Withdrawal rates have always been a passion of mine. Beyond a natural desire to judiciously plan and manage my own early retirement, I’ve also been fascinated with withdrawal rate mechanics from a purely intellectual perspective. Not only is it an interesting topic, but I also like how there’s still lots of room for new development.
So rather than simply defaulting to the same methodologies of the classic Bengen or Trinity papers, I’ve always strived to build upon their solid intellectual foundations while bringing my own unique perspective to the table. By tweaking the default assumptions like the retirement length, home country, or available fund options, I believe people can gain a greater appreciation for investing options far more interesting than the old 4% rule of thumb implies.
That desire for deeper understanding is only matched by my persistence in seeking ways to improve my calculations, and one project I’ve been working on for a while now is the way I calculate safe and perpetual withdrawal rates. I just released an important new update to the Withdrawal Rates tool, and to best explain it I think it might help to step back and start from the beginning.
So if you’ve ever used Portfolio Charts to research your own retirement portfolio or just want to stay up to date with the latest in withdrawal rate calculation techniques, read on.
How safe withdrawal rates are calculated
If you’re new to the topic or just want a quick refresher, safe withdrawal rates look at the historical sequence of returns for a portfolio and measure the amount of spending that would have ended with at least $1 in the account even in the worst economic situation on record. That spending is measured as a percentage of the original portfolio value, and the assumption is that the original budget is adjusted by inflation every year, not the ever-changing portfolio value. The idea is to identify a safe level of comfortable spending power that you can count on to last.
Different people researching safe withdrawal rates have different assumptions and methodologies, and it’s important to understand them when comparing numbers. But traditionally speaking, most researchers you’ve heard of use real-world sequence of returns rather than theoretical models, assume that retirement lasts 30 years, and tend to focus on large cap stocks and short or intermediate bonds (usually in the United States) where there is the most easily available data.
So as a quick baseline, let’s start with the Classic 60-40 portfolio and plot the 30-year safe withdrawal rate for someone who retired in 1970.
One datapoint may be interesting, but it definitely doesn’t tell the whole story. So to begin expanding the possibilities, the first thing to point out is that the traditional 30-year retirement assumption is pretty arbitrary. It was originally based on the idea that a normal retirement age is 65 and a normal lifespan is about 95, which to be fair is a reasonable place to start. But not everybody is normal. Some work well into old age, while others wish to retire early.
Importantly, safe withdrawal rates vary a LOT by the length of retirement. How much? Here’s the same chart tracking the real-world safe withdrawal rates of a 1970 retiree for every retirement length we have data for.
While we’re here, allow me to point out a few things.
- First, a SWR of 100% makes perfect sense if your retirement is for only one year. My withdrawal rate calculations assume that spending for the year is deducted at the beginning of the year in order to cover all expenses. So whatever you have at the start of the year, you just spend it all if you don’t need anything to last beyond one year.
- Second, keen observers may be wondering why a database from 1970 through 2023 has 55 retirement samples instead of 53. The answer is that 1. there are 54 full years from Jan 1970 to Dec 2023, and 2. as we just discussed, the income for the first retirement of one year actually comes in year zero before any investing returns even start.
So even though the shape of the curve may not be what you expected, the endpoints do make sense as a sanity check. And to build on the curve shape, it’s important to think beyond single start years.
While retirement performance starting in 1970 may be interesting in a dry historical sense, on its own that information is not very useful to anyone today. After all, economic conditions are very different now than they were in 1970, and if you retire today your own sequence of returns will be very different. So a more useful exercise is to chart the same curves starting not just in 1970, but in every year since 1970.
Now we’re getting somewhere and can start to see the big picture. Safe withdrawal rates are chaotic and uncertain, but they follow a few basic rules like a waterfall flowing into a river. They all start at the same point, fall rapidly as retirement lengths grow, always flow downhill, and eventually level out over time.
While you’re thinking about waterfalls, this is a good point to call out a nice bit of safe withdrawal rate research that I really like from Michael Kitces. In his article What Returns Are Safe Withdrawal Rates REALLY Based Upon?, Kitces points out that long-term safe withdrawal rates are largely determined by portfolio performance in the first 15 years of retirement. I think the above chart serves as a nice visual representation of his observation, as each line effectively turns the corner and flattens out substantially by the 15-year mark.
Next, let’s zoom in from the extremes and look at the exact same data within the range most people are interested in.
Here we can better appreciate the spread of outcomes. The red point marking the 30-year SWR starting in 1970 is the same as before, but clearly it’s not the most important number. What we really want is the minimum 30-year outcome since 1970 marked by the orange point. That’s what Bill Bengen called the SAFEMAX withdrawal rate. Others prefer statistical metrics like the 5th percentile outcome likely similar to our red dot, but I’ve always liked Bengen’s methodology for its simplicity and clarity.
So following Bengen’s basic methodology, I have always defined safe withdrawal rates as the minimum on record. And expanding it to multiple timeframes beyond a fixed 30 years, I chart all of the blue lines and seek out the minimum. Here you can see that with the orange line.
The importance of withdrawal rate projections
One of the things I love about visualizations is how they allow you to see patterns that people staring at a sea of raw numbers often miss. And by thinking beyond fixed retirement lengths living only on the vertical gray line and looking at the full blue curves, an important aspect of withdrawal rates jumps out.
How do you account for the endpoints?
The shorter blue lines are from more recent start dates, and their numbers are no less important than the long-term results. While the last chart may not seem like a problem, consider the situation of a 100% Total US Stock Market portfolio.
See that lighter blue line at the bottom that ends at 25 years? That represents the safe withdrawal rates of a person who retired in 2000. Even though we do not yet have 30 years of data from that cohort, we know for certain that the ultimate 30-year SWR will set a new all-time low. How do we know? Because it’s already at the lowest 30-year withdrawal rate level and, just like water, WRs always flow downhill with more data.
To illustrate that mechanic in a way that’s easy to understand, here’s a quick animation of how the lines change each year as new data becomes available. Even if you’re not 100% accurate to a perfect tenth of a percent, I wager you can predict line trajectories pretty well even with no information about future portfolio returns.
Think for a moment about any retirement study that focuses solely on full 30-year timeframes. That 4.4% withdrawal rate may feel secure for a total stock market portfolio, but we know for a fact it’s already outdated by a particularly poor start date less than 30 years ago. It’s only a matter of time before the SWR is revised down.
Wouldn’t you like to know that before starting your retirement?
To be fair, in this example the difference still feels in the range of rounding error that is perhaps not worth worrying about. Portfolio Charts also expands asset options way beyond simple US-based stocks and bonds, so let’s scan the portfolios and talk about a more extreme case. Here’s the same chart for the Ivy Portfolio.
Here you can really start to appreciate the issue with ignoring newer data in long-term withdrawal rates. If you only look at the minimums for each full timeframe, the SWRs would look like this.
See the problem? By fixating only on the points on the vertical line and ignoring the trajectories along the way, you run the risk of drastically overstating the safe withdrawal rate that’s obviously coming.
How to predict withdrawal rates
This is the point in the explanation where opinions depart on the best way to proceed.
Many withdrawal rate studies just ignore the issue entirely, fixating on single retirement periods and reporting the worst directly measurable case on record. I respect the consistency and dedication to historical accuracy, but I dislike the idea of ignoring the clear warning signs for impending safe withdrawal rate reductions. There has to be a middle ground somewhere.
Others attempt to fill in the gaps using more recent performance datapoints by feeding them into Monte Carlo simulations that run tens of thousands of randomized scenarios. That may sound impressive, but by definition absolutely none of those scenarios correctly model even a single real-world historical result. And depending on the bootstrapping assumptions, the model may even introduce its own bias and influence the outcomes. Simulations have their place, but as a fan of accurate histories over theoretical models it’s not my favorite technique in this application.
I’ve always thought there must be a better way, and my method to date has been to stick to real-world data and project withdrawal rates using simple curve matching techniques.
In this case the projection method finds the lowest endpoint and parallels the long lines that came before. This has been my method of choice for estimating long-term withdrawal rates from more recent start dates.
But three things have always bothered me about this method. First, you can see the a “hitch” in the line from it’s original endpoint where it clearly isn’t a smooth trajectory with the same curvature. Second, it doesn’t account for other endpoints that may cross that worst-case line in the future. And third, I’ve always wanted a more elegant approach that references the core portfolio data rather than detached curve matching tricks.
Well, after a lot of research I’ve settled on a new system. Here’s a snapshot of the same portfolio using my new method.
See the improvement?
Using the new method, every solid blue line is known historical data and every dotted line is a projection. Those projections are way more smooth and accurate than before, and rather than just projecting one particularly low endpoint, I now check them all.
Well, almost all. Remember how we talked about how the first 15 years largely determine a portfolio’s ultimate trajectory? To prevent whiplashing projections resulting from too little data where the youngest blue lines have yet to “turn the corner”, I only project retirement scenarios with at least 15 years under their belts. But once there’s enough data to work with, the chart takes it from there.
Here’s how it works under the hood. For each portfolio I measure the longest inflation-adjusted CAGR we have data for. You can see that number for yourself by using the Heat Map and seeking the longest-term real CAGR at the top right when maximizing the black “years held” box. When there is at least 15 years of real-world sequence of returns data to use in the withdrawal rate calculations, that’s what is always prioritized. And when we reach a retirement endpoint with no more data, I assume that the portfolio will continue to grow at the same very long-term real CAGR.
Of course no projection is perfect, but the beauty of this method is that it will only get more accurate with time. And in the meantime, it offers valuable context that I believe is critically important for modern retirees. Consider the three 30-year Ivy Portfolio SWRs I illustrated above with the orange points, all calculated from the exact same dataset:
30-year Ivy Portfolio SWRs
No projection: 5.8%
Old projection: 4.8%
New projection: 4.3%
I would argue that the data with no projections could be considered misleading and sets people up for failure. The old projection system was a big improvement, but the new one is more thorough, accurate, and conservative.
While I’ve focused on safe withdrawal rates in this discussion, it’s also important to note that the new calculations also apply to the perpetual withdrawal rate numbers. For those not familiar with the term, instead of spending a portfolio down to zero, perpetual withdrawal rates are designed to maintain the initial inflation-adjusted principal. Due to the new system that projects every start date, many of the perpetual numbers may be noticeably lower than before! So if that metric is important to you, be sure to check your favorite portfolio for the most up-to-date perpetual withdrawal rates.
My goal with Portfolio Charts has always been to steer people in the most educated and responsible direction given the best data available. Put it all together, and I’m happy to say that the new projection system is the next important step in my larger mission.
Investing for the long term
I realize that’s a lot of background information, but hopefully it explains my thought process in a clear manner that helps you trust the underlying Withdrawal Rates methodology. Now that we’re past the heavy stuff, I’m also excited to discuss another cool new feature.
Did you notice in the above examples how withdrawal rates for a given portfolio tend to approach a steady long-term floor over time? You weren’t imagining things! It’s a real phenomenon, and thanks to the new projection system I can also quantify that limit. Let’s call it the Long-Term Withdrawal Rate.
Technically speaking, the long-term withdrawal rate is the halfway point between the projected 60-year SWR and PWR for a given portfolio. Basically, it’s the number that both metrics approach given enough time. And in more actionable terms, it’s my new favorite target for early retirees or for investors of any age who want to emulate an endowment and make their money last forever.
To illustrate an interesting characteristic of long-term withdrawal rates, I’m going to break out a different visual. Here’s the top half of the Retirement Spending chart that tracks account values over time in retirement. The spending rules are set to the same assumptions as the classic Withdrawal Rates calculations, the portfolio is the same as the Classic 60-40 shown in the Withdrawal Rates chart above, and the three images simply alternate between using the 30-year safe, perpetual, and long-term withdrawal rates for the same portfolio.
Pay attention to the bottom line at the 30-year mark.
- At the 4.4% safe withdrawal rate, you can see that in the worst case the portfolio lasted exactly 30 years. That matches the traditional SWR definition perfectly.
- At the 3.4% perpetual withdrawal rate, the worst-case line touches the horizontal line tracking the initial inflation-adjusted principal. That’s how PWRs work.
- At the 3.7% long-term withdrawal rate, the worst-case line doesn’t fail or remain completely untouched at 30 years. Instead, it trends horizontally within a “normal” range of safe portfolio variability. After all, portfolios go up and down all the time. Just because it’s below its initial value does not necessarily mean it’s on an unsafe track.
So if SWRs are potentially too aggressive for early retirees and PWRs are arguably too conservative, LTWRs are a nice balance in terms of sustaining a portfolio long-term without going overboard. I would argue it’s the Goldilocks retirement metric — not too hot, not too cold.
The long-term withdrawal rate is just right.
Everything affected
Between the projection updates and the new long-term withdrawal rate data, there’s a decent amount of new information around Portfolio Charts. Here’s everything that is affected:
- The Withdrawal Rates chart contains all new data.
- The Portfolio Matrix and Risk and Return comparison charts both reflect new safe and perpetual withdrawal rates.
- The Financial Independence chart now uses the long-term withdrawal rate as a savings target rather than the old 40-year SWR metric. It’s a little more conservative and in line with the overall financial independence theme.
- Each Portfolio page and the My Portfolio tool is up-to-date with new info.
- Owners of the Toolkit should check their email for a new spreadsheet download link that updates everything to match the main website. (UPDATE: Make that two, as astute users already found one error that I’ve already fixed. Keep the feedback coming!)
With so many pages being touched, it’s possible that there’s a mistake somewhere. So if you see something that looks off, please let me know.
What do your own retirement projections look like?
Even the most fascinating numbers are only as useful as your ability to put them to good use. The cool thing about withdrawal rates is that they relate to perhaps the most high-impact financial goal that you can aim for in your life — freedom. The freedom to say no without fear. To pursue interests without regard to their financial reward. And to generally live life on your terms.
As you browse the updated data and think about your own retirement prospects, be sure to think beyond any old caricature of retirement you have in your head. Your life is what you make of it. How can things like safe and long-term withdrawal rates using intelligent asset allocation empower you to take charge today? If you’ve ever wanted to set yourself up to think primarily about impact rather than income, truly planning for the long term may open your eyes and lessen your burden.
I first studied withdrawal rates in depth to plan my own course change, and I feel blessed that my path of greatest impact involves sharing what I learned with you. So I hope you enjoy the new data. But most of all, my true desire is that some of you may use the concepts along with your own empowered initiative to change your life for the better.
Trust me on this. Truly internalize the new and improved Withdrawal Rates chart, and your own path forward gets a lot more fun.
Join the conversation
