The Global Withdrawal Rates chart identifies the most efficient retirement portfolios based on the worst retirement scenarios in multiple countries. Use this to explore unique portfolio options that worked well in a wide variety of economies, find a portfolio that looks beyond typical home country bias, and quickly validate ideas across oceans and borders.
Chart
Modify the inputs in black to update the image /// Instructions
If the chart doesn’t load after a few seconds, refresh your browser. Because of the large dataset, it can also be a little slow sometimes. So give it a few seconds to update when you change a setting.
Click to open
The interactive charts are sophisticated tools that push the limits of some mobile devices. If it doesn’t work, don’t give up! Visit this page on a laptop or desktop for the full experience.
Overview
Traditional withdrawal rates numbers like those found in famous retirement studies or the Withdrawal Rates tool typically study the retirement performance of a small handful of portfolios in one country at a time. The Global Withdrawal Rates chart takes a much broader approach and looks at the worst case scenarios for up to 8008 portfolios in 10 countries simultaneously.
Similar to the Portfolio Finder, it plots the 30-year safe withdrawal rate versus the ulcer index for every portfolio option to help visualize the full portfolio space in terms of retirement performance and pain. And adding its own new spin, it also studies the numbers from a truly global perspective.
By searching the full cloud of portfolio options, one can get a good feel for the types of portfolios that have been the most consistent for investors around the world.
Featured Discussion
Methodology
Asset Definitions
The Global Withdrawal Rates chart studies a narrowed set of asset options from the normal full Portfolio Charts collection. These assets are selected so that the data definitions cleanly translate across domestic, foreign, and global markets.
- Domestic Stocks — large cap blend
- Foreign Stocks — large cap blend, developed
- Domestic Bonds — 10-year treasury bonds (or equivalent gilts, bunds, etc.)
- Foreign Bonds — intermediate treasury bonds, developed, investment grade, unhedged
- Domestic Bills — 3-month bills, aka interest-bearing cash
- Commodities — broad global commodities such as a fund like GSG
- Gold — physical gold or an ETF like GLD that tracks the same
Data Notes
The data reflects realistic investment options for real-life investors rather than purely academic numbers. For example:
- Foreign bonds follow standard index fund methodologies and properly track investment-grade countries while excluding junk bonds.
- Recognizing the preponderance of certain types of index funds outside of the US, foreign stocks and bonds follow developed ex-US data for US investors and developed world data for all others. Because of the small percentage of any non-US country in the larger cap-weighted world index, the effect of overweighting the domestic market with domestic and world overlap is generally minor.
- For all data series the returns also account for the modern expense ratio for each asset. These fees are unique to each domestic market and represent what you’ll find if you shop for the required index funds today. Some things like commodities funds typically have higher expenses than stock funds, so it’s important to be honest about the fee impact. You can find more info on expense ratios in the Fund Finder.
- The calculations do not account for taxes. When interpreting the results, be sure to include your own estimated taxes in the expenses that your portfolio must cover every year.
Portfolio Combinations
Instead of simply looking at a small handful of portfolio options, the Global Withdrawal Rates chart finds every possible combination of those 7 assets (repeats allowed) using 10% intervals. That’s 8008 portfolios ranging from simple options like 100% domestic stocks to more complex mixes like 60% domestic stocks, 20% foreign bonds, 10% bills, and 10% gold.
Scenarios
When most people studying withdrawal rates talk about expanding scenarios, they generally end up looking backward for the oldest historical data possible. When looking at one country, that means that with a century of data you have at most 100 scenarios. And when every scenario is from the same country (especially a prosperity outlier like the United States), the economic conditions are probably less diverse than you realize.
The Global Withdrawal Rates tool takes a different approach and looks wide instead of deep for retirement scenarios. By looking at 40+ retirement years since 1970 in 10 different countries, it studies 400+ retirement scenarios. So the database contains not only many more scenarios of the typical long-term retirement study but also 10x the country diversity.
The countries studied represent the 10 largest developed markets (as defined by most index funds), so the data properly samples economic contemporaries with established economies. They cover 4 continents, 6 currencies, and some particularly unique economic outcomes in places like Japan, Italy, and Spain.
Spending Strategy
All calculations use the default fixed budget assumption. At the beginning of retirement, the total portfolio value is multiplied by the withdrawal rate to determine the annual budget. That money is deducted from the portfolio at the beginning of the year and set aside in a non-interest-bearing account to cover all expenses. The next year, the new spending budget is determined by adjusting the original budget by inflation, NOT by the new portfolio value including investing returns.
While generally considered a conservative system for retirement spending relative to other variable options, this is a nice baseline for capturing the ability of a portfolio to maintain purchasing power. And it’s the default assumption for most retirement studies you’ll read about.
Calculations
For all 8008 possible portfolio options, the tool calculates the withdrawal rate that depleted the portfolio in 30 years in all 400 historically accurate scenarios around the world. The calculations also account for more recent scenarios with at least 15 years of data and estimate where they’ll end up at the 30-year mark. The minimum number on record for each portfolio is its global withdrawal rate.
Alongside the withdrawal rate, it also finds the Ulcer Index for each portfolio in every country. The ulcer index is a metric invented by Peter Martin that quantifies the relative pain of owning a portfolio considering the length, depth, and frequency of all drawdowns. It’s my favorite measure of relative portfolio risk that captures the emotional investing experience far better than cold standard deviation numbers.
The Global Withdrawal Rate chart plots the minimum SWR vs. the maximum UI for every portfolio. It allows investors to select the assets and countries to be included or excluded from the analysis, and also provides several search tools to identify specific asset allocations in the portfolio cloud.
Calibration
Like testing a new measuring device to make sure it works correctly, whenever you see a new piece of research it’s helpful to compare the results to other respected studies. The short story is that I can accurately replicate their calculations given the same assumptions, and the numbers here should apply reasonably well to scenarios not wrecked by war. The long story may teach you something about their work as well.
Portfolio Charts vs. Bengen
The OG of retirement research was a man named William Bengen who wrote the paper Determining Withdrawal Rates Using Historical Data (pdf) back in 1994. In that paper, he looked at 5 combinations of large cap stocks and intermediate bonds in the US since 1926.
Studying every retirement start date on record (and using a similar projection method to my own to extend the start dates to 1976), Bengen found that 50% stocks was the sweet spot for the best withdrawal rates and that 4% was a good rule of thumb for 30 years. That’s where the famous 4% rule and its associated portfolio assumption comes from.
Looking at the same country and asset options, my method studies 11 portfolios instead of just 5. Here’s what that looks like.
Safe Withdrawal Rates
Bengen assumptions
Not only am I right on target with the 4% rule, but the ideal portfolio also matches his finding within the margin of error (his closest portfolio option was 50% stocks). To explore the numbers more precisely, here’s how his specific calculations compare to my own.
Portfolio-depleting withdrawal rates in the worst start year
50% US stocks, 50% US treasuries
Years to Failure | Bengen 1927-1976 start | PC 1970-2009 start |
---|---|---|
28 | 4.25 | 4.10 |
30 | – | 4.01 |
33 | 4.00 | 3.90 |
That’s a pretty darned good match, especially since our studied timeframes only overlap for a few years. For reference, Bengen’s worst case was the retirement period starting in 1966 while mine was in 1973. So the 1970’s largely determined both worst case scenarios. And a little surprisingly, my numbers are actually a bit more conservative.
There are a few possible reasons for that:
- My data accounts for expense ratios while his does not. Add those back in, and the results are even closer.
- Bengen deducted expenses at the end of the year instead of the beginning, which allows withdrawal rates to be a little higher.
- Differences in data sources can affect the numbers even for the same asset.
Take those three things into account, and I would say that my calculations match Bengen’s extremely well.
For those who naturally fixate on very long data histories for the most accurate withdrawal rate data, that fact may surprise you. After all, his data started in 1927 and covered the Great Depression while mine only started in 1970. So other than just double-checking my own calculations, I think there’s a good lesson here.
The US has been lucky enough historically that the painful-but-not-crippling 1970s were already near the bottom of the worst cases on record. And because of that local reality, looking back another 43 years really doesn’t make much difference in safe withdrawal rates. So if you truly want to explore worst case scenarios, you have to stop thinking only in terms of timeframe depth and start thinking about country breadth.
There’s more to investing research than just obsessing over US markets.
Portfolio Charts vs. Pfau
If I had to pick one retirement researcher who influenced me the most, it would be Wade Pfau. In 2010 he published a paper called An International Perspective on Safe Withdrawal Rates: The Demise of the 4% Rule? that took Bengen’s methodology and expanded it in two important ways.
First, Pfau applied the same idea outside of the US and calculated the 30 year SWR for 17 different countries using data since 1900. And second, he looked at every combination of stocks, bonds, and bills in 1% intervals. That’s 5,151 portfolios instead of just 5. By finding the specific portfolio that supported the highest SWRs in each country (an assumption he refers to as “perfect foresight”), Pfau upended the standard ideas about the appropriateness of fixed allocation advice for all people.
This is the visual that opened my eyes.
Maximum Sustainable Withdrawal Rate for 30 Years
By Percentage Allocation to Stocks, start dates: 1900-1979
Clearly the home country has a drastic effect not only on safe withdrawal rates but also on the ideal portfolio composition. In countries like the Netherlands 80% stocks worked best, while in others like Switzerland just 20% stocks was ideal.
While we’re here, also look at all of those bottom lines with the lowest safe withdrawal rates. Pay attention to the very strong bias towards stocks in those countries compared to many of the others. And put a pin in that thought, as it will come in handy with the Cederburg example below.
With my new calculation technique, I can largely replicate Pfau’s assets and portfolio optimization methodology to check my numbers against his for many countries outside of the US. In his paper, he offers a nice table that shows the 30-year SWRs for perfectly optimized portfolios of stocks, bonds, and bills in each country. Here’s how mine stack up for the 10 countries where we both have data.
30-year Safe Withdrawal Rates by Country
SWR (worst scenario year)
Country | Pfau 1900-1979 start | PC 1970-2009 start | PC-Pfau |
---|---|---|---|
Canada | 4.42 (1969) | 4.35 (1973) | -.07 |
United States | 4.02 (1969) | 4.23 (1973) | .21 |
United Kingdom | 3.77 (1900) | 4.16 (2000) | .39 |
Australia | 3.68 (1970) | 3.86 (1970) | .18 |
Netherlands | 3.36 (1941) | 4.62 (1999) | 1.26 |
Spain | 2.56 (1957) | 2.76 (1970) | .20 |
Italy | 1.56 (1944) | 3.26 (1970) | 1.70 |
France | 1.25 (1943) | 4.19 (1970) | 2.94 |
Germany | 1.14 (1914) | 4.65 (2000) | 3.51 |
Japan | 0.47 (1940) | 3.94 (1973) | 3.47 |
In the column on the right I list the measurable difference between our calculated SWRs. Half of them are quite good, with my numbers since 1970 over-estimating the longer-term SWRs since 1900 by an average of just 0.2%. But the other half in bold are not very close at all, and I think the reason for that is worth discussing.
Pfau was very transparent with his work and also provided the worst case years with his data. Look at the bolded countries and years. What do they all have in common? These worst case scenarios all resulted either from losing a World War or being directly occupied by an invading force. Clearly, being decimated by war is also devastating for retirement scenarios. Whether that’s a good baseline for your own decisions today is up for debate, but it’s an important fact to be aware of.
Another interesting calibration note from the Pfau data is the portfolio composition. Looking at the percent stocks numbers in his line chart, I can also compare that with my own ideal portfolios in each country. For the countries where our SWRs match well, our ideal percentage of stocks also match within +/- 10% most of the time. In some places like Spain it’s more than that, but still within the flat section of the line where the percent stocks doesn’t affect the withdrawal rate all that much anyway. And again, the outliers are primarily in those war-torn countries.
So here’s my balanced take.
If you want to know the behavior of portfolios in the countries most affected by the chaos after WWII, the Portfolio Charts data is clearly not sufficient for the job. I highly recommend Pfau’s work for that task. Dig deep enough, and you may come to the conclusion that no withdrawal rate will safeguard you if an atomic bomb drops on your home. And it’s true. War is terrible, and in those situations a comfortable retirement is the least of your worries.
But if you want to study portfolio ideas in some really rough scenarios — just not your entire market being wiped out — I feel pretty good about my numbers. Looking not only at safe withdrawal rates but also ideal portfolio compositions in non-war scenarios, I think the Portfolio Charts data matches the Pfau data reasonably well. At the very least, my global withdrawal rates are a good data-driven place to start that is far more conservative than numbers based only on outcomes in the United States.
And for those reading this who are concerned about the 1970 start date and want to know the effect of using more years of data, the numbers suggest that discounting my SWRs by about 0.2% should be in the right ballpark to match much longer-term timeframes in stable countries.
Portfolio Charts vs. Cederburg
On the topic of portfolio composition, another calibration point I’d like to explore is how my numbers compare to the conclusions from a new paper that has been making the rounds by Aizhan Anarkulova, Scott Cederburg, and Michael O’Doherty called Beyond the Status Quo: A Critical Assessment of Lifestyle Investment Advice. To save time, and simply because he’s been most public talking about the group work, I’ll refer to it as the Cederburg study.
The stated goal of the paper is to test the effectiveness of popular lifecycle funds consisting of shifting percentages of stocks and bonds over time. As part of that process, the authors clearly have strong opinions about the importance of using data free of typical US bias.
While I applaud their effort, in this context it’s important to understand that their chosen methodology is quite unique. From pooling randomized data from 38 countries and treating it all as fungible to utilizing actuarial tables for variable retirement length, their simulation system is nothing like the classic withdrawal rate methodologies. So replicating their withdrawal rates is a project for another day.
That said, the conclusion they draw that has gotten the most attention is the idea that the ideal retirement portfolio from a global perspective has no bonds at all. Specifically, when looking at different mixes of domestic stocks, foreign stocks, and domestic bonds, a portfolio of 35% domestic stocks and 65% foreign stocks achieved the best results using their mixed-country data model*. That’s a fairly controversial finding that goes against most retirement advice, and has inspired much debate among people interested in these things.
I figured this would be a good test to see how my own numbers compare, so I narrowed down the asset options to the three that they considered and let the cards fall where they may.
Global Withdrawal Rates
Using Cederberg study assumptions
(*) If you read the full Cederberg text, you may note that the main section also talks about two portfolios that contain bills. Their 35% conclusion comes from the Internet Appendix that looks at more granular portfolio combinations and does not consider bills at all. See Table B.II.
Ignore the SWR number and look at the asset allocation. Out of 66 possible portfolios, a portfolio of 20% domestic stocks and 80% foreign stocks was the best option. That’s already really close to their conclusion, and 30% domestic comes in at #6 just barely below that same point. So yeah, my method looking at 10 different countries largely matches their broad portfolio composition findings.
Does that mean you should rush out and load up on foreign stocks? Well, not so fast.
The relevant part of the Cederburg paper looks at three possible assets — domestic stocks, foreign stocks, and domestic bonds. When looking at worst case scenarios, I’ve found that the options one provides as foreign relief valves in times of high domestic pressure have a significant effect on the results. Why the bias to foreign stocks? And what effect does that have on the analysis?
Here’s what the ideal portfolios look like with different combinations of assets under consideration including a high-quality foreign bond index that only invests in investment-grade countries.
Global Withdrawal Rates
With different asset options
See the problem?
If you just look at domestic stocks and bonds, the ideal portfolio is 30% stocks and 70% bonds. The SWR is just miserable in the worst case of Spain in the 70’s. If you only allow an investor to choose a foreign stock fund, then of course that will look better than the other options.
If you allow foreign bonds instead of foreign stocks, however, then the ideal stock allocation actually goes down to just 20%. And if you allow the option of both foreign stocks and bonds, then the ideal portfolio is still only 60% stocks. Clearly the asset options matter.
There’s also an issue in this type of analysis with country sensitivity. To see how one worst-case country affects the numbers, here’s the same analysis with the default Cederburg asset assumptions excluding Spain.
Global Withdrawal Rates
Cederberg assumptions, excluding Spain
We’re back to 60% stocks again.
From what I can see, the only way to achieve the result of 100% stocks is to look at the data for terrible economies like Spain in the 1970s and artificially restrict the outside options to foreign stocks only. Is that really a fair summary of options available to investors?
Looking at Spain in this example and post-WWII Japan in the Pfau example, perhaps the true lesson here is not that bonds as a whole are undesirable but simply that investing in the bonds of a wrecked economy is a predictably terrible idea. This is also a good reminder that with any analysis looking at so much history, single summary numbers can sometimes mask important information.
So in conclusion, my own global withdrawal rate methodology also supports the high-level portfolio conclusions of the Cederburg paper when using the same asset assumptions. However, context matters. And for my own work, I allow freedom of exploration to help identify any limiting assumptions and seek better options.
Speaking of exploration, once you look at thousands more portfolio options than the Cederburg paper studied, the conclusions also could be worth revisiting.
Disclaimer
The tradeoff for having many more asset classes to model than the original studies is that there are fewer years of available historical data. The data here only goes back to 1970. There may be times before 1970 when a portfolio would have failed when a more recent one did not, and the resulting withdrawal rate would be lower. For more information on how that affects the numbers, see the calibration section above.
The withdrawal rates shown do not account for taxes, and one should note that asset classes like gold also have different tax treatment than stocks and bonds. Considering taxes, your personal withdrawal rate may be lower than the one shown.
And as always, keep this in mind:
Past results are no guarantee of future performance
Use this tool as a comparative guide to the effects of asset allocation on withdrawal rates, not as a guarantee of success. Just because something did great in the past does not mean it will continue to do so on your own personal timeframe. I personally believe withdrawal rate research is a wonderful way to help set financial goals and guidelines, but one should never put their life savings in the hands of a single back-tested number. Flexibility, intelligence, and determination will beat mechanical withdrawal rates every time!
Discussion
Join the conversation
Articles
Insights that reference Global Withdrawal Rates