Thursday, March 1, 2012

Spending Flexibility and Safe Withdrawal Rates

The new March 2012 issue of the Journal of Financial Planning is out. 

A couple of noteworthy items...

In the letters to the editor, Dick Purcell has a letter included discussing the article I wrote for the January 2012 issue.  That letter had its genesis as a comment at this blog! Making comments here is useful :)

Second, an article I wrote with Michael Finke and Duncan Williams of Texas Tech University appears in this issue.  It is "Spending Flexibility and Safe Withdrawal Rates." Traditional safe withdrawal rate research focuses on finding the highest withdrawal rate possible with a sufficiently small probability of failure. This ignores the tradeoff that lower withdrawal rates means less spending and less enjoyment in retirement. The more you spend, the more you can enjoy your early retirement, but the higher is your chance of running out of funds later in retirement. We try to balance this tradeoff by considering retirees who have different amounts of external income floors (such as Social Security) and different degrees of personal flexibility about how much spending fluctuations they are willing to endure.

The findings we describe are starting to get discussed at internet discussion boards and personal finance blog sites. This is primarily because Glenn Ruffenach wrote a column at SmartMoney juxtaposing the findings of this approach with the 1.8% withdrawal rate I described last August in the Journal of Financial Planning. The comments are along the lines of: these moronic researchers can't figure out if the safe withdrawal rate is 1.8% or 7%. A think a lot of confusion stems from this throwaway line in the introduction of Mr. Ruffenach's column: "a safe withdrawal rate for some individuals could be as much as 7%."

That 7% appears in Figure 3 of our article. But it is not correctly described as a safe withdrawal rate. As we add in this article, this 7% withdrawal rate has a 57% chance of failure over a 30-year retirement. It's not safe. But what is does is maximize the overall expected lifetime satisfaction for a fairly flexible retired couple who has a secured income base of $20,000 from Social Security. This is how the couple best balances spending more early in retirement with the tradeoff that they may have to spend less later in retirement.


Third, though not a part of the journal issue, the journal website now includes a Podcast interview with me. The topic is mostly about my article that will be in the April issue next month, but I also talk about this article and all of the articles I've had in the JFP.


  1. I wish I had the ability to write something of substance like others who comment here, but unfortunately I can't.

    What I can do however is express my gratitude for the efforts you put into your blog. I'm just a lay person (DIY investor living off my portfolio, not a financial planner) who has become a big fan of your work. Thanks for all you do to share your research.

    1. Thank you very much for the nice comment and for reading my blog!

  2. (moved from
    to here)

    Wade –-

    In answer to your question, I have a two-part answer: same spirit followed by step backward.

    I think your article in the new JFP lays out decision considerations very much in the spirit of Joe Tomlinson’s approach –- most fundamentally, your addressing of tradeoff between shortfall risk and withdrawal rate, with the analysis considering longevity probabilities, allocations, perhaps a guaranteed income source . . . I think that’s a very valuable contribution.

    Something about it that’s of special interest to me is that, while your investor example is from the wealthier minority that this whole field of study seems to be developed for, I think your article’s approach has applicability for the neglected majority of the population with less wealth. I think that’s worth a LOT!

    There is however one aspect of the article that troubles me: introduction and use of the so-called “coefficient of risk aversion.” I did not see any description of how that measure is to be determined for an investor, but I saw enough to set me aflame: what appears to be acceptance of that measure’s method of determination and use in portfolio theory.

    In my opinion, our duty is to do the very best we can to INFORM the investor of consequences of alternatives, for her choice –- inform her BEFORE her choice, so she can make an INFORMED CHOICE. I mean the consequences for which she invests, which she understands -– dollar purchasing power for her future needs and goals. This is INFORMED choice.

    The way it is done in portfolio theory is the opposite: Somehow conjuring the investor’s choice of a tradeoff between two misleadingly labeled technical measures she does not understand, before the consequences of her choice or alternatives are revealed or determined, enabling the mathematician to use her uninformed choice to pretend to calculate an optimum and its consequences. This is UNinformed choice.

    In the math at the end of the article, under the references, it appears that the “coefficient of risk aversion” is used as an input for calculation of consequences, which means the critical decision was somehow made before the consequences were revealed or calculated –- the UNinformed approach.

    So I like your article very much up to the point where the “coefficient of risk aversion” is introduced –- and very much dislike the latter part based on that factor. I would prefer that the latter part of the article be devoted to informing the investor of dollar-probability consequences of alternatives so she can make an informed choice.

    Dick Purcell

    1. Dick,

      Thanks as always.

      I want to respond partly to this by quoting from my review of Joe's article:

      "How to analyze the appropriate asset allocation and product allocation? There are two basic ways. Are innovations for the framework to evaluate retirement income strategies best guided by using various criteria and risk metrics (such as failure probability or value of bequests) which allow direct retiree decisions for evaluating tradeoffs related to spending and bequests, or should the framework specifically incorporate the formal mathematical concepts of utility maximization to help guide retiree choices, perhaps under the recognition that the human mind is not equipped for making decisions among abstract and uncertain tradeoffs?

      First let me give a brief answer to that. I think both approaches are useful and mutually reinforcing. Retirees should be free to choose as they wish to find the proper balance among competing tradeoffs. Most retirees will not be happy letting their decisions be guided by an indecipherable black box. They will want to feel they are empowered to make decisions. But with utility, their decisions can also be “double checked” to make sure they sufficiently understood the task of choosing among abstract and uncertain possibilities and found something matching a reasonable set of preferences."

      Beyond that, I think one of your key points is that you could appreciate Joe's article more because he took a lot more care than us to pinpoint proper values for his parameters, whereas we did not. Is that right?

      That's fair enough.

      I think, you might be happier if our article stopped at Figure 2.

      That figure tells a lot to help a retiree make an informed choice. It is not a probability of failure, but it is a percentage of retirement to expect having to spend without any remaining wealth. Combining this with the outside income floor, retirees can investigate different spending outcomes and the probabilities for each.

      What we do after that is just try to show how some stylized retirees might decide among these tradeoffs, to give some ballpark ranges. Retirees can compare their own decisions to what the utility analysis shows just to make sure they fully understand the problem and are getting something that sounds reasonable.

      Thank you for the comment that this can also be applied for folks with a lot less wealth. Just ratchet the nest-egg down to $100,000 or whatever, and the same sort of analysis can proceed.

      About that coefficient of risk aversion, I really would like to give it a new name for this context. It is: spending flexibility.

    2. Wade –-

      Thanks for your reply. Yes, I appreciate greatly that your article’s approach, unlike most of what’s done in this field, is applicable to the majority of The People who face retirement with less than $1 mil.

      On the matter in contention, I think you have put the spotlight on a question of greatest importance to the American public: should The People rely on economists’ guidance derived from the economists’ application of what they call “utility.”

      My answer is NO!

      It is not application of the CONCEPT of utility that I object to. I praised Joe Tomlinson’s application of utility, which in stunning contrast to prevailing economist malpractice was actually based on the investor’s utility -- dollar purchasing power for her future needs and goals.

      What I object to is economists’ use of some abstract measure NOT based on the victim’s utility, but instead conjured by the economists to enable completion of their pseudo-optimization math.

      Consider how “utility” (NOT!) is conjured in the portfolio theory of Markowitz and Sharpe, carried into contamination of current university investment “education” by their agent Bodie through his BKM textbook. It’s based NOT on the victim’s utility, which of course is dollar purchasing power for her future needs and goals. Instead it’s based on pretending to determine her tradeoff preference between two short-term technical measures she does not understand –- return-rate arithmetic mean and standard deviation –- presented under the disguise of the most deceptive pair of mis-labels in all of economics, “expected return” (which she should not expect) and “risk” (diverting her fear from her REAL risk of dollar shortfall for her future needs and goals).

      The poor victim has NO IDEA how this choice relates to her utilities of investment alternatives! NO IDEA what her choice will mean, or what other choices would mean, for her utility -- dollar purchasing power for her future needs and goals.

      This is NOT in the investor’s interests. It is ABANDONMENT of the victim's interests to boost the careers of the likes of Markowitz, Sharpe, and their agent Bodie.

      I don’t object at all to valid use of utility as a part of execution of the excellent purpose your article laid out. The way to do that is first calculate, and show the investor, probabilities for dollar purchasing power results of competitive allocation-and-withdrawal alternatives. Then use utility to mathematize the INFORMED preferences she expresses among these alternatives. This is the way Joe Tomlinson did it.

      Dick Purcell

    3. Dick,

      Thanks for your further explanations. I think Joe's article is great, and I do get that he did a better job justifying his choice of parameters, but I am still struggling with why you approve of his approach so much more than ours.

      I think your discussion of single period utility from MPT is a detour, because that is not what we are doing.

      Ours is based on a lifetime consideration: the utility from a lifetime spending path. And while we could have motivated this better, it can be informed by investor preferences. Essentially, we ask retirees to answer questions like:

      Consider that you will, on average, have an income of $60,000 for 80% of the rest of your life, and an income of $20,000 income for 20% of the rest of your life. What is the smallest amount of guaranteed income that you would be willing to receive for the rest of your life to eliminate the uncertainty and income fluctuations? If you are risk neutral, it is .8 x 60000 + .2 x 20000 = $52,000. But most people are risk averse and would be willing to accept less than this to eliminate the uncertainty and to eliminate the bad outcome. How much less? Well, the answer a person gives can be used to back out a risk aversion coefficient. And in our paper we are just looking at how the results work out for a couple different risk aversion coefficients.

      I don't see this is all that different from what Joe was doing, though he did explain it better and justified it with survey results.

      Just to be clear, those utility adjusted dollar amounts in his Table 2 and 3 are dollar amounts that come from his formula for utility (gain^alpha for positive values, with the further loss aversion adjustment for negative values)

      We also show utility-adjusted dollar amounts in our Figures 3-5.

      I do welcome your criticisms and comments, as ours was just a first step and needs much more work to be developed into a comprehensive retirement strategy evaluation scheme. I do want it to be useful, so I do want to understand your criticisms better.

      Thank you, Wade

  3. Wade –

    First and foremost, I want to re-emphasize that I think the framework your article presents is a most important advance beyond the prevailing focus on just failure rate. It offers paths to much fuller investor benefit from wealth-at-retirement. And unlike most approaches to retirement spending, it is suited for the neglected majority of folks who enter retirement with well under $1 mil.

    It is only on the matter of use of “utility,” in the form of investor “coefficient of risk aversion,” that I have trouble. On that matter:

    1. You’re right that I did not understand that in the article's approach, an investor’s coefficient of risk aversion is to be determined as described in your reply to me just above, and thought that instead it was to be determined as in modern portfolio theory (MPT), from which I think it is most important to be DISsociated.

    But I think the article presents grounds for that misunderstanding. Upon re-reading it, the article still seems to me to express respect for and imply application of the utility theory of MPT. At the very top, the article starts like this:

    Executive Summary

    • Shortfall risk retirement income analyses offer little insight into how much risk is optimal, and how risk tolerance affects retirement income decisions
    • This study models retirement income risk in a manner consistent with risk tolerance in portfolio selection in order to estimate optimal asset allocations and withdrawal rates for retirees with different risk attitudes

    Further on there are more expressions of respect for utility as used in portfolio theory and likening the article's approach to that of MPT.

    In the article there is discussion of presenting an investor an example of constant versus varying retirement incomes. But from the article I did not understand that each investor’s coefficient of risk aversion was to be determined that way. I thought that was just the authors talking to the readers.

    It’s my opinion that wherever “utility” or “risk aversion” is to be proposed for guiding individual investors, it is MOST important to DISsociate it from the way those terms and measures are used in prevailing portfolio theory. There they are used to divert the focus from the investor’s future purchasing-power needs and goals in favor of financial industry and academic interests.

    2. Even if I had understood that each investor’s “risk aversion” is to be assessed as described in your reply just above, I’d prefer more focus on trying to communicate the investor’s particular alternatives to the investor, in probabilities for HER dollar purchasing power. I view even the best utility approach as only a last-resort backup approach. One reason, not the only one, is that I don’t believe an investor’s utility is a smooth curve – that instead, each investor's utility curve has sharp-curve elbows, at places unique to that investor.

    My priority is: for communication to the investor, try everything. Try my scrollable probability-distribution graphs, of distributions representing HER dollar alternatives. If they don’t work, try the five-pictures approach, depicting HER dollar alternatives. For example, one set of pictures has three enjoying life on the yacht and two living in the cardboard box under the Sixth Street bridge. The other set has all five pictures of life in the trailer park. Investor, choose your picture set.

    I think our first obligation is to do everything we can to inform the investor, for her informed choice. Anything we do to derive utility from examples that don’t fit the investor’s specific case, and then use that utility inside the black box, is a sadly inferior approach.

    And there’s the danger that the academics love the utility approach, optimizing choices for mathematical curves instead of people. The "scholarly" literature is full of it.

    Dick Purcell

  4. Dick,

    Thanks for your detailed comments! This will help as we try to make further improvements and design a better system that meets individual needs!

  5. Wade –

    After further re-readings of your and Joe Tomlinson’s articles and your comments here on your blog amplifying what you have in mind for assessing an investor’s utility, I agree with you that the approach you intend in your article is very much in line with Joe’s.

    At the very end of Joe’s article, he calls for further work and better methods for determining utility for the particular investor. Clearly we three all agree with that. It’s the very opposite of the ignore-the-investor’s-future diversion into Greeky mathematical abstractions of MPT "utility” taught by Markowitz and Sharpe, with which the Bodie BKM textbook floods university investment “education.”

    I’d just urge that the whole focus, including utility curve development if necessary, be based on the particular investor’s financial future rather than some abstract examples dreamed up in some ivory tower. A person may have very different attitudes toward risk in different situations or scenarios – one person very risky in poker but risk-averse in bungee-jumping, another person very risky in bungee-jumping but risk-averse in poker. So I think it’s essential to focus on the particular investor ‘s actual financial outlook.

    If that’s done, for many investors we may not even have to retreat to the backup approach of “utility”!

    Dick Purcell

    1. Thank you Dick, and we will try to do this. I think the way to proceed is to think of risk aversion instead more as flexibility about changing spending levels. A risk averse individual translates into someone who wants to take very small the chance that their income could significantly drop even if it means giving up some of the upside, while a more risk accepting person would be willing to accept a small chance that income could fall significantly to have a better chance to spend greater amounts.

    2. Wade, I like your first sentence. But I fear the rest of what you said could lead some folks who don't understand your thinking to proceed in the opposite direction.

      I reject the notion that a person has a particular risk aversion that is the same for all situations. I think that when we classify people on a scale of risk aversion, we end up optimizing for mathematical curves instead of for people.

      I am saying a person's "risk aversion" preferences do or may DEPEND ON THE SPECIFIC SITUATION -- have kinks and elbows for particular situations, have different steepnesses for different situations. To serve people instead of mathematical curves, we should (a) show a person alternatives for her particular situation; (b) try to help her decide without resort to the utility notion, so the decision comes from her instead of from a black box; and (c) if (b) cannot be done, as a last resort try to determine her utility for that particular situation.

      I think we agree on this. But with the dominance of this field by high priests of optimizing for mathematical curves instead of people, we have to keep shouting our preference for serving people.

      Dick Purcell

  6. Wade -
    Thanks for your colleagues and your for this research. As someone in the real world who will be retiring in the very near future I appreciate practical research and advice. I understand the need for theory but how does that get applied? Maybe Scott Burns over simplifies these ideas but people in the real world want to know how to apply something to their specific situation.
    Thanks Matt Bly

    1. Matt,

      Thank you for your important question.

      It prompted my newest blog entry.

  7. So, is the next step for helping "real world" people lie in developing a questionnaire, possibly with an Excel calculation of specific numerical inputs, which will tell someone what number to use for their personal coefficient of risk aversion?

    And I think it is important to have some sort of "disclaimer" when using standard terms to express your non-mainstream framework. Dick often jumps on this, but OTOH you just cannot avoid using certain terms because they express something calculated. The important thing about your (Wade) articles is the different framework for interpreting & combining those outputs to express the "risk of running out of money."


    1. Thanks Bongleur!

      Yes, unlocking someone's personal coefficient of risk aversion is an important task.

      I've been thinking that in this context, "risk aversion" really just becomes "spending flexibility"

  8. What do you think about Bengen's new article where he takes the first 18 years returns from a 1969 retiree and grafts on the returns starting in 2000? He starts thinking about what happens if the deadly sequence of returns happens in the middle of retirement.

    1. Hi Bongleur,
      Thanks for stopping by.
      I'll tell you what I think. It is actually my new blog post for today:

  9. Nice article. What do you think about the utility trailers from toronto? I have had pretty good experience with them...