When to Worry, When to Wing It: Withdrawal Rate Case Studies – SWR Series Part 47

August 18, 2021 – In my post two weeks ago I outlined my approach to retirement planning: In light of significant uncertainty in retirement, I like to do a more careful, robust, and scientific analysis. Not because I could ever undo any of the existing uncertainties but because I don’t want to add even more uncertainties through “winging it” in retirement.

But how much detail is really required? I can already hear objections like “you can never know your future spending month-by-month, so why go through all this careful analysis with a monthly withdrawal frequency?” To which I like to answer: Well, maybe that’s the part where you can indeed use the “wing it” approach! So, today I want to go through a few case studies and learn how much of a difference it would make in my safe withdrawal strategy simulations if we a) carefully model the whole shebang in great detail, or b) just wing it and use a rough average estimate for the spending path. For example…

• Does the intra-year distribution of withdrawals matter? In other words, how much of a difference does the withdrawal frequency make: monthly vs. quarterly vs. annual?
• What if there are fluctuations in my annual withdrawals around the baseline average budget, due to home repairs, health expenses, etc.?
• What if those fluctuations have an upward bias?
• What if there is a slow (upward) creep in withdrawals?
• What about nursing home expenses later in retirement?

Where can I safely wing it? And which are the ones I should worry about? Let’s take a look…

1: Intra-year fluctuations in withdrawals? Wing it!

In my withdrawal rate simulations, I always assume that you a) withdraw monthly and b) have a flat withdrawal profile in every year. That’s a slight deviation from the Trinity Study. They only have annual return data, hence, they are bound to make their model annual and thus the withdrawals have to be annual as well: at the beginning of each year, to be precise. I don’t quite get why anyone would do that, i.e., withdraw an entire year’s worth of retirement spending all in one big chunk and then slowly draw that down over the next 12 months, but so be it. Though I can also poke a few holes into my assumption: What if your withdrawals aren’t exactly flat throughout the year? People might spend more than one-twelfth of their annual budget during the summer months and then again during the holiday traveling season. And people might have large “chunky” expenses during specific months due to property tax bills and/or estimated federal income tax payments in January, April, June, and September.

So, this raises the question: How much of a difference does it make if we alter not the total annual withdrawals every year but the distribution of the spending and thus the withdrawal amounts intra-year? Let’s take a look at the following three scenarios with a 4% withdrawal rate, 75% Stock, 25% Bond portfolio over 30 years:

1. Baseline: 12 equal monthly withdrawals of 0.333% of the initial balance (adjusted for inflation) at the beginning of each month. The standard 4% Rule simulation in my toolkit.
2. Quarterly withdrawals: only 4 annual withdrawals at the beginning of month 1, 4, 7, 10, 13, 16, 19, 22,…, 355, 358, worth 1% of the initial balance (adjusted for inflation) and zero withdrawals in all the other months.
3. Annual withdrawals: Starting in month 1 and then on each retirement annivesary, withdraw 4% of the initial balance (adjusted for inflation). Zero withdrawals in all the other months.
4. Lumpy Monthly withdrawals: Imagine we start with a $2m portfolio and target a 4% withdrawal rate, i.e., $80k per year. Instead of withdrawing exactly $6,666.67 per month, assume we withdraw $5,000 in most months (months 1, 3, 4, 6, 7, 9, 10, 12) but then you also have some lumpy expenses (summer travel, holiday expenses, estimated tax payments, etc.) amounting to $10k, $12k, $8k, $10k in months 2, 5, 8, and 11, respectively. See the bar chart below. So, we still withdraw monthly but in installments of between 0.75x to 1.80x the monthly average. Quite a bit of spending volatility: a standard deviation of about $2,600 (levels) or 67% (month/month percentage changes).

In the spirit of the post last week, SWR Series Part 46, let’s see how the final distribution after 30 years differs using the three withdrawal frequency assumptions. See the chart below. Again, we note that there is significant uncertainty over your final net worth balance and there is nothing we can really do about that. But the good news is that the final distribution looks the same regardless of the withdrawal frequency. There are some slight changes in the extreme tail probabilities, i.e., the probability of running out of money slowly rises from 1.5% to 1.73% to 2.42% when we go from monthly to quarterly to annual withdrawals. But for the most part, the distributions look the same. The “Lumpy Monthly” scenario is even closer to the baseline and even has a slightly lower failure probability.

In the table below, I also assemble some more stats. The top half of the table again has some of the same data as in the histogram charts. But in the lower half, I also like to condition on expensive equity valuations (as in today), i.e., what are the SWR stats conditional on a CAPE>20 and the S&P 500 at an all-time high. The model here again assumes a 30-year horizon and a 25% target for the final portfolio when calculating the failure probabilities and failsafe withdrawal rates. If you’re a regular reader, you may recall that this is a nice universally useful assumption to capture both traditional retirees who have a 30-year horizon with a 25% final value target for a bequest but are also applicable to an early retiree who has to bridge 30 years until Social Security and company pensions kick in. In that case, the 25% final value target is to supplement the pension and government program.

The failure probabilities of the 4% Rule under that scenario are at about 22.5% for all scenarios. The failsafe withdrawal rate goes from 3.584% to 3.581% (quarterly) to 3.504% (annual) and 3.585% (monthly with fluctuations). And I should mention it here again, the reason why I display the SWR/fail-safe with so many digits is that because I like to showcase how little of a difference the withdrawal frequency makes. I don’t advocate pinning down your personal SWR to the closest 0.001%. That would be down to a $20 precision in the annual budget for a $2m portfolio. Not a good idea! Probably it’s ideal if you round your budget to the nearest $1,000/year = 0.05% of a $2m portfolio.

A quick side note: Why is it that by shifting to annual withdrawals, you get a slight deterioration in the failsafe rate? The earlier you withdraw the funds the less time they have to accumulate. Thus, on average, you should get better growth of your portfolio by withdrawing small sums more frequently. Hence, the deterioration in the average and median final net worth numbers when you reduce the frequency of withdrawals.

But how about the tail events? That can go either way. I looked at the more detailed results (not displayed here for brevity) and I noticed that during the Great Depression, you could have benefitted(!) from annual withdrawals because you withdrew a bigger chunk from the portfolio right before the 1929 crash. On the other hand, annual withdrawals would have hurt the 1965 retirement cohort because early on you still had moderate asset growth until the 1973 meltdown. The net effect of annual withdrawals mixing together the 1929 and 1965, 1966, and 1968 cohorts was slightly negative.

2: Fluctuations in annual withdrawals? Wing it!

Some critics have pointed out – correctly – that your spending will likely not be flat over the years. If you are a homeowner you will likely agree that home repairs and renovations are not spread out exactly uniformly over the years. Same with cars and other high-valued items; you may have to repair or even replace your car in one year and then have no major car expenses the next. Does that mean that my retirement withdrawal simulations are all moot? Well, let’s put that to the test.

Let’s compare the following scenarios:

1. Baseline as before: Monthly withdrawals, flat throughout the entire 30-year retirement. For example, if you have a $2m portfolio and a 4% withdrawal rate, you’d withdraw $80k every year. And that’s withdrawn monthly, in $6,666.67 installments.
2. Withdrawals are still monthly and level in every year, but they are 25% higher in odd years (including year 1) and 25% lower than the baseline in even years. Thus, the withdrawals jump back and forth between $100k in odd years and $60k in even years. That’s a quite a bit of fluctuation in annual withdrawals, by the way, -40% and +67% annual changes!
3. The same as before but we reverse the pattern: -25% in odd years (including the first year) and +25% in even years. So, to use the same example above again, this is someone who withdraws $80k per year on average, but will do so by withdrawing $60k in odd and $100k in even years.

Let’s see how the results shape up under the three scenarios. Below is the same table format. True, by frontloading your withdrawals (+25%/-25%) you get a slightly lower final net worth, and by starting with the lower of the two withdrawals (-25%/+25%) you get a slightly larger final net worth than under the baseline. But the differences are not really that large, considering the significant spending variability (+40%/-67%). Conditional on high equity valuations, the results are also almost identical. The fail-safe rates are all very close to each other, around 3.58%. The failsafe amounts would be about 0.7% lower or higher depending on whether we start with the high or low spending shock in the first year. Not really anything to sweat.

Also notice that, if we don’t really know ex-ante whether we will get the high or low spending shock in the first year and we model this as flipping a coin determining the first-year spending shock, we’d average over the two scenarios +25%/-25% and -25%/+25% and that average will land smack back at the baseline outcome. See the table below:

Thus, even sizable annual spending variability doesn’t appear to make much of a difference in withdrawal stats, as long as the fluctuations are centered around a baseline mean. Get your baseline budget right and beyond that, be my guest, just wing it!

Summary so far

So, from scenarios 1 and 2, what have I learned? Why don’t we see a much larger impact from deviating from the flat spending assumption? Very simple: Retirement success or failure is due to three major causes: First, Sequence of Return Risk. Second, Sequence of Return Risk. And third, Sequence of Return Risk. No, but seriously, if you recall from Part 14 and Part 15 of this series, Sequence Risk comes from an extended and deep drawdown in your portfolio which causes you to withdraw relatively more shares while the portfolio is down. You’d then draw down the portfolio so much that even an eventual recovery can not rescue your underwater nest egg. Specifically, during some of the historical bear markets, equities occasionally took more than a decade to recover in real terms (see my 2019 post “Who’s afraid of a Bear Market” for more fun facts). If the market is down for that many years, it doesn’t really matter so much how you distribute and spread out your withdrawals month after month. If the frequency of withdrawal fluctuations is much faster than the major market cycles, then spending fluctuations don’t have too much of an impact on Sequence Risk, if the average spending is roughly your annual budget.

But what if we didn’t get our baseline budget right? That brings me to the next case study…

3: Fluctuations with an upward bias? Worry!

Let’s keep the annual fluctuations from the previous case study with the following twist: Instead of fluctuating +/-25% assume that in odd years we hit our budget and withdraw the target, e.g., $80k if we started with a $2m portfolio. But in the even years, we overshoot and withdraw $100k. Think of this as a retiree who calibrated his or her budget to the best-case scenario where you have no surprise expenses. But then every second year, in the even years, reality catches up with you and “sh!t” happens, i.e., you have home repairs, you have much higher travel expenses because you attend your third cousin’s wedding, you have health expenditures, you replace your car, you upgrade your home, etc. The possibilities are endless.

That’s easy to simulate. I simply run the same sims as above but alternate between 0% and +25% spending shocks. Otherwise, the assumptions are exactly the same as in case study 2. The simulation results look a bit worrisome. You will shift the final net worth distribution significantly to the left. Not more “uncertainty” if measured as the standard deviation, but more risk of running out of money. The unconditional failure probability goes up from 1.5% to over 9%. The safe withdrawal amount conditional on expensive equities shrinks by 10%. Or, if keeping the withdrawal rate the same you would exacerbate your failure probabilities. This is something to worry about!

4: Retirement Spending Creep? Worry!

Another retirement variation and idiosyncrasy: what if your withdrawals creep up slowly, i.e., spending grows at a rate faster than inflation? The reasons for this spending creep are endless. Here are a few:

1. You’re “keeping up with the Joneses”: The average U.S. resident increases spending not just in line with inflation but also in line with average per capita GDP. That’s easily 1-2% annually. If you retire and commit to spending only your initial budget plus CPI inflation every year, keep in mind that your neighbors, friends and relatives around you are advancing their budget every year and you will fall behind. If you’re fine with that, more power to you. But not everyone wants to fall behind by 35% or 81% respectively when their peers are raising their standard of living by 1% and 2% per annum, respectively.
2. Your personal inflation rate is higher: Recall that the U.S. CPI measures the price index of an average consumption basket for an average urban consumer. Your personal rate can differ wildly. Just taking a look at all the different subcomponents of the CPI and how much variation there is in the cross-section raises the issue that if your basket is overweight on the higher inflation items (health care, rent, etc.) and underweights the low-inflation items (electronics) then your personal CPI might increase much faster than the overall CPI. So forget about “Keeping up with the Joneses” – merely keeping up with your own basket might necessitate raising your withdrawals faster than the CPI!
3. You no longer want an iPhone3: Sometimes people ask me whether the CPI makes adjustments for advances in product quality. The answer is, yes, definitely. And that can pose a problem. Imagine the average mobile phone becomes 30% cheaper every year. If you bought a phone for $600 two years ago, then it should cost around $294 today. But if your phone breaks you would likely not buy that same phone again. Instead you will likely buy a new state-of-the-art phone, costing maybe even more that the old one two years ago, say, $700. So, while the mobile phone category in the CPI might have a 30% annualized price drop, your out-of-pocket expenses for this category is up 8% per year. And again, the difference between the CPI and your personal spending changes comes from the fact the agency that constructs the CPI number (Bureau of Labor Statistics), does not compare the average mobile phone price in 2021 with the average phone price last year. It compares phones with identical charactestics across time! So, in other words, this is a form of “keeping up with the Tim Cookses”, i.e., raising your real spending because you buy better and better electronic devices.

So, let’s see what happens if we raise our real spending by 1% or 2% a year. That certainly has an impact on your withdrawal states. The unconditional failure probability of the 4% Rule goes up from 1.5% to 7.7% to almost 19%, when raising the withdrawals by 1% and 2% a year, respectively. Conditional on expensive equity valuations you go from 22.5% to 49% to 58%. Ouch! The failsafe initial withdrawal amount goes from about 3.6% to 3.2% and 2.85%, respectively. That’s a 10.6% and 20.4% drop, respectively, from the baseline with a flat spending profile. Spending creep is definitely something we cannot ignore! I certainly worry about that in my personal retirement planning!

5: Nursing Home Expenses? Wing it (mostly)!

Do you remember the Paula Pant podcast with Suze Orman? In a nutshell, Suze Orman proposes that you’d need at least $5 million or even better $10 million to hedge against all the uncertainties in retirement, including the possibility of nursing home expenses; up to $350k a year, her claim, not mine. My reply: When I retired at age 44, nursing home expenses were the last problem on my mind. It’s not like I said my Goodbye’s at the office on June 1, 2018, and then moved into a nursing home that same afternoon. I’d like to delay that move for at least another 40 years! Nursing home expenses most often hit retirees in their 80s or older. Irrelevant for at least another 30+ years for me and 40+ years for my wife.

But that doesn’t mean that we can completely ignore nursing home expenses. A large enough expense shock occurring decades into your retirement can still jeopardize your retirement. It all depends on how large the shock is and how many years into the future we’re talking about! Here are the scenarios I like to study:

1. Baseline: a 50-year retirement horizon with flat expenses and no nursing home stay. I assume a 75/25 asset allocation as before but a zero final net worth. One can think of this as a 40-year old early retiree who wants to ensure the money lasts until age 90.
2. A nursing home stay during the last 3 years of retirement (ages 87-90) with an additional annual cost worth 1/20 of the initial portfolio value. So, for example, if you start with a $2m initial portfolio we assume that the nursing home adds $100,000 in annual withdrawals on top of the baseline withdrawals. So, if you assume a 3.5% withdrawal rate from a $2m portfolio and thus $70,000 annual withdrawals, we would jack up the withdrawals to $170,000 during the last three years. That’s a pretty decent budget – not as much as the $350,000 annually for Suze Orman’s Ritz Carlton nursing home, but still quite generous!
3. A nuring home stay as before, but we assume that the expenses span the last ten (!) years of the investment horizon.
4. The same scenario as before, i.e., the nursing home stay begins in year 41, but only lasts yntil the end of year 43 and the retiree passes away after that. The idea here is that in most cases, a nursing home sty doesn’t last much longer than 3 years. Sure, an 80-year old may have an unconditional life expetancy of 10 years, but conditional on moving into a nursing home, that life expectancy usually melts down significantly. According to this source, most nursing home stays are 3 years or less. Thanks to Fritz at The Retirement Manifesto for the link!

Let’s take a look at the results. A nursing home stay in the final three years of your 50-year retirement horizon has a pretty negligible impact on your final net worth distribution. No surprise here. The unconditional failure probability of the 4% rule goes up from 1% to 2%. The market-peak conditional SWR stats slightly deteriorate with a failsafe withdrawal amount lower by 2.57%. In contrast, spending the entire final decade in the nursing home is certainly very impactful. The failsafe withdrawal amount is almost 10% lower than in the baseline. But a stay in the facility for a full ten years is also quite unlikely. Conditional on entering the facility in year 41, you probably also face a much lower life expectancy. If you live for only another 3 years, you have pretty much the same stats as in the baseline. In that sense, we can certainly “wing it” with respect to the nursing home expense risk.

Side note: Though it’s the first time I write about the nursing home issue in the context of the SWR Series, I have briefly looked at some simulations in a 2018 post in response to the Suze appearance on the Afford Anything podcast. Similar results: Not much of a difference in safe withdrawal rates especially if you account for the fact that ending up in the nursing home also likely lowers your life expectancy.

Bonus: Boredom in Retirement? Wing it!

Before I wrap up today’s post, I like to reiterate something I’ve pointed out numerous times before. But since not every reader has read all posts and listened to all the podcasts I’ve done over the years, here it goes again: when I announced my plan for retirement, especially early retirement, people inevitably asked me if might get bored. Well, that is one retirement uncertainty where I was willing to “wing it”. And I have so far. We traveled very extensively in 2018 and 2019. Never came even close to feeling bored then. Even with the recent travel restrictions and being cooped up at home, I never exactly felt bored. But I noticed that I have some extra bandwidth. Even back in 2018 when I retired, I knew that this would happen. But I had no idea what exactly I would want to do about it and I never really planned for it. I just winged it. And sure enough, starting in 2021 I started pursuing some other projects I’m passionate about. Winging the boredom part is much easier than winging an underwater portfolio!

Conclusion

There you have it. The retirement planning lesson for today: Never Say Never Wing It!

Specifically, some of our idiosyncrasies can be safely ignored, such as normal spending fluctuations (case studies 1 & 2). But getting your baseline budget right might be something you want to spend some time on. Fudging the numbers here and basing your average budget on the best-case scenario, i.e., lowest possible spending and not factoring in unexpected budget busters like home repairs, might be a recipe for failure (case study 3).

Another major headache is the issue of spending creep as in case study 4. Hey, it’s just a percent of spending creep per year, how much of a difference in the failsafe can that make? Well, it would be wise to lower the initial retirement budget by over 10%. And a 20% cut if you plan spending increases by 2% over inflation. And this is simulated over a 30-year horizon. Longer horizons would demand an even more drastic cut.

Somewhere in between is the nursing home scenario, case study 4. What makes the nursing home expense much less of an issue is that it’s the antithesis of Sequence Risk; it’s too far in the future to really matter. For early retirees with several decades before that big-ticket item might hit you, you can safely wing it today. The average traditional retiree might want to worry a little bit more but we still use the power of compounding to deal with this issue.

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!

Title picture credit: Pixabay.com

Technical Appendix

For those interested in replicating the results here, how would I simulate quarterly withdrawals in the SWR Simulation Google Sheet? Or any of the other deviations from flat withdrawal patterns? Very simple, here’s the trick.

First, go to the tab “Cash Flow Assist” in the Google SWR Sheet. Then set the scaling, which is set to 1.00 in the baseline (hence, a flat spending path!) to the pattern you like to model. For quarterly withdrawals set the scaling to 3.00, 0.00, 0.00, 3.00, 0.00, 0.00, 3.00, …

For annual: Set the scaling to 12.00 in months 1, 13, 25,… and zero in all other months.

For the -25%/+25% annual spending pattern, set the scaling to 0.75 in odd calendar years (= months 1-12, 25-36, 49-60,…) and to 1.25 in even calendar years (= months 13-24, 37-48, 61-72,…):

And likewise for the spending creep: Simply start with a scaling of 1.00 and then aply geometric growth to the scaling factors by 1% and 2%, respectively.