Monday, February 12, 2018

Unraveling Retirement Strategies: Constant-Dollar Spending (4% Rule)

Sustainable Withdrawal Rate (SWR) strategies are based on the work of William Bengen[1], whose research uncovered sequence of returns risk. The basis of the strategy is that there is a constant amount of spending from a stock and bond portfolio that would have been “safe” in 95% of historical 30-year periods of stock returns.

SWR had a colorful beginning. Peter Lynch of Magellan Fund fame posited that a retiree should be able to invest in stocks and spend about 7% of her portfolio forever. Scott Burns quickly showed that a 7% withdrawal rate was far riskier then Lynch imagined.[2]. (The culprit, as Bengen would show, is sequence of returns risk. The order of market losses is more important than the average return.)

Lynch’s response was, “OK, but surely there is some percentage that would work?”

Bengen found that the safe withdrawal percentage rate in the U.S. was historically 4% to 4.5% over rolling 30-year periods of market returns, hence the “4% rule."

Wade Pfau later showed that number only worked in the U.S. and a handful of other countries[5]. More recent work by Pfau suggests that the number at present is perhaps 3% to 3.5% — a sizable range. A range of 3% to 4.5% may sound small but it’s the difference between a safe spending amount of $30,000 and $45,000 a year per $1M of savings. Regardless, it’s well below Lynch’s 7% assertion.

The basic process for implementing this strategy is for a retiree to calculate 4% (or 3% or 4.5%, depending upon who you choose to believe — I'd go with Pfau) of her total investment portfolio value the day she retires and to spend that dollar amount (not percentage) for the rest of retirement, increasing it annually by inflation.

A retiree with a million dollar portfolio who agrees with Pfau's 3.5% could spend $35,000 the first year of retirement. If inflation ran 2% that year, he could spend $35,700 the following year.

This strategy will result in two possible outcomes when simulated, though actual retirees might not behave this way. The strategy will produce a constant, inflation-adjusted income amount from the portfolio until the retiree dies or the portfolio is depleted, whichever comes first. SWR practitioners try to minimize the latter outcome to “only 5% or 10%” of retirements, or 1-in-10 to 1-in-20.

With all due respect to Bengen, whose research exposed sequence risk — an important contribution — I consider this strategy irrational and believe that it probably makes sense only for households with so much savings that they don’t need it. The idea that we can spend an amount calculated at the beginning of retirement and continue spending it regardless of what happens to our financial situation over perhaps 30 years is not only risky but irrational.

Constant-dollar spending strategies are risky for households that aren't wealthy and unnecessary for those that are.
[Tweet this]

As retirement progresses, our finances and life expectancy will change, our finances for better and for worse. Life expectancy constantly declines. Constant-dollar spending is a strategy to ignore any of this new information after the original spending amount is calculated.

Larry Frank, David Blanchett and John Mitchell published research[3] showing that our spending should be “dynamically” adjusted as our finances change over time but you probably don’t need to read the research to understand this.

Instead, have this conversation with your 75-year old wife. “How much can we spend this year, Walter?”


“Thanks, but how do you know that, Dear?”

“It’s the amount we could safely spend a decade ago based on our life expectancy and wealth back then.”

Let me know how that conversation works for you.

Wade Pfau and Jeremy Cooper wrote a great explanation of the strategy (and others) for Challenger Ltd[4]. Granted, mine is more cynical, but they’re decades from retirement and I’m a decade into it. I have more skin in the game.

Here are some considerations I have learned by running a gazillion retirement simulations that you probably won’t find elsewhere without a lot of digging.

SWR strategies are based on the assumption that future market returns will be similar to historical returns but there is a lot of reason to believe they will be lower in the future. Even if they are similar, we don’t know the expected market return, whether it changes over time or even the distribution of those returns. We pretend they are normally distributed, largely because it makes the math easier, but empirical evidence shows that there are far more extreme market events than a normal distribution would predict. That’s a lot of uncertainty on which to bet one’s retirement.

Many retirees who like this strategy believe that they are creating their own annuity without committing a large sum of money to an insurance company. There is a specious similarity to an annuity in that both provide constant income but SWR is like an annuity from an insurance company with a 5% to 10% chance of going out of business.

Further, annuities provide maximum lifetime consumption while SWR strategies are economically inefficient. It’s expensive to tie up 96% of your wealth so you can safely spend 4% of it[6].

Self-annuitizers may expect that have more liquidity with a SWR strategy than they actually have. In the same way I explained that the liquidity of TIPS ladders can be illusory, spending from an SWR portfolio to meet large, unanticipated expenses means spending income you were counting on for the future. Certainly TIPS ladders and investment portfolios are more liquid than annuities but there is a price to pay in the future for spending them early.

A high spending percentage increases the probability of outliving the portfolio and reduces the expected terminal portfolio value, which is often part of a planned bequest. At the other extreme, a low spending percentage decreases the odds of portfolio depletion but increases the expected terminal portfolio value. This has some interesting implications. The following charts show the relationship.

Retirees with limited wealth relative to their spending needs may need to spend a larger percentage and accept a greater portability of outliving their savings. They are less likely to accumulate a large terminal portfolio.

Retirees with adequate savings can play it safe with a low withdrawal rate but they are more likely to die with a large portfolio. This is a good thing for households with a strong bequest motive but not so good for those without. Without a bequest motive, a retiree who plays it safe with a low withdrawal rate may find late in retirement that she has an extra million bucks she could have spent to enhance her life.

This may be ideal for a household with so much wealth that they can easily afford their desired standard of living and still expect a large portfolio to leave to heirs but then one must ask why a household this wealthy needs a SWR strategy, at all.

As I explained in Retirement Income and Chaos Theory, constant-dollar spending strategies are probably chaotic when the portfolio is sufficiently stressed. It’s a bit like hanging around the event horizon of a black hole in that scenario where just a little bit of bad luck can nudge your strained household finances into an irreversible downward spiral.

In a post entitled, "Time to Retire the Probability of Ruin" back in April 2015, I wrote that we should stop basing retirement plans on this metric. A short time later, Moshe Milevsky wrote a better piece entitled, "It’s Time to Retire Ruin (Probabilities)"[7].

Constant-dollar spending strategies are quite unpopular among economists and most of the researchers I know. They are sometimes popular among planners who charge fees based on assets under management and retirees who see them as an annuity alternative.

I am none of those and not a proponent of constant-dollar spending strategies. I have many reasons, but the most basic are that it is irrational to ignore new information that comes available as retirement progresses and financially unsound to attempt to derive constant income from a volatile portfolio. The strategy is risky for retirees who are not wealthy and unnecessary for those who are.

Ultimately, it isn't possible to re-create an annuity with a one- or two-person risk pool. Rational strategies to spend from a volatile portfolio will suggest that we spend more when our portfolio grows in proportion to our spending and life expectancy and demand that we spend less when it shrinks. That requires a variable spending rule strategy, which I will address next.


[1] Conserving Client Portfolios During Retirement, William Bengen.

[2] Dangerous Advice from Peter Lynch, Scott Burns.

[3] An Age-Based, Three-Dimensional Distribution Model Incorporating Sequence and Longevity Risks,  Frank, Mitchell, and Blanchett.

[4] The Yin and Yang of Retirement Income Philosophies, Pfau, Cooper.

[5] Journal The 4 Percent Rule Is Not Safe in a Low-Yield World, Wade Pfau.

[6] The 4% Rule - At What Price?, Scott, Sharpe, and Watson.

[7] It’s Time to Retire Ruin (Probabilities), Moshe Milevsky.


  1. good post. i hope it gets the wide readership it deserves.

    i came across an example somewhere [unfortunately i can't remember where] that illustrated the silliness of swr.

    A and B are the same age and each have $1million saved. A retires this year, B decides to wait a year.

    year 1 starts with A withdrawing $40k. B withdraws nothing and adds nothing.

    the market goes down 20%.

    at the start of year 2, A's portfolio is 0.8* 960k = 768k.
    at that same moment B's portfolio is 0.8* 1000k = 800k.

    in year 2, using a 4% swr, A withdraws an inflation adjusted 40k. meanwhile B has a bigger portfolio, but following the same rule can only withdraw 32k.

    every year thereafter B will have a bigger portfolio than A, but is only "allowed" a smaller withdrawal.

    clearly there is something wrong with this "rule."

    1. Yes, what's wrong with the rule is that it ignores the new information available, in your example, at the beginning of year two.

      Thanks for the comment.

  2. Hi Dirk -

    Do you have a suggestive, rule-of-thumn math function to determine portfolio withdrawal rates for a typical person? I.e.

    $ withdrawn this year = F( portfolio_performance_this_year, living_needs, current_age, probability_of_death, risk_tolerance, extra_info)

    Or do you have a step by step procedure to determine this function in mathematical terms?

    1. Hi, Victor. Thanks for writing.

      Great question but with a longish answer. This question pertains more to variable spending strategies than constant-dollar spending. I'll write about those soon. The difference is that my answer below refers to a spending rate that we re-calculate annually.

      The "safe" withdrawal rate is a function of expected market returns, expected market volatility and remaining life expectancy. The "safe" withdrawal amount further depends on your portfolio's performance.

      This can be seen in a paper written by Moshe Milevsky. I provided an Excel spreadsheet some time ago that makes the calculation (you can still download it here).

      Before you do, however, you might want to read the paper in footnote [5] above in which Milevsky later seems to regret any contribution he may have made to the constant-dollar strategy.

      The percentage you can safely withdraw grows over time as your life expectancy increases but the size of your portfolio varies up and down so your future withdrawal amounts vary and are as unpredictable as your portfolio returns.

      You ask if "this year's living needs" are a factor. Excellent question. They are not.

      Most spending strategies try to estimate how much you can safely spend without running out of savings. They don't consider how much you will need to spend or if the amount you can expect to spend will be greater than your expected expenses.

      Even if we could perfectly predict how much we can safely spend throughout retirement (we can't) our expenses are largely unpredictable and expense shocks are completely unpredictable.

      Lastly, though you phrase your question as a function, the answer is not deterministic. We can only estimate how much you can probably spend. These probabilities are based on some pretty uncertain assumptions, like the future rate of returns of our portfolios and future inflation.

      Thanks for asking!

  3. Fascinating topic, Dirk. In reality, folks react when the market goes through a bear, and I can't imagine any retiree not reducing spending when they see their portfolio dropping 20%. I enjoy your thinking, and look forward to your variable spending post. I'm retiring in June 2018, and am planning a "floor/ceiling" approach based on trailing returns. We'll see how it plays out....

    1. I can’t, either, Fritz, but that may just be a failure of our imaginations.

      If they do reduce spending, then that’s a different strategy.

      There comes a point as a retiree’s finances head toward failure at which he has no choice but to keep spending to pay non-discretionary bills even though he sees that will eventually be fatal. (The alternative is to have his power and water service disconnected this month.)

      I have spoken with retirees who fell into this trap.

      I no longer believe that everyone can see ruin coming in time to jump out of the crashing plane before it’s too late. However, I was discussing this with two other researchers recently and the consensus was that this is the primary risk-mitigation plan for middle income retirees.

      Thanks for the comment!

  4. Using Social Security and Pension as "floor portfolio", could one structure upside portfolio with mostly dividend paying stocks so one never need to sell shares? Sure dividends are not guaranteed but at least aren't those dividend king/aristocrats more safe than the insurance companies selling annuities (itself is very expensive anyway)?
    I tend to think those strategies are better than MPT (bond/stock mix) and 4% rule.

    1. Sure, you could do that but I wouldn't recommend it.

      If a growth company earns a $1.00, it will likely skip dividends and reinvest most or all of that dollar in its company's expected high future growth rate. A slower growing company has to pay out a dividend because its growth rate can't compete. So, maybe they pay you $0.02 on the dollar in dividends and your stock price immediately falls by the amount of the dividend. Maybe you're happier with 98 cents plus two cents than with a dollar, but they're pretty much a wash to me.

      The other problem is that investing only in value stocks like this doesn't give you adequate diversification. You should buy growth and income, small and large cap, domestic and international.

      As Wade Pfau once explained, spending-dividends-only is not a valuable strategy.

  5. Thanks for your post. As always I enjoy the thinking behind it. I agree it is not wise to make fixed spending rules with a volatile portfolio. There are so many future possibilities. Rather than calculate the expected return of ultimately unknowable futures, perhaps it is possible to balance one unknown against another and thereby cancel them out.

    1 Legacy plans used to balance the risk of health care emergency / long term care.
    So these funds are set aside to be used for which comes first. If it's long term care, the legacy recipients at least didn't have blubbering granddad sitting in den.

    2 Increase for inflation is balanced with trend to decrease spending as we age and decreasing life span
    So use inflation increases for first 10 years and not there after.

    3 Increasing Annuity income allows for higher stock return and volitility
    So delay SSI until 70 by spending from portfolio

    4 Sequence of Risk return is balanced by decreased stock percentage at retirement.
    So if poor returns early in retirement spend more from bond portion bucket.

    5 Fraud and decreased mental acuity poor decisions balanced with annuity %
    So transition more % of bond portfolio into annuity% driven income stream.

    6 Lack of spousal interest is balanced with an adaptable plan and trusted advisor.
    So make a will, write information down, find a trusted person or company.

    7 Uncertainty of future is balanced with plan that is changable with situation.
    So enjoy today and be thankful for every day you are vertical.

    1. Some good thoughts here but I wouldn't say "rather than."

      A good retirement plan needs to minimize the risk of failure and be able to survive those scenarios in which the improbable occurs. Simulating future scenarios accomplishes the former but the latter has to be analyzed as you suggest. Those scenarios are too improbable to be adequately explored by simulation. Still, you can't throw out the former.

      A few additional comments. You have to be very careful assuming that spending will decline as you age. It does for some (especially those with resources appropriate for their spending) but not others (undersavers, for instance). For both groups, spending spending is quite volatile year-to-year. And lastly, for those whose spending does decline, that decrease is typically 1.5% or so a year and not nearly enough to offset other than very mild inflation. I never count on this when planning.

      Second, if you spend more from bonds then you increase your equity allocation. If your problem is poor equity returns then you're making it worse.

      Lastly, I don't see a rational alternative to a plan being "changeable." On the other hand, many retirees seem to plan on mitigating risk by figuring out how to fix the problem if and when one arises. In many cases, by the time they figure out they need to change it will be too late.

      Thanks for the comments!

  6. Great article and I'm looking forward to the next installment. I may be incorrect, but it seems to me that most of these SWR discussions are based on an asset allocation that is the S&P 500 and US Treasury bonds. How does the math change when you have US large cap, small cap, international large cap, small cap, emerging markets, REITS, TIPs and US treasuries.

    We are 60/40, stocks / bonds, but 50% of the stock portion is international. And both US and the international stock exposure is titled slightly toward small cap and value. We use DFA funds primarily through our advisor. We also have 2 Stoneridge funds which, theoretically, are less correlated to other asset classes.

    Maybe that type of analysis is too complex for the models to use- looking at historical returns for each asset class for example. Most models that I've seen just let you plug in an expected return without any consideration of volatility that might be dampened by a widely diversified portfolio. It seems that such a portfolio would reduce the chances of sequence of returns risk some.

    1. Mike, I don't think it would have a big impact. All of the models I use, build or read about use a volatility estimate. It doesn't matter what components make up the portfolio if you use a representative mean and sigma for their combination.

      The problem is market uncertainty. We don't know the underlying expected return, volatility or even the distribution. Asset correlations change.

      While the diversification you describe is clearly a good thing to minimize portfolio risk, it doesn't make the projections any less uncertain to guess a different mean and variance.

      Sequence risk is a problem caused by periodic spending from a portfolio. You can mitigate it by allocating more to bonds and less to stocks, so long as you don't allocate too much to either. I can't think of how diversifying within the equity allocation would help much.

      Thanks for commenting.

    2. I did a quick follow-up analysis this morning. My models often assume a 5% expected real return with 12% standard deviation because other research does, as well. It's easier to compare. However, I also frequently use returns and risk from for broadly diversified equity portfolios.

      For a 50% IFA equity portfolio the long term stats are mean=8.14% (nominal) and sigma 11.15%. Inflation averaged 3.01% annually, so that gives me a real mean return of 5.14%, not very different than 5% and 12%.

      Using Milevsky's formula, 5/12 calculates a p(ruin) 15% with a 4% annual withdrawal and a 30-year life expectancy while 5/11.2 estimates 13.44%.

      Assuming a 15-year life expectancy, 5/12 estimates a p(ruin) of 6.6% compared to 5.94%.

      Given that the confidence interval for these estimates is huge due to the limited amount of historical data available, this difference seems fairly negligible.

      Changing the overall equity allocation, life expectancy and withdrawal rate assumptions seem to have a much larger impact on survivability
      than equity sub-allocations.

      That's far from a thorough analysis but it suggests that broader equity diversification may provide a lower probability of ruin but probably isn't the most important factor. Still, it's a good idea for other reasons.

    3. I have found that the "Portfolio Charts" website is useful for understanding how different portfolio components can impact long-term performance without relying on the normal distribution in a Monte Carlo approach. It uses an actual return database but that only dates back to the early 70s. This doesn't have the Great Depression in it, but is still useful as it has a period with high inflation as well as the big bond and stock market from the past 35 years, so sequence of returns does show up.

      Basically, it shows that having a more diversified portfolio does increase the safe withdrawal rates, but it is not going to push an SWR from 3.5% to 6%. My main takeaway is that once you have tossed some international into the equities, the more complex portfolios don't provide that much extra benefit, so asset allocation overall is much less important than the other personal finance decisions in your life.

    4. This comment has been removed by the author.

    5. Sorry, I should restate that. Lowering mean-variance always helps but equity sub-allocations don't lower it enough to make a big difference. Changing your stock-bond allocation would also lower mean-variance and have a significantly greater impact on sequence risk. Once you've done that, you're tweaking.

  7. I listened to the "Masters in Business" interview with William Sharpe recently where he called withdrawals during retirement the most complex problem in finance today. Given the increase in lifespans and retirement periods over the past three decades, I have considered Bill Bengen and some of the subsequent dynamic withdrawal researchers viable candidates for a Nobel Prize in Economics due to the importance to society of the findings regarding sequence of returns and potential solutions. Unfortunately, this is unlikely to occur given how practical the research has been.

    Similarly, the development of the concept of the low cost index fund in the 60s with initial implementation in the early 70s has totally changed investment for the masses for the better over the past several decades.

  8. cycnicalanddisgustedFebruary 14, 2018 at 7:29 AM

    I would take exception to only one statement "We pretend they are normally distributed, largely because it makes the math easier".
    Having spent a couple of years in defense analysis I discovered that Normal distribution are A. not easy, B. not accurate, and C. very popular.
    I suspected the popularity of selecting Normal distributions had its basis in two ideas. One, it was complicated enough to justify the analysts existence and two, it was simple enough to calculate on a computer.
    I showed many examples where a uniform distribution was easily as accurate, much easier to calculate, and certainly easier to demonstrate.
    Could the financial analyst select Normal for similar reasons?
    By the way, my ideas were not well accepted by the defense analysis community.

    1. I love the name. Every time I see a notification pop up on my screen saying that I have received a "cynical and disgusted" comment, I have to smile.

      I'm not really seeing a disagreement here. When I say it makes the math easier, I mean easier than treating the distribution as non-parametric.

  9. Wondering if you can comment here or in a future post and clarify the difference between (and the relative pros and cons) of the retirement buckets/time segmentation strategy (which Wade Pfau says is a hybrid of probability based and saftey first) vs the safety first strategy (which I think you espouse)? And on another note, is there a way to sign up to receive an email alert for all future comments on one of your posts? Thanks.

    1. I plan to post on time segmentation (TS) strategies in the near future.

      I would say that floor-and-upside is my preferred strategy for many households, though some households will have little or no upside portfolio and very wealthy households probably don't need a floor. I don't like to make blanket statements like "floor-and-upside is always the best strategy" or "everyone should delay claiming Social Security benefits" because there are always scenarios in which those wouldn't be the best strategies. In the majority of households that follow my blog, however, I suspect they usually are the best strategies.

      I discussed TS strategies with Wade recently and, while he has written about them, he has written about nearly all valuable strategies, so I wouldn't consider that an endorsement. TS has known issues (no strategy is perfect), primarily cash drag.

      Regarding the email, you can sign up to receive all my posts by email on the right side of the blog. If you write a comment and someone replies to it, I believe you will also receive an email notification of that reply. I don't know of a way to be notified when someone comments on a post that isn't a reply to one of yours. I'll check around. Sorry.

      Thanks for the note!

    2. Thanks. So is 'safety first' the same as 'floor and upside'? And 'time segmentation/retirement buckets' is not considered floor and upside? Trying to clarify all this.
      There's no place to click to request to receive further comments after commenting on your blog, as I've done with other blogs....
      Thanks again.

    3. Let me quote from that post. "The floor-and-upside strategy for financing retirement is sometimes referred to as “safety first” and derives from The Theory of Life-Cycle Saving and Investing."

      The primary goal of TS is to match the duration of liabilities to the duration of investments, not necessarily to provide a floor. An annuity, for example, would be an unlikely component of a pure TS strategy. I'll explain that in a future post. I don't consider it a safety-first strategy.

      Correct, there is no such place to click to receive further comments on a blog post, though if someone replies to your comment, I believe you will be notified.

    4. Thanks. And just so you know, I wasn't notified when you replied to my comment.

    5. Thanks. Good to know. Apparently, only the author gets notifications. Sorry I couldn't help.

    6. When you publish a comment, do you see a check-box to the right of the PUBLISH and PREVIEW buttons labelled "Notify Me"? If you see that, then checking the box should notify you when someone responds to your own comment.

  10. Hi Dirk,

    I just recently discovered your blog and have enjoyed reading.

    Regarding your analysis of SWR strategy, I tend to agree with a point made by Michael Kitces in one of his blog posts that SWR is, in effect, no different from a "floor-with-upside" approach (which you seem to prefer), and that they they are essentially just two ways to skin the same cat.

    I take your point that withdrawals from a volatile portfolio will never be quite as safe as an annuity in terms of building a floor, but if a conservative withdrawal rate is chosen, I believe the risk is small and worth the potential (and very likely) upside: Better liquidity, much better chance of accruing a large portfolio that can be tapped at higher rates later (once it becomes apparent that you weren't one of the unlucky few cases where sequence of returns risk nailed you early in retirement), and the chance to leave a bequest to children, charitable causes, etc. One can never totally hedge against the unknown, but I believe many folks find this a compelling risk-reward proposition.

    I also tend to agree with Wade Pfau and others who believe 4% is probably too optimistic looking forward. However, a 3% withdrawal should weather all but a cataclysmic economic event-- the likes of which could very well challenge the solvency of insurance companies paying annuities as well.

    As a side note: while interesting, I don't believe Pfau's research into how the 4% rule has fared in other countries is particularly relevant to U.S. retirees. For starters, market performance is not randomly generated. It is always tied to the productivity and macroeconomic circumstances of the individual country in which one happens to live. To suggest that there's a global mean to which the U.S. might be expected to revert since we're above it is like suggesting that Togo is due to catch up any day now since it's so far below. Secondly, the low-performing countries (below 3% SWR) in Pfau's data were nearly all losing parties in a world war (or war-ravaged countries such as France) and the failure cases are clustered around these catastrophes. If you expect the U.S. to suffer a similar fate within the next 30-50 years, then all bets are off, including the survival of the aforementioned insurance companies.

    Lastly, I think economists sometimes can't see the forest for the trees. Their goal is efficiency: Spending as much of the portfolio as possible without depleting it before dying. I think most retirees, by contrast, just want a comfortable, sufficient retirement, and won't feel cheated if their cushion becomes a pile of cash left to their children or a charitable cause when they go. The fact that retired folks overwhelmingly reject the advice to annuitize a large chunk of their portfolio could indicate that they're all out to lunch, or it could indicate that economists don't understand the motivations and desires of most retired folks. I think it's the latter.

    Best regards,

    1. Dave, I respect your opinion, but disagree. I know retirees who have gone bankrupt with an SWR strategy. I also have great respect for Michael Kitces, but don't agree that SWR is another form of floor-and-upside and I don't believe most researchers agree, either.

      I think Pfau's international research is indeed relevant. I think it shows that the U.S. won the 20th century economically and there is no reason to believe it will win the 21st. Countries, like companies, grow old and their economy has to slow, just like that of a small growth company.

      Do you have data to support your statement about most retirees not caring if they leave a lot behind? Because, I just worked with two families and both wanted to maximize consumption and not worry about a bequest.

      Regardless, what most retirees want is irrelevant to your own personal plan desires.

      A 3% withdrawal rate should weather most storms but what if it doesn't weather yours? I'm not willing to take that risk when I don't have to.

      As I said, I respect your opinion and I hope that works out for you, but I don't find much I agree with.

      Thanks for writing, Dave.

    2. Bill Bernstein looked at a variant of the SWR calculation a almost two decades ago in his "The Retirement Calculator from Hell" series. In Part III, he looked at the question of how safe that 95% probability of not running out of money was. His conclusion was that anything over 80% was likely to be over-precise compared to the data set:

      The US had a major civil war in the mid 1800s. Where is it written that there won't be one 30 years from now? Almost nobody suspected that the biggest economic, financial, and military conflagrations were likely a decade before they occurred.

      So while I don't think Togo is a valid comparison to the US, the European countries certainly are potential waypoints over the next century. We have major advantages over those countries, such as the past proven ability to assimilate immigrants in and make them Americans (Canada is similar) unlike many European and Asian countries. However, we do do go through waves where this is less possible. Who knows when a wave will not actually recede and go back to building our strength?

      We also have a big moat which reduces the potential for invasion, but nuclear weapons on missiles in more and more hands makes that moat less important. An EMP explosion or two would likely crater our economy for years. It could be our version of the Blitzkrieg hitting Holland, Belgium, and France.

    3. That series is what got me interested in retirement planning.

      I think the culprit here is one that, unfortunately, is common in retirement strategy thinking – equating portfolio failure with retirement finance failure.

      "Poor investment results" doesn't even make the top-ten list of reasons cited for elder bankruptcies.

      I try to explain to clients who want a portfolio survival rate of 90% to 95% that there is perhaps an 80% chance of their finances failing and that failure may have nothing to do with portfolio failure. Huge medical expenses, for example, could lead to bankruptcy even in a good market.

      Thanks for the comment!

    4. I hear an interesting undertone from many people who claim that no Social Security or pension will be there for them in the future. As a result, they are entirely focused on their portfolio to save them.

      I have a different take. We plan on deferring my SS to 70 as mine should be close to the maximum available, recognizing the current actuarial year 2034 21% haircut challenge. My spouse has a NYS pension that is 90% funded using somewhat optimistic assumptions, so it is one of the better pension funds out there.

      Between our SS and pension, we should have a comfortable middle-class floor income even if there are 20% haircuts down the road on both. This floor income should be largely unassailable in bankruptcy and would be difficult to steal in its entirety through hacking financial accounts. This income alone should cover all of our normal fixed costs and then some.

      Our portfolio will provide the "fun money" as well as long-term care security etc. The portfolio will have something like a 50/50 stock-bond split with a mix of US and international equities and the fixed income side focused on high quality bonds and some cash.

      I am under no illusion that if the social, economic, and financial system were to collapse to the point that SS and NYS pensions take 50% + haircuts, that my portfolio would be oblivious to this. Instead, I think we would be looking at a 1930s type of investing scenario or worse. So I throw massive loss of SS etc. into that 20% bucket that Bernstein discusses where all hell breaks loose and the outcome of everything becomes uncertain. So I don't worry about it because I have no idea how to effectively hedge against it short of becoming a billionaire and buying my own island high enough to avoid rising seas.

    5. I agree with your analysis. I think there is only a small chance that Social Security will go away, though a better chance that benefits will be reduced. If it does, what will you do? For most households, it is effectively impossible to finance retirement without it.

      I am a proponent of looking at retirement from at least two perspectives. First, plan the upside so the probability of bankruptcy is quite low, then have a backup plan (a floor) in case the low probability scenarios come to pass.

      Thanks for sharing!

  11. In all the discussions of SWR I have not seen any definitive information of what asset allocation it should be based on. A 70/30 allocation will permit a whole different withdrawal rate than a 30/70 split. What should asset allocation be based on different scenarios - expenses covered by pensions and SS, no pension but 80% of expenses covered by SS etc. etc.

    1. "In all the discussions of SWR I have not seen any definitive information of what asset allocation it should be based on."

      Not sure what you've been reading, but all of the research I read considers asset allocations. So does Bengen's Book, Conserving Client Portfolios During Retirement, which you can find at Amazon.

      Having said that, as my post points out, constant-dollar SWR strategies suggested in that book are a bad idea.

      "A 70/30 allocation will permit a whole different withdrawal rate than a 30/70 split. What should asset allocation be based on different scenarios - expenses covered by pensions and SS, no pension but 80% of expenses covered by SS etc. etc."

      Are you looking for the withdrawal rate that you need or the withdrawal rates that various allocations might support? Those issues after your hyphen determine the former. If you have a high floor, then you will need to spend less from the upside portfolio and vice versa. I would think a more logical approach would be to start with the amount of spending you need from the upside portfolio, not with asset allocations.

      A "good" asset allocation depends on what you are trying to achieve.

      If you are trying to find the optimal asset allocation to grow your portfolio, good luck with that. Gordon Irlam, who developed, found that the optimal asset allocation to achieve this in one of his studies fell within a 95% confidence interval of 10% to over 80% equities. If your future holds great market returns, something like 80% would be great. If poor returns lie ahead, 10% might be optimal. Since future returns are uncertain, there is no way to know which will work best for you.

      Are you, instead, trying to find an asset allocation that lets you sleep at night, whose volatility won't tempt you to sell at the worst time? I have suggested Bernstein's guidance, which you can determine my taking the worst portfolio decline you could accept in a bear market and allocating twice that percentage plus 10% of your portfolio to equities. (Be aware that your estimate of how big a loss you can stomach tends to be significantly higher in good times than when you actually start losing tons of money in a bear market.)

      I find that calculations tend to be higher than most people can stomach. It suggests 85% equities for me, but I can only stomach the volatility of a 40% or 50% equity allocation. I've received similar comments from readers. So, again, it depends on what you are trying to achieve, portfolio growth or less stress.

      Thirdly, perhaps you are trying to find the asset allocation that is most likely to support the spending rate you choose. The charts in Bengen show that the best allocations tend to be in the middle of the range, with portfolio survivability dropping off below 30% and above 70%.

      Most simulations I run show the same. When I analyze all failed-portfolio scenarios in a simulation, the majority of them occur below 30% and above 70% equity allocations.

      Lastly, you have to consider the rest of your retirement plan. What is your remaining life expectancy? Are you single or married? How much total wealth do you have?

      My "elevator answer" to this question is to allocate 40% to 60% to equities depending on your risk tolerance as suggested by Bernstein.

      Or you can do what Harry Markowitz, the father of Modern Portfolio Theory, did with his retirement portfolio – throw up your hands and invest half in stocks and half in bonds. There are a lot of arguments that end up somewhere around that allocation.

      Frankly, I believe market uncertainty is so great that trying to find the perfect asset allocation is a waste of time. Just don't pick a really bad one. Good luck and thanks for writing!

    2. A couple of years ago, Meb Faber did an analysis looking at a number of published asset allocation models. He just looked at them from an accumulation standpoint from about 1972 to 2013. The basic conclusions was that most of the models clustered fairly close to the traditional 60/40 and that fees were more important than any reasonable asset allocation model. He took the best performing model by Mohamed El-Arian and showed that by adding 1.25% fees it would under-perform most asset allocation models and at 2.25% fees it became the worst performing model.

      So a relatively simple diversified asset allocation with 40% - 60% equities and low fees is probably as much certainty as possible in this business.

  12. Wealth Meta wrote a calculator that does exactly what this post is talking about (sets initial withdraw amount, then adjusts for inflation). It lets you play with the asset allocation and does historical back testing. Worth fiddling with to understand what is going on.

    1. This comment has been removed by the author.


    2. I publish this link with some reluctance as I describe in a reply to the following post. These calculators provide a VERY shallow understanding of the issues.

      Here are a few concerns with this one. It works with a fixed length of retirement rather than simulating random life expectancies. It uses a constant-dollar withdrawal amount. It simulates 90 scenarios when 10,000 would be more appropriate. It provides mean outcomes when, given the number of expected outliers, the median would be more valuable (and probably lower). A distribution of outcomes would be even better.

      Play with these simulators if they help but put very little confidence in the results.

  13. As I am on the brink of retirement and now have to invest a large amount of cash into a portfolio ((UK SIPP) bond yields make this an interesting time. Your articles are always clear and I value them. One thing though, a failure metric based on the number of years unfunded would seem a good parameter but I never seem to see it. Seems to me failure by 1 year is much different from failure by 5 years yet both contribute equally to the single failure probability.


    1. This comment has been removed by the author.

    2. Paul, you are correct. One of the oldest criticisms of Bengen's and similar studies, which are based on the probability of ruin, is that they consider a 30-year retirement and treat a plan that supports 9 years or 29 years equal failures and plans that support 30 years or 40 years equal successes. This is one of the reasons I have suggested abandoning P(ruin) as a metric, as has Moshe Milevsky.

      In simulations of tens of thousands of scenarios, there will be many scenarios that come up just short and many that barely survive. I suspect that they average each other out.

      The answer is to look at the distribution of outcomes rather than draw a line in the sand and claim victory just on the other side. The answer also includes randomizing life expectancies instead of assuming a 30-year retirement – a portfolio that fails after 20 years isn't a problem if you only live 19.

      My simulations always provide distributions of outcomes and random lifetimes.

      Bottom line, a simple SWR simulation with fixed lifetimes and a single hurdle for success like P(ruin) isn't very useful. That's why I don't recommend calculators like the one mentioned in another comment to this post. They provide a very shallow understanding of the problem.

  14. Enjoyed studying this, very good stuff, thanks.

  15. This comment has been removed by a blog administrator.

  16. This comment has been removed by a blog administrator.

  17. This comment has been removed by a blog administrator.