June 16, 2023 – I wonder if I’ll ever run out of material for the Safe Withdrawal Series. Fifty-eight parts now, and the new ideas come faster than I can write posts these days. This month, I initially planned to write about the effects of timing Social Security in the context of safe withdrawal simulations. But one issue keeps coming up. It’s almost like a personal finance “zombie” topic that, after I thought I put it to rest once and for all, always comes back when you least expect it. It’s flexibility. If we are flexible – so we are told – we don’t have to worry much about sequence risk. We can throw out the 4% Rule and make it the 5.5% Rule. Or the 7% Rule or whatever you like.
Only it’s not that easy. In today’s post, I like to accomplish three things:
- Provide a simple chart and a few back-of-the-envelope calculations to demonstrate the flexibility folly.
- Comment on a recent post by two fellow personal finance bloggers and showcase some of the weaknesses of their approach.
- Propose a better method for modeling flexibility and gauging its impact on safe withdrawal amounts. Hint: it uses my SWR Simulation tool!
Let’s take a look…
Why flexibility is overrated: One chart and a simple back-of-the-envelope calculation
Before I even get too far into the weeds, let me briefly demonstrate the intuition for why flexibility is overrated:
1: A simple chart to showcase the flexibility fallacy
Imagine we’ve determined that over a certain retirement horizon, a fixed 4% initial withdrawal rate is indeed the historical failsafe, i.e., for the historical worst-case retirement date, likely in 1929 or around 1964-1968, a 4% initial withdrawal amount would have exactly depleted the portfolio.
The flexibility crowd now tells us that we can start with, say, a 5.5% initial withdrawal rate and then subsequently just be flexible and make some small adjustments, like temporarily curbing consumption and/or getting a side hustle, etc. Of course, we already know that if 4% is the failsafe, then consuming 5.5% over the entire retirement horizon will not be safe (red line). And using the purple line withdrawal path, you still withdraw more than under the failsafe 4% figure every year. Thus, you will still fail. It’s a mathematical certainty – no simulations necessary. So, we know for sure that if we start above 4%, then the flexible withdrawal path must cross to below 4% at some point to make up for the initial excess withdrawals (see the green line).
So, if you like to raise your withdrawals early on, you may face some very deep and potentially prolonged spending cuts later in retirement. It’s like squeezing a balloon!
Let’s look at an example with some concrete numbers…
2: A simple numerical example
Imagine we have a FIRE enthusiast couple with a 40-year horizon, an 80/20 portfolio, i.e., 80% stocks and 20% intermediate government bonds (10-year US Treasurys), and a zero final value target. The retirees have an $80k per year budget and saved $1,450,000 so far. Well, applying the 4% Rule, the portfolio target is $2,000,000, so they are still $550,000 short of reaching FIRE. But don’t let your hearts be troubled: flexibility to the rescue: Our retirees read on the interwebs that “if you’re flexible, you can raise your SWR to 5.5%.” $80k divided by 0.055 gives you a savings target of $1,454,545, and rounding that down to $1,450,000, our FIRE couple reached their savings target, potentially years before they’d ever imagined. What’s not to love about flexibility, then? Well, the math doesn’t add up.
First, the 4% Rule doesn’t even work over a 40-year horizon. The historical failsafe would have been 3.43% for the cohort that retired right before the September 1929 stock market crash. A 4% withdrawal rate would have had a 7% failure rate overall (6.94% post-1926) and a 25.07% failure rate conditional on the Shiller CAPE above 20. Today’s Shiller CAPE is just under 30, by the way.
So, 4% is not very safe. A 5.25% withdrawal rate would have had a two-thirds failure rate if the initial CAPE had been elevated, and a 5.5% initial rate would have failed 75% of the time (not displayed in the table, but take my word for it). Conditional on an elevated CAPE, not even accepting a modest failure probability gets you only slightly closer to 4%. At 1%, 2%, and 5% failure probability, we’re looking at 3.63%, 3.66%, and 3.73%, respectively.
So, retiring on an $80k p.a. budget with only $1.45m, you might need a whole lot of flexibility. How much flexibility? Well, there is a wide gap between reality and what is often advertised as the necessary degree and duration of flexibility. For example, flexibility is often marketed as a very short-term thing, where you curb your discretionary spending, but only during bear markets, which usually last only 1-2 years. Compare that to reality. At a 3.43% SWR and with a $1.45m portfolio, your actual safe consumption level would have been only about $49,800. That’s a whopping $30,200 a year below your $80k budget – every year for 40 years, not just during bear markets! So, if you have the flexibility to curb your consumption by about 38% every year during your entire retirement, then go ahead. But I would find that very unappealing.
Another way to gauge how much you need to curb your withdrawals is to compute the failsafe portfolio you’d need for that $80k-a-year lifestyle. It’s $80k divided by 0.0343, i.e., $2.332m, not $1.450m. We have a shortfall of about $882k. And we will not overcome that kind of a shortfall by simply eliminating our bar and restaurant budget and working as an occasional barista on the weekends. $882k looks like multiple years of full-time employment for most people. If you currently have a well-compensated job you don’t completely hate, then you might as well keep working for a few more years and avoid this flexibility trap.
Why 5.5% isn’t the new 4%: My opinion on the MadFientist & Nick Maggiulli flexibility approach
In a recent post on the MadFientist blog, together with Nick Maggiulli of the Dollars and Data blog, the two authors (let’s abbreviate them MF&NM) proposed a “discretionary” withdrawal rule where you change the withdrawal amounts based on equity drawdowns. The idea is that, in retirement, you should have the flexibility to vary your discretionary spending if needed. If you can occasionally reduce or even eliminate your discretionary spending, you can start with a higher initial withdrawal amount. Say, if the stock market is at or at least near an all-time high (i.e., within 10%), you withdraw your full retirement budget. If the stock market is in “correction territory,” i.e., between 10% and 20% off the recent high, you cut 50% of your discretionary budget. And if you’re in “bear market territory,” defined as 20% or more below the recent high, then you completely eliminate the discretionary budget.
Let’s take the following example: imagine 50% of your expenses are mandatory and the remaining 50% are discretionary. We use the same example as above: a 40-year horizon and an 80/20 portfolio. MF&NM now proclaim that a 5.5% initial withdrawal rate is feasible. Let’s put that to the test.
Let me first recreate their results. Since the post on the MadFientist blog provided only few details, I’ll have to make some assumptions, like:
- As always, I use the S&P500 (and predecessors) total return index for equities and the 10-year U.S. Treasury benchmark bond index for the diversifying asset.
- I adjust the equity index with CPI inflation to determine the real drawdowns. Notice that this is a conservative estimate on the drawdowns because if we base the all-time high on the month-end values only, we miss some much higher index values intra-month. So, my drawdowns might be shallower and shorter than what you find when you factor in daily closing index values or even intra-day values.
- I use monthly return data from 1871 to 2023. My results will differ slightly from the MF&NM results because their annual return data will likely miss some of the historical worst-case scenarios. For example, the annual return data won’t capture the August 31, 1929 retirement cohort, often one of the worst retirement cohorts on record.
Let’s look at the historical drawdowns time series; see the chart below. The top is the S&P 500 cumulative return, adjusted for CPI inflation. The usual disclaimers apply regarding the historical data in the pre-S&P 500 and pre-Composite index era. We see a nice steady drift of about 7% annualized. But it was a bumpy ride! The index has spent considerable time in the correction and bear market territories, see the chart on the bottom!
In fact, if I plot the percentage of months that each retirement cohort had spent in the three consumption scale buckets (100%=close to peak, 75%=correction, 50%=bear market), we notice some very unpleasant issues:
- The average retirement cohort since 1925 (about the time when MF&NM started their simulations) got to spend the full amount only 51% of the time. About 14% of the months, you were in a correction, and during the remaining 35%, you were in bear market territory.
- Thus, the 5.5% withdrawal rate applies only about half the time, the 4.125%(=5.5%*0.75) withdrawal rate applies 14% of the time, and 2.75%(=5.5%/2) applies 35% of the time. So, the weighted average withdrawal amount isn’t anywhere close to 5.5% but only about 5.5% * (0.51+0.75*0.14+0.5*0.35) = 4.35% of the initial portfolio; that’s a 21% haircut. It would have been nice if MF&NM had pointed this out in their post!
- The prevalence of deep and extended bear markets has increased since the 1920s, so by extending my study to that early period, all the way back to 1871, I may get slightly different overall results from MF&NM.
- Even though the 40-year distributions over the three buckets are roughly the same across cohorts, different cohorts have very different experiences over the first 15 years of retirement. If you were unlucky enough to retire close to the market peaks in 1929 or between 1964 and 1977, you would have spent the majority of your first 15 years in retirement with a zero discretionary budget. For example, in 1929, you’d have spent twelve out of the first fifteen years in retirement scraping by and spending only on mandatory categories without any discretionary budget. It would have been nice if MF&NM had pointed this out in their post! So, the narrative that flexibility is just a short-term inconvenience goes out the window. And good luck finding a job if we ever experience a repeat of the 1982 or even 1932 job market!
- The discretionary withdrawal rule doesn’t eliminate Sequence Risk. The “bad luck” cohorts in 1929, 1964-68 are all the “usual suspects,” i.e., the cohorts that retired right around their prominent market peaks.
But it gets even worse. In the table below, I display the success probabilities of different baseline safe withdrawal rates, i.e., without discretionary spending cuts.
A 5.5% baseline WR had a 92.8% success probability in my simulations. Compared to 98% in the MF&NM table. Their success probabilities are much more aligned with my 30-year simulations. I’m not sure why. Three explanations:
- They might have done a little switcharoo and accidentally shifted to a 30-year horizon.
- They might have used a different spending rule with an even bigger average haircut than my 21% to push up the baseline consumption by a quarter point. For example, I use only monthly data for the equity returns, and the drawdowns are only relative to the monthly closes. If you pin the drawdown rule to the daily closes or intra-day highs, you will generate steeper drawdown stats. But also slightly higher baseline withdrawal values. It’s the squeeze-the-balloon effect again!
- Because I use the pre-1925 data with cohorts that had less severe discretionary spending reductions, some of my SWRs are quite low. The failsafe withdrawal rate was 4.84.
But in any case, with a 4.84% failsafe, and after we apply the 21% haircut to account for all the months when we have to reduce consumption by 25% or even 50%, we’re left with only 3.82%. Sorry: not 4%, and certainly not 5.5%. We can’t miraculously take a sub-4% safe withdrawal rate and turn it into a 5.5% rate, either. You can put lipstick on a pig, but it’ll still be a pig.
Just a side note: there is a widespread myth in the FIRE community that the failsafe withdrawal rate no longer drops when extending the retirement horizon beyond 40 years. MF&NM allude to this, and Kitces also has an article making this whacky claim. Let me break the news for everybody: Say, if 1929 generated the lowest sustainable withdrawal rate over 40 years, as it often does, then that rate would have exactly exhausted the portfolio by 1969. If you tag on another 10 or 20 years of retirement, you must reduce the initial rate to have sufficient funds left in 1969 to fund the additional retirement years. It’s a mathematical certainty! It’s also an empirical reality: the failsafe further declines between 40 and 60 years, albeit slower (5.23 to 4.84% to 4.57% to 4.42%). This artifact is true for both the discretionary rule as well as the plain old fixed safe withdrawal rate analysis, Trinity-Study-style. For your enjoyment, I also enclose the same table for the fixed safe withdrawal rates over 30, 40, 50, and 60-year horizons. The failsafe drops from 3.64% to 3.43%, to 3.26%, and to 3.14% as we move from 30 years to 60 years. It’s indeed possible that SWRs don’t change much when targeting capital preservation and going from 40 to 60 years. But with capital depletion, you still see noticeable declines between 40 and 60 years!
But let’s move on! Next, I plot several case studies with actual withdrawal amounts in the chart below. The blue dots are the annualized monthly withdrawal rates generated by the MF&NM discretionary method—all rates as a percentage of the initial portfolio (adjusted for inflation). The red line is the 12-month moving average, and the black line is the fixed withdrawal rate. Both the fixed and the discretionary withdrawal rates are computed to deplete the portfolio over 40 years exactly. Notice the timing assumption: The September 1929 cohort would have started withdrawing on August 31 of that year.
A few observations:
- Notice how the discretionary method would not have succeeded using a 5.5% initial withdrawal in 1929, 1965, and 1968. Those three cohorts would have required a baseline withdrawal amount equal to 5.00-5.25% of the initial portfolio. In other words, 5.5% would have run out of money despite the discretionary spending pattern.
- When using a fixed SWR, 1929 is the worst retirement start date over a 40-year horizon; SWR=3.43%. Again, notice how the discretionary rule would have required most of the first retirement half to cut your discretionary spending completely or at least by 50%. I don’t call that flexibility. That’s either back to work or deprivation!
- Both cohorts in the 1960s would have generated a safe withdrawal rate just above 3.5%. The discretionary method would have started above 5% but dropped to 2.5% because of the steep real equity drawdowns in the 1970s.
- In 1972, right at the market peak, a 4% Rule would have indeed succeeded. And the discretionary method could have even pulled off a withdrawal rate North of 5.5%. But the MF&NM method would have withdrawn below 3% for almost the entire first decade. Not a very appealing strategy!
Comparing variable withdrawal paths: a utility-based model
I noticed that the arithmetic average withdrawal amount over 40 years using the discretionary method is slightly above the fixed withdrawal rate. Thus, admittedly, the MF&NM method might offer a modest hedge against Sequence Risk. By definition, you withdraw less when stocks are down and more when stocks are rallying. But the mean withdrawals over a 40-year horizon are only a very poor measure of my personal preferences. Here are the two reasons:
- Time preference: I don’t like the idea of withdrawing less during early retirement and then backloading the withdrawals later in retirement when I’m in my 80s. I prefer the other way around!
- Risk aversion: I prefer a stable and predictable consumption path rather than a volatile one. In other words, the mean disguises the crazy fluctuations as I plotted in the case studies for the 1926, 1965, 1968, and 1972 cohorts.
Now, how do we compare withdrawal paths that are not fixed? We now have 480 moving parts, and it sounds impossible to compare two competing withdrawal paths. Well, it’s actually very simple; this is a well-known problem in economics and finance. We use a utility function to model time preference and risk aversion. The risk aversion comes in through a concave utility function and the time preference through discounting. Then, the utility of a path of T withdrawals w(0),…,w(T-1) is
As is customary in much of economics and finance, I use a simple Constant Relative Risk Aversion (CRRA) function of the form:
Notice that for gamma=1, this reduces to just a plain (natural) log-utility function, compliments of L’Hôpital’s rule. Since utility is just a unit-free, hard-to-interpret measure, we can also translate the utility of any volatile withdrawal path back into one fixed number equal to a “fixed-consumption-equivalent” number, i.e., calculate a fixed and level withdrawal amount w_bar that would have given you the same utility as the volatile one:
Now we can calculate this consumption equivalent utility for all the MF&NM discretionary spending rule paths as well as the fixed safe withdrawal rates. For the latter, the w_bar is obviously just the fixed rate itself. I also assume that beta=0.96 (p.a.), i.e., you care 4% less about next year’s utility than this year’s utility. I also consider four different gamma values.
- gamma=0 implies linear utility, thus, risk-neutrality. This will not apply to most folks except for maybe Sam Bankman-Fried. But it’s a good benchmark.
- gamma=2 implies a very high degree of risk tolerance. For example, when I optimize glidepaths, a gamma of 2 will often imply an optimal equity weight of 100% for the entire accumulation phase. Very few investors will have that kind of risk tolerance, so I view this as a good lower bound on gamma.
- gamma=3.5 implies a moderate risk aversion. Most actual glidepaths used in target date funds by Fidelity, Vanguard, T. Rowe Price, etc., look like they came out of dynamic programming code optimizing stock/bond glidepaths and using that sort of utility function. Thus, that’s likely a pretty decent assumption for the average U.S. investor.
- gamma=5.0 implies a very low risk tolerance. In my glidepath optimization research, I found that the shift out of equities and into bonds starts much earlier than in most industry target date funds. I use 5.0 as the upper bound for gamma and the lower bound for risk tolerance of most investors out there.
Let’s take a look at the time series of the relative consumption-equivalent utility numbers; see the chart below. The way to read this chart is to note that, for example, for the Jan 1925 cohort, a risk-neutral investor would have preferred the discretionary withdrawal path over the fixed path; by about 3%. But with a modest risk aversion (gamma=3.5), the discretionary spending pattern would have been equivalent to close to 14% under the fixed withdrawal rate rule every single month. Thus, looking at the chart, we find that assuming risk-neutrality, the discretionary spending rule improved your retirement utility by maybe 5% on average. But using a more realistic parameter for risk aversion, we find that folks with high risk tolerance are still worse off with the MF&NM rule by about 4% on average. Moderate risk-aversion investors lose about 10%, and very risk-averse investors lose about 15%. I would stay away from this discretionary rule!
Here’s a better way of modeling flexibility
It’s one thing pointing out holes in other people’s analysis. But as a professor of mine always used to say, “It takes a model to beat a model.” So, instead of just dumping on other people’s work, let me propose how to account for flexibility properly. No shiny objects, no hiding skeletons in the closet. Just complete transparency and plain and easy-to-understand analysis.
Let’s stay with the numerical example but extend the horizon to 50 years. Most people who retire in their early-to-mid forties might want to plan for a retirement lasting that long, especially when accounting for joint survival probabilities.
The safe withdrawal rate is now down to 3.26%, which means the safe withdrawal amount out of a $1.45m portfolio is only $47,303.
How do we get to $80k/year with flexibility? It won’t be easy, and there isn’t one single solution that gets us there. But here would be six steps to accomplish our task:
Step 1: Account for Social Security.
First, let’s assume that both spouses are 45 years old when they retire. Assume spouse one claims benefits at age 62 (month 205 of retirement) and expects to receive $1,000 a month, while spouse two claims benefits at age 70 (month 301 of retirement) and expects to receive $2,500 per month. This could be the typical spousal lifetime benefits maximization outcome where the higher earner defers up to age 70, and the lower earner claims ASAP. See opensocialsecurity.com for a useful tool.
I enter those values in my Google Sheet (see Part 28 for the link and manual), specifically in the tab “Cash Flow Assist.” Accounting for those benefits, we can shift up the SWR as a percentage of the initial assets to 3.85%, or $55,845 a year. Still far away from $80k, but we are just getting started!
Step 2: Account for lower expenses later in retirement.
People often point out that many retirees don’t keep a level consumption profile. Most retirees slow down and spend less later in retirement. Assume that starting at age 75, the two retirees only spend 90 cents on the dollar and, at age 80, only 80 cents on the dollar relative to the initial baseline. The way I can model this in my worksheet is to change the scaling in column “S” in the “Cash Flow Assist” tab. Change that to 0.90 in months 361-420 and 0.80 in the subsequent months. The SWR as a percentage of the initial assets is now 3.92% or $56,855. I’m amazed at how little of a difference this makes. Another retirement myth is busted: If you’re in your 40s and you account for lower spending later in retirement due to slowing down and traveling less, it doesn’t make much of a difference in the SWR calculations! But also note that this result works both ways: higher medical and nursing home expenses later in retirement won’t make a big difference either!
Step 3: Account for the current equity drawdown.
Despite the recent recovery in the stock market, we’re still significantly below the Jan 2022 highs. Conditional on a 10-15% drawdown from the most recent peak, we can now raise the SWR to 4.50%, or $65,276 annually. A caveat: even with the drawdown, we’re still at a very elevated CAPE ratio. But it’s been pointed out that the CAPE ratio is difficult to compare across time, so probably the equity drawdown is the better valuation metric for our SWR purposes.
Notice how my approach operates very differently from the MF&NM methodology. Even with a 5.5% baseline withdrawal rate, you’d still have to curb the initial discretionary spending by 50% because we’re currently still in the “correction territory,” between 10% and 20% off the recent peaks (about 6% below the peak in nominal terms, but about 14% when adjusting for CPI inflation). Then, the MF&NM methodology only affords you a 4.125% initial withdrawal rate as of June 2023, much lower than what I would have recommended. So, don’t call me the conservative retirement planner! Actually, MadFientist and Nick Magiulli are currently recommending a lower initial withdrawal amount than even I would! You should also read Part 54, where I outline that with slightly depressed equity valuations, you can raise your initial SWR to well above 4%, even almost 5%. Significantly higher than the 4.125% current recommendation that the MF&NM model would currently recommend.
Step 4: You’re fine with a 2%-5% failure probability.
Nothing is certain these days, so why would you target an absolutely certain and safe retirement (at least if measured by historical return patterns)? If you’re fine with crossing your fingers and hoping that your following your retirement date the market doesn’t repeat the 2-5% worst historical retirement cohorts, then you can increase your withdrawal amount somewhat. I personally prefer to work off the failsafe probabilities, but maybe others are more comfortable with that kind of risk. At 2% and 5% failure probabilities, we’re now looking at $67,114 and $68,789, respectively or 4.63% and 4.74% of the initial portfolio. It’s not really much of a difference.
Step 5: Spending reductions and/or side hustles.
Instead of promising you 5.5% and hiding the gnarly spending reductions behind a big curtain, the more honest approach is to ask yourself: how much flexibility am I willing to offer to increase my baseline retirement budget? Well, let’s assume that our two retirees are OK with lowering their withdrawals by up to $2,000 per month for a maximum of five years. Say, half of that comes from a reduction in discretionary spending and the other half from a side hustle. Let’s input a $ 2,000-a-month inflow in the “Cash Flow Assist” tab. Reading off the safe withdrawal amounts from the table, using the 2% and 5% failure rates, and conditional on a 10-15% equity drawdown, I get $72,067 and 73,403, respectively. Not quite there at $80k, but considering that we started at $47,303, we’ve already closed 80% of the gap. And by the way, the $2,000 in spending reductions are not set in stone. You only keep that as an option. If the market performs well over the first year or so in retirement, you may rerun the safe withdrawal toolbox, and maybe at that time, the $80k per year becomes sustainable again without the side hustle!
Step 6: More sacrifices
Given that $80k is quite a big chunk of money, there will probably be some expenses that can be cut in the long run. If we can find $700/month in spending cuts starting in year six and going forward, we can push the safe withdrawal amounts for the 2% and 5% failure rates to $78,942 and $80,276 annually. Maybe forego a hobby later in retirement. Move to a cheaper area. $8,400 a year out of an $80,000 budget is not the end of the world. And again, this $700 spending reduction is only optional if the market moves against us. With a high probability, we can still plan to enjoy an $80k annual retirement budget without the side hustle or this sacrifice. But even in that worst-case scenario, we’ll still have a generous discretionary spending budget.
There we go; we reached the $80k target. Notice that the significant retirement budget increases came from sources that MF&NM completely ignore: Social security income later in retirement and more attractive equity valuations. Another valuable option not even mentioned here would be liquidating your primary residence later in retirement, either directly or through a reverse mortgage.
In contrast, throwing in some side gigs and spending reductions here and there didn’t make a huge difference. And it’s not for lack of trying: Step 5 lowered withdrawals by $120,000 and Step 6 by a total of $378,000. But even that doesn’t raise the safe withdrawal amount by 37.5% (=moving from 4% to 5.5%). Thus, if you want to rely on spending reductions only, then most retirees probably don’t have the patience to suffer through prolonged stretches of deep spending cuts inherent in the discretionary spending rule proposed by MF&NM!
It’s a much better and more honest approach to explicitly gauge how flexible you can be, i.e., how much spending reduction and/or side hustle income and for how long. Then put that all into the Big ERN Google Simulation sheet and see if your version of flexibility makes any noticeable difference in the baseline sustainable withdrawal amount. Maybe there are some folks out there who would be OK living without discretionary spending for 12 years. And then, go ahead and plug that into the SWR Sheet. But I suspect that it is unpalatable to most retirees.
Conclusions
Flexibility is overrated. Still. Again. I thought I had debunked it in Parts 9, 10, 23, 24, and 25. The proponents of flexibility – and this is true for all of them, not just MF&NM – do a really good job disguising the following skeletons in the closet:
1: They often won’t tell you explicitly how long you must be flexible. Every time we deviate from the fixed withdrawal amount, it’s no longer enough to show me a summary table like MF&NM or in the Guyton-Klinger research. I need to see the time series of withdrawals, especially in the worst-case historical cohorts, to gauge if I like the flexible and volatile path more than the fixed withdrawal path. Just one table won’t cut it, folks! I need to know the depth and length of drawdowns, not just the peak consumption! I suggest people also present the utility-based stats, not just some misleading initial withdrawal rates!
2: In defense of MF&NM, I admit they did a good job laying out how deep the spending cuts may be. In contrast, the Guyton-Klinger spending rule research paper is not very clear on the depth of the spending cuts if you start with an aggressive initial withdrawal rate. The casual reader may incorrectly infer that the GK rule only needs one or two 10% steps down. But I’ve demonstrated that GK with a 5-6% initial withdrawal rate would have required long and deep spending cuts. They look very similar to the MF&NM spending rule!
3: MF&NM advertise their safe withdrawal rate as 5.5% without telling you that right now, as of June 15, 2023 (and certainly as of late May when they published their article), with the S&P 500 still about 12% below its CPI-adjusted all-time high, you’d only withdraw 4.125%, not 5.5%. And during much of 2022, you would have only withdrawn 2.75%, forgoing your entire discretionary budget. That’s much less than a fixed withdrawal rate conditional on a modest market drawdown. See Part 54 for details!
4: While it’s commonly accepted that withdrawal amounts should ideally subside later in retirement, when retirees slow down in their 70s and 80s, some of the flexible and discretionary spending patterns go exactly against that. The 1929 cohort that retired right at the stock market peak would have withdrawn only 3.15% of the initial portfolio value annually in the first half of retirement. But 5.10% in the second half. The opposite of what most retirees aim for. The same is true, qualitatively, at least – for all the other worst-case historical cohorts.
Therefore, in light of all of the evidence, let’s put this flexibility nonsense to rest again.
Thanks for stopping by today! Please leave your comments and suggestions below! Also, make sure you check out the other parts of the series; see here for a guide to the different parts so far!
All the usual disclaimers apply!
Picture Credit: wikimedia
Brilliant. Thank you.
You’re welcome!
Thanks for the post, as usual time well spent reading it.
Have you ever looked into the methods to set withdrawal suggested by McClung in “Living Off your Money”? It appears to combine modest flexibility with modest increase in WR.
Brgds
Yes, he has. https://earlyretirementnow.com/2017/04/19/the-ultimate-guide-to-safe-withdrawal-rates-part-13-dynamic-stock-bond-allocation-through-prime-harvesting/
No, Prime Harvesting is about AA/sort of glidepath. What I refer to is in Chapter 4, “Surveying and Selecting a Variable-Withdrawal Strategy.
Yes, I’ve looked at McClung’s book. The rules studied in Chapter 4 are similar to the Guyton Klinger rules, see parts 9,10.
Again, McClung is subject to the same constraint as I displayed in the first chart: If you start higher, you squeeze the balloon and you will have to undershoot the fixed withdrawal amount at some point.
Thanks. Of course there are the same fundamental constraints, I am more looking to squeezing the balloon less if you want. E.G. setting initial withdrawal a bit (~0.2%?) higher than SWR, and then with a relatively high floor.
As I do not have USD as currency and plan to invest internationally I am uncertain if and how to use CAPE.
I’m also curious (concerned might be too strong a word!) about using VT rather than VOO…, since diversification is the only free lunch, and because I don’t believe that US hegemony will necessarily continue for the rest of this century, I’m more comfortable with VT (72%), AVUV (5%), AVDV (3%), and BND (15%) with 3 mo treasuries for my 5% cash, to last 40-60 years at a SWR.
Do you have any comments on relative risk of using un-historically proven VT rather than VOO (s&p500) for long retirements, Karsten? If you’ve already commented on this topic, apologies if I haven’t seen them to date!
This is tangentially related to using CAPE when broadly diversified internationally, which is why I ask it here. Does that still make sense (CAPE > 20 and also using drawdowns of 10-15% of VOO [or of VT, maybe 8%-13% to be more equivalent?] when adjusting my/our SWR)?
Oops, just read: https://earlyretirementnow.com/2016/05/05/emergency-fund/
Guess I’ll be moving that 5% cash into BND or VTIP! Though my current ~ 5% APR isn’t bad t’all, but I’m feeling market-timey about buying more bonds right now anyway, so I will probably do that.
I’m not saying you shouldn’t hold short-term instruments (money market, 3m T Bills, etc.). In retirement you can certainly have that as part of your diversifying assets. While accumulating, I just did 100% equities.
Check out SGOV. In usual times it doesn’t yield enough to justify the effort of clicking the button, but these are not usual times and now SGOV is near the top of the yield curve. Look at the 30 day SEC yield (~5.2%), not the last 12 months yield.
Yeah, that’s a good fund. It yields a bit more than most brokerage money market funds. The 12-m rolling yield is deceiving for obvious reasons. Good call on using the 30d SEC yield!
When the US stock market tanks then all countries feel the pain. ButI’ve played around with adding international stocks (see the feature in the SWRR sheet). I have monthly MSCI data only since 1970. It would have helped your SWRs for those cohorts. At least somewhat. So, I don’t blame you for adding some non-US stocks.
Your first chart, the “simple chart to showcase flexibility fallacy” assumes that the retirement period starts at “the historical worst-case retirement date, likely in 1929 or around 1964-1968.” In those cases,, a 4% initial withdrawal amount would have exactly depleted the portfolio. But in scenarios where we’re not starting retirement at the absolute worst time, the portfolio grows over the retirement period, sometimes into the millions. In my personal planning using historical data and a beginning withdraw amount of $54,000 annually from a portfolio of roughly $1.3 million (which is very close to a 4% withdraw rate) the range of outcomes 30 years later is between $620,000 left to having over 10 million in today’s dollars. I don’t think either outcome is particularly likely – the first assumes that I retire at the absolute worst time in history to start retirement, and the second assumes that I start at the absolute best time. But the average amount left after 30 years is around 4 million in today’s dollars. Flexibility assumes that I’m not likely to be retiring at an absolute worst time, and realizing that if I’m retiring at a more historically average moment to begin retirement, that my savings are likely to actually grow during the course of my retirement, so I can spend a little more than the absolute minimum that would guarantee success had I retired at the worst possible time. Am I actually starting with a 5.5% withdraw rate? No – but I will be revisiting my portfolio over time, and if I see that my savings are growing and that there’s an excess, that’s when I would consider using some of it.
Let me guess: this is a tool that starts simulations in 1970, conveniently ignoring all the historical bad cohorts. And all your final portfolio values are in nominal dollars.
Yeah, sure, if you believe we will never go through a 1930s and 1965-1982 stretch of returns and inflation again, then my analysis is not for you. Sure, withdraw 5-6% and be happy.
Regardless, surely you agree that 4% being the “failsafe” does not = 5.5% by definition not being safe as you implied. And therefore you do not by definition need to spend less than 5.5% at some point. Only if the sequence turns out that way, which it may or may not.
If your current expenses include a mortgage perhaps in a HCOL area it seems even more likely that your income needs post-mortgage and 70+ and 80+ will be much less than .9 and .8 of current as well. All depends on how lavish your disposable budget is.
5.5% is so ridiculously high that I’d bet that if you start retirement today, 5.5% will not survive 40 years.
You can always hope for the best, but that’s not an attractive strategy for me and most of my readers here.
Also, you explicitly explicitly model your mortgage in my SWR sheet to see how much influence you have from 1) nominal payments only, i.e., no CPI adjustments to mortgage payments and 2) paying off the mortgage x years into retirement. See Part 57. We should not use the mortgage as some BS red herring to handwave ourselves from 4% to 5.5%. You’d need to supply some concrete numbers to convince me.
Likewise with the 0.8 and 0.9 scale: almost no impact going from 1 to 0.9 and 0.8. Going from 1 to 0.75 and 0.5 will not do much either.
“Let me guess: this is a tool that starts simulations in 1970, conveniently ignoring all the historical bad cohorts. And all your final portfolio values are in nominal dollars.”
No, I use two different calculators. Both have data starting in the 1870s, although I choose to use data starting at 1902 (the year of my grandfather’s birth – I’m sentimental that way, and not convinced of the relevancy of data from the late 1800s, when economic conditions were fairly different from more modern times). My projected portfolio values are inflation adjusted. (I also assume that I might live to 102, just to avoid longevity risk.)
“if you believe we will never go through a 1930s and 1965-1982 stretch of returns and inflation again”
Because data from both periods is included in the calculators I use, I believe my portfolio would withstand conditions as bad as these – I only have to believe that we won’t see anything _worse_. (My philosophy on this is somewhat similar to my reasoning for putting all the spare cash I had into the market during the Great Recession: I figured, ‘Either the government will do something to help the economy, or we’re all going to be broke anyway, so I might as well invest.’) If in the future we go through a period worse than the 1930s … we all have bigger problems, which remaining employed would not likely have spared me from.
“Sure, withdraw 5-6% and be happy.” … Did I say I was doing that? No, what I said was my withdraw rate was “very close to a 4% withdraw rate.” At the same time, if I stick with that withdraw rate (and adjust for inflation annually, which I actually tend to resist doing), in 25 years I won’t be at all surprised to see my inflation-adjusted portfolio to have tripled in value. And that’s the problem with retirement investing: we would be foolish to make a plan with less than 100% chance of success using past data, but in doing so, many retirees will see their portfolio grow during the course of their retirement. In some cases, by very large amounts. Which means that in all likelihood, many of us have “oversaved” and could be living better.
I retired a few years earlier than I’d planned to when I figured out that with what I had already saved, the only difference that working 5 more years would make is that I would die with more money in the bank.
Please share the link to those calculators. and all the inputs. According to my data, there’s no way a 4% rate would have survived a retirement starting 8/31/1929 or many of the retirement dates between 1964 and 1968. Unless I see some evidence I must assume that that 1) the calculators are trash, 2) you made a mistake or 3) you made some other hidden assumptions to fudge the numbers, i.e., home sale later in retirement, pensions, inheritance, etc.
[Edit later that day]:
I think I now understand where your confusion lies. I looked at your initial comment again and you wrote:
So, maybe you ran the simulation since 1929 and all the worst-case scenarios were included. But if you now exclude the worst-case scenarios then you end up with final portfolio values $600,000-$10,000,000. That’s brilliant. It’s like saying you don’t have to worry about plane crashes. Just don’t fly with the planes that crash.
Ah! You found my error! I did indeed include some extra income: Social Security (not to deliberately ‘fudge the numbers,’ but because I’m simply used to thinking of my whole retirement plan as one comprehensive thing, rather than isolating separate parts of it). Although my starting withdraw rate from my investment portfolio is around 4.1%, once I start receiving SS, the withdraw rate drops to about 2.7% for the same annual income. There’s the mistake in my numbers.
If I exclude SS completely from the projections in my calculator and start with a 4.1% withdraw rate, my plan only succeeds in 68% of past scenarios (going back to 1902). Still, it ends up being quite a range of results, from “you would have run out of money in about a third of retirements,’ to ‘in the absolute best case, you would have ended up with $38 million in today’s dollars.’ The ‘average’ result is that the portfolio would have grown to about 6 million in today’s dollars, which is over 4 times the starting value. (This is if I live to be 102, which is a 47-year retirement.)
That’s what I meant by “in scenarios where we’re not starting retirement at the absolute worst time, the portfolio grows over the retirement period, sometimes into the millions.” We don’t know when we start our retirement if we’ve picked a bad moment in history to do so, but to ensure the success of a retirement plan, we would be smart to assume that we might be retiring at a terrible moment to do so. To build a successful plan, it has to stand up to the worst history threw at people – the eve of the Great Depression, or 1966. But the reality is that we are probably retiring at a more average moment, in which our portfolios would actually provide all the income we need, and actually grow faster than inflation. To borrow your metaphor, we have to craft a plan that assumes the plane we’re on will crash (because it _might_), although there are good odds that it will land safely.
“According to my data, there’s no way a 4% rate would have survived a retirement starting 8/31/1929 or many of the retirement dates between 1964 and 1968.” One of the calculators I use is FIREcalc (FIREcalc.com ). Even if SS income is excluded from my total plan, starting with a withdraw rate of 4.1% (in round numbers it’s $54k of income from a $1.3 million portfolio, invested 90% in equities, and adjusting the spending annually for inflation), I’m getting a success rate of 68%, with 1902 as my data start year to include both the Great Depression period and the 60s. (It doesn’t allow me to set a specific date like 8/31/1929 as a starting point.)
A second (and very simple) calculator I use is https://www.cfiresim.com/ – in this one, if you input 2022 as the retirement start year and 2069 as the end year, with a $1.3 million starting portfolio invested 90/10 equities/bonds and use ‘historical data – all’ (and CPI historical for inflation), with no other outside income it results in a 93.5% success rate (“Failed 8 of 123 total cycles.”) – This link should give you both the inputs and the results tabs already done. https://www.cfiresim.com/3f54efb3-4baf-4855-92f3-55b2fcb6641d
Both of these calculators say they use historical data; but neither finds that a 4% withdraw rate would have failed in 100% of historical cases, including the Depression and the 60s. So I’m naturally curious about the discrepancy.
Good! Mystery solved. Thanks for confirming, that 4.1% does not survive a 47-year retirement.
A few words about the $38m (max) and $6m (average) figures:
Max final value: I can guarantee you that you will not end up with $38m final value, CPI-adjusted. The “shoot-the-moon” outcomes in historical simulations were all generated by the retirement cohorts around the 1920, 1932 or 1982 market bottoms. The CAPE in 1920: 4.9. The CAPE in 1932: 5.6. The CAPE in 1982: 6.6. The CAPE today: 30. Unless you believe we have another roaring stock market recovery like following the historically worst bear market bottoms you will not repeat that. See Part 46, where I post the final value distribution, conditional on the initial CAPE. You will see that the distribution is not as optimistic as what you believe, considering we’re above 20 in the CAPE. All the positive outliers occur when the CAPE is low. (albeit for a 30y horizon, but the same will hold for 47y as well)
(From Part 46: Final Net Worth: 30 years of withdrawals, 75/25 portfolio, 4% p.a. withdrawals. Conditional on the initial Shiller CAPE.)
Mean final value: because of the extreme skewness due to a few extreme outliers, the mean is a meaningless measure (pardon the pun). And again, due to the CAPE environment today, it’s completely nonsensical to calculate historical mean final values and assume you can replicate those in today’s environment when the CAPE is still way elevated compared to historical means.
Finally: cfiresim an all the other annual simulation tools will not capture the the 1929 worst-case scenario because the peak equity market occurred in August/September. That’s another reason I don’t like cfiresim.
But again, if I’m wrong and the stock market rallies by more than 700% in CPI-adjusted terms over the next 9 years (like it did between the 1920 bottom and the 1929 peak), I will surely raise my withdrawals. But I wouldn’t bet my retirement security on such an insane stock market forecast.
Well, I didn’t think it would grow to 38 million, because I don’t believe that I’m retiring at the very best time in all of history (just as I don’t believe we’re at the very worst).
But when you say “Thanks for confirming, that 4.1% does not survive a 47-year retirement,” you say it as though it’s definitely doomed to fail in any historical scenario. But when I remove SS from the FIREcalc calculator, and simply put in a $54,000 starting annual withdraw from a 1.3 million portfolio invested 90% in equities with a 47-year retirement period, starting with data from 1902, set to constant spending power, it says, “FIRECalc looked at the 75 possible 47 year periods in the available data … For our purposes, failure means the portfolio was depleted before the end of the 47 years. FIRECalc found that 24 cycles failed, for a success rate of 68.0%.”
I wouldn’t say that succeeding in 51/75 cycles confirms that 4.1% does not survive a 47-year retirement. Maybe I started my retirement at a period that’s like one of those 24 failing cycles. But, luckily, I don’t actually think that I’ll live to 102, and I have several backups to my overall plan in case things go south.
Correct. I should have written “4.1% is not a failsafe rate” to be 100% clear. Otherwise, someone will claim an SCR of 10% is not a failure because there are some historical cohorts that made it 47 years. Sorry for the confusion!
Interesting article as always. Whilst I don’t disagree with your dire warnings about following the approach discussed, I am not sure I agree with the statement in the first section that it’s a mathematically certainty that you will fail if withdrawing those higher rates – looking at your own data, it looks like there were retirement dates in history where you could have withdrawn those kind of rates throughout and still not run out. Therefore if you happened to retire at one of those lucky times you would be fine.
If your plan has a 50% chance of failing, then I guess there is a 50% chance that it will succeed. I guess what you are showing is that when it fails it usually fails big and that this is often glossed over.
The flaw to me on the approach is about saying that you are prepared to eliminate your entire discretionary spend for multiple years at a time – to me that’s not going to be much fun.
Let me write it again: It’s mathematical certainty that the SWR is lower for a specific retirement date when you go from 40 to 60 years. So, the 1929 cohort has a lower SWR over 40 years than 60 years. This is true for every single cohort. So it is a mathematical certainty that failsafe is lower.
You seem to refer to the fact that there were cohorts with SWR much higher, i.e., 10% and more in 1982. That’s a completely different issue and I never claimed otherwise. Also even in that cohort with the very high SWR, it will decline as you go from 40 to 60 years.
In retirement planning, most folks are not happy with only a 50% success rate. Even for much less significant issues I’d not accept a 50% failure rate:
Are you late for job interviews 50% of the time?
Are you late for your kid(s)’ recitals/graduation ceremonies/birthdays/etc. 50% of the time?
Do you miss flights 50% of your time?
Do you miss funerals and weddings 50% of your time?
But if you’re fine with a 50% failure rate in retirement, you can still use my tool. You can look up the median-WR that would have created exactly that 50% success. See the table in the section “2: A simple numerical example,” the last row under “50.00%”
Thanks Ern for your very thoughtful analyses. Do you ever analyze using reverse mortgage or equity in home while spending every penny of investments? Trying to limit legacy while optimizing or maximizing spending.
If you own a house and don’t plan to leave a bequest, you should consider a reverse mortgage, selling the home later in retirement and renting afterward. Another idea: if you plan to leave a specific bequest size, you just stay in the house, make the house part of the bequest and simply reduce the financial assets you leave to your heirs. They can always sell it at the step-up basis.
Glad to see I wasn’t the only one bothered by that MadFientist post. They make it sound like everyone has flexibility and the younger the more flexibility you have which is exactly the opposite. MadFientist should be more careful what he “preaches” for the community since he’s a celebrity in the medium.
I like your approach better ERN. Congrats
Good, I’m glad we think alike! 🙂 Thanks for the kind words!
Hi Ern! Great info. Wonder if you would do work on inflation hedging for retirees. For example, we often hear about the dreadful 1966 scenario, but inflation-indexed securities didn’t exist back then. How would the presence of those securities change that scenario, and what’s optimal overall?
I’ve been thinking about that issue. The problem is that we don’t have TIPS returns pre-1997. One could try to backfill with estimated TIPS returns. Essentially, write a model for how the real rates would have evolved pre-1997.
But it’s a lot of work for eventually finding the obvious answer: If you had had access to TIPS in the 70s and early 80s you would have done very well.
My understanding is that TIPS doing well is only for the case of **unexpected** inflation. Given Volcker raised the 11.2% fed funds rate from 1979 to 20% in June 1981, I suppose by “early 80s” you specifically mean (up through May 1981-ish)? I’m not sure if expected inflation wasn’t already anticipated by bond markets much earlier than that, like prior to Reagan’s election, in 1979 or earlier? ¯\_(ツ)_/¯
Anyway, your comment roughly stands, but it got me thinking about how useful TIPS are now, given we truly don’t what the next 5+ years will bring inflation-wise…, last week bringing some relief in sight at least.
I’m not a market timer, really I’m not! I do entertain what if thinking about it though!! 😉
It’s risky: maybe the inflation shock is over. I think there’s already a lot of disinflation built-in, that will hit in a year or so. See my post here: https://earlyretirementnow.com/2023/05/17/may-2023-market-musings-monetary-policy-and-inflation/
But to hedge against the other tail risk where inflation rebounds again and the Fed has to go to 8% FFR, you’re better off with TIPS than nominal bonds, for sure.
TIPS turned out to be a terrible investment in 2022 during the exact sort of scenario people buy TIPS for. Turns out duration risk trumps the TIPS adjustment. Imagine correctly foreseeing the inflation of 2021-2022, buying TIP in response, and then experiencing a total returns of -12.24%!
https://finance.yahoo.com/quote/TIP/performance/
Re: inflation/deflation in the future, the -0.5%/month disinflation we all think is good news today will suddenly become bad news as CPI falls below 2% a few months from now. The Fed, which takes 6-12 months to pivot their policy, shows no signs of being concerned and is still fighting the inflation monster that died last year. Their focus is on “core” inflation, which more heavily weights housing and transportation, which are lagging the trend for their own reasons. If they *started* worrying about too-low inflation today, the first rate cuts might arrive in April or May 2024, but they haven’t started worrying about that yet. Real risk-free rates might hit 4-5% before the next rate cuts and that will accelerate us into a deflationary recession.
If this is the forecast, call-protected, 6% yielding corporate bonds with lots of duration are looking good. Pivot into stocks when the SHTF.
Everything was a terrible investment in 2022. VTIP was down only 4.58% (dividends included) in 2022. Compared to double-digit losses in other long-term Treasury ETFs and ~20% down for most equity funds. [edit: Sorry, VTIP is the short-term TIPS fund. Yeah everything was awful due to duration. CPI-adjustments didn’t help much]
But you’re right, this recent inflation shock was a disappointment for TIPS. Real yields went up from just above 0 to 1.5-1.75% for the 10y, 20y and 30y durations. Even with the CPI adjustments lots of investors are underwater.
I was hoping for this one! Thank you Big Erin! As per usual great analysis and commentary.
Glad you found this helpful! Thanks for the kind words!
One thing I noticed reading this – which is admittedly more of an emotional argument than an empirical one (though our emotions certainly matter, which is a big part of why we choose risk according to our risk tolerance, which is often as much emotional as it is practical) is that I personally would feel stressed out and feel that I’m needlessly taking high risk by continuing my fixed withdrawal rate throughout a deep and prolonged depression when I could fairly easy cut discretionary spending somewhat instead. Sure, sure, the math was there before I retired to put me at ease, but what if I’m in the “worse than all previous cohorts” retirement cohort? I would actually sleep better at night reducing my withdrawal rate slightly, and accepting the associated cuts to discretionary spending than continuing fixed withdrawals and trusting the math just so I could eat out more or take an extra vacation each year.
This preference seems like it would significantly impact the calculation of my “consumption equivalent utility” function.
The conclusion of this, is that if I do indeed cut spending slightly during market draw-downs, I can correspondingly increase it slightly during bull runs. Advertising this as “a higher SWR with flexibility” may indeed be misleading, but if I squeeze the balloon less during high sequence risk times, I can indeed squeeze it more other times – and sleep better. And to be clear, I plan to start with a SWR of ~3%, so I’m not using the idea of flexibility to save less, I’m using it to reduce SRR.
What you describe sounds like a CAPE-based rule. Imagine your portfolio goes down, but the equity and bond valuations now look more attractive. So, you withdraw a larger % out of a smaller portfolio, but still require a small reduction in spending (i.e., the net effect is still negative). That’s totally legit and is also the optimal thing to do. And will likely give you high utility. Likely higher utility than under the MF&NM method where you ignore equity valuations. The biggest difference would be if you retire at the bottom of a bear market. MF&NM recommends only 2.75%. But with the CAPE Rule you’d easily get to 5.5% right at the beginning of retirement.
Great post as usual. I agree that flexibility is overrated. I think there is more potential to mitigate sequence of return risk via the portfolio. I think between tactical asset allocation and alternative, non-correlated asset classes there is some potential to boost SWRs. The problem is the lack of historical data.
One thing I disagree with is saying equity drawdown is the better valuation metric for our SWR purposes. Even though CAPE can vary across time and country, I still think it’s a more fundamental measure than drawdown. Is a 15% drawdown from CAPE of 90 in Japan 1989 the same as a 15% drawdown from a CAPE of 20?
Agree. A better approach would be to add more asset classes. Real Estate comes to mind. Trading options (see my relevant work on the topic) can easily add 1.5% expected return to your portfolio. There is your higher SWR. And it doesn’t require eating rice and beans and staying for 12 years.
About CAPE vs. drawdown: I hope someone can develop a reliable dynamic estimate of a “normal” CAPE. The last time it dipped below 20 was between 2008 and 2011. Historically, 15 was the median and everything above 20 was considered overvalued. There has been a level shift in what we consider a high vs. low CAPE. So, over the last few decades, the drawdown is likely the more reliable metric. The idea here is that if the market is already down by 20+% peak to October 2022, do we believe that we should still worry about another Great-Depression-size drop on top of that?
Thanks for quantitatively arguing against that MadFientist article I had a problem with. I want my flexibility to be used to enhance my retirement and not a requirement to survive it.
Glad we had the same feeling about that article. And good point: I’m not against flexibility. I like the flexibility to raise my withdrawals and spend more! 🙂
Flexibility is overrated… for the middle class. For the wealthy early retiree, it makes a ton of sense! For example, in my own simulations, I model never giving myself at the beginning of each year either a 20% raise, a 20% cut, or leaving the budget alone, and then adjusting for CPI. A person following this strategy will sometimes take a 60% budget cut! And to your point in the article above, they may be below their initial withdrawal rate for years (although it’s really rare to be 60% down for years.)
So what kind of crazy person would want to do this? A person who wants to know, with math, how much they can raise their budget as their situation improves. a 4% initial withdrawal rate with a flat spending strategy quickly turns into a 2% withdrawal rate in the modal case. That’s why it works! That’s the secret to all high initial withdrawal rates! They always decline.
So if you force your withdrawal rate down over time, you can start with about 5.5% as long as you crank it down to 1.2% over the next 43 years. Hopefully that’s done through the portfolio rising faster than the budget, but if it doesn’t, it’s time to be flexible.
To be clear, this is a terrible strategy for someone who’s thinking about $80K in today’s dollars as their annual spend. But folks reading this blog are often in big tech, big law, or big finance. Those folks might be retiring with $5M-$10M relatively early, and be thinking about generational wealth, so a 100+ year investment horizon.
With that kind of budget, the person has a choice. They can sustain that $150K to $300K a year with no caveats with a 3.3% flat withdrawal strategy and a pretty low failure rate. Or they can start a bit higher initially ($250K to $500K), be flexible, and spend much more money over time and leave twice as much to their children, on average, as they started with, after inflation.
And I’d hope that a $250K annual spend included a bunch of fun money! Ideally, it’s going to target at least a third of the budget as fun money, donations, and other “trivial to cut” spending. An additional third ought to be easy to postpone, like preventative maintenance, new vehicles, and upgrades to houses. Only a third should be structurally difficult to cut.
In other words, the key to a flexible spending strategy is a flexible (and high) budget!
This is the optimal strategy for the top 10% or so of earners, IMO. It’s not for the middle class. But if you are frugal and have a giant windfall or a consistent high earner, you can retire in your 40s or 50s with a fat but flexible budget.
A lot of problems go away if you’re seriously wealthy. Or, to be precise, the problems will then be different ones.
But I still wonder if those rich folks get too used to the comfort of a $500k/y lifestyle and then feel depressed about scaling back again. Hedonic treadmill?
High earners are no different from the rest of the population in terms of money habits. The absolute majority spend what they earn. Those with a FIRE mindset who retire with 5+ millions already have more money than they accustomed to spending.
ERN is attempting to thread the needle between under and over saving, but in reality, almost everyone undersaves, and a small minority oversaves. Threading the needle folks are a rounding error.
Good summary of my thought process! Thanks!
That said, the “almost everyone under-saves” applies to the population as a whole. Within the FIRE community, I got the impression that a lot of folks are shortchanging themselves and over-save. Most people who FIRE in their 40s and 50s can likely push the SWR a little bit higher than undet the dumb 4% Rule of Thumb if they expect large pensions and Social Security.
Not so fast J M. I never earned a lot of money. I saved a lot (at one point 90% of w-2 wages) and had little luck with investments. Otherwise everything that I have in my life I earned with modest salary. I retired at 42 last year. Ironically my income now is more than it was every before I was actually working. My savings is still pretty high because I don’t like to waste my money.
To me it sounds like to you need a little motivation: So here is it: Save your money. Spend a lot less. Don’t give up and you will be just fine (I mean it).
Oh, yes! retirement is great!
Many roads lead to Rome! 🙂
Thanks Big-Ern, the data doesnt lie! Good insights.
I’m assuming this whole article is for those that don’t plan to use a CAPE-based withdrawl rate as proposed in the SWS and spreadsheet which already has a dampened version of flexibility based on equity valuation (by definition!). And of course the additional benefit of recalculating SWR based on remaining portfolio every month is you are guaranteed not to have catastropic model failure 1/2 way through retirement! I would way rather suffer in proportion to the rest of everyone else’s retirement portfolio;-)
The CAPE rule works, too. I thought it’s best to compare the volatile spending rule to a fixed rule to start with. But I’m confident the CAPE rule also beats the MF&NM method.
Here is what I love about this post, you challenged other contributors in the community not for their intent or even the final result of ~$80k/year, but their methods and rigor behind the result. Not to be contrarian, but to be complete and to force people to understand the underlying assumptions. Bravo as usual.
Something I do all the time at my job is to run experiments and one potential idea is figure out the worst possible SWR and see what that life is like for 6 months. Really live it and feel it. If you can live that way for 5-10 years, then maybe extreme flexibility would work fine for you.
Thanks! Agree 100%. Most people will have trouble living on 50% for 6 months. Especially in the FIRE community where you’ve already cut out all the excess it will be hard to go down another 50%! Most people will give up after 6 months. 12 years for sure!
Interesting post!
I have some questions:
What you are pointing out fits what I’ve seen from calculators: that if you are very unlucky and retire in a really bad year, you have to accept that that means you might have to live at the minimum end of your flexibility scale for 10+ years.
To me though, the biggest argument for incorporating flexibility is not the worst case scenario, it’s how inefficient a fixed spending amount is if you don’t get very unlucky.
My experience with playing around with different calculators and flexible withdrawal strategies A LOT is:
Failure rate is basically driven by the minimum withdrawal.
But: the failure rate hardly changes if you increase the withdrawals in good years (within reason, let’s say up to 30,40% increase).
So my analysis is that while flexibility does not magically reduce the amount of money you need to be comfortable when unlucky, it seems really wasteful to not add an flexible ‘if things go well’ withdrawal logic on top.
In other words would you agree with: Make sure you are OK with your minimum for a long time, but add 30% flexibility on top to account for things going well.
Another question unrelated question: I love that you analysed how much higher the withdrawal can be depending on far away from the ATH we are.
Looking at your table, would you agree a reasonable approximation rule is: if you have a 100% historic success rate, then you can increase the safe withdrawal rate by as about much as the current S&P is away from the ATH?
Agree: you should increase your withdrawals in the event the market cooperates. So, flexibility certainly works well when ratchet up the withdrawals.
Re your question, I wouldn’t say that it’s exactly 1-for-1, but close. It would work as a rule of thumb, yes.
Is there a SWR series post on this upside flexibility? E.g. if 3.5% is the SWR (perhaps CAPE adjusted) for retirement length, the logical withdrawal adjustment is to the higher of inflation or 3.5% of the portfolio after withdrawals (as if you were starting that year). What proportion of cohorts would see their portfolio exceed 60% real growth and achieve 5.5% SWR (or 4%, 5%) on the original value anyway?
(And guess that analysis is another counter to the “one more year” mentality.)
Of course. You can gauge what’s the final value after x years when withdrawing only 4%. There’s a tab in my Google Sheet.
You can also check how CAPE-base rules would have cranked up withdrawals during long bull market.
From my own experience, my portfolio is up even after 5 years of withdrawals. I don’t really spend more now. I’d rather save the excess and build more defenses against unforeseen risks, though.
Glad you wrote this post. I was pretty surprised when I saw the MadFientist post as I remembered his blog being pretty decent. Reminded me of the yield shield crowd when I read it. The basic idea that retiring much earlier with flexible withdrawals has a reasonable chance of having to take a big hit to spending for many years to make up for it is so basic. I think there is a selling a dream aspect to these kind of posts (including the FIRE in your 30s at 4% swr crowd) where they want to make reaching FIRE seem easier than it is.
Glad you pointed this out, because that was exactly my impression. The MF&NM article was a bit of click-bait and you get the impression that the authors, just like the yield shield folks, probably don’t even apply their method themselves. Everyone is in the business of “selling a dream” and that 5.5% bumper sticker sticks with a lot of gullible people in the FIRE community. Let’s hope not too many people get burned by this.
Karsten,
Great article. I appreciate how you use basic logic, mathematics, and modeling to make your points. (unlike others that use only language).
I have a slightly different take of flexibility. Starting with a 40 year retirement, the appropriate SWR fail-safe rate for the equity/bond risk tolerance. (3.43% for 80/20 equity/bond portfolio). Every 5 years (of more frequent if desired) recompute / revise the SWR fail-safe calculation adjusting the (3) key parameters:
– reduce the number of retirement years (from 40 to 40-5=35)
– update to current CAPE
– adjust the equity/bond ratio to match the shorter retirement period or changed risk tolerance.
The result is higher retirement consumption during or after high performance times (let’s have a party) and lower spending during poor performing markets.
I would love to get your thoughts on this approach.
Yes, we should certainly adjust the spending along the way. Maybe every 1 year you see what’s the new recommended CAPE-based withdrawal rate. Also, account for depletion in conjunction with CAPE (see Part 54 for the mechanics) and the shortening of the horizon will help you a bit with cushioning the withdrawals if there’s a big drop.
I noticed the current version of the sheet now incorporates the cash flow sheet in the CAPE calculations. Not sure when they happened but thought it deserved a call out.
Thanks. I noted that change in Part 54, i.e., how to factor in partial asset depletion and supplemental flows.
There are very few articles in the personal finance community that make me nervous because I perceive they would cause people harm. That 5.5 percent flexible article is one of those. Great rebuttal.
Yeah, good point. Some folks in the FIRE community may not read their post carefully enough and conclude that 5.5% is the new 4%.
There are very few articles in the community that make me worry they will cause people harm. That 5.5 percent article on flexibility is one of them. Thanks for the great rebuttal.
Thank you for another great article. Never mind 5.5%, I am even concerned about a 3-3.5% SWR. The 3-3.5% SWR seems to assume that we will not see any future event much worse than what we have seen in the past 100 years or so. 100 years seems like an inadequate data set when we are looking at a 50–60-year planning horizon so is that a safe assumption? What will the SWR rate be if we see an event in the coming years that is say 30% worse than 1929? What are the odds of such an event? Is there something about modern capital markets that makes the odds of this close to zero? This is the main concern that keeps me with a bad case of just another year syndrome so hoping you can provide some insights to help those with these concerns pull the trigger on EA.
You can test against withdrawals in non-US countries, Italy, Switzerland and Japan I think are the edge cases where 2.8%-3.2% (30-35x) is the SWR with some consistent blend of global and domestic equities and bonds. This somewhat covers the beyond 1-in-100 scenarios for the US.
To make this relevant for us in the USA, though, we should avoid countries like Japan and most of war-torn Europe. I don’t think Canada will declare war on us and destroy most of our assets.
Agree You bring up another valid concern: model risk. What if returns patterns in the future are consistently worse than in the past? As an additional hedge, I always think that I should err on the side of caution and maybe withdraw SWR-x% to account for that unquantifiable risk.
But clickbait always entices people to look for solutions that exaggerate the SWR. Caution isn’t a good way for you to viral! 😉
I’m so glad we have big ERN in our community to really math things up and debunk crazy posts like that of MF.
You must have a ton of enemies in the FIRE community now, but it’s necessary. Keep up the great illuminating work that you do to help us.
Thanks for the kind words. I feel I have gained a lot of friends, too. That’s all I care about! 😉
Outstanding as usual … The comments are as fun to read as the article!! Hopefully MF&NM will chime in, I love the debate =).
I would love that, too! Crossing my fingers!
Thanks for writing this, and once again (for the 6th time now??!!) debunking the idea we can just wave our hands and be flexible in early retirement.
I appreciate your dedication to trying to keep people from making serious errors in their retirement planning, and the rigor in which you present the evidence.
That said….man, if I were you, I would have a hard time not just posting something like “Reply to MF: Flexibility still doesn’t work-please see parts 9, 10, 23, 24, and 25, as well as this spreadsheet which contains open access to related formulas and source data” and calling it good.
Or just disengaging completely, say “F it! If people want to hitch their retirement to hopes and dreams instead of math, let them” , and then ride off into the sunset.
You’re definitely a better person than me, because I would have given up long ago.
So yeah, thanks for taking the time to put all this together. It’s truly a service to everyone trying to retire early. I’m sorry that this must get frustrating. Everyone is always going to want to have their cake and eat it too; there will always be another pollyanna withdrawal method to disprove.
Apologies for the continuing comment. Realized I was a bit negative there.
I’d like to point out that though I’ve never met Brandon from the MF, I’ve read every post of his, and have gotten quite a lot from many of them. His material has probably helped lots of us out there.
While we can never know someone, especially from just reading their blog, it would seem that Brandon is an intelligent person with the best of intentions.
I could theorize why he wrote this latest article, which appears to be mathematically unsound, but I’d likely be wrong. I’m guessing he probably believes what he wrote in this latest article, and was trying to help his readers. But at the end of the day, he’s still human like the rest of us. This stuff is complex, and he may have stepped a bit outside of his circle of competence. This seems to be a danger for lots of us.
I earned an advanced degree in my former field, and am legally considered an SME in my niche. I’m often tempted to use my statistical training to model things that have nothing to do with my expertise. On good days I remember that I know a lot about one small of circle of the world, but that doesn’t mean I have a deep understanding of unrelated fields. It’s tempting to overstep one’s knowledge base and drift into that perceived circle of competence.
Reading your material helps me remember I don’t know what I don’t know. I hope others do the same.
Noted Good point. Let’s not assume malice.
Thanks for your kind words!
I love your comment. After blogging for so long, I no longer get as upset as in the beginning. Hey, I get new topics to write about. So, there’s some upside.
But I agree, it feels like that old saying “The lie has already gone around the world before the truth even puts on its shoes” and this context you wonder, how many people have deluded themselves with the 5.5% rule and now don’t even want to hear about my careful, rational rebuttal?
Nice read as always. Looking forward to more. What’s your opinion on stock valuation currently? Ben Carlson has a nice analysis saying that stocks are more or less fairly valued nowadays: https://awealthofcommonsense.com/2023/06/how-cheap-or-expensive-is-the-stock-market-right-now/
Thanks again for your great work.
Yeah, thanks, I saw that post before. I agree that stocks are not as crazy overvalued as before, hence my allowance for a higher SWR. The relative valuation, 1/CAPE minus cash yields now looks pretty bad though. Some other measures look pretty average again. So, we certainly don’t have to target the historical worst-case scenarios today.
So if I was starting my FIRE now, do you think I can set a 4% and forget? I don’t intend to keep looking at my finances more than once a year during fire so I need something save and that I dont need to adjust
With the market a bit down from the peak and interest rates high and the CAPE also not insanely high anymore, I certainly propose that 4% works again as a flat, fixed withdrawal amount (adjusted for CPI). See also https://earlyretirementnow.com/2022/10/12/dynamic-withdrawal-rates-based-on-the-shiller-cape-swr-series-part-54/
Excellent perspective and analysis as usual. Appreciate all the work you do!
I was wondering how this strategy compares with the CAPE based strategy you have written about before. Using the simulation tools, it seems the CAPE strategy can also result in prolonged periods of low withdrawals – particularly for the 1920s and 1960s cohorts. Maybe this is a common trade off with any variable rate strategy?… Are we to conclude that the fixed (real) withdrawal strategy is something you now favor given all the issues with “flexible” strategies?
I haven’t done that yet. The challenge with the CAPE-based startegy is that the simple version of it does not plan depletion. So, it’s hard to compare the CAPE-based version when there’s money left over at the end with the two scenarios I studied here which both exactly deplete the portfolio.
But, yeah, it’s on my to-do list to come up with a utility-based approach for CAPE-based rules.
A bit confused by that comment. Isn’t that the purpose of the “Final Value (%)” cell (b12) on the CAPE based rule sheet of your SWR toolbox? Maybe i’m misunderstanding something here.
Yes and no.
Yes, you can check how much of a difference capital depletion would have made in today’s situation.
But I never ran any simulations where I went through every past cohort and followed them over time to study how the changing CAPE and portfolio value and rolling through the life expectancy and cash flow parameters over time impacted the actual withdrawals.
The only thing I can do is to assume capital preservation, then you can just look at the calculations for the historical drawdowns in that tab.
Somewhat related observation: I’ve been playing around with your SWR 2.0 toolbox google sheet and noticed that, for my scenario, the SCR for a 0% equity drawdown is 2.71% while for a 15% drawdown its 3.41%. If I do that math using a hypothetical $1M portfolio that is down 15% to $850k, you’d end up with a slightly higher initial withdrawal amount when its down 15% vs back to $1M ($27k vs $29k). I could understand them being equal but not sure I understand why your SCR would actually be slightly lower once your portfolio has returned to its previous high?
Regardless, it brings up an interesting realization. Assuming you don’t contribute any additional money to your portfolio, your initial withdrawal amount won’t actually increase until the market hits a new high (in real terms).
Another odd thing I noticed: I’m modelling a scenario where retirement lasts 600 months and there’s 70% value remaining. No additional cash flows. Per this article, I’d expect that SCRs would rise slightly if I reduce the length of retirement. But when I reduce to 500 months, the SCR for the 0% drawdown case goes down slightly to 2.70% (vs 2.71%) and the SCR for the 15% drawdown scenario goes to 3.28% (vs 3.41%). Why would that be the case? I’d expect the opposite to be true.
You’re probably using a 100% equity portfolio, so the current drawdown vs. the model drawdown move about 1–for-1. Yes, in that case, you’d ratchet up the withdrawals only when you reach a new all-time-high, CPI adjusted.
About the second question: There is a strict monotone negative relationship between the horizon and the SWR if you’re using capital depletion. It’s a mathematical certainty.
If you have a large capital preservation target (100% or even 70%) you can encounter situations where over 50y, a certain SWR would have failed, but over 60y you recover enough to reach your 70% target.
Hmm- actually I was using a 60/40 portfolio. But maybe I don’t understand the logic by which the SCR rate by equity drawdown table is calculated? Does it just take the SCR calculated for the 0% chance of failure and then adjust that by fixed percentages to calculate the SCR for each percent of equity drawdown shown ?
Asked another way, how should I interpret that table when NOT modeling a portfolio where equites are 100%?
Sorry, then I misunderstood your original comment. A 60/40 portfolio usually generates SCRs much higher than 2.71%. With a very long horizon and a high final value target, it’s also possible to get very low initial SCRs. Sorry about the confusion.
The conditional SCRs displayed in the table are the failsafe SCRs conditional on the equity drawdown.
Hello, thank you for your blog post and the thoughtful analysis. I am in the same boat as you – prefer to having a fixed SWR with little to no (<1%) probability of failure using historical sequences. So far, I have also gathered 3% to be a SWR for a US stock/bond portfolio which is consistent with the 3.25% you are calculated. However, lately, I have been come across some analysis in the FIRE space using size, factor, geographic, and asset diversification to get a better SWR – in the range of ~5%. E.g., Weird Portfolio, Golden Butterfly, etc. which you can find on https://portfoliocharts.com/portfolios/. Have done similar analysis using portfolios other than US stocks and US bonds?
My SWR toolkit allows you to use all sorts of other asset classes:
non-US stocks
small-cap vs. large caps
value stocks vs growth stocks
gold
short-term T-Bills
30y bonds
portfoliocharts has probably a few more asset classes, but keep in mind that the simulations are only annual and start in 1970. It’s a totally useless tool because you can’t capture the bad cohorts in 1929 and 1964-1968.
I agree flexibility is overrated, but it isn’t worthless.
Many discretionary goods/services do get cheaper during recessions/stock market meltdowns when even folks who are still working cut back on those too in tough times. So if you have some slack in your budget, it can be easier to cut back on these things while still consuming them.
For example, briefly in Spring 2020, I was able to book 10-11 month out domestic airfare/hotels for up to 80% off of the usual price. I also got some insane deals on luggage too at the time.
In 2008/09, Groupon became popular way for restaurants to offer excess inventory of dining at a heavily discounted price since so many people were cutting back on eating out.
Also people who are able to retire in their 30s or 40s, often end up using their time to do some kind of freelance part time work like yourself that brings a small amount of supplemental income. For example, In tough times, educational demand often spikes, perhaps an early retiree with a strong STEM background could offer tutoring services part time for a year or two until the market partially comes back.
“but it isn’t worthless”
True. Never claimed otherwise. But keep in mind that all those price changes are already factored into the CPI. Unless you have wildly different spending percentages than the CPI basket, this is nothing you can use to improve your situation.
And I certainly make some money on the side. But it’s not 50% of my expenses. Not even 25%. So, again, the better way is to model explicitly what you think you can pull off in the flexibility department. Then see that it doesn’t make a huge difference on your initial SWR. Which in turn means that the 5.5% proposed by MF&NM would require a lot more flexibility than most people can pull off.
Fascinating analysis! My withdrawal rate will be around 1% since we saved and invested well over a long period of time and we seem to find optimum happiness from a fairly modest cost lifestyle. But the thing that always occurs to me when safe withdrawal rates are discussed is that the whole exercise is based on a tiny data set of only a few decades. I did a lot of modeling in my career of complex systems using all kinds of algorithmic tools as well as early artificial intelligence software. The one thing that was nearly universal was that regardless of how much data you included in your training set, when it was time to use the model to predict future results reality had shifted and the input data was now outside the parameters the model had been calibrated on. In those cases the results were pure nonsense. I fear the basic premise of basing future performance on past results is just as flawed. The only conceivable basis for doing so is we have no other tools available. The idea that 2023-2073 will bear any resemblance to 1929 or 2000 has no logical or scientific basis and stretches the boundaries of reason. Imagine if you had based a safe withdrawal rate as a Japanese citizen on 1950 to 1990 data. The Nikkei rose by over 15% annually during that period. However since then it has averaged a negative rate of return. Every dominant world economy faces those potential doldrums particularly if they accumulate massive debt, aging populations and low birthrates. I have no answers but the problem is real.
Good point. 1%, if that’s all you need to be happy, will be extra safe, not just against historical worst-case scenarios but even much worse.
I doubt that we will see anything as bad as Japan 1990-current or Japan and much of Europe during WW2 because the US economy is a bit more resilient and we will not be invaded. But I get you: bad policies, big debt burdens, populist policies, etc. will have an effect on productivity growth. Better play it safe. But I don’t think that a SWR under 2.5% is necessary even in that disaster scenario.
What about modelling flexibility by recalculating the SWR each year and seeing how much income fluctuates by and what happens in this instance if you take an SWR+x% withdrawal each year?
IE Start of Y1 you work out the 40 year swr and take this plus 0.5%. Start of Y2 based on the the new balance and 39 years you again calc the SWR and take this plus 0.5% etc.
“recalculating the SWR each year and seeing how much income fluctuates”
That’s what a CAPE-based rule would do. See Part 18.
You can also use the fixed-rate approach and tie your withdrawals to the equity drawdown. In that case, conditional on an equity drawdown you might not have to reduce your spending much. And if the market keeps rallying you will walk up your withdrawals accordingly.
All your calculations are pure math based and this is an accurate mode to do all calculations.
Still, in reality there are a few good options which can help or mitigate an eventual crash in the market and help anybody with an early retirement to survive the bad years.
In my opinion, either if you consider 3.5% or 4% or 5.5% withdraw rate implementing or considering these options will really help:
-If you need an income of 50K (5% return on 1mil) from your investments why not make the savings to provide you 55k (5% return on 1.1mil). The extra 5k will give you a little extra safety.
-If you retire at 55 why not try a part time for a few years before full retirement. This will really help you to asses you plan and to make the transition easier.
-Are you able to produce some extra income still enjoying a full retirement? Maybe some consulting, hobby, internet activity…
-Why not use the bucket option? Save 2 years of expenses in high return bonds or T-bills. These days 5% return is available.
-Do you really need the 50K (or whatever amount you are planning for) in retirement? Would you be able to leave only on 40k if necessary? If yes then you have more flexibility.
I think that 3.5% rate of return is too conservative and you will sacrifice your life for next generations.
Personally I am targeting a 5% withdrawn with the option to leave ok on 3.5% if is necessary.
Anyway Erin, your articles are full of good information and are really helping people to get ready for the retirement.
There is one reality. There isn’t a separate reality for the non-math retirees. Most of the things you propose are helpful but will not magically raise the SWR from 3.43% to 5.5%.
Also, for the 1,000,000th time: as I outlined in Part 55, a bucket approach does not raise your SWR. Even the #1 bucket strategy authority in the FIRE community, Fritz Gilbert doesn’t claim that a bucket approach creates market timing alpha.
If I’m to use a 3% SWR, I’ll just put everything into CDs and bonds and do that. If the market future returns won’t cover a 4%+inflation, why bother investing in the first place? It’s always that…you’re too conservative or too “flexible”…what’s the right SWR number for God Sake!!!?! Do we really need to read 55 posts to find that out? How many nerds are out there?
Let me try to understand: you accumulate for 10+ years. And now you don’t have the patience to spend a few hours studying to determine if your early retirement plan is sound? And you don’t want to spend a few hours trying to find out if you may be able to withdraw much more and/or retire earlier than with a 3% WR?
If so, I can’t help you. My series is not for you. Please find your retirement advice elsewhere. Good luck!
Thank you for the math and explanations. I fear that many “young”, healthy FIRE folks underestimate future healthcare costs and caregiving costs. Medicare is not free, you have to pay for it, and there are many things not covered or are only partially covered. Even if you purchase a supplemental insurance, that is a cost with many remaining out of pocket costs if you need uncovered medicine or treatment, equipment, such a lift chair or a shower chair, or home renovations to be handicapped accessible. Assisted living and senior communities require upfront “buy-in” with monthly fees, not counting “extra” for tasks not covered such as personal care or laundry. Even if healthcare becomes more socialized in the future, your tax burden would likely go up to pay for it. Just some food for thought from a “seasoned” healthcare professional.
That’s another fly in the ointment. You can’t be flexible later in life when you face healthcare and nursing home costs. Or strictly speaking it will require upward flexibility – the opposite of what MF&NM propose.
That said, t’s also many years and decades in the future, so not really correlated with Sequence Risk. But still a big headache.
Agree…at a minimum, it’s probably a good idea to model about $2k/month in additional expenses after the age of 80 ( I think your default SWR spreadsheet already includes something like this IIRC). Ideally this additional cost will be partially offset by social security income. I’m also targeting a high capital preservation percentage as an additional hedge.
But, it’s definitely a “slippery slope” once you start trying consider all the “What ifs”…you can talk yourself into never retiring or targeting a very low withrawal rate. I think you covered this psychological element in one of your posts and its very true. IIRC, your advice was basically to assume that 1/2 of your social security will be paid and also to include additional medical expenses in the cash flow tab after a certain age.
Anyway, after modelling various scenarios using your spreadsheet, with different captial preservation targets, using CAPE based rule (and not), I’m starting to hone in on a plan…likely 3% initial SCR adjusted for inflation. Given my portfolio (60/40), my time horizon (600 months), capital presevation targets (70%) and current market valuation (down 10% off the all time high), that’s what the SWR spreadsheet suggests. Using the CAPE based rule with similar params suggests a starting SCR of 3.6% but I’d rather not have to worry about potentially adjusting my spending downwards in future years. Instead, I might consider applying the CAPE rule each year as a sanity check as well as a way to gauge whether its a good time for extra “one off” purchases (buying a new car, going on a nicer vacation, etc)…the “upwards flexibiltiy” you talk about in this post.
Thank you for all your work in creating these blog posts. It’s so refreshing to read an early retirement blog that approaches the topic with rigor and well considering rationales beyond just hand waving and feel good platitudes like “just be flexible”. While the other early retirement blogs can be fun to reads as “mind candy”, I kind of feel like i’ve binged on junk food if I read too many of them. Your blog provides a well grounded counterpoint. I’ve read through all your posts, and am now going back and re-reading them again. I don’t always quite follow all the math, but appreciate the way that you summarize the key points. Having the spreadsheet to play with also helps put things in perspective.
Thanks for the feedback!
I took out the default supplemental flows because folks too often didn’t realize the flows were there and then got unexpected results. Now the default is $0. If you need to model higher expenses later, you’d have to enter that by hand! 🙂
Yeah, and that looks like a good plan. With a 70% final value target you need to curb initial withdrawals. Makes sense.
Cool- yeah probably better to leave the supplemental flows blank by default. I’ll admit that I got confused by that at first (the results weren’t making sense and then I happened to notice those cash flows).
Thanks! I’m glad I reset everything to 0. 🙂
How about a 3.4% (of the initial retirement portfolio) withdrawal floor, updated annually for inflation (typical fixed model) but then allowing for up to 5% of actual current portfolio value when above the floor (good market years). This seems to back-test well.
Have you backtested it? How about 1929 or 1968? You will probably hit the lower bound for an extended period (10+ years early-on in retirement). But if that’s still a happy retirement (it would be for me) then that seems legit.
Good call, I was using 1970+ simulator, using the complete set had to back off to 3.25% initial amount floor with a 5% of portfolio method. This would still be a nice retirement and consistent with your work on SWRs (not surprising). Point of the 5% of portfolio is to enable Fat FIRE type spending (extra travel, upgrades, etc.) when times are good. Yes, it can stay at the floor for a number of years – but we should be planning on a WR in the 3’s anyway based on all your work.
Yeah, probably apply a haircut if you want to hedge against a repeat of the earlier market events in 1929 and 1960s. No sim engine with 1970+ data will capture that kind of risk! But thanks for confirming!
You only have flexibility when your portfolio creates more income than you spend. Once you are spending it all, flexibility is gone. Investors believe they will get historic levels of return from stocks. With the massive amounts of debt and lack of population growth, globally, that is unlikely. To be safe, your portfolio needs an income cushion to continually grow the principal. That is difficult at a valuation level where the S&P 500 pays a 1.5% yield. A pile of mutual funds are not likely to do it anymore.
Yeah, that’s all worrisome. Maybe it all works out, but I share your concerns and like to play it safe and keep an extra cushion.
Fantastic job in detailing all of this out for us mere math mortals! The only thing I’d say though is I believe it’s healthy as well to consider the probability of your own death in the equation. I’m 52 and any scenario I plug into any calculator has a zero percent chance of me running out of money at 75 even if I double our current spending (been retired 9 years now), but going by the actuarial tables I’ve got a 8% chance I’m dead.
I’ve lost three people close to me under the age of 68, one only 55 this year all due to cancer. Each one two years ago was planning for the future and living like they were gong to be around forever. Too many people croak with way too much money, and yes probably more life’s a bit of a struggle financially at the end, but typically those are not the people reading this article.
Anyway what I’m saying is if you’ve got a 10% probability you’ll be dead, accepting a 10% probability you’ll run out of money is probably prudent. Easy for me to say admittedly it’s easy for me to say as I’ve overshot our needs financially, but if I hadn’t I think I’d prioritize experiences over being 100% safe. Just food for thought, love your work.
But you’d have a stressful retirement if you experience some bad Sequence Risk. I’d rather work a little longer and then retire worry-free.