Does your advisor understand the hidden risks of the 4% rule?
A Common Problem in Need of a Common Solution
This blog is the first part of a series. To read part 2, click here. To read part 3, click here.
Regardless of their financial literacy or prior professional expertise, retirees face some of the hardest questions in portfolio management each and every day: how much can you withdraw from a portfolio and for how long? Even mathematical and analytical clients are challenged by these questions, leading to lots of conversations by financial advisors and investment advisors around withdrawal rates. Life is different than in the accumulation stage, a stage where contributions and compound interest worked together to grow an account; during the distribution stage retirees need to withdraw funds even while the stock and bond markets continue to gyrate.
Portfolio management science tells us that your asset allocation, that is the mix of stocks and bonds, and your withdrawal rate are two of the larger factors that determine retirement success. There are two major studies that advisors and the investment community point to as watershed moments in withdrawal rate research: the SAFEMAX research by Bill Bengen, and the article Retirement Savings: Choosing a Withdrawal Rate that Is Sustainable by Cooley, Hubbard and Walz (you may recognize this better if we call it by its informal name, the Trinity study). These two studies from the 1990s may not be household names across the US, but they are part of the basis for the “4% Rule” that is often used by retirees and investment advisors as a guiding light when modeling retirement.
In this post, we’ll investigate this rule that is not a rule and help you determine whether your advisor is using all the research available on retirement income planning. Let’s start by focusing on the better-known Trinity study and the basics of the 4% Rule.
Origins of The 4% Rule
For the Trinity study, the authors considered five different portfolios (100% equity, 75% equity/25% bonds, 50% equity/50% bonds, 25% equity/75% bonds, 100% bonds) and four different retirement horizons (15, 20, 25 and 30 years). They then tested a number of different withdrawal rates ranging from 3% to 12%, adjusted each year for inflation to maintain the retiree’s lifestyle. The authors found that a 4% withdrawal rate had a 98% chance of success with a portfolio of 100% stocks over a thirty-year horizon – this is one birthplace of the 4% Rule.
Now that we’ve briefly discussed what the authors did do – let’s discuss what they did not do. In the study, the withdrawal rate is based on the initial portfolio balance and the amount is only adjusted for inflation, annually. In other words, they project that a retiree can live off of the same exact amount of money (adjusted for inflation) for thirty years and never change their spending. This ignores the reality that accidents and one-time expenses can be unpredictable. I think the authors took this shortcut for computational and modeling ease. Other research from retired Berkley Professor Gordon Pye¹ does model emergency spending, and the conclusion is that with emergency spikes in spending a 3% withdrawal rate is safer. That is a 25% decrease!
A second concern with the Trinity study is their use of index data from 1926 to 1995 to represent an investment portfolio. They are not alone, as this study has been replicated many times with more recent data and varying assumptions. The use of index data makes sense from an academic standpoint – you want the longest history of data you can possibly get to construct as many thirty-year retirement horizons as possible, and there are not enough mutual funds or ETFs that have been around long enough for an individualized comparison. However, it is inappropriate for a retiree to directly compare because while you can invest in a fund that tracks the index, or buy the underlying stocks in an index, you cannot simply invest in an index like the S&P 500 or NASDAQ. There are also fees, expenses, taxes, and transaction costs that the Trinity study skips over for the simplicity of making an academic model.
The 4% Rule Revisited
To partially address my concerns and share some insights, we recreated the research & study of the 4% Rule for a more realistic and practical approach in which investors hold real mutual funds with real fees and expenses². To do this we have to sacrifice the historic data³ that has only one “path” (a starting investment + all the returns that happen over thirty years) but spans seventy+ years. Instead we used some computation methods called bootstrapping to create more alternate-reality paths as a substitute for the one path through history. Specifically, we studied VUSTX (the Vanguard Long-Term Treasury Mutual Fund) and VFIAX (Vanguard 500 Index Mutual Fund), replacing the indexes. These low-cost mutual funds were available to investors – in fact the Vanguard 500 Index fund is the first index fund that came to market.
In the original Trinity study, they have 70 annual returns and about 17 of those annual returns include recessions. The study kept the sequence of returns intact, capturing the true-to-history returns for each thirty-year horizon, creating 41 holding periods each spanning 30 years from which to calculate success. The first thirty-year holding period spans from 1926 – 1956, the second being 1927 – 1957 and so on. Most authors replicating the study have also replicated this method for returns. Yes, that means the “4% Rule” was based on only 41 historical starting points! This doesn’t invalidate the results, but it is a concern for generalizing into an infinite future.
One way I felt I could improve on existing research was by using some methods from financial econometrics to create more data points and analyze the outcomes. I wanted to see if the theory held for a different data set. We picked daily adjusted close prices from 3/13/1993 – 4/14/2022 to construct monthly rolling returns (giving us a lot more data), and the bootstrapping method allowed us to create more than one path of history.⁴ Our use of monthly returns gives 100x more data points compared to using only annual returns, but it differs in how frequently there is a recession.⁵ There are 7,338 monthly rolling returns in our sample, about 700 of which are recession monthly returns. Stay tuned for more research & analysis on how we increase the length and number of recessions and how it affects retirement planning & spending in our next post in this series.
For this project, I constructed 100,000 paths that each last 30 years and then computed the probability of success. To keep comparisons as simple as possible we defined success the same way as the Trinity study – ending with any positive balance (even $1.00) would be a success, though we fully acknowledge this is not in line with many retirees’ goals. If your goal is to leave an inheritance or donate to charities, understand that these withdrawal rates might be too optimistic for your individual case. Don’t let these studies & interpretations substitute for doing a proper financial plan with a fiduciary! With that in mind, given our changes to the methodology, does the 4% rule still hold?
Yes! But I’d like to highlight worst case scenarios, to put the numbers from the table above into a fuller perspective. Instead of looking at just the big picture, let’s drill down and look at the individual paths. When I look at the 10 worst scenarios from the 100,000 paths for each portfolio/allocation combination with a 4% withdrawal rate, the results tell me the probabilities in the table above might be understating the true risk a retiree would face.
For example, there is the conservative 20/80 asset allocation that invests domestically and buys long-term treasury bonds. This portfolio runs out of money in only 181 months (about 15 years) in the worst case scenario. The balanced 60/40 asset allocation’s worst path is running out of money in 265 months (about 22 years).
Making the 4% Rule Work for You
So what does this all mean now? Overall, we see that the high fixed-income allocations have high probabilities of success from the tables, but looking at the worst case scenarios they run out of money the soonest and end with the largest loss after thirty years. That doesn’t necessarily sound conservative. On the other hand, the higher equity allocations (60 or 80% equity) usually run out of money last, compared to the other allocations in these worst case scenarios, and have less loss.
Why is this the case? The higher fixed-income portfolios are less volatile or risky than the high equity portfolios (which is why they’re referred to as conservative allocations), but with less risk there is less potential for return. This can affect retirees especially: when there is an unfortunate sequence of returns (like early losses) it’s more difficult for a high fixed-income portfolio to make up those losses later on while still taking distributions.⁶ One way to handle this is to ask your advisor about bucking another common practice: having more stock, though it seems more risky, looks to be one way to protect against running out of money.
All in all, our study finds similar conclusions to the original Trinity study – the 4% Rule, under all of the assumptions outlined above, holds with a high probability. However, the biggest drawback of all of these studies so far is that sequence of returns risk is still hiding beneath the surface. There are going to be possible scenarios out there that our data simply cannot reflect no matter how many paths we construct, and there will be bad luck.
Now, just because our data cannot perfectly reflect every possible scenario does not mean we are powerless against the unknown. Fellow financial planner Michael Kitces outlines the biggest danger zone for retirees in his blog “The Portfolio Size Effect and Using a Bond Tent to Navigate The Retirement Danger Zone,” published in 2016. According to Kitces, the biggest risks to a retiree’s portfolio happen towards the end of the accumulation phase and at the very start of retirement when you start withdrawing – he identifies the size effect as the culprit – which is part of the sequence of returns risk we describe here.
So how do we deal with sequence of returns risk and the size effect and prepare our clients for retirement? We code them into our models and watch what happens, then we repeat and improve on the same study outlined above. In part 2 of our series, we will code in the danger zone Kitces discusses. Instead of relying on probabilities or chance, we will simulate severe losses and recessions at the beginning of each retirement and observe what it takes for the portfolio to recover. Using these results we can start outlining a plan for a much safer withdrawal rate.
Anessa Custovic, PhD
¹“The Effect of Emergencies on Retirement Savings and Withdrawals,” Journal of Financial Planning (Nov. 2010).
²All data for the comparison study was pulled from Yahoo Finance’s database of historical data, which includes a large number of publicly-listed investments.
³Mutual funds and ETFs do not have extremely lengthy historical data as index funds do. We sought to include funds that 1) had a decently long history and 2) are still traded today in order to keep it realistic.
4The Trinity study and many similar studies use historical returns as is, keeping the order of returns exactly the same. We rolled daily returns to create monthly returns, using non-traditional periods other than month end to month end. We computed 30 years of monthly returns randomly from the historical series repeatedly to generate 100,000 possible return sequences given the original monthly data. While we do not have historical data as long as the original study, we have more possible sequences of returns. I chose 100,000 paths because it’s large enough to pick up most sequences of returns given the number of possible combinations of returns from our sample.
5The proportion of recession days we have is smaller due to the dates of our data, which include the dot-com bubble, great financial crisis, and COVID crisis. Each of the 7,400 monthly rolling returns in our real-life mutual funds have the same likelihood of being chosen for any given path. Comparing it to the Trinity study, we can review about 2,400x more paths for every portfolio allocation (i.e., 60% equity/40% bond, etc.)
6For example, someone retiring in the first month of the Great Recession immediately takes a big loss right as they start withdrawals and has to stomach months of compounding losses while taking out money simultaneously. It’s more likely this retiree will run out of money compared to someone retiring in the last month of the Great Recession; waiting just a few months to retire would decrease that chance.
