Insights September 2015 Financial Modeling Topical articles, tutorials, case studies and news Welcome Welcome to our August edition of Insights Financial Modeling, Americas. We begin with a brief announcement of Towers Watson Unify, a new software solution that enables you to integrate your financial modeling and reporting applications into an automated and governed process for greater efficiency and control. The case study in our next article looks at how we implemented RiskAgility FM at Samsung Fire & Marine Insurance (SFMI) in Korea. It describes some of the challenges faced by SFMI and how RiskAgility FM s modeling capabilities overcame them to provide an integrated modeling process at the forefront of financial modeling. The following article considers implications for liability models in MoSes or RiskAgility FM when adopting a Least Squares Monte Carlo approach to proxy modeling. Finally, we introduce Star ESG RN, a new addition to the STAR ESG product suite that produces risk-neutral economic scenarios. With many clients now starting to migrate to RiskAgility FM, let me draw the attention of MoSes users to our statement on Microsoft FoxPro support on page 15. Dominique Lebel Americas Financial Modeling and Reporting Product Group Leader (Life) Towers Watson Unify: Integration. Automation. Governance. Towers Watson is proud to announce Towers Watson Unify, a new software solution that enables you to integrate your financial modeling and reporting applications, including Towers Watson and third-party software products, into an automated and governed process. What would this mean for your company? Seamless integration allows you to have easier and more powerful configuration, and control of individual software products used in your financial modeling and reporting processes. Automated workflows can be executed at the press of a button, scheduled or triggered by external systems for improved efficiency and productivity. Workflows can also be defined to require review and intervention at key checkpoints by authorized staff, and these checkpoints can be conditional based upon predefined thresholds. Best practice governance, including reviews, approvals, conditional processes, user authentication and audit logs, ensures that your financial modeling and reporting processes are conducted properly and tracked thoroughly. Towers Watson Unify provides exceptional capabilities for integration, automation and governance to raise your company to a higher standard of risk and capital management. This allows your highly trained staff to spend less time on routine production tasks and more time using the results to deliver value-added insight and competitive advantage to business stakeholders. Let us show you how your company can rise to a higher standard of performance in risk and capital management with Towers Watson Unify. Visit our website, contact your local Towers Watson representative or email us to learn more. In this issue 1 Towers Watson Unify: Integration. Automation. Governance. 2 RiskAgility FM Implementation Case Study 6 Least Squares Monte Carlo Implications for Projection Models 10 STAR ESG RN Risk- Neutral Economic Scenario Generation 15 End of Microsoft Support for FoxPro 16 Financial Modeling News
RiskAgility FM Implementation Case Study Overview SFMI has historically used MoSes for liability cash-flow projections for pricing, business planning and liability adequacy testing purposes. However, asset modeling of long-term insurance sat outside of the MoSes system. SFMI wanted to migrate from their existing systems to a RiskAgility FM asset/liability modeling (ALM) system to enable greater integration with their processes and to be at the leading edge of financial modeling in this area. This article describes the migration to RiskAgility FM. Background of the Migration to RiskAgility FM SFMI is the largest property & casualty (P&C) insurer in South Korea, accounting for 25% of the market. SFMI provides comprehensive services including auto, fire, marine and liability insurance, as well as long-term insurance. Long-term insurance is one of the unique features of Korea s P&C market. The base policy is a long-term insurance policy that covers accidental death or disability. A majority of policies have riders attached, and these riders can cover diagnosis, hospitalization or operation due to disease, long-term care or critical illness. Long-term business represents more than 60% of SFMI s total block of written premium. Given its magnitude, SFMI has an ongoing need to model the assets backing their liabilities accurately and efficiently. SFMI uses a number of tools to manage the longterm business, including setting the strategic asset allocation, matching the asset and liability duration, and forecasting the future available capital. In order to use these tools, SFMI needs to have a financial system that projects both asset and liability cash flows simultaneously, including the interaction between them. SFMI s system for valuing their long-term business used MoSes for the liability cash-flow projections but did not use MoSes for projecting asset cash flows. One of the main motivations for moving to RiskAgility FM was to provide seamless integration of model asset data, liability data and assumption tables stored in the SQL Server with the modeling system, which RiskAgility FM supports. Modeling Overview Towers Watson provided SFMI with a standard RiskAgility FM ALM Project (in MoSes terminology, a Standard Application, which was called the central asset/liability model, or C-ALM ) and modified it to model SFMI s block of long-term business. While developing the ALM model, there were some challenges that made the modeling more complicated. This article focuses on the asset side of the ALM, since the liability modeling side was still in development at the time of the writing of this article. Several challenges surfaced, including: Too many model points for one specific asset class (over 80,000 model points) Foreign exchange hedge account implementation We now turn to some of the RiskAgility FM features used to overcome these challenges. 2
RiskAgility Standard ALM Project SFMI s RiskAgility FM ALM Project is an extension of the original C-ALM and uses many advanced new input features of RiskAgility FM to simplify the reading of data. We illustrate how reading data is simplified by comparing MoSes C-ALM and RiskAgility FM C-ALM data readings: Data MoSes C-ALM uses.dbf files to read the data. RiskAgility FM links directly to an Excel workbook or a database through the Input Manager, meaning conversion tasks are no longer needed. Data can be managed more efficiently and better organized because of this direct linkage. In addition, if the ALM Project reads in the data from a database, then the risk of managing data can be transferred to a third party. New input features have been used and further modifications performed to automate the data input processes. Model structure The MoSes C-ALM required a number of steps to read in data and model structure construction using user defined, sub-model arrays. This needed to be coded within the model, which could become complicated. The arrayed sub-model in RiskAgility FM s ALM template is set as data-driven. By changing from user defined to data-driven, the multiple steps required to construct arrayed sub-models in the MoSes C-ALM Standard Application are no longer necessary. Assumptions Under the MoSes C-ALM Standard Application, the assumption tables had to be in.tbl format. Under RiskAgility FM, most of the assumptions are changed to be read in directly from Excel. This linkage is controlled through the Input Manager, eliminating the additional step in MoSes of exporting the assumptions in.tbl format, which makes it easier to manage assumptions. The next two sections of this article discuss how RiskAgility FM s features overcame the challenges outlined above to meet SFMI s aim of having an accurate and efficient ALM process. Seriatim Run Solution to Having Too Many Model Points for One Specific Asset Class One of the types of assets that SFMI holds is loans. SFMI has issued thousands of existing loans, and while the total block of loans contributed to a significant portion of the total assets, each individual loan s value was relatively small. Loans could not be grouped since each loan had its own characteristics: different interest rates depending on the economic conditions, different interest rate reset timing, and different prepayment penalties and repayment start times. RiskAgility FM s Standard ALM Project s projection mode is period-by-period processing (Figure 1). In this mode, each asset and liability model point is projected simultaneously for each time period. The next period is then projected after the current period s projection of all assets and liabilities is finished. When asset and liability cash flows are stored in memory, the model can do calculations related to the interaction between assets and liabilities in the current period. However, since all of one period s cash flows are stored in memory, period-byperiod processing is highly memory intensive. Figure 1. Period-by-period mode Time T=0 T=1 T=2 T=3 T=4 T=5 Asset MP 1 Asset MP 2 Asset MP 3 Liability MP 1 Liability MP 2 Fund 3
Period-by-period processing is necessary when doing ALM modeling because of the interactions between the assets and liabilities, such as strategic asset allocation, setting the liability crediting rate according to asset returns and allocations, and dynamic lapses. In contrast, policy-by-policy, or seriatim, processing is used in a valuation model. In seriatim mode, a model projects the cash flows for one model point at a time (Figure 2). The next model point would be projected once the current model point is finished. Since a seriatim mode projection releases memory once the model point has finished, it is less memory intensive. Given SFMI has tens of thousands of existing loans, each of which has to be run individually, it was not feasible to run them under period-by-period mode. Towers Watson and SFMI solved this problem by assuming that the loans are held to maturity. This is a valid assumption since the loans are not actively traded assets (as they are not securitized). If the loans are assumed to be held to maturity, we can then leverage the seriatim projection mode to reduce memory usage and run time. In SFMI s RiskAgility FM model, the ALM Project uses RiskAgility FM s hybrid projection mode to optimize the model s efficiency. The following illustrates how the ALM Project works under the hybrid projection mode: In the initial stage, the existing loans are projected through seriatim mode using runseriatim functionality. The results of this run are stored in a SmartArray to be used in the next run. Then the parallel mode projection begins. Under this mode, assets other than existing loans plus the liabilities are projected. The cash flows for the existing loans are read in from the SmartArray populated during the previous step. Bundle Modeling Solution to Bundle Modeling Solution to Implementing the FX Hedge Account Korean regulations dictate that FX risks associated with foreign bonds owned by an insurer need to be hedged. To model this, we used a bundle model as part of the ALM Project. A bundle model is a submodel that combines multiple assets to act effectively as one asset by aggregating the cash flows from other models. The bundle model was one of the features in the standard C-ALM as a core function, and it was enhanced and modified for this project. Figure 2. Seriatim mode Time T=0 T=1 T=2 T=3 T=4 T=5 Asset MP 1 Asset MP 2 Asset MP 3 Liability MP 1 Liability MP 2 Fund 4
Under the standard ALM Project, different asset types are modeled in the different sub-models. It is possible to combine the different assets using the aggregator, which is the external function summing asset cash flows or asset values by categories or funds in a standard ALM Project. However, the resulting combination of assets will not act as one asset without processing it through a bundle model. A convertible bond offers a simple example of using the bundle model. It gives the bondholder the option to convert the bond into company stock at a predetermined date. A bundle model can combine a stand-alone bond and a stand-alone stock to model the convertible bond. The bundle model can be extended to other complicated assets such as structured assets, which are a combination of a straight bond with several derivatives. Alternatively, dynamic hedging for variable annuities can be modeled in the bundle model by combining the risky asset and the risk-free asset. SFMI hedges the FX risk of its foreign bonds using FX swaps. In SFMI s ALM Project, the foreign bonds are modeled in the bond model, and FX swaps are modeled in the interest derivative model. The bundle model combines these assets and constructs the FX hedges. The FX hedge bundle model allows SFMI to: Assess the effectiveness of the various FX hedging strategies. One of the strategies that SFMI considered was a roll-over hedging strategy. Rollover hedging is the strategy used when the foreign bond has a longer maturity than the FX swap, and hence, the FX swap needs to be repurchased under the economic and foreign exchange conditions at the time of its maturity. This means that SFMI is exposed to FX risk when making this repurchase. SFMI uses the bundle model to assess this risk. Pass information from one asset sub-model into another. Capital gains and losses due to FX changes for both the foreign bonds and FX swaps are calculated and then netted off. Any capital gains or losses unrelated to FX changes would be booked as unrealized capital gains or losses and would be realized only when they are sold. The bundle model stores information for each of its components. In this case, the model passes the bond s acquisition costs to the bundle model, which in turn passes it down to the FX swap model. The bundle model acts as a catalyst in passing the information from one sub-model to another. The bundle model should be projected before each of the bundled components is projected since it acts as a controller of each component. Project Success Global collaboration and the setting of clear, predefined objectives has led to the RiskAgility FM implementation project s successful run to date. Expertise on local business requirements came from Towers Watson s Korean, Hong Kong and German offices. During the system design phase, the German office provided the bundle model and a U.S. office provided information on hybrid projection mode practices. This phase of modeling took six months to implement. 5
Least Squares Monte Carlo Implications for Projection Models Over the past couple of years, Least Squares Monte Carlo (LSMC) has emerged as a solution for producing proxy models for complex products with options and guarantees. Towers Watson has developed software that helps modelers create these proxies and embed them into their liability projection models. This software is called RiskAgility Proxy Modeler, and some of its key features are outlined at the end of this article. This article considers the implications for projection models that need to produce the calibration inputs for LSMC. LSMC implementation can require that models produce over 40,000 real-world scenarios, each of which requires multiple inner-market consistent scenarios. Every one of these scenarios will have different economic and demographic assumption inputs and present many challenges to deliver them. We will briefly go through the LSMC process before a more detailed analysis of the modeling to put the changes that are required to the liability model in context. The LSMC Process An Overview The objective of LSMC is to construct a proxy function that can be used to value the balance sheet under arbitrary stresses or to value complex products such as those with options and guarantees. This proxy is a function of risk factors, where the function is found using regression techniques. The risk factors should be chosen such that they describe movements in the liability adequately to ensure that the calibrated proxy gives good approximations to the true liability values. Therefore, all factors that affect the liability value should be included in the set of risk factors. Risk factors can be divided into two broad categories: market risks (e.g., yield curve principal components, and equity volatility) and nonmarket risks (e.g., mortality). Each of the sets of risk factors, where a single set of risk factors is described as an outer scenario representing a single real-world state, should then be run through the liability model. All outer scenarios are contained within the Outer Scenario File (Figure 3). To simulate the liability model result for an outer scenario, the market risk factors are used to calibrate an economic scenario generator (ESG). The ESG then projects a number of risk-neutral scenarios (the inner scenarios ) into the future, producing an inner scenario file. The liability model must then be run using the inner scenario file and the nonmarket risks from the outer scenario as inputs to value the liability for this state in the liability model. The output from these liability model runs can then be regressed against the inputted risk factors to produce the proxy function (Figure 3). Outer scenario: a set of risk factors representing a single real-world state Inner scenario: a single ESG scenario produced by calibrating to the outer scenario Figure 3. LSMC modeling process Outer scenario file: x40,000 combinations of risk factors Market factors Nonmarket factors ESG x2 inner scenarios x2 inner scenarios x2 inner scenarios. (x40,000) Liability model Proxy function 6
Running Through All Sets of Inner Scenario Files There are two possible approaches: Approach 1: Use a single nested stochastic scenario file combining all of the small inner scenario files into one aggregated file. This means that no code changes are required to the liability model for the scenario files. Approach 2: Use a separate file for each inner scenario, and modify the model to loop over the scenario files. The Iteration Map: Mapping Model Iteration Numbers to Inner and Outer Scenarios The recommended approach to ensure that the model runs through all sets of inner scenarios is to create an iteration map to map the model iteration number to an inner and an outer scenario number. An iteration map allows the user to vary the number of inners for each outer, as we have shown in Figure 4. For each of the two approaches to running all inner scenario files (outlined above), the iteration map has a different role: The user will need to reassign each of the inner scenario numbers upon aggregation into one file under Approach 1. The iteration map can be used to achieve this end. Under Approach 2, the iteration map will be used to identify the market-consistent scenario to use. This is achieved by looking at the outer scenario number to find the ESG run output file to reference (which is based on the set of economic risk factor stresses in the outer scenario). The correct scenario to pull from this file is the inner scenario number. For both approaches, the iteration map can be used to identify the nonmarket risk factor stresses to be used. This is achieved by referencing the outer scenario number row in the outer scenario file. Modifying Your MoSes/RiskAgility FM Model for LSMC The recommended approaches for RiskAgility FM and MoSes are different. RiskAgility FM For RiskAgility FM, we recommend using the single-file method (Approach 1). The iteration map referenced above can be used to assign scenario numbers in the combination process. The combined file should then be set as an external source in the Input Manager. The scenario map and the outer scenario file should also be set as external sources in the Input Manager. These can then be used for the derivation of stressed demographic assumptions. MoSes Due to its reliance on FoxPro, MoSes cannot read files larger than 2GB. Therefore we do not recommend the single-file approach. Instead, the looping method (Approach 2) should be used. Adapting a MoSes model to work with the looping method is simple. Only two columns within the code need be modified and an EXTERN added to get it working; all other existing code will work as before. These two columns are the startup column and the initialize column. Looping over the scenario files requires that we can open and close the schedule class because the model will need to open and close the c. 40,000 files in sequence. This requires replacing the existing schedule class in the code with a new class, Closeable Schedule, to enable the model to complete this step. Figure 5 shows that the start-up column will need to be modified once code for the Closeable Schedule class has been added. Figure 5. Modified start-up column Figure 4. Iteration map 7
The code for this class, which inherits from the schedule, can be added by navigating to Utilities- >External DLLs->Externs. A window named External DLL Settings will appear, and the user can input the code necessary to create a closeable schedule. This code is available from us upon request. Figure 6. Existing code works as before As the new class inherits from the old schedule class rather than having been created from scratch, all of the syntax referencing columns from the schedule file in the application is still valid (Figure 6). Therefore, all that remains is to ensure that the MoSes application picks up the values for the iteration correctly. The following code (Figure 7) in the initialize column achieves this by referencing the iteration map table referenced above. Setting Assumptions for Nonmarket Risk Factors The nonmarket risk factors, namely any demographic factors, may require some additional assumptions set up before the liability model can be run. Consider the case of a mortality risk factor where a 10% stress has been applied in the outer scenario file. If the liability model is currently set up to read in directly from a mortality table without applying any multiplicative adjustment, then it would be difficult to apply the stress without recreating the mortality table for every different outer scenario. This is both impractical and time-consuming. The most natural approach to this problem is to build the functionality to stress the assumptions based on the outer scenario into the liability model. For insurers yet to do this, although this requires some investment in modeling, this investment does lead to significant benefits in simplifying and automating the capital modeling process. Figure 7. Initalize column example 8
Preparing for LSMC If you are considering moving to a LSMC solution, which should provide robust results for valuing complex products and valuing arbitrary stresses to their balance sheet, then there is a need to consider the implications on adaptation requirements to the liability model. This article demonstrates that these adaptions may not be as difficult as anticipated. Indeed, many companies could benefit from implementing some of these techniques even if they are not immediately considering using LSMC because it would facilitate a future transition. About RiskAgility Proxy Modeler RiskAgility Proxy Modeler encapsulates all of the processes listed in Figure 3 apart from the liability model section, where it is designed to integrate with both RiskAgility FM and MoSes. Key Features of RiskAgility Proxy Modeler Automates curve fitting and LSMC Automatic algorithms calculate terms included in the loss function Eliminates expert judgment from the selection of risk stresses and loss function fit Automates validation metrics Integrates with Towers Watson software products RiskAgility EC, RiskAgility FM and STAR ESG to form a fully integrated capital modeling suite Works with both profit and nonprofit business Flexible modular design allows out-of-model adjustments, and ability to mix and match with existing software investments including third-party ESGs To receive a copy of the code necessary to create the closeable schedule class, please contact: patrick.penzler@. 9
STAR ESG RN Risk-Neutral Economic Scenario Generation Overview of STAR ESG RN In this article, we look at STAR ESG RN, the latest addition to the STAR ESG product suite, and how it can be used for risk-neutral economic scenario generation. STAR ESG is a sophisticated software toolkit providing financial modeling and risk analytics to the insurance, pension and banking communities. It is used by our clients to manage and report on risk exposures in excess of US$4 trillion and forms a key component of the advice driving our fiduciary mandates. STAR RN adds the capability to generate risk-neutral, market-consistent scenarios. Risk-neutral scenarios are a key input to any application requiring marketconsistent valuation of options and guarantees, including: Solvency II balance sheet MCEV U.K. realistic balance sheet Hedging Guarantee pricing Economic Scenario Generation Generating stochastic economic scenarios requires a number of building blocks. One fundamental decision is the selection of a model for each asset class of interest. There are many such models, each with its own pros and cons. These models must be calibrated to market data, which for some model types can be a complex optimization problem. Next, to generate stochastic scenarios, we need to use the model to project forward each variable by Monte Carlo simulation. Finally, we need to validate the scenarios to confirm that an acceptable fit has been achieved. For STAR ESG RN, we have implemented advanced technical models for the major asset classes, including interest rates, credit spreads, equities and property. At each stage of the process of generating scenarios calibration, simulation and validation a number of mathematical techniques are used to improve performance. Figure 8 illustrates this process and the main steps involved at each stage. Together, these aim to provide: Fast calibration High-quality fit to a wide range of market data Realistic simulations Rapid convergence to targets Figure 8. End-to-end scenario generation process in STAR RN Market data Calibration Simulation Calibration: Powerful nonlinear optimization routines and sophisticated option-pricing formulas allow straightforward calibration of multifactor models. Users can apply weights to optimize the fit. The resulting calibration can be viewed both graphically and numerically. See Figure 9. Validation Market data High speed Automated Weights fitting Optimizes model parameters Illustrates (expected) fit quality Runs stochastic differential equations Convergence testing on the fly (resampling) Martingale test Volatility testing (i.e., fit to option surface) Correlation testing Scenarios 10
Figure 9. Example inputs Simulation: Scenarios are generated by Monte Carlo simulation of the models implemented in STAR RN, using variance-reduction techniques to minimize simulation and discretization errors. 11
Validation: The user interface includes a wide range of diagnostics, which can be viewed in flexible graphical or numerical formats. Some examples are shown below in Figures 10 and 11. Figure 10. Example Martingale test Figure 11. Example validation of volatility surface Validation of scenarios is an important step in demonstrating to management, regulators and auditors that the scenarios are sufficiently robust. All stochastic modeling involves a degree of approximation, and it must be demonstrated that these are within a reasonable tolerance. Having the validation output in the ESG interface allows users to refine the scenarios where necessary and rapidly assess any errors. 12
Model Implementation in STAR RN The underlying mathematical models used in STAR RN have been selected to make use of existing academic research and provide high-quality scenario sets. For interest rates, STAR RN uses the (SABR) Stochastic Alpha Beta Rho implementation of the LIBOR Market Model (LMM) for interest rates. This is an evolution of LMM, which is one of the most popular frameworks for modeling interest rates, meaning that it is straightforward to transition from an LMM interest rate model to a SABR interest rate model. SABR solves a number of problems that are inherent in other versions of the LMM approach: It can ensure nominal interest rates do not become negative, a real benefit because scenarios are more realistic and your liability model will not produce unexpected results when negative interest rates are encountered. The scenarios are better behaved because the explosive interest rate issue that commonly affects other LMM implementations is significantly reduced. This issue often causes inaccuracies in valuations and may cause liability models to fail. SABR LMM avoids the need for manual adjustments, increasing the accuracy of valuations. Equity and property assets are modeled using the stochastic volatility Heston model, which combines excellent fit-to-market option data across both term and moneyness dimensions with realistic dynamics. Corporate bonds are modeled using the Arvanitis, Gregory, Laurent model. This model extends the popular Jarrow, Lando, Turnbull framework and allows stochastic modeling of credit spreads for different ratings as well as migrations between ratings. STAR RN includes variance reduction techniques that reduce the errors that arise from using Monte Carlo simulation. This improves the quality of the stochastic scenarios, leading to more accurate valuation results. It also has the benefit that the ESG validation tests, such as the Martingale test, are more likely to pass. Consequently, it is easier to gain approval from auditors, regulators and senior management. These techniques include: The ability to run a finer time step within the model to reduce discretization error without affecting the desired output A feature that allows the user to specify a maximum tolerance for error in Martingale convergence tests, which the ESG will target by carrying out convergence tests during a run Antithetic sampling of random numbers 13
Mitigating Explosive Interest Rates in SABR LMM The LMM framework can be implemented with various assumptions for the underlying distribution of the risk drivers. A common assumption in basic versions of LMM is a lognormal distribution, which can lead to explosive interest rates. In this phenomenon, interest rate paths in certain scenarios increase to exceedingly high values, causing problems for downstream calculations. SABR LMM substantially reduces this issue. Figure 12 illustrates this by comparing the maximum interest rate observed in two sample sets of scenarios, one generated using a standard lognormal LMM and the other using SABR LMM. Figure 12. Maximum projected 25-year spot rate (EUR calibration on December 31, 2013) Years Standard LMM SABR LMM 10 31% 11% 20 522% 14% 30 1382% 15% 40 2384% 15% 50 >10^70% 16% The distribution of interest rates generated by SABR LMM is shown by this fan chart. 14
Licensing STAR RN Our clients have two ways to access STAR RN: License the ESG software directly, which is deployed with quarterly calibration updates, and comprehensive training and support. Full documentation is provided, including a user guide, detailed technical manual and calibration document describing the models, assumptions, limitations, calibration and validation processes. Purchase pre-generated simulations, which are tailored to each client s specifications. These are accompanied by documentation showing the calibration and validation of the simulation set. The STAR ESG suite is supported by a team of more than 50 consultants who have academic and commercial experience in this field. Specializations range across economics and econometrics, capital and risk, actuarial advisory, asset pricing, trading, risk management, and systems engineering. An Integrated Solution Risk-neutral ESGs are becoming more widely used as a number of regulatory frameworks and insurers are starting to use marketconsistent balance sheets as one of their bases. Expectations for the quality and performance of ESGs are increasing, both from insurers and regulators. STAR RN offers an integrated solution to generating risk-neutral scenarios that will help to meet these challenges. End of Microsoft Support for FoxPro Microsoft extended support for Microsoft Visual FoxPro 9.0 Professional Edition ended on January 13, 2015. As this was the last version of FoxPro supported by Microsoft, Towers Watson can no longer source any security patches or fix any issues with FoxPro that may impact the operation of MoSes on current or future operating systems. Towers Watson will continue to maintain and support MoSes for the foreseeable future where FoxPro continues to operate as documented. Towers Watson released our new product, RiskAgility FM, in April 2014 (/riskagilityfm) to supersede MoSes. RiskAgility FM has no reliance on FoxPro: In fact, it is data agnostic, which allows for process gains in both inputs and outputs from your system. We have facilities in place to help you run your existing MoSes models in RiskAgility FM, take advantage of its new features, as well as a no-fee license-exchange program. For STAR RN, we have implemented advanced technical models for the major asset classes including interest rates, credit spreads, equities and property. At each stage of the process of generating scenarios calibration, simulation, validation a number of mathematical techniques are used to improve performance. 15
Financial Modeling News Viewpoints Q&A: The Rise of Risk Management as a Business Strategy Partner According to Towers Watson s Global Insurance ERM Survey of almost 400 insurers worldwide, insurers are most satisfied with ERM when they engage the risk management function as a business strategy partner. In this roundtable discussion, Towers Watson panelists share their insights based on experiences helping insurance companies advance their ERM programs. Principles-Based Reserving Is Closer to Going Live: Are You Ready? After over a decade of hard work, it now appears likely that principles-based reserving will go live on January 1, 2017. For life insurance companies, a new phase of work needs to begin, and in short order. Infographic: How to Prevail Over Today s Life Insurance Industry Challenges Life insurance industry challenges are looming, but with the right technology and financial models, top- and bottom-line growth can be improved. Emphasis A global magazine providing thought leadership for the insurance industry Contacts Software Support Visit https://software.support. or email software.support@. Holly Starkey holly.starkey@ Dominique Lebel dominique.lebel@ Carlos Gonzalez carlos.gonzalez@ Follow us on Twitter @towerswatsonins. About Towers Watson Towers Watson is a leading global professional services company that helps organizations improve performance through effective people, risk and financial management. With 16,000 associates around the world, we offer consulting, technology and solutions in the areas of benefits, talent management, rewards, and risk and capital management. Learn more at. Copyright 2015 Towers Watson. All rights reserved. TW-NA-2015-43821. August 2015. /company/towerswatson @towerswatson /towerswatson