Analysis of Proposed Principle-Based Approach

Transcription

1 Milliman Client Report Analysis of Proposed Principle-Based Approach A review and analysis of case studies submitted by participating companies in response to proposed changes in individual life insurance statutory reserve and risk-based capital methodologies. September 24, 2009 Prepared for: The Society of Actuaries Prepared by: Milliman, Inc. Ken Joyce, FSA, MAAA Craig Roberts, FSA, MAAA Karen Rudolph, FSA, MAAA William Sayre, FSA, MAAA Uri Sobel, FSA, MAAA 2009 Society of Actuaries, All Rights Reserved. The opinions expressed and conclusions reached by the authors are their own and do not represent any official position or opinion of the Society of Actuaries or its members. The Society of Actuaries makes no representation or warranty to the accuracy of the information.

2 FORWARD Milliman would like to acknowledge those who participated in the development of this research paper, especially the actuarial task force (ATF) members who provided the case study modeling results. We would also like to thank the Society of Actuaries Project Oversight Group (POG) who provided advice and guidance during the course of the project. We also thank the research coordinators from the Society of Actuaries for their efforts in organizing and coordinating the project. SOA RESEARCH COORDINATORS Jan Schuh Ronora Stryker PROJECT OVERSIGHT GROUP Mary Bahna-Nolan Susan Deakins Bob Diefenbacher Jerry Enoch Mike Leung John Palmer Craig Reynolds Steve Strommen Kenneth Vande Vrede

3 TABLE OF CONTENTS EXECUTIVE SUMMARY OBJECTIVES 5 KEY OBSERVATIONS 5 KEY COMPARISONS 6 BACKGROUND REFERENCES 7 ELEMENTS FRAMING THE RESEARCH 7 RESEARCH LIMITATIONS 7 METHODOLOGY RESEARCH APPROACH 9 ROLES OF PARTICIPANTS 10 QUALITY CONTROL 10 PEER REVIEW 10 TERM INSURANCE BLOCK DESCRIPTION 11 FORMULAIC STATUTORY COMPONENTS 12 DETERMINISTIC COMPONENTS 12 COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY 14 RESULTS BY CALENDAR YEAR 16 STOCHASTIC EXCLUSION TEST 17 STOCHASTIC RESERVE 18 STOCHASTIC RESERVE DISTRIBUTION 20 MINIMUM RESERVE 21 C3 PHASE III CAPITAL COMPONENTS 22 OBSERVATIONS 23 UNIVERSAL LIFE INSURANCE BLOCK DESCRIPTION 24 FORMULAIC STATUTORY COMPONENTS 25 DETERMINISTIC COMPONENTS 26 COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY 27 RESULTS BY CALENDAR YEAR 31 STOCHASTIC EXCLUSION TEST 33 STOCHASTIC RESERVE 35 STOCHASTIC RESERVE DISTRIBUTION 37 MINIMUM RESERVE 38 C3 PHASE III CAPITAL COMPONENTS 39 OBSERVATIONS 40 TRADITIONAL WHOLE LIFE INSURANCE BLOCK DESCRIPTION 41 FORMULAIC STATUTORY COMPONENTS 41 DETERMINISTIC COMPONENTS 42 COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY 43 RESULTS BY CALENDAR YEAR 44 STOCHASTIC EXCLUSION TEST 46 STOCHASTIC RESERVE 46 STOCHASTIC RESERVE DISTRIBUTION 48 MINIMUM RESERVE 49 C3 PHASE III CAPITAL COMPONENTS 50 OBSERVATIONS 51 3

4 DEFERRED ANNUITY BLOCK DESCRIPTION 53 FORMULAIC STATUTORY COMPONENTS 53 DETERMINISTIC COMPONENTS 54 COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY 55 RESULTS BY CALENDAR YEAR 57 STOCHASTIC EXCLUSION TEST 57 STOCHASTIC RESERVE 57 STOCHASTIC RESERVE DISTRIBUTION 58 MINIMUM RESERVE 59 C3 PHASE III CAPITAL COMPONENTS 59 OBSERVATIONS 60 PRACTICAL ASPECTS OF THE RESEARCH WORK 61 SUGGESTED REFINEMENTS 63 APPENDIX A VM-20 AND C3 PHASE III OVERVIEW 64 CHANGES TO VM-20 SINCE DATE OF RESEARCH 66 APPENDIX B STOCHASTIC EXCLUSION TEST DEFINITION 67 APPENDIX C ATF DATA REQUEST 68 APPENDIX D ATF DOCUMENTATION REQUEST 78 APPENDIX E Q & A FROM ATFS TO RESEARCH TEAM 81 APPENDIX F ADDITIONAL DETAIL 30 YEAR TERM DETERMINISTIC RESERVE 86 APPENDIX G ASSUMPTION DETAIL FOR TERM 87 APPENDIX H ASSUMPTION DETAIL FOR UNIVERSAL LIFE 102 APPENDIX I ASSUMPTION DETAIL FOR TRADITIONAL WHOLE LIFE 121 APPENDIX J ASSUMPTION DETAIL FOR DEFERRED ANNUITY 136 4

5 EXECUTIVE SUMMARY OBJECTIVES In collaboration with the National Association of Insurance Commissioners (NAIC), the Life Reserves Work Group (LRWG) and the Life Capital Work Group (C3WG) of the American Academy of Actuaries are working to finalize a principle-based reserve and capital methodology for life insurance products. The objective of this research is to serve as a field test for the proposed principle-based approach to statutory reserves and required capital (PBA). It is hoped this research will provide the working groups and the life insurance industry as a whole with modeling results using actual data across various product types and from a company s inforce block, to facilitate a better understanding of: The impact of the proposed methodology through preliminary modeling results; Challenges of implementing the methodology; and Elements of the methodology that may benefit from refinement before adoption by regulators. The reader can find details corresponding to the impact on Term Insurance beginning on page 11; Universal Life Insurance beginning on page 24; Whole Life Insurance beginning on page 41; and Deferred Annuities beginning on page 53. The section titled Practical Aspects of The Research Work discusses challenges of implementing the methodology and begins on page 61. The section titled Suggested Refinements begins on page 63. KEY OBSERVATIONS The field test results indicate the cash value floor requirement of VM-20 has a material impact on the Per Policy Reserves and thus the Deterministic Reserves for term insurance, universal life and whole life policies. Its significance depends in large part upon the competitiveness of the product, especially for term insurance products. When compared to current statutory minimum reserve requirements, this research test bed suggests a wide range of possible outcomes. The range is influenced by: Sufficiency of the premium to support benefits, expenses and profit Maturity of the block Level of margins used in the valuation Characteristics of the asset portfolio supporting the policies The value of margins, as measured by the difference in Seriatim Reserves, can be significantly muted by the application of the cash value floor. Determining the direction and ultimate impact to the entire modeled block of margins that are established on each individual risk factor will be difficult. Detail provided as part of this research on the universal life policies show that although a margin increases the Seriatim Reserve, the Per Policy Reserve may not follow suit, due largely to the influence of the cash value floor. The direction of the margin (e.g. to increase or decrease the rate of lapse) could vary on a contract level and may even vary on a policy year level. With one exception, the tested term and whole life blocks indicate stochastic exclusion test results less than 4%. The universal life blocks exhibited stochastic exclusion test ratios both greater and less than the 4% threshold, demonstrating that a block of universal life insurance with secondary guarantees will not necessarily fail (ratio greater than 4%) the stochastic exclusion test. Influencing factors include the strength of the guarantee, characteristics of the assets supporting 5

6 EXECUTIVE SUMMARY the policies, underlying product charges and loads and the level of assumed lapses when the secondary guarantee is in play. The result of the stochastic exclusion test is not always an indicator of whether a stochastic excess exists for a block. Of the six universal life blocks subject to the research, four blocks failed the stochastic exclusion test with ratios of 5.0%, 5.6%, 7.5% and 25.4%. These four blocks had no stochastic excess over the deterministic reserve. Two blocks passed the exclusion test with ratios of 1.7% and 2.5%, one of which did have a stochastic excess over deterministic, though the stochastic excess accounts for only 4% of the total VM-20 minimum reserve. For the tested life insurance blocks, where the C3 Phase III Stochastic Amount is larger than the comparable formulaic RBC amount it is larger by less than 10%. For deferred annuities, though the research included only one block calculating the C3 Phase III Stochastic Amount, it is approximately the same or slightly larger than the comparable formulaic RBC amount. KEY COMPARISONS This research allows for comparison between a principle-based methodology and the current formulaic statutory methodology for reserves and risk-based capital. Term insurance is generally thought to be largely insensitive to economic conditions and the research supports this belief. For aggregate inforce term blocks, the principle-based reserve amounts determined in this research vary widely from the comparable statutory reserve amount. The calculated amounts provided for this research show Deterministic-to-Formulaic results ranging from 0% for a relatively newer, less competitive, 30-year level premium term product to 82% for a more mature, competitive 20-year level premium term product. Because the principlebased amount reflects the risks inherent in each term product, we conclude that the term blocks contributed to this research effort carry varying levels of risks to their respective companies. With respect to risk-based capital elements, this research indicates that current statutory reserve levels for recently issued term insurance are high enough that proposed C3 Phase III requirements would result in no additional C3 requirement above the statutory reserve amounts. Universal Life with secondary guarantees is generally thought to be a product type that will, depending on the strength of the secondary guarantee to the policyholder, require stochastic analysis. The research results for the stochastic exclusion test support this belief. The full stochastic analysis, however, produced only one block with a stochastic excess over the Deterministic Reserve. The Deterministic-to-Formulaic reserve ratio for universal life ranges from 67% to 116%. This indicates a wide range but in contrast to term insurance results, the current statutory reserve is bracketed by the range for the Deterministic Reserve. With respect to riskbased capital elements, this research indicates that current statutory reserve levels for recently issued universal life are high enough that proposed C3 Phase III requirements would result in no additional C3 requirement above the statutory reserve amounts for five of the six tested blocks. Participating Whole Life products were the focus of the whole life section of this research; three participating whole life blocks and one non-participating block were submitted. The Deterministic Reserve ranges from 78% to 100% of the Formulaic Statutory Reserve. All four blocks would qualify for the stochastic modeling exclusion. If the full stochastic analysis were performed, one of the four blocks would demonstrate a stochastic excess over the Deterministic Reserve. With respect to risk-based capital elements, three of the four blocks would require no additional C3 Phase III amount above the formulaic statutory reserve amount. 6

7 BACKGROUND REFERENCES The portion of the NAIC s proposed Valuation Manual relating to life insurance reserves is VM-20. VM-20 is the section which defines the minimum statutory reserve standards for life insurance policies subject to a principle-based reserve valuation. At the time the research was initiated, the most recent draft version of VM-20 was dated September 22, It is this version we refer to throughout this report. For the risk-based capital component of this research, we refer to the September 2008 draft version of Report of the American Academy of Actuaries C3 Life and Annuity Capital Work Group. This is referred to as the C3 Phase III report. Other publications frequently noted in this report are listed below. Canadian Institute of Actuaries Educational Note - Margins for Adverse Deviations, Committee on Life Insurance Financial Reporting, November Report from the Preferred Valuation Basic Table Team, a subgroup of the American Academy of Actuaries/Society of Actuaries Joint Preferred Mortality Oversight Group, March Descriptions, definitions and formulas for all aspects of this research are not included in this report. The reader is referred to the documents listed above for reference. A brief summary of VM-20 and C3 Phase III is included in Appendix A. ELEMENTS FRAMING THE RESEARCH The requirements of VM-20 are expected to impact new business issued after the operative date of the Valuation Manual. In this research, the data presented assumes the requirements apply to the inforce subject policies. This allows for comparisons of current statutory requirements to the proposed PBA methodologies for a block of inforce business. The data used in this analysis was generated by contributing companies. Milliman has relied on information supplied by contributing companies. We reviewed the information provided for reasonableness and consistency with the requirements of VM-20 and the C3 Phase III report, however, we performed no audits or independent verification of the information furnished to us. The results presented in this report have been expressed in a way to preserve the confidentiality of the contributing companies by normalizing the submitted results. Life insurance results are normalized using a unit basis of $1,000,000 of insurance amount inforce and rounding to the nearest dollar, deferred annuity results are normalized using a unit basis of $10,000 of statutory reserve inforce and rounding to the nearest dollar. RESEARCH LIMITATIONS This research report is intended to facilitate a better understanding of the impact of a principlebased approach. It is not intended to provide any recommendations in regards to the proposed methodologies under the PBA framework. It is not intended as a comprehensive overview of the impact of PBA on the industry as a whole. The research team also recognizes there may be critical aspects of PBA not reflected in this analysis if such aspects were not developed at the time of the research. Most notably, the net premium reserve element, which developed in early 2009, is not considered here. 7

8 BACKGROUND The research began with a call for industry participants willing to establish actuarial models to execute a field test of the proposed PBA methodology. A study participant is referred to as an actuarial task force (ATF). In the end, 12 contributing companies provided results for four product types: term, universal life, whole life and deferred annuity. The valuation date for this research was September 30,

9 METHODOLOGY RESEARCH APPROACH Each ATF was provided the same data request memorandum referencing the September 2008 draft versions of VM-20 and the C3 Phase III (C3P3) report. Each research member kept in close contact with their assigned ATF groups. As the ATFs began to develop results, the research team provided them with a reporting template, to ensure consistency of the submitted data. The ATFs also received a template for documentation of the key aspects of their actuarial models. Though intended as a field test in which each ATF was expected to interpret the VM-20 and C3P3 requirements, there needed to be some common framework with which to start. The research data request memorandum provided this framework. Each ATF was asked to accommodate the data request memo as much as reasonably possible. The valuation date of September 30, 2008 was recommended in order to synch up with typical asset adequacy testing valuation dates. The ATFs were directed to use inforce asset and liability models as long as the assets being used in the model were intended to support the liabilities being valued. The starting yield curve was inherent in the scenarios provided, therefore each ATF used the same starting yield curve. The ATFs were provided 1,000 scenarios intended to represent a set of 10,000 from the Academy s Economic Scenario Working Group scenario generator. Also provided were 16 stochastic exclusion test scenarios. For purposes of this research, the research team also provided limited guidance regarding assumptions and particularly assumption margins. To expedite the research, the models were considered to fall into one of three groups. 1. Model assumptions are prudent estimate assumptions and the ATF can identify and describe margins on each assumption as required by the proposed regulations; or 2. Model assumptions are anticipated experience and the ATF is willing and able to determine margins for each assumption as required by the proposed regulations; or 3. Model assumptions are anticipated experience and the ATF will use margin guidance provided by the POG or research team. The participating ATFs frequently used a combination of the three options above, depending on the experience assumption. With respect to mortality assumptions used in this research, the ATFs determined the credibility of their company s experience and blended it with industry experience if applicable, using a credibility blending method. The industry experience is represented by the 2008 Valuation Basic Tables. To determine the industry experience table appropriate to the underwriting expectations of each risk class, the research team directed the ATFs to the Report from the Preferred Valuation Basic Table Team, a subgroup of the American Academy of Actuaries/Society of Actuaries Joint Preferred Mortality Oversight Group, March For guidance on mortality margin setting, the research team directed ATFs to the Canadian Institute of Actuaries Educational Note Margins for Adverse Deviations, Committee on Life Insurance Financial Reporting, November This educational note provides for a margin of N additional deaths per thousand where a range for N is provided. The ATFs were not precluded from taking other approaches, however. Margins used by the individual ATFs are detailed in the Appendices. With respect to lapse assumptions used in this research, the ATFs either used their company s specific margin on lapses or, where no specific margin existed, utilized guidelines set in the Canadian Educational Note, as directed by the research memo. The Canadian Educational Note suggests a low (5%) to high (20%) margin, where the chosen margin level should reflect the estimation error and adverse deviation of the risk factor. The ATF should determine whether a non-level margin that varies by duration is appropriate. The margin should be added or subtracted 9

10 METHODOLOGY from the anticipated experience assumption in a way that increases, not decreases the reserve amount. This proved a challenge for some lines of business, particularly for the lapse assumption. The expense levels used in the research were fully allocated expenses and included taxes (except federal income tax), licenses and fees and assume the company is a going-concern business. Inflation was recognized in the models, but not future expense improvements. Low (2.5%) to high (10%) margins, per the Educational Note, were suggested where a margin did not exist. In modeling the specifics of dynamic policyholder behavior and non-guaranteed elements participants were directed to follow the requirements outlined in VM-20. Asset portfolios were based on each company s current September 30, 2008 inforce asset listing appropriate for the block being modeled. The data request memo suggested a schedule of default charges for reinvestment assets, which some ATFs chose to use for both inforce and new assets. Other ATFs chose to use their company s current default charge assumption for the inforce assets. For all ATFs, the spread over US Treasury rates on reinvestment assets was specified by the research team. All but two ATFs used this specification. Complete details of the data request can be found in Appendix C. ROLES OF PARTICIPANTS The POG provided management, oversight and direction for the project, meeting with the research team periodically by phone. The SOA staff coordinators facilitated the research through scheduling and communications. The ATFs provided the modeling expertise and the research team provided management of the ATFs, analysis of the research data, and this report. QUALITY CONTROL Through maintaining frequent contact with the ATFs, the research team members were able to provide guidance on questions that arose during the modeling effort. As results were produced, the research team reviewed results together with the ATF. Ultimately, the research team requested and received specific model output intended to enable the researcher to review the ATFs calculation of the deterministic amounts and the stochastic amounts. As each ATF s results were finalized, their result package was cross-reviewed by another researcher. PEER REVIEW The POG provided peer review of this report. 10

11 TERM INSURANCE BLOCK DESCRIPTION In this section, research results are presented for the modeled term insurance products, followed by a section of observations and discussion. The results have been normalized using a unit basis of $1,000,000 of insurance amount inforce and rounding to the nearest dollar. Four blocks of term insurance are modeled as part of this research. The blocks uniformly include 20-year level term, but two include 30-year level term and one includes 10-year level term to age 95 products. We include the 10-year and 30-year level premium products in our results, but our primary focus is the 20-year product. Premiums are guaranteed at issue of the contract. After the level premium period, the guaranteed premiums follow an annually increasing pattern. Basic information on the four blocks is shown below. TERM BLOCK Risk Class Structure Issue Years Projection Period Level Premium Period Period Following Level Premium Period T-1 2 Nonsmoker 1 Smoker Oct. 1, 2008 through Sept. 30, , 20, 30 Not included in calculations (100% lapse at end of level period). T-2 3 Nonsmoker 2 Smoker Oct. 1, 2008 through Sept. 30, Included in calculations after shock lapse of 75% for two years following level premium period. T-3 3 Nonsmoker 2 Smoker Oct. 1, 2008 through Sept. 30, , 30 Included in calculations after shock lapse of 97-99%. T-4 3 Nonsmoker 1 Smoker Oct. 1, 2008 through July 31, Not included in calculations (100% lapse at end of level period). It is important to note that T-2 and T-4 treated all inforce policies as subject to the Valuation of Life Insurance Policies Model Regulation (i.e. Triple-X) in this analysis, even if issued prior to 1/1/

12 TERM INSURANCE FORMULAIC STATUTORY COMPONENTS To facilitate comparisons to current formulaic reserve requirements the components of the current NAIC Statutory Reserves are shown in Table I. Note that each block listed includes varying numbers of issue years in its inforce. TABLE I STATUTORY RESERVE AND RELATED AMOUNTS TERM BLOCK Level Premium Period Basic Reserve Deficiency Reserve Gross Reserve Reinsurance Reserve Credit Net Deferred Premium Net Reserve for Comparison to VM-20 T-1 10 $ 2,809 $ 155 $ 2,964 $ 22 $ 1,754 $ 1,188 T ,682 1,090 3, ,081 2,669 T , , ,254 2,877 T , , ,561 T , ,374 T ,627 4, ,105 T , , ,028 DETERMINISTIC COMPONENTS Modeling results for the deterministic components of VM-20 are shown in Table II followed by a discussion of Items #1 through #6. TABLE II VM-20 DETERMINISTIC COMPONENTS TERM BLOCK Level Premium Period Item #1 Sum of Seriatim Reserves Item #2 Sum of Per Policy Reserves Item #3 Sum of Per Policy Reserves ignoring Reinsurance (FN1) Item #4 Sum of Per Policy Reserves using Anticipated Experience Item #5 Item #4 with Mortality Margin only Item #6 Item #4 with Lapse Margin only T-1 10 $ (1,897) $ 266 $ 297 $ 128 $ 241 $ 139 T-1 20 (710) T-1 30 (4,986) T ,840 7,867 8,762 5,307 5,973 7,165 T N/A T-3 30 (1,203) 1,079 N/A T ,947 N/A 1,040 1,725 1,129 FN1: N/A implies reinsurance is not applicable. Item #1, the Seriatim Reserve, is used as a step in the calculation of the Deterministic Reserve and equals the present value of cash flows for each policy, discounted at the net asset earned 12

13 TERM INSURANCE rate for the model segment under a single deterministic scenario (Section C.2 of VM-20, September 2008). A negative Seriatim Reserve at the valuation date indicates the premiums are sufficient to fund future benefit and expense obligation cash flows. The four blocks tested represent a variety of competitive profiles. For example, at the time of pricing, block T-1 had a higher profit margin than T-4, which in turn had a higher profit margin than T-2. Item #2 is the sum of Per Policy Reserves for the block of policies being valued. The Per Policy Reserve is the greater of the Seriatim Reserve and the policy s cash surrender value, adjusted for reinsurance. In the case of term insurance, the cash surrender value can be $0, depending on the term length and issue age of the policy. On the valuation date, all of the policies included in this analysis had a cash surrender value of zero. For purposes of this research, we are treating Item #2 as the Deterministic Reserve, although in a total company situation, the Deterministic Reserve would equal the sum of the Per Policy Reserves over all policies subject to VM-20. It is interesting to note the effect the cash value floor has on Deterministic Reserves evaluated for an inforce term insurance block. For a competitively priced block, the cash value floor has little impact (e.g. Block T-2). However, for a less competitively priced term block, the floor has a significant impact. The applicability of this observation is limited to the insurance products evaluated in this research. Item #3 is the same as Item #2, but ignoring reinsurance, if any. Item #4 is the same as Item #2, but excluding all margins on risk factors. Item #5 is Item #4, with the mortality assumption being the only risk factor with a margin. Item #6 is Item #4, with the lapse assumption being the only risk factor with a margin. 13

14 TERM INSURANCE COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY Table III-A summarizes comparisons of the VM-20 Deterministic Reserves to Formulaic Net Statutory Reserves as well as the impact of reinsurance and the margins applied to mortality risk and lapse risk factors. The percentages are based on normalized results before rounding. TABLE III-A COMPARISON OF DETERMINISTIC TO STATUTORY RESERVES/ MARGIN ANALYSIS A B C D E TERM BLOCK Level Premium Period Item #2 / Net Statutory (FN1) Item #3 / Direct Statutory (FN2) Impact of All Margins (FN3) Impact of Mortality Margin (FN4) Impact of Lapse Margin (FN5) T % 25% 107% 88% 8% T % 33% 242% 176% 33% T % (FN6) 0% 457% 217% 17% T % 89% 48% 12% 35% T % N/A 265% 194% 41% T % N/A 235% 148% 62% T % N/A 87% 66% 8% FN1: Item #2 from Table II divided by Net Reserve for comparison to VM-20 from Table I. FN2: Direct Statutory = Gross Reserve less Net Deferred Premium from Table I. FN3: [Item #2 - Item #4] / Item #4, from Table II. FN4: [Item #5 - Item #4] / Item #4, from Table II. FN5: [Item #6 - Item #4] / Item #4, from Table II. FN6: Detail surrounding this result is found in Appendix F. Column A in Table III-A demonstrates a wide range of Deterministic Reserves as compared to Statutory Net Reserves. Contributing to this wide range are premium guarantee period, mix of business by duration and competitive profiles of the blocks. For those blocks with reinsurance, the difference between the VM-20 Deterministic Reserve with and without reinsurance cash flows is considerably larger than the current statutory reinsurance reserve credit. For example, block T-1 has a total statutory reserve credit of $60 whereas, under VM-20 the implied reinsurance reserve credit (on a deterministic basis) is defined as the difference in Deterministic Reserves with and without reinsurance cash flows (Items #2 and #3 in Table II), then block T-1 has a $233 reserve credit. Likewise, block T-2 has a statutory reserve credit of $229, while VM-20 values this credit at $895. Both block T-1 and T-2 employ YRT reinsurance arrangements in this analysis. Note that at the time of this research, the NAIC s position on reinsurance within VM-20 and specifically the reinsurance reserve credit is not settled. Columns C, D, and E of Table III-A quantify the relationships of risk factor margins in the modeling work. Column C quantifies the impact of all margins as a percentage of Item #4 in Table II. For example, in block T-3, 20-year term, a margin of 265% is added to Item #4, the Per Policy Reserve using anticipated experience, in arriving at the Deterministic Reserve of $844. This 265% has components of mortality (194%; column D) and lapse (41%; column E). Assumption margins 14

15 TERM INSURANCE in this field test cover a wide range from just under 50% to over 400% as computed by these ATFs. Detail on the methods used by the ATFs to establish margins is found in the assumption detail section. In most cases, mortality plays the largest role in the aggregate margin impact. 15

16 TERM INSURANCE RESULTS BY CALENDAR YEAR Comparisons by calendar year of issue are provided below for the 20-year level premium blocks. The Statutory Reserve net of deferred premium and reinsurance reserve credit is compared to the VM-20 Deterministic Reserve. The ratio shown is Deterministic-to-Statutory. TABLE III-B CALENDAR YEAR ANALYSIS FOR 20 YEAR LEVEL PREMIUM T-1 T-2 ISSUE YEAR Statutory Deterministic Ratio Statutory Deterministic Ratio % % , % % , % 1, % , % , % ,326 1,086 82% , % % % , % % % Total 2, % 9,561 7,867 82% T-3 T-4 ISSUE YEAR Statutory Deterministic Ratio Statutory Deterministic Ratio , % % % % % % % % % % % % 16

17 TERM INSURANCE Total 1, % 6,028 1,947 32% Statutory methodology amortizes an initial expense allowance over the life of the contract. The principle-based approach recognizes actual expenses when incurred. As a result, the Deterministic Reserve in the year of issue will reflect any remaining initial costs of commissioning or underwriting. In all but one case, the 2008 issue year ratio is higher than renewal years. Documentation for T-3 indicates there are no first year or renewal percent of premium or commission expenses on these policies, explaining why its 2008 ratio is lower than the 2007 ratio. STOCHASTIC EXCLUSION TEST The Stochastic Exclusion Test is defined in Section C.4 of VM-20 and in the C3 Phase III proposal. For reference, it is also described in Appendix B. As of the September, 2008 version of VM-20 and C3 Phase III requirements, a test result less than 4.0% indicates the block qualifies for the stochastic modeling exclusion. TABLE IV-A STOCHASTIC EXCLUSION TEST TERM BLOCK Level Premium Period Ratio T % T % T % T % T % T % T % The relative age of the business in the block, as well as the competitive profile of the business, influences whether the block meets the Stochastic Exclusion Test criteria. Newer blocks exhibit a higher test ratio. Block T-3, the newer more competitive 30-year level term product, does not meet the pass mark for stochastic exclusion. This result aligns with the original development of the pass mark, which was set to potentially allow exclusion for guarantee periods no longer than 20 years. With regard to the results for T-3, in practice, the company would likely test the two guarantee periods of the product line together. Combining the field test results this way produces a test ratio of approximately 2.94% for this product. The result depends on the company s distribution of business between the guarantee periods. ATF T-2 was able to provide data to facilitate a demonstration of the Stochastic Exclusion Test ratio by calendar year of issue. This data illustrates the impact of the age of the block on the test result. 17

18 TERM INSURANCE TABLE IV-B CALENDAR YEAR ANALYSIS OF SET (T-2) ISSUE YEAR Stochastic Exclusion Test Ratio Insurance Amount in $MM % $ % % 1, % % 1, % 2, % 2, % 2, % 3, % 2, % 2, % 2,008 Aggregate 2.61% 23,567 STOCHASTIC RESERVE For purposes of this research, Stochastic Reserve calculations were performed by all ATFs regardless of the outcome of the Stochastic Exclusion Test. The ATFs were asked to calculate the Stochastic Reserve based on two discounting methods (1) using the net asset earned rates of the model segment for the scenario, and (2) using 105% of the one-year U.S. Treasury interest rates for the scenario. As of the writing of this report, VM-20 has adopted the latter methodology with this guidance: Guidance Note: The use of different discount rate paths for the Seriatim and Scenario Reserves is driven by differences in methodology. The Seriatim Reserve is based on a present value of all liability cash flows, with the discount rates reflecting the investment returns of the assets backing the liabilities. The Scenario Reserve is based on a starting estimate of the reserve, and assets that support that estimate, plus the greatest present value of the accumulated deficiencies. Here, the discount rates are a standard estimate of the investment returns of only the marginal assets needed to eliminate either a positive or negative deficiency. Components of the Stochastic Reserve determination under both discounting methods are provided in Table V-A and Table V-B. For policies that qualify for exclusion from stochastic modeling using the Stochastic Exclusion Test, VM-20 requires a Modified Deterministic Reserve be used to determine any stochastic excess over the Deterministic Reserve. Table V-A shows the Modified Deterministic Reserve for those block with a stochastic exclusion test ratio less than 4.0%. The Modified Deterministic Reserve is described in Section C.4.1 of VM-20. Results shown are for the 20-year level premium products for all term blocks. The fact that term block T-1 has a $0 Stochastic Reserve amount is reasonable when compared with the negative seriatim reserves for this block shown in Table II. 18

19 TERM INSURANCE With one exception, there are only minor differences in the Stochastic Reserves under the two different discounting methods. TABLE V-A STOCHASTIC RESERVE ANALYSIS 20 YEAR LEVEL PREMIUM PERIOD TERM BLOCK Deterministic Reserve S.E.T. Ratio Stochastic Reserve Modified (70CTE) Deterministic Reserve NAER 105% UST T-1 $ 692 < 4.0% $ 692 $ 0 $ 0 T-2 7,867 < 4.0% 7,911 8,145 8,394 T < 4.0% T-4 1,947 < 4.0% 1, VM-20 requires the aggregate starting assets to be within 2% of the final aggregate minimum reserve. For purposes of the research, ATFs were asked to iterate to a starting asset amount within 2% of the 70CTE Stochastic Reserve. All the ATFs were able to find a starting asset amount within or reasonably close to the +/- 2% requirement. Iterating to this 2% limit was one of the most time-consuming aspects of the work for most ATFs. TABLE V-B STARTING ASSETS 20 YEAR LEVEL PREMIUM PERIOD TERM BLOCK NAER as discount Starting Assets Starting Asset Ratios 105% UST as discount Starting Assets Starting Asset Ratios T-1 $ 0 N/A $ 0 N/A T-2 7,840 96% 7,840 93% T % % T % % 19

20 TERM INSURANCE STOCHASTIC RESERVE DISTRIBUTION Distribution of the 1,000 scenario reserves for blocks T-2, T-3, T-4 are shown below. 20

21 TERM INSURANCE 21

22 TERM INSURANCE MINIMUM RESERVE The Minimum Reserve required by VM-20 is the Deterministic Reserve plus the excess, if any, of the Stochastic Reserve over the Deterministic Reserve. Table VI shows these two components of the VM-20 requirements for the 20 Year Level Premium Term blocks. It also shows the ratio of the VM-20 Minimum Reserve to the current Formulaic Net Statutory Reserve from Table I. TABLE VI MINIMUM RESERVE 20 YEAR LEVEL PREMIUM PERIOD TERM BLOCK Deterministic Reserve Excess of Stochastic Reserve over Deterministic Reserve VM-20 Minimum Reserve VM-20 as a percentage of Net Statutory Reserve less Deferred Premium T-1 $ 692 $ 0 $ % T-2 7, ,911 83% T % T-4 1, ,947 32% These results suggest a wide range of possible VM-20 outcomes when compared to current Statutory Reserve standards. The range is influenced by: Sufficiency of the premium to support benefits, expenses and profit Maturity of the block Level of margins used in the valuation Characteristics of the asset portfolio supporting the policies 22

23 TERM INSURANCE C3 PHASE III CAPITAL COMPONENTS All of the ATFs modeling term insurance used the same models and assumptions for purposes of field testing the C3 Phase III (C3P3) capital components, except federal income tax payments were included and the projections considered a working reserve equal to the cash surrender value (or $0, if none). Although NAIC RBC requirements for a reporting entity are aggregate measures, it is useful to compare the PBA total asset requirement (TAR) under the C3P3 proposal to an approximation of the block s contribution to the C3, or interest rate risk, component of the current RBC calculation, where this comparable amount is determined as defined in the footnotes below. For the business segments subject to this research, the C3P3 TAR is shown in Table VII in the column labeled C3P3 Stochastic Amount. The C3P3 Reported Amount, shown in the final column of Table VII, is equal to the C3P3 Stochastic Amount less the net statutory reserve (formulaic). To the extent the business segment being modeled in this research is aggregated with other business segments, the Reported Amount shown below may actually contribute an offset (i.e. carry a negative value) to the total Reported Amount. For purposes of this report, the C3P3 Reported Amount for the business segments subject to this research is floored at zero. TABLE VII C3 PHASE III - TERM TERM BLOCK Level Premium Period Starting Assets (FN1) Starting Assets as Pct of Statutory Reserve Net of Reinsurance (FN2) C3P3 Stochastic Amount Comparable Formulaic RBC Amount (FN3) C3P3 Reported Amount (FN4) T-1 10 $ 2, % $ 1 $ 2,956 $ 0 T , % 405 3,768 0 T , % 591 4,152 0 T , % 7,541 10,256 0 T , % 377 1,381 0 T , % 0 4,126 0 T ,648 94% 2,735 6,058 0 FN1: C3P3 requires the starting assets be at least equal to 98% of the reserve and other liabilities on the policies being valued. FN2: Starting Assets divided by Gross Reserve less Reinsurance Reserve Credit from Table I FN3: For purposes of this report, the comparable RBC amount is determined as [ % * (1-.35)] * (Statutory Reserve net of reinsurance less Policy Loans) from Table I FN4: Equal to C3P3 Stochastic Amount less the Statutory Reserve net of reinsurance from Table I, but not less than zero. If these blocks of term insurance are thought of as being subject to VM-20 at the time C3P3 becomes operative, then the VM-20 Minimum Reserve from Table VI (for the 20 year level term periods) and the C3P3 Stochastic Amount from Table VII provide a pro-forma C3P3 capital requirement on a full principle-based platform. For term, this results in values less than $0 for all but T-4, which is $788 ($2,735 1,947). 23

24 TERM INSURANCE OBSERVATIONS The field test results for term insurance provide insight into the range of results to be expected when valuing level premium term insurance under a principle-based methodology. Summarized below are several key observations presented in detail above. The cash value floor requirement can have a material impact on Per Policy Reserves for term insurance. Its significance is greater for less competitive premium products than for more competitive premium products. The cash value floor was referred to by some participants as a secondary margin since, for certain policies, the Seriatim Reserve is a negative value. The first layer of margin brings the Seriatim Reserve from anticipated experience assumptions to prudent estimate experience assumptions. The required cash value floor provides a second layer of margin to the result. The Deterministic Reserve for term insurance is expected to be less than the Formulaic Statutory Reserve at most policy years. In early issue years, the Deterministic Reserve may exceed the Formulaic Reserve due to early issue and commission expenses. This research provided Deterministic-to-Formulaic results ranging from 0% for a relatively newer, less competitive, 30-year level premium term product to 82% for a more mature, competitive, 20-year level premium product. With the exception of the first policy year, the pattern of Deterministic Reserves generally follows the pattern of Formulaic Reserves. The results of this research suggest a wide range of possible outcomes under VM-20 requirements when compared to current statutory minimum reserve standards. The range is influenced by: Sufficiency of the premium to support benefits, expenses and profit Maturity of the block Level of margins used in the valuation Characteristics of the asset portfolio supporting the policies Ignoring one high outlier result, margins accounted for 48% to over 200% of a Per Policy Reserve based on anticipated experience. The Stochastic Exclusion Test ratio is generally less than 4% for term insurance and depends upon the length of the premium guarantee and maturity of the block in question. Earlier policy years may have test ratios in excess of 4%, but these quickly fall to less than 4% for later years. There is little, if any, excess of Stochastic Reserve over Deterministic Reserve in the field tested blocks of term insurance. However, one block exhibited a higher Stochastic Reserve than modified Deterministic Reserve while still meeting the Stochastic Exclusion Test criteria. The credit for YRT reinsurance ceded is greater under principle-based reserving than under current formulaic statutory requirements since reinsurance benefits and premiums beyond the current policy year are recognized. Principle-based risk-based capital requirements for the C3 component of term insurance are less than current formulaic C3 requirements. The blocks subject to this research would qualify for the stochastic modeling exclusion, if the company chose that path. 24

25 UNIVERSAL LIFE INSURANCE BLOCK DESCRIPTION In this section, research results are presented for the modeled universal life insurance products, followed by a section of observations and discussion. The results have been normalized using a unit basis of $1,000,000 of insurance amount inforce and rounding to the nearest dollar. Six blocks of universal life insurance are modeled as part of this research. The blocks include single life and joint life (UL-4) business. All blocks include a secondary no-lapse guarantee (NLG) as part of the policy provisions, some operating on a shadow account design and others on a stipulated premium design, as noted below. UL BLOCK UL-1 UL-2 UL-3 UL-4 UL-5 UL-6 Risk Class Structure First two eras: 1 NS, 1 S Last era: 3 NS, 1 S 4 Nonsmoker 2 Smoker 4 Nonsmoker 2 Smoker 3 Nonsmoker 1 Smoker 5 Nonsmoker 3 Smoker 3 Nonsmoker 2 Smoker Issue Years Projection Period Oct. 1, 2008 through Sept. 30, 2038 Oct. 1, 2008 through Sept. 30, 2087 Oct. 1, 2008 through Sept. 30, 2127 Oct. 1, 2008 through Sept. 30, 2127 Oct. 1, 2008 through Sept. 30, 2038 Oct. 1, 2008 through Sept. 30, 2038 Type and Length of Secondary Guarantee Period Three eras of UL product, only one having a 9 year specified premium NLG feature Shadow Account over Lifetime 0% Shadow Account; Long Term guarantee provided by a No Lapse Guarantee rider 0% Shadow Account; Long Term guarantee provided by a No Lapse Guarantee rider 0% Stipulated premium; 10 Years for most ages, 5 years for older issue ages Shadow Account over Lifetime 0% Lapse Assumption when Secondary Guarantee is in Play Most policies are at or beyond the NLG period. Baseline lapse assumption applicable during NLG. Most policies are at or beyond the NLG period. Baseline lapse assumption applicable during NLG. Each ATF was directed to use a projection period long enough to capture material cash flows. 25

26 UNIVERSAL LIFE INSURANCE FORMULAIC STATUTORY COMPONENTS To facilitate comparisons to current formulaic reserve requirements, the components of the current NAIC Statutory Reserves are shown in Table I. Note that each block listed includes varying numbers of issue years in its inforce. TABLE I STATUTORY RESERVE AND RELATED AMOUNTS UL BLOCK Basic Reserve Deficiency Reserve Gross Reserve UL-1 $ N/A $ N/A $122,672 UL-2 55, ,312 UL-3 50,036 4,348 54,384 UL-4 29, ,731 UL-5 146, ,718 Reinsurance Type Reinsurance Reserve Credit Net Reserve for Comparison to VM-20 Cash Value YRT Risk Transfer $ 853 $121,819 $ 115,204 YRT Risk Transfer 0 55,312 30,793 YRT Risk Transfer 0 54,384 38,232 YRT Risk Transfer 0 29,731 16,961 YRT Risk Transfer , ,773 UL-6 78,168 31, ,189 None N/A 109,189 50,015 26

27 UNIVERSAL LIFE INSURANCE DETERMINISTIC COMPONENTS Modeling results for the deterministic components of VM-20 are shown in Table II followed by a discussion of Items #1 through #6. TABLE II VM-20 DETERMINISTIC COMPONENTS UL BLOCK Item #1 Sum of Seriatim Reserves Item #2 Sum of Per Policy Reserves (FN1) Item #3 Sum of Per Policy Reserves ignoring Reinsurance (FN2) Item #4 Sum of Per Policy Reserves using Anticipated Experience Item #5 Item #4 with Mortality Margin only Item #6 Item #4 with Lapse Margin only UL-1 $ 59,675 $116,088 $116,195 $115,965 $116,114 $115,582 UL-2 31,662 36,971 64,346 34,907 34,988 35,020 UL-3 19,006 39,711 45,497 38,811 38,925 39,486 UL-4 24,611 28,281 40,158 22,959 23,131 28,205 UL-5 124, , , , , ,824 UL-6 114, ,075 N/A (FN3) (FN3) (FN3) FN1: In all but one case, Item #2 is at least as great as Cash Value from Table I. The one exception is due to modeling variances that would not likely arise in a production valuation exercise. FN2: N/A implies reinsurance is not applicable. FN3: ATF did not calculate. Item #1, the Seriatim Reserve, is used as a step in the calculation of the Deterministic Reserve and equals the present value of cash flows for each policy, discounted at the net asset earned rate for the model segment under a single deterministic scenario (Section C.2 of VM-20, September 2008). A negative Seriatim Reserve at the valuation date indicates the premiums are sufficient to fund future benefit and expense obligation cash flows. Item #2 is the sum of Per Policy Reserves for the block of policies being valued. The Per Policy Reserve is the greater of the Seriatim Reserve and the policy s cash surrender value, adjusted for reinsurance. For purposes of this research, we are treating item #2 as the Deterministic Reserve, although in a total company situation, the Deterministic Reserve would equal the sum of the Per Policy Reserves over all policies subject to VM-20. Item #3 is the same as Item #2, but ignoring reinsurance, if any. Item #4 is the same as Item #2, but excluding all margins on risk factors. Item #5 is Item #4, with the mortality assumption being the only risk factor with a margin. Item #6 is Item #4, with the lapse assumption being the only risk factor with a margin. 27

28 UNIVERSAL LIFE INSURANCE COMPARISON OF DETERMINISTIC COMPONENTS TO FORMULAIC STATUTORY Table III-A summarizes comparisons of the VM-20 Deterministic Reserves to Formulaic Net Statutory Reserves as well as the impact of reinsurance and the margins applied to mortality risk and lapse risk factors. The percentages are based on normalized results before rounding. TABLE III-A COMPARISON OF DETERMINISTIC TO STATUTORY RESERVES/ MARGIN ANALYSIS A B C D E UL BLOCK Item #2 / Net Statutory (FN1) Item #3 / Gross Statutory (FN2) Impact of All Margins (FN3) Impact of Mortality Margin (FN4) Impact of Lapse Margin (FN5) UL-1 95% 95% 0% 0% 0% UL-2 67% 116% 6% 0% 0% UL-3 73% 84% 2% 0% 2% UL-4 95% 135% 23% 1% 23% UL-5 99% 99% 0% 0% 0% UL-6 116% N/A (FN6) (FN6) (FN6) FN1: Item #2 from Table II divided by Net Reserve for comparison to VM-20 from Table I. FN2: Gross Reserve from Table I. FN3: [Item #2 - Item #4] / Item #4, from Table II. FN4: [Item #5 - Item #4] / Item #4, from Table II. FN5: [Item #6 - Item #4] / Item #4, from Table II. FN6: Table II data not available. Column A in Table III-A compares Deterministic Reserves to Statutory Net Reserves. The range demonstrated is 67% to 116%. Column B in Table III-A compares the sum of Per Policy Reserves ignoring reinsurance to the gross Statutory Reserve. The majority of the universal life blocks were modeled with mortality risk transfer YRT reinsurance arrangements. The data submitted by the ATFs allows for some analysis with respect to the impact of modeling reinsurance cash flows in a principle-based regime. In some cases, the impact is quite different under principle-based than under statutory rules. Note that at the time of this research, the NAIC s position on reinsurance within VM-20 and specifically the reinsurance reserve credit was not settled. The analysis in Table III-A.1 below defines the principle-based reserve credit as the difference between the Deterministic Reserve with and without reinsurance cash flows. 28

29 UNIVERSAL LIFE INSURANCE TABLE III-A.1 IMPACT OF REINSURANCE MODELING UL BLOCK Deterministic Reserve (Item #2) Deterministic Reserve without Reinsurance Cash Flows (Item #3) Difference Statutory Formulaic Reinsurance Reserve Credit UL-1 $ 116,088 $ 116,195 $ 107 $ 853 UL-2 36,971 64,346 27,375 0 UL-3 39,711 45,497 5,786 0 UL-4 28,281 40,158 11,877 0 UL-5 144, , Comments provided by each ATF regarding reinsurance modeling follow: UL-1 UL-2 The mortality risk of each policy was 50% reinsured on a YRT basis. A reinsurance margin was not tested. YRT reinsurance was modeled. Our YRT reinsurance does not impact the statutory reserve because monthly premiums are paid in arrears. There is no reinsurance reserve credit. We included a 5% margin for reinsurance expense by increasing YRT premium rates. UL-3, 4 Many of the modeled policies have a percentage of the mortality risk reinsured to a pool of carriers on a YRT basis. The ceded percentage for each model cell is the actual ceded percentage for the policy. The premium rates modeled are the treaty rates. There are no statutory reserve credits taken for YRT reinsurance because the premiums are paid quarterly in advance. The modeling indicates that with prudent estimate mortality assumptions, the reinsurance provides a benefit that reduces the principle-based reserve. When run with pricing-type assumptions, the company does not see the same magnitude of benefit from the reinsurance. In fact, the reinsurance is generally perceived as a cost. We did not include any margins on the reinsurance cash flows. UL-5 A very limited portion of the business is reinsured, approximately 0.4% by face amount. YRT premiums are paid for reinsuring 90% of the net amount at risk on this portion of the total face amount. 29