Research and Reality: A Literature Review on Drawing Down Retirement Financial Savings

Abstract

How do, could, and should retirees draw down their financial savings? This article reviews over 100 papers on this topic from the perspective of individuals, families, governments, and financial institutions. Three significant conceptual/methodological weaknesses in the existing literature are identified: (1) analysts have examined a limited range of self-managed drawdown strategies; (2) nearly all have ignored home ownership, pensions, debt, and government taxes and transfers when quantitatively evaluating alternative drawdown strategies; and (3) there is a well-acknowledged gap between the behavior implied by economic models and that of real-life individuals, particularly when it comes to voluntary annuitization. Expanding the set of drawdown strategies evaluated (e.g., including larger payouts when life expectancy is reduced after the onset of a significant health condition, or using savings as bridge income to delay the take-up of Social Security payments), refining the income concept used, and more exact modeling of the trade-offs underlying individual decision-making will likely increase the appeal of self-managed drawdown strategies and help resolve the “annuity puzzle” that has long dominated this line of research. It may also lead to advice and financial products that will better meet the needs of retirees.

Acknowledgments

This project began in response to a call for proposals by the Society of Actuaries’ Pension Section, and we gratefully acknowledge their generous financial support. The various drafts of this article were reviewed and feedback provided by a Project Oversight Group; we wish to thank each member: Steve Siegel, Kevin Binder, Eric Keener, Carol Bogosian, Ian Genno, Gavin Benjamin, Andrew Peterson, Cindy Levering, Dylan Porter, Eric Keener, Judith Chan, Michael Archer, and Tom Lowman. Special thanks to Chen He for his assistance in tracking down the cited papers, as well as Lars Osberg, Sven Sinclair, and an anonymous reviewer for their valuable feedback. B.-J. M. gratefully acknowledges the financial support of the Natural Sciences and Engineering Research Council of Canada and the Social Sciences and Humanities Research Council of Canada.

Notes

See Brown (2008) for a summary of studies reporting the mortality improvements and declines in labor force participation at older ages for OECD countries.

This is aside from “Guaranteed Withdrawal Minimum Benefit Products” (see the Appendix), which are limited in Canada.

Literature on the topic of retirement savings drawdown generally only considers nominal annuities (that is, a single-premium immediate nominal annuity). The likely explanation for this concentration is that the majority of the literature is by American researchers, and most privately available payout annuity products in the U.S. market are fixed in nominal dollars (Brown 2009a). Consequently the term “annuity” in this article refers to this type unless otherwise stated.

This term is also known as a “phased withdrawal strategy,” “self-insure,” or, when the payments replicate those from a life annuity, “self-annuitization.” This article uses the term “self-managed drawdown strategy.”

For empirical evidence of the low rates of voluntary annuitization, see Milevsky and Young (2007) and Brown (2009a) for the U.S. experience, and James and Song (2001) for an international perspective.

See the Appendix for a list and description of the major types of annuity products.

The section 3 “How could” reviews the disadvantages of annuitization that could explain this aversion.

For instance, if wealth W is invested in a bond with a rate of return of R, then it will grow to W(1+R) after one period. If the same wealth were instead used to purchase an actuarially fair life annuity, it would grow to W(1+R)/(1−q) if the consumer survives (where q is the probability of death in that one-year period) (Brown et al. 2008).

Ameriks et al. (2001) showed that a life annuity could allow a risk-averse individual to invest remaining, nonannuitized, wealth more aggressively since the annuity's guaranteed income stream would substitute for the conservative assets in the portfolio, such as cash or bonds. See the section 3 “How Could” under “Hybrid Strategies” for more discussion.

The vast majority of past published studies have evaluated alternative drawdown strategies by measuring the pre-tax income obtainable from retirement savings alone; see the section 4 “How Should” under “Discussion on Research Methodologies.”

Risk aversion is the degree to which an individual is not willing to take on financial risk (volatile income stream in this case) in return for a potentially greater return.

This explanation could be considered rational or behavioral, depending on whether the individual is justified in believing that he or she could, in fact, achieve a higher return than that offered by the annuity.

Source: U.S. Census Bureau, Survey of Income and Program Participation, 2008 Panel, Wave 10. For definition of “Net Worth”, see http://www.census.gov/people/wealth/about/faq.html.

Based on the 2005 Survey of Financial Security weighted household data. We calculate net financial assets as the sum of registered financial assets and nonregistered financial assets (checking accounts, GICs, trusts, etc.), less nonmortgage debt (credit card, lines of credit, car loans, etc.).

Source: U.S. Census Bureau, Survey of Income and Program Participation, 2008 Panel, Wave 10.

These are the mortality improvements that affect the entire population and therefore cannot be diversified by insuring a larger group of annuitants.

Malcolm Hamilton proposed this explanation via personal correspondence in May 2011.

Malcolm Hamilton, May 2011.

The study was published in Warner and Pletter (2001).

Using stochastic simulation, Doyle and Piggott (2003) found that, for countries with a mandatory DC state pension system, the cost of the government's minimum pension guarantee was considerably higher if the citizens followed a self-managed drawdown strategy than had they purchased either a fixed or variable annuity.

Note that these drawdown strategies are those examined in this line of literature. The section 4 “How Should” proposes others (under “Discussion on Research Methodologies”).

In an attempt to simplify previous findings regarding these strategies, we do not include the numerous assumptions underlying each study's analysis.

Strategies no. 1 (ii) and no. 3 (ii) are slight modifications of a strategy described in Avery and Morrison (2009), which is given in no. 4 (ii).

A table summarizing previously published studies that have examined the optimal time to annuitize (including the researchers’ assumptions and specifications) was given in Blake et al. (2003) (see Tables 3 and 4 therein). This table was then updated and extended in Horneff et al. (2008) (see Table 1 therein).

As explained in the section 3 “How Could” under “Annuitization,” systematic longevity risk is the risk of mortality improvements that affect the entire population and therefore cannot be diversified by insuring a larger group of annuitants.

Although the term “consumption” is used in this line of research, the amount actually modeled is generally pre-tax income generated by retirement savings alone.

The meaning of “additive separability” is that the utility of one period is not affected by a change in the utility of another period. Davidoff et al. (2005, p. 1585) relaxed this feature of the standard utility function by assuming that individuals may exhibit an “internal habit,” meaning “it is not the level of present consumption, but rather the level relative to past consumption that matters for utility.”

Some examples include Milevsky and Robinson (2000), Albrecht and Maurer (2002), Young (2004), Milevsky et al. (2006), Bayraktar and Young (2009), and Wang and Young (2010).

Proposed by Joseph Tomlinson via personal correspondence in October 2012.

When the specified period is lifetime, this measure then becomes the probability of lifetime ruin.

Lars Osberg suggested this drawdown strategy through personal correspondence in May 2013.

See bullet 2k “Insufficient personal savings” in the section 3 “How Could” under “Annuitization.”

Ameriks et al. (2001) and Young (2004) both recommended conservative investments for retirees with shorter time frames, and more aggressive investments for longer time frames.

For instance, to explain the absence of taxes and government provisions, the common assumption has been that “benefits are taxed as ordinary income; therefore taxes will not change the desirability of voluntary annuitization or systematic withdrawal from a self-managed retirement account” (Dus et al. 2004, p. 6).

Other research has similarly found that annuitization is less desirable in the presence of means-tested government transfers (Butler et al. 2011) and a minimum consumption floor (Pashchenko 2013).

For example, Sinclair and Smetters (2004), Ameriks et al. (2011), Pang and Warshawsky (2010), and Peijnenburg et al. (2011a).

They found that the expected present discounted value of the benefits for a single-premium life annuity was approximately 20% lower than the expected present discounted value of the premiums for someone from the general population.

For instance, the Globe and Mail recently commented that the Canadian federal government's abandonment of a CPP expansion, which was allegedly due to provincial preferences, was not publicly favored according to polls (Curry 2011).

See Shapiro (2010) for an extensive review of existing financial products for retirees, as well as those envisioned for the future.

Regarding employer DB plan benefits, Sondergeld and Greenwald (2007, p. 16) advised retirees that “(w)hen retiring from a job with a DB pension plan, seriously consider the consequences of taking a lump sum (if offered) instead of a lifetime income from the pension plan.”

For instance, while OAS benefits are subject to income tax, GIS benefits are not. In addition, registered wealth (RRSPs and the RRIFs that replace them for older retirees) and other wealth (Tax-Free Savings Accounts) are taxed differently at payout since the former has tax-deductible contributions while the latter does not. In the United States, a similar distinction exists for Roth IRAs and “standard” IRAs.

In fact, empirical evidence suggests quite the opposite, that spending drops significantly after retirement. This phenomenon is referred to as the “retirement consumption puzzle.” For studies that examine the possible causes behind this trend, see Brzozowski and Lu (2005) in Canada and Aguiar and Hurst (2005) and Hurd and Rohwedder (2005) in the United States.

Horneff et al. (2008) did not, however, examine all of the strategies that we have listed in the section 3 “How Could” under “Self-Managed Drawdown Strategies.”

Proposed by Joseph Tomlinson via personal correspondence in October 2012.

In Canada, for example, informal caregivers were responsible for 70% of the hours of support provided to elders with a long-term health problem (Lafrenière et al. 2003).

Information on the vulnerability of seniors to fraud and scams can be found in the SOA report by Iannicola and Parker (2010), a report by the Canadian Task Force on Financial Literacy (Government of Canada 2010) and Fishman (2010).

AIR is “the benchmark investment return that must be earned by the underlying portfolio before payments can actually increase” (Milevsky 2002, p. 25). If the AIR is i and the experience rate of return is r, then the payment will increase by the ratio (1+r)/(1+i).

Dus et al. (2004) found that by choosing to self-manage their retirement savings, young retirees reduced their expected shortfall risk and improved their expected payouts.

See Broadbent et al. (2006).