Life After Full-time Work Blog

Learn about preparing for life after full-time work through posts from Don's upcoming book.

#127 Decumulation: The State Of The Art Today

Five aspects of generating income for life


In January I took part in a virtual meeting of the Advisory Board of the World Pension Summit (WPS), thinking about possible topics for the projected 2021 WPS conference, usually held in October in The Hague. (Actually half the Advisory Board – the spread of members around the world requires two virtual meetings at widely different times of the day.) One topic that was among the top 8 in a pre-poll of members was decumulation, and I was asked to provide a bit of background before our discussion.

The members are a terrific group to be in touch with, as they bring expert perspectives from around the world (unlike most conferences, which tend to be national in scope), and so it’s possible not only to see issues from different perspectives, but also to extract the principles that define the topics when you eliminate the problems caused by each country’s regulations.

In this blog post I’ll use my introduction as the base to discuss today’s state of the art of decumulation issues and principles.


There’s a natural quote to start with. Nobel Prize winner Bill Sharpe called decumulation “the hardest, nastiest problem in finance.” It’s a combination of two questions. How much can you sustainably withdraw from your pension pot? And what’s a sensible asset allocation? Nobody has yet found the optimal solution.

Baby Boomers in the millions are retiring these days, with no satisfactory answer. I’m even older than them, being a World War 2 baby, so there certainly wasn’t any guidance for me when I retired.

So I’ve thought about this for many years. And I’ve reached the conclusion that there are basically four ways to generate sustainable income for life. None of them is superior to the others. They all have pros and cons. And of course you can use combinations of them. (See for more on this subject.)

  1. Buy an immediate lifetime income annuity.
  2. Draw down an amount each year that is based on your future longevity at the start of each year. (In the States this is the formula for the required minimum distribution or RMD.)
  3. Buy longevity insurance (a deferred annuity), and draw down an amount each year from your remaining assets that is designed to last until your longevity insurance kicks in. (This is simply not available in most countries.)
  4. Assume death at some advanced age, and draw down an amount each year that is designed to last until that advanced age.

As a matter of interest, the approach I use for my wife and me is the fourth one, and I wanted to use well-founded principles, not just something someone concocted recently. And I found those not-new principles from everything I’d learnt from defined benefit and collective defined contribution plans.

So that’s my introduction of our discussion of decumulation. I hope we can discuss, not the specific approaches I’ve mentioned, but more generally the view from 30,000 feet, about how important and how timely a topic decumulation is, and how suitable for a conference.


That was how I introduced the topic at the meeting. For this blog post, let me identify five fundamental sub-topics and tell you what the state of the art is today on each of them (and I’ll mention some names, in case the WPS organizers are looking for informed speakers):

  • Deal with longevity uncertainty
  • Compare longevity uncertainty with investment uncertainty
  • Desirable solutions differ across different individual retirement goals and situations
  • Would default decumulation options be helpful in pension plans?
  • Individual behavioral idiosyncrasies complicate the subject.


  1. Deal with longevity uncertainty

One member made the profound statement that we experience mortality individually, but we price it collectively. What this means, in practice, is that if we want to self-insure our longevity uncertainty, we have to budget for living a very long time, whereas if we pool our longevity uncertainty with others we need only pay for average longevity. And further, that longevity pooling is therefore the most efficient or least expensive way to pay for longevity risk. (For more on this subject, see T07 in Freedom, Time, Happiness (FTH for short): .)

In my list of four ways to decumulate, Numbers 1 and 3 are forms of longevity risk pooling. Discussions T05 through T09 in FTH discuss them in detail. A huge practical problem, though, is that deferred annuities are available in very few places in the world, so there’s no risk pooling that way. And immediate annuities guaranteed by life insurance companies typically are based on fixed interest rates, which are close to 0% now and likely to remain at that tiny level, meaning that many retirees would like to pool their longevity risk while still taking some investment growth risk.

The result is that there’s a search for ways to pool only longevity risk. And the jargon word that encapsulates longevity risk pooling is tontine – a pooling of assets that are paid out, along with investment earnings, to the survivors from time to time, without the (added cost) guarantee that an insurance company’s annuity provides. It’s possible to arrange tontine payments over time to increase as members die (the “mortality credit” to the survivors) or to anticipate those credits and attempt to even out the payments over time. In short, many forms are feasible, combined with many forms of investment. It’s too bad that they have a bad reputation because of corrupt practices by US insurers in the early 20th century – any instrument can be abused, but that doesn’t mean the instrument is itself bad. So I’m glad that there is today a lot of interest in exploring the possibility of establishing well-run tontines, though there’s inevitably caution about practical aspects.[i]

[STOP PRESS: I understand that QSuper, a major Australian superannuation fund, will shortly be announcing the availability of a lifetime income product that pools longevity risk without annuity-type guarantees. More details as soon as they’re available.]

Another approach being discussed today involves SeLFIES, a proposed new kind of bond (see which would pay the owner a real income for 20 years, with ownership retained by the owner and not pooled. While this doesn’t hedge longevity risk, which essentially remains self-insured, it’s a different approach to ensuring income for a specific (and, for many people, acceptably long) period in retirement.[ii]

  1. Compare longevity uncertainty with investment uncertainty

If we don’t know how long we’ll live, there’s uncertainty in knowing how much to set aside for retirement income that will last for life. And if we don’t know the return our investments will provide, that too creates uncertainty in knowing how much to set aside. Which form of uncertainty has the bigger financial significance?

In possibly the only original work I’ve ever done (!), I found that for a male aged 60, longevity uncertainty caused less of a financial problem than the uncertainty from being 100% invested in fixed income. As that hypothetical male grew older, the gap between the two forms of uncertainty narrowed and then reversed. For a 75-year-old male, longevity uncertainty created more financial uncertainty than being 100% invested in equities. Since we typically are unwilling to take that much investment risk, logic suggests that from age 75, males should be even more unwilling to take longevity risk, and should therefore eagerly seek longevity risk pooling. (See T04 in FTH for a more comprehensive discussion of this topic.)

This is a very important topic that hardly ever gets discussed. Perhaps one reason is that Bob Collie published an even better explanation,[iii] but Bob’s work too is rarely cited.

By the way, I know referring only to a male’s age is unhelpful. So let me add that the 75-year-old male’s situation also faces an 80-year-old female, and also a couple in which the male is aged 85 and the female 81. Those are ages after which (if you have average life expectancy as shown in the need for seeking longevity pooling becomes very great.

  1. Desirable solutions differ across different individual retirement goals and situations

People are different. Their goals are different, and their situations are different. So there’s no one-size-fits-all optimal solution for everyone.

Some people have bequest motives, others don’t. For those who do, there could be a specific asset or a specific amount in mind, or perhaps it’s “whatever is left after I’m gone.” Ways to take each of these into account in retirement planning can be found at Once that is dealt with, we can then focus on retirement income for life.

Is your pot (which may or may not include your home: see big enough to support your desired standard of living? Check which “wealth zone” you’re in (see because your choices could vary a lot.

  • You may have the luxury of seeking or not seeking growth, if you’re in the Bequest or Endowed Zone.
  • If you’re in the Lifestyle Zone, you have difficult choices to allocate your assets across growth-seeking and safety-oriented assets and an annuity. You might decide on Solution #4 above (assuming no deferred annuities are available) and divide your pot into growth and safety buckets. Or, if you’ve divided your desired lifestyle into essentials (what you feel you simply cannot live without) and nice-to-haves, you might decide that the essentials are so important that you lock them in via an immediate annuity; then the balance can be divided between growth-seeking and safety-seeking to avoid too much volatility in your nice-to-have consumption, but presumably much greater relative growth-seeking than if you hadn’t bought the annuity. This is the Zone with the most difficult choices.
  • If you’re in the Essentials Zone, this is where you have little choice and much hardship, as there isn’t enough even for your essentials.

A series of three recent papers that shed further light on all these choices is by Dr Geoff Warren and two co-authors.[iv]

  1. Would default decumulation options be helpful in pension plans?

Default options are often very useful, the idea being to nudge people into doing what they probably want to do anyway. The default acts a form of positive reinforcement of a sensible choice, without taking away the ability to decline the default option. Defaulting contrasts with other approaches like education, compulsion and neutrality. In fact, education by itself simply isn’t a viable stand-alone approach in the foreseeable future.

Pensions are familiar territory for default options, particularly since defined contribution plans became first popular and then dominant, requiring all sorts of choices from workers: whether to join the plan, how much to contribute, what asset allocation to use, and so on. By and large these defaults work very well.

In most ways decumulation is a much more difficult decision than joining a plan and choosing a contribution rate, and more difficult also than asset allocation during the accumulation stage, since decumulation requires not only asset allocation decisions but also decisions about how much can be sustainably withdrawn. From that perspective, then, default decumulation options would be extremely helpful, and those who feel sufficiently educated would have the practical ability to reject a default option and replace it with their own design.

The problem, of course, is that there is no accepted superior approach to decumulation. The four approaches mentioned at the start all have pros and cons; none dominates the others, so there’s no obvious stand-out. In addition, since we know (from the discussion of sub-topic #3) that desirable solutions differ as people’s situations and goals differ, that makes it even more difficult to identify the best default option.

Perhaps all we can do at this stage is to have many experts work on the problem, identifying the general shapes of situations and goals that align best with combinations of the different approaches. Then we can discuss the work, and see if we can start to agree on what might fit. My guess is that we’ll probably end up with a set of perhaps three different combinations as default options, each representing a sensible (I hesitate ever to use the word optimal with such a complex problem) approach to each of three different situations – rather than a single default option into which we’d like to place most prospective retirees.

  1. Individual behavioral idiosyncrasies complicate the subject

All of that ignores that fact that people have emotional reactions to many aspects of the different approaches. Some of these reactions are valid emotions, even if they seem to push us into worse expected outcomes than purely rational thinking would lead us into. Other reactions are just plain wrong. For example …

  • Most people underestimate their longevity. They may only remember what life expectancy at birth was, when they were born; or perhaps they’ve updated that number to current life expectancy at birth. But that’s just plain wrong (at any rate, for those in at least average health) because the expected age at death increases as one ages, since aging makes one a member of a more elite group of survivors.
  • Although an immediate annuity has been shown to be the optimal approach for a retiree with no bequest motive, most people prefer to retain some flexibility to adapt to changing economic conditions and changing life circumstances. The desire for flexibility is a natural and reasonable emotional goal that isn’t taken into account in the standard financial utility function.
  • Most people have a negative reaction to the word “annuity” but a much more positive one to the expression “guaranteed income for as long as you live.” This suggests that it isn’t just financial notions that any default options will need to encompass, it’s also the ways in which they’re expressed.


I remember that there are three stages in the evolution of ideas. First there’s the leading-edge stage, in which there are a few pioneers, and lots of innovation and testing. Then some variants of those ideas survive and become mainstream, and there are many practitioners involved. Eventually the idea becomes a commodity: it’s accepted wisdom, and everybody uses a common form. I mention this because today pretty much all angles regarding decumulation are in the first, or leading-edge, stage. It’ll be a long time before they’re commoditized.



Tough issues, not yet settled in any generally accepted form.


[i] Experts on tontines include Dr Moshe Milevsky and Richard Fullmer, if you’re looking for further reading.

[ii] SeLFIES are the brainchild of Nobel Prize winner Robert Merton and Dr Arun Muralidhar.

[iii] See An earlier version of this post contained an incomplete reference.

[iv] The one explaining the principles is found at


I have written about retirement planning before and some of that material also relates to topics or issues that are being discussed here. Where relevant I draw on material from three sources: The Retirement Plan Solution (co-authored with Bob Collie and Matt Smith, published by John Wiley & Sons, Inc., 2009), my foreword to Someday Rich (by Timothy Noonan and Matt Smith, also published by Wiley, 2012), and my occasional column The Art of Investment in the FT Money supplement of The Financial Times, published in the UK. I am grateful to the other authors and to The Financial Times for permission to use the material here.

12 Responses to “#127 Decumulation: The State Of The Art Today”

  1. Thomas Philips says:

    Annuitization is a very powerful idea, but has proven problematic as some insurers have made their portfolios more risky (particularly after being taken over by private equity managers), so that rewards accrue to the shareholders and managers, while risks are borne by annuity holders. An excellent discussion of both the strengths and weaknesses of annuitization can be found on pages 181-187 of Andrew Ang’s textbook “Asset Management – A Systematic Approach to Factor Investing”.

    A nation-wide approach to annuitization that is integrated with Social Security might prove both simpler and more effective in practice. Whenever a worker leaves a job or retires, she should be allowed to transfer some or all of the vested balance in her retirement savings plan to the Social Security Administration for the eventual purchase of an annuity. As Social Security is already set up to handle individual accounts, this is a simple feature to add, and the transferred amount can be placed in a separate account for the eventual benefit of the retiree and invested in accordance with a standard investment policy.

    Now citizens can obtain full pension portability and annuitization along with the benefits of longevity risk pooling, full transparency and minimal investment costs. Arun Muralidhar and I proposed this solution as part of a larger proposal for Social Security reform in our 2008 FAJ Paper “Saving Social Security: A Better Approach” (, and I continue to think that it addresses most of the problems associated with annuitization. In fact, our approach can be thought of as a nationwide tontine in which participation is voluntary, and even if only a modest fraction of the population participates, most of the benefits of longevity pooling will be realized.

    • Don Ezra says:

      Thanks very much, Tom, for both the idea and the reference. I hope some of the excellent ideas in circulation will lead to action. Meanwhile, not being involved in public policy discussions, I simply react rather than advocate. But there are times I wish the ideas can be brought to fruition, as I’d like to be a participant!

  2. Mike Clark says:

    Good day, Don

    As ever, you invite us to engage with the most interesting problems.

    Let me start here. This issue spans the ownership of risk by the public sector and private sector. The state has a balance sheet, and a Chief Risk Officer role (currently well hidden in some countries!) as well as its “bright future” CEO role.
    Let us ask the question: “Assuming the state is inclined to let the private sector manage as much risk as possible, leaving as little as possible for the state (via government) what would the state like the default to be?” The answer will clearly differ by country. US and UK have quite an individualistic culture (and thus political economy, I suggest), Japan and Sweden quite collective.

    So yes there is no universally economically dominant solution (and cannot be?), since we need to look at this through a risk preference lens. The decisions for an individual (and thus the default which may be nudged their way) depend on the risk preferences of a nation which in turn will influence the private sector offerings and how they integrate with the public sector underwriting of the residual risk. In the end, society (us individually and collectively) bears all risks!

    Some brief undeveloped thoughts:
    *You have omitted the fine idea you taught me many years ago, to use some capital to buy a call option on that deferred annuity. Still wise, yes, even if no product, yet?
    *The state has an interest in wealth creation. So too much annuitization, with “prudent” financial regulation that means insurers are herded towards bonds, suppresses risk taking. Not good for entrepreneurial activity. A weaker form visible in UK DB schemes? (“What is the optimal risk preference for a nation?” More prosaically “Taking the risk preference as given, how well is it being implemented?”
    *I have a strong preference for tontines to be open. Intergenerational. Collective DC? On the menu, of course, then some excellent governance issues. Not only for the board of the financial organisation, but for the policymakers who makes the rules in which it sits!

    Enough for now!

    Best, Mike

    • Don Ezra says:

      Thanks, Mike, your thoughts are always fascinating and deep. These days I focus on the individual much more than on advocating ways for society to change. And (as you say) different countries will probably move in different directions. So my focus is on principles that enable individuals to cope better, given the rules of their society. Hence the importance of work like the linked one by Geoff Warren. As for tontines, my guess is that they will indeed be intergenerational (open pools), unless something suddenly happens to cause a sudden seismic shift in reasonable expectations. Finally … my goodness, you have a long memory! Yes, that notion of an option to purchase a deferred annuity was one I developed when I was much younger and was thinking about first principles, having just turned 60 and wanting to develop the best principles I could, to see how best I could cope with eventual decumulation. Of course these options aren’t available. But my (at the time) unconstrained thinking dreamt up the option, because (a) it locked in current interest rates, in case they declined over time, (b) it locked in current longevity, in case it grew, and (c) if my own health started to decline, I would only be out the cost of an option rather than the deferred annuity. So yes, still desirable, in principle!

      • Mike Clark says:

        Thanks for the reply Don. Yes, I am still at the “changing society” stage. Must send you soon a piece I have just written under the Actuaries for Transformational Change (ATC) banner.

        Distinction understood. Of course I could gently suggest that part of an individual making the best of things as they are is to help improve them for their children and grandchildren when they have gone. That probably includes spending some effort/resources on tweaking those rules!

        Best, Mike

        • Don Ezra says:

          Thanks, Mike, and yes, you’re absolutely right — despite my intense focus (which you’ll recognize as the way I work, after all our years of working together) it’s a human imperative to make things better for our children and grandchildren — thanks for making that point so gently!

  3. I have been reading your blogs with interest, being involved in investment management for over 40 years. Would you be interested to discuss how to organize the decumulation phase for informal workers in developing countries?
    We have 45000 customers now in Ghana. Life expectancy after age 60 is 16 years. For now I structured a 20-year annuity to leave room for innovative solution to cover longevity risk.


    • Don Ezra says:

      Thanks very much for this information. I’m happy to discuss this; let’s use my email rather than the website. Meanwhile, you’re using the fourth approach I listed above, which makes a lot of sense. Out of curiosity I checked, and derived some probabilities. For a 16-year life expectancy, lengthening the planning horizon to 20 years increases the probability of not outliving the horizon to roughly 67%; it increases to 75% at 21 years and to 90% at 26 years, for what that information is worth.

  4. Ted Harris says:

    Personal genome sequencing can add further clarity. The cost is now below $1000. While adding information to a decumulation plan, particularly as to morbidity and mortality, there are still many uncertainties because of the overwhelming number of permutations and combinations, as well as epigenetic influences and the likelihood, or not, of a characteristic being “expressed”. While this information can be useful to an individual, it could result in “reverse” adverse selection in a group arrangement.

    As the cost of identifying and analyzing vast quantities of information continues to drop, we will know more and more about our personal “Blueprint” (book by Robert Plomin), but there are so many other external variables – so we can never develop a precise answer. That is also true of the assets we use to fund our plan.

    We continue to engage in risk reduction, but risk elimination is forever elusive. Our industry remains intact!

    • Don Ezra says:

      Thanks, Ted, love it! The concept I’m more familiar with is “biological age” which Moshe Milevsky introduced me to. That would be a better input into longevity calculations and the planning horizon than is chronological age, which (Moshe’s book reminds us) simply counts the number of times we’ve been around the sun. Now I have to get “Blueprint” — thanks for the reference.

  5. Thomas Philips says:

    I just learned that Alicia Munnell published a simple (and useful) decumulation rule 10 years ago: Spend your dividend and interest income, as well as your Required Minimum Distribution (RMD) as specified by the IRS each year. The IRS uses life expectancy when determining RMDs and so does all the math for you! You can read her paper at While the strategy is clearly U.S. centric, it’s a surprisingly simple and effective solution for people who don’t want to bother with the complexities of decumulation.

    [Added subsequently] It’s a good thing that you moderate your comments, because it wasn’t Alicia Munnell, but Wei Sun and Anthony Webb who created this rule! Apologies for the error – would you kindly correct my post. I believe their paper later appeared in a collection edited by Alicia Munnell, but credit for the idea should go to Sun and Webb.


    • Don Ezra says:

      Thanks very much, Tom — particularly as I hadn’t seen the original article to which you provide a link. This is a variant on approach #2 in my blog post. What’s interesting is that in Post #33 I mentioned that approach #2 has a tendency to generate sharply declining withdrawals near the end of the longevity table. So the Sun/Webb approach (which, their article suggests, was developed to overcome a too-low early withdrawal rate if only the RMD is used) would then generate an even sharper decline near the end of the longevity table. For most people my guess is that this won’t matter; but it does suggest to me that, as time passes, if the person or couple involved is still in good health, they might consider changing to something along the lines of approach #4.

Leave your question or comment