SHARING THE LONGEVITY RISK BETWEEN ANNUITANTS AND ANNUITY PROVIDER

Transcription

1 SHARING THE LONGEVITY RISK BETWEEN ANNUITANTS AND ANNUITY PROVIDER ANNAMARIA OLIVIERI University of Parma (Italy) Meeting of the Working Group on Risk ESSEC Business School Paris, 18 February 2013 * Based on joint work with Ermanno Pitacco, University of Trieste (Italy) Paris, 18 February 2013 p. p. 1 1/26

2 Motivation Traditional life annuities imply financial and longevity guarantees Longevity guarantees Basic (and essential, so far) feature of the life annuity: pooling longevity guarantee On the other hand: pooling unavailability of money for bequest life annuities not attractive in the early years after retirement The value of longevity guarantees has increased substantially in the latest decades, because of major unanticipated mortality improvements (+ reduction of interest rates) Their pricing requires stochastic models not yet fully reliable or not yet easily tractable Further: if applied, currently the fees for these guarantees would be high, and make the product even less attractive than how it is already perceived Paris, 18 February 2013 p. p. 2 2/26

3 Motivation (cont d) Arranging the post-retirement income: Alternative solutions, implying a different sharing of the longevity risk + Self-annuitization longevity risk for the individual Group self-annuitization (GSA) Mortality/longevity linked annuities Traditional life annuities longevity risk for the annuity provider + Paris, 18 February 2013 p. p. 3 3/26

4 Motivation (cont d) We focus on mortality/longevity linked annuities: the annuity benefit is adjusted either to The mortality eperienced by a specific pool of pensioners/annuitants, or The mortality eperienced by the general population, or New projected life tables Literature: Lüty et al. [2001]: adjustment to updated mortality forecasts Richter and Weber [2011]: link to actual mortality eperience van de Ven and Weale [2008]: adjustment so to transfer the aggregate longevity risk Denuit et al. [2011]: adjustment to the unepected longevity improvements eperienced by a given reference population Paris, 18 February 2013 p. p. 4 4/26

5 Motivation (cont d) Our aim: 1. To investigate the features, and check the feasibility, of alternative arrangements in terms of benefit trajectories 2. To assess the value of practicable solutions, in particular from the point of view of the annuity provider 3. To quantify the longevity risk charged to individuals and the annuity provider In this presentation: Step 1 Several possible linking solutions to mortality/longevity are considered, and compared We admit participation to etra-returns on investments Paris, 18 February 2013 p. p. 5 5/26

6 Linking arrangements Longevity risk Idiosyncratic (process or insurance risk): some individuals live longer than others temporary and subject to pooling effects Aggregate: on average, people live longer than epected permanent, no pooling effect General purpose of linking arrangements: sharing losses (or profits) originated by aggregate longevity risk, so that 1. Annual payouts are close to a target amount model calibration based on more direct measures of longevity 2. The portfolio reserve is close to the available assets more suitable for a transfer of the aggregate longevity risk only Smoothing effects, so to avoid to share with the individuals temporary profits/losses (namely, due to the idiosyncratic risk or financial returns) adjustments every k years (say, k = 3,5) & multiperiod (namely, k-years) longevity measures Paris, 18 February 2013 p. p. 6 6/26

7 Linking arrangements (cont d) Basic items in defining the linking rule: 1. Choice of a mortality dataset 2. Choice of the adjustment coefficient of the benefit Eamples of mortality datasets Actual number of surviving annuitants: N +1,N +2,... Actual number of survivors in a reference cohort: L +1,L +2,... Epected number of surviving annuitants, according to (the initial) information F 0 (for eample: F 0 = life table at time 0): E[N +1 F 0 ],E[N +2 F 0 ],... Paris, 18 February 2013 p. p. 7 7/26

8 Linking arrangements (cont d) Epected number of survivors in the reference cohort, according to (the initial) information F 0 : E[L +1 F 0 ],E[L +2 F 0 ],... Epected number of surviving annuitants, according to (the current) updated information F 0,t: E[N +t F 0,t],E[N +t+1 F 0,t],... for eample: F 0,t = {F 0 ;N +1,...,N +t 1 } Epected number of survivors in the reference cohort, according to (the current) new information F t : E[L +t F t ],E[L +t+1 F t ],... for eample: F t = new projected life table at time t Paris, 18 February 2013 p. p. 8 8/26

9 Linking arrangements (cont d) Definition of the adjustment coefficients Various approaches possible. In particular, the definition can be: retrospective: directly involving observed mortality, in terms of N +1,N +2,... (indemnity-based), or L +1,L +2,... (inde-based) prospective: relying on updated mortality forecasts, e.g. E[L +t F t ],E[L +t+1 F t ],... Longevity measures based on: Number of survivors Actuarial quantities, such as portfolio reserve and assets, actuarial value of an annuity Paris, 18 February 2013 p. p. 9 9/26

10 Linking arrangements (cont d) Mortality/longevity adjustment Group Self-Annuitization (GSA)-like: assets portfolio reserve every k years: b t = b t 1 A t V [P] t }{{} individual benefit at time t coefficient of adjustment, γ t If A t V [P] t because of financial returns, or mortality eperience, or issue of a new life table γ t 1 V [P] t + = A t Apart for intermediate times, all the risks are borne by the individuals, as a pool Paris, 18 February 2013 p. p /26

11 Linking arrangements (cont d) Mortality/longevity adjustment based on the number of survivors: life table set at time 0 survival rate in life table F 0 every k years: b t = b t 1 E[N +t F 0 ] N +t } {{ } annuitants, out of initial N m [1] t If N +t E[N +t F 0 ] (i.e.: if t p t p F 0 ) m [1] t 1 = b t 1 t p F 0 t p realized survival rate annuitants bear a part of the aggregate longevity risk (but possibly also some idiosyncratic longevity risk) life table risk with the provider not necessarily V [P] t + = A t, but through the adjustment the difference should be reduced Paris, 18 February 2013 p. p /26

12 Linking arrangements (cont d) Variant: current life table m [2] t = E[N +t F t ] N +t Variant: life table at time 0, updated to the mortality eperienced in the portfolio in the time-interval (0, t) m [3] t = E[N +t F 0,t] N +t Variant 2 (potentially) implies more longevity risk for the individuals, variant 3 possibly less Paris, 18 February 2013 p. p /26

13 Linking arrangements (cont d) The coefficients m [1] t, m [2] t, m [3] t are indemnity-based, as they compare the epected survival rate with the survival rate realized in the pool Variants: inde-based adjustment coefficients m [4] t = E[L +t F 0 ] L +t survivors, out of initial L, in a reference population = b t 1 t p F0 t p survival rate realized in the reference population m [5] t = E[L +t F t ] L +t In principle, less idiosyncratic longevity risk is transferred to individuals Basis risk borne by the annuity provider Paris, 18 February 2013 p. p /26

14 Linking arrangements (cont d) Further variants: Ratios based on the survival rates of different life tables m [6] t = E[N +t F 0 ] E[N +t F t ] = t p F0 tp F t m [7] t = E[N +t F 0 ] E[N +t F 0,t ] m [8] t = E[N +t F 0,t] E[N +t F t ] Idiosyncratic and part of the aggregate risk retained by the provider Paris, 18 February 2013 p. p /26

15 Linking arrangements (cont d) Mortality/longevity adjustment unfunded liabilities -like: assets accumulated at the return credited to policyholders, according to the financial participation arrangement every k years: b t = b t 1 Af t V [P] t }{{} m [9] t Because of financial guarantees, it may result A f t A t However, since V [P] t + = A f t, then V [P] t + should be close to A t The risk transferred to individuals is mainly the aggregate longevity risk Paris, 18 February 2013 p. p /26

16 Linking arrangements (cont d) Mortality/longevity adjustment based on the actuarial value of the life annuity: every k years: b t = b t 1 E[a K +t F 0 ] E[a K+t F t ] } {{ } m [10] t The aggregate longevity risk connected to the possible update of the life table is transferred to individuals Consequences for the provider: individual reserve: if, for eample, m [10] t < 1, then: V t + = b t E[a K+t F t ] < V t = b t 1 E[a K+t F t ] however, the portfolio reserve V [P] t + = N +t V t + retains the uncertainty on the number of survivors thus it may turn out V [P] t + A t Paris, 18 February 2013 p. p /26

17 Linking arrangements (cont d) Matching the mortality/longevity linking to a participation to the etra-return on investments: Every year { } 1+ηt g t b t = b t 1 ma 1+i,1 } {{ } f [1] t interest rate i guaranteed year by year (g t : realized yield; η t : participation rate; ma{η t g t,i }: yield credited to policyholders) Every k years b t = b t 1 f [1] t m [ ] t Variant: f [2] t such that the (annual) interest rate i is guaranteed on average every k years Paris, 18 February 2013 p. p /26

18 Benefit trajectories In all cases: Entry age = 65 Mortality/longevity adjustments every k = 5 years Maimum age for mortality/longevity adjustment (apart from the GSA): 95 (i.e., time 30) One cohort A ω A 0 : remaining assets at cohort s ehaustion, as a percentage of the initial assets (= premiums) Traditional pricing: A 0 = N E[a K F 0 ] Eperienced mortality: 90% of the best-estimate (as at time 0, i.e. F 0 ) Paris, 18 February 2013 p. p /26

19 Benefit trajectories (cont d) Case a: Case b: Case c: Best-estimate life table unchanged No etra-return on investments New projected life table at time 10, yielding a higher life epectancy No etra-return on investments New projected life table at time 10, yielding a higher life epectancy Etra-return on investments: 2%; 1.80% credited to policyholders (every year, through arrangement f [1] t ) Paris, 18 February 2013 p. p /26

20 Benefit trajectories (cont d) Linking based on the number of survivors, indemnity based t m [1] t = t p F 0 t p m [2] t = t p F t t p m [3] t = t p F t p case a case b case c case a case b case c case a case b case c ,t Paris, 18 February 2013 p. p /26

21 Benefit trajectories (cont d) Linking based on survival rates, obtained from different life tables t m [6] t = t p F 0 tp F t m [7] t = t p F 0 tp F 0,t m [8] t = t p F case a case b case c case a case b case c case a case b case c ,t tp F t Paris, 18 February 2013 p. p /26

22 Benefit trajectories (cont d) Linking based on assets and portfolio reserve. Linking based on the actuarial value of the annuity unfunded liabilities GSA actuarial value t m [9] t = Af t γ V [P] t = A t m [10] t V [P] t = E[a +t F τ ] E[a +t F t ] t case a case b case c case a case b case c case a case b case c Paris, 18 February 2013 p. p /26

23 Benefit trajectories (cont d) Total benefit adjustment up to age 95 & final value of assets b 95 b 0 A ω A 0 case a case b case c case a case b case c no long. adj. 100% 100% % 8.55% 8.55% 8.93% m [1] t = t p F 0 m [2] t = t p F t m [3] t = t p F t p 84.53% 98.03% % 2.98% 7.58% 11.53% t p 84.53% 99.27% % 2.98% 9.46% 6.55% 0,t m [6] t = t p F 0 t p 87.06% % % 4.84% 9.53% 6.38% 100% 98.76% % 8.55% 6.69% 13.87% tp F t m [7] t = t p F 0 tp F 0,t t = t p F 0,t m [8] tp F t m [9] t = Af t V [P] t γ t = A t V [P] t 87.06% % % 4.84% 9.53% 6.38% % 97.81% % 12.52% 5.75% 16.32% 72.66% % % 0.49% 2.30% 24.81% 72.66% % % 0.02% 0.18% 0.82% m [10] t = E[a +t F 0 ] E[a +t F t ] % 87.98% % 8.55% 9.47% 56.73% Paris, 18 February 2013 p. p /26

24 Some remarks Main issues: Due to the systematic nature of aggregate longevity risk, it is necessary to sort out the pricing and hedging problems Appropriate policy designs may realize a satisfactory trade-off between risk and price of the longevity guarantees In particular, it is important to match in a convenient way mortality/longevity with financial participation arrangements Net steps: Assess the cost and risk of the alternative designs Solvency vs value created Self-annuitization vs partial longevity guarantees, and timing of the annuitization Paris, 18 February 2013 p. p /26

25 Thank you! Contact: Paris, 18 February 2013 p. p /26

26 References M. Denuit, S. Haberman, and A. Renshaw. Longevity-indeed life annuities. North American Actuarial Journal, 15(1):97 111, 2011 H. Lüty, P. L. Keller, K. Binswangen, and B. Gmür. Adaptive algorithmic annuities. Mitteilungen der Schweizerischen Aktuarvereinigung, 2: , 2001 A. Richter and F. Weber. Mortality-indeed annuities: Managing longevity risk via product design. North American Actuarial Journal, 15 (2): , 2011 J. van de Ven and M. Weale. Risk and mortality-adjusted annuities. National Institute of Economic and Social Research. London. Discussion Paper No. 322, Available online: Paris, 18 February 2013 p. p /26