What Vanguard Gets Right (and Wrong) About the 4% Rule


Vanguard has always been one of the most respected voices in the investing industry, and its values have defined a whole generation of DIY investors. Jack Bogle started the first index fund in 1976 and took a lot of heat for that decision among his contemporaries who felt that “Bogle’s Folly” was destined to fail. But he was a true visionary, and his dedication to low costs and customer-centric business practices helped build the company into the financial titan it is today. So when Vanguard speaks, people listen.

Thanks to its selection of low-fee index funds, Vanguard also owns a permanent soft spot in the hearts of the small niche of investors who consider themselves members of the FIRE movement. For the uninitiated, FIRE is shorthand for Financial Independence Retire Early. It’s a community of young, motivated people who seek to leave the work treadmill far before traditional retirement age through maximizing the happiness utility of each dollar spent and setting up their investments to do the daily work paying the bills so they no longer have to. Pioneers like Bogle made that system achievable for everyday people outside of stuffy Wall Street boardrooms, and resourceful individuals like Pete Adeney (aka Mr Money Mustache), Jacob Fisker, and Vicki Robin have long served as role models for scores of prospective early retirees.

Not everybody funds their retirement the same way. Some people like Buffett-style stock picking while others prefer predictable regular income from dividend-paying stocks. But the most well-known method by far in the FIRE community is the rule of thumb collectively known as “the 4% rule”. In a famous paper written in 1994, an investment adviser named William Bengen determined that 4% was the maximum initial withdrawal rate for basic stock & bond portfolios that would have still not completely run out of money even over the worst rolling 30-year retirement period on record. Scores of subsequent studies and articles have since been written on the topic, and the concept has become so ubiquitous that many early retirees take the 4% rule as somewhat of an article of faith. I personally think there’s much more to the story, but I totally respect the history of the topic.

So when Vanguard recently released a white paper titled “Fuel for the F.I.R.E.: Updating the 4% rule for early retirees“, it makes sense that it generated lots of attention. It suggests that the old 4% rule is far too aggressive for modern retirees, which not only caused a bit of concern among the FIRE community but also elicited a round of “I told ya so’s” from FIRE skeptics who look down at early retirement and put lifetime-work on a pedestal. When a boring corporate white paper gets buzz like that, you know they hit on something people care about. And as a card-carrying member of the FIRE movement myself, I figured it was worth a read.

What I found certainly got me thinking, and I know it’s getting serious when I fire up the spreadsheets. So allow me to do a bit of a book report and talk about withdrawal rates.

The importance of assumptions

The Vanguard paper focuses heavily on Bengen’s work that I mentioned above, and it can be summarized as a detailed critique of 5 key assumptions baked into the research that established the 4% rule:

  1. The use of historical returns as a guide for future returns
  2. A fixed retirement horizon of 30 years that is designed for traditional retirement age
  3. An idealized academic methodology that ignores investing fees
  4. A lack of asset diversification beyond simplistic US stocks and bonds
  5. A fixed withdrawal method that ignores more flexible options

While I may not agree with every prescribed solution in the paper, I think they make a lot of great points about the diagnosed shortcomings of traditional withdrawal rate studies. In fact, I’ve written about the same issues at length.

Browse Portfolio Charts for a little while, and you’ll find articles about how withdrawal rates work that outline the critical importance of diversification in retirement portfolio performance. There are others that demonstrate just how shortsighted it is to judge a portfolio solely by its average return or percentage of stocks. There are tools for calculating withdrawal rates over different time horizons and even information on how to project future failure points for more recent start dates without resorting to guessing estimated returns.

Very early retirees might be interested in how to find the perpetual withdrawal rate that will last forever rather than simply focus on depleting a portfolio to zero over increasingly longer timeframes. For those thinking beyond simplistic spending assumptions, there are also articles about how smart retirement planning is about more than just avoiding failure and a Retirement Spending tool that allows people to study the retirement performance of any portfolio and spending strategy they can think of.

Of course, most fans of asset allocation are well-educated on the importance of minimizing fees. To take it one step further, the Fund Finder is an interactive tool that helps investors identify the lowest-cost fund options for any portfolio — including for investors outside of the United States and looking far beyond the default Vanguard options.

And speaking of investors outside of the US, Portfolio Charts is even pretty unique in its ability to address one more assumption that Vanguard overlooks. Not every investor lives in the United States! Between inflation, exchange rates, and very different local stock and bond markets, withdrawal rates can vary quite a bit around the globe even for the exact same portfolio.

To be clear, I’m certainly not the only person to write about these issues. Smart people like Wade Pfau, Michael Kitces, Karsten Jeske, and even the great William Bengen himself have also contributed to our evolving collective withdrawal rate knowledge for years in their own unique ways. One reason I provide all of the above information is to offer a bunch of links to my own analysis that may help you to expand the way you think about how to finance your own retirement. But for the purposes of this discussion, I also recognize that I may need to lay a bit of a foundation in terms of my own retirement research to hopefully lend a bit of credibility to what I have to say next.

As much as I agree with the majority of what Vanguard talks about in their paper on FIRE and the the 4% rule, I believe they completely miss the mark on two things. First, I think they grossly underestimate the typical educated retiree in the FIRE community, as all of these things are regular topics of conversation among those who advance past the very basic early concepts. But that’s a relatively minor quibble compared to my second issue that falls more directly in my wheelhouse.

To my eyes, the Vanguard analysis has a data problem.

To explain, let’s go back to the top item on the Vanguard list and talk about their views on historical returns.

When assumptions meet reality

The key methodological point that underscores the entire paper is the one they choose to discuss first — the idea that historical data is deceptive when predicting future returns for withdrawal purposes. Vanguard claims that William Bengen’s original 4% research “relied on the premise that historical returns are good indicators of the returns that investors can expect in the future.” They definitively explain that this is not the case, and as evidence they compare the real CAGR* of US stocks and bonds since 1926 to their own proprietary model of expected 10-year returns.

(*) They call it the “average historical return” in the paper, but discuss that it’s the geometric CAGR in the fine print.

And the result certainly isn’t encouraging.

US Stocks

Average real return: 7.5%

10-year forecast: 2.4%

US Bonds

Average real return: 2.4%

10-year forecast: -0.3%

I pulled the inflation-adjusted numbers above from the text of the paper, but the summary chart that Vanguard offers for the section is also pretty direct. Vanguard clearly prefers their internal model for estimated future returns over historical data when it comes to withdrawal rate research.

As Vanguard summarizes the analysis, “Past performance is no guarantee of future results. Relying on historical returns to predict future ones may make retirees overly optimistic about their probability of success.” While they don’t come right out and say it, they strongly imply that the numbers prove that the old 4% rule is based on an antiquated average return and no longer applies today.

It’s indeed pretty alarming. But the thing is, it’s not only completely wrong in terms of withdrawal rate methodology but also provably misleading.

Vanguard’s claim that Bengen’s withdrawal rate analysis depends on using past data to project future returns is a bit of a strawman argument, as Bengen says nothing of the sort. In fact, right on the first page of his famous original paper on the topic he argues the exact opposite. After describing the experience of a hypothetical adviser failing his clients by recommending spending based on historical average returns in a variable world of uncertain outcomes, he offers the following clear explanation:

“The logical fallacy that got our hypothetical planner into trouble was assuming that average returns and average inflation rates are a sound basis for computing how much a client can safely withdraw from a retirement fund over a long time. As Larry Bierwirth pointed out in his excellent article in the January 1994 issue of the this publication (“Investing for Retirement: Using the Past to Model the Future”), it pays to look not just at averages, but at what actually has happened, year-by-year, to investment returns and inflation in the past. He demonstrated that the long-term effects of certain financial catastrophes, such as the Depression or the 1973-1974 recession, can overwhelm the averages. Such “events” cannot be ignored, and the client should be made aware of them.”

William Bengen

Determining Withdrawal Rates Using Historical Data

In short, Bengen’s entire philosophy is to use historical data NOT to study past or future average returns, but to stress test withdrawal strategies using known worst-case scenarios. When using that process, future expected returns are actually irrelevant in terms of SWR calculations unless they become the new nightmare historical outcome that trumps all others that came before. And despite the instinct of nervous investors to always suspect that “this time is different” and the end is near with every new item of negative market news, with enough history at your fingertips that’s almost certainly not the case.

So with that in mind, let’s circle back to Vanguard’s quoted numbers. The first thing to note is that they are two very different statistics (the average real return since 1926 and the forward-looking 10-year expected return in 2020) and can’t necessarily be compared apples-to-apples. But the thing that really got my attention is what Vanguard did NOT explore — whether the expected 10-year return using their fancy-sounding Vanguard Capital Markets Model accurately predicted past withdrawal rates.

Now I don’t have insight into the details of Vanguard’s projection methodology, so I can’t speak for the specific calculations. I’m sure they’re very sophisticated models by people way smarter than I am. So to give them the benefit of the doubt, I’m going to make a bold assumption and concede that Vanguard has a true crystal ball and their projections are always 100% accurate. Beyond giving them a ton of credit, that actually makes backtesting pretty useful to test their assumptions about the importance of future 10-year returns to eventual withdrawal rates. After all, assuming they’re always right means that the historical record should perfectly match any similar projections Vanguard made in the past. So with the benefit of hindsight, how do Vanguard’s current expected returns compare to similar super-accurate projections we can find in the historical record?

Let’s start with stocks.

Studying the quoted Vanguard timeframe, each gray vertical column shows the forward-looking 10-year real CAGR for the total US stock market for every year since January 1927. Think of them as 100% accurate Vanguard estimated returns for the next 10 years at every point in time. The horizontal black line represents the CAGR over the full timeframe.

Next, look at the horizontal red line. That represents Vanguard’s current estimated 10-year return for stocks, which they imply jeopardizes the 4% rule because it’s so far below the average. I took the liberty of highlighting each real-world 10-year return that fell below that mark. There are 18 in total, meaning that 21% of the time the 10-year return for stocks was even lower than Vanguard estimates it today.

And yet, the 4% rule still worked.

Remember, it was calculated from the exact same set of data!

Next, let’s look at bonds.

Here you can really see the returns mountain and downslope since the early 80s that bond skeptics talk about when they question the utility of historical bond data. But you can also see the vast valley of terrible returns before that. Even with the gloomy 10-year Vanguard forecast of a -0.3% real return in bonds over the next 10-years, the actual 10-year real return of bonds fell below that mark 21 times since 1926, or about 25% of the time. In fact it’s even worse than that, as the annualized real return of bonds between 1933 and 1981 was a crazy-low -0.6% not just for 10 passing years but for nearly five decades of investing misery.

And yet, the 4% rule still worked.

It’s also not just a matter of fortunate timing where one asset picked up the slack when the other experienced a poor ten years. Consider the case of an unlucky retiree leaving their job in 1971, the beginning of a decade of skyrocketing interest rates and inflation that crippled stock and bond returns at the same time. An accurate 10-year forecast in 1971 would have predicted a 10-year annualized real return of 0% for stocks and -2.6% for bonds. To put that into context, here’s how it compares to Vanguard’s expected returns as of today. Which starting situation would you expect to experience worse withdrawal rates?

10-year expected real returns in 1971

Stocks: 0%

Bonds: -2.6%

10-year expected real returns in 2020

Stocks: 2.4%

Bonds: -0.3%

What if I told you that the measurable 30-year withdrawal rate of a 60/40 mix of stocks and bonds starting in 1971 was a healthy 4.8%? Really let that sink in, and current projections won’t look so scary.

Long story short, the low expected 10-year returns that Vanguard appears worried about today are nothing compared to even more dire 10-year timeframes in the past. When it comes to withdrawal rates, those terrible expected returns are not unusual aberrations that still need to be accounted for. In fact, they are already baked into the numbers by design. So the implication that moderately lower than average expected 10-year returns today mean that the 4% rule must be revised downward is simply not backed up by the facts.

Vanguard gets it right when they say that past results can’t predict the future. But history is pretty darned handy for testing the usefulness of a predictive model, and in this case I’m less than impressed with the results.

There’s more to life than a solid theory

So if 10-year returns have frequently been way worse historically than Vanguard expects today and the 4% rule survived just fine, why exactly are they suggesting that now is the time to revise portfolio expectations? That’s a good question.

I certainly can’t speak for Vanguard, and it’s possible that they just took the easy path by discounting the historical record out of hand rather than accurately explaining their own proprietary model. But I will say that it’s not the first time I’ve seen this same argument and usually the issue is much more mundane. Vanguard may be the 7-trillion pound heavyweight in the investing industry, but when discussing withdrawal rates they appear to fall into an all-too-common trap that I see all the time. By intuitively believing that withdrawal rates work as a function of the average rather than the worst-case, they imply that lower-than-average expected returns correspond to a reduction in the 4% rule rather than simply the natural variability that safe withdrawal rates already account for. But they just don’t work like that.

In addition, I’ve found in my own engineering experience that sometimes the smartest academics get so caught up in their own models that they lose sight of how those models track real-world results. Sure, it sounds pretty impressive to feed a Monte Carlo simulation with carefully calculated expected returns, standard deviations, and asset covariances and run 10,000 simulations to determine how often one expects portfolios to fail. But in their rush to dismiss the historical record in favor of their preferred proprietary methods (which I’m sure they hope to offer advisory clients at a very reasonable price), in this case Vanguard appears to have neglected to validate those predictions against known historical withdrawal rate data. While I’m no PhD quant, I know how my engineering boss might feel about a mathematical model I’m fixated on that can’t be replicated in a real-world test. When it’s you against nature, reality isn’t the problem.

And honestly, some of it is just common sense. Looking at the charts one more time, even a layperson can understand one major problem with leaning too heavily on 10-year expected returns as a model input — they change all the time.

Picture the same Vanguard analyst running an identical simulation in a few years once expected returns have jumped from the paltry red-line lows to well over double-digits. Will they proclaim that the 4% rule is now the 8% rule? And will you believe them? Why or why not?

From my perspective, that type of persistent average chasing is exactly the erroneous thought process that Bengen sought to remedy by using history to remain focused on the worst-case scenario. While there’s a place for expected returns analysis in terms of setting near-term expectations, when it comes to long-term retirement and your life savings is on the line I’d argue that Bengen’s more conservative approach is far more steady and dependable. Perhaps if Vanguard writes a future paper that takes Bengen’s advice to heart and compares their predictions against the historical worst-case scenarios rather than simply the long-term average, they might reach a different conclusion.

In the meantime, I’m just thankful for the data to study, the tools to independently evaluate ideas, and the avenues to share what I learn with the world. Please don’t get me wrong — it’s not about taking down Vanguard. I’m always open to new information, and for all I know I’ll be writing my own new article in the future revising my views on this topic. From the big dogs offering index funds all the way down to the girl sitting at her desk thinking hard about handing in that final retirement notice, we in the greater FIRE community are in it together.

I’m happy to do my part advancing the whole team forward, and I hope you are, too.

Are you thankful for the contributions towards the understanding of FIRE?