Products and Strategies for the Decumulation of Wealth during Retirement: Insights from the Literature

Abstract

The question how individuals should (optimally) annuitize their wealth remains of high relevance in light of longevity risk and volatile capital markets. In this article, we first present traditional and innovative products and strategies for the decumulation of wealth during retirement, based on a review of 72 selected academic articles in peer-reviewed journals. We further identify relevant factors that generally influence the conception of these products from the retirees’ perspectives, and derive implications for product developers, before concluding with avenues of future research. Our results indicate that innovative suggestions often comprise tontine-like structures, exploit actuarial and accounting smoothing in various ways, defer annuitization to higher ages, or combine it with long-term care options, for instance. Key areas of future research in this field include the consideration of both insurer and retiree perspectives in the analysis of products, using behavioral considerations when evaluating the retirees’ perspective, and taking into account the impact of costs or expenses. While recent articles increasingly consider these aspects, manifold opportunities for future research remain.

ACKNOWLEDGMENTS

The authors thank two anonymous reviewers for valuable comments on an earlier version of the article.

Discussions on this article can be submitted until January 1, 2024. The authors reserve the right to reply to any discussion. Please see the Instructions for Authors found online at http://www.tandfonline.com/uaaj for submission instructions.

Notes

1 There is also a field of research dedicated to the general question of behavioral perception and biases in this context, as well as the annuity puzzle, e.g., based on surveys or experiments, such as Agnew et al. (2008) or Beshears et al. (2014). We refer to Alexandrova and Gatzert (2019) for a systematization of the related literature on annuitization decisions, as we focus on the different products and strategies for the decumulation phase.

2 Hanewald et al. (2013), Horneff et al. (2008).

3 Stamos (2008, 65).

4 Bravo and de Freitas (2018, 212–13).

5 Idiosyncratic longevity risk can be characterized as “the risk that is reduced most by pooling” as “these risks are independent across individuals in the pool” and can be measured by “the variability in the average survival time in the pool.” Bernhardt and Donnelly (2021, 102) define it as “the risk that the distribution of future lifetime is not observed perfectly among the annuitants.” Systematic longevity risk is arising “from dependent risk factors that impact the survival probabilities of all individuals in the pool to a greater or lesser extent” (Qiao and Sherris 2013, 960) and as “the risk that the wrong distribution of future lifetime has been chosen” in Bernhardt and Donnelly (2021, 102), for instance. Qiao and Sherris (2013, 951) further point out that “Dependence between individual lives created by common drivers of mortality improvements has the potential to undermine pooling effectiveness, not only because of declining payments, but also because of systematic longevity risk that cannot be reduced with increases in pool size.” We further refer to Piggott et al. (2005) for additional explanations.

6 See MacDonald et al. (2013, 36) and Hu and Scott (2007, 77ff.) on such “period-certain” annuities.

7 Stamos (2008, 65), Hanewald et al. (2013).

8 Bravo and de Freitas (2018, 212–13), Horneff et al. (2008, 397).

9 Notional defined or non-financial defined contributions (NDC) are similar to classical pay-as-you-go public pension schemes; however, the benefit payments are regularly adjusted based on several factors (economic growth, life expectancy, etc.), and hence they do not necessarily guarantee a minimum pension level. For further studies, we refer to Holzmann et al. (2012).

10 Hanewald et al. (2013).

11 MacDonald et al. (2013, 7ff.).

12 Stamos (2008), Horneff et al. (2008).

13 Hu and Scott (2007, 77): A period-certain annuity thus combines a series of zero-coupon bonds for the first period and a “longevity” annuity afterward, whereby individuals are willing to buy the combined product rather than the individual components.

14 Bravo and de Freitas (2018).

15 Hanewald et al. (2013), Pitacco (2016).

16 Proposed by Milevsky (2005) and further discussed in Pitacco (2016).

17 Milevsky (2005), Gong and Webb (2010).

18 Pitacco (2016).

19 The “income drawdown option” allows the retiree “to choose when to convert the final capital into pension” (Gerrard et al. 2004, 321).

20 Pitacco (2016), Horneff et al. (2008).

21 Introduced by Huang et al. (2009) and further considered in Pitacco (2016).

22 “For example, in the UK, accumulated pension assets have to be mandatorily annuitized by age 75. Germany’s ‘Riester’ plans provide a tax inducement if life annuity payments begin to pay out at age 85 (withdrawn amounts must either be constant or rising, prior to annuitization). In the USA, of course, annuitization is not compulsory for 401(k) plans; as a result, most retirees roll them over to an Individual Retirement Account and manage the funds themselves, subject to the tax laws requiring minimum distributions to begin at age 70 1/2” (Horneff et al., 2008, 398).

23 Horneff et al. (2008).

24 Gong and Webb (2010, 210–11).

25 Pitacco (2016).

26 Gong and Webb (2010, 211): “We calculate that a household planning to smooth consumption through its retirement would need to allocate only 15% of its age 60 wealth to an ALDA with payments commencing at age 85, holding the remainder of its wealth in unannuitized form to finance consumption from age 60 to 85.”

27 Di Giacinto et al. (2014).

28 Horneff et al. (2008, 396).

29 Di Giacinto and Vigna (2012, 498) and also Gerrard et al. (2004), who state that investors with “not too high risk aversions” will prefer income drawdown option over immediate annuitization (with sufficiently good risk-reward assets).

30 Gong and Webb (2010).

31 Gong and Webb (2010), Horneff et al. (2008).

32 Pitacco (2016).

33 ifa (2021, 49ff.).

34 Bravo and de Freitas (2018), Jørgensen and Linnemann (2012), Maurer et al. (2013; 2016), Pitacco (2016).

35 See, e.g., Maurer et al. (2013).

36 See Bravo and de Freitas (2018), Bohnert et al. (2015), Maurer et al. (2013). The participation scheme in the surpluses may depend on country-specific regulations; e.g., in Italy, the investment return in excess of the guaranteed rate is paid out entirely, while in other countries only partial payouts are required (Bravo and de Freitas 2018).

37 Bohnert et al. (2015), Bravo and de Freitas (2018), Maurer et al. (2013), Pitacco (2016).

38 Albrecht and Maurer (2002), Alexandrova et al. (2017), Bravo and de Freitas (2018), Guillén et al. (2006), Maurer et al. (2013; 2016).

39 Alexandrova et al. (2017), Maurer et al. (2016).

40 Maurer et al. (2013), Guillén et al. (2006), Bohnert et al. (2015).

41 Jørgensen and Linnemann (2012, 139).

42 Bohnert et al. (2015), Maurer et al. (2013).

43 Pitacco (2016, 93).

44 See Pitacco (2014), Gatzert et al. (2012), Gatzert and Klotzki (2016).

45 Pitacco (2014, 45), Gatzert et al. (2012).

46 Pitacco (2014), Gatzert and Klotzki (2016).

47 Pitacco (2014, 44ff.).

48 Pitacco (2016, 90–91), Brown and Warshawsky (2013), Chen et al. (2021a), and the references therein.

49 Hieber and Lucas (2022).

50 Pla-Porcel et al. (2016).

51 Gatzert and Klotzki (2016, 59).

52 See Eling and Ghavibazoo (2019), who provide a review of the LTC literature.

53 Pitacco (2014, 45ff.).

54 Chen et al. (2021a).

55 Gatzert et al. (2012), Gatzert and Klotzki (2016).

56 In this context, Brown and Warshawski (2013) use empirical Health and Retirement Study (HRS) data to identify favorable poolings of risk groups that increase the number of eligible individual that are currently rejected for LTC insurance due to underwriting constraints.

57 Certain underlying theories, i.e., compression, equilibrium, and pandemic theory, address relations between total life expectancy and disability-free life expectancy (see Pitacco 2014, 142ff. and references therein).

58 Pitacco (2016, 90).

59 Gatzert and Klotzki (2016), Brown and Warshawsky (2013).

60 See Eling and Ghavibazoo (2019), Gatzert et al. (2012), Pitacco (2016, 92).

61 Gatzert and Klotzki (2016, 59).

62 Denuit et al. (2011), Pitacco (2016), Bravo and de Freitas (2018).

63 Denuit et al. (2011).

64 Bravo and de Freitas (2018).

65 Bravo and de Freitas (2018), Pitacco (2016).

66 Bravo and de Freitas (2018), Pitacco (2016).

67 Barrieu et al. (2009, 21–26, 32). See also Denuit et al. (2011) and the references therein for discussions on the securitization of longevity risk.

68 Milevsky and Salisbury (2015, 92), Weinert and Gründl (2021, 53).

69 E.g., Chen and Rach (2019), who propose tontine payoffs with minimum guaranteed annuity payments, or Donnelly and Young (2017), who propose a product that includes a tontine’s basic mortality-sharing mechanism but smoothes mortality effects via a transfer to an insurer.

70 Milevsky and Salisbury (2015, 92).

71 Bernhardt and Donnelly (2019), Valdez et al. (2006).

72 Bravo and de Freitas (2018).

73 Milevsky and Salisbury (2015), Weinert and Gründl (2021).

74 Weinert and Gründl (2021).

75 Bernhardt and Donnelly (2019, 172ff.).

76 Bernhardt and Donnelly (2019, 172ff.).

77 Bernhardt and Donnelly (2021).

78 Milevsky and Salisbury (2015, 92) also explain pooled annuity funds as lying “between a tontine and an annuity.”

79 A comparison between pooled annuity overlay fund by Donnelly et al. (2014) and the equitable income tontine by Milevsky and Salisbury (2016) along with an in-depth description of the products is provided in Bräutigam et al. (2017).

80 Starting with a tontine structure and then switching to an annuity (“tonuity”)—see Chen et al. (2019b)—which may not only be favorable for retirees but also for the insurer to reduce solvency capital requirements, or vice versa as a (less favorable) “antine” (Chen et al. 2020).

81 Chen et al. (2020).

82 Bravo and de Freitas (2018).

83 Bernhardt and Donnelly (2019).

84 Chen et al. (2020, 120) further show that when combining a tontine with an annuity, the annuity component becomes more important for retirees with a higher risk aversion, and that an increasing pool size leads to a stronger emphasis on the tontine component.

85 Milevsky and Salisbury (2015, 92).

86 Qiao and Sherris (2013, 951), Chen and Rach (2021; 2022).

87 Milevsky and Salisbury (2015, 92), Weinert and Gründl (2021, 51).

88 Valdez et al. (2006); an in-depth analysis of a fixed up-front and time-varying proportional fee in tontines for service providers (e.g., an insurer) is conducted in Chen et al. (2019b).

89 Expenses do arise from setting up the fund, advertising it, investment management fees, and administration tasks, including the notification and verification of deaths of pool participants as well as the resulting redistribution of funds among the surviving pool members (Donnelly et al. 2013, 68).

90 E.g., in case of the product by Fullmer and Sabin (2018); see also Weinert and Gründl (2021, 54).

91 Bernhardt and Donnelly (2019), Piggott et al. (2005), Valdez et al. (2006).

92 See Chen et al. (2021b).

93 Bernhardt and Donnelly (2019).

94 Milevsky and Salisbury (2015, 99).

95 Chen et al. (2020, 98).

96 Chen et al. (2021b).

97 MCPT represents a suggested extension of CPT that considers annual changes in contract values instead of only the terminal wealth distribution (see Ruß and Schelling 2018, 1105ff.).

98 Weinert and Gründl (2021, 52, 81)

99 Weinert and Gründl (2021).

100 Chen et al. (2021c).

101 Boon et al. (2020, 421ff.).

102 See, e.g., Bräutigam et al. (2017, 415–419) for such a comparison.

103 See Donnelly et al. (2014) for an introduction, and Donnelly (2015) for a more comprehensive study.

104 Milevsky and Salisbury (2016).

105 Which is why Fullmer and Sabin (2018) suggest “individual tontine accounts” as an alternative.

106 Valdez et al. (2006).

107 Milevsky and Salisbury (2016, 596).

108 Qiao and Sherris (2013, 972) propose external reinsurance to pool members to ensure a narrower band for future payments (which implies a guarantee, as opposed to the already-mentioned case that refers to Donnelly et al. 2013).

109 van Bilsen and Linders (2019), Gatzert and Schmeiser (2013).

110 ifa (2021), Gatzert and Schmeiser (2013).

111 See, e.g., Jørgensen and Linnemann (2012), Guillén et al. (2006), and Jørgensen (2007), who explore the TimePension mechanism in detail.

112 Van Bilsen and Linders (2019).

113 Boado-Penas et al. (2020).

114 Huang et al. (2014), Balter and Werker (2020).

115 Gatzert and Schmeiser (2013).

116 Van Bilsen and Linders (2019).

117 Jørgensen and Linnemann (2012, 139).

118 Gatzert and Schmeiser (2013).

119 Gatzert and Schmeiser (2013), Boado-Penas et al. (2020).

120 Jørgensen and Linnemann (2012, 140).

121 See ifa (2021, 22ff.) for a review of the German innovative decumulation product landscape, Bohnert and Gatzert (2014) for a formal and numerical analysis of dynamic hybrid products applied to the accumulation phase, and Gatzert and Schmeiser (2013) for a discussion.

122 Linnemann et al. (2015, 74).

123 E.g., as a concave or convex function of age, see Linnemann et al. (2015, 75).

124 E.g., MacDonald et al. (2013), Horneff et al. (2008), Piggott et al. (2005).

125 Albrecht and Maurer (2002), Hanewald et al. (2013).

126 Hanewald et al. (2013, 92ff.), Albrecht and Maurer (2002).

127 Waring and Siegel (2015).

128 Schmeiser and Post (2005, 240–42, 254). The authors also point out that the costs for asymmetric information with respect to the retiree’s health as well as other transaction costs can be saved within the family.

129 Di Giacinto et al. (2014), Schmeiser and Post (2005).

130 Horneff et al. (2008).

131 See Waring and Siegel (2015) and Forsyth et al. (2020, 443), respectively.

132 E.g., MacDonald et al. (2013), Pitacco (2016), Bravo and de Freitas (2018), Albrecht and Maurer (2002), Forsyth et al. (2020), Horneff et al. (2008).

133 See, e.g., Milevsky and Salisbury (2015), Qiao and Sherris (2013), Bravo and de Freitas (2018), Valdez et al. (2006).

134 See Bernhardt and Donnelly (2021). Moreover, Donnelly (2015) and Milevsky and Salisbury (2016) discuss different ways of “equalizing” payoffs within (heterogeneous) pools of annuitants by means of actuarial fairness or equitability.

135 E.g., Chen et al. (2020, 120), Weinert and Gründl (2021).

136 E.g., Donnelly and Young (2017), Chen et al. (2019b), Chen and Rach (2019).

137 Bernhardt and Donnelly (2019).

138 E.g., Alexandrova and Gatzert (2019) and MacDonald et al. (2013) on rational and behavioral influencing factors regarding annuitization decisions, and Eling and Ghavibazoo (2019) on drivers for demand and supply of LTC products.

139 Horneff et al. (2008).

140 If they have no access to small pools, see Stamos (2008). Stamos (2008, 56) also points out that pooled annuity funds are generally very effective in insuring against longevity risk, even for small pool sizes.

141 E.g., Chen et al. (2020, 2021b).

142 Horneff et al. (2008), Gerrard et al. (2004) in the context of DC pension plans.

143 Horneff et al. (2008).

144 Bernhardt and Donnelly (2019), Hanewald et al. (2013), Konicz and Mulvey (2015), Dagpunar (2021), Chen and Rach (2022), Schmeiser and Post (2005), and for a review on annuitization decisions in this regard, see Alexandrova and Gatzert (2019).

145 Hanewald et al. (2013).

146 If both products have a death benefit component, annuities can exhibit a higher expected lifetime utility than tontines, where (safety) loadings play an important role as well; see Chen and Rach (2022). The authors show that adding death benefits can be beneficial for both policyholder (CRRA) and insurer (in terms of solvency capital requirements).

147 Konicz and Mulvey (2015).

148 E.g., Weinert and Gründl (2021) in the context of a portfolio of standard annuities and tontines.

149 Agnew et al. (2008), Beshears et al. (2014).

150 Brown et al. (2017).

151 Chen et al. (2021c).

152 E.g., Fullmer and Sabin (2018), Milevsky and Salisbury (2015; 2016), Piggott et al. (2005), Valdez et al. (2006).

153 Hieber and Lucas (2022).

154 E.g., Weinert and Gründl (2021), Chen and Rach (2019), Chen et al. (2019b), Chen et al. (2020).

155 Weinert and Gründl (2021), Chen et al. (2020, 2021b).

156 Donnelly et al. (2013), Weinert and Gründl (2021).

157 Qiao and Sherris (2013), Donnelly (2015), Milevsky and Salisbury (2016), Bräutigam et al. (2017).

158 Qiao and Sherris (2013).

159 Qiao and Sherris (2013).

160 Donnelly et al. (2014) and Donnelly (2015), with the latter especially focusing on the actuarial fairness of GSAs.

161 See Milevsky and Salisbury (2016) for their conception of this product, and Bräutigam et al. (2017) for a respective comparison to similar tontine-like approaches (pooled annuity overlay fund).

162 As “income drawdown option,” see, e.g., Gong and Webb (2010), Horneff et al. (2008), Gerrard et al. (2004), Di Giacinto and Vigna (2012).

163 Di Giacinto and Vigna (2012) find that the costs of suboptimality lie between 6 and 40% in terms of loss of expected present value of consumption from retirement to death, depending on risk aversion. See also Di Giacinto et al. (2014) for further optimality analyses in this regard.

164 Milevsky (2005), Gong and Webb (2010), Horneff et al. (2008).

165 Gong and Webb (2010, 211): “We calculate that a household planning to smooth consumption through its retirement would need to allocate only 15% of its age 60 wealth to an ALDA with payments commencing at age 85, holding the remainder of its wealth in unannuitized form to finance consumption from age 60 to 85.”

166 Gong and Webb (2010, 211): “The intuition is simply that the household is suffering much less actuarial unfairness, but getting almost as much longevity insurance. An ALDA also dominates an optimal decumulation of unannuitized wealth.”

167 Gong and Webb (2010, 211). Horneff et al. (2008, 396) show that such “integrated strategies can enhance retirees’ well-being by 25%–50% for low/moderate levels of risk aversion when compared to full annuitization at retirement.”

168 Gong and Webb (2010, 211).

169 Gong and Webb (2010, 211).

170 Horneff et al. (2008, 396).

171 Di Giacinto and Vigna (2012).

172 E.g., Huang et al. (2014).

173 See, e.g., Hu and Scott (2007), as well as Chen et al. (2019a), who further show that loss aversion is the most relevant factor when deciding about buying immediate annuities, and the probability transformation is most important when deciding about deferred annuities.

174 E.g., Eling and Ghavibazoo (2019), Chen et al. (2021a), Pla-Porcel et al. (2016), and the references therein.

175 Eling and Ghavibazoo (2019), Brown and Warshawsky (2013).

176 Surpluses achieved through a steady accrual of investment, mortality and expense profits can be significant in case of lock-in characteristics (see Bohnert et al., 2015, 93) and provide an alternative solution with regard to higher medical expenditures, without directly recurring to, e.g., LTC or other combined products (see Pitacco 2016, 89ff.).

177 Maurer et al. (2013; 2016), Alexandrova et al. (2017).

178 See Jørgensen and Linnemann (2012) and Linnemann et al. (2015) on TimePension, started in 2002 and introduced 2009 in the Danish market according to the latter.

179 See the product proposed and studied by Boado-Penas et al. (2020), whereby the insurer manages the underlying assets (without offering choices to the retiree) where a “corridor volatility smoothing method” is used. The smoothing scheme is described as follows: “The smoothing mechanism in the retirement phase […] consists of interaction between two layers: the collective fund and an additional safety layer. This allows to increase the pensions if the underlying funds go up and prevents, to some extent, decreasing the pensions if the funds go down. In the extreme case that the actual market situation becomes not manageable for the insurance company and the safety layer is empty, the pensions would need to be reduced” (Boado-Penas et al. 2020, 7).

180 See van Bilsen and Linders (2019), where liabilities are hedged through the investment strategy, but without including guarantees or longevity risk.

181 Jørgensen and Linnemann (2012), Linnemann et al. (2015).

182 van Bilsen and Linders (2019).

183 See, e.g., Hanewald et al. (2013) or (for fees in tontines vs. annuities) and Chen et al. (2021b). Petrichev and Thorp (2008, 1144) further calculate that “delaying retirement by five years reduces required wealth by 5% or less. On the other hand, 25% more wealth is needed to maintain the relativity of the pension with average earnings, as compared with consumer prices, and around 6% more wealth is needed for each 1% increase in investment management fees.”

184 ifa (2021, 22ff.), Bohnert and Gatzert (2014).

185 Bravo and de Freitas (2018) discuss ways to hedge longevity risk along with advantages and disadvantages; see also Barrieu et al. (2009) and Boon et al. (2020).

186 Denuit et al. (2011), Bravo and de Freitas (2018).

187 In different ways; examples include Bravo and de Freitas (2018), Chen and Rach (2019; 2022), Chen et al. (2021a; 2021b), Denuit et al. (2011), Gatzert and Klotzki (2016), Guillén et al. (2006), Huang et al. (2009), Maurer et al. (2016).

188 See, e.g., Forman and Sabin (2015) for a discussion of such practical aspects in case of tontine pensions.

189 E.g., Weinert and Gründl (2021), who use MCPT to study optimal fractions of tontines and annuities, Hu and Scott (2007) and Chen et al. (2019a) using CPT, or Agnew et al. (2008), Beshears et al. (2014), and Brown et al. (2017), who use experiments/surveys. Other articles consider the retiree’s perspective by, e.g., simulating account balances (Guillén et al., 2006) or annual average pension benefits (Brown and Warshawsky, 2013; Jørgensen and Linnemann, 2012), which is equally of importance for policyholders.

190 See Alexandrova and Gatzert (2019) for a respective review as well as, e.g., Agnew et al. (2008), Beshears et al. (2014), or Brown et al. (2017) for such behavioral studies.

191 E.g., Hu and Scott (2007), Chen et al. (2019a), Bär et al. (2021).

192 See, e.g., a current working paper in this regard on a CPT-based comparison of selected retirement products by Chen and Rach (2021).

193 E.g., Forman and Sabin (2015), Chen et al. (2021b) on fees in tontines and annuities, and Chen et al. (2020) (in terms of “safety loadings”).

194 Donnelly et al. (2014).

195 Exceptions include, e.g., Bär et al. (2021), who explicitly include income from social security in their analysis of optimal retirement saving decisions using threshold-based utility functions, and observe a significant impact, while Weinert and Gründl (2021) and Chen et al. (2020) consider portfolios of tontines and annuities.

196 Schmeiser and Post (2005).

197 Gatzert and Klotzki (2016), Hieber and Lucas (2022), Pitacco (2016).

198 See, e.g., Brown and Warshawsky (2013) for an actuarial study of prices and expected benefits based on empirically estimated transition probabilities, Chen et al. (2021a) for a simulation study of long-term care options from the perspective of the insurer and the policyholder based on CPT, Pla-Porcel et al. (2016) for a proposal of embedding (public) long-term care annuities in retirement pension systems, and Levantesi and Menzietti (2012) on solvency capital requirements for longevity and disability risk in life annuities with long-term care under parameter risk.

199 A review of research specifically on long-term care insurance along with a comprehensive presentation of future research opportunities in this regard can be found in Eling and Ghavibazoo (2019).

200 E.g., Horneff et al. (2008; 2020), Konicz and Mulvey (2015), Schmeiser and Post (2005).

201 See Regulation (EU) 1286/2014 on key information documents for packaged retail and insurance-based investment products (PRIIPs).