Sunday, January 15, 2012

Safe Withdrawal Rates: Have I been barking up the wrong tree?

As a percentage of retirement date assets, what is the highest amount you can withdraw, while adjusting this amount for inflation in subsequent years, without running out of funds for a sufficiently long period of time?
This is a question plaguing retirees and near-retirees as lifespans and retirements lengthen, and as traditional defined-benefit pensions are falling by the wayside. Since William Bengen’s seminal investigation of the US historical data in his 1994 Journal of Financial Planning article, answers have tended to center around 4%, with perhaps 4-5% being a comfortable range for many retirees and planners.
For retirement planning research, I paid my dues by looking at three different methods (international data comparisons, speed of wealth depletion in the years since retirement, and impact of market valuation and yield measures at the retirement date) which each question whether recent retirees can still consider 4% to be safe over a 30-year horizon despite its historical success for retirements beginning up to 1980.
The question of safe withdrawal rates certainly does matter. My title is overly provocative. A 2011 Financial Planning Association survey described by Jonathan Guyton indicates that 75% of surveyed financial planners either always or frequently use systematic withdrawals with their clients. For them, the safe withdrawal rate is relevant. But my point is that safe withdrawal rates may be less important than I earlier thought.
That’s because for a lot of people involved in the retirement income debate, even when using a well-diversified portfolio of stocks and bonds (which is the recommendation in the safe withdrawal rate literature), the only safe withdrawal rate is 0%. We are deluding ourselves to think otherwise.
An alternative approach to retirement income planning that is gaining traction is the “guaranteed floor / upside potential” approach. With this approach, you first build a floor of very low-risk guaranteed income sources to serve your basic spending needs in retirement. The guaranteed income floor is built with Social Security and any other defined-benefit pensions, and through the use of your financial assets to do things such as building a ladder of TIPS or purchasing single-premium immediate annuities (SPIAs). GLWBs could also play a role here. Not all of these income sources are inflation adjusted, and you do need to make sure your floor is sufficiently protected from inflation, but this is the basic idea.
According to the Retirement Income Industry Association (disclosure: I am on their Academic Advisory Board), a fundamental goal of retirement planning is to “first build a floor, then expose to upside.” This is also the approach Moshe Milevsky has in mind when he recommends that you “pensionize your nest egg” and it is the way that Zvi Bodie suggests you can “risk less and prosper” in retirement.
Once you have a sufficient floor in place, you can focus on upside potential. With any remaining assets, you can invest and spend as you wish. Since this extra spending (such as for nice restaurants, extra vacations, etc.) is discretionary, it won’t be the end of the world if you must stop spending at some point. You still have your guaranteed income floor in place to meet your basic needs no matter what happens. With this sort of approach, withdrawal rates hardly matter. (Another note from the Jonathan Guyton article linked above: he suggests that retirees may not be as blasé as I’ve just made it appear about their abilities to fund even their ‘discretionary’ expenses).
“Safe withdrawal rates” and “guaranteed floor / upside potential” are really two competing approaches to retirement income planning. How can we reconcile them?
I think they can be reconciled by broadening some aspects of safe withdrawal rate studies. First, these studies ask the question about how much can be withdrawn over time from a risky portfolio, and so they do not directly incorporate other income sources such as Social Security, annuities, or pensions. Second, these studies focus on the probability of failure (also called shortfall risk, it is the probability of running out of wealth while still alive), without giving consideration about what is lost in terms of life satisfaction by using a lower withdrawal rate and spending less. The fact that with low withdrawal rates, people will typically leave behind a large pot of wealth (unless their retirement matches the worst-case scenario) is not something included in the analysis.
I have been part of a research effort with Michael Finke and Duncan Williams of Texas Tech University to incorporate the sustainable spending “safe withdrawal rate” framework into the “guaranteed floor / upside potential” framework. We look at safe withdrawal rates after adding other income sources from outside the retirement portfolio (such as Social Security and pensions). We also leave behind the safe withdrawal rate objective of worrying only about a low failure rate (low shortfall risk), and instead try to balance the competing tradeoffs between wanting to spend and enjoy more while one is still alive and healthy, against not wanting to run out of portfolio wealth while still alive (the guaranteed floor will always be in place though). We find that someone with flexibility about how much they can spend and with more outside sources of income may be willing to accept rather high failure rates as a part of balancing these competing tradeoffs.* Life is indeed a balancing act. Our study, “Spending Flexibility and Safe Withdrawal Rates,” can be downloaded now as a working paper, and it has been accepted for publication in an upcoming issue of Journal of Financial Planning.
Moshe Milevsky and Huaxiong Huong earlier produced a related work, and what we think of as “spending flexibility,” they call “longevity risk aversion.” Joseph Tomlinson will also have a paper exploring these sorts of concepts in an upcoming issue of Journal of Financial Planning.
I do have an article on safe withdrawal rates in the current January issue of Journal of Financial Planning. I generalize the framework for safe withdrawal rates to include any assets or assumptions, and it think it can really serve as a final word from me about traditional safe withdrawal rate studies (except, I still want to look at variable withdrawal rates).
But the next generation of studies with a more comprehensive view toward withdrawal rates and retirement income is already on the way.
* This finding puts me in the rather odd position of having first recommended rather low withdrawal rates in earlier research, but now suggesting that much higher withdrawal rates (and chances for failure) may be acceptable in certain circumstances. If you are confused about this, Doug Nordman wrote a nice summary of these two competing viewpoints called, “Is the 4% withdrawal rate really safe?”

Update: just to clarify, when I talk about "safe withdrawal rates," I refer to the classical approach to safe withdrawal rates using a diversified portfolio of stocks and bonds. William Bengen and the Trinity study suggest stock allocations in the neighborhood of 50-75%. So it's a matter of how much can be safely withdrawn from a volatile portfolio. The "floor/upside" approach rejects this sort of view about the matter: volatile portfolios are inherently not safe.


  1. Wade, this sounds like further good thinking in your work. But it also sounds to me to be getting worse in a problem this field already suffers: Doing better and better for the better off, while leaving the rest of the people further behind.

    (I hasten to add that I am not criticizing YOU on this matter. I know from prior discussion that you share this concern.)

    But –- what percent of the people have enough to meet their “needs” with only a fraction of their funds invested at 0% or near-0% return?

    In a response to a recent Michael Kitces blog about “the rest of the people,” David Jacobs estimated that the financial planner market is maybe the wealthier 20%.

    What about all the rest of the people??

    I don’t mean to be criticizing the good work underway that you have outlined in this current blog. But I think that as a nation we have a far bigger, neglected retirement planning problem of inadequate guidance for the rest of the people. It threatens the futures of lots of people –- and enough people that it is a serious problem for the whole economy, potential danger for ALL of us.

    Dick Purcell

  2. Dick,

    Thanks for sharing. Yes, this post is written from the perspective for whom the safe withdrawal rate question is even relevant. For altogether too many people, the entire topic of safe withdrawal rates will not be a realistic part of their life. They just haven't saved enough to be where they want to be when withdrawing 4% of their savings.

    I think we discussed this some at Bogleheads, and I'm not sure what the best general advice would be for people who are falling well short of their retirement goals, but for whom either saving more or delaying retirement are not viable options. They could either make a Hail Mary pass with a stock-heavy portfolio, or they could lock-in what they have with low-risk investments or annuities.

    The income floor / upside approach would suggest that those way behind where they want to be would look more at annuities as an option. Probably that would be the wise thing to do. Isn't this also your view?

    For people further away from retirement, I do agree that the "risk less and prosper" approach is going to be very tough for many people. I'm working on a book review now about the approach.

    At any rate, thank you indeed for the reminder that retirement planning research must focus on meeting the needs of everyone, and not just the well to do.

    1. Hi Wade,
      My way of applying the floor/upside idea is to match retirement spending liabilities (rock bottom essentials, nice to have, somewhat deferrable -like house repairs) with allocations to income assets that have similar risk and cash flow timing properties. That fits in the range of assets from social security (in Canada CPP/OAS/GIS), through LTC insurance, stocks, bonds, reverse mortgages etc.

  3. Hi Wade,
    Interesting blog post on trying to reconcile "safe withdrawal rates" versus ""guaranteed floor/upside potential." I'm hopeful that utility analysis like you are doing with Messrs. Finke and Williams and perhaps my own work will help with this reconciliation. In my article for the February JFP, I've included an income annuity along with systematic withdrawal using regular investments as choices, and demonstrate where one approach is favored over the other based on utility measurements. So perhaps that's working toward a reconciliation.
    One thing that surprised me in my research was how badly a portfolio of TIPS performed in a utility sense when variable mortality was introduced. I need to read Zvi Bodie's new book, but I wonder how he deals with longevity risk.
    I had some involvement with the FPA retirement survey that Jonathan Guyton writes about. I recall looking at crosstab results that showed planners who had a larger percentage of clients with a significant chance of running out of money had a greater inclination to use the "guaranteed floor" approach.
    I can understand where Dick Purcell is coming from with his comment. We do need to find ways to deliver better products and planning services to people of average means. Unfortunately, all the incentives for those in financial services point to serving upscale markets.
    This is a really good blog. I hope it becomes more of a forum for discussion.
    Joe Tomlinson

  4. Canadian Investor: I was trying to reply to you earlier, but something was wrong with the blog. Thank you for some additional details about how to make an income floor. I hadn't mentioned much about that. Matching assets to liabilities is an important part of the story. Thank you.

  5. Joe: Thank you for stopping by. I just posted my new blog about GLWBs. I'm glad to see your Advisor Perspectives article is out now.

    In Zvi Bodie's new book, he does emphasize annuities more. I do very much agree that building a TIPS ladder for the guaranteed floor with a specific end date in mind for spending can be a big problem if you live longer than that. He has backed away from calling TIPS "safe and worry-free" in the new book and is more clear about some of these points.

    Thanks for the nice comment about my blog. Michael Kitces is doing a great job of generating discussions on his blog, but I think this is generally harder for blogs because most people will read it through an email notification or RSS feed and them don't know about any interesting comments that are made. But there are online discussion boards such as the Bogleheads Forum ( with very large readership that can be great places to discuss these issues. I've chatted a lot with Dick there.

    From there, I know that Dick does NOT like utility maximization. We had an earlier discussion, but it was at a stage where I only had read Finke and Williams earlier work, and that work was not yet public, so I could only say something like "there is some great new work using utility analysis to come up with some very interesting insights about the retirement income process." I was constrained about being specific.

    Now that your paper will be out in a few weeks, and Finke, Williams and I have had another look at it with the article I linked in my post, it may be time to revisit the issue in a discussion. Dick, I hope we can persuade you that utility maximization is not just physics envy, but can help provide some real world dollars and cents insights for real people.

    Joe, I don't want to get you addicted, but a really good thread at Bogleheads which evolved into a discussion about utility by page 2 or 3 of the comments is "Redefining Risk"

  6. Wade,

    Another great post. Thanks. We have created a Retirement Income Cash flow system (RICS) that insulates the first 2 years of living expenses (cash or other near term investments). The next 5 years (years 3 -7) are covered by a bond ladder or a diversified intermediate term bond portfolio. Anything over year 7 is invested for "growth" (purchasing power protection). We target an overall mix of 60% stock/40% bond but it can vary depending on the clients risk tolerance, etc. We also closely monitor the safe withdrawal rate and institute Guyton's cuts/freezes/raises methodology. Lastly, we simulate and stress test the portfolio using our financial planning software (Monte Carlo simulations).

    By insulating the first 7 years of cash flow, we can withstand upwards of a 7 year bear market without having to withdraw from the growth portion of the portfolio. Much of this is modeled after Evanky & Katz's cash flow system outlined in "Retirement Income Redesigned" and combining other retirement income thought leadership (Kitces, Guyton, etc).

    I look forward to more research on using annuities to guarantee a retiree's "floor." There certainly is no perfect system and no way to issue a guarantee. Glad to see you are working on this. We -as planners - need it, as does the public.

    Finally, I agree with Dick but, unfortunately, many people forgave saving for the future in lieu of living for the present. They are now reaping what they have sowed. That comment certainly doesn't apply to all people however, I believe it should be our mission, as planners, to help people realize the long-term ramifications of their current spending habits. We need our clients to save as much as possible so they can be in the "20%." The folks who are now able to work with a planner didn't get their by accident.

    Thanks for allowing me to share my thoughts. I very much appreciate your blog and your research.

    Brad Owen

    1. Brad,

      Thank you very much for sharing your system. It sounds quite reasonable. I really want to have a look at variable/dynamic withdrawal rates soon, along the lines of Guyton's decision rules and Bengen's ceiling/floor approach. Thus far, I've only looked at "set it and forget it" approaches, mostly constant inflation-adjusted withdrawals.

      I'm just curious, the FPA survey of planners discusses three strategies:

      1)systematic withdrawals
      2)time diversification

      Your asset allocation approach certainly seems to be more along the lines of time diversification, though you are treating it as an asset allocation approach to feed into dynamic systematic withdrawals. Would you call it (1) or (2), or both (1) and (2)? Does this distinction really have any meaning?

      Again, thank you for sharing. Since I am not a planner myself, it certainly helps to know the approaches/interests of planners to help guide what I should be simulating in research.

    2. Wade,

      I'd call it a combination of (1) and (2) (and maybe even (3)!). We use 3 "time diversification" buckets but pay attention to the overall asset allocation when those 3 buckets are combined. We've tried to combine the thought leadership of 3 or 4 different systems in to our approach.

      From a planners perspective, I can say that our clients like the time diversification approach. Behaviorally, it allows them to see where their income will be coming from each year (at least for the first 7) and also allows us to take on more risk with the growth bucket b/c they know they won't need to tap in to it for 7 years. Lots of nice behavioral benefits there.

      Thanks again for all your leadership in this area. Very much needed!


    3. Thanks a lot Brad!

      It really helps me to see your approach. This year I want to focus much more on simulating different retirement income strategies, including time diversification and Guyton's decision rules and so on. Have you seen Stephen Huxley and J. Brent Burns' Asset Dedication approach to time diversification?

      I liked that book a lot.

    4. Paul Samuelson has to be rolling over in his grave. The so-called bucket approach, is it not another manifestation of the myth of time diversification? is the clearest explanation I have encountered. If you think the risk of holding a risky asset declines with time you are simply a victim of Siegel and Bogle most likely. Sadly you are also wrong. Risk and Uncertainty: A Fallacy of Large Numbers is Samuelson's original paper. Cassandra

    5. I share your concern, thank you. The US historical data is too short and too exceptional to have a good idea about the success of different strategies.

  7. I think the whole 4%/SAFEMAX research is good from a historic and academic point of view but I still feel uncomfortable that it does not take into account the current markets better. As you mentioned before, we wont know if your 1.8% SAFEMAX estimate for 2010 is true until 2040! Another issue is that I doubt any retiree would be disciplined enough to have a million or more in the 2-4 investments that most modeling has (bond index, S&P 500, maybe small cap/foreign). Everyone pretty much has a mix of several fund classes and sources and thus their return and alpha/beta will be quite different. I would like to see research move in the direction of: If your portfolio is getting x% return this year and y% the year before, what is the safe amount to withdraw? I think SAFEMAX has to be localized to the retirees actual returns and based on current market conditions more.

    Also, while I see more recommendations for income annuities, the issue I see with them planning wise is how then do you accommodate for inflation? If you needed $100 of income this year and you get that all from a fixed income annuity, the next inflation raise of say 3% ($3) has to come from investments. Then the next year that 3% amounts to $6.09 from investments. The compounded growth rate of that contribution has never been modeled before. At the end of 20 years or so you need about $200 ($100 annuity + $100 inflation) of which, your $3 starting withdrawal from investments has grown 33 times to $100 (vs just 2x overall). The whole SAFEMAX approach has a hole when you add any fixed income annuity AND want to still match inflation for the whole amount.

    1. Thad, thank you. Yes, this year I am hoping to look more at variable strategies. Not just the fixed inflation-adjusted withdrawals I've been looking at thus far.

      About your point on annuities, it is important, but not insurmountable.

      On this issue, you may like to see:

      John Ameriks, Robert Veres, and Mark J. Warshawsky's December 2001 article from the Journal of Financial Planning, "Making Retirement Income Last a Lifetime".

      I blogged about it here:

      The link I had for the article is not working now.

      But it looks at the case where one annuitizes half of their assets with an non-inflation-adjusted annuity and then uses withdrawals to maintain the same total inflation-adjusted amount. They find favorable results for partial annuitization to improve the overall success rate for 4.5% inflation adjusted total withdrawals.

  8. Very interesting discussion. One of the assets you mention for constructing the "floor" is a ladder of TIPS. A question about that:

    What do you do when the TIPS you are holding increase significantly in value, as they would have done in 2011 for no particular reason (certainly not inflection)? Do you continue to hold them till maturity, or do you do something to harvest your good fortune?

    Charlie B

    1. In the above, I typed "inflation" but some mysterious force in my computer changed it to "inflection".

    2. Charlie,

      Let me preface this with: I am not a financial planner, so this is just my opinion, and also you may have some unique circumstances which call for another answer...

      But generally, I would advice against trying to harvest gains from your bond ladder.

      You must have created a TIPS bond ladder because you wanted to safely protect a floor of spending.

      By selling it off and not keeping it until maturity, you expose yourself to risk. If you sell and interest rates go up, then you will benefit from the move. But if you sell and interest rates go down, then you will lose from the move. That's because, you still want to protect your income floor, and rebuying the bond ladder will become even more expensive and you will be ensured a reduction to your future spending.

      I think TIPS yields are just as likely to continue going down, than to go up. That's not a prediction, I'm just saying I wouldn't be surprised if TIPS yields continue decreasing. I wrote about this in a column at Advisor Perspectives:

      When you talk about harvesting gains, you still have to buy something. TIPS have gained, but likewise any other fixed income product that you might switch to will also be rather expensive, so you don't get any particular benefit from the move.

      You are just adding risk when you don't need to.

      You were specifically talking about a TIPS ladder. If you own a TIPS mutual fund and it has grown to represent a higher percentage of your assets than you desired to have, then rebalancing some by selling some of the TIPS fund is no problem. But I don't think you were talking about this.

      Best wishes, Wade

    3. I am not fully versed on your 1.8% SWR prediction for 2010 retirees but would like to understand what led you to this projection;

      And since valuations have only gone up since then, and interest rates have gone down, is your prediction for the 2011 and 2012 retiree even more dire than 1.8%...?

    4. Newt, I just provided my updated views about this here:

      Thanks for the comment.

  9. Ialways appreciate the post which is full of its swing color.
    pension loan
    pension release