The Ultimate Guide to Safe Withdrawal Rates – Part 15: More Thoughts on Sequence of Return Risk

Welcome back to our Safe Withdrawal Rate Series! Last week’s post on Sequence of Return Risk (SRR) got too long and I had to defer some more fun facts to this week’s post. Again, to set the stage, I can’t stress enough how important Sequence of Return Risk is for retirement savers. In fact, after doing all this research on safe withdrawal rates (start series here, and also check out our SSRN research paper) if someone asked me for the top three reasons a retirement withdrawal strategy fails I’d go with:

  1. Sequence of Return Risk,
  2. Sequence of Return Risk,
  3. and let’s not forget that pesky Sequence of Return Risk!

Huh? Isn’t that lame? Surely, low average returns throughout retirement ought to be included in that list, right? Or even top that list, right? That’s what I thought, too. Until I looked at the data! Let’s get rolling and look at some more SRR fun facts.

Low average returns are less of a problem than Sequence of Return Risk!

To see how much or how little average returns impact SWRs, let’s look at the following example of a retiree with a 30-year horizon and an 80/20 equity bond portfolio. I calculate the SWRs with the following assumptions:

  • All simulations are in real terms (CPI-adjusted withdrawals and CPI-adjusted asset returns)
  • Capital depletion (final value =$0). Results would be qualitatively similar results if I use capital preservation, which would be more applicable for early retirees with a longer than a 30-year horizon.

As a warm-up, let’s look at the table below: three case studies for different retirement starting dates. It displays the SWR that would have exactly exhausted the portfolio over 30 years, the 30-year average return of an 80/20 portfolio (0-30 years, point to point CAGR), as well as the average returns over the 5-year windows: 0-5 years, 5-10 years, … 25-30 years.

SRR Table04
Three case studies: Safe Withdrawal Rates of an 80% Stock, 20% Bond portfolio (30-year horizon, $0 final value) vs. realized real returns.

Retiring in December 1968 would have afforded you an SWR of only 3.8%. And that would have exhausted your capital over 30 years! But the average return over the 1968-1998 period would have been a staggering 6.16%. The reason why you still ran out of money after 30 years is that you had low returns early on and the strong returns, even double-digit real returns in years 15-30, came too late.

If you had retired only ten years later you would have experienced a very similar 30-year average return: 6.03%. But the SWR would have been a staggering 9.12%. The strong returns came during the first 20 years of the simulation and the weak returns during the last 10 years that cover the recessions in 2001 and 2008/9 wouldn’t hurt your SWR anymore. Also, October 1955 is an intriguing case study: Only very underwhelming average returns: 3.45% over the next 30 years, but a very healthy SWR of 5.72%.

Of course, with case studies, we can go only so far. In the chart below, let’s plot the entire SWR and average return time series. Wow, the SWR and the average return are only very slightly correlated. Even if you knew the average return of your portfolio mix over the next 30 years, you’d have a hard time pinning down an appropriate SWR. Some of the lowest 30-year returns in the late 1800 and early 1900s actually coincide with relatively decent SWRs of around 4% and as high as 7%!

SRR Chart01
The path of the Safe Withdrawal Rate has only little in common with the 30-year average return of the underlying portfolio! Note: the x-axis is the retirement start date, so it goes through March 1987!

Very interesting! There have been plenty of examples where the 30-year SWR far exceeded the 30-year rate of return. SRR is a risk that can go both ways; sometimes it helps the retiree sometimes it hurts the retiree. We knew that already from last time: the beneficiaries and losers are the savers and retirees. If the retiree benefits then the saver loses and vice versa. It’s a zero-sum game!

Some more serious SRR analysis

Let’s do some more sophisticated statistical analysis. We run two linear regressions to “predict” the SWR based on future returns. 1) knowing only the 30-year average return and 2) knowing the average returns during the six windows: 0-5 years, … , 25-30 years. Of course, “predicting” the SWR is a bit of a misnomer. Nobody knows the future returns. This is more of a thought experiment of how well you could have pinned down the SWR if you had known future returns. Or, let’s call it accounting for the SWR in hindsight, rather than predicting it.

The results are in the table below:

  • Knowing only the average returns over 30 years we get a pretty underwhelming regression fit. An R^2 of only 0.31, so knowing the 30-year return explains only 31% of the variance in the realized SWR. We already saw the poor fit/correlation from the chart above, but this is the statistical and quantitative confirmation. But note that the slope coefficient on the average return is positive and statistically significant. And for statistics wonks, yes, this is a Newey-West adjusted t-stat to account for the overlapping windows.
  • Knowing the returns in the 6 separate windows you get an almost perfect fit: close to 96% of the variation in the SWR are explained by the average returns in the 6 windows. What’s more, the slope coefficients for the different windows are very different and all extremely highly statistically significant. They may sum up to roughly the same number as in the univariate regression, but the earlier windows get a much larger slope coefficient. By weighting the 6 different time windows differently we now get an almost perfect fit. Precisely what I mean by SRR matters more than average returns: 31% of the fit is explained by the average return, an additional 64% is explained by the sequence of returns!
SRR Table03
Linear regression result: SWR on future realized returns. Knowing only the average returns over the next 30 years is not very informative (low R^2). The distribution of returns over the 30 years matters a lot more! Note: Due to overlapping windows the t-stats are Newey-West heteroskedasticity-adjusted.

If you’re unlucky you can get screwed twice by Sequence of Return Risk!

What’s even worse than getting screwed over by SRR in retirement? Very simple: first you get screwed over by SRR while saving and then again while withdrawing money in retirement. Let’s look at the following hypothetical retirement saver who starts saving $5,000 in 1959 and does so for 15 years. Then he withdraws $4,000 from the portfolio during retirement. During the last few years of the accumulation and the first few years of retirement, the 1973/4 recession hits. You get hammered twice because that’s exactly the kind of return profile that you want to avoid while saving (high return early on, low returns later) and retiring (low returns early on, high returns later).

To see how much this saver/retirement cohort lost from SRR let’s plot the actual portfolio value that’s subject to SRR and the hypothetical portfolio value had this person experienced the exact average monthly return that prevailed during those 30 years (orange line). Without SRR you would have about 25% more in your portfolio at the beginning of retirement. After 15 years of retirement, you would have seriously depleted your portfolio. Without SRR you’d be about 100% better off! So, SRR hurt you both while saving and during retirement. Bummer!

SRR Chart03
Investing $5,000 per month in an 80% Equity, 20% Bond portfolio for 15 years, then withdrawing $4,000 per month for 15 years. All values are real, CPI-adjusted numbers. Actual returns vs. experiencing the 30Y average throughout reveals that Sequence Risk can cost you both during the accumulation and retirement!

Are there ways to alleviate Sequence of Return Risk?

Last week after posting the first part of the SRR blog post, two commenters had suggestions on how to overcome or at least alleviate the SRR problem. Both of them are brilliant ideas. But only one of them works.

1: Use the Bogleheads-endorsed VPW rule.

Instead of withdrawing a fixed amount regularly, why not withdraw a certain percentage of the principal. This can be a constant percentage or the age-dependent withdrawal percentage in the VPW rule. Now, the final net worth looks like it’s independent of the order of returns:

(1-w) ∙ (1+r1) ∙ (1-w) ∙ (1+r2) = (1-w) ∙ (1+r2) ∙ (1-w) ∙ (1+r1)

Since the final net worth is the same, does that mean SRR is irrelevant for people who apply the VPW? Not really. Because the second withdrawal will now differ depending on which of the returns is larger. Imagine r1>r2. Then your second withdrawal is higher when you experience r1 first than when you experience r2 initially. So, you can’t hide from SRR. If you try to equalize the final portfolio value through VPW then SRR hits you through the withdrawal amounts! If you try to equalize the withdrawal amounts then SRR hits you in the final portfolio value. Pick your poison! The VPW has its pros (and cons) as we showed here, but it can’t eliminate SRR!

2: “Mortgage” your retirement.

For retirement savers there is one sure-fire way to avoid missing out on strong equity returns early during the accumulation phase: Borrow against your future retirement account contributions and invest the whole loot as one big lump-sum payment without further contributions in the future (i.e., use your future retirement contributions to pay down the margin loan). Sounds crazy? Two researchers from Yale found that this is a way to “diversify across time,” which is just another way of reducing SRR. As we showed last week, a lump-sum investment is not subject to SRR!

If this sounds too extreme to you, I’d have to agree. There are multiple reasons why this is not workable. For example, since I work in a very volatile industry I’d not be comfortable borrowing against future earnings. But there are ways to at least alleviate the SRR problem:

  • Hold a 100% equity portfolio early on. Don’t bother about holding any bonds.
  • Don’t bother about paying down low-interest debt when young (mortgage, low-interest student loans). Put every last dollar you can scrape together into the stock market early on. (of course, high-interest debt, especially debt with interest higher than your expected equity return should be paid down as early as possible)
  • Once you have a critical mass of equity investments, then tackle your low-interest debt.

The benefit of this method is twofold: 1) you gain from the higher expected return in equities over fixed income investments and 2) you alleviate the SRR by spreading around the equity risk more evenly across time. It’s a win-win!

We hope you enjoyed today’s post. Please leave your comments and suggestions below!



145 thoughts on “The Ultimate Guide to Safe Withdrawal Rates – Part 15: More Thoughts on Sequence of Return Risk

  1. It isn’t surprising why low average return isn’t even among the top 3 because SRR dwarfs all other risks. If you end the first decade of retirement with at least 70% of your starting stash, then you should put any problems in the rest of the retirement. Else, your SWR needs a serious adjustment or a side income stream becomes necessary. Good analysis as always, big ERN!

    I recently did a similar backward-looking exercise where I took my current dividend portfolio, recast it using 2008-09 data and analyzed quantitatively what is the portfolio level dividend cut would’ve been. If that cut is lower than my safety margin in dividend consumption for living expenses, then it means the Great Recession can be experienced with no sale of shares at the bottom. Selling equities at the bottom is precisely what aggravates SRR and causes retirement failure.

  2. Another great read! I look forward to your posts as a momentary distraction from work every week! Michael Kitces had an interesting article a while back that reached a similar conclusion and found the highest correlation to a successful 30-year SWR was the first 10 years of returns, but I like how you have blown that out even further by 5-year periods.

    Question – you clearly know a lot more about statistics than about every other early retirement blogger out there. The mother of all retirement models would be to have a Monte Carlo feature in your withdrawal calculator that applied Monte Carlo simulations to the lower expected future returns. I would expect that might have a data table to do multiple simulations, and it would take a while to calculate, but it would be pure gold. Do you have any plans to enhance your already great calculator in this manner in the future? Maybe you think Monte Carlo doesn’t translate well to SWR prediction – but I’d be interested to hear your views on that as well.

    1. I am thinking about doing some MC simulations. But so much depends on the expected returns. One can show everything and nothing. Also it’s hard to simulate the mean reversion feature of equity returns with a simple Monte Carlo simulation.

        1. Wow, nice article. But what are the assumptions in the Monte-Carlo experiment? Is this re-sampling from actual returns or simply log-Normal returns (what are mu and variances/co variance?)
          1.73-2.65% seems awfully low even for a pessimist like me. One thing a Monte Carlo sim can’t replicate very well is the sharp recovery after a deep drop, like 2008/9. Maybe that’s why the SWR came in so low?
          Thanks for sharing!

          1. Log Normal. (Except for the “Low Yields/High Valuations” model which is more complicated.) The numbers are all for a 60-year retirement with 50% of the portfolio left at the end, which makes it substantially longer & more conservative than most things people graph.

            The lack of mean reversion (which *seems* to happen most of the time…at least outside of Japan) in Monte Carlo definitely pushes down the withdrawal rates a fair amount on longer time horizons. Since I *mostly* believe in mean reversion, it is one reason why I take monte carlo with a large grain of salt.

              1. I used 1) US historical numbers, 2) “reduced historical numbers” i.e. lower expected return and higher standard deviation, in an attempt to mimic “but but but future returns will be lower” :), and 3) world historical numbers.

                1. Ok. Well what were the specific numbers? E.g., equity expected return 4.0% real, bond expected return 0.5% real, risk 15.0% and bond 6.0%. Correlation -0.25.
                  Not sure why you are so hesitant to share those. If someone tells me about a MC experiment and doesn’t tell me what the parameters are, I have a hard time interpreting the results. The return parameters are the one part that can’t be a “state secret.” 🙂

                2. Wasn’t hesitant to share them, just didn’t know that’s exactly what you were after (and some people are bored by that level of detail) :). I didn’t use correlations because the median + standard deviation were for a blended portfolio. Research Associates publishes 3 estimates. I used their “medium volatility” forecast which is: 4.9% mean, 10% standard deviation. The US historical come from Blanchett’s paper 2012 paper on Optimal Withdrawal Strategies: 5.2% and 11.82%. The “reduced” numbers are just those but -.5% mean and +2% stdev: 4.7% and 13.82%. The world numbers came from Dimson-Marsh-Staunton. They used to publish them annually in the Credit Suisse yearbooks but stopped in 2012 for some reason, so I used the 2012 numbers.

                3. OK, got it. Yeah, the MC simulations I did are about in line with those numbers. One thing I noticed about Monte Carlo: There is less Sequence of Return Risk. In actual numbers there is mean reversion in returns: if returns were low for 5 years they are more likely to snap back. In a side-by-side comparison, I found that the R^2 of regressing SWR on 30-Y average returns is 0.65 in Monte Carlo, vs. 0.31 in the actual return data!

            1. Parametric Monte Carlo simulations are quite problematic for retirement planning IHMO. Without a fully parameterized stochastic model of the stock market based on rigorously tested principles and theory, such a simulation is simply garbage in–garbage out as assumptions will likely preordain the results. A nice read along this line of argument:

              Perhaps my extreme pessimism is because Dr. Andrew Lo’s recently released book entitled “Adaptive Marketings” is still haunting me:

              Dr. Lo’s book is a great read and, on a tangent, I was intrigued by his Volatility-Controlled Value-Weighted Indexing in Chapter 8 as a step towards dynamic asset allocation between stocks and bonds.

              1. Agree: With MC sims it’s garbage in, garbage out. I much prefer doing the historical simulations, who also have their shortcomings, but at least they capture more realistic moves over the business cycle.
                Interesting read the book by Lo! I put that on my AMZN wishlist!

                1. Hi ERN, great series!!!
                  McClung uses in his book Bootstrapping Simulation. What do you think about this kind of simulation?
                  Also in this post, you focus on 30-year retirement. How is it different for a longer horizon retirement?

                2. Bootstraping is fine but unless you do it with longer blocks (as opposed to simple re-sampling) you’ll lose the mean-reversion properties of returns and varying S/B-correlations over time.

                  In this post, I used 30 years. All results easily extend to longer horizons.

  3. It seems like there should be an insurance product to alleviate some of this risk. Annuities theoretically could play this role, but seem to have their problems for whatever reason. One would think that as the FI community grows over time and there are enough people in different phases of the life cycle, some type of insurance product against SRR would be feasible and make sense.

    1. I’m not OP, but you can always hedge your downside with put options, futures or CFDs. Alternatively, you could swap your variable equity returns for a fixed return, although I’m not sure that option is available to retail investors.
      The whole problem with this approach is that you start getting into timing the market, which you have a high probability of getting wrong and underperforming the market.

      1. Good point! Option markets are very liquid for maybe 3 months into the future. But you don’t need to hedge short-term volatility. This is more about a prolonged downturn. Hard to hedge against, both for big investors and retail investors!

    2. That would be an application for the FinTech world: Match retirement savers and retirees to hedge out the SRR, which is exactly a zero-sum game. Of course, you could also do the same with long-dated equity options. But the market for those is pretty thin.

  4. Thanks for the shoutout about the “mortgage your retirement” comment! For a look at what happens when this strategy goes very, very wrong, there’s an “entertaining” (well, entertaining like watching a train wreck) thread over at Bogleheads documenting a young graduate who decided to take on tens of thousands of dollars in credit card debt and dump it into the market leveraged at 5-10x… a few months before the financial crisis. He covered his margin calls with ever-more-desperate sources of credit and ended up hundreds of thousands of dollars in the hole by the time the market bottomed out. Luckily the story has a happy ending — as of his last posts in 2015 it looks like he made it out of the hole without defaulting and has rebuilt a substantial portfolio.

    1. Oh my! What a cautionary tale! That’s why I prefer the “Mortgage your Retirement”-Light version as described in the post. Never invest and leverage income from years and decades in the future! Maybe, maybe, maybe if you have a safe government job. But not when working in finance!

  5. “And for statistics wonks, yes, this is a Newey-West adjusted t-stat to account for the overlapping windows.”

    You rock!

    “…SRR matters more than average returns: 31% of the fit is explained by the average return, an additional 64% is explained by the sequence of returns!”

    I guess this is quite intuitive, as a straight line is always going to be a worse fit than a series of breaks in the trend line. This links quite nicely back to your post on the CAPE rule for SWR and why it worked so well, and reminds me of the formula used by Jack Bogle to explain decade-long returns:

    All is useful as a basic thumb rule of what may lie ahead to adjust SWR/portfolio allocation acordingly.

  6. I feel like the plots in Part 3 showing different optimal asset allocation for different levels of CAPE, and this post both indicate that there should be a great benefit to actively adjusting asset allocation using valuation or trend-following. Active management tends to give up some returns but generally does a good job of reducing drawdowns. I would absolutely love to see a post that examine SWR using some form of tactical asset allocation.

    1. Ha, brilliant suggestion. I have been mulling over how to write a tactical asset allocation post. Conceptually it’s easy, practicing it is really hard. The Prime Harvesting (Part 13) has some valuation flavor in it. But it’s really naive.
      But I agree, we have to eventually go to a double-active SWR rule: The allocation changes in response to valuation/momentum/volatility and the withdrawal rate changes in response to valuation. That’s the holy grail and I’m still thinking about how to do that right. 🙂

    2. Darrow Kirkpatrick looked into using CAPE10 and was in favor of it:

      I reran his tests using some different data & assumptions and was less impressed:

      And Wade Pfau has written a paper about using CAPE10 “Long-Term Investors and Valuation-Based Asset Allocation” which found that “On a risk-adjusted basis, valuation-based strategies provide comparable returns but with substantially less risk than a 100 percent stocks buy-and-hold strategy, and comparable risks but with much higher returns than a 50/50 fixed asset allocation strategy.”

      I haven’t seen any retirement research that looks at stuff like trend following, dual momentum, etc, etc. Though the non-retirement research on it looks interesting.

      1. jp6v,
        Given your postings about Prime Harvesting on, I am wondering what are your thoughts on McClung advocating the use of iCAGR20 and Tobin’s Q as additional inputs along with CAPE10 when he forms a “retirement-valuation metric” to set an initial WR in Chapter 9 of his book?

        I would ask Big ERN the same question, but I suspect he has something coming for us in his forthcoming “holy grail” post he casually mentioned in a previous reply. 😛

        1. Unfortunately, I don’t have any strong feelings one way or another. I’m dubious about CAPE10, so using other metrics seems like a good way to limit the potential error. McClung wrote that valuations seem too vague to use but too useful to ignore — which seems a good description of the current situation.

          I think a better course is to explicitly factor in expected returns in your future withdrawals. If you use something like VPW it has a parameter for the “rate”. The VPW spreadsheet just uses the historical average for equities and bonds. But I think using current expected returns is a better approach. The trick is then….how do you figure out those expected returns?

          Research Affiliates and Robeco are two firms that put out detailed expected returns guides. RA is the more conservative of the two. At the moment RA forecasts that’s MY portfolio (i.e. my exact set of holdings) has an expected (real, geometric) return of 2.55%. I check their website and update it every 5 or 6 months.

        2. I’m a believer in valuation. I would like to mix in some momentum and vol control as well. Still thinking about how to best combine this systematically to manage actively. It’s the withdrawal rate and the stock/bond mix.

  7. One thing I’ve been wondering about recently is whether things like margin, portfolio lines of credit, stand by reverse mortgages, or home equity lines of credit could play a role in minimising (or totally alleviating) sequence risk. Something like “if the market crashes then for the next 5 years your withdrawals come out of leverage” in order to protect the portfolio. Yes, you’d be paying interest on that. The numbers (should) be small enough that margin calls aren’t a problem. I dunno….I have to think about it more.

    1. That’s a great point. And the answer is yes. I think this is the equivalent to Prime Harvesting (start with high bond share, then live off bonds for a while during recession). Using credit would merely mean you start with more equities and start using credit once you hit the 0% bonds.
      As you mention: interest rates are higher on credit, of course. But you also have less opportunity cost from low bond returns during good times.

  8. Fascinating. Those early years in retirement are so crucial.

    One way to combat having to “sell low” from the portfolio in that first decade is to start with a partial retirement. Find away to have some income — at least enough to cover some of your expenses until you get to a point where you feel beyond comfortable that your portfolio could weather any storm. For example, you could start a blog 😉 Or maybe a brewery.


    1. Thanks for stopping by, Dr. PoF!
      I like the idea of the part-time smooth transition into RE. Of course, the problem still persists if the market plunge happens once we go from part-time work to full RE. But at least we would have bridged a few more years and not touched the principal during that time. That could make all the difference in the long-term!

  9. Amazing how big of an impact SRR has. That first chart showing the difference from retiring in 1968 compared to 1979 is startling!

    That is an interesting concept about mortgaging your retirement. Conceptually it makes sense, but seems risky. If you use margin early on and SRR backfires on you, the financial stress and thought of BK would be daunting.

    Also, the mortgaging your retirement is only helpful in the accumulation phase. But would you suggest in the retirement phase to tap home equity to cover cost of living in the first decade of retirement if SRR hits you wrong? Then, instead of draining your investments at the low, you could sit on them while they bounce back and repay the home loan with later good year.

    1. Thanks! It turns out that for folks who are far along on the path to FIRE, this mortgage your retirement business is too late already. We already have a sizable equity stash. This is more an issue for people who start out with $0.00. But even then it’s way too risky, even for someone like me. But the simple rule like don’t accelerate the home mortgage paydown and start aggressively with 100% equities when young (but without leverage), sounds like something workable!

  10. ugh. Goodness. I read this and my blood pressure rises just a little as my date gets closer and closer. But I think it’s good to have these things in mind. Here are my practical takeaways and I’m interested in your opinion: 1) keep my largish HELOC open for flexibility. We can’t time the market, but when there is a profound drop in market prices where things drop below the historical mean don’t sell shares to meet cash flow requirements, borrow at low interest rates. Even if I didn’t have a speck of cash flow from another source, this should sustain things for 3-4 years. 2) If a severe 2008-2009 market correction happens, consider getting a part time gig to meet some of those cash flow needs to ride out the SWR risk. 3) As time goes on in your retirement, the SWR risk drops?

  11. Nice work, although I would have liked to see a more thorough discussion of how variable withdrawal methods mitigate (if not eliminate) SoR.
    About VPW, I would suggest that you read the full discussion starting here: I see the point of the author, although I also see why some might disagree with his classification of risk categories.
    In any case, something like VPW (or G-K or others) does mitigate SoR quite a lot, and this is really worth studying in more depth. Yet another reason for which nobody in their right mind should use a constant (fixed) withdrawal method.

    1. I looked at the discussion. I agree 100% with the user “longinvest”
      I looked at the different scenarios (+10%/+10%, -15%/+42%, +42%/-15%) in my post and the IRR when using VPW with a 10% withdrawal very period would be: +10%, +8.7%, +11.7%, respectively.
      So, the IRR is much lower (due to lower withdrawals when returns are low initially) when you’re unlucky and they are much higher than 10% if you are lucky with high returns early on.
      As I wrote before: there is no hiding the SRR. Sure you can hide the SRR by looking only at the final value. But that’s not proper, in my view.

      1. I tend to agree. I understand longinvest’s point but feel that it is almost semantics. “Sequence of Returns Risk only means what happens to the final portfolio value”. Well, yeah, okay that *is* how people use it. But that’s, in large part, because so much retirement research is based on SWRs and final portfolio value is the only thing to really look at. If you told a retiree “hey, the market crashed and your annual withdrawal is 40% less this year….but you’re not exposed to sequence of return risk, honest!” …. they’d probably punch you in the face for being a pedantic smart alec.

          1. OK, but for some of us, it’s actually quite reassuring to have a simple way to ensure the preservation of the portfolio value from a known cause of ruin (or as you point out, the 3 top causes of ruin). I think in particular this may appeal to a particular class of retirees, those who have sufficient resources and willingness to have a very large chunk of variable expenses in their budget.

            I’m the one who made the original comment on the previous post about VPW. I now understand the error, so thanks for the clarifications.

            It may be shortsighted to focus on portfolio value rather than on the total lifetime withdrawals. But as jp6v’s comment hints, so much of the commentary and research focuses on final portfolio value as the measure of success (or ruin) that it is a very natural way for us non-experts to think about the problem.

            VPW, or other PMT based or variable withdrawal schemes seem to put the sequence risk where we can see it and to keep it out of the portfolio value.

            However, I’d be very glad to learn that I am mistaken or that I am placing too much importance on this. I would not want to distort my decision making to preserve an illusory advantage. I’m on the verge of my own ER and so these are not abstract decisions.

            Thanks again for all of the great work and contributions.

        1. Yes, I don’t disagree with you guys, even if I understand longinvest’s point of view (he’s the author of VPW, by the way). Personally, the only thing that matters to me is the full sequence of withdrawals (hence my purchasing power) and even with VPW, a bad sequence of returns does impact the sequence of withdrawals, true. But the salient point is that, irrespective of the way one defines SoR, variable withdrawal methods (VPW or else) do mitigate the issue in a significant manner. Would be great to see ERN hone a bit more on this, with some quantification of the potential gains.

          1. As I outlined in the simple numerical example: the difference in IRR carries over to the VPW rule. It’s intuitive that the lowest IRR is a little bit higher than under the fixed % rule.
            That makes sense because you automatically lower your withdrawal at the bottom of the market. So, one could argue that VPW can theoretically alleviate a little bit the SRR.

  12. I’m having a hard time understanding mortgage your retirement. If you borrow and get hit early on, then you’d be in a big hole. I don’t see how that reduce SRR.
    I agree with your light version, though. I prefer to keep our mortgage and use the money to invest in the stock market.
    Good read.

    1. Hi RB40!
      The only thing the Mortgage your Retirement rule accomplishes is that you get the exact same IRR as the market over the investment horizon. It won’t be higher (as would be the case with an early drop) and it won’t be lower (as would happen with an early surge and late drop). But you still get the market risk over the investment horizon. That is something you can never avoid.
      But again: I wouldn’t have the stomach for going all in at the beginning of my career.

  13. Nice work ERN…
    The idea to borrow against my future earnings is interesting, yet too extreme for me. I even hold cash as emergency fund!

    SRR is on my mind as well. It sometimes makes me think that I should somehow time my move into Re: By that I mean, only go full work free when there has been a serious crash, a year or 2 of recovery and I still meet the 4 pct rule. That way. I could try to time being in a cohort that has no SRR risk. Maybe a weird and risky approach as there is no guaranteed that with that I eliminated the SRR (random walk of markets, could crash again the next day).

    That being said, our current view on the whole FIRE has changed sharply. It is more an approach where we do not wait to be FI to work less. We actually start to do this now. We are fully aware that it means working longer. It also might mean that we are less exposed to the SRR. In an ideal world, we work enough to cover the basic costs of living and the extras come out of the portflio. That should be way less than a 3pct withdrawal rate!

    agreed, it is not a pure form of FIRE, maybe it is just the way it would work for me, given my constraints…

    1. I like that hybrid approach of using the growing stash of wealth to slowly afford to work less. Instead of going from 100% work to 0 work.
      Also, that mortgage your retirement viewed as more of a thought experiment. To justify high equity share early on. I wouldn’t recommend this to most people. 🙂

      1. Going from 100% to 0% could be too much for me. Slowly going down and do a soft landing sounds better to me. (Mainly due to the fact that from 100 to 0 still takes me about 10 years…)

  14. “But at least we would have bridged a few more years and not touched the principal during that time. That could make all the difference in the long-term!”
    Hi. Given the above comment, I’m wondering why my Vanguard planner said, after running my #s, that my working another year or 2 or 3 makes negligible difference in my probability of success. Might it be because I’m 61 years old so not talking about 30 years before retirement? Hard to imagine that starting to draw down my portfolio now vs in 3 years wouldn’t impact my retirement success . Would appreciate your thoughts.
    Also, would appreciate any rules of thumb or guidelines for folks who aren’t finance or economic wonks, and who are less than 5 years from a ‘normal’ retirement (at age 65). Any comments about a SWR (is 3% safe for a 35-40 yr retirement)? Is a 50/50 allocation too conservative? Can’t I just spend from fixed income during a market downturn, and then rebalancing would be buying stocks when they’re cheap? Thanks for any help for us non-finance types!

    1. There is no simple rule fo thumb for this because I don’t know all of your parameters. Imagine you never expected any Social Security or pension. A horizon of 35 years vs. 33 years vs. 32 years would probably not make that much of a difference.
      Very different story if you are trying to hack Social Security: Claim benefits as late as possible (age 70) for the higher earning spouse. Now you are trying to bridge 9 years until you receive pretty substantial benefits. Just as an example, you expect the equivalent of 3% of today’s net worth of SocSec benefits. You could probably go as high as 5% or 5.5% initial withdrawal rate and then reduce to 2-2.5% once SocSec kicks in after 9 years. If you hold off retirement for another 2-3 years and you now face only 6 or 7 years of high withdrawals you can significantly increase your success probability or increase your withdrawal rate.

      And one other thing to consider: Even if you do not increase your retirement rate because the nest egg grows for 2-3more years, the retirement income will be higher.

      Why not play around with the Google sheet: See part 7 here:

      And the Google Sheet:


      1. Many thanks. I’ll take a look at those Google sheets. And yes, I’m planning to hold off on social security until age 70. Vanguard planner knows that I’d be taking more from my portfolio from retirement (planning was based on retiring at 63 1/2) until 70, but said working a few more years wouldn’t much impact probability of success…… Reading your series is causing me to question the entire VG analysis.
        If you could clarify something for me:
        We want to avoid retiring when the market is down/low returns early on. But high valuations are also problematic. What am I not understanding (a lot I’m sure)? These seem contradictory to me. Thanks for clarification.

  15. Hi

    I would just like to commend you on this very in depth analysis of safe withdrawal rates. the amount of thought time and effort that must’ve been put into these 15 posts is huge and very informative. I am well away from having to think about withdrawing from my portfolio, but I may have to bookmark this series and have a look at it again in 5 years 🙂

  16. I suppose this column might be interesting to statisticians. But it doesn’t tell me what to do.

    Earlier in your series you explained why 3.25 to 3.50 percent is a SWR, depending on this and that. That was helpful. I inferred that 3.00 percent is even safer, based on the same assumptions. But public equity prices continue to climb and I suspect we are now in, or are fast approaching, the scary slice of CAPE Shiller that implies SRR is about to hit us like a freight train. So what do we do now if we are age 60 or so and almost retired?

    1. Check out the exchange between Qawsed and me. I included a link to the Google sheet to play with your own simulations. The problem is that there is no obvious one size fits all solution. 3.5% is good for early retirees. Traditional retirees can go higher. Especially when they take SocSec at age 70 and have to bridge only a few years with their savings. If the withdrawals are much lower after age 70 when SoSec kicks in then your initial withdrawals can be much higher than 4%. Probably higher than 5%. I recently did a study for the podcast where someone wants to retire at age 51 and expects a generous Social Security benefit at age 70. 5% initial WR, then reduced by SocSec at age 70, is doable.

  17. In some ways, being able to pay off a low interest rate mortgage or invest in a portfolio of less than 140% equities is a luxury. i.e. you don’t have to maximally leverage your life in order to meet your financial goals.

    Some people would rather drive a Tesla and vacation in Paris than be able to have a paid off home or be able to take less risk with their investments. To each their own.

    But the best way to deal with sequence of returns risk in my opinion is to simply maximize flexibility and adaptability. The means saving a lot of money, minimizing fixed expenses, and keeping an eye on the portfolio as you spend from it and making adjustments accordingly.

  18. Serious kudos for a great series ERN. The level of thought and effort are exceptional. I’m not in a position very different from you in many ways it seems, and have researched and struggled with many of the questions you address. Mathematically ER is on the cusp of viable for us — but with uncertainty and concerns.

    This series has added significantly to thinking about two of my three largest concerns: 1) what is a reasonable target initial withdrawal rate – i had been working with a 3.3-3.4% target from more simplistic math but lack the deeper statistical skills you have to be confident in that, so this has been invaluable work you’ve contributed to many of us! 2) Sequence risk scares the heck out of me, and the one thing that would drive me to “one more year”-itis. I’ve been concerned how much downside risk to account for here (including engineering spending flex options), but without a clue how to start to even think about it never mind quantify it, so thanks for tackling this topic and putting thought and analysis behind it. 3) Healthcare costs, but…….yeah. 🙂

    Great job and thanks again for publicly contributing your time, skills and work!

  19. Hi, only recently came across the whole ‘FIRE’ scene, so now trying to consume as much information as possible. Thanks for writing this series of articles. Have found them very informative.

    As someone still relatively early on in the acquisition phase, I have been thinking more about the savings side of the equation and while i know trying to time markets in general is a fools errand, i wonder whether you had any thoughts on using something like CAPE to adjust the allocation of your investments. Was thinking along the lines of mechanically diverting a proportion of regular savings to ‘safe’ assets (high interest cash accounts, premium bonds etc.) when valuations look expensive, then re-allocating that pot towards equities as markets fall / valuations look better, or perhaps making a lump sum on every x% draw down in equities…Will fire up excel and do some simple analysis.

    1. If you find anything let me know. The experience from the late 90s is that if you shift out of equities when the CAPE gets high you may miss the tail end of the rally because you move out too early. Even under the Bogle Scenario (part 16) where the CAPE slowly converges back to normal, your returns will be higher with equities than with bonds. So, in the accumulation phase, using CAPE to time the asset allocation is not that helpful.
      You might find the CAPE helpful in timing what equity market (US vs. non-US) you find more attractive. That might work…

  20. I’ve been trying to learn what I can about SORR. My question is what happens to a portfolio say 3.3m , and a swr of 3% and a given AA. Starting on day 1 you withdraw 60 months of living expense or 500K in this case and put it into some low risk vehicle say a CD ladder thereby creating a homebrew self funded annuity. You let the portfolio ride with attention to re-balance or even glidepath to a different AA as you grow older. This would give you 5 years of living expense with an unmolested portfolio. At 5 years again retire and re-evaluate your initial retirement amount and again withdraw 5 years based on the new calcs or stick with 3% and live another 5 years. etc, ie various possible scenarios with their own risk. You would only open the portfolio 6 times in 30 years instead of 30 times. The likelihood of opening the portfolio in recession would also be reduced since recessions don’t typically last 5 years. Re-balancing tends to reduce SORR or so I’ve read. Do you mitigate SORR to some extent by limiting the sequence to epochs longer than 1 year? SORR to me looks like compounded risk during a series of withdrawals.

    Your initial retirement would be during a period of adequate funding since you wouldn’t retire unless forced in the middle of a recession. It might be possible to “fudge the sequence” if your subsequent annuity funding occurred at an inopportune time by one offing a years living expense thereby placing you in a difference sequence.


    1. That strategy may or may not work. What if you’re unlucky and your first 5Y window runs out exactly a the bottom of the market? You’d then withdraw a big chunk and miss the boat on the rebound. But as I wrote in the glidepath posts, (here and here), it could be worthwhile to hedge the initial 5 years with some bond holdings and then switch more and more into equities over time. This glidepath approach works particularly well when the CAPE is elevated (>20).
      Thanks for stopping by!

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.