and technology: solution options must seek to expand choices rather than becoming a barrier to market or a constraint to choices
Innovation includes leveraging new/changing modeling, data, technology, and business applications: solution options must encourage and not constrain the ability for community to innovate around these complex risk assessment and management challenges
Transparency includes the ability to understand the genesis of modeling output and analytics, and the differences between varied modeling approaches, data requirements, and risk analytics: solution options must seek to expand transparency, balancing that with respect to potentially proprietary intellectual property (IP) across the community
Efficiency issues span the entire catastrophe modeling process from data collection and coding, to manipulation for use in different models, and on to managing resulting results and analytics: solution options must push for improving efficiency without compromising the values of the other three solution precepts

3.2 Accomplishments

As an exploratory and survey project, the Project Team engaged with generalized goals as expressed in the Project Charter (see Appendix). Key accomplishments related to the aims set forth in the Charter include the following.

Expanded the knowledge base for the participants and their respective firms regarding the specifics around OED design, origins, and evolution through focused educational sessions and actual experience working with that standard
Established a working relationship with the IDF ITWG to start synchronizing interests and efforts on interoperability, including participating in the technical review of data conversion technology prototype addressing CEDE to OED conversion
Assessed current OED capabilities and approach to describing simple and complex exposure aspects including basic exposure information (location, composition, values)more detailed and complex exposures including secondary modifiers (For the purposes of this document, secondary modifiers include construction and composition details supporting various modeling approaches, which are peril-specific, often differ by provider, and may be IP-sensitive), and complex policy conditions.
Assessed operational considerations such as exchanging data in OED format, describing/packaging complex portfolios, and identifying prospective technical support needs
Assessed interoperability considerations, including manual conversion of data between formats, aspects of developing and supporting data conversion technology, and respecting IP considerations
Assembled coordinated recommendations for progressing this body of work (presented below) with coordinated efforts while seeking to minimize operational and trading disruption

3.3 Summary Observations

Following are key observations resulting from the Project Team education, work with the ITWG, and the OED use profiling and coding work.

Overall, OED is comprehensive in terms of describing most exposures and contracts, especially commonly used policy forms and reinsurance structures
In its current form, OED does not clearly address some aspects common to current catastrophe modeling practices, such as describing builders risk, workers compensation and some watercraft/marine exposures
Similarly, while OED has the framework for addressing a variety of geographic/administration identification schemas, construction and occupancy classifications, and other exposure aspects, there may be need for commonly used cross-reference tables and classification schemas to either be directly included, or otherwise supported/cross-referenced for common usage without misinterpretation
The Project Team did not have the opportunity to directly assess complex portfolio hierarchies and/or reinsurance structures, and recognized that assessment of these aspects will benefit market implementation
Having coded data in OED, the Project Team agreed that there are clear challenges in fully addressing some aspects of exposure description (such as secondary modifiers) and complex contract terms (such as peril and/or geography-specific conditions), especially when working with multiple model constructs
Regarding interoperability, the Project Team recognized that the ideal solution for using a common standard would be for model developers to automatically consume the common standard
In the absence of models directly consuming a common standard, there is a need for mapping/conversion technology and supporting assets, and – importantly – model providers should be responsible for specifying mapping/conversion between their format and the common standard to avoid misinterpretation of data (e.g. by a third- party developer)

Implementing and using a common standard must not be bound by any one technological framework or stack, and need to be adaptable over time to keep up with technological changes
Additional assets will be helpful for market adoption, including reference implementations and sample data, ongoing support documentation, and possibly technical assets such as APIs or other tools

In supporting the consolidation of observations and recommendations, AIR Worldwide provided the following to ensure clarity of their perspective as a leading model, data, and service provider in the catastrophe modeling discipline:

“AIR will continue to support our clients and the industry through thought leadership on making it easier to consume and transform exposure data. AIR has historically both opened up our standard and supported interoperability with data sharing agreements and supporting the creation of OED. We have made an in-kind contribution of opening up our CEDE formats and can further support clients with this as needed. We will actively participate in these discussions to help the industry meet its goals.

We believe that any solution for common data formats must not stifle innovation. AIR will continue to develop state of the art models for supporting our clients and will evolve the CEDE formats to be able to support the needs of our clients.

With regard to the POC, we believe the problem statement and the end goal need further clarification. It must be acknowledged that as the CEDE has evolved over time, the OED may no longer be representative of the latest version of the CEDE. Ultimately, we believe that if we are able to develop and maintain a solution that satisfies the needs of the industry, that solution must have input and buy-in from all relevant parties.”

4. Recommendations

In creating these recommendations, the Project Team is hopeful to maintain and build the momentum already existing in the marketplace and advance these efforts around common standards and interoperability. Accordingly, the Project Team is making recommendations in five areas: Organization, OED Application, OED Evolution, Interoperability, and Choice and Innovation. The graphic below presents the conceptual construct for these recommendations, each of which is described in more detail in the body of this section.

4.1 Organization

The community should unify the current diverse investments and initiatives in the marketplace regarding common standards and interoperability by forming a single consortium or collaborative to oversee carrying the work forward. Otherwise the community risks duplication of effort and likely confusion in approaching the same problems through different and possibly conflicting efforts.

Such an organization should include a Steering Body to oversee and address needs and concerns in advancing the work, guide requirements, set priorities, and manage resources. The Project Team also recommends an organizational model with an independent executional entity to carry the work forward, including program management, resource/product development and management, and involving community working groups as may be needed.

It is important to note the Project Team is comprised of catastrophe modeling practitioners and providers, and will defer to other more specialized entities with expertise in collaborative organization management regarding specifics in addressing this recommendation.

4.2 OED Application

The community should start using OED in market areas where it can support business activity today. As noted above, there are gaps in applying OED for use across all area of the market currently using catastrophe modeling to transact business. Additionally, the Project Team identified challenges using some model-specific details, and describing some complex financial structures (NB as cited earlier, the Project Team did not have the opportunity to assess handling of complex portfolio and reinsurance structures).

This recommendation envisions applying OED for use in geographic and/or peril market areas with lower requirements/opportunities to leverage highly-detailed data. An example of a candidate area might be simple residential business as opposed to high-value and complex commercial or industrial exposures. It may be useful for the community to pilot this approach with limited/dedicated participants to initially engage in trading with OED, scope operating and modeling issues, and start building best practices while minimizing potential disruptions to trading activity.

4.3 OED Evolution

The community should use the organizational model recommended above to address OED design, development, curation, and evolution by leveraging experience from OED use in market. Under this recommendation, the recommended Steering Body drives definition of core standard attributes with industry participants and providers. This may include defining core data for different modeling purposes, ongoing design/evolution of OED components, technology options, and extensibility to new areas of risk and lines of business.

It is further recommended that under this model, the insurance/reinsurance industry participants should drive the definition of contracts: how they are described and how they operate when activated by appropriate risk aspects (i.e. causes of loss). This would ensure mutually understood fidelity of contact description, enhance abilities for technical evolution through common standards, and provide a practical design framework for extension into other lines.

Similarly, it is further recommended that provider participants (model developers and associated service/data providers) should drive common definition/description of attributes needed for peril modeling and risk analytics. There will be a need for consensus around core information needed to describe exposures and common/important peril-specific attributes. This would in turn influence standards for initial data collection at the point of original submission, further ensuring data and interpretation fidelity across the industry value chain.

4.4 Interoperability

The community should encourage/assist/support model providers in consuming OED directly, and in the absence of that, and to address exceptional situations, continue the ITWG work to develop conversion and mapping assets with model providers owning mapping definitions/interpretations to ensure fidelity of treatment. Under this recommendation, the organization recommended above would serve as a central resource for addressing and managing these efforts, including provider community collaboration and support.

4.5 Choice and Innovation

The community should promote choice and innovation by helping the community of model developers and service/data providers continue to advance risk analytics capability by protecting their unique IP. This could include supporting secure exchange of model/provider- specific detailed data, including IP-sensitive/proprietary aspects as part of common standard data exchange process.

This recommendation requires additional consideration and assessment, and presupposes successful advancement of the previous four recommendations. It is therefore further recommended that this be revisited once progress has been made to best leverage that progress, and out of respect for possible currently unrecognized issues and challenges.

5. Conclusion

The intent of this project has been to identify issues, clarify options, and create recommendations for implementing an open common exposure data standard and assessing model interoperability issues. In the course of this work, experienced catastrophe model developers and practitioners have surveyed the practicalities of using OED as a common standard. While OED provides a solid core for describing property exposures for catastrophe risk analysis, it does have descriptive and operating limitations, and further development and evolution is needed to extend its utility across market segments.

The tactical examination of using OED also highlighted strategic issues for the broader community of underwriters, intermediaries, and providers. These include organization and governance around developing and curating a common standard, market adoption, and model- specific differentiation including addressing IP concerns. The Project Team recommendations position these issues for advancing this work cooperatively across the market. It is the Project Team’s perspective that the best way to progress this work is by engaging it in day-to-day market practice, thereby illuminating practical priorities and engaging innovation at the operational level.

The Project Team would like to express sincere gratitude to the CMOS Workgroup participants, The Institutes, Oasis, Nasdaq, and the IDF ITWG for their engagement and participation in this work. We very much appreciate the interest and support expressed across the catastrophe modeling community for this endeavor.

6. Appendix

6.1 Project Management and Participants

6.1.1 Project Management and Facilitation

The collective CMOS participants determined to collaborate with The Institutes to ensure appropriate compliance, governance, and to provide executional guidance and management resources where needed. Additionally, the initiative engaged and independent consultant with catastrophe modeling and insurance industry experience to facilitate project execution, communications, and work product development.

The Institutes
Anthony Sasso, Head of Operations, RiskStream Collaborative
Susan Jakiela, Program Manager
Ivan Avila, Program Manager

Cravens Consulting
M. Mark Cravens, Principal

6.1.2 Project Team

In assembling the Project Team, CMOS participants agreed the importance of representation from different catastrophe modeling community constituencies to include insurers, reinsurers, intermediates, and model and service providers, and to limit the number of firms represented thereby streamlining logistics. Following are the firms represented and their project participants.

AIR Worldwide
Gatayri Natarajan, Vice President, Product Management Andrew O’Donnell, Senior Engineer II

Aon
Dan Dick, Executive Managing Director Meredith White, Managing Director

Chubb
Donald Parkes, Senior Vice President Mary Wong, Catastrophe Management

Guy Carpenter
Susan Denike, Managing Director GC Analytics
Ming Li, Managing Director
Taronne Tabuchi, Vice President, Senior Risk Engineer

KatRisk
Guy Morrow, Managing Member Kaia Morrow, Data Analyst

Renaissance Re
Andrew Dhesi, Senior Vice President Dail Rowe, Senior Scientist

Xceedance
Rakesh Dhawan, Account Management

6.2 Project Charter

Following is the Project Charter as adopted by the Project Team in the 9-June-2020 team meeting.

The purpose of this Proof of Concept is to explore implementing a target state of operations around a common property catastrophe exposure data standard, and resources supporting data standard/model interoperability. This exploration will be time-bound to complete prior to August 2020 unless otherwise agreed by the PoC team members. Key dimensions addressed in this work include:

The value and issues arising from identifying and maintaining an open common exposure data standard
The needs/process for common conversion/translation resource development and maintenance, including ensuring data fidelity and IP governance
The potential impacts on and prospective benefits for cat modeling operations efficiency (e.g. workflows)
The efficacy of various technologies in supporting these processes
The requirements for establishing, implementing, maintaining, and governing the above
The synchronizing of the CMOS initiative with other work ongoing in the industry and cat modeling community

It should be noted that this PoC charter and the work resulting should support the broader charter for the CMOS initiative as that becomes established.

6.3 Common Standard Selection Criteria

The criteria below are verbatim from the Open Modelling Collaboration - Open Data Standards Reportv1.pdf report of fall 2019. The report was provided to the CMOS team at the outset of the project that was provided gratis to the project team as part of the ongoing work undertaken by the IDF. In that, the criteria for an open common data standard are as follows.

Available to everyone, with no subscription/fee involved
Non-commercial license terms attached which clarify usage terms
Accessible – in a usable format and technology options
Transparent - Documentation provided
Owned by the industry with governance by a cross-representative steering board

including vendors, brokers, re/insurers.
Mechanism for contributions, change requests and updates is clear and documented

After review, the Project Team voted to adopt these criteria in the 9-June-2020 meeting in order to ensure synchronization with other efforts in the market. Those initiatives had also previously designated OED as the selected standard meeting these criteria, and the Project Team similarly voted to adopt OED as the common standard for the purposes of this project.

For more information about Helix, email helix@theinstitutes.org.

Initiative for Catastrophe Modeling Operating Standards (CMOS) Report on Operational Survey and Proof of Concept Project