Insights. Financial Modeling. Topical articles, tutorials, case studies and news

Transcription

1 Insights September 2014 Financial Modeling Topical articles, tutorials, case studies and news Welcome We open this issue with an article on the use of predictive modeling for the life insurance business. Having developed and refined these techniques over many years in the property & casualty (P&C) insurance space, we are now seeing an increased demand for assistance in implementing these techniques with life insurance companies. Continuing the previous edition s feature on RiskAgility FM (financial modeling), we have included a feature on migrating models from MoSes. This article provides tips to help modelers get started and to lay the groundwork for making the most of RiskAgility FM. Our final article outlines some of the new software support services that we are rolling out to clients to more efficiently solve any problems that you may encounter with our software suite. Dom Lebel Americas Financial Modeling and Reporting Product Group Leader (Life) GLMs in Life Insurance The use of predictive modeling techniques to refine the pricing of insurance risks has been standard practice in the P&C insurance sector for many years. However, life organizations have traditionally used simple one-way techniques and so, for many life insurance providers, predictive modeling techniques represent a new era of modeling. The aim of a predictive model is to derive a meaningful relationship between the modeled event (e.g., death or surrender) and various explanatory factors (e.g., age, policy duration and postal code group). This provides enhanced understanding of the underlying risk drivers and is a vital input in pricing business in competitive sectors especially where competitors are pricing on a segmented basis (e.g., pricing payout annuities according to ZIP/postal code or fund size). Traditional approaches have attempted to explain the modeled event using a very small number of explanatory variables. They fail to allow for the effect of correlations in the data, and this can often lead to material misunderstanding (and hence, mispricing) of the relevant risks. In recent years, many life insurers have started to use multidimensional predictive techniques such as generalized linear models (GLMs), which allow a number of factors to be analyzed simultaneously. The factors can include data held by the provider such as customer age and policy duration, as well as externally available data such as geodemographic, lifestyle or investment market. By capturing a greater number of variables, and allowing for correlations and interactions between variables, GLMs identify and quantify the true effect of risk factors on the modeled event. In this issue 1 GLMs in Life Insurance 6 Migrating MoSes Applications to RiskAgility FM 10 RCS Software Support Services 11 News From the World of Financial Modeling

2 Figure 1. What are GLMs? Phase 1 Phase 2 Phase 3 Age Gender Smoker GLM Probability of dying Affluence Other What Are GLMs? A GLM is just one type of a vast family of multivariate regression models. It seeks to model the observed event as a function of some explanatory factors (e.g., age, sex and duration). Using a GLM with a log-link function would give the following, an intuitively simple structure: Modeled quantity = Base level for observed population Factor 1 (based on age) Factor 2 (based on gender) Factor 3 (based on duration). The modeled quantity could be, for example, the probability of death or the lapse propensity. As P&C insurers have found for many years, these relatively simple models can provide greatly improved pricing granularity, allowing insurers to price good risks and bad risks far more accurately, and minimize the risk of antiselection. Key Advantages of GLMs The technique is widely accepted in the P&C insurance sector. GLMs allow appropriately for correlations in data. Ignoring these can produce significant inaccuracies in rates. GLMs can allow easily for interactions in the factor relationships for example, in mortality analyses, the age curve generally differs by gender. GLMs allow for the nature of the random process underlying claim experience. This allows the output to incorporate information about the likely variability of the model results. GLMs are robust, transparent and easy to understand. GLMs are grounded in statistical theory and offer a practical method for insurance companies to attain satisfactory profitability and a competitive advantage. 2 towerswatson.com

3 Mortality Analysis The use of GLMs in analyzing a life insurer s mortality experience is becoming increasingly widespread. By fully identifying and quantifying all available explanatory factors beyond what is typically assumed by a traditional one-way mortality analysis, the insurer can develop premiums that accurately reflect the relative risk characteristics of the pool of underlying policyholders. Rating factors typically found to be significant in a GLM mortality analysis are summarized in Figure 2. Figure 2. Predictive factors in a mortality analysis Age Personal data Medical information for example, BMI Gender Typical rating factors found to be significant in addition to time period factor (for example, calendar year) Geographical data Socio-geodemographic Portfolio-derived clusters Population mortality clusters Amount 130% 130% 120% 150% Policy data Duration 125% Product type/options 120% Company data Distribution channel Where traditional analyses are restricted by concerns about time trends in the data, the use of calendar year of exposure as a factor allows these trends to be captured explicitly. This permits the use of many more years of data than would otherwise be considered safe and allows companies to absorb more from their data. This is particularly relevant in relation to mortality investigations where mortality improvement assumptions need to be quantified (as demonstrated by Figure 3, where mortality is decreasing year after year). By introducing interactions into the GLM, for instance, interactions between pension or annuity amount and calendar year of exposure, or between sex and year of exposure, such trends can be considered in more detail than might otherwise be feasible. Figure 3. GLM result for calendar year (screen shot from our Emblem software) Rescaled Predicted Values Calendar Year Exposure Model prediction at base levels Model prediction + 2 standard errors Model prediction - 2 standard errors 3 towerswatson.com

4 As well as using GLMs to automatically fit to the most appropriate mortality rates, GLMs can also be used to model the observed mortality effect in excess (or below) the one predicted by any given standard mortality table. Figure 4. U.K. mortality scores An individual s ZIP or postal code, shown to be a valuable predictive proxy for underlying health and health behavior, has been successfully used in recent years as a rating factor in GLM analyses. This can help better reflect the mortality characteristics of the underlying data. Figure 4 shows mortality scores ranging from high (red) to low (blue) across the U.K. Retention Analysis With increasing competition in the life sector, and correspondingly reduced product profit margins, life insurers are increasingly turning to their in-force books in a search for higher profits. What are the drivers underlying policyholders lapse or surrender decisions? GLMs provide a natural tool to answer such questions, given their ability to analyze large numbers of factors. Such an analysis can reveal how previously overlooked parameters, such as age, socioeconomic characteristics or distribution source, may affect policyholder behavior. The ability of GLMs to correctly allow for correlations in data and for possible interaction effects between factors allows firms to develop more accurate behavior models. For instance, firms can see how the duration effect might differ between different commission structures or between different distribution channels. Proper consideration can also be given to any geographical/socioeconomic effect by either overlaying socioeconomic ZIP or postal code categorizations, or studying the observed high- and low-retention postal codes derived from portfolio experience. Additionally, where a key measurable is fund changes rather than policy movement, a GLM can explicitly capture the effect on the benefit amount. As with a mortality analysis, it is possible to improve understanding of how observed calendar-year-ofexposure effects in a portfolio relate to external factors such as relevant economic indices. As with a mortality analysis, it is possible to improve understanding of how observed calendar-year-of-exposure effects in a portfolio relate to external factors such as relevant economic indices. It is worth nothing that one of the potential constraints to using GLMs in the life sector having too few observed events generally disappears with lapse and surrender analyses, since most companies will have sufficient volumes of responses to support a meaningful analysis. 4 towerswatson.com

5 Figure 5. Predictive modeling case studies Postal code pricing A large annuity writer was concerned about its changing mix of business as the market grew more competitive. Our predictive model of its experience split out this changing mix from other factors. We then enhanced the model to provide a predictive postal code-based rating basis to ensure the company avoided antiselection. The company licensed our software as a result. Health insurer A leading health insurer was concerned that it was falling behind best practice in its claim modeling. We developed a series of predictive models for claim inception and termination, in conjunction with licensing of our proprietary software. The client estimated an improvement in its loss ratio from this work on the order of 5 million to 10 million annually from the first line of business involved. Policyholder behavior Better rating Health insurer A leading insurer was concerned about the lack of understanding of its policyholder behavior for savings. Conventional analytics were failing to provide any insight. We developed a predicative model of its experience, with particular reference to socioeconomic clusters, allowing accurate pinpointing of the key high-value, high-surrender segment. A leading health insurer had been working on basic predictive models of its experience but had found basic software packages unsatisfactory. The company tested our proprietary software on several of its products and data sets, and concluded that the increased understanding and speed of the modeling were essential in order to retain its strong niche position. Figure 6. History of GLMs in life GLMs help better mortality analysis GLMs help better retention analysis GLMs for price optimization Life Conclusion Over the last decade in the insurance industry, we have seen GLMs transform from a technique used almost exclusively in the P&C insurance practice to one considered an optimal solution for experience analysis of life insurers. By combining the right GLM modeling tools, good data and analytical skill, life insurers have been able to gain and retain a competitive advantage. And with a global trend toward increased sophistication of risk models, behavior models and optimization, we expect the rise of the life GLM to continue. About Emblem Towers Watson s Emblem is a predictive modeling platform that uses GLM techniques to reveal the underlying patterns of portfolio experience, including mortality, lapses and other probabilistic events. Emblem s ability to fit large, complex data sets within seconds improves the quality, speed and efficiency of a predictive analysis process. Key Features of Emblem Ability to generate complex, industry-leading predictive models Focused numerical algorithm, efficient memory management and recycling of models make Emblem fast Intuitive user interface and powerful graphical display of results Numerous options for statistical analysis and diagnosis Saddle Builder functionality for in-depth analysis of factor interactions Ability to find hidden complexities within data Comparison and testing of multiple models to ensure consistent model reliability 5 towerswatson.com

6 Migrating MoSes Applications to RiskAgility FM In the last edition of Insights Financial Modeling, we introduced our new financial modeling software, RiskAgility FM. Although this is completely new software, one of the key drivers for developing it was to provide the capability to migrate MoSes applications into RiskAgility FM so that your company can leverage investments you ve made in your existing models. Migrating MoSes applications translates the input, output, code and projection structures used in MoSes into the new Input Manager and Run Manager approach used in RiskAgility FM. In this article, we provide techniques that can be used to make the migration to RiskAgility FM smoother and more effective, and that will allow you to take advantage of many of the new features right away. The techniques discussed below apply primarily to the migration of custom applications that were built specifically for individual companies. Companies that use the MoSes U.S. Life Modeling Suite would migrate to both RiskAgility FM and the U.S. Library, which requires a number of different migration considerations. What Is a RiskAgility FM Migration? When we think about migrating a MoSes application into RiskAgility FM, we usually consider it in two stages: Importing the MoSes application into RiskAgility FM as is and verifying the results The adaptation of the application to make use of the new features in RiskAgility FM Importing the MoSes Application When you import a MoSes model into RiskAgility FM, the software will automatically set up the structure necessary to run the application. All assumption sets, projection tasks, model classes and other considerations will be converted into their equivalents in RiskAgility FM without any user interaction. In some cases, the import process of the MoSes application may identify areas that need changes to allow the resulting RiskAgility FM project to work effectively. If this is the case, an Import Project Log will be shown after the import. Using the New Features When you have successfully imported the MoSes application into RiskAgility FM and are happy that it is producing the correct results, you will immediately benefit from many RiskAgility FM features. These 6 towerswatson.com benefits include access to the new development environment and more efficient usage of grid resources. You may want to consider making additional changes to take advantage of other powerful new features in RiskAgility FM. This is much more specific to the application being upgraded and will often benefit from proper planning. Changes can be made to the RiskAgility FM project either in one attempt or step by step as the model is developed further. While modifying the project to use the new features is dependent on the individual application s use of specific input functionality and how comprehensive you want the changes to be, for a reasonably standard project, we would expect this process to be counted in days or weeks, rather than months. Because the approach to using inputs and managing the projection process in RiskAgility FM is different from MoSes, we suggest getting a good understanding of how RiskAgility FM works before starting to make changes. In the remainder of this article, we will go through some common areas that may need fine-tuning when first importing a MoSes application into RiskAgility FM. Preparing the MoSes Application for Import The first step before importing a MoSes application into RiskAgility FM should be to clean the existing MoSes application components. Some simple considerations can make the import process smoother. When setting up the Input Managers of imported MoSes applications, RiskAgility FM uses the categories in MoSes to name and populate the input pages. If the variables in the MoSes model have categories assigned, then the Input Manager will already have a suitable structure. We recommend you remove unused items from the MoSes application. Ideally, the application used in the migration should be as clean as possible and contain only those items needed for future reporting. The review should include the following items, and those no longer required should be deleted from the MoSes application: User groups: Review the variables linked to data in each user group If all user groups link to the same data variables, then there is no need to migrate more than one user group Make note of the tasks included in each user group, as you may want to move these around once migrated into RiskAgility FM

7 Move all projection tasks required into one user group, and delete all other user groups Projection tasks and other tasks Assumption sets Wild cards and wild-card sets Importing and Building the Application Once the MoSes application has been cleaned up, the next step is to import it into RiskAgility FM. The import process will leave the original MoSes application untouched and create a new folder containing the RiskAgility FM application. First, a note about how applications are organized in RiskAgility FM: RiskAgility FM distinguishes between solutions and projects, and a solution can contain several projects. Possible uses for multiple projects could be several versions of a model to be used as part of an analysis of movement, or as several different corporate entities. The solution directory should contain everything necessary to run the projections associated with the projects. To import a MoSes application, choose Project from the File/New menu and select Import MoSes application. Enter the location and name for the RiskAgility FM solution and project, and click OK. In the next dialog, enter the location of the MoSes application folder. The impact of the check box Display default variables in the Input Manager is described in Box 1. The subsequent import of the MoSes application is fully automated. Once the MoSes application has been imported, an Import Project Log will be shown (Figure 7). This contains information about the actions RiskAgility FM has taken during import. For example, RiskAgility FM only creates a single assumption set on an input page if there are a number of MoSes assumption sets that contain the same information for the variables allocated to that input page. This prevents unnecessary duplication of assumption sets, and this will be noted in the Import Project Log. Default Variables In MoSes applications, a variable can be assigned a value in the input tab under Task View, and this value can be overwritten in the code. This coding approach made it difficult for the user to understand which value was actually used in the model and is now considered suboptimal from an audit perspective. In RiskAgility FM, a variable can either have its value set in the Input Manager or within the code, but not both. The variable property Set Value in Input Manager determines whether the variable takes its value from the Input Manager. If the property is set to Yes, and you try to assign a value to the variable within the code, the following compilation error will be shown: binary = :no operator found which takes a righthand operand of type <type> (or there is no acceptable conversion) Therefore, a choice needs to be made as to whether the variable will be set in the code or in the Input Manager. This decision will depend on each individual variable and the context it is used in, and may require additional variables to be created. Box 1 describes how RiskAgility FM determines the initial setting of the variable property Set Value in Input Manager during the import process. Once the MoSes application has been cleaned up, the next step is to import it into RiskAgility FM. The import process will leave the original MoSes application untouched and create a new folder containing the RiskAgility FM application. Box 1. The setting Display default variables in the Input Manager influences how RiskAgility FM treats variables that are set to default values When importing a MoSes application, RiskAgility FM uses a set of rules to determine the initial setting of the property Set Value in Input Manager. These initial settings can be influenced from the Import MoSes Application screen, where there is a check box named Display default variables in the Input Manager. If variables in MoSes have the Display in Assumptions property set to Yes but are only set to default values within the task view in MoSes, the abovementioned check box determines whether these variables appear in the Input Manager. Selecting Display Default Variables in the Input Manager will create input variables for these variables and set the variable properties to take its value from the Input Manager. Not ticking the check box will mean these variables do not appear in the Input Manager and can be changed in the code. Figure 7. The Import Project Log summarizes the actions RiskAgility FM has taken during import for example, assumption sets containing the same values are combined 7 towerswatson.com

8 Missing Returns Some applications contain code that relies on a variable, for example, the product code, only taking a certain set of values, say, TI (term insurance) or WL (whole life). Consider the following sample code for renewal expense: if (prod _ code == TI ) return exp_ren_ti(t); if (prod _ code == WL ) return exp _ ren _ wl(t); record with a product code other than TI or WL If the data contain a record with a product code other than TI or WL, the result of this column is undefined and could produce a nonsensical answer. The new compiler in RiskAgility FM detects this potentially dangerous code and reports the compile error: Not all control paths return a value. To correct the code, there are two solutions: 1. Generate an error: throw NonFatalError( Unknown product code + prod _ code); 2. Return a value e.g. zero or NO_AVG: return 0.; Your choice of these solutions depends on your specific situation. Correcting the code might change the results (e.g., if a previously unnoticed data problem now becomes a point skip). Therefore, it might be useful to correct the code in the MoSes version of the application prior to importing it into RiskAgility FM. Multiple Definitions MoSes global variables can cause problems in RiskAgility FM. This only applies to models where global variables are declared in the EXTERNS section of the start-up column. The RiskAgility FM compiler will detect these situations, and the code will need to be corrected before the project can be built. There are two general rules to follow that should fix most of the errors: All declarations apart from exactly one should use the extern keyword. For instance, use double global_var; in the EXTERNS section in exactly one model class and extern double global_var; in all other model classes. Never assign a value in the EXTERNS section; see Listing 1. 8 towerswatson.com Listing 1. The RiskAgility FM compiler recognizes dangerous code and reports an error // this will give a compiler error START EXTERNS extern long max _ errors = ; END _EXTERNS // this is correct START _EXTERNS extern long max _errors; END _EXTERNS max _errors = ; Since the syntax for global variables in MoSes has sometimes been perceived as being complex, we designed RiskAgility FM to use project variables instead. Project variables are global variables that are model-class independent and can be referenced directly by formulas in any model classes within a project. They can easily be created in the RiskAgility FM interface. Unsupported MoSes Functionality Some MoSes features have not been widely used by clients or have been superseded by newer and improved functionality over the software s evolution. We have used this opportunity to simplify the underlying system code, which aids with performance, future maintenance and development of the platform. Features that are no longer supported include layers and Product Feature Assumptions. Most functionality that is commonly used in the majority of client models is still available in RiskAgility FM. Most MoSes applications imported to RiskAgility FM will continue to run without needing changes. For example, DBCurser and DBTable functionality still work, but we expect that use will disappear over time. For advice on using alternative functionality, please contact your Towers Watson consultant. Variables of type variant are a special case. In MoSes applications, variants are commonly used when a variable needs the flexibility to be specified in a number of different data formats, ranging from a single value to a table structure that could include any number of selectors. When variant tables are used, the data are formatted in the MoSes proprietary msv format. In RiskAgility FM, variant data flexibility has been incorporated into the Input Manager, and the variant variable type is not supported. Any variable that was a variant in the MoSes application will have its type changed to floating point during import into RiskAgility FM. The imported code will still include variant table reading parameters, which are no longer valid in RiskAgility FM. An entry will appear in the Import Project Log to show that a variant variable has been converted to a floating point variable.

9 For each of these variables, external sources and lookup definitions should be established in the Input Manager so the model will access the data in the required formats. Any code that uses these variables will need to be reviewed and changed to ensure proper syntax and use of Input Manager features. Running the Projection Once the application has been built, the next step is to verify the results. First, we need to look at the inputs of the projection. Data MoSes was designed to cover the whole financial modeling process, from data creation to report generation. However, clients are increasingly using data warehouse technology and have dedicated teams to generate the data. Therefore, we have focused design efforts for RiskAgility FM to be a financial modeling calculation engine that supports a large variety of file formats. So whatever format your data are in, RiskAgility FM is likely to be able to read it. Among the supported file formats are Excel files (Excel 2007 and later), text files and database files (Microsoft SQL Server, Microsoft Access, FoxPro). Note that RiskAgility FM requires the first row to contain headers in any nondatabase external source that is to be used as a data file. The reading of external data is also no longer contingent on the value of the product field, so the data file does not need to contain this field. When a MoSes application is imported into RiskAgility FM, the name of the data file is kept, but the extension is automatically changed to.xlsx. However, this is just a placeholder. The data will not be converted. Therefore, the data links need to be reviewed. The decision as to which file type is best suited is company-specific and depends on your data process. As an illustration, we will look at how we can copy the contents of a FoxPro DBF data file into a commaseparated value (CSV) file that can be used in RiskAgility FM to reproduce the results. To do this, create an Other Task of type FoxPro Procedure in MoSes, and enter the code shown in Listing 2 into the FoxPro Procedure field. Replace the <dbf_file> in the code with the full path of the MoSes data file and <csv_ file> with the full path to the new RiskAgility FM data file to be created. Listing 2. FoxPro code to convert a FoxPro DBF file into a CSV USE < d b f _ fi l e > COPY TO < c s v _ fi l e > TYPE csv CLOSE ALL If your company has a Microsoft Visual FoxPro license, you can also enter these commands directly into Visual FoxPro. Other options include using Microsoft Query in Excel to read the contents of the DBF data file into Excel and saving it as an.xlsx file, or creating an external source of type database and supplying a connection string to directly read the FoxPro DBF data file. Assumptions RiskAgility FM does not force you to use the new Input Manager approach to read assumptions. You can initially still use the standard MoSes tablereading approach using table variables in your RiskAgility FM project. This means that it should be enough to copy the assumptions into the directory structure under the solution directory. Make sure the wild cards are set correctly so that RiskAgility FM can locate the assumptions. Comparing the Results Once the data have been set up and the assumptions have been copied, the projections can be run from the Run Manager. RiskAgility FM outputs the results in CSV files. The files are named: <run page>_<projection>~<model object name>.csv Therefore, there is one file per model object. CSV files are an industry standard file format and should integrate easily into your reporting process. For those companies using MoSes Excel reporting, we can provide an add-in for Excel with equivalent functionality to read RiskAgility FM output. Please contact us for more information. 9 towerswatson.com

10 Summary Migrating MoSes applications into RiskAgility FM is a two-step process: Step 1: Import the MoSes application into RiskAgility FM. This is an automated process where the software sets up all the components you need to run the model. The process can be made more effective by following a few simple steps. Most MoSes applications should be able to run without changes, and some applications may need very straightforward changes. Step 2: Make use of the powerful new features of RiskAgility FM. This may require some effort but can be done step by step. It requires planning and a good understanding of the new features in RiskAgility FM, but can make your processes more streamlined, quicker, easier to audit and more robust. We can help you find the right approach for your company. Our consultants have gained a great deal of experience and knowledge during the development of RiskAgility FM, when many of our own and client applications were migrated. We can help you effectively modify MoSes applications, and take advantage of the extensive and powerful new features of RiskAgility FM. If you would like to learn how to make best use of the many improvements in RiskAgility FM, please contact one of the consultants listed on the last page. RCS Software Support Services We have continually sought to maximize the customer software support experience and make this our primary goal. Challenged by the success and rapid global expansion of our enterprise products, we are striving to stay on the leading edge of product support and technical service delivery. We are pleased to announce the implementation of a cutting-edge software support process that provides global coverage while maintaining a single, holistic view of all customer support issues. Key features of the new support process are: Implementing a new client portal. The portal will enable a client to raise issues and check the status of existing ones, search the knowledge base for common issues, and download new software releases and software license files (to follow shortly), replacing DVDs and dongles where appropriate. The online portal is now available globally. During the annual license renewal, clients will be invited to register to use the portal. Alternatively, you can request access from your local Towers Watson software representative. However you wish to log a ticket with us, you will receive a reference number of the form TKxxxxxxxx. Developing new services such as 24x7 support. We have created a new first-line support call center that is now able to process voice, and portal queries covering the standard regional 9 a.m. 5 p.m. (Monday Friday) support and has the ability to provide enhanced levels of service up to full 24/7 support (where the client has registered for Enterprise Support). The team will also be available to support future products and services as they become available. Skilled/experienced local second-line support. Regional technical support and consulting teams are available to support you in planning and implementing our enterprise software environments. This team will also pick up escalated issues logged through the first-line team, providing local knowledge to the client and application. Quickest route to resolution. We have developed processes that are continuously improving to meet growing client expectations. Tickets requiring specialized knowledge are identified and the correct resource employed as quickly as possible, and efficient escalation of complex technical queries are sent to our technical consulting and services team or the development team. 10 towerswatson.com

11 Continuous training, monitoring and assessment. The global support team will continuously undertake training as part of their weekly routine to improve their knowledge of commonly reported issues, and will follow a comprehensive quality assurance process to monitor performance and proactively identify areas for improvement. The first-line team will be audited annually against ISO standards. The vision for the support team is to resolve user queries as efficiently as possible using a collaborative global support team. We are confident that the new support service will help complement our enterprise software products. Regional telephone details, the self-service portal address and addresses to access the new first line support can be found at towerswatson.com/en/services/tools/contact-ussoftware-solutions. We welcome any feedback on the services provided by Towers Watson Software Support. Please direct any feedback directly to the support team or to Duncan Betts, head of Software Technical Support at duncan.betts@towerswatson.com. News From the World of Financial Modeling Solvency II: Back on track This Insights focuses on two particular aspects of Solvency II: the Own Risk Solvency Assessment (ORSA), which now includes the intermediate step of the FLAOR (Forward-Looking Assessment of Own Risks), and the underwriting and reinsurance opinions to be provided by the actuarial function Life supplementary reporting The European Embedded Value (EEV) principles were published 10 years ago in May. Since then, we have seen unprecedented change in life supplementary reporting from EEV, to CFO Forum Market Consistent Embedded Value Principles (MCEV), to balanced scorecards. In our 10th annual survey, we reflect on where we are today and what is next for life supplementary reporting Emphasis A global quarterly magazine providing thought leadership for the insurance industry 11 towerswatson.com

12 Contacts Software Support Visit or Holly Starkey Dominique Lebel Carlos Gonzalez Follow us on 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 more than 14,000 associates around the world, we offer consulting, technology and solutions in the areas of benefits, talent management, rewards, and risk and capital management. Copyright All rights reserved. NA September 2014 towerswatson.com /towerswatson