Disaster Risk Finance Analytics Project Development of core open source Disaster Risk Finance quantitative tools Terms of Reference 1. Background Developing countries typically lack financial protection against the impacts of natural disasters and rely on ex-post measures (e.g. budget reallocations, donor assistance) to attempt to meet financing needs. Disaster Risk Finance (DRF) can increase the ability of governments, homeowners, businesses, agricultural producers, and low income populations to respond more quickly and resiliently to disasters. Developing country governments are facing increasing amounts of disaster risk data from historical disaster loss databases and catastrophe risk models. However, by itself this risk information (e.g. average annual losses, probable maximum losses) is unsuitable for financial analysis to inform DRF decisionmaking. While catastrophe risk data lay the ground for DRF solutions, there is a need for analytical solutions that bridge the gap between this data and informed financial decision making. In response to these challenges, Governments are increasingly requesting advisory services to manage their fiscal costs of natural hazards. Addressing this demand, the Disaster Risk Financing and Insurance Program (DRFIP), a partnership of the World Bank s Finance & Markets Global Practice and the Global Facility for Disaster Reduction and Recovery (GFDRR) has supported many countries over the past decade, pioneering innovative financial tools and strategies to increase post-disaster financial response capacity and build catastrophe insurance markets. The DRFIP has an analytics function which produces a variety of quantitative outputs to help World Bank clients develop DRFI strategies. These include probabilistic financial disaster risk assessments to help clients to understand their financial risk arising from natural disasters; cost-benefit analyses of different types of financial instrument (e.g. reserve funds, credit options, insurance); quantitative support to financial instrument design; and technical pricing of financial products. The DRFIP analytics function applies expertise in actuarial and probabilistic catastrophe risk modelling, and economic analysis to deliver this support. The DRFIP is undertaking a four-year DRF Analytics Project to improve the understanding and to increase the capacity of governments to take informed decisions on DRF based on sound financial analysis. The objective of the project will be achieved through four outcomes: (1) Governments understand their financial risk related to natural disasters; (2) Governments employ efficient financial/actuarial analysis, such as cost-benefit analyses, in the development of DRF strategies; (3) Improved financial capacity to meet financial needs immediately following natural disasters; and (4) Increased capacity in Governments to monitor and evaluate DRF strategies. Under this project a suite of interactive DRF quantitative tools will be developed which can be adapted and applied to support capacity building and decision making in countries. Tools will be available in both spreadsheet format and as a web based application (on a web based platform) and will be used to support capacity building and decision making by Governments, donors and other stakeholders. 2. Scope of this consultancy DRFIP is now seeking support from a suitably experienced and qualified consultancy or a consortium (herein referred to as the Consultant ) in the specification, development and delivery of these tools in
spreadsheet format (Phase A). This contract is for Phase A only; a Phase B contract for development of the tools in a web based application will be contracted for separately following completion of Phase A. Methodologies underlying the tools will be of a complex quantitative nature; the user audience will include those from non-technical backgrounds. Therefore, engaging visual representations and clear succinct descriptions of complex results is paramount. The tools must be robust and generic enough to capture a large number of countries while also remaining simple to use and easy to interpret the results. This consultancy will leverage the existing portfolio of tools and methodologies developed by the DRFIP analytics function to create four interactive DRF analytics tools. Tool development will draw on current international best practice in user interface design, user experience, visualization, transparency and quality in coding, and probabilistic, financial, actuarial and economic analysis. Flexibility in the specification and design of the tools is required and the consultancy is asked to include a 10% contingency to allow for changing requirements. This consultancy will include a data collection exercise from publicly available reliable information. The final tools will use a set of prepopulated disaster data and macroeconomic data prepared by the consultant. The development of all tools will follow a cycle of: i. Specifying the problem ii. Specifying a tool to address the problem iii. Data gathering iv. Developing the tool v. Verification and validation of the tool vi. Independent audit, user testing and formal acceptance vii. Feed back to the development cycle Verification and validation will be completed by the Consultant providing a detailed report of all checks performed. The spreadsheet based tools should be subject to independent audit (to be organized by the Consultant); the independent audit team should be completely separate from the team completing the development, validation and verification. Formal acceptance will be completed by DRFIP. Tasks under this consultancy are set out in Table 1 with further detailed requirements outlined in Annex 1. Table 1: Tasks under this consultancy TASK I: EMERGENCY FUNDING ASSESSMENT TOOL The purpose of this tool will be to (1) estimate the potential emergency response need following a natural disaster event based on historical loss information (e.g. based on fiscal losses or number of people affected); and (2) quantify the fiscal funding gap based on the funding that is assumed to be available immediately following a disaster. TASK II: RISK FINANCING EVALUATION TOOL The purpose of this tool will be to evaluate the cost effectiveness of DRF instruments and strategies in funding costs/losses based on a given risk profile. The tool will conduct economic and financial analysis of DRF instruments (based on default strategies or strategies specified by the user).
TASK III: RISK TRANSFER OPTIMIZATION TOOL The purpose of this tool will be to estimate the optimal combinations of risk transfer and risk retention (from a purely financial/statistical perspective) to finance a defined emergency response need. This tool could either be standalone or integrated within the Task II deliverable, as an additional feature. TASK IV: FINANCING CRISIS RESPONSE TOOL The purpose of this tool will be to estimate the financial cost of rapidly making payouts to households affected by a crisis. The outputs from this tool should be in a format which enables easy transfer into the tools developed under TASK II and III above. Additional tasks may be added to this contract as an extension if and when required and agreed between the DRFIP team and the consultancy. Sub-Tasks for Tasks I, II, III and IV outlined in Table 1 are outlined in Table 2. Table 2: Sub-Tasks under this consultancy (for each Task I to IV) Sub- Task. A1 A2 A3 A4 A5 Sub-Task Review of existing tools and methodologies Specification development Data compilation Tool development in a spreadsheet Verification and validation of spreadsheet tool. Details PHASE A The DRFIP team will provide the Consultant with relevant tools and analysis developed internally. Specifying the problem and the most appropriate tool to address this, is the most challenging part of this process. The Consultant will work collaboratively with the DRFIP analytics team to develop an appropriate specification for the Excel based tool. The specification delivery by the consultant will include a word document and a mocked up Excel version of the tool input and output interface. A database of suitable historical event data will be compiled by the Consultant from available public information (from one or more sources). A database of suitable macroeconomic data will be compiled by the Consultant from available public information (from one or more sources). A tool will be developed in MS Excel. The MS Excel tool should be developed using best practice in coding, user interface design and experience. Delivery will include both alpha and beta versions of the MS Excel tools for user testing. Verification and validation of the tool will be completed by the Consultant. The consultancy is expected to follow international best practice and standards in verification and validation of tools build in MS Excel. There will be a feedback loop to sub-task 4. to fix for any issues identified as part of this sub-task.
Sub- Task. A6 Sub-Task Independent Audit, User Testing and Formal Acceptance PHASE A Details The independent audit team (organized by the consultant) should be completely separate from the team completing the development, validation and verification. Output from the independent audit to be delivered to the DRFIP team. User Testing will include testing by both WBG teams and client countries, to be organized in collaboration with the DRFIP. There will be a feedback loop to sub-task 4. to fix for any issues identified as part of this sub-task. Formal acceptance will be completed by DRFIP following the independent audit and user testing. This will include the following: Methodology document including (but not limited to) full details of logic, assumptions, data, and reliances and limitations. User manual/guide. Guidance notes to include details of assumptions used and tool limitations. A7 Development and delivery of Verification and validation report (documenting the associated Tool documentation functional tests performed and user acceptance tests). Data maintenance plan for the historical event and macroeconomic databases. Training material, including o PowerPoint slides o Working examples for training sessions This sub-task should be completed in concurrence with subtasks 2-6 above. A8 Tool delivery and training The Consultant will deliver the tools along with all associated documentation. The Consultant will deliver one half day training session per tool to WBG staff not involved in the project. A9 Post Delivery Support Support should be available for 6 months post-delivery. 3. Project timeframes, Payment Schedule and Delivery Dates The duration of the contract will be for 13 months from mobilization. Project delivery dates are outlined in Table 3; a more detailed project plan and staffing plan will be agreed with the Consultant within 2 weeks of contract signing (with drafts to be discussed during an inception meeting when the contract is signed). The Consultant is expected to conduct at a minimum three full-day meetings during the course of this project to present and discuss the status, progress and challenges of the assignment. In addition, training sessions on outputs for the World Bank should be arranged, based on feedback from the World Bank and
agreed during the inception phase. These meetings could be held at the World Bank offices in Washington DC or at the consultant s location; the location can be agreed between the consultant and DRFIP. The cost for travel by the firm to Washington DC will be borne by the DRFIP. Table 3: Project delivery dates PHASE A Sub-Task Num. Sub-Task Description Timeframe A1, A2 Review of Existing Tools To be completed within 1 months of Specification Development contract inception. A3, A4 Data compilation To be completed within 3 months of Tool Development in a Spreadsheet contract inception. A5 Verification and validation of To be completed within 4 months of spreadsheet tool contract inception. A6 Independent Audit, User Testing and To be completed within 6 months of formal acceptance contract inception. A7 To be completed in conjunction with subtasks A2-A6 above. Development and delivery of associated Tool documentation and Final completion within 7 months of Maintenance Plan contract inception. A8 Tool delivery and training to WBG To be completed within 7 months of team contract inception. A9 Post Delivery Support For 6 months post delivery Table 4: Payment Schedule Sub-Task Num. Consultancy Payment Schedule A1 A4 45% of total consultancy A5 A8 45% of total consultancy A9 10% of total consultancy For additional sub-tasks added under this contract, deadlines and payment schedules will be determined at the point of extension of contract and confirmed as updates to Tables 3 and 4. 4. Technical Proposal Requirements and Selection Criteria The Consultant will have proven expertise in developing user friendly interactive MS Excel based tools for use in financial decision-making. The Technical Proposal shall be presented electronically in Adobe PDF format, in English. The main body of the Technical Proposal, including figures and tables, shall describe the project team, its organization and qualifications, and a technical and management plan for performing the project including a detailed task and subtask schedule and statement of person-months per task and subtask. A title page, signed cover letter indicating the contact person, abstracts of one page and table of contents shall begin the Technical proposal. Key personnel CVs may be appended. The body of the Technical Proposal shall not exceed 30 pages excluding the cover page, abstract, table of contents, and CVs. Please note the following items that should be included in the bidding documents:
1. The Consultants bid should price each task (I IV) separately as tasks could be negotiated separately; 2. The Consultant should provide unit costs for consultants and activities which will be used in the event of contract adjustments or additional work; 3. An appendix detailing the proposed methodology for conducting validation and verification is required, consistent with international best practice in the relevant domains of: finance and/or actuarial science; and software/coding for quantitative tools within Microsoft Excel; and 4. An appendix providing detailed examples (and visualizations) of interactive tools previously delivered in spreadsheet format [noting UI/UX is paramount]. Technical criteria for selection, and associated weights, are indicated in Table 5. Table 5: Technical criteria for selection Technical Criteria Weight (%) Technical qualifications and expertise in complex risk modelling and financing analysis, including working in a probabilistic framework. 20 Technical qualifications and experience in the development of robust interactive spreadsheet based tools for non-technical users and for use in an operational context 30 Specialist experience in user experience and user interface design 20 Proposed process for detailed and documented verification and validation process 20 Project Management Plan and project manager qualifications and experience 10 Total 100 The firm will be selected following the World Bank s Guidelines: Selection and Use of Consultants by the World Bank for Operational Purposes. Minimum qualifications of the firm to be selected for the required assignment include: 1. Technical qualifications and expertise in complex risk modelling and financing analysis, including working in a probabilistic framework; 2. Proven experience in the development of robust software or other quantitative tools for nontechnical users and for use in an operational context; 3. Expertise in user interface (UI) and user experience (UX) design; 4. A strong project management function with a proven track record of delivery; 5. A staffing plan and skills mix that best represents implementation of the activities listed in the TOR; 6. Demonstrated experience working with different stakeholder communities to integrate knowledge and expertise into a product; 7. Excellent documentation and presentation skills required along with demonstrated experience in writing reports/user-guides of a technical nature suitably translated for non-technical audience; 8. Experience of working with and in developing countries is an advantage but not a prerequisite; 9. Scientific and technical background in natural hazard (and climate) risk and disaster risk management preferable, but not essential.
5. Resources to be provided to the Consultant The DRFIP team will provide the following support to the Consultant for the purposes of this assignment as appropriate and relevant to the delivery of the individual tasks: Prior tools and documentation produced by the World Bank to support DRF decision-making relevant to each tool identified under tasks; Expert advice and inputs from the DRFIP and other World Bank team actuaries, catastrophe risk modelers and software engineers, as required. This includes collaboration on specification development and development of methodologies. The World Bank team will also review and signoff the proposed Verification and Validation process. DRFIP will provide resources for user testing and formal acceptance of the tools. These collaborations will take the form of discussions; the Consultant will be responsible for all production tasks. [Note, some user testing and validation will be completed by DRFIP but this should not reduce the validation and verification, independent audit or user testing completed by the consultant] 6. Implementation Arrangements and Accountability The Consultant shall directly report to the World Bank task team lead for this project. 7. Delivery and Licensing All reports shall be provided in Microsoft Word or PowerPoint format to be agreed with the consultant. All tools will be provided in a format agreed at project outset in writing. The outputs produced under this assignment will be owned by the World Bank, which will become the principal institution to own and maintain the data for its own purposes. All outputs (including code) should be copyrighted to the World Bank Group. 8. Warranty Period The warranty period shall last for one year after the final delivery date. Contingent upon normal, diligent use on the part of DRFIP, the Consultant shall, at no additional cost, rectify errors and defects, replace defect parts of equipment and rectify errors in software governed by this Agreement and which the DRFIP has complained about before the expiry of the warranty period. No damages or other remedies for breach of contract may be claimed for defects that are rectified pursuant to the warranty. 9. Language All communications and outputs shall be in English, unless stated otherwise for sub-tasks added subsequently to this contract. The Consultant is responsible for all translation work.
Annex 1: Requirements of Task I, II, III and IV TASK I: EMERGENCY FUNDING ASSESSMENT TOOL The purpose of this tool will be to (1) estimate the potential emergency response need following a natural disaster event based on historical loss information (e.g. based on fiscal losses or number of people affected); and (2) quantify the fiscal funding gap based on the funding that is assumed to be available immediately following a disaster. A database of historical loss information will be prepopulated and sit behind the tool and be available for selection (e.g. by country and by peril) [Note: Version 1 of the tool should capture at a minimum 4 key perils.] The database will be compiled based on publicly available data, on e.g. the number of people affected historically by disasters, from one or more sources e.g. historical data from EM-DAT and Desinventar. The data will initially be cleaned by the consultant. [The database should be structured such that it can easily be updated over time if/when improved data becomes available.] A user will have the option to overwrite the default historical data (pulled down from the prepopulated database) where more appropriate data exists. Historical data will be scaled/trended to present day using macroeconomic inputs. A database of macroeconomic inputs needed for scaling/trending will also be pre-populated by the consultant from public sources and sit behind the tool. A risk profile will be fitted to the historical data selected by the user, subject to a minimum number of data points. An assumption for the emergency response cost per person will be required. The tool will include a default assumption which should be capable of being overwritten by a user. A user will input the funding that is currently or might be available immediately following a disaster. This will be evaluated against the fitted distribution to show the funding gap. The outputs are expected to include at a minimum visualization and qualitative descriptions of the following: o Estimated annual emergency response liability (average and realistic disaster scenarios e.g. 1 in 5, 10, 20, 50 years) alongside the most recent and most severe events from the historical dataset; o Estimated emergency response liability over a 5/10-year time period; o Uncertainty in the estimated emergency response liability (plotting a confidence interval chart with different scenarios under different distributions); o Funding gap for both the estimated annual emergency response liability and for the most recent and severe events from the historical dataset; and o As an advanced user option, the full loss simulations will be output to an Excel file (e.g. 10,000 simulations of the risk distribution) Tools to include qualitative statements about the reliability and uncertainty of the results due to data, distribution choice etc. [Since the calculations will be automated no expert judgement on the choice of risk distribution will be possible by the user] Specific consideration by the consultant in conjunction with the DRFIP team will need to be given to the following points: Reliability of publicly available information and appropriate cleaning of such data; Fitting of multi-peril versus single peril losses; Appropriate tests to use to choose the best fit distribution from defined set of distributions; Choice of pre-defined distributions for different perils and multi-perils; Advanced user functionality to fit a distribution for the tail (e.g. extreme event theory, Pareto distribution);
Choice of realistic disaster scenarios (simple reference scenarios either country-specific such as the worst historical event to date and the losses it would lead to today, or return-period specific such as 1-in-30 or 1-in-50 which would be highlighted along the distribution and would allow simple, visual comparison between countries). TASK II: RISK FINANCING EVALUATION TOOL The purpose of this tool will be to evaluate the cost effectiveness of DRF instruments and strategies in funding disaster costs/losses based on a given risk profile. A loss risk profile will be an input to the tool. [This could be a loss distribution fitted under Task I or from a user input (e.g. input of 10,000 loss events).] A framework 1 for evaluating sovereign DRF strategies was published in 2016 and this framework could be used as the basis in evaluating specified DRF strategies. The tool will analyze the cost of financing emergency response needs under 1 3 default alternative DRF strategies. The user will input the DRF strategies from a selection of available DRF instruments including reserves, contingent credit, budget reallocations, parametric insurance and ex-post debt. Each default DRF strategy can include one DRF instrument or a combination of instruments, and will be pre-defined by the consultant, in agreement with the DRFIP team. The methodology needed to evaluate the cost of financing the response need will be based on a number of assumptions. Default assumptions will be pulled from a pre-populated database. The consultant will compile the database of appropriate macroeconomic assumptions. The user will have the option to vary some of the key assumptions from their default values under an advanced user function. The outputs are expected to include at a minimum visualization (including graphical representation) and descriptions of the following: a) Annual liability including averages and realistic disaster scenarios; b) Uncertainty (e.g. fan charts) of estimated liability over a 5/10-year time period (average and realistic disaster scenarios); c) Funding gap for the annual liability under each selected DRF strategy; d) Average annual cost of financing the liability under each DRF Strategy; e) Cost of financing a small, medium and large disaster under each DRF Strategy. The severity/frequency of small, medium and severe disasters will be pre-defined based on return periods. A user will have the option to select alternative events of different frequency/severity; f) Average annual cost savings of DRF strategy B & C relative to DRF strategy A; g) Annual cost savings of DRF strategy B & C relative to DRF strategy A for a small, medium and severe events; h) Sensitivity results (of d, e, f and g) for the key parameters/assumptions. Specific consideration by the consultant in conjunction with the DRFIP team will need to be given to the following points: Instruments (type and size) to include in 3 default DRF strategies; Choice of default realistic disaster scenarios Default macroeconomic assumptions needed to model the opportunity cost of each instrument Different user access levels 1 A Framework to Evaluate Sovereign DRF Strategies was developed; the paper can be found here. This framework was then applied in five practical case studies; the paper can be found here.
TASK III: RISK TRANSFER OPTIMIZATION TOOL The purpose of this tool will be to estimate the optimal combinations of risk transfer and risk retention (from a purely financial/statistical perspective) to finance a defined liability. This tool could either be standalone or integrated within the Task B deliverable, as an additional feature. A framework 2 for evaluating sovereign DRF strategies was published in 2016 and this framework could be used as the basis in estimating optimal combinations of risk transfer and risk retention. The optimal solution(s) will be derived as a function of risk appetite inputs (for which a default characterization will be used, such as no losses more than 1% of GDP 90% of the time for instance) The optimal solution(s) could be presented as a graph showing attachment point as a function of uncertainty tolerance. Discussions will take place with the DRFIP team to agree on the actual definition of optimal solutions to be used and on the format of its representation. The risk transfer will be based on simple parametric insurance contracts, for which default assumptions (which can be overwritten) will be pre-populated and used (e.g. cost of diversified capital). The risk retention assumptions will be based on pre-populated macroeconomic data. The liability will be defined by the user and will be subject to a maximum limit. TASK IV: FINANCING CRISIS RESPONSE TOOL The purpose of this tool will be to estimate the financial cost of rapidly making payouts to households affected by a crisis (referred to as a crisis response mechanism). The country for which the crisis response mechanism will apply will be selected from a drop-down menu. The user will then be prompted to enter in the number of households they wish to protect with the crisis responsive mechanism in each of the defined administrative boundaries (level II), if a crisis was to occur. The user will enter the amount of resources (in monetary terms) to be provided to each individual household in the event of a crisis (e.g. $50). The user will define the perils covered by the crisis response mechanism based on a default list within the tool. The complete list to be captured by the tool includes: drought, flood, earthquake, cyclone, tsunami, volcano; however this will vary depending on each country. The user will define under which conditions for each peril the mechanism will respond to a crisis as follows: o For each peril, the user will be prompted to enter for what type of crisis they wish to trigger a response of the mechanism. For example, options could be given to the user for a light (1 in 3 year event), moderate (1 in 5 year event), or severe (1 in 10 year event) crisis. o For the selected triggering event, the full number of HHs entered in the province will receive the value of the crisis assistance Hazard data will be gathered at admin level II for each of the countries for which the model will operate. In the initial phase, it is proposed to focus on 10-15 countries (to be provided by the DRFIP) for the analysis. This hazard data must have historical values for 15-20 years, as well as capture the geospacial correlation of losses over time. This will be gathered by DRFIP from globally available sources (e.g. ThinkHazard). 2 Through a DRF Impact Appraisal Project recently completed by the DRFIP, a Framework to Evaluate Sovereign DRF Strategies was developed; the paper can be found here.
Using the hazard data, the model will generate 10,000 simulations of payouts for the selected country at admin level II. For each of these simulations, should a crisis greater than the triggering threshold be reached, it will be assumed that a payout will be made to the entered number of households in that sub-region The model will then output a distribution of the $ value of the losses (using the user inputs). The outputs are expected to include at a minimum visualization and qualitative descriptions of the following: o Estimated annual crisis response cost (average and realistic disaster scenarios e.g. 1 in 5, 10, 20, 50 years); o Estimated crisis response liability over a 5/10-year time period; o Uncertainty in the estimated emergency response liability (plotting a confidence interval chart with different scenarios under different distributions), and; o As an advanced user option, the full loss simulations will be output to an Excel file (e.g. 10,000 simulations of the risk distribution at admin level II) Tools to include qualitative statements about the reliability and uncertainty of the results due to data, distribution choice used for the simulation etc. [Since the calculations will be automated no expert judgement on the choice of risk distribution will be possible by the user] Specific consideration by the consultant in conjunction with the DRFIP team will need to be given to the following points: Reliability of publicly available hazard information and appropriate cleaning of such data; Fitting of multi-peril versus single peril losses; Appropriate tests to use to choose the simulation technique (bootstrapping, distribution fitting, Bayesian methodology); Advanced user functionality to fit a distribution for the tail (e.g. extreme event theory, Pareto distribution).