This blog post provides additional details associated with a couple of interesting papers I mentioned over the past few weeks. The first is by Sexauer and Siegel proposing a self-executed low-/no-risk approach to retirement planning which aims to deliver DB plan like outcome. The second is Sheikh, Roy and Lester’s dynamic decumulation strategy which proposes an integrated quantitative black-box approach including not just the usual age, wealth, lifetime income and risk tolerance but also the utility/satisfaction that a retiree receives from a particular withdrawal strategy. The first one (very conservative) you could probably execute in some form yourself, while the second you’d have to rely on an ongoing JPMorgan implementation.
(I)-Highlights of “A pension promise to oneself “ by Stephen Sexauer and Laurence Siegel
In a nutshell
With the demise of DB pension plans, Sexauer and Siegel propose a way of building a “pension promise to oneself” within the context of DC-like plans, or at least consider it as a benchmark to cost of building a safe pension. They propose specific “no risk” and “low risk” mechanisms to accumulate required assets, to deliver a target retirement income by a combination of high savings rates (20%+) invested in TIPS (real 0% assumed return) and longevity insurance. The pension target “promises” to deliver 70% income replacement rate indexed for first 20 years of retirement, followed by a nominal income stream thereafter based on the 20th year indexed income until death. In the current capital market environment a multiplier of 22 is needed to calculate the assets necessary to deliver a given income stream (i.e. $2.2M needed to be accumulated at age 65 retirement to deliver $100,000 indexed for the first 20 years of retirement followed by a constant nominal income stream starting after age 85, equal to income in 85th year). If death occurs before 20th year assets remain for an estate (the remaining part of the unused TIPS ladder), if death occurs after 20 years there are no residual assets but income continues for life from the longevity annuity. Even assuming a risky approach during accumulation, the required savings rates are much higher than currently typical practice.
Given the “pension catastrophe that we are facing”, and with the realization that “the battle to save DB plans in their original form is almost entirely lost”, Sexauer and Siegel propose a self-implementable approach to “achieve better DC plan outcomes”.
They define the components of a Personal Pension Plan as determining the: (1) required income stream in retirement (e.g. 70% of final working year income reduced by known/expected pension payments like Social Security), (2) retirement multiple (RM, which multiplied by the target retirement income defines the required assets necessary to deliver the income), (3) savings rate during working life which invested in low-risk assets will deliver the required assets for the retirement savings and income specified in (2) and (1), respectively.
The authors inject a sense of realism (which some of the retirement planning papers seem to forget) when they note that, just as companies’ DB plans make adjustments to changing circumstance (add more money or change benefit promises), individuals should be ready to make Personal Fiscal Adjustments (PFAs- such as reducing consumption and/or working longer/harder); “without such adaptive behaviour, the pension promise to oneself is almost doomed”!
In its simplest and minimum risk form, TIPS are used during the pre-retirement accumulation phase, and a laddered TIPS portfolio is used for the initial 20 years of retirement starting at age 65 followed by a “nominal deferred annuity” (longevity insurance) starting at age 85 equal to the last TIPS ladder payout. At DCDBbenchmark.com there is a monthly updated view of what a $100,000 payment would buy for a 65 year old male (for female or joint would buy less) in the market at the end of the current (and previous) month(s). For example at the end of February 2014 a $100,000 investment using this decumulation strategy will deliver $4,568 in year 1 and growing to $6,784 in year 20, followed by a continuing nominal payment of $6,784/year in year 21 and thereafter until death from the deferred annuity (longevity insurance).
Note that the Retirement Multiple RM is just the inverse of year 1 yield $4568/$100,000= 4.568%, so in this particular point of time RM=21.89; 85% of the assets are used to generate the first 20 years of indexed income while the longevity insurance consumes the other 15% of accumulated assets.
Furthermore they note that this self-created pension plan is essentially riskless (if you consider insurance company providing the longevity/deferred annuity going out of business as zero risk event).
The authors also explore a “risk taking” version of their approach, but only for the accumulation phase, given that once a person has retired “he cannot work longer or save more….the only available PFA if postretirement risk does not work out is to cut consumption”. The RM is the same when using a risky portfolio for accumulation, though you may be required to save less in each period, but you will have to make up for any return shortfall relative to planned return by either increasing savings later or reducing expected retirement income.
The authors demonstrate how one might determine the target retirement income, using professions with dramatically different human capital (real wage) growth rates: a teacher (1.92%/yr), a sanitation worker (1.05%/yr) and a software developer (heavily profiled from 10%/yr in early years to -13%/yr approaching retirement). Typically the approach is to start with: “initial salary”, number of working years assumed to be 40, some annual human capital growth profile appropriate to individual, target 70% of final year income replacement, an “initial savings rate” accompanied by an annual savings increase equal to some “percentage of real (inflation adjusted) portion of raise saved”. (The percent of raise saved approach is derived from the “Save More Tomorrow” program which proved very successful in getting people committed into gradually increasing the savings rate over time.) For the risk-free accumulation model a risk free rate of zero is assumed (which is about what is available today for TIPS), while for the risky model a 2% real rate is assumed (perhaps a 50% equity content assuming 4% equity risk premium?). Note that the lifetime savings goal is calculated by multiplying the RM by the 70% of final salary reduced by available other pension income (e.g. Social Security)
What is shocking is that the commonly stated target of 70% income replacement rate (even including Social Security) using realistic assumptions, lead to very high required savings rates, and you must remember that the model assumes 40 years of working and savings!
The authors comment on the teacher example that “Although 20% is not enough as a career average savings rate, it is a great start, whereas the tiny 6-9% rates often seen in employee benefit programs are not even in the right ballpark and will lead to penury in the employees’ old age unless supplemented by massive (and tax-disfavored) additional personal savings.”
The pension promise to oneself can be done but it is no cakewalk to achieve a 70% final year income replacement rate even factoring in Social Security. You can build your own personal pension plan: required savings rates are high (>20% with a risk free accumulation model), number of working/saving years are high (40), you must have a mechanism to deal with longevity risk (deferred nominal annuity/longevity insurance starting at 85 in this instance), you must be prepared to make Personal Fiscal Adjustment (PFA) as circumstances change (no plan can realistically be expected to remain unchanged over 40 years of saving followed by 35 ears of retirement).
(II)-Review of “Breaking the 4% rule” by Sheik, Roy and Lester
This JPMorgan paper argues that they have a superior decumulation approach than the “4% rule” (draw 4% of assets in year 1 and then adjust for inflation annually) by dynamically adjusting withdrawals and asset allocation to reflect market risk, longevity risk and personal circumstances. Specifically the paper considers: expected lifetime utility, age, lifetime income sources and wealth to dynamically manage withdrawals and asset allocation during retirement. Note that in this paper the authors assume that there is zero preference to leaving an estate (i.e. you spend your last penny when you die), though the methodology could accommodate a bequest preference.)
The authors argue 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 paper compares three decumulation strategies: (1) the ‘4% rule” (sole objective is to “preserve purchasing power”), (2) Required Minimum Distribution (RMD is the IRS based method starting at age 70.5 requiring an “Annual Withdrawal Amount= Portfolio Value/Remaining life expectancy”; RMD “strives to avoid depleting the portfolio assets prematurely” by using age only as determinant of withdrawal), and (3) Dynamic decumulation method (proposed here) of “maximizing expected lifetime utility” (by dynamically setting withdrawal rate and asset allocation factoring in: age, wealth, lifetime income, and risk profile.)
The authors list the five factors driving their dynamic decumulation approach: (1) differences in individual preferences to magnitude (decreasing marginal utility for incremental income) and timing ($1 received at a younger age valued more than at an older age) of withdrawals, (2) level of wealth and lifetime income (higher downside income risk tolerance for those with greater wealth and lifetime income sources), (3) age and life expectancy (using “survival weighted utility of withdrawal reduces attractiveness of strategies that defer income” to end of life), (4) market randomness and extreme events (using JPMorgan’s proprietary Non-normal Framework; JPM 2014 capital market expectations are (Return/Volatility): for US large cap (7.5%/14.75%), US Aggregate bonds (4.25%/6.5%) and inflation (2.25%/1.5%)), and (5) dynamic nature of the decision-making process (real people “adapt their withdrawal rates and asset allocations to changes in the market climate and their personal circumstances). Lifetime income appears to be factored into the model by transforming it to a bond-like investment. These five factors are combined into an optimization program to calculate the “optimal asset allocation and withdrawal rate at each age with the goal of maximizing lifetime utility”.
The authors argue that (as fuzzy/subjective it might be or as variable it might be from individual to individual) “maximizing lifetime utility” is a better measure of success than “probability of failure” for a given asset allocation and withdrawal rate, because the latter doesn’t account for utility (income timing/magnitude preferences), risk of excessive wealth accumulation or “degree of potential shortfall” in retirement. They, in fact, caution that “the choice of utility function is not easy or clear cut, and even then there is debate around the proper choice of parameter values…there is still considerable subjectivity in measuring and evaluating the utility of different withdrawal strategies”. (This utility function, and associated parameters, may be the strength (new twist) and the weakness of the approach.)
The authors’ recommendations to retirees (some common sense, others less obvious or even controversial) are as follows (no tax considerations, changes annually and varies with risk tolerance):
-use “dynamic post-retirement financial planning, adapting their asset allocations and withdrawal rates to evolving circumstances and portfolio experiences” (common sense)
-“increasing age allows retirees to increase their withdrawal rates and decrease their equity allocations” (common sense)
-“greater lifetime income, from sources such as Social Security, pensions or lifetime annuities, allow retirees to increase their withdrawal rates and equity allocations” (perhaps but it probably depends heavily on required incremental income to meet basic needs)
-“higher initial wealth suggests retirees lower their withdrawal rates and increase their fixed income allocations” (perhaps more controversial and likely strongly influenced by utility function and parameters, and again depends on basic and discretionary income targets not discussed in the paper)
To illustrate the proposed approach and corresponding recommendations, three case studies were looked at in detail:
…and for comparison, the Appendix also contains the following recommendations for “more risk averse” investors; notice that withdrawal rates are not that changed, but bond allocation is significantly higher (except for the 2nd investor couple with much higher wealth who are already treated very conservatively)
Considering the results of three case studies the authors compare the initially mentioned three decumulation approaches and conclude that the:
-“4% rule”: has high risk of excess wealth accumulation and premature asset exhaust, but provides steady real income stream (60% equity allocation)
-RMD: has higher income variability than “4% rule” but has low risk of excess wealth accumulation and asset exhaust (60% equity allocation)
-JPM dynamic decumulation: compared to RMD the odds are for higher income early in retirement and lower income variability, but still low risk of excess wealth accumulation and asset exhaust
The paper also uses the concept of “certainty equivalents” (which “measure how much lifetime real income individuals would accept in lieu of trading all their retirement assets and following a particular strategy”) as a measure of how effective a strategy is for risk averse investors. Their conclusion is that RMD and JPM dynamic decumulation are both better than the “4% rule” (pretty low bar, as I doubt that anyone uses that nowadays; it is more of a reference point) and that the JPM dynamic approach is slightly better with this measuring stick than the RDM approach for the examples given. (However given the subjectivity of the utility function and the associated parameters, many would call the latter two approaches about the same based on the numbers quoted in the article).
The authors conclude that “Our proprietary model helps investors and their financial advisors adapt appropriate withdrawal rates and bond exposure on an annual basis, tailored to investors’ age, wealth, lifetime income and risk profile”
Few if any financial advisors would seriously consider nowadays the set-it-and-forget-it for 30-years “4% rule”, or for that matter to disregard: age, wealth, (other) lifetime income or risk tolerance (and estate and/or defensive reserves to cover higher healthcare expenses in last couple of years of life) in setting an investor’s decumulation strategy.
Still this is a very interesting paper in that it tries to tackle the decumulation problem in an integrated quantitative manner, explicitly discussing and including all the various (obvious) factors like age, wealth, and (other) lifetime income which come into play in the decumulation decision, but then attempt to bring the more subjective “expected lifetime utility” into consideration as well. This proprietary (and certainly inaccessible to vast majority of DIY investors or even their advisors, without JPM help) black box approach then generates the upcoming year’s withdrawal rate and asset allocation.
Given that this is a proprietary approach, with somewhat opaque black box characteristics, it would be difficult to predict how this would perform over a 20-40 year retirement period. However, if I understand correctly, each year a fresh recommendation is generated for the next 12 months reflecting the changing circumstances of the individual/couple, the changing capital market expectations and model refinements. Given that the recommendation is only for one year, there is opportunity to make annual corrections to (overly aggressive or defensive) previous recommendations. This is consistent how a financial planner would approach an annual recommendation.
This could turn out to be a great marketing tool for JPM if it can pull in new retirement savings (AUM) and, given that recommendations likely to change annually, keep them there for many years. The question is, assuming that the advice is of similar quality as that available elsewhere from competent advisers/planners, will the cost of implementing with proprietary JPM funds will detract from outcomes.
Still some might find the recommended withdrawal rates and/or the asset allocations (e.g. 5.9% withdrawal rate and 83% stock allocation for 65 year old couple, or 72% stock allocation for the 80 year old female with $20K other income), especially without understanding explicitly the minimum required incomes to meet basic/musts for these investors in a 2008-2009-like market drop of 40-50%, somewhat aggressive. For an 80 year old female with about a 10 year life expectancy 72% stock allocation and 9% draw might only make sense if income derived from the assets is mostly for discretionary expenses and the $20K lifetime income meets majority of basic needs, or this female’s utility for the income derived from assets drops dramatically (e.g. two years of draw without a market recovery after 50% drop in equity prices would half assets and income; unlike the first paper discussed, this one makes no mention of Personal Fiscal Adjustment (PFA) which might be necessary, even though this is a dramatically more aggressive approach to draw and asset allocation).
Others might be uncomfortable with the capital market expectations used to generate the recommendations. But as indicated earlier, annual updates to the model will provide opportunities to fine tune, as long as market forces have not ravaged the size of the portfolio.
The relatively high initial withdrawal rates recommended will be found attractive by many. Whether high withdrawal rates and aggressive asset allocations will be sustainable on an ongoing basis will no doubt be determined over time. As the name suggests, this is a dynamic decumulation strategy which will be adjusted annually using the black-box (and presumably a responsible adviser). Over the longer term, the success or failure of what is proposed will be heavily influenced by market returns less the associated costs (probably in the range 1-2.5% to cover) to access the advice and the actively managed JPM funds that come with it.
Of course, dynamic decumulation decisions are the only way to go, and all the factors discussed (and likely some others) should be brought to bears for a specific individual. But I am not in a position to judge whether the specifics of this models are good, bad or indifferent, and how it will perform over time.
P.S. I read this paper a number of times but I must have missed or misunderstood whether: lifetime income referred to is real or nominal (probably real at least for Social Security), what risk tolerance means in simple terms and what risk tolerances were used for the three case studies, no mention of minimum /basic/musts expenses that must be covered or desired additional discretionary income, asset allocation determined by taking as much risk as you need or can tolerate? how does the JPM proprietary non-normal distribution reflect downside risk, would a more realistic/appropriate utility function incorporate a more U-shaped time preference for income during retirement reflecting cost of increased healthcare during last 3 years of life?