RETIREMENT INSIGHTS. Breaking the 4% rule. Dynamically adapting asset allocation and withdrawal rates to make the most of retirement assets

Transcription

1 RETIREMENT INSIGHTS Breaking the 4% rule Dynamically adapting asset allocation and withdrawal rates to make the most of retirement assets

2

3 FOREWORD Over the past decade, retirees have been forced to navigate the dual investment challenges of extremely low interest rates and elevated market volatility. Many have relied on the popular 4% rule to draw down their portfolio assets, but this approach developed in the very different market climate of the 1990s has increasingly been called into question in terms of providing a truly sustainable retirement income stream. As a result, the rapidly growing number of investors preparing to enter retirement may wish to reconsider their withdrawal options. Our research suggests investors and their financial advisors should look beyond the static rules of the past when seeking to achieve stronger results from their retirement income withdrawal strategies. A portfolio-based solution using a more robust withdrawal rate framework may help investors better address their retirement funding needs by embedding market risk, longevity risk and evolving personal circumstances in a way that a cash-flowbased approach simply cannot. As we have evaluated various withdrawal strategies and have taken into account new criteria, we have found that: Maximizing expected lifetime utility (i.e., potential derived satisfaction) serves as a more effective benchmark of retirement withdrawal success than typical measures, such as probability of failure. Focusing on utility offers a way to quantify how much satisfaction retirees receive from their portfolio withdrawals. This can help potentially increase investors level of income when they are most apt to enjoy their retirement dollars, while still avoiding the risk of premature portfolio depletion. A dynamic approach to managing withdrawals and asset allocations provides a more effective use of retirement assets than traditional approaches. Adapting to changes in market conditions and investors specific situations over time can help maximize the expected lifetime utility generated by retirement assets. This type of dynamic strategy may help provide greater payout consistency and reduce the likelihood of either running out of money or accumulating excess wealth that is unlikely to be used by the investor. Age, lifetime income and wealth all provide key insights into how to adjust investors withdrawal strategies throughout retirement. Holding all other factors constant, higher initial wealth levels suggest individuals lower their withdrawal rates, while also increasing their fixed income allocations. Greater lifetime income, through Social Security, pensions and/or lifetime annuities, allows individuals to increase both their withdrawal rates and equity allocations. Increasing age allows individuals to increase their withdrawal rates, while also decreasing their equity exposure. J.P. MORGAN ASSET MANAGEMENT 1

4 Based on this analysis, a dynamic withdrawal model may offer substantial advantages to help investors make the most of their assets throughout their retirement years. For more information about the J.P. Morgan Dynamic Retirement Income Withdrawal Strategy or any of the other topics covered in this paper, please contact your J.P. Morgan Asset Management representative. Sincerely, Abdullah Z. Sheikh, FSA Research & Analytics Asset Management Solutions Global Multi-Asset Group S. Katherine Roy, CFP Chief Retirement Strategist J.P. Morgan Asset Management Anne Lester Senior Portfolio Manager Asset Management Solutions Global Multi-Asset Group 2 BREAKING THE 4% RULE

5 TABLE OF CONTENTS 5 Overview 7 Comparing different retirement income withdrawal strategies 9 Developing a dynamic decumulation model 14 Dynamic retirement income withdrawal applications 25 Conclusion 26 Appendix J.P. MORGAN ASSET MANAGEMENT 3

6 4 BREAKING THE 4% RULE

7 OVERVIEW The numbers are staggering. Over the past 20 years, retirement planning has largely been about how to maximize asset accumulation efforts to ensure as many Americans as possible are ready to retire with sufficient savings. Today, with 76 million baby boomers transitioning to retirement and an average 10,000 people projected to reach age 65 per day over the next 20 years, 1 retirement income withdrawal strategies have now also begun to receive much greater attention, as more investors re-examine the best way to tap into their retirement nest eggs. In an ideal world, one could simply match all future expenses in retirement with the proceeds from a portfolio of assets derived from pre-retirement savings, thereby safeguarding adequate income and removing the risk of funding shortfalls. This is difficult to achieve in practice, however, for two main reasons: 1) Future expenses are unknown Since no one knows how long they will live, it is not possible to forecast the exact costs associated with living, traveling and receiving health care in retirement, an issue further complicated by how variable these expenses can be over time. 2) Future assets are unknown A portfolio of non-cash assets, such as equities or bonds, will fluctuate depending on economic and market environments, making future account values unknown at the outset. The unpredictable nature of these two critical inputs highlights some of the weaknesses in conventional approaches to retirement income planning. Indeed, there is mounting evidence that the static withdrawal rules of thumb that may have worked well enough in the past likely do not offer the most efficient usage of retirement assets. The 4% rule in particular has faced increased scrutiny, prompted by the current prolonged low interest rate environment and the negative impact fixed withdrawals had on shrinking retirement account balances in the wake of the 2008 financial crisis. Looking ahead, we believe investors may want to consider a dynamic approach to retirement income withdrawals. Based on our research, periodically adjusting withdrawal rates and portfolio asset allocations in response to changes in personal wealth, age and lifetime income 2 may offer greater probability of retirement income funding success, not just in terms of the amount of dollars received but also with retirees overall satisfaction with their withdrawals. 1 Pew Research Data; December The term lifetime income refers to any income that is guaranteed and will last for life, such as Social Security, pensions and/or lifetime annuities. J.P. MORGAN ASSET MANAGEMENT 5

8 In this paper, we examine: The pros and cons of different retirement income withdrawal strategies How a dynamic withdrawal model can help enhance investors retirement experience Case studies that illustrate real-world portfolio applications and potential advantages of a dynamic withdrawal framework Our goal with this analysis is to help investors develop more robust solutions to their post-retirement income and asset allocation needs. Working with their financial advisors, retirees can apply this framework and incorporate their unique circumstances and risk profiles into a dynamic withdrawal strategy specifically designed to secure their income needs and better weather the constantly evolving nature of the financial markets. 6 BREAKING THE 4% RULE

9 Comparing different retirement income withdrawal strategies Evaluating various retirement income withdrawal approaches starts with defining the objectives of a successful postretirement decumulation strategy. Most would agree that the primary focus of a prudent withdrawal approach is to maintain a careful balance between managing lifestyle risk and longevity risk, two critical, if at times conflicting, goals. The key is to generate and withdraw enough from retirement savings to finance expenses at a level that maintains a sustainable post-retirement living standard, while avoiding the uncertainty of running out of money. It is also important to remember, however, that there is an emotional value to retirement income, as well as a monetary one. As such, we recommend adding a third goal to this list: maximizing how much utility value investors receive from their withdrawals. This offers a more holistic perspective around the withdrawal planning process to help capture the full potential of retirement assets. For example, the risk of running short of money is easy to grasp, but pulling out too little is also problematic, in our opinion. Setting aside specific bequests and/or legacy aspirations, an excess level of unused wealth at death may only represent assets that could have been utilized to enhance the richness of an individual s retirement experience, perhaps significantly. Hence, we believe the aim of a withdrawal strategy should be to exhaust retirement assets in the most efficient manner possible, while mitigating the risks of premature portfolio depletion. In order to incorporate the concept of utility value into the withdrawal process, we focus on a metric known as expected lifetime utility, which quantifies the collective perceived satisfaction of all withdrawals received in retirement. While this measure can arguably be somewhat subjective, it draws from well-established microeconomic principles that offer important insights into the emotional aspects of investing. A robust withdrawal framework also needs to take into account a wide variety of personalized factors that may influence the optimal income strategy. Specific individual investor characteristics, such as age, wealth level, level of lifetime income (e.g., Social Security, pensions or lifetime annuities), life expectancy and risk preferences provide key inputs into determining an appropriate asset allocation strategy and the resulting withdrawal rate that can be realistically supported. In addition, adequate decumulation sustainability requires accounting for a broad range of market scenarios and planning for as many future (and historical) trajectories of the financial markets as possible, including severe downside scenarios such as those experienced in the financial crises of All of these considerations provide useful context to compare different retirement income withdrawal strategies. Consider the two most common approaches: The 4% rule dictates that individuals withdraw 4% of their initial portfolio value in the first year of retirement and annually increase that amount by the inflation rate of the preceding year to maintain purchasing power (alternatively, inflation increases may be based on historical rates or longterm averages). This approach focuses on withdrawals only and does not recommend a particular asset allocation. Dollar amounts are determined strictly by portfolio value, without any additional consideration to individual retiree characteristics around wealth, age or lifetime income. The required minimum distribution (RMD) approach is based on the amounts that the U.S. government requires retirees to withdraw from traditional Individual Retirement Accounts and employer-sponsored retirement plans beginning at age 70½. Annual withdrawals are determined using the following equation: Withdrawal amount = Portfolio value / Remaining life expectancy The RMD approach also does not recommend a particular asset allocation, nor does it account for retiree wealth or lifetime income, though it does consider age. J.P. MORGAN ASSET MANAGEMENT 7

10 EXHIBIT 1: OVERVIEW OF VARIOUS RETIREMENT INCOME WITHDRAWAL APPROACHES Strategy Primary benefit 4% rule Strives to preserve purchasing power Customized to individual Accounts for market uncertainty Sets withdrawal rate Sets asset allocation No No Yes No RMD Strives to avoid depleting portfolio assets prematurely Age only No Yes No J.P. Morgan Dynamic Strategy Strives to maximize expected lifetime utility Age, wealth, lifetime income, risk profile Yes Yes Yes Source: J.P. Morgan Asset Management. For illustrative purposes only. As shown in Exhibit 1, these common approaches have pros and cons. The 4% rule is simple to understand and easy to apply, but it has come under fire in the past several years given that it does not hold up particularly well against longevity risk in extremely volatile markets, especially when a portfolio loses significant value during the early years of retirement. The RMD methodology is better at ensuring that retirees do not deplete assets early, since it incorporates life expectancy into its basic formula. However, it can be unstable in terms of minimizing lifestyle risk, since it directly translates portfolio volatility into income volatility. For example, a 15% asset decline automatically translates into a 15% RMD spending cut, potentially difficult to manage on a fixed budget. Given these shortcomings, J.P. Morgan began to explore how investors might better satisfy their retirement income needs. It seemed intuitive that a more dynamic approach to withdrawal rate and asset allocation modeling could provide the flexibility and customization necessary to minimize both lifestyle risk and longevity risk, while extracting more potential from a portfolio through a wider range of market cycles. The result of this detailed research is a withdrawal framework that is much more adaptive to the dynamic aspects of postretirement decision making and the uncertain nature of PLACING UTILITY VALUE ON RETIREMENT ASSETS The fields of microeconomics and behavioral finance offer practical insights into how investors tend to view and respond to decisions about money and investing. Applying these principles to a withdrawal strategy may help investors enjoy richer and more fulfilled retirements by acknowledging the emotional elements behind the retirement income planning process. financial markets, including the risks of extreme events such as the 2008 financial crisis. The J.P. Morgan Dynamic Retirement Income Withdrawal Strategy offers withdrawal rate and asset allocation suggestions, both of which are proactively tailored to retirees evolving circumstances and portfolio experiences on an ongoing basis. For a given age, level of wealth and level of lifetime income, our framework dynamically determines an appropriate asset allocation and corresponding withdrawal amount for the upcoming year. The resulting detailed output is based on sophisticated portfolio modeling that has largely been out of reach for most investors. (For a more detailed description of this process, please refer to the Appendix.) This highly personalized approach is also unique in the sense that it has been specifically built around the concept of maximizing expected lifetime utility. Incorporating individual characteristics and preferences as key inputs offers a disciplined yet dynamic withdrawal approach to help retirees extract greater financial and personal value from their portfolios. This process seeks to help compensate for the magnitude of income shortfalls and extra spending that may transpire throughout retirement, while reducing the risks of both outliving assets and excess wealth accumulation. The result is a retirement income planning approach designed to help achieve broader investment success across the entire retirement horizon. 8 BREAKING THE 4% RULE

11 Developing a dynamic decumulation model To develop a cohesive withdrawal framework, we identified five key factors to address in our decumulation model. Each of these variables can have significant impact on optimizing a post-retirement withdrawal solution. Following are brief descriptions of each factor with the rationale for its inclusion. Factor 1: Individual preferences for magnitude and timing of withdrawals In order to evaluate the satisfaction, or utility, a retiree receives from a given withdrawal strategy, we rely on a utility-based methodology to help measure and evaluate the emotional value of each dollar paid out. We apply a utility function that can model retiree preferences for both magnitude and timing of any income received to help quantify this core input. Withdrawal magnitude: The utility function around withdrawal amounts is concave (or curving in, as shown in Exhibit 2) with respect to increases in retirement income. In other words, retirees get less satisfaction from each additional dollar of income withdrawn above a certain point. This is consistent with human behavior, whereby the first dollar of income received carries more marginal satisfaction than each additional dollar. Withdrawal timing: The utility function around when a retiree receives money includes a discount factor that applies to each future withdrawal. This discount factor increases as withdrawals take place in the future, as illustrated in Exhibit 3. As a result, there is a time preference for withdrawals made earlier in retirement. This is also consistent with human behavior, whereby income received earlier (e.g., today) is more attractive than income received in the future (e.g., tomorrow). It is worth noting that this time preference for earlier income is strictly emotional in nature and distinct from the time value of money, which instead accounts for the monetary discounting of dollars (i.e., dollars received in the future compared to their present purchasing power). Factor 2: Levels of wealth and lifetime income It seems logical that different levels of wealth and lifetime income, such as Social Security, pensions and life annuities, should generate different suggested asset allocations and withdrawal rates in retirement. For example, retirees with higher initial wealth levels have a greater ability to withstand negative shocks, such as unanticipated medical, living or travel expenses due to inflation or personal emergencies. Similarly, individuals and couples with significant lifetime income have a lower risk of poorer outcomes later in retirement than retirees with less lifetime income, since this secured income baseline ensures at EXHIBIT 2: MARGINAL UTILITY DECREASES FOR EACH ADDITIONAL DOLLAR OF INCOME Utility $0 $10,000 $20,000 $30,000 $40,000 $50,000 Spending Source: J.P. Morgan Asset Management and Essays in the Theory of Risk Bearing, Markham Publ. Co., Chicago, EXHIBIT 3: TIME PREFERENCE FOR EARLIER WITHDRAWALS, BASED ON UTILITY OF $100 SPENT AT VARIOUS AGES Present utility Age Source: J.P. Morgan Asset Management. For illustrative purposes only. J.P. MORGAN ASSET MANAGEMENT 9

12 least a minimum standard of living even in scenarios of high longevity and/or poor financial market returns. Consequently, our framework recognizes that downside risk is greater for those with less initial wealth and lower or no lifetime income, relative to those with greater wealth and higher levels of lifetime income. Factor 3: Current age and life expectancy Life expectancy at any given age measures the number of years, on average, retirees can expect to live based on the mortality experience of the U.S. population. Our model uses standard Actuarial Life Tables from the U.S. Social Security Administration. 3 Data is available for males and females, providing the opportunity to derive the probability of survival for couples as well. 4 To illustrate the concept of life expectancy, Exhibit 4 shows the conditional probability of survival for males, females and couples to various ages, assuming a current age of 65. In the case of couples, there are two probabilities: at least one spouse surviving to a given age and both spouses surviving to a given age. As would be expected, a retiree s probability of survival decreases as age increases. EXHIBIT 4: SURVIVAL PROBABILITY FOR 65-YEAR-OLD MALES, FEMALES AND COUPLES TO DIFFERENT AGES Probability (%) Males Females Couples (one member) Couples (both members) the survival weighted utility of $1 projected to be withdrawn at age 120. This is because there is a 97% chance that at least one spouse will survive to age 75 but close to 0% odds that at least one spouse will still be alive at age 120. Survival weighting utility of withdrawals reduces the attractiveness of strategies that defer income withdrawal to the end of an individuals or couple s lives, as the utility value of a dollar spent later is significantly less than that of a dollar spent at a younger age, simply because there are greater odds the later dollar will never actually be used. In some cases, mortality also affects lifetime income. For example, in the case of Social Security or joint and survivor annuities, benefits are often reduced on the death of a spouse. 5 Our framework accounts for this reduction in lifetime income. In such cases, the lower amount is somewhat offset by a higher utility derived from a dollar of spending for an individual versus a couple. That is, in order to compensate for the corresponding lower expenses associated with one individual versus a couple, we attribute a higher utility for the same dollar amount of spending by an individual. 6 Factor 4: Market randomness and extreme events lnvestors and financial advisors typically employ a scenariobased approach when considering market uncertainty. A scenario-based approach evaluates how different withdrawal strategies might fare in a variety of economic and financial market environments. This approach recognizes that a solution that meets income goals across a wider range of possible scenarios is more robust and preferable to one that meets retirees needs in only a small, specific set of market conditions Age Source: J.P. Morgan Asset Management and Social Security Administration. For illustrative purposes only. Our framework weighs the utility of withdrawals at a given age by the probability of survival to that age to arrive at a survival weighted utility of withdrawal. Thus, each period s withdrawal utility is discounted by the probability of being able to actually obtain it. For example, a 65-year-old couple s survival weighted utility for $1 projected to be withdrawn at age 75 is higher than 3 For further details, please see 4 To calculate the probability of at least one spouse surviving to a given age, use the formula P(A or B) = P(A) + P(B) P(A and B). For ease of computation, we assume A and B are independent so that P(A and B) = P(A) x P(B). 5 Our model assumes a one-third reduction in lifetime income upon the death of a spouse, an approximation based on the current Social Security formula rules applied to a couple with one working spouse. The exact reduction in Social Security benefits will likely vary depending on the unique circumstances of the couple. 6 To maintain parity in utility functions, we ensure the utility of $0.66 (twothirds of $1) spent by the surviving spouse is equal to the utility of $1 spent by a couple. For further details, please refer to the exact specification of the utility function in the Appendix. 10 BREAKING THE 4% RULE

13 To capture the future uncertainty associated with the financial markets, our model applies a forward-looking simulation that generates 10,000 random equity and bond returns. 7 We use J.P. Morgan 2014 Long-Term Capital Market Assumptions for the purposes of calibrating expected performance, as shown in Exhibit 5. The firm offers market assumptions that cover a wide array of asset classes. Any number could be included in the model output, but we focused on two asset classes with very different risk/reward profiles to help simplify our analysis, while still effectively illustrating how retirees might adjust risk exposure. EXHIBIT 5: J.P. MORGAN 2014 LONG-TERM CAPITAL MARKET ASSUMPTIONS U.S. large cap stocks U.S. aggregate bonds Inflation Expected annual compound return 7.50% 4.25% 2.25% Expected volatility 14.75% 6.50% 1.50% In order to account for the downside risk of extreme market events, such as the 2008 financial crisis, our model applies J.P. Morgan s proprietary Non-Normal Framework. 8 The key to this framework is recognizing that many traditional financial models are based on three incorrect assumptions around what is generally assumed to be market normality: Traditional assumption #1: Returns are independent from period to period. In reality, some asset class returns exhibit marked serial correlation, frequently influenced by prior periods. Traditional assumption #2: Asset returns are normally distributed. In reality, asset returns are not normally distributed, as can be empirically observed in a broad range of past scenarios. Traditional assumption #3: Asset class relationships are linear. In reality, asset class correlations show significant breakdown in times of market stress. Our Non-Normal Framework strives to correct these shortcomings by using advanced statistical methods designed to capture the long-term downside risk associated with market anomalies. This offers an analysis that can model empirically observed non-normality, taking into account more than just one-standard-deviation events and better reflecting their observed frequency. This revised set of assumptions and statistical techniques is designed to help gain a better picture of downside risk and provide a more realistic framework for identifying and testing potential investment solutions. Factor 5: Dynamic nature of the decision-making process Evolution of an individual s wealth in retirement will depend on actual realized expenses and the performance of portfolio assets. As discussed earlier, although retirement expenses and portfolio performance can be estimated, they cannot be forecasted with any amount of detailed precision. For example, it is possible to broadly estimate living, travel and health care expenses, but each of these also closely depend on unknown factors such as inflation, travel frequency and destinations and general retiree well-being and related health care needs. Similarly, portfolio performance will depend on future economic and financial market performance, which is completely unknowable in the present. Luckily, individuals in retirement are not obliged to adopt a static withdrawal approach but can, in fact, adapt their withdrawal rates and asset allocations to changes in the market climate and their personal financial situations. If one year s experience is better than expected (e.g., actual expenses are lower and/or portfolio performance is better than anticipated), an individual could conceivably increase the following year s withdrawal slightly to enjoy the surplus or decrease portfolio equity risk and increase fixed income allocations to lock in the gains for income later in retirement. Conversely, if one year s experience is worse than expected, an individual could decrease the withdrawal the next year or increase portfolio equity exposure to target higher expected returns and potentially higher income in the future to make up for the shortfall. Our framework easily allows for this dynamic nature of postretirement withdrawal planning. Retirees can change both their asset allocations and withdrawal rates at the end of each year, in response to their actual experience over the course of that year. We believe this improves the efficiency of our post-retirement income solution, compared with more static approaches. 7 We believe 10,000 simulations should reduce simulation error sufficiently to draw robust inferences from the results. 8 For full details about this model, please see J.P. Morgan Asset Management s Non-Normality of Market Returns: A Framework for Asset Allocation Decision Making (2009). J.P. MORGAN ASSET MANAGEMENT 11

14 Putting it all together: Framework methodology Our dynamic model effectively combines these five key factors into a single cohesive framework. To achieve this, we apply concepts from the field of dynamic programming, a body of work that specifically deals with finding optimal solutions to multiperiod problems with several decision variables at each stage. This process arrives at customized asset allocation and withdrawal rate considerations through the complex, integrated analysis of the retiree s specific investment profile and an incredibly broad spectrum of possible market scenarios, as illustrated in Exhibit 6. Critical inputs include: Functional form and parameter values of the utility function used to describe retiree preferences Life expectancy for individuals and couples from 60 onwards Initial levels of wealth at retirement and lifetime income levels 10,000 simulations accounting for non-normal financial market returns for stocks, bonds and inflation Our model then calculates optimal asset allocation and withdrawal rate solutions at each age with the goal of maximizing expected lifetime utility. This is achieved through a backward induction process, beginning at age 120 and working toward age The end output of this process is a set of tables that detail an optimized asset allocation and withdrawal strategy that changes over time, seeking to maximize the investor s expected lifetime utility of retirement assets. In our opinion, this metric offers a more meaningful, holistic success measure than typically applied benchmarks, such as probability of failure, which focuses solely on the risk of running out of money too soon. Expected lifetime utility: A better quantifier of success Measuring a withdrawal strategy s effectiveness based on how much satisfaction a retiree receives from his or her portfolio assets represents a significant paradigm shift. This changes the focus to evaluating the maximum potential utility value that can be received each year, while also accounting for the probability that a retiree will live long enough to actually obtain it. In contrast, probability of failure calculates, for a given asset allocation and withdrawal rate, the likelihood of early depletion of retirement assets over a fixed or variable time horizon, usually life expectancy. Although incorporating life expectancy is an improvement from a fixed time horizon calculation, a probability of failure metric still suffers from two major drawbacks: First, it does not incorporate utility theory. Probability of failure metrics do not allow for individual preferences for either the magnitude or timing of income. Behavioral studies have shown that individuals receive less satisfaction from each additional dollar of income and tend to exhibit a strong time preference for current over deferred withdrawals. A probability of failure metric fails to acknowledge these nuances and assumes a fixed satisfaction from each withdrawal dollar throughout retirement. Ignoring these behavioral tendencies may potentially lead to unrealistic assumptions and solutions, such as advocating constant real withdrawal strategies for the duration of retirement. 9 In the final year of life, 120 based on our life tables, we assume the optimal solution is for retirees to spend all of their remaining wealth (excluding any bequest amounts). Other constraints and objective functions can also be introduced, depending on the specifics of the individual. Please refer to the Appendix for technical details of framework and algorithm. EXHIBIT 6: J.P. MORGAN DYNAMIC MODEL KEY INPUTS CALCULATION DYNAMIC OUTPUT Define utility function parameter values to describe retiree preferences Define life expectancy from 60 onwards Set levels of retirement wealth and lifetime income Generate 10,000 simulations accounting for non-normal financial market returns Solve for optimal asset allocations and withdrawal rates at each age to maximize expected lifetime utility = Customized asset allocation and withdrawal rate considerations based on age, wealth, lifetime income, gender, marital status and risk profile 12 BREAKING THE 4% RULE

15 Second, it does not adequately allow for the risk of excess wealth accumulation or any degree of potential shortfall in retirement. Probability of failure metrics only consider downside risks in a binary fashion, the consequence of which may be advocating strategies that are overly conservative, either with low withdrawal rates and/or fixed income heavy portfolios. This may risk unnecessarily reducing retirees standard of living, as well as resulting in a potential accumulation of assets later in retirement for which investors ultimately have little need or use. A utility-based metric eliminates these problems. To be fair, even this preferred approach has drawbacks. For example, the choice of utility function is not easy or clear cut, and even then there is debate around the proper choice of parameter values to be used in calibration. The concept of utility also cannot be measured directly but rather relies on drawing inferences from human behavior. Accordingly, there is still considerable subjectivity in measuring and evaluating the utility of different withdrawal strategies. However, while this measure is by no means perfect, we believe it incorporates much more realistic assumptions about human preferences and retirement dynamics than traditional metrics. These insights provide significant advantages in developing a withdrawal model better structured to meet the real-world needs of retirees. ACCOUNTING FOR BEQUESTS AND LEGACIES The J.P. Morgan Dynamic Strategy can be customized to accommodate retirees specific bequest and legacy aspirations. Because this model strives to exhaust retirement assets in the most efficient manner possible, retirees simply must segment whatever dollar amount they hope to leave behind away from their withdrawal portfolios. This ensures an adequate amount will be reserved to fulfill their wishes rather than hoping enough assets are left in their estates to meet these needs. THE IMPORTANCE OF PLANNING FOR EXTREME MARKET EVENTS The concept of investment fat-tail risk refers to the relatively smaller probability of experiencing extreme performance outcomes that may occur less frequently but remain well within the realm of possibility. The challenge is that these severe events are completely random and may happen far more often than most people are willing to contemplate. Just ask retirees unlucky enough to suffer recently through two devastating bear markets and a decade of flat equity returns. J.P. MORGAN ASSET MANAGEMENT 13

16 Dynamic retirement income withdrawal applications Our main recommendation to retirees through this research is to be dynamic in their post-retirement financial planning, adapting their asset allocations and withdrawal rates to their evolving circumstances and portfolio experiences. To this end, we analyzed how a dynamic withdrawal strategy may broadly apply to changes in three core investor characteristics: 10 Age Lifetime income Wealth level To aid in this understanding, we first analyzed our model s output based on changes to each individual factor, holding all other factors constant (see Exhibit 11 for a summary). Portfolio suggestions were based on equity and bond allocations. 1. Impact of aging As individuals get older, life expectancy decreases based on standard mortality table calculations, albeit in a non-linear fashion due to survivorship bias. 11 Lower life expectancy also implies a shorter investment horizon for the portfolio. This suggests that, if the goal is to utilize retirement savings more fully, the effects of increasing age with regards to withdrawal rates and asset allocations should be as follows: For a given level of wealth and lifetime income, older retirees may have a higher withdrawal rate recommendation than younger retirees, as there is less time remaining to spend retirement savings. Older retirees should be more conservative in their asset allocations, with higher allocations to fixed income, as there is less time to recoup any potential portfolio losses to finance future withdrawals. Exhibit 7 shows how our dynamic framework details this precise relationship between aging, withdrawal rates and asset allocations. In this example, the suggested first year withdrawal rate for a 65-year-old couple with $1 million in retirement savings and $50,000 in lifetime income is 5.9%, with a suggested 17% total portfolio bond allocation. For a 75-year-old couple also with $1 million in retirement savings and $50,000 in lifetime income, the suggested withdrawal rate jumps to 7.7% per year, while the suggested bond allocation increases to 24%. BOTTOM LINE: Increasing age allows retirees to increase their withdrawal rates and decrease their equity allocations. EXHIBIT 7: UTILITY-BASED WITHDRAWAL RATES AND ALLOCATIONS ACROSS AGE (HOLDING LIFETIME INCOME AT $50,000 AND WEALTH AT $1 MILLION CONSTANT) Withdrawal rate (%) Withdrawal rate (LHS) Age Retiree age Suggested withdrawal rate Bond allocation (RHS) % 17% % 21% % 24% % 28% % 31% % 34% % 36% Suggested bond allocation Source: J.P. Morgan Asset Management. For illustrative purposes only. 10 The optimal withdrawal rate and asset allocation is also a function of class of utility function and the choice of parameters. 11 For every year that an individual ages, life expectancy decreases by less than a year. So while 65-year-old males have a life expectancy of 17.5 years, 66-year-old males have a life expectancy of 16.8 years, greater than what is obtained by subtracting one from the life expectancy of the 65-year-old males. This is because survivors are generally healthier than those that did not survive in the previous year, and thus the life expectancy of these survivors is based on a generally healthier population Bond allocation (%) 14 BREAKING THE 4% RULE

17 2. Impact of lifetime income As discussed earlier, retirees with little or no lifetime income have a greater risk of poorer outcomes later in retirement than retirees with higher levels of lifetime income. This suggests that the effects of increasing lifetime income levels with regards to withdrawal rates and asset allocations should be as follows: For a given age and wealth level, withdrawal rates at higher lifetime income levels should be greater than at lower lifetime income levels. This is because the higher secured income floor reduces the likelihood of extremely poor outcomes in retirement due to excessive withdrawals. Retirees with greater levels of lifetime income should be more aggressive in their asset allocation, with larger equity allocations, as a higher proportional part of their overall retirement income stream is protected from any negative ramifications from potential negative equity returns. Exhibit 8 shows how our dynamic framework quantifies the precise relationship between lifetime income levels, withdrawal rates and asset allocations. In this example, the suggested first year withdrawal rate for a 65-year-old couple with $1 million in retirement savings and $20,000 in lifetime income is 5.5%, with a suggested 38% total portfolio bond allocation. For a 65-yearold couple with $1 million in retirement savings and $40,000 in lifetime income, the suggested withdrawal rate jumps to 5.8% per year, while the suggested bond allocation decreases to 24%. BOTTOM LINE: Greater lifetime income, from sources such as Social Security, pensions or lifetime annuities, allows retirees to increase both their withdrawal rates and equity allocations. 3. Impact of wealth There are two main drivers to consider in terms of the impact different wealth levels may have on the retirement withdrawal planning process. First, higher retirement wealth levels require a smaller proportion of withdrawals to be spent on basic necessities, such as food, shelter and health care, allowing potentially more to be spent on non-essentials, such as travel and entertainment. Second, utility theory suggests asymmetric attitudes of individuals with regards to gains and losses. In a retirement context, reducing withdrawals by a certain dollar amount carries more emotional distress than the pleasure obtained from increases of a similar amount. This effect is more pronounced for wealthy retirees, as a specific EXHIBIT 8: UTILITY-BASED WITHDRAWAL RATES AND ALLOCATIONS ACROSS LIFETIME INCOME LEVELS (HOLDING AGE AT 65 AND WEALTH AT $1 MILLION CONSTANT) Withdrawal rate (%) Withdrawal rate (LHS) Bond allocation (RHS) $0 $10 $20 $30 $40 $50 $60 $70 $80 Lifetime income (thousands) Lifetime income per year (in U.S. dollars) Suggested withdrawal rate $10, % 46% $20, % 38% $30, % 30% $40, % 24% $50, % 17% $60, % 12% Source: J.P. Morgan Asset Management. For illustrative purposes only. percentage decline in wealth, due for example to poor equity performance, leads to a larger dollar and utility reduction compared to less wealthy retirees. Keep in mind that this concept is distinct from wealthier retirees financial ability to weather volatility, at least in terms of securing minimum baseline levels of living essentials (e.g., market losses are unlikely to result in wealthier retirees going hungry, but it may prompt more at-home dinners than eating out at fancier restaurants). Instead, it focuses on the greater emotional distress of experiencing larger dollar losses. This suggests that the effects of increasing wealth levels with regards to withdrawal rates and asset allocations should be as follows: For a given age and level of lifetime income, withdrawal rates at higher wealth levels should be less than at lower wealth levels. This is because the actual dollar amount withdrawn is substantially higher and the satisfaction derived from greater withdrawals does not increase Suggested bond allocation Bond allocation (%) J.P. MORGAN ASSET MANAGEMENT 15

18 proportionally once a reasonable lifestyle level has been achieved. Utility theory suggests that beyond that point reducing withdrawals by a certain dollar amount carries more distress than the pleasure obtained from increases of a similar amount. This effect is more pronounced for wealthy retirees, as a specific percentage decline leads to a larger dollar and utility reduction compared to less wealthy retirees. Wealthier retirees should be more conservative in their asset allocations, with larger fixed income allocations. Poor portfolio returns carry greater pain than commensurate upside performance, and the same percentage decline will result in a higher dollar loss for a larger portfolio compared to a smaller portfolio. Exhibit 9 shows how our dynamic framework captures the precise relationship between retirement wealth, withdrawal rates and asset allocations. In this example, the suggested first year withdrawal rate for a 65-year-old couple with $1 million in retirement savings and $50,000 in lifetime income is 5.9%, with a suggested 17% total portfolio bond allocation. For a 65-yearold couple with $2.5 million in retirement savings and $50,000 in lifetime income, the suggested withdrawal rate falls to 5.5%, while the suggested bond allocation increases to 38%. BOTTOM LINE: Higher initial wealth suggests retirees lower their withdrawal rates and increase their fixed income allocations. EXHIBIT 9: UTILITY-BASED WITHDRAWAL RATES AND ALLOCATIONS ACROSS WEALTH LEVELS (HOLDING AGE AT 65 AND LIFETIME INCOME AT $50,000 CONSTANT) Withdrawal rate (%) Wealth level (in U.S. dollars) Withdrawal rate (LHS) Suggested withdrawal rate Bond allocation (RHS) $500 $700 $900 $1,100 $1,300 $1,500 $1,700 $1,900 $2,100 $2,300 $2,500 Wealth levels (thousands) $500, % 0% $1 million 5.9% 17% $1.5 million 5.7% 28% $2 million 5.6% 34% $2.5 million 5.5% 38% Suggested bond allocation Source: J.P. Morgan Asset Management. For illustrative purposes only Bond allocation (%) EXHIBIT 10: SUMMARY OF HOW INDIVIDUAL FACTORS MAY AFFECT WITHDRAWALS AND ASSET ALLOCATIONS Factor Withdrawal rate Equity exposure Increasing age Higher lifetime income Higher initial wealth level EQUITY ALLOCATIONS AND RISK PROFILE SUITABILITY The J.P. Morgan Dynamic Model solves for a suggested asset allocation factoring in lifetime income as a bond-like investment. Additionally, the equity allocation is reflective of the risk capacity individuals have based on their mortality at various ages. For example, at age 65 a male has a 78% chance of living 10 more years, whereas at age 90 the probability is 5% for living 10 more years. As a result, the model may recommend a higher allocation to equities than may be more commonly applied when making a recommendation for investable wealth alone based on a client s risk profile. 16 BREAKING THE 4% RULE

19 Multidimensional aspects of a dynamic withdrawal strategy Although an analysis of each of these individual factors provides useful background, in practice age, wealth and lifetime income are likely to evolve concurrently throughout retirement. Hence, it is important to understand how our model adapts withdrawal rates and asset allocations when more than one factor changes at the same time, since this is how individuals typically experience retirement. Exhibit 11 offers specific examples of the J.P. Morgan Retirement Income Withdrawal Tables generated by our dynamic framework. These provide actual suggested allocations and withdrawal rates for different combinations of age, wealth and lifetime income. To illustrate how these tables function, consider three hypothetical situations and the optimal withdrawal and asset allocation suggestions implied by our framework for each: Case study 1: 65-year-old couple with $1 million in retirement savings and $50,000 in lifetime income. Our suggested withdrawal rate is 5.9% for the next year, with a suggested bond allocation of 17% and the remaining 83% invested in equities. Case study 2: 60-year-old couple with $2.5 million in retirement savings and $20,000 in lifetime income. Our suggested withdrawal rate is 4.8% for the next year, with a suggested bond allocation of 47% and the remaining 53% invested in equities. Case study 3: 80-year-old single female with $500,000 in retirement savings and $20,000 in lifetime income. Our suggested withdrawal rate is 9.1% for the next year, with a suggested bond allocation of 28% and the remaining 72% invested in equities. EXHIBIT 11: SAMPLE J.P. MORGAN DYNAMIC RETIREMENT INCOME WITHDRAWAL TABLES Couples (same age) $20,000 lifetime income Optimal withdrawal rate Optimal bond allocation Couples (same age) $50,000 lifetime income Optimal withdrawal rate Source: J.P. Morgan Asset Management. For illustrative purposes only. Optimal bond allocation J.P. MORGAN ASSET MANAGEMENT 17

20 Males $20,000 lifetime income Optimal withdrawal rate Optimal bond allocation Males $50,000 lifetime income Optimal withdrawal rate Optimal bond allocation Females $20,000 lifetime income Optimal withdrawal rate Optimal bond allocation Females $50,000 lifetime income Optimal withdrawal rate Optimal bond allocation Source: J.P. Morgan Asset Management. For illustrative purposes only. 18 BREAKING THE 4% RULE

21 Stronger projected withdrawal outcomes Next, we wanted to evaluate how our framework might stand up to the rigors of real-world retirement funding. We compared the customized considerations of the J.P. Morgan Dynamic Strategy to the static output of the 4% rule and RMD model, assuming a fixed 60% equity/40% bond allocation across retirement for the latter two approaches. We applied these three withdrawal strategies to the previous three case studies, running each hypothetical investor example through 250,000 simulations across the entire retirement horizon until death. For each simulation, the market return each year drew from the 10,000 possible equity, bond and inflation scenarios used in our optimization process, covering the gamut from strong rising periods to severe declines. Each withdrawal strategy was assessed in two ways. First, we analyzed the distribution of portfolio values over time to understand how successful the withdrawal approach was in addressing key risks in retirement. Second, we numerically evaluated how effective each strategy was at maximizing retirement income for risk-averse investors. TAX CONSIDERATIONS Our model is focused strictly on optimizing withdrawal rates and asset allocations in retirement. It does not take into account any applicable taxes, due to the highly variable nature of different tax situations. ONGOING ANNUAL TABLE UPDATES These tables may change in any given year, as new research around retirement income withdrawals emerges and J.P. Morgan Long-Term Capital Market Assumptions and the U.S. Social Security Administration s Actuarial Life Tables evolve. ADAPTING TO SPECIFIC RISK PROFILES Our model output can also be adjusted to accommodate different risk profiles. Sample tables for investors comfortable with either greater or less market risk can be found in the Appendix. Our first test applied a straightforward measure: Does the strategy provide a steady stream of income while effectively avoiding both premature depletion of portfolio assets and excess wealth accumulation? The results of our analysis are illustrated in Exhibits 12 and 13, which respectively show the distribution of potential withdrawals and corresponding remaining portfolio values from ages 65 to 105 for each of the three withdrawal strategies. ABOUT THE BOX-AND-WHISKER CHARTS Values are presented in a box-and-whisker chart format, with the box marking the range of the 25th, 50th (median) and 75th percentile outcomes, from top (best) to bottom (worst). The whiskers reaching out from the top and bottom of the box show the range of the distribution of outcomes up to the 5th and down to the 95th percentiles. As the dispersion of consumption streams and values increases, the box-and-whiskers become more elongated, indicating a wider range of possible outcomes and, by extension, greater income and portfolio uncertainty. Percentile 5% 25% 50% 75% 95% J.P. MORGAN ASSET MANAGEMENT 19