Post-NAIC Update/PBA Webinar

Transcription

1 Post-NAIC Update/PBA Webinar Donna Claire, FSA, MAAA, CERA Chair, American Academy of Actuaries Life Financial Soundness / Risk Management Committee (AKA PBA Steering Committee)

2 Agenda for Webinar Fall LHATF Update: Larry Bruning, Chairperson of NAIC's Life and Health Actuarial Task Force Life Risk-Based Capital Update: Philip Barlow, Chairperson of NAIC's Life RBC Working Group Credibility: Stuart Klugman, member of the Credibility Practice Note team Mortality Example: Mary Bahna-Nolan, Chairperson of the Academy's Life Experience Subcommittee Questions 2

3 Fall 2010 LHATF Update Larry Bruning, FSA, MAAA Chair, Life and Health Actuarial Task Force (LHATF) Chief Actuary, Kansas Insurance Department 3

4 LHATF Update Mortality Table Update Payout Annuity - Projection Scale and Margins being worked on with completion date of December 2010 Preneed and Guaranteed/Simplified Issue Data call December 2010 with submission due date March 2011 and expected draft experience table early 2012 American Academy of Actuaries Non-forfeiture Improvement Work Group Work continues with final report to LHATF March 2011 Amendments to the Standard Non-forfeiture Law for Life Insurance Conference calls will be scheduled to finalize amendments and expose a draft Amendments to Valuation Manual Amended VM-00 and exposed for comment Amended VM-31 and exposed for comment Amended VM-20 and exposed for comment 4

5 LHATF Update Statistical Agent/VM-51 VM-51 Subgroup chaired by NY has completed Mortality Experience Data Formats Requested that Society Of Actuaries assist in developing data formats for other experience data Feedback Loop for PBA Discussed objectives, scope of work and NAIC structure Formed a subgroup chaired by SC to develop a white paper for exposure Discussed the need to respond more quickly to requests for comment on International Issues Formed a subgroup chaired by KS to develop draft responses for LHATF consideration. NY will draft thoughts on Insurance Contracts. 5

6 LHATF Update Received update from Oliver Wyman on potential methodology changes to C-3 Phase 2 and AG 43 Three methodology changes are being considered and two will be tested Data results of testing projected in November/December timeframe LHATF will consider proposals after reviewing data testing results LHATF considered a referral from B Committee regarding contract reserves as defined in the MLR rebate calculation with no changes recommended LHATF adopted minutes of conference calls and the AHWG s Report 6

7 LHATF Update NAIC Impact Study of VM-20 Towers Watson selected as consultant Contract and Strategy on Confidentially is nearing completion Letter and Survey sent to 60 Company Universe October 12 Survey response deadline October companies (75 product variations) selected by October 29 Educational Sessions Planned Academy Webinar on Mortality & Credibility October 26 One or two day training for Actuarial Teams in Kansas City week of November 8 TW will be generating the 10,000 scenarios TW will be setting up capacity to run Models internally TW developing draft of specific metrics TW setting up communication via address: vm20.impactstudy.com TW setting up secure website for sharing testing data among Regulators & Teams 7

8 Life Risk-Based Capital Philip Barlow, FSA, MAAA Chair, NAIC Life Risk-Based Capital Working Group Associate Commissioner, Insurance Bureau Dept of Insurance Securities & Banking of Washington, D.C. 8

9 Life Risk-Based Capital Working Group Deferred the project to change the RBC treatment of mortgage loans to A sub-group has been working with the ACLI on modeling requirements. A call will be held to discuss the modeling, once the subgroup finishes its work. Released the Derivative Risk Mitigation proposal with a an amendment to adjust TAC for a timing mismatch on the accounting between the option and the asset. Also referred the mismatch issue to SAPWG. Adopted the proposal to exclude the value of non-u.s. affiliates from both the ACL-RBC and TAC effective for The WG will distribute a list of outstanding issues on C3P3. Once interested parties have the opportunity to review and provide input, the WG will address issues by conference call between now and the next NAIC meeting. The WG will also take responsibility for coordinating with LHATF on issues relevant to both reserves and capital. 9

10 Other NAIC actions The Capital Adequacy Task Force exposed for 30 days an Academy report on RBC treatment for Deferred Tax Assets. The exposure ends November 8. The CADTF also requested approval for a model law change to modify the trend test for Life RBC to begin at 300% instead of 250%. The C3 Phase II Results Subgroup held an organizational call and will resume work. The initial focus will be on the information provided by Oliver Wyman. The subgroup expects to have several open calls in the near future. 10

11 Limited Fluctuation Credibility Stuart Klugman, FSA, CERA Member, Academy s Credibility Practice Note Work Group Staff Fellow, Education, SOA

12 Agenda Why look at limited fluctuation? How limited fluctuation works A few insights How it applies to VM-20 Using amount of insurance 12

13 Why look at limited fluctuation? There are two kinds of formal credibility in common use: Limited Fluctuation Greatest Accuracy All texts point to greatest accuracy as being superior But, it requires more information Also, the formulas are more complex Finally, limited fluctuation is intuitive, particularly in the context of actuarial applications 13

14 Review of limited fluctuation A data set is assigned full credibility provided the probability that the relative error in the estimate is less than r% with probability at least p% r and p are completely arbitrary The standard formulas assume that the estimate has a normal distribution 14

15 Derivation Let m be the unknown quantity and let X be the estimator of m based on a sample of size n Then X can be assigned full credibility (that is, used as is) provided X m Pr r /100 p /100 m Suppose X has a normal distribution with mean m and variance 2 /n 15

16 Derivation, continued X m p /100 Pr r /100 m Pr rm /100 X m rm /100 rm /100 X m rm /100 Pr / n / n / n rm /100 z / n z n rm /

17 Derivation, completed In the full credibility standard, z is the (1+p/100)/2 percentile from the standard normal distribution Example suppose r = 3% and p = 90% Then there is full credibility if: n m ( / m) 2 17

18 Example Suppose you want to estimate the mortality probability at a specific age. Further suppose that the true probability is It is reasonable to assume these observations are independent and identically distributed Bernoulli trials. Then the mean is 0.01 and the variance of one observation is 0.01(0.99) = The standard for full credibility is a sample size of 3007(0.0099)/(0.01) 2 = 297,693 18

19 Isn t the standard usually in terms of the number of deaths? Consider the example. Suppose, instead of trying to guess at the true value, we use the data to estimate it (yes, this is circular). What if we had 50,000 observations and 500 deaths? Then the estimate is 0.01 and we end up at the same place. Now assume that the variance equals the mean (by noting that 1 q is near 1, or assuming Poisson in place of Bernoulli). Let n be the sample size and d the number of deaths 19

20 Example, continued There will be full credibility if n > 3007(d/n)/(d/n) 2 This simplifies to d > 3007 and thus if there are more than 3007 deaths, the estimate has full credibility 20

21 Isn t that a bit odd? If I have really good mortality, I will need to have a larger sample size to achieve full credibility. That doesn t seem right. The reason is that full credibility is about relative error. If I have 10% mortality, then a 3% error margin means my estimate need only be between 9.7% and 10.3%. With 1% mortality, I now need to be between 0.97% and 1.03% to achieve full credibility. 21

22 Partial credibility The square root rule is commonly used. Take the square root of the ratio of the actual number of deaths to the number required for full credibility. 22

23 VM-20 If less than 30 deaths, use the standard table. Using the 3007 standard (not enshrined in the document), this means that if credibility is less than 10%, don t bother. No credibility method is specified; it only has to be based on a statistical method consistent with accepted actuarial practice The experience data is to be blended with the standard table 23

24 VM-20 margins The margin is a percentage increase to the credibility weighted result. The adjustment can vary by age. It is to cover two kinds of risk. Random fluctuation risk due to sampling. Let f be a measure of the margin needed for this risk. Company variation risk due to inherent differences between the company and the selected table. Let c be a measure of the margin needed for this risk. The total margin is Z*f + (1 Z)*c 24

25 Random fluctuation risk The VM-20 draft says this should be higher for cases with low credibility. It should also be between 1% and 10%. When multiplied by Z, this should have the following pattern: With low credibility the margin is small With high credibility the margin is small In between the product increases, then decreases. VM-20 does not set a pattern for f, but the final result should look like the above 25

26 Company variation risk The value of c should be constant. That is, how a company inherently differs from a standard table has nothing to do with the standard table or the amount of data. Note that with no credibility, all the margin is determined by c. This is similar to when standard tables with margins were mandated. As less reliance is placed in the standard table, this part of the margin formula will be reduced, eventually reaching zero when full credibility is attained 26

27 When mortality is by amount When using deaths as the standard for full credibility, there is an assumption that the experience rates are by policy, not by amount When using amounts, there will be less credibility assigned to the experience data Why? 27

28 Why? Credibility is about the accuracy of the estimate. When by amount, there is an additional random factor is the death from a high value policy or a low value policy? If by chance there are the expected number of deaths, but they happen to be for smaller amounts, the estimate will be understated. So why use amounts? 28

29 Why amounts? If there is no difference in mortality by amount, there really is no good statistical reason to use amounts. They add noise, not accuracy. If there is a difference, then the mix of sales has an impact. As long as the future mix will be similar, this will provide a more accurate estimate of future mortality. 29

30 Formulas? They are in Credibility Theory Practices at 30

31 Mortality Example Mary J. Bahna-Nolan, FSA, CERA, MAAA Chair, Life Experience Subcommittee 31

32 Deterministic/Stochastic Mortality Assumption Seven step approach to determine prudent estimate mortality assumption 1. Determine credibility segment 2. Determine mortality segment 3. Determine company experience mortality rates and credibility procedure (if not using simplified method) 4. Determine applicable industry mortality table 5. Determine credibility adjusted mortality rates (if not using simplified method) 6. Determine margin 7. Determine prudent estimate mortality rates 32

33 Step 1: Determine credibility segments Used to determine: Qualification for simplified method Credibility factor for use in determining overall margin for prudent estimate mortality assumptions Experience within credibility segment must be based on all in force claims data with study period within previous 3 years; study period can range from 3 10 years 33

34 Step 1: Determine credibility segments cont d Each segment shall consist of policies with similar underwriting methods and mortality experience characteristics Acceptable to group of policies with different types of insurance into the same credibility segment if the underwriting methods and mortality experience characteristics are similar for all the policies E.g., All fully underwritten flexible premium universal life policies May remove from the credibility segment(s): any policies for which the experience is reflected through adjustments to the prudent estimate mortality rate assumptions. For example: impaired lives Policies in which there is a reasonable expectation that policyholder behavior will lead to mortality results that vary significantly from those that would otherwise be expected such as additional mortality due to anti selective lapse at end of level premium term 34

35 Step 2: Determine mortality segments Used to determine separate: Credibility adjusted experience rates and Prudent estimate mortality tables Within each credibility segment: Group the policies company expects to have similar underwriting method and mortality experience E.g., All preferred non tobacco risks of the fully underwritten flexible premium universal life policies If actual experience data not available or has limited credibility: May use actual experience of one or more mortality pools in which the policies participate May include data from other sources if available and has the same underwriting and mortality experience characteristics to similar policies within the credibility segment 35

36 Example Steps 1 and 2 Credibility segment = all fully underwritten traditional life mortality (term and non-par WL) Nontobacco Mortality segment1 = preferred and super preferred nontobacco-traditional life Mortality segment2= residual nontobacco-traditional life 36

37 Step 3a: Determine if qualify for simplified method Method to determine anticipated experience assumption differs slightly depending on level of credibility Initial credibility for simplified method determined by # of claims for the particular Credibility Segment <30 use simplified method > 30 use credibility adjusted method 37

38 Example Step 3a Determine if qualify for simplified method For our credibility segment, # claims = 1,285 1,285 > 30 do not qualify for simplified method 38

39 Step 3b: Determine experience mortality rates With > 30 claims in credibility segment Choose credibility method/procedure With < 30 claims in credibility segment Choose industry mortality table Determine experience mortality Choose industry mortality table Credibility weight/blend with experience mortality 39

40 Step 3b: Determine experience mortality rates cont d If the credibility weighted method is used: Determine Credibility procedure Must be based on statistical method Statistical method must be consistent with standard actuarial practice Cannot use rule of thumb approach based on # of claims For example, Limited Fluctuation Method Determine credibility for each mortality segment and mortality experience cell Determine Credibility Factor (CF) for credibility segment 40

41 Step 3b: Determine experience mortality rates cont d Experience mortality Must be reviewed at least once every three years and updated as needed Includes the in force and claim data pertaining to the study period for all policies currently in the credibility segment or that would have been in the credibility segment at any time during the period over which experience is being evaluated Study period must be between 3 and 10 years Use actual mortality experience directly applicable to the credibility segment, when available May use actual experience data of one or more mortality pools in which the policies participate under the terms of a reinsurance agreement provided that the policies in the credibility segment have underwriting methods and mortality experience characteristics similar to those of the policies in the pool; and the aggregate pool data are a available to the company 41

42 Example Step 3b Experience mortality study for credibility segment Duration (All Ages Exposure Actual Claims Expected Claims Actual to Expected Ratio Mortality Rate Confidence Interval Percent Combined) Count Amount Count Amount Count Amount Count Amount per 1000 Min Max Credibility 1 600, ,720, , , % 2 555, ,700, , , % 3 700, ,000, , , % 4 500, ,000, , , % 5 350,000 95,000, , , % 6 275,000 45,000, , , % 7 195,000 30,000, , , % 8 88,000 15,000, , , % 9 29,000 8,000, , , % Company ABC Mortality Study Experience period: January 1, 2005 to December 31, 2009 Combined Traditional Life by Duration Gender: All Tobacco Status: Nonsmoker Underwriting Classes: All, excluding substandard Expected Basis: 2008 VBT RR100 ANB Credibility Factor Total 3,292, ,420,000 1, ,755 1, , % 42

43 Example Step 3b Experience mortality study for mortality segment Male, Preferred NT, Company ABC Mortality Study Experience period: January 1, 2005 to December 31, 2009 Combined Traditional Life by Age Gender: Male Tobacco Status: Nontobacco Underwriting Classes: Preferred and Super Preferred Expected Basis: 2008 VBT RR100 ANB Age (All Durations Exposure Actual Claims Expected Claims Actual to Expected Ratio Mortality Rate p/1,000 Confidence Interval Percent Combined) Count Amount Count Amount Count Amount Count Amount Count Amount Min Max Credibility ,315 5,447, , % 193.8% % ,700 16,144, , , % 148.7% % ,000 24,056, , , % 93.7% % ,500 35,262, , , % 78.6% % ,360 49,114, , , % 79.5% % ,900 61,383, , , % 78.8% % ,970 38,740, , , % 79.6% % ,700 29,070, , , % 87.5% % ,425 23,510, , , % 92.1% % ,760 12,029, , , % 93.4% % Total 996, ,758, , , % 85.7% % , ,497, , , % 79.2% % 43

44 Example Step 3b Experience mortality study for mortality segment Male, Preferred NT, by duration Company ABC Mortality Study Experience period: January 1, 2005 to December 31, 2009 Combined Traditional Life by Duration Gender: Male Tobacco Status: Nontobacco Underwriting Classes: Preferred and Super Preferred Expected Basis: 2008 VBT RR100 ANB Duration (All Ages Exposure Actual Claims Expected Claims Actual to Expected Ratio Mortality Rate per 1000 Confidence Interval Percent Combined) Count Amount Count Amount Count Amount Count Amount Count Amount Min Max Credibility 1 185,430 67,405, , , % 2 179,920 47,359, , , % 3 199,470 68,263, , , % 4 150,540 41,055, , , % 5 105,020 32,395, , , % 6 82,510 17,055, , , % 7 58,490 12,030, , , % 8 26,470 5,835, , , % 9 8,780 3,360, , % Total 996, ,758, , , % 85.7% % 44

45 Example Step 3b Experience mortality study for mortality segment Male, Preferred NT, by duration cont d Duration (All Ages Exposure Actual Claims Expected Claims Actual to Expected Ratio Mortality Rate per 1000 Confidence Interval Percent Combined) Count Amount Count Amount Count Amount Count Amount Count Amount Min Max Credibility 1 185,430 67,405, , , % 2 179,920 47,359, , , % 3 199,470 68,263, , , % , ,730, , , % Total 996, ,758, , , % 85.7% % 45

46 Example Step 3b Experience mortality study for mortality segment Male, Preferred and Super Preferred NT Issue Age A/E % credible A/E % credible % Experience Credible % overall 38 By Duration 15% 15% 18% 16% 14% 11% 12% 9% 4% 0% 0% 0% 0% 40% 39 Dur 1,2,3,4+ 15% 15% 18% 29% 40% % Credibility Used 15% 15% 18% 29% 31% 32% 34% 36% 38% 40% 40% 40% 40% 40% % A/E, E=2008 VBT RR % % % Weighted % VBT % % % % % % % % % Issue Ages dur 4+ All Age, All duration A/E & credibility Actual to Expected Ratio All Ages % All Ages Durations 1,2,3,

47 Example Step 3b Experience mortality study for mortality segment Male, Preferred and Super Preferred NT cont d Overall Preferred and Super Preferred NT mortality Mortality Experience Factor as % of 2008 VBT RR overall Expected split between classes = 50% Super Preferred, 50% Preferred Verified based on experience of cases written Confirmed by reinsurance underwriting assessment Assume 5 percentage point reduction in mortality for super preferred Use conservation of total deaths principle Mortality Experience Factor as % of 2008 VBT RR overall Super PNT -5% PNT 5%

48 Example Step 3b Experience mortality study for mortality segment Male, Preferred and Super Preferred NT cont d Best estimate mortality factors as % 2008 VBT RR100 Credibility blend with expected table (RR100) Mortality Experience Factor as % of 2008 VBT RR overall Super PNT -5% PNT 5% % Experience Credible overall Credibility Used 15% 15% 18% 29% 31% 32% 34% 36% 38% 40% 40% 40% 40% 40% Best Estimate Mortality Experience Factors = (EF * CF) * (1-EF)* overall Super PNT PNT

49 Step 4: Determine applicable industry table Determined for each mortality segment Applicable industry tables currently the 2008 VBT Tables If none of the 2008 VBT tables appropriately reflect the risk characteristics of the mortality segment, company may use any well established industry table based on similar or appropriate risk characteristics of policies within the mortality segment May use underwriting criteria scoring procedure (UCS) and Underwriting Criteria Scoring Calculator Found on the SOA website, life/2008 score calc.aspx May use alternative method to determine applicable basic table Must be based on sound method(s) Must document reasons approach to selection more appropriate for mortality segment than that from applying UCS Calculator 49

50 Step 4: Determine applicable industry table cont d May make adjustments to account for: Factors not recognized or quantified in the UCS algorithm but factor into expected mortality for particular mortality segment Experience of underwriting staff Number of underwriting exceptions Simplified underwriting May adjust table resulting from calculator up or down 2 tables 50

51 Example Step 4 - Determine applicable industry basic table Using UCS Calculator: Super Preferred NT RR80 Nontobacco Preferred NT RR90 Nontobacco Standard NS RR110 Nontobacco Smoker RR100 Tobacco Determine no further adjustments to table warranted 51

52 Step 5: Determine credibility adjusted experience rates If qualify for simplified method: Credibility adjusted experience rates equal applicable industry table Credibility adjusted experience rates are not actually defined under the simplifed method If do not qualify for simplified method: Blending experience mortality with applicable industry table based on credibility of underlying experience data Where experience data is only partially credible or credible for only a period of time: Grade into applicable industry mortality rates using reasonable method consistent with accepted actuarial practice 52

53 Example Step 5 Determine credibility adjusted mortality experience rates Male, Super Preferred NT - RR100 Anticipated Mortality Experience Issue Duration Age SPNT Fctr Blending with RR80 Credibility Factor for Male, SPNT = 40% Issue Age Duration Credibility for Qx Segment

54 Example Step 5 Determine credibility adjusted mortality experience rates cont d Issue Duration Age PNT Fctr Blending with RR90 Credibility Factor for Male, PNT = 40% Issue Age Duration Credibility for Qx Segment

55 Step 6: Determine margin Mortality margin in the form of a percentage increase applied to the Anticipated Experience Assumption 2 components to margin Random fluctuation risk Company variation risk 55

56 Step 6: Determine Margin cont d Random fluctuation risk Covers deviations in the mortality experience resulting from periodic variations of the experience from the mean random fluctuation from the expected results of credible component of a company s mortality Ranges from 1% to 10% 56

57 Example Step 6a Determine margin for random fluctuation risk Set to be 6% Considerations: Though durational experience has limited credibility, overall is close to fully credible Mortality and underwriting are homogonous 57

58 Step 6: Determine Margin cont d Company variation risk Covers deviations from a selected industry mortality due to differences in underwriting practices and the demographics of the underlying insured lives. Fixed amount which varies by issue age Ranges from 21% at younger ages to 9% at older ages For companies with full credibility = 0% 58

59 Step 6: Determine Margin cont d Company variation risk percentage margin table Issue Age Load Issue Age Load <45 21% % % % % % % % % % % 77+ 9% % 59

60 Example Step 6b Determine margin for company variation risk Margin chosen from table based on issue age Issue ages = 21% Issue ages = 20% Issue ages = 19% 60

61 Step 6: Determine Margin cont d Margin within each mortality segment = CF x margin for random fluctuation risk + (1 - CF) x margin for company variation risk where, CF = Credibility Factor determined for the credibility segment determined using the selected credibility procedure (If using simplified method, CF = 0) Margin should be increased to reflect situations involving greater uncertainty 61

62 Step 6: Determine Margin cont d Examples of situations involving greater uncertainty: Reliability of the company s experience studies is low due to imprecise methodology Length of time since the data was updated or other reasons Underwriting or risk selection risk criteria associated with the mortality segment have changed since the experience on which the credibility adjusted experience rates are based was collected Data underlying the credibility adjusted experience rates lack homogeneity Unfavorable environmental or health developments are unfolding and are expected to have a material and sustained impact on the insured population Company s marketing or administrative practices or market forces (for example, the secondary market for life insurance policies) expose the policies to the risk of anti-selection 62

63 Example Step 6c Determine margin NT risks = CF * margin for random fluctution risk + (1-CF) * margin for company variation risk Issue Age Margin = 97% * 6% + (1-97%) * 21% = 6.5% = 97% * 6% + (1-97%) * 20% = 6.4% = 97% * 6% + (1-97%) * 19% = 6.4% Decide to increase margin by 1% due to turnover in chief underwriter and some senior staff 63

64 Step 7: Set Prudent Estimate Mortality Take credibility adjusted mortality experience rates Increase by margin Make any adjustments for: Anti selective mortality Substandard risk Any other adjustments 64

65 Example Prudent estimate for mortality = Credibility adjusted mortality rates * (1 + margin) Once determined, further adjust for anti-selective mortality Prudent Estimate Mortality Rates Male, Super Preferred NT Issue Age Margin Duration

66 Questions? 66