The Context

Investors fear return uncertainty and drawdowns associated with owning relatively risky asset classes, such as equity. The fact that greater risk is associated with greater expected return does not preclude the possibility that realized returns may be far less than a low-risk asset could provide, even with horizons as long as 5 to 10 years. Fear prompts the average investor to sometimes act against his own best interest. Therefore, the average investor’s portfolio often underperforms a static benchmark, even before fees. The average investor tends to increase allocation to riskier assets after the market has already significantly risen and decrease allocation after a significant decline.

Given that financial planning in the context of retirement is a multidecade endeavor that can last 50 years or more, investors face an additional challenge. Financial planners cannot be concerned solely with “managing” the behavior of investors facing short-term return uncertainty, which remains an important challenge, but must also be concerned with how to address longer-term uncertainties. For example, there is significant uncertainty about the assumption for long-term expected returns, real as well as nominal. Some lucky investors may accumulate much of their wealth before retirement during an exceptionally strong bull market (e.g., 1982 to 2000), whereas less fortunate investors may have been planning to retire in 2008 or early 2009, just after the most dramatic global liquidity crisis since the Great Depression and a significant decline in interest rates. Moreover, average expected returns represent only part of the story. Some investors may have recorded above-average returns at a time when their accumulated wealth was already significant, whereas others may have recorded above-average returns when their accumulated wealth was low. The implication of the interactions among savings patterns (in accumulation), withdrawal patterns (in decumulation), and timing of above- and below-average portfolio returns is important to understand, particularly with respect to how the interactions should affect the allocation policy.

Furthermore, investors face significant risks in the transition period—say, the last 5 to 10 years—from accumulation to decumulation. Considering the uncertainty in long-term expected returns, patterns of returns, and patterns of savings, it is unlikely that a financial plan established when an individual is 30 years old can remain static thereafter. The plan must be revaluated periodically, which requires an effective feedback mechanism or tool. The investor has many more options available before retirement, however, although some may not necessarily be pleasant. In the event of lower than expected accumulated wealth, the investor could decide to save more, postpone retirement, and/or adjust retirement plans. These options may be unavailable or may be harder to implement once the decision to retire has been made. The investor must implement a transition strategy that reduces the likelihood and/or significance of the unplanned adjustments that may be necessary as the targeted or desired retirement date draws near.

In addition, longevity remains uncertain, and retirement plans often are based on expected longevity. Even though the median life expectancy for a 65-year-old individual in the United States is 83.3 years for men and 85.9 years for women, a significant percentage of people will live past age 90. Furthermore, as someone ages, the older she is expected to live. A dynamic issue, life expectancy becomes even more complicated in the context of a couple—one or both individuals could live a very long time. This possibility should also influence the allocation policy over time and the potential need for longevity insurance.

Finally, governments in many countries have implemented policies and programs to support the retirement effort. The most important program in the United States is Social Security, but its purpose is to provide only a minimum level of inflation-adjusted income, not to sustain the standard of living that an individual had before retirement. Other programs seek to encourage savings and facilitate wealth accumulation, such as 401(k)s, traditional and Roth IRAs (Registered Retirement Savings Plans and Tax-Free Savings Accounts in Canada), and health savings accounts (HSAs). Most investors, however, underestimate the savings effort required to maintain the standard of living to which they are accustomed, cannot implement a comprehensive and coherent adaptive retirement plan, cannot optimize across all relevant parameters, and lack access to the feedback mechanism needed to make the appropriate adjustments over time.

Secure Retirement and Other Literature

Many books on retirement planning have been published in recent years. Almost exclusively, they cover general issues of retirement preparedness, public policy, asset allocation and asset location principles, and lifestyle recommendations. These books provide useful guidelines and simple investment rules of thumb. Recommendations are often based on relatively simple analytics, sometimes illustrated in a single-period context, and/or are supported by referring to more in-depth studies published in academic journals or business research. For example, CFA Institute published a comprehensive literature review titled Longevity Risk and Retirement Income Planning,[1] which aggregates much of the relevant literature.

Currently, however, no book links the academic and business research on the most relevant dimensions of retirement planning—such as risk in accumulation and decumulation, longevity risk, longevity products, asset allocation, taxation, and measures of utility—and tests this research within an integrated and realistic empirical framework. Furthermore, some aspects of retirement planning have been insufficiently researched. For example, although how to better manage risk in decumulation is currently a focus of interest, the nature of risk in accumulation and how it should affect allocation policy are still not well understood by most investors.

Improving retirement planning is therefore a complex challenge that requires a comprehensive and integrated theoretical and empirical effort. As Robert Merton said in a 2017 interview, “The retirement problem is a global problem. The good news is, finance science can be used to solve it. Design things on finance principles, rather than institutionally…. If you design on financial principles, it will work everywhere in the world.”[2] Designing a solution to the retirement problem based on financial principles is the goal of Secure Retirement. Although written in the context of US (and, at times, Canadian) investors, its principles are universal. Secure Retirement is written not only for sophisticated investors and financial advisers but also for those interested in creating more informationally efficient web-based retirement platforms. Its findings can be applied to all investors, although the investments of individuals with super-high net worth require greater complexity with respect to assets and asset classes, insurance products, taxation, and legal framework.

The Content

The material of Secure Retirement is concentrated in seven core chapters, Chapters 2 to 8. The introductory chapter (Chapter 1) frames the main issues and the scope of this book. The concluding chapter (Chapter 9) explores improvements that should eventually be integrated into this effort but would require new research and highly advanced quantitative methodologies, such as machine-learning models and algorithms.

Secure Retirement relies on a building-block approach. Most arguments are supported in the following way:

• A logical argument and its expected consequence are presented.
• The argument is supported using scenario analyses.
• The argument is further supported in the context of a simulation environment.

Chapter 2 explains the relevant dimensions of retirement planning in the absence of uncertainty. Chapters 3, 4, and 5 increase our understanding of risk in the context of accumulation (Chapter 3), decumulation (Chapter 4), and the transition between these periods (Chapter 5). They also discuss what allocation strategies and tools are appropriate in each context, incorporating the effects of Social Security and annuities. Chapter 6 discusses the different methodological processes that can be used to apply and calibrate a comprehensive retirement approach that considers all three phases of the retirement process. It also presents the different measures of utility and satisfaction that may be appropriate in different contexts, such as when investors have excess wealth and intend to use this wealth for a specific purpose. One such measure is the percentage of PIO (preferred income objective) target, which measures the average percentage of the preferred inflation-adjusted retirement income target expected to be achieved across all scenarios and over time. This measure uses adjusted mortality assumptions to account for the concerns of retirees who may live much longer than the median mortality age. Secure Retirement emphasizes sustained retirement income rather than wealth accumulation.

As of Chapter 6, the retirement framework is designed around a simple context characterized by few asset classes and no taxation. Chapter 7 adds broader considerations related to asset allocation and choice of asset classes, taxation and asset location, life insurance, the relevance of variable annuity products, the appropriate income replacement ratio, the role of reverse mortgages, the choice of mortality assumptions, and the complexity of a household.

Finally, Chapter 8 presents a fairly comprehensive model of retirement planning and applies it to our prototype investor, named John. An integrated simulation engine was specifically designed for this chapter. Chapter 8 illustrates how successive improvements made to our framework change John’s expected retirement income distribution and the expected utility derived from consumption (such as his percentage of PIO target). The model is applied when John is 30 years old and then again when he is only 5 years from retirement. In a real-life context, the model should likely be reapplied periodically, perhaps yearly, especially if circumstances (income, health, financial conditions, and so on) change dramatically.

Secure Retirement is not about active management of specific asset classes. Although we have conducted research on the effect of using factor-investing approaches to modify the distribution of equity returns (but not the expected return), the book’s emphasis remains on risk management and long-term financial planning.

Some Major Observations

Secure Retirement covers too many aspects to properly address them all here. Some important observations can be made, however. Consider the following five aspects.

First, the effect of return volatility on cumulative wealth in the context of periodic savings contributions during the accumulation phase differs greatly from its effect in the context of static wealth (implying no new savings contribution). Volatility can enhance the expectation of cumulative wealth and improve its expected distribution when the current level of wealth is small relative to the present value of future savings contributions. This dynamic can lead to higher portfolio returns even in a low market return environment. Therefore, without increasing expected downside risk, volatility provides support for very high levels of equity exposure for a period that can extend up to 15 years before retirement.

Second, we can use our understanding of risk in the context of accumulation to derive more-effective allocation glide paths between higher- and lower-risk asset classes. In addition, our analyses show that a dynamic allocation strategy—such as that proposed by Giron, Martellini, Milhau, Mulvey, and Suri (2018), inspired by the constant proportion portfolio insurance (CPPI) approach—can significantly reduce yearly drawdowns without unfavorably affecting the distribution of expected retirement income.[3] Furthermore, we show that the portfolio turnover resulting from the implementation of their approach, an issue ignored in their original research, can be significantly reduced.

Third, our research does not support the frequent argument that purchasing single-premium annuities in a low interest rate environment is inadvisable. We found that incorporating an annuity component allows us to maintain a higher equity allocation in the remaining liquid asset portfolio. We do not advise substituting annuities for equity or even for a balanced portfolio, but they may be substituted for a portion of the fixed-income component. This approach considerably alleviates the argument against annuities in a low interest rate environment even though their expected duration is longer than that of most fixed-income portfolios. In fact, an annuity strategy should be implemented gradually because a market crisis (leading simultaneously to lower yields and low equity prices) occurring in the final years leading to retirement can significantly affect this strategy’s efficiency.

Fourth, we examined the question of asset location—that is, which assets should be placed in which type of accounts according to their respective tax status. Should higher expected return assets, taxed at a lower rate than other assets, be allocated to the taxable account in priority or to the tax-deferred/tax-exempt accounts? Those advisers who recommend allocating higher expected return assets first to tax-deferred/tax-exempt accounts support their argument through the return-compounding benefits of higher expected return assets. Those who recommend allocating these assets to the taxable account argue that although taxation lowers net expected returns, it also reduces risk; in addition, the tax rate applied to interest income is higher than the rate applied to equity (dividends and capital gains). Our research shows the appropriate answer is more subtle than either recommendation but generally supports allocating riskier assets first to tax-deferred/tax-exempt accounts at long investment horizons.

Finally, overall simulation results obtained in a comprehensive setting show that coherent and disciplined allocation and risk management processes can significantly enhance investors’ well-being. Consider the case of our prototype investor, John. What started as a percentage of PIO target of 64.9% for the bottom half of all scenarios generated when John was 30 years old—the top half of all scenarios scored close to 100%—ended with a score of 90.1% when John reached the age of 60 and all efficiency improvements were implemented. Furthermore, the worst expected yearly drawdowns over the plan’s lifetime were reduced by 30%.

Our work shows that a well-designed financial planning tool is essential to support the work of investors and financial planners. The challenge is simply too complex to be handled by simple rules of thumb, current generic robo-advisers, or simple Excel spreadsheets. Eventually, implementing machine-learning algorithms to improve longevity assumptions and asset allocation decisions will further improve the results.