I simply value my privacy just like you. But more importantly, the site is not about me and the numbers do not care who I am. Like any good engineer who points to the verified measurements rather than to himself, I created Portfolio Charts to allow good data to speak for itself with no intermediary required.

Honestly, you should not trust me (or any financial personality) simply because of fame, credentials, or an impressive-sounding title. I take great care to document my data and assumptions so that who I am should not matter. That said, my engineering background does offer a unique perspective on how to study data, understand uncertainty, and evaluate risk that even the best CFP should appreciate.

As one part of a well-diversified asset allocation, I believe gold has proven to be measurably good at reducing overall portfolio volatility. I also do invest in gold with my own money. But no, I’m not a “goldbug” like the guys you hear in late-night commercials pushing bullion.

Most of the portfolios on Portfolio Charts contain no gold at all. I have also pointed out several times that while gold can be very useful in small to moderate percentages, investing too much in gold is pretty much the worst thing you can do as an investor.

So it’s all about balance. And if you just don’t care for gold, that’s also great! There are plenty of good portfolio options that don’t require it.

Portfolio Charts is purely an educational resource and should not be considered personal advice. I am happy to share good data to help you make an educated decision, but I have no professional qualifications to tell you how to invest your hard-earned money. If you need help, I recommend seeking out a professional adviser who can look at your specific situation and provide personal advice.

Everything on the site is free with no subscription required. There are no ads, affiliate links, data harvesting, or corporate sponsors. I make money through the optional generosity of patrons — memberships, product sales, and donations. If you like what I do, your support is greatly appreciated!

Many people correctly point out that past performance is no guarantee of future returns. While that is an accurate warning, some people take it a step too far and argue that backtesting is irrelevant or even deceptive for planning purposes. I believe that is reductive and throws away valuable context.

Proper backtests that look at the full range of historical outcomes (including the best, worst, and everything in between) can be extremely educational for understanding the underlying nature of a portfolio. Some (like stock-heavy portfolios) are naturally volatile, while others (like cash-heavy portfolios) are calmer. No portfolio is perfectly predictable, but not all portfolios are equally unpredictable.

While there are never guarantees in investing, finding a portfolio that would have met your needs even in the not-so-good times to invest will allow you to invest with confidence. Portfolio Charts is uniquely good at doing just that.

The numbers behind the assets are a personal collection compiled from many reputable public sources including annual fund reports, index providers, academic research, and a good number of my own calculations. The goal is to present accurate, consistent, and actionable data that dutifully tracks real-world performance. I document everything on the Data Sources page.

Like other tools and studies, the timeframe covered is simply a matter of data availability.

Retirement studies often default to very long datasets by convention. The Trinity and Bengen studies used data going back to 1926 not because that was a magical date for safe withdrawal rates but because that’s what was readily available from Ibbotson Associates for the indices they studied. Wade Pfau and Michael Kitces used data since the early 1870s because that’s what was freely available from Robert Shiller. And Pfau later did the same analysis for developed international markets since 1900 because that’s what was available from the DMS dataset. All three sources have limited information on diverse asset classes beyond large cap stocks, bonds, and T-bills.

Likewise, modern competing backtesting tools that have very wide breadth (down to thousands of specific tickers) often only have as much data as the youngest fund in the portfolio. It is not uncommon to only have a decade of data if you’re lucky, and having numbers all the way to 1970 would be considered a rare luxury.

Good data for different markets is very difficult to find. If you know where to find reliable older data, please contact me.

All numbers on Portfolio Charts are adjusted for inflation. I believe this is the best way to study all investing returns because it reflects the true change in purchasing power of your investments. Many other tools report nominal numbers which will always be higher. So when comparing numbers to outside data, make sure they are also adjusted for inflation.

All calculations assume that dividends and interest are reinvested and that the portfolio is rebalanced once a year.

Taxes are impossible to calculate for every person in a global audience, so they are not included. Fund fees are also excluded, although the Assets section can help you find applicable ETFs with the lowest fees. When planning for your own budget, be sure to account for taxes and fees as a line item that the portfolio must support.

The underlying data is updated once a year in early January.

While single long-term averages starting before 1975 may certainly be skewed by a one-time surge in the gold price for reasons that won’t repeat, the increase was actually not that unusual historically and many portfolios containing gold did just as well after it ended.

But even more importantly, the Portfolio Charts backtesting methodology is designed to specifically see past one-time bubbles in any asset — not just gold — by studying all start dates simultaneously. When you look at the best case, worst case, and everything in between, one-time price events are obvious, not hidden.

Unlike most tools that cater heavily to investors in the United States, Portfolio Charts is fairly unique in its ability to study portfolios from the perspective of investors around the world. Change the home country and it translates the numbers to the currency and inflation of where you live. This will allow you to study how portfolios performed relative to your local purchasing power.

Beyond calculation assumptions, the home country also affects two other things on the site: asset options and portfolio definitions. The Assets section offers ETFs for each asset type that are available where you live. And the standard portfolio definitions translate the design intent to preserve the domestic/international framing of the original portfolio idea. But if you don’t care for the automatic translation, you can also model any iteration you want using the Charts.

Portfolio Charts is a little different from other backtesting tools in how it focuses on indexes instead of tickers. Indexes are the underlying collections of securities that individual tickers are designed to track, and focusing on the big picture not only allows us to ignore the noise in minor fund differences but also expand data histories to look back far earlier than your favorite fund probably existed.

This methodology section explains more about how to understand the standard assets in Portfolio Charts. And for a head start translating your own portfolio to the Portfolio Charts asset paradigm, try the Assets page. Type any ticker or ISIN into the search box (look for Search right below the main image), and it will show you exactly which asset class that fund belongs to. The database does not include every fund on the market, but if your ETF is a standard index fund matching a site asset there’s a good chance it’s in the list.

When searching specifically for worst-case scenarios, the sampling rate matters. The more datapoints you have to search, the more opportunities you have to identify a new worst case.

Portfolio Charts uses annual data. Many similar sites use monthly data, and a few use daily. Even when studying identical portfolios with the exact same source, referencing annual data will report the mildest drawdowns and daily data will report the deepest.

That said, sampling granularity generally affects all portfolios equally. So even coarse annual drawdown data is still very valuable for relative portfolio comparisons. If the absolute number is most important, I recommend supplementing the drawdown numbers here with external sites that use more granular data.

In addition, be sure to check that the source you are reading accounts for inflation. All numbers on Portfolio Charts account for inflation.

Higher returns do help, but it’s more complicated than that. Safe Withdrawal Rates are highly influenced not only by average returns but also by annual volatility. All things being equal, higher volatility lowers the SWR. So two portfolios with equal average returns may have drastically different withdrawal rates based on the underlying volatility. And sometimes portfolios with lower average returns but lower volatility can still have higher SWRs. That’s how you get interesting results like this:

This plots the returns and calculated SWRs of a variety of different popular portfolios (data through 2025). Note that the Total Stock Market had the second-highest return but the lowest SWR. Also note that none of the top five SWR portfolios have more than 50% stocks. Diversification plays a larger role than many realize.

Basically, volatility and downside risk are just as important as long-term returns percentages in SWR calculations. One should look beyond returns numbers that mask the underlying volatility and consider the positive effects of diversification when studying withdrawal rates.

Please don’t! At least not yet. A good retirement plan has many features and contingencies in place other than a simple safe withdrawal rate. You should never put your faith solely in a SWR number from any source, here or elsewhere. Make sure you have a good understanding of both your proposed portfolio and also your proposed lifestyle before you jump to any conclusions.

The most important takeaway is that different portfolios never considered by the various retirement studies may have different withdrawal rates. So while the tools here can only look back as far as the data availability allows, the results may be more relevant for your own personal asset allocation. Consider the withdrawal rates as a maximum starting point, and assume that there are times in the past and the future when they are lower than what you see here. Be smart about it, and plan conservatively.

Preferring other studies for their longer data histories (provided you also follow their investment assumptions) is perfectly reasonable. The one thing I would caution against is blindly following a withdrawal rate calculated for a very specific set of funds that you do not personally own. Not all stocks and bonds are created equal, and individual assets differ far more than you may realize.

I believe the idea that one should use the most data possible to find the worst-case scenario is well-reasoned. However, there is more than one way to look for those scenarios. The traditional way is to look at deep US data going back as long as possible, but one could argue that focusing solely on the US is also limiting. Another approach is to look at broad scenarios across many countries.

For people who prefer the most data possible, the Global Explorer accomplishes this by studying safe withdrawal rates since 1970 across a dozen different countries. That actually provides far more distinct retirement scenarios to study (both in quantity and genuine economic variety, as countries like Japan and Spain have experienced truly tough times compared to the US) than even the very longest traditional retirement study going back to the late 1800s in the United States alone. And it also covers more than 43,000 portfolio options instead of just a handful.

The Withdrawal Rates chart does not simply display the results for the small handful of 30-year retirement periods. It calculates the lowest safe withdrawal rate over ever-growing timeframes based on how withdrawal rates mathematically decay over time. That allows it to not only project 30-year withdrawal rates for investing periods originating less than 30 years ago, but to also predict the results for even longer retirement periods. The end result is a best estimate that provides conservative figures based on known poor retirement start dates.

/// How it Works ///

It depends on the asset allocation, and most retirement studies are very limited in that regard. But we can certainly compare several portfolios apples-to-apples to get a feel for how the timeframe affects the numbers.

First, let’s compare the 30-year SWRs from the Withdrawal Rates calculator to those from three different retirement studies using more years of data.

William Bengen used data since 1926 for a 50/50 portfolio of large cap stocks and intermediate treasuries. He found that an initial withdrawal rate of 4% lasted 33 years while 4.25% lasted about 28 years, and concluded that 4% was a good rule of thumb for 30 years. The tools here directly calculate that the 30-year SWR since 1970 was about 4.2% for the same portfolio. Let’s go with his rounding and say that the Portfolio Charts numbers are 0.2% higher.

Wade Pfau used the same Bengen methodology with large cap stocks and T-bills starting in every year since 1870. He found that the 30-year SWR was about 4%, while the tools here calculate that the 30-year SWR since 1970 was 4.3% for the same portfolio. So the Portfolio Charts numbers are 0.3% higher.

Michael Kitces used a 60/40 portfolio of large cap stocks and intermediate treasuries with data since 1871. Kitces concludes that the 30-year SWR for this portfolio was 4.5%. Interestingly, the SWR calculated here for the same portfolio since 1970 was 4.2%. So the Portfolio Charts number is actually 0.3% lower than his very long term number. The difference is likely a result of our different sources for the same asset data, a point that Kitces himself notes can affect SWRs by about half a percent. This is also why his numbers differ from Bengen’s by about half a percent for very similar portfolios.

Looking at the three studies, the Portfolio Charts numbers since 1970 vary from known SWRs calculated using much longer timeframes by about 0.3% for the same portfolios.

Next, let’s set the Portfolio Charts data aside and look directly at the data from the Pfau and Kitces studies.

The horizontal colored lines are my own addition to mark the SWRs calculated over different timeframes. The orange line marks the low-point since 1870, which is right at 4%. The blue line marks the low point since 1970 with a SWR of about 4.3%. So Pfau’s data also shows a 0.3% difference for a SWR calculated since 1870 and one calculated since 1970.

No colored lines this time, but you get the idea. Look at the lowest point, and compare it to the one starting in 1973. The difference is in the same 0.3% ballpark.

In summary, the numbers calculated here since 1970 vary by only about 0.3% from a variety of different reputable retirement studies with much longer data sets. Also, both the Pfau and Kitces charts independently demonstrate that the absolute lowest SWRs since the 1870s were only about 0.3% lower than those since 1970 for the portfolios they studied. Finally, 0.3% is within the known margin of error caused by variation in data sources for the same index.

I believe that’s a reasonable amount of error for general research purposes, but I always recommend that people plan conservatively. For investors understandably concerned about the timeframe differential, note that simply using the more conservative 40-year results available on this site makes up the difference for the error between various studies. An investor using the Perpetual or Long-Term withdrawal rate would be in even better shape.

Calculating the 30-year SWR for a traditional portfolio consisting of large cap blend stocks and intermediate bonds since 1970 usually differs from much longer studies by about +/- 0.3%. (See “How much of a difference does the 1970 start date make…” for more info.)

One example of an asset allocation with lots of asset diversification is the Ultimate Buy and Hold portfolio. It specifically calls for 10 different types of stocks (including REITs) and two different types of bonds. Reducing the assets to the restricted options used in most longer studies, the same portfolio would be described as 60% large cap blend and 40% intermediate bonds.

Since 1970, the measurable 30-year SWR for the true Ultimate B&H portfolio was 5.1%. The SWR over the same timeframe for the simplified 60/40 version was 4.1%. So replacing the diverse collection of stocks and bonds with over-simplified data understated the SWR by 1%. Note that this difference also varies quite a bit by portfolio and by country. In some cases, the error from questionable asset substitutions can be much higher!

While every portfolio is different, the error from only using large cap blend and intermediate bonds to model all types of stocks and bonds can be many times higher than the error from only using data since 1970. So when choosing any withdrawal rates study to model your own portfolio, look at your specific assets and think about how both sources of error might affect the results.

One should not assume the worst year will be the same for all portfolios. SWRs are calculated by studying every possible start year and identifying the single worst one for each individual portfolio. 1966 was the worst year for the old SWR studies that only looked at the S&P 500 and a broad bond fund, but other more diverse asset allocations may have had very different worst years.

One can use similar methodologies to find the worst year for other portfolios. For example, compare the various 30-Year SWRs for the Classic 60-40 and Permanent Portfolio:

The worst retirement year for the Classic 60-40 (in the data we have available) was in 1973, at the start of a decade of extremely high inflation (that negated stock gains) and rapidly rising interest rates (that killed bonds). Data from multiple reputable sources shows this was only slightly better than retiring in 1966 (see the “How much of a difference does the 1970 start date make…” question above). The worst retirement year for the Permanent Portfolio was 1980, the peak of the gold spike that preceded an 80% drop in the gold price. Not all assets are created equal, and different portfolios may perform better or worse in different economic conditions.

Without speaking to the methodology of the outside source, the technical answer is that both may be right about the raw numbers. But the very important caveat is that a knowledge of history is required to properly interpret gold data.

It’s common for withdrawal rate researchers to want to use as much historical data as possible. What many people miss, however, is the unique history of gold that makes data prior to 1971 completely irrelevant to modern investors. Unlike some other assets where people argue about how its premium may have changed over time, it’s also not a simple matter of opinion. The unique legal history of gold that explicitly defined how it is priced at all has profound effects on safe withdrawal rates.

Long story short, for a century prior to 1971 gold had exactly zero nominal return due to the gold standard that defined how currencies work. And it also wasn’t just a US thing, as the Bretton Woods treaty fixed the price of gold globally. To learn more, read these two articles about the gold history and how using inapplicable data suppresses withdrawal rates.

For the record, all data here on Portfolio Charts goes back to 1970. While that timeframe is simply a result of overall data availability for the many assets I track, a nice side effect is that the dataset accurately reflects gold performance in the modern era. The end result is that the data here applies to modern investors, while backtests using gold data before 1970 do not.