For this discussion, the relevant characteristic of probability-based vs. safety-first is that the probability-based approach uses a total returns investing strategy with a diversified portfolio of financial assets aiming to support the entire lifestyle spending goal for the retiree. The safety-first approach, on the other hand, differentiates between essential expenses and discretionary expenses. Volatile assets (with higher expected returns but also higher volatility and greater downside risk) should not be used to finance essentials. As Modern Retirement Theory describes it, essentials should be covered by assets that are secure, stable, and sustainable.
The important question is: why use risky assets as part of the retirement income strategy? Dirk frames the probability-based justification as being that people like to dream they will be able to increase their spending in the future. I agree that this is the implication of the strategy. The logic of this is explained in Stock and Watson's Floor-Leverage Rule, as they argue that the desire to have sustainable spending and also equity investments for greater upside seems to be the reason for using a probability-based 'safe withdrawal rate' type of approach. But they argue that this approach is convoluted, and if these are really the goals of a retiree, the retiree would be better served by first locking in a safety-first style safe spending floor with the majority of assets, and to then invest very aggressively with remaining assets with the hope of being able to raise the spending floor further in the future.
Dirk rightly points out that in practical terms there really isn't any point to do this. His posts are about how investing more aggressively when you are 65 so that you might be able to enjoy a higher standard of living when you are 80 really doesn't make all that much sense.
I agree. But the point of this post is that I'm not convinced that this is the way most people are thinking about retirement spending. My thought is that people are logically wanting to 'amortize' their future (speculative) market gains by spending more today. I think this is the way matters are considered in the original Bill Bengen research and also in follow-ups like Jonathan Guyton's decision rules. The emphasis was about how if you are willing to cut spending in the future when markets underperform, then you could spend more today. Let's get today's spending higher.
There are two ways to be aggressive with a retirement strategy. Spend more today, or invest assets more aggressively for greater upside potential. In this context, I think the following figure basically explains the logic behind the probability-based approach:
The more aggressively one wishes to spend, the more aggressive is the asset allocation which minimizes the probability of failure. [Note, this particular figure is based on historical average returns, and it is not the figure I would suggest using in today's market environment to decide on a withdrawal rate].
And so, for those with a more aggressive lifestyle goal, the reason to invest more aggressively is to improve the odds that the goal will be met, accepting all of the accompanying downside risk this creates. I'm not saying this is the right decision to make, I'm only saying that my understanding of the logic of probability-based approaches is that this is what people actually have in mind.
In some sense, this post may be trying to make a distinction about something relatively minor and unimportant. But I think this is an important question: how to retirees who accept the probability-based approach and who do not otherwise have large bequest goals think about the decision to invest for upside?
Is it with the hope that they will be able to continue increasing their spending as they age?
Or, is it with the hope that this will justify them spending more today than they could otherwise lock-in with safe assets, since their portfolio will hopefully enjoy the upside they seek?
Hi Wade,
ReplyDeleteA great post once again! Hopefully the profession will begin to trend towards mixing both the probability based and safety first concepts – instead of which one is better than the other. I’ve always found with most things in life a combination is often better than either or.
A couple of points that Frank, Mitchell and Blanchett found in their series of papers that look at how cash flows are impacted when probability based simulations are serially connected to simulate a person aging through retirement. First, seeking higher returns also results in higher volatility. Thus, what returns seem to give (higher WR), volatility takes away (lowers the WR). Second, spending more today results in lowering the portfolio balance for tomorrow’s spending – which linear thinking suggests needs to be compensated for by seeking higher returns. This brings us back to the first point – higher returns with their volatility lowers the WR. So seeking an allocation based on the retiree’s time remaining resolves this circular result – by establishing first and foremost the length of the distribution period most probable for the retiree at their present age.
As to your great questions: Retirees should be taught the larger picture about shepherding their resources with balance in mind rather than chasing returns with greater volatility or trading off tomorrow for today (the classic struggle people always have between spending versus saving). Returns first compensate for any inflationary effect while markets give or take some spending ability on the margins. Social Security and/or pensions provide some liability matching for housing as an example (the main component of most retiree liabilities when liability matching expenses, and possibly more living expenses for many. Thus, the debate between safety first and probability based approaches is often about discretionary expenses in many cases.
I may have strayed from your main point – but doing so was to bring into focus what the real conversation should be about – managing resources regardless of source for the remaining probable lifetime in a prudent manner. The combination of safety first and probability based methods should also look at how much safety first comes from non-portfolio sources. Trying to provide absolute safety, or the other extreme total dependence on markets, tends to emphasize the con of either.
A great post to get thoughts moving towards how to combine both approaches rather than either/or Wade.
Though I am often accused of being a safety-first guy, and I do lean that way, I actually see a continuous spectrum from safety-first to probability-based strategies and feel like my job as an adviser is to place the client on the appropriate point along that spectrum for her given circumstances. As a retiree, that's what I do with my own plan.
DeleteThanks Larry. Considering that spending more today increases the probabilty of needing to spend less tomorrow, and also the vagaries of asset allocation and portfolio volatility, this really is a dynamic problem.
DeleteIMO you are correct: "with the hope that this will justify them spending more today than they could otherwise lock-in with safe assets, since their portfolio will hopefully enjoy the upside they seek." Of course some may be motivated by the goal of providing for heirs or other beneficiaries.
ReplyDeleteAs I see it, the problem is adopting any formulaic approach -- probability-based or safety-first. Instead each individual should adopt a dynamic planning approach that is flexible in both spending and investing and which is continually updated. Of course broad probabilities must always be considered, but no formula can do the job for you. There ain't no science that is better than that.
As always, I find this commentary interesting but frustratingly blind to a reality of modern retirement investing in the US: the required minimum distribution. One is free to model any level of withdrawals before age 70 1/2, but to be useful to current and future retirees who have most or all of their retirement savings in tax-deferred IRA accounts, the withdrawal or spending plan must account for RMDs after that age. Early withdrawals from IRAs could be used to spend, to rollover to a Roth, or to invest in a taxable account. Various combinations of these actions surely will lead to varied probabilities of success, but it's not obvious which might be best.
ReplyDeleteRMDs require only that one performs the relatively simple task of computing an additional annual tax liability. Otherwise RMDs should have no impact on retirement planning.
DeleteBut with the sudden bolus of money that arrives via the RMD, there is the question of "now what?" Do I spend the money whimsically, put it back in the stock market, buy a bunch of dividend stocks, purchase short-term Treasuries, or what? I don't give a ding-dang about a legacy, but I sure don't like frittering money away, either. So, my RMD's have a large impact on my retirement planning.
DeleteOf course there is the psychological aspect to retirement planning. It may be tempting to use the increased after-tax cash flow in a whimsical manner, rather than to follow your plan. Otherwise RMDs simply become routine cash-flow investment planning.
Delete"Follow your plan" is, more often than not, the answer to these kinds of questions. Why would your decision to spend the money whimsically, buy Treasuries, or buy dividend stocks change because you had to withdraw the money and pay taxes on it? Why not pay the taxes (if owed) and then reinvest the net amount the same as it was before?
DeleteI fear we may all be straying from the issue. I don't think "the hope that this will justify them spending more today than they could otherwise lock-in with safe assets" is being questioned, certainly not by me. I'm thinking past that.
ReplyDeleteThe retiree in question has already decided to take additional risk to spend more than she could with safe assets. Maybe she could spend 3% with safe assets and 4% with a 60/40 portfolio, for example, and decides on a pure systematic withdrawal strategy or floor-and-upside. The "upside potential" I'm referring to is upside potential for increasing wealth or increasing spending beyond 4%, not beyond a safe payout.
As Money magazine so often told us, "you can spend 4% of your initial portfolio and still end up with a huge portfolio". (Which is mostly true-- you might.)
My post was about what you do with that extra wealth beyond what is needed to support your initial spending rate. I see three basic purposes. First, non-recurring desires such as a bequeath or money to cover long term care, for example, make perfect sense.
Second, to gradually increase your spending throughout retirement as your wealth grows, by spending the same percentage of a growing remaining portfolio balance with SW, or by buying more flooring with floor-and-upside.
Wade offered a third alternative: to spend more in early retirement with hopes of making up for it with stock gains later in retirement. While I agree that SW suggests a higher initial withdrawal rate when higher market returns are expected in the future, a typical retiree with SW would increase spending only after her portfolio grew, not in anticipation of future growth.
My post suggested that investing in stocks to increase spending over time (my second alternative above) isn't attractive because a) retirees "typically" spend less as retirement progresses and b) by the time you earn the additional spending it may be too late to enjoy spending it.
There is only one question for me here: how many retirees who choose systematic withdrawals do so in hopes that they will increase their spending over time beyond their initial retirement year spending? What I tried to point out in my post is that benefit may not be as attractive as it first appears.
Dirk,
DeleteI agree with this. Retirees should think carefully about justifying greater stock investments as a way to live a more lavish lifestyle at 75 than at 65.
My perspective as a retired super-saver: I know how much I've historically spent each year; take that and multiply by some factor (I use 1.66), add in taxes and medical care, and set aside a portion of our portfolio so that a 64/40 mix and 2% withdrawal rate meets this (wildly inflated) estimate of needs. Now what do I do with the leftovers? Early bequests? Charity? Procrastinate? We've chosen the latter -- invest it aggressively for now and decide what to do in 10 years or so.
DeleteThere does seem to be a lot of straying. Investments in equities rather than safe assets are made in order to increase returns. Increased returns are generally desired both for the present and the future. Regardless of what formulas may be used, equities involve risk, but, at least in theory, safe assets do not.
ReplyDeleteBeyond that, I am at a loss regarding what debate exists. Maybe Dick is saying that use of a proper SW eliminates the equity risk provided the individual defers spending of any additional equity earnings????
No debate, really. Just a question. Do retirees who invest in equities do so because they hope that their spending can increase from the initial withdrawal amount over time if their portfolio grows, or is that NOT a significant reason they would choose SW? Your third sentence suggests the former.
DeleteExcellent post, Wade. I have been gravitating to the safety-first camp based on wading through the literature on life-cycle finance. My conceptual view is: take care of basic spending need first by optimizing Social Security and filling in with SPIAs, which leaves discretionary spending to be dealt with. With discretionary the asset allocations and spending patterns will depend on risk aversion and the personal discount rate. Risk averse people can't handle much change from year to year in discretionary spending so they will favor fixed income portfolios resulting in lower average spending than someone who can tolerate more volatility--a world cruise one year and a visit to relatives the next--and these more risk tolerant individuals will optimize with more aggressive investment allocations. How much optimal spending gets front loaded depends on both risk aversion and the personal discount rate--low risk aversion and a high personal discount rate goes with front loading the retirement spending. Although the argument can be made that coming up with utility functions and personal discount rates involves too many assumptions, it makes sense to me as a conceptual approach and, along with stochastic dynamic programming, it all falls into place nicely. There's actually a lot of common sense buried under the complicated math that many economists use.
ReplyDeleteThanks Joe,
DeleteIt's going to be worth exploring these issues further. The research has been done, but I'm not sure if clear explanations of the logic are available for widespread public understanding.
You've got the utility maximization model best explored in Moshe Milevsky's "Planet Vulcan" article.
And you've got the dynamic programming approach which most recently entered public consciousness through that JP Morgan report.
There is a lot of good stuff in these models.
When I first read the Cotton article, my reaction to it was in between the way Wade interprets it and the way he suggests it might be re-interpreted. I don't believe that those advocating a higher allocation to risky assets are thinking that they will be able to increase their spending in ten years, nor do I think that they are necessarily thinking that they want to spend more this year. My interpretation is that they are thinking that in the next two to five years, if outcomes are good, they can spend some of the excess; for example, given a 30% up year with a 50% stock portfolio, and thus looking at about a 15% gain, perhaps they can take that cruise next year that they would like but can't swing this year or perhaps they can provide some help to a college-age grandchild.. Now, one can argue as to whether that's appropriate or not, but it seems to me to make more sense psychologically that either of the other interpretations.
ReplyDeleteI agree with you. I wasn't suggesting, though, that a retiree might not WANT to spend money sooner than ten years, or even that he might not be able to. Rather, that given the volatility of stocks, amounts we would want to spend in less than 7 to 10 years are probably better invested in our bond portfolio. If we used a time-segmentation strategy, they would be.
DeleteDirk asks in his blog post of 3-20-14, << Is the promise of continually improving your standard of living as retirement progresses a significant incentive for you to invest in equities after you retire?>>
ReplyDeleteI'd answer, "The promise of continually SECURING my standard of living as retirement progresses is a significant incentive for me to invest in equities now that my wife and I are retired." That is, we want to live no worse than we do now (and possibly even a bit better over the years as we continue to age), taking into account that either or both of us might need to spend large but unknown amounts on assisted living sometime in the future. Both of us are in good health, and we try hard to stay that way, but the thought of having to pay $6-$12K/m to support one or both of us in some comfortable assisted-living facility scares the hell out of us. And, the thought of having to be moved from an expensive, comfortable facility to some $3-$4K/m Medicaid-accepting facility if we could no longer afford the expensive facility, is even more disconcerting. That was the option our father/father-in-law faced just 3m before his death. It was a heart-breaking option to have to face for a loved one who already had suffered the loss of independence.
So, we never, ever want to run out of the money we need to self-fund good, expensive care. At the same time, we want to live retirement independently and comfortably as long as we can, as we are now.
Francis,
DeleteThanks for sharing,
This sort of introduces a 4th category. Investing in equities not to raise the standard of living today or in the future, but as the best shot of being able to have enough growth in the portfolio to fund uncertain and potentially large long-term care expenses.
In this sort of case, minimizing the probability of "failure" might be the appropriate goal, as without growth you are not expecting to have enough to cover these expenses. Only high growth will make this goal possible, and only equities give a shot at achieving this. Thank you.
Wade, I have a different perspective on this. Saving for possible high LTC expenses and mitigating longevity risk are two very different animals. Retirement income is relatively certain. We'll need it every year we live and we know roughly how much it will cost. LTC costs can range from nothing (for about 40% of retirees) to enormous (for a few) and may occur decades in the future. Attempting to tweak your income strategy to cover both liabilities can't address either optimally.
DeleteThis should be an insurance problem, but LTC insurance doesn't work, at least not well. Saving for it after you retire will be an enormous challenge, but if I were going to try to do that, I'd match the liabilities with separate portfolios.
This seems like a strong case for purchasing LTC insurance as an alternative to trying to rely mostly on volatile assets like stocks to fund potential assisted care expenses. A LTC policy would do a better job than stocks at "securing" a standard of living that could be interrupted by assisted care expenses, possibly sooner rather than when stocks have had years to produce "high growth." Stocks may take many years to grow substantially and could be down significantly just when LTC expenses occur. And LTC policy payouts are usually tax-free.
DeleteOne thing that people often forget is that when one moves into a LTC facility and they have a spouse/partner remaining at home, the living expenses at home often drop substantially (perhaps on the order of 25%) during that time; this frees up money that can help offset the LTC charges. And if one lives alone and enters LTC for an extended period, many of the expenses of living at home disappear -- especially if the house is sold -- and can be applied to the LTC costs.
I have a good LTC policy and for my retirement plan, I plug the numbers into my financial projections under the assumption that the last two years of my life and that of my spouse will be spent in a nursing home. Furthermore, the plan includes the sale of our home when the second person enters a nursing home. I think that if more people plugged potential liabilities like long term care explicitly into their financial projections -- along with the payouts that a good LTC policy would provide -- the planning process might be less intimidating and people might feel less compelled to rely on volatile stocks for health care security.
I think Francis Gilbert has it right.... we plan for a future. The better off we are the more likely we will be able to achieve the financial goals to reach that future. If you don't have any kids or grand kids that is one situation. If you have a handicapped dependent you have a different set of rules for planning. If your current retirement income is insufficient go back to work and save some more. There are alternatives between $12K per month and $3K per month. My neighbor across the street hired a live-in house keeper for about $4K/mo and another $1K in expenses. Lived at home until he passed and had friends and neighbors around to visit because he was easy to find.
ReplyDeleteEach individual or couple is so different you can't have one rule. But, if you have to depend on investments to secure some part of your standard of living you are in trouble. Work more, increase risk in the hope you can have reduced risk in the future.... it is all a balancing act.
Something I don't see written about very much is using declining withdrawal percentages from an equity-based account (60/40 to 80/20) on top of a floor of Social Security, pension, annuities etc.
ReplyDeleteWe are fortunate enough to have had decent paying jobs and we have had the discipline to save a fair amount of money over the years. I envision our strategy in retirement (starting @ 8-11 years from now) to use Social Security, a small pension, and annuities from about 20-30% of our savings as a floor. I expect we will probably buy our savings to buy some TIPs to cover the first few years until age 70 so that we can delay claiming SS until then to maximize those benefits.
After that I am thinking of basing a withdrawal rate from the savings as simply the dividend and interest income from the remaining savings + 2-3% of principal for the first few years and then declining that percentage over time as our income demands decline. The dividend and interest income should increase with inflation and we would still have substantial assets sitting there in case of a major medicale or long-term care demand.
We would get a pretty steady income floor that would slowly rise with inflation from the Social Security, pension and annuity (at least some purchased to adjust with inflation) while the savings withdrawals would fluctuate some with the markets, but shouldn't be highly variable as a percentage of total income.
The approach you are taking is generally called the essential-discretionary approach. It's a bit different from the total returns approach where a volatile portfolio is used to fund the entire lifestyle spending goal.
DeleteWith your approach, you have essentials covered, and your stocks are used for discretionary expenses where there is some flexibility to raise or lower expenses.
This sounds fine. Jonathan Guyton does believe that this approach may create a smaller chance to being able to fund your lifestyle goal, but one's risk aversion is an important consideration here.
Given the U-shaped typical spending model that I have seen, it appears to me that this could actually maximize lifestyle goals, since the higher spending is likely to occur in the first 5-10 years. As we have seem, reliance on equities for guaranteed assets and spending over a 5 year period is fraught with problems. So relying on equities for only a discretionary spending component in the first few years while using their proven long-term inflation hedge for managing major risk items in the later years seems to be matching the risk profiles of the assets and needs fairly well.
DeleteIf the withdrawal rates from a volatile portfolio are dictated by 30 year plus time horizons, then it seems that a substantial chunk of the first 10-15 years may be best managed by using the actuarial payout calculations of SS, pensions, and annuities to maximize those payouts where the insurers are betting the average person won't make it past 80 instead of assuming that everybody will have to be paid until they are 95. Unless the expenses associated with those strategies, the overall payout ratios should be higher than a similar portfolio managed to pay money until 95 or later.
In our personal case, I expect that we will end up with a total base "essential" regular "paycheck" that will be about 3x our SS payment with much of it inflation adjusted. That should cover the essentials and a bunch of discretionary spending. We would still be left with about 50-70% of the original at-retirement asset base at age 70 to grow in a relatively high (60%-80%) equity portfolio to generate discretionary income and future major risk items.
Why more research is not done on the 'essential- discretionary' approach? To me this is the most logical way to approach retirement. If I have the basic needs (home,food,health) covered by social security, pension(if available),dividends and interest, I have the flexibility to postpone my discretionary needs on bad years. This might give me the ability to hold higher percentage in stocks. If JNJ drops by 30% in a bad year, I will not be forced to sell as the dividends which I rely on is uninterrupted and is in fact growing.
ReplyDeleteThe problem with this kind of research is that it is purely academic and does not consider the practical ways a retiree can be flexible in withdrawal rates based on market conditions if their portfolio is structured accordingly.
While the essential-versus-discretionary approach may be more intuitive to some, real-world evidence shows that the systematic withdrawal approach is also very practical.
Deletehttp://www.fpanet.org/system/getAsset/?id=440EA706-DC3F-B39D-7209A76C46D656CA
Yes, Jonathan Guyton argues along these lines as well. I was just reviewing some video interviews about this with him and Michael Kitces, and hopefully I'll get them posted to the blog this week. Thanks.
Delete