Session 36 PD, Mortality Assumption Setting for Pension Actuaries. Moderator: Andrew J. Peterson, FSA, EA, FCA, MAAA

Transcription

1 Session 36 PD, Mortality Assumption Setting for Pension Actuaries Moderator: Andrew J. Peterson, FSA, EA, FCA, MAAA Presenters: David T. Kausch, FSA, EA, FCA, MSPA Laurence Pinzur, FSA

2 Session 36 - Panel Discussion: Mortality Assumption Setting for Pension Actuaries Moderator: Andrew J. Peterson, FSA, EA, FCA, MAAA Presenters: David T. Kausch, FSA, FCA, MAAA Laurence Pinzur, FSA

3 SOCIETY OF ACTUARIES Antitrust Notice for Meetings Active participation in the Society of Actuaries is an important aspect of membership. However, any Society activity that arguably could be perceived as a restraint of trade exposes the SOA and its members to antitrust risk. Accordingly, meeting participants should refrain from any discussion which may provide the basis for an inference that they agreed to take any action relating to prices, services, production, allocation of markets or any other matter having a market effect. These discussions should be avoided both at official SOA meetings and informal gatherings and activities. In addition, meeting participants should be sensitive to other matters that may raise particular antitrust concern: membership restrictions, codes of ethics or other forms of self-regulation, product standardization or certification. The following are guidelines that should be followed at all SOA meetings, informal gatherings and activities: DON T discuss your own, your firm s, or others prices or fees for service, or anything that might affect prices or fees, such as costs, discounts, terms of sale, or profit margins. DON T stay at a meeting where any such price talk occurs. DON T make public announcements or statements about your own or your firm s prices or fees, or those of competitors, at any SOA meeting or activity. DON T talk about what other entities or their members or employees plan to do in particular geographic or product markets or with particular customers. DON T speak or act on behalf of the SOA or any of its committees unless specifically authorized to do so. DO alert SOA staff or legal counsel about any concerns regarding proposed statements to be made by the association on behalf of a committee or section. DO consult with your own legal counsel or the SOA before raising any matter or making any statement that you think may involve competitively sensitive information. DO be alert to improper activities, and don t participate if you think something is improper. If you have specific questions, seek guidance from your own legal counsel or from the SOA s Executive Director or legal counsel. 2

4 Presentation Disclaimer Presentations are intended for educational purposes only and do not replace independent professional judgment. Statements of fact and opinions expressed are those of the participants individually and, unless expressly stated to the contrary, are not the opinion or position of the Society of Actuaries, its cosponsors or its committees. The Society of Actuaries does not endorse or approve, and assumes no responsibility for, the content, accuracy or completeness of the information presented. Attendees should note that the sessions are audio-recorded and may be published in various media, including print, audio and video formats without further notice. 3

5 Selection of Base Mortality Rates DAVID T. KAUSCH, FSA, EA, MAAA October 12, 2015

6 Topics Applicable ASOPs Recap of RP-2014 tables Mortality experience studies Case studies 5

7 Applicable ASOPs 6

8 ASOP No. 35 Selection of Demographic and Other Noneconomic Assumptions for Measuring Pension Obligations Measurements of pension obligations includes valuations, generally not individual benefit calculations Many considerations including (among other things) Purpose of the measurement Characteristics of the obligation Characteristics of the covered group Must be a reasonable assumption (among other things) Appropriate for the purpose of the measurement Reflects actuary s estimate of future experience Has no significant bias 7

9 ASOP No. 35 Disclosure Requirements Disclose each significant assumption (e.g., mortality) Disclose mortality improvement assumption (even if assumption is zero future improvement) Uncertainty about the occurrence or magnitude of future mortality improvement does not by itself mean that an assumption of zero future mortality improvement is reasonable Beyond the ASOP Scale MP-2014 represents the RPEC s current best estimate of future mortality improvement in the U.S. 8

10 ASOP No. 12 Risk Classification A Risk Classification System is a system used to assign risks to groups based upon the expected cost or benefit of the coverage or services provided Applicable to plan populations or subgroups General Considerations (among other things) Relationship between risk characteristics and expected outcomes, causality is not required Differences in rates should reflect material differences in expected cost for risk characteristics Consider interdependence of risk characteristics Actuarial Considerations Adverse selection Credibility (ASOP No. 25, next slide) Practicality 9

11 ASOP No. 25 Credibility Credibility is a measure of the predictive value in a given application that the actuary attaches to a particular set of data Credibility Procedure The evaluation of subject experience for potential use in setting assumptions without reference to other data The identification of relevant experience and the selection and implementation of a method for blending the relevant experience with the subject experience 10

12 Credibility Analysis ASOP No. 25 Actuarial Considerations Whether the procedure is expected to produce reasonable results Whether the procedure is appropriate for the intended use and purpose Whether the procedure is practical to implement when taking into consideration both the cost and benefit of employing a procedure 11

13 Credibility Analysis A data set can have full, partial, or no credibility This is not always a precise science Generally, Full credibility assumption may be based on actual experience Zero credibility no weight given to actual experience Partial credibility gray area in between Case studies will go into this in more depth 12

14 Recap of RP-2014 Tables 13

15 Recap of RP-2014 Base Tables RPEC published Mortality Tables for Annuitants and Employees RP-2014 (sometimes referred to as the Total Dataset tables) Blue and White Collar Bottom and Top Quartile Disabled (one set for all ages) All tables are amounts based meaning Exposures and actual deaths are weighted by salary for Employee tables Exposures and actual deaths are weighted by benefit for Annuitant tables Additional set of tables published based on headcounts 14

16 Recap of RP-2014 Base Tables Employee Tables have rates from ages 18 to 80 Annuitant Tables have rates from ages 50 to 120 Disabled Tables have rates from ages 18 to 120 Unlike RP-2000, no Combined tables published RPEC suggests Use separate pre-retirement and post-retirement tables, or Develop plan-specific Combined table based on retirement expectations 15

17 Recap of RP-2014 Base Tables Blue Collar defined as at least 70% of plan population either hourly or union White Collar defined as at least 70% of plan population both salaried and non-union Bottom Quartile lowest amount quartile in total dataset (salary for employees, benefit for annuitants) Top Quartile highest amount quartile in total dataset (salary for employees, benefit for annuitants) Tables are intended to be applied to entire group 16

18 Dollar-Weighted and Headcount- Weighted Rates The main RP-2014 tables were developed on a dollar weighted basis Employees actual deaths and exposures weighted by salary Annuitants actual deaths and exposures weighted by benefit The purpose of dollar weighting is to better reflect impact of mortality in pension valuations Evidence suggests salary is a factor in mortality Pension benefits are usually stratified by salary Mortality gains and losses are weighted by benefits 17

19 Dollar-Weighted and Headcount- Weighted Rates OPEB valuation mortality gains and losses are also weighted by benefits However, OPEB benefits are generally not stratified with salary Headcount-weighted rates may be more appropriate SOA released each table with a headcount-weighted RPH version 18

20 Dollar-Weighted and Headcount- Weighted Rates General observations about the tables Headcount weighted tables have higher mortality rates for healthy employees and annuitants at most ages This is to be expected since generally higher pay/benefits correlate with lower mortality rates so when given more weight (i.e., dollar weighted) they bring rates down Impact on annuities varies with age and interest rates 19

21 Dollar-Weighted vs. Headcount- Weighted Annuities at 6% Male Immediate Annuities Female Immediate Annuities Age RP-2014 RPH-2014 Ratio Age RP-2014 RPH-2014 Ratio % % % % % % % % % % % % % % Factors are based on age in 2015 with fully generational MP-2014 projections 20

22 Mortality Experience Studies 21

23 Mortality Experience Studies Large plans may use experience studies Need to adjust for mortality improvement during the study For base tables, goal to get A/E close to 1 Additional analysis needed 22

24 Mortality Experience Studies Weight exposures and actual deaths Headcount Salary/Benefit Liability (may be a bit circular) Choice may depend on purpose of measurement Understand method used in published table when comparing to experience 23

25 Case Studies 24

26 Case Study 1 The XYZ Company has sizeable populations of both Salaried and Hourly employees The XYZ Pension Plan provides ongoing FA5 benefits to both Salaried and Hourly participants In the same plan, with the same underlying benefit formula No lump sum provision (other than small lump sum cash-outs) The XYZ Postretirement Medical Program provides a flat dollar subsidy; only for eligible Hourly retirees XYZ s actuary has performed a mortality experience study covering the years Analysis of annuitant mortality only; active mortality not credible 25

27 Case Study 1 Benefits-weighted A/E* results Salaried Annuitants Hourly Annuitants Total Annuitants RP-2014 RP-2014 WC RP-2014 BC Male 76% 96% N/A Female 87% 104% N/A Total 77% 97% N/A Male 114% N/A 97% Female 117% N/A 106% Total 115% N/A 101% Male 103% N/A N/A Female 108% N/A N/A Total 104% N/A N/A * Expected based on corresponding RP-2014 rates adjusted back to experience years with Scale MP

28 Case Study 1 Headcount-weighted A/E* results Hourly Annuitants RPH-2014 RPH-2014 BC Male 109% 98% Female 112% 104% Total 110% 100% Note: Flat postretirement medical benefit structure implies A/E analyses based on RPH-2014 rates! * Expected based on corresponding RPH-2014 rates adjusted back to experience years with Scale MP

29 Case Study 1 Numerous options for (gender-specific) base mortality rate assumption 1. Single set of rates for the both the pension plan and the postretirement medical plan 2. Single set of $-weighted rates for the pension plan and a separate set #-weighted rates for the Hourly postretirement medical plan 3. Separate sets of $-weighted rates for the Salaried and Hourly populations within the pension plan and a separate set #-weighted rates for the Hourly postretirement medical plan 4. Combine any of the above with credibility weighting 28

30 Credibility Analysis Full Credibility With enough data, may use 100% of experience for mortality rates Need to consider for each q x May wish to consider graduation of raw qs Suppose we want to be confident with probability p (e.g., 90%) that the observed q is within in r (e.g., 5%) of the actual q Keep in mind that r and p are subjective *The next few slides are based on a paper by Benjamin cited on References slide 29

31 Credibility Analysis Additional underlying assumptions Suppose we assume deaths are iid Suppose headcount weighting Then the number of observed deaths needed for full credibility is [(Z 1-(1-p)/2 )/r ] 2, (Z is from a standard normal) In our example with r=5% and p=90%, the number of observed deaths needed for full credibility is [(Z 1-(1-p)/2 )/r ] 2 = [(Z 0.95 )/0.05 ] 2 = [1.645/0.05 ] 2 = 1,082 for each q! Formulas are different for benefit weighted calculations 30

32 Credibility Analysis Few if any plans will see 1,082 deaths at each age and gender Options for other plans Combine experience (e.g., 5-year age brackets) keeping in mind iid assumption and test each age bracket for credibility Scale the published table by the ratio A/E from the experience study 31

33 Case Study 2 Statewide ABC Retirement System performs a 5-year mortality experience study Purpose of study is to update and select assumptions for annual actuarial valuations Total number of female deaths (all ages) in experience study period observed is 1,260 Total number of expected female deaths based on RP (adjusted as needed for study) is 1,215 The ratio of actual to expected deaths is 104% Is it reasonable to scale the RP-2014 female table by 104%? 32

34 Case Study 2 First, consider dollar weighting instead of headcount weighting (consistent with the development of the RP-2014 rates and the purpose of the assumption) Total actual dollars of benefits of deaths was $21.15 million Total expected dollars of benefits of deaths was $19.89 million Dollar weighted A/E ratio is 106% Is scaling RP-2014 by 106% a reasonable assumption? 33

35 Case Study 2 Rephrase the question in terms of credibility theory The parameter in question is f = A / E, the ratio of actual deaths over expected, the proposed scaling factor for the standard table On a headcount weighted basis we saw before that 1,082 deaths were needed for full credibility Therefore, 1,260 actual deaths implies full credibility Does the answer change on a dollar weighted basis? 34

36 Case Study 2 We are trying to determine if we have enough data so that PP ff ff < 0.05 > 90% Where ff is the true ratio of actual to expected Dollar weighted formulas differ (see References) Calculation results vary for each case Number of deaths needed for full credibility: 1,042 Dollars of benefits of deaths needed: $17.06 million 35

37 Case Study 2 We can conclude that f=106% is fully credible But is r=5% good enough? When the parameter is a mortality rate q, which may be 1-2%, being within 5% means an error of % When the parameter is the ratio f, which may be near 100%, being within 5% means a ratio of % May want to consider different r (and/or p?) To be sure f is within r=1% of the true ff, we need Number of deaths: 26,000 Dollars of benefits of deaths: $426 million 36

38 Partial Credibility Another alternative is to weight the actual experience and the published table If q x are the pertinent RP-2014 rates, with insufficient number/dollars of deaths for full credibility, can solve for Z to get appropriate weighted qs Z fq x + (1 Z)q x AAAAAAAAAAAA DDDDDDDDDDD Where ZZ = = 1,260 = 22% NNNNNNNNNNNN DDDDDDDDDDD 26,049 This results in using 22%*106% + 78%*100%= 101% of the standard table 37

39 Partial Credibility So what is the right answer 104%, 106% or 101%? Depends on the purpose of the measurement, the credibility of the data, and what we specifically mean by credibility (e.g., r and p) ASOP No. 25 leaves room for Professional Judgment (Section 3.4) Not always a precise mathematical process Can give actual experience full, partial, or zero weight based on a credibility procedure or without a rigorous mathematical model 38

40 References Practice Notes (American Academy of Actuaries): Selecting and Documenting Mortality Assumptions for Pensions Recently updated Q & A s Credibility Appendix Credibility Practice Note (Life Valuation Subcommittee) Selecting Mortality Tables: A Credibility Approach (SOA Research Project) 39

41 Disclaimers and Acknowledgement Readers are cautioned to examine original source materials and to consult with subject matter experts before making decisions related to the subject matter of this presentation This presentation expresses the views of the author and does not necessarily express the views of Gabriel, Roeder, Smith & Company Thank you to Francois Pieterse and Rebecca Stouffer who checked and peer reviewed this presentation 40

42 Selection of an Appropriate Mortality Projection Scale LAURENCE PINZUR, FSA

43 Mortality Projection Topics Quick review of the RPEC_2014 projection model Scale MP-2014 assumption set Alternate assumption sets Demonstration of software available on SOA website Note: Unlike RP-2014 mortality rates, RPEC published a single mortality improvement model. I.e., no collaror amount-specific versions of MP

44 RPEC_2014 Model & Assumption Sets RPEC_2014 Model + Committee- Selected Assumption Set Scale MP-2014 Rates User-Selected Assumption Set + Other 2D Rates 3

45 RPEC_2014 Model & Assumption Sets The RPEC_2014 Model 1. 2D W-H Graduation of Historical SSA Mortality Data 2. 2-Year Step-Back 3. Double Cubic methodology for smooth interpolation between historical MI rates and assumed Long-Term MI rates Assumptions 1. Long-term MI rates 2. Length of Convergence Periods Horizontal Diagonal 3. Blending percentages for horizontal and diagonal interpolations 4

46 Historical MI Unsmoothed SSA mortality data through Males Ages Calendar Years 5

47 Historical MI Graduated SSA mortality data through Males Ages Calendar Years 6

48 Assumptions Interpolation process starts with MI rates in Males 1 Ages 3 Assumed Long-Term Rates Calendar Years CP* 2 * CP = Convergence Period 7

49 MP-2014 Assumptions The committee-selected assumption set: 1. Long-term rates (LTR): Flat 1.0% through age 85 Sight linear taper to 0.85% at age 95 Then linear decrease to 0.0% at age Convergence periods: 20 years for both age/period and cohort effects (starting in 2007) 3. Blending of horizontal and diagonal interpolations: 50%/50% 8

50 MP-2014 Assumptions Focus on 1.0% LTR Assumption (through age 85) Average annual decline in the age-sex adjusted death rate in the US has been 1.07% from 1900 to 2011 and 0.97% from 1982 to 2011 * Considerable variation by age group and observation period Clear reduction in MI at oldest ages Many reasonable arguments why the future levels of MI in the US will decrease/increase/stay the same The most recent Technical Panels of the Social Security Advisory Board have recommended that the SSA increase its long-term MI assumptions to 1.0% -- or higher. * Long-range Demographic Assumptions For the 2015 Trustees Report (SSA OCACT; July 2015) 9

51 MP-2014: Heat Map (Males) 10

52 MP-2014: Heat Map (Females) 11

53 Alternate Assumption Sets 95 Heat Map based on SSA s Methodology* Males Males Social Security assumes a quick transition away from recent MI rates. Values of near-term MI rates are based entirely on horizontal (age-only) averages and so is the their projection into the future; i.e., no projection along year-of-birth cohort lines * Data for years prior to 2009 were smoothed 12

54 Alternate Assumption Sets Heat Map based 10-Year Convergence Periods Males No variation in assumed MI after CY years of historical year-ofbirth cohort effects disappear! 13

55 Alternate Assumption Sets Heat Map based 20-Year Convergence to SSA LTR and 50%/50% blend of horizontal and diagonal Males Year-of-birth cohort effects are restored Assumed flat LTR age segments are apparent on the right-hand edge 14

56 Alternate Assumption Sets MP-2014 Heat Map Males 15

57 Implementation Tools Develops 2D MI rates using the RPEC_2014 model with alternate assumption sets Develops one-dimensional MI rates from Scale MP rates (similar to Scale BB from Scale BB-2D) Resulting 1D rates still need to be applied generationally; i.e., no static projection functionality Study/Pension/research-2d-1d-con-tool.aspx/ 16

58 Questions 17