Lifecycle Portfolio Choice with Tontines

Transcription

1 Lifecycle Portfolio Choice with Tontines Jan-Hendrik Weinert 1, Ralph Rogalla 2, and Irina Gemmo 3 1 Department of Finance, Goethe University Frankfurt House of Finance, International Center for Insurance Regulation (ICIR), Theodor-W.-Adorno-Platz 3, Frankfurt am Main, Germany 2 School of Risk Management, Tobin College of Business, St John s University, 101 Astor Place, New York, NY 10003, United States 3 Department of Finance, Goethe University Frankfurt House of Finance, International Center for Insurance Regulation (ICIR), Theodor-W.-Adorno-Platz 3, Frankfurt am Main, Germany February 28, 2018 Preliminary Extended Draft: Please do not cite or circulate without the author s permission Abstract We derive the optimal life-cycle portfolio choice and consumption pattern for a CRRA gain-loss utility maximizing investor, facing uncertain labor income, risky capital market returns, mortality risk and an age varying level of living standard. In addition to stocks, bonds and deferred annuities, the individuals have access to tontines. Tontines are cost-efficient financial contracts providing age-increasing, but volatile cash flows, generated through the pooling of mortality without guarantees, which can help to match increasing financing needs at old ages. We find that tontines can generate significant welfare gains in individuals portfolios. We expect the stake in tontines to increase in age and a maximum level of tontine investment for moderate levels of total financial wealth for any age. However, higher risk aversion reduces tontine investments. At retirement, deferred annuities account for the majority of total financial wealth, supplemented by a notable level of tontine investment. Keywords: Tontines, Life Insurance, Consumption, Annuities, Mortality, Retirement Planning, Consumption-portfolio choice, Deferred life annuities JEL Classification: D14, E21, G22, I31, J10, L51 Corresponding author

2 1 Introduction Demographic change and the associated shift in the age structure of the population make it increasingly difficult to secure the funding of pay-as-you-go pension systems in many Western societies. Declining birth rates and simultaneously increasing life expectancy worldwide 1 lead to an increase in benefit recipients of the statutory pension insurance with a simultaneous decrease in contributors. The associated increase in the so-called pension ratio makes it more difficult to maintain pay-as-you-go pension systems, such as the public pension system in Germany. In return, funded pension products and private pensions are gaining in importance. This challenge of an aging society is intensified by increasing funding needs in old age 2. The medical advances of the past decades cause that a multitude of diseases and ailments can now be cured, that would have lead to death 50 years ago 3. However, these medical measures and treatment methods are often associated with enormous costs and increase especially in old age, when afflictions pile up. Thus, e.g. a costly, elderly-friendly conversion or expansion of the home will be necessary, which allows the longest possible and independent living in the familiar environment. Furthermore, very high costs for ambulatory and inpatient care are incurred in old age. However, specialized health insurance often depends on the level of care and includes derogations so that soft factors and uninsured aspects are not covered. These include, for example, costly items to maintain the standard of living (e.g., dependence on taxi services because of visual impairment or the use of high quality meal-on-wheels services or shopping delivery services), or the delivery of high-quality nutritional services beyond the statutory level (e.g., massages or home help). This raises the need for a pension product that is suitable to meet the rising capital requirements at retirement age at low cost. These considerations lead to the principle of tontines, which has been adapted to current conditions to meet the requirements of the 21st century. 1 According to the Worldbank (2015), worldwide life expectancy at birth has increased between 1960 and 2015 from 52.5 to above 71.7 years. The increasing lifetime will cause the number of people over 80 years old to almost double to 9 million in Germany by the year 2060 according to forecasts by the Statistisches Bundesamt (2015). In the future, it is therefore very probable that very high ages of 100 years and even more will be achieved by a large number of people. 2 According to the medicalisation thesis motivated by Gruenberg (1977), the additional years that people live due to demographic change are increasingly spent in bad health condition and disability. Jagger et al. (2016) find a significant increase in life expectancy between 1991 and 2011 in England. In those additional years of life, the demand for care products and medical service increases over-proportionately. Coming from 2.6 million nursing cases in Germany in 2013, Kochskämper (2015) estimates between 1.5 and 1.9 million additional nursing cases in Germany in the year 2060 due to demographic change. By the year 2030, the demand for stationary permanent care will increase by 220,000 places in Germany. A study by Standard Life (2013) shows that an 85-year-old person on average has six times higher total spending than a comparable 65-year-old person. 3 For example, the invention of penicillin and other antibiotics in the 1940s and 1950s or the ability of medical treatment of cardiovascular diseases in the 1960s lowered the mortality rates for all age groups. 1

3 The tontine is a financial product developed by its eponymous inventor, Lorenzo de Tonti, in the 1650s for long-term public funding of the French state. In their original form, each tontine owner received a lifelong annual pension in exchange for a one-off payment to the French State. The shares of deceased tontine members are spread among the survivors, increasing their pension payments. Payments to surviving tontinists therefore increase over time. According to this mechanism, the last survivor receives the pension payments from everyone else. A tontine thus offers an age-increasing payout structure without the need of guarantees, which is generated by the diversification of mortality between policyholders. This peculiarity makes tontines appear extremely interesting against the backdrop of an increasing need for capital in old age, since a small amount of funds can generate very high payouts in old age. As a result of these developments, tontines are becoming increasingly important. McKeever (2009), Milevsky (2015) and Li and Rothschild (2017) work up the historical development of tontines, Sabin (2010), Milevsky and Salisbury (2015) and Milevsky and Salisbury (2016) focus on the actuarial fair and optimal payout structure of a tontine and Chen et al. (2017) combine a tontine and an annuity into a common product. In Weinert (2017b) the tontine is supplemented with a cancellation option and its effects on the other tontinists is determined as well as the fair cancellation amount. Weinert (2017a) estimates the cost of a tontine compared to traditional life insurance products from an economic as well as an regulatory perspective. In Weinert and Gründl (2016), tontines are studied for suitability as a complement to traditional retirement products, taking into account the aforementioned demographic challenges. The authors analyze whether a tontine is suitable for serving the increasing financial needs of elderly people. However, that article aims to find the fundamental effects of tontinization and therefore, the model assumes that individuals care about gains and losses in wealth in a simplified framework without capital markets. To analyze this topic with a holistic approach, we include the tontine in a standard life-cycle framework 4 and identify the optimal portfolio structure considering gains and losses in consumption. Due to the over proportional increase in cost with age to maintain a constant level of living standard, we incorporate an age increasing external habit into the model and derive the optimal consumption, saving and portfolio choice pattern for a CRRA gain-loss utility maximizing investor, facing uncertain labor income and risky capital market returns. First results indicate that it is optimal for medium wealthy individuals to increase the stake in tontine investments with age to maximize expected lifetime utility. In Section 2, we introduce the life-cycle model applied to find the optimal consumption and 4 See for example Horneff et al. (2010), Hubener et al. (2013), Maurer et al. (2013) and Horneff et al. (2015). 2

4 portfolio allocation into stocks, bonds, deferred annuities and tontines. We solve the model numerically and present the our expected results in Section 3. We discuss the optimal life-cycle asset allocation for the base case. Further, we vary the calibrations of key parameters and determine the expected life-cycle profiles. 2 The Model We assume that an individual has no bequest motive and has access to capital markets by investing in risk-free bonds, risky stocks, risky tontines and deferred life annuities. Utility: The expected discounted sum of futures utilities is E 0 β t tp u (C t, X t ), (1) t=0 where β > 0 is the subjective discount factor, u ( ) denotes the gain-loss CRRA-utility function, and C t describes the consumption in period t. Throughout our model, we assume an individual with a constant entry age ω. Therefore, we drop the index indicating the entry age for all variables in the following. Hence, we denote an individual s conditional survival probability with tp. 5 X t is the external habit and represents the individual standard of living. The recursive definition of external habits follows Yogo (2008) and is given by X t+1 = X φ t C1 φ t, (2) where φ [0, 1) is the degree of persistence. G t+1 = C t+1 /C t is the consumption growth and D t = C t /X t is the consumption-habit ratio. As shown in Appendix A.2.1, the log consumptionhabit ratio follows an AR(1) process: d t+1 = g t+1 + φd t. (3) Following Yogo (2008), we assume that individuals exhibit gain-loss utility according to u (C, X) = Υ (v (C) v (X)) (4) 5 The full expression for the conditional survival probability of an ω years old at contract inception of surviving t more years would be tp ω. 3

5 and CRRA utility v (C) = C 1 γ. (5) The CRRA gain-loss function according to Tversky and Kahneman (1992) is defined as z 1 θ 1 θ for z 0 Υ (z) = (θ [0, 1), η > 1) (6) η z 1 θ 1 θ for z < 0 with the degree of loss aversion η and the degree of diminishing sensitivity θ. Therefore, u (C t, X t ) = ( C t η C t ) 1 θ X t 1 θ X t 1 θ 1 θ for C t X t for C t < X t (7) with the marginal utility of consumption u C = C γ t ( C t ηc γ t C t ) θ X t for C t > X t X t θ for C t < X t. (8) Stocks and Bonds: The bond return R f is constant over time. The risky stock return in t, R St, is assumed to be serially independent and identically log-normally distributed with expected return µ S and volatility σ S. Annuity: We denote the annuity with. At each point in time before retirement t < K, the individual can purchase life annuities for a non-refundable premium A t. The premium of an annuity, which pays out a constant life-long real payout L K starting at retirement K is A t = L Kξ t, t < K. (9) The annuity factor for an individual aged t + ω results as ξt = (1 + ϕ) K 2 p j j=t R (K 1 t) f Ω K i=1 K 1+i j=k 1 p j R i f, t < K, (10) where ϕ is the loading charged on the annuity premium, p j is the one year survival probability of an ω + j year old, and ω + Ω is the maximum attainable age. However, there is no possibility 4

6 to surrender the annuity contracts. alternative: We denote the annuity with. At each point in time before retirement t < K, the individual can purchase life annuities for a non-refundable premium A t. The premium of an annuity, which pays out a constant life-long real payout L K starting at retirement K is A t = L K K tξ t, t < K. (11) Following Milevsky (2006), the deferred annuity factor for an individual aged t + ω is ( ) bγ ( λ + Rf 1 b, exp { } ) κ+k t K tξt b = (1 + ϕ) { ) exp κ ( λ + Rf 1 exp { } } (12) κ b where ϕ is the loading charged on the annuity premium, Γ (, ) is the incomplete Gamma function, λ captures accidental deaths, b is the dispersion coefficient and κ = ω m, whereas m denotes the modal value of life. Tontine: We denote the tontine with. The individual can purchase immediate tontines for a nonrefundable premium T t at any time. Based on Weinert and Gründl (2016) and Weinert (2017a), the tontine provides an age increasing, normally distributed payout with mean µ t+τ = T t q t+τ R τ f (13) and standard deviation σ t. q t denotes the one year death probability of an ω + t year old and the tontine premium is T t = = = Ω+ ω τ=t Ω+ ω τ=t Ω+ ω τ=t q τ T t R (τ t) f L t,τ [ L t,τ ξ t,τ ] 1 q τ R (τ t) f 1 q τ R (τ t) f (14) with tontine payout L t,τ. The left side of the subscript of L, denotes the timing of investment in the tontine, whereas the right side of the subscript denotes the timing of the tontine payout resulting from this investment. However, there is no possibility to surrender the tontine 5

7 contracts. LTC insurance: In an alternative scenario, we assume that the individual has the opportunity to invest in long-term-care (LTC) contracts, which cover some of the old age expenses. The pricing of the LTC contract follows Levikson and Mizrahi (1994). Labor Income: Labor income (t < K) is determined as exp {h (t)} P t Ψ t for t < K Y t = ζ exp {h (t)} P t for t K. (15) where the permanent component P t depends on its value in the previous period and innovation N t : P t = P t 1 N t. (16) The function h (t) describes the empirically calibrated hump shaped income profile during work life and Ψ t displays a transitory shock. ln (N t ) and ln (Ψ t ) are normally distributed with mean zero and standard deviation σ n and σ ψ. During retirement (t K), the individual receives a constant pension, where ζ is the constant replacement rate. Wealth accumulation: The available wealth W t can be invested in bonds B t, stocks S t, annuities A t (if not retired, t < K), tontines T t and consumption C t. The budget constraint is: B t + S t + A t + T t + C t for t < K W t = B t + S t + T t + C t for t K. (17) B t + S t compounds to the financial wealth. Individual disposable wealth in t + 1 is W t+1 B t R f + S t R St+1 + t τ=0 L τ,t+1 + Y t+1 for t < K B t R f + S t R St+1 + t τ=0 L τ,t+1 + L t+1 + Y t+1 for t K. (18) B t R f + S t R t+1 describes the value of financial wealth in t + 1 and Y t+1 is the labor income (before retirement) or retirement income (from retirement on). t τ=0 L τ,t+1 is the sum of the tontine income of all periods invested in the tontine, whereas L t+1 is the sum of annuity income of all periods invested in the annuity: 6

8 L {t+1} k = L {t} k + A t /ξt for t < K L t+1 = L t for t K (19) A t /ξ t is the additional annuity payment purchased in period t. Short selling is not allowed, thus B t, S t, A t, T t 0. (20) Mortality dynamics: We employ the Gompertz-Makeham law of mortality 6. The objective conditional survival probability of an ω year-old individual of surviving t more years is 7 { ( ) ( λ)} tp = exp λt + 1 e t b b λ. (21) with λ = λ + 1 b e κ b. (22) where λ captures accidental deaths, b is the dispersion coefficient and κ = ω m, whereas m denotes the modal value of life. We maximize Equation (1) with respect to the optimal consumption C t subject to Equation (2), Equation (7), Equations (11)-(20) and Equation (21). 2.1 Calibration Base case: For the base case calibration of the model, we follow Yogo (2008) and set the degree of persistence to φ = 0.68 and the degree of risk aversion to γ = 1. The latter results in a log-utility calibration of Equation (7), which has the advantage of scale invariance of utility and therefore does not depend on the absolute size of the variables. In accordance with Tversky and Kahneman (1992), we set the degree of diminishing sensitivity to θ = 0.12 and the degree of loss aversion to η = Following Weinert and Gründl (2016), we set β = 1, because we assume that the future states are as important as present states for an individual who aims to secure the future standard of 6 See Gompertz (1825) and Makeham (1860). 7 See Appendix??. 7

9 living 8. Following Horneff et al. (2010), we set the entry age to ω = 20 years at t = 0 and assume a maximum attainable age of 105 as in Weinert and Gründl (2016), which corresponds to Ω = 86. We set the retirement age to 65 which corresponds to a retirement time of K = 46. In the base case, we assume no annuity loading and therefore set ϕ = 0 according to Horneff et al. (2010). Furthermore, we assume a high replacement rate of ζ = 0.68 in the base case following Horneff et al. (2010). Following Gourinchas and Parker (2002), σ ψ = 0.15 and σ n = 0.1. Furthermore, the risk free bond return R f = , the expected stock return µ S = 0.06 and the volatility of stock returns σ S = 0.18 as in Horneff et al. (2010). We assume no correlation between stock returns and income shocks. We calibrate the tontine as in Weinert and Gründl (2016) with the calibration of the tontine volatility as shown in Table 2 in Appendix A.1. The complete set of chosen parameters is shown in the Appendix A.1 in Table 1. Table 3 summarizes the endogenous model variables. 3 Preliminary Results In line with prior lifecycle studies, our preliminary results indicate that young individuals with low cash on hand have little incentive to diversify their investment into less-risky assets and invest their disposable wealth exclusively in stocks. With increasing cash on hand, they invest an increasing fraction into risk-free assets. We find the tontine investment to be very low for young individuals, since it realizes small returns at low ages. An increase in age and cash on hand results in the typical life-cycle pattern of decreasing stock investments as well as in an increase of the tontine investment. However, investors with a particularly high level of cash on hand increase their tontine investments less than investors with a medium level of cash on hand. Investors with a very high level of cash on hand are more likely to hold a sizable fraction of their wealth in deferred annuities, as they can afford to wait another 45 years before receiving any payoffs. With depreciating human capital, investors are more likely to purchase bond-like financial assets. Due to an increasing mortality credit, deferred annuities are becoming the most attractive form of risk-free investment, and they crowd out bonds for any given level of cash on hand by the age of 60. For the same reason, the tontine gains attractiveness as the individual ages, resulting in an increasing fraction of investment in tontines. Overall, the availability of tontines can increase the expected lifetime utility compared to a situation without tontines. 8 Parsonage and Neuburger (1992) and Van der Pol and Cairns (2000) provide empirical evidence for a subjective discount rate of zero for the discounting of future health benefits. 8

10 References Chen, A., Hieber, P., and Klein, J. (2017). Tonuity: A novel individual-oriented retirement plan. Working paper, available at SSRN Cocco, J. F., Gomes, F. J., and Maenhout, P. J. (2005). Consumption and portfolio choice over the life cycle. The Review of Financial Studies, 18(2): Gompertz, B. (1825). On the nature of the function expressive of the law of human mortality, and on a new mode of determining the value of life contingencies. Philosophical Transactions of the Royal Society of London, 115: Gourinchas, P.-O. and Parker, J. A. (2002). Consumption over the life cycle. Econometrica, 70(1): Gruenberg, E. M. (1977). The failures of success. The Milbank Memorial Fund Quarterly. Health and Society, 55(1):3 24. Horneff, V., Maurer, R., Mitchell, O. S., and Rogalla, R. (2015). Optimal life cycle portfolio choice with variable annuities offering liquidity and investment downside protection. Insurance: Mathematics and Economics, 63: Horneff, W., Maurer, R., and Rogalla, R. (2010). Dynamic portfolio choice with deferred annuities. Journal of Banking & Finance, 34(11): Hubener, A., Maurer, R., and Rogalla, R. (2013). Optimal portfolio choice with annuities and life insurance for retired couples. Review of Finance, 18(1): Jagger, C., Matthews, F. E., Wohland, P., Fouweather, T., Stephan, B. C., Robinson, L., Arthur, A., Brayne, C., Function, M. R. C. C., Collaboration, A., et al. (2016). A comparison of health expectancies over two decades in England: results of the cognitive function and ageing study i and ii. The Lancet, 387(10020): Kochskämper, Susanna und Pimpertz, J. (2015). Herausforderungen an diepflegeinfrastruktur. IW-Trends Vierteljahresschrift zur empirischen Wirtschaftsforschung, 42(3). Levikson, B. and Mizrahi, G. (1994). Pricing long term care insurance contracts. Insurance: Mathematics and Economics, 14(1):1 18. Li, Y. and Rothschild, C. (2017). Adverse selection and redistribution in the irish tontines of 1773, 1775 and Working Paper. 9

11 Makeham, W. M. (1860). On the law of mortality and the construction of annuity tables. The Assurance Magazine, and Journal of the Institute of Actuaries, 8(6): Maurer, R., Mitchell, O. S., and Rogalla, R. (2010). The effect of uncertain labor income and social security on life-cycle portfolios. Technical report, National Bureau of Economic Research. Maurer, R., Mitchell, O. S., Rogalla, R., and Kartashov, V. (2013). Lifecycle portfolio choice with systematic longevity risk and variable investmentlinked deferred annuities. Journal of Risk and Insurance, 80(3): McKeever, K. (2009). Short history of tontines, a. Fordham J. Corp. & Fin. L., 15:491. Milevsky, M. A. (2006). The calculus of retirement income: Financial models for pension annuities and life insurance. Cambridge University Press. Milevsky, M. A. (2015). King William s tontine: why the retirement annuity of the future should resemble its past. Cambridge University Press. Milevsky, M. A. and Salisbury, T. S. (2015). Optimal retirement income tontines. Insurance: Mathematics and Economics, 64: Milevsky, M. A. and Salisbury, T. S. (2016). Equitable retirement income tontines: Mixing cohorts without discriminating. ASTIN Bulletin: The Journal of the IAA, 46(3): Parsonage, M. and Neuburger, H. (1992). Discounting and health benefits. Health economics, 1(1): Sabin, M. J. (2010). Fair tontine annuity. Social Science Research Network Working Paper Series. Standard Life (2013). The retirement smile. news/new_articles/2013/260613newdrawdowntransferoption.xml. Statistisches Bundesamt (2015). Bevölkerung Deutschlands bis 2060: 13. koordinierte Bevölkerungsvorausberechnung. Publikationen/Thematisch/Bevoelkerung/VorausberechnungBevoelkerung/ BevoelkerungDeutschland2060Presse pdf? blob=publicationfile. Tversky, A. and Kahneman, D. (1992). Advances in prospect theory: Cumulative representation of uncertainty. Journal of Risk and uncertainty, 5(4):

12 Van der Pol, M. M. and Cairns, J. A. (2000). Negative and zero time preference for health. Health Economics, 9(2): Weinert, J.-H. (2017a). Comparing the cost of a tontine with a tontine replicating annuity. ICIR Working Paper Series, (31/2017). Weinert, J.-H. (2017b). The fair surrender value of a tontine. ICIR Working Paper Series, (26/2017). Weinert, J.-H. and Gründl, H. (2016). The modern tontine: An innovative instrument for longevity risk management in an aging society. ICIR Working Paper Series, (22/2016). Worldbank (2015). Databank. Life expectancy at birth. indicator/sp.dyn.le00.in. Yogo, M. (2008). Asset prices under habit formation and reference-dependent preferences. Journal of Business & Economic Statistics, 26(2):

13 A Appendix A.1 Tables Variable Description Value Source β Subjective discount factor 1 Parsonage and Neuburger (1992) Van der Pol and Cairns (2000) Weinert and Gründl (2016) t Time Horneff et al. (2010) ω Entry age 20 Horneff et al. (2010) φ Degree of persistence 0.68 Yogo (2008) θ Degree of diminishing sensitivity 0.12 Tversky and Kahneman (1992) η Degree of loss aversion 2.25 Tversky and Kahneman (1992) γ Degree of risk aversion 1 Yogo (2008) R f Risk free bond return Yogo (2008) µ S Expected stock return 1.06 Horneff et al. (2010) σ S Volatility of stock return 0.18 Horneff et al. (2010) K Retirement time 46 Horneff et al. (2010) ϕ Loading charged on annuity premium 0 Maurer et al. (2010) Horneff et al. (2010) Ω Maximum time of an individual in the 86 Weinert and Gründl (2016) model σ t Standard deviation of tontine returns in t Table 2 Weinert and Gründl (2016) h ( ) Income shape profile Cocco et al. (2005) Horneff et al. (2010) P t Permanent component of income in t Cocco et al. (2005) Horneff et al. (2010) Ψ t Transitory shock in t Horneff et al. (2010) N t Innovation in t Horneff et al. (2010) σ n Standard deviation of ln (N t ) in t 0.1 Gourinchas and Parker (2002) Horneff et al. (2010) σ ψ Standard deviation of ln (Ψ t ) in t 0.15 Gourinchas and Parker (2002) Horneff et al. (2010) ζ Constant replacement rate 0.68 Cocco et al. (2005) Horneff et al. (2010) λ Accidental death 0 Milevsky (2006) b Dispersion coefficient 11.4 Milevsky (2006) m Modal value of life 82.3 Milevsky (2006) Table 1: Calibration base case 12

14 Age σ t Age σ t Age σ t , , Table 2: Simulated standard deviation of tontine returns for every age for a tontine investment of EUR 10,000 13

15 Variable Description tp Conditional survival probability C t Consumption in t X t External habit in t u ( ) Utility G t+1 Consumption growth D t Consumption habit ratio Υ ( ) Gain-loss function v ( ) CRRA-utility u C Marginal utility of consumption R St Risky stock return in t A t Annuity premium in t T t Tontine premium t L K Constant real annuity payout beginning from K ξt Annuity factor for individual aged ω + t p t One year survival probability of an ω + t year old individual µ t Tontine payout mean in t q t One year death probability of an ω + t year old individual L t,τ Tontine payout in τ for tontine premium paid in t ξ t,τ Tontine discount factor in τ for premium paid in t Y t Labor/retirement income in t W t Available wealth B t Bond investment S t Stock investment λ t Objective hazard function of an ω year old of dying at the age of ω + t F (t) Probability of an ω old individual of dying before the age of ω + t f (t) PDF Table 3: Endogenous model variables A.2 Proofs A.2.1 Log Consumption-Habit Ratio ( ) Ct+1 ln (D t+1 ) = ln X t+1 ( = ln C t+1 X φ t C1 φ t ) ( = ln (C t+1 ) ln = ln (C t+1 ) ln X φ t C1 φ t ( X φ t ) ) ln ( C 1 φ t ) = ln (C t+1 ) φ ln (X t ) (1 φ) ln (C t ) = ln (C t+1 ) φ ln (X t ) ln (C t ) + φ ln (C t ) ( ) ( ) Ct+1 Ct = ln + φ ln C t ln (D t+1 ) = ln (G t+1 ) + φ ln (D t ) X t d t+1 = g t+1 + φd t 14