You might have noticed that I haven’t published any guest posts for a while. I even explicitly state on my contact page that I’m no longer interested in publishing any guest posts. But every once in a while you make an exception to the rule. David Graham, actually, Dr. David Graham (FIPhysician), has been on a roll with a bunch of top-notch guest posts on other personal finance blogs; first writing for the White Coat Investor blog on Roth Conversions, then two guest posts on Physician on FIRE, first on Asset Location in Retirement, and then on Buffer Stock and Bucket Strategies to alleviate Sequence Risk in retirement. All really important topics! And after sending a few emails back and forth with the good Dr. Graham we agreed on a topic for him to publish a guest post here on the ERN blog, Instead of using backward-looking historical return windows, as I would normally do in my SWR Series, why not check the sustainability of the 4% Rule with forward-looking return projections? Vanguard and a lot of financial companies publish those every year. Sounds like an interesting exercise! So, without further ado, please take over Dr. Graham…
As we all know, ERN is the man when it comes to Safe Withdrawal Rate (SWR) and FIRE. Reading between the lines, he has a different opinion on SWRs for a 60-year retirement vs. a more traditional 30-year plan. Obviously, using only historical data, it is more difficult to study SWR with rolling 60-year stock and bond returns than 30-year periods. Nevertheless, FIRE often subscribes to the 4% rule despite a prolonged period of income demands on the accumulated nest egg. In order to further understanding of the 4% rule over a longer than usual planned retirement, let’s visualize the 4% rule over a 60-year period and see what we can learn.
Case Presentation
Assume a $1 million dollar portfolio is invested via a Three Fund Portfolio. There is $500,000 in a brokerage account, $400,000 in a 401k, and a Roth IRA worth $100,000. As for asset location, the Roth is invested only in US Stock. The brokerage account and 401k split the rest of the assets in a 60/40 or 90/10 stock/bond asset allocation. Withdrawals are from the brokerage account first, the 401k second, and the Roth last.
[ERN: This also assumes flat/constant asset returns in line with the Vanguard 2019 Global Market Outlook, more on that below. Also, within the 3-fund portfolio, there’s an 80/20 split between U.S. stocks and International stocks (developed and emerging). So, for example, the 60/40 portfolio has 48% U.S. Stocks, 12% International Stocks and 40% U.S. Total Bond Market Fund.]
60/40 Portfolio with 4% SWR
In Figure 1, at age 31 there is a 4% withdrawal from the account. That is $40,000 (4% of $1M). Assume inflation is a fixed 2% per year in this example, so the next year, $40,800 is withdrawn. A common misconception of the 4% rule is that you use inflation to adjust the 4%. You don’t. Only in the first year do you use 4% to calculate the withdrawal amount. In subsequent years, use inflation to increase the dollar amount. In this instance the 2% yearly inflation increase is $40,800, $41,616, $42,448, etc.
Since investments in the portfolio increase in value, the slow inflation of the withdrawal amount stays close to 4% withdrawal percentage for about a decade. Then, the withdrawal percentage increases rapidly. After 44 years, it depletes the portfolio. Monte Carlo simulation predicts a mere 18% chance that this plan will succeed.
Assumptions, Assumptions
In this simulation, assumptions are everything. Bengen’s original 4% rule and the subsequent Trinity study use historical data. As we all know, past performance is not indicative of future results. Recently, Vanguard released assumptions for 10-year returns. I utilize these assumptions in this scenario. For full disclosure, Big ERN cautioned me about using 10-year projections for results over 60 years.
[ERN: True! I don’t think it’s unrealistic or even outright wrong to assume low returns for the entire 60-year horizon. Certainly, I personally hope that high valuations today (stocks and bonds!) cause low returns only over the medium term, say, over the next 10 years, before we revert back to more average returns, say, around 6.7% real for equities and 1.5% real for (government) bonds. But I can also rationalize persistently low asset returns going forward, certainly much lower than previously experienced. A lower trend GDP growth, lower population growth, high government debt causing higher future taxes, etc. may all lead to leaner real asset returns going forward. Over several decades! Again, I hope that will not materialize but it’s a good exercise to see how a 60-year retirement would perform under persistently low returns. As a “kick the tires” exercise!]
Nonetheless, it is interesting to look at to see how different SWR perform given these assumptions. Again, there are limitations on historical data when looking at such prolonged time frames.
He also states that people don’t usually run out of money due to low average returns, but because of Sequence Risk. True, but as Kitces points out, prolonged mediocre returns can be worse than a market crash.
[ERN: Agree. A nice example would be the 2008/9 crash, which was actually not that bad from a Sequence Risk perspective. The drop was swift, and so was the recovery. A 60/40 portfolio with a 4% Rule would have recovered quite nicely from that bear market. Contrast that with the 1965/66 retirement cohorts: they experienced lackluster returns until the early 1970s, then multiple recessions. Those made the mid-60s some of the worst retirement cohorts on record. But just to be sure, that case would still qualify as Sequence Risk, in my book!]
In addition, assume a 2% inflation rate year over year in this scenario. Obviously, inflation is lumpier than that, and inflation adds additional complexity to resulting conclusions.
Future (Expected) Returns
Instead of using historical data, let’s assume that Vanguard’s assumptions hold on average for 60 years. They have a nifty Infographic with the following assumptions for 10-year returns: US equities 4-6%, US bonds 2.5-4.5%, International equities 7.5-9.5%. [ERN: these are all nominal returns, always keep in mind the big distinction!] For this example, we will use 5%, 3.5%, and 8.5% returns, respectively.
[ERN comment 1: People may find the 5% return expectation for U.S. Stocks way too low. And remember, that’s only 3% after inflation! But historically, the 10-year return expectations were about in-line with 1/CAPE. With a Shiller CAPE at around 30, that would put the real expected return at around 3.3%, roughly in line with the Vanguard estimate.]
[ERN comment 2: I personally find the 3.5% expected return gap between international and domestic stocks a bit large. 1-2% seems more realistic. Two weeks ago, when I looked at the return expectations produced by JP Morgan and Bank of New York Mellon, the gap between U.S. all-caps and international stocks was closer to 1% (BNYM) to 2% (JPM). Even with only 12% portfolio weight in international stocks, the roughly two additional percentage points mean 0.24% extra annualized return, which will make a big difference over 60 years!]
Above you can see what a difference a small change in SWR makes over time. Now, we start with a 3.25% withdrawal from the portfolio. Note the bump around age 52. Reviewing cash flows, the brokerage account runs out of money at that time and the 401k is tapped resulting in increased tax payments.
[ERN: I love this level of detail. Normally, we don’t take into account the tax implications. I did a more personalized and detailed analysis in most of the case studies I did in 2017/18, looking at both asset allocation, asset location and taxation. But in the context of the Safe Withdrawal Rate Series, I normally assume you simply withdraw 1/(1-TaxRate)xTargetConsumption. So you have to gauge your average tax rate in retirement and distinguish between gross and net withdrawals!]
Over time, the withdrawal percentage slowly increases. Monte Carlo simulation results in a 44% chance of success.
Remaining Portfolio in the 60/40 Example
Above, see how much money is left after 60 years when you use a 60/40 portfolio. In green, with a 4% SWR, you run out of assets. In blue, the lower 3.25% SWR grows slowly over time. $1.6M remains after 90 years (mostly in the Roth). [ERN: This is in nominal dollars, so don’t get too excited about the $1.6m: that’s worth only about $500k in real inflation-adjusted dollars!!!]
Trinity Study and More Aggressive Asset Allocation
We know from the Trinity Study that more aggressive asset allocations performed better over longer periods of time. What if the portfolio is aggressively invested in stocks?
Now we can see how a more aggressive portfolio affects a 4% SWR. Here, it takes almost 50 years before inflation of the withdrawal amount starts to affect the withdrawal percentage. If you are more aggressive in your stock to bond ratio, you can tolerate a higher SWR. Monte Carlo simulation supports a 34% chance of success.
Starting with a more aggressive portfolio and a lower SWR, you see the SWR never increases above 3.25%. Portfolio value continues to increase over time and Monte Carlo predicts a 53% chance of success.
A brief word about Monte Carlo simulation. Generally, when doing financial planning, the goal is to get Monte Carlo between 80-90% of success. More than that, and a plan is a failure because you die with too much money. Less than that, and there is a good chance that your plan will fail due to running out of money. Despite Monte Carlo simulation’s obvious issues, it is interesting to note—given such a long time frame—success rates of all these plans are quite low.
In green, you see the remaining portfolio balance of 4% SWR with the 90/10 portfolio. Only the 3.25% SWR with a 90/10 portfolio demonstrates significant growth over time, potentially ending up with over $6M in assets.
Lessons Learned for SWR over 60 years
Other obvious limitations, aside from the assumed returns noted above, are too numerous to mention in this methodology. However, FIRE folks should be interested to see how 4% vs 3.25% SWR affects portfolios over 60 years. We also know from the Trinity Study that there is a sweet spot for asset allocation. You want to have enough stocks for longevity planning and to fight off inflation.
ERN points out that Sequence of Return Risk (SORR) is the main concern early on and suggests being conservative early in early retirement and then using a rising equity glidepath. Kitces also has some interesting work on planned allocation strategies with equity glidepaths or bond tents. And ERN wrote about glidepaths using historical simulations in Part 19 and Part 20 of the SWR Series.
How Does this Simulation Hold Up to ERN’s Model?
ERN has the best data on the internet, seen above. This table might just be the coolest thing on the internet in relation to FIRE. Note that 30 vs. 60 years is very predictive of any portfolio asset allocation surviving. With a 4% withdrawal rate over 60 years, however, 100% stocks succeed 89% of the time and 50% stocks only 65%. Also, obvious to all, the higher the stock percentage, the better the outcome.
Conclusion
Sixty years is a long time. Most of us aren’t even that old, and it is hard to conceive of what the world will look like then. The SWR for a planned 60-year retirement is not 4%. The rule of 25x is a good starting point, but not the final solution to early retirement.
* * *
About the author: David Graham, MD is an Infectious Diseases physician and a Registered Investment Advisor in the state of Montana. His blog can be found at FiPhysician.com
[ERN: Thanks, Dr. Graham! I learned a lot! Based on my own research, I’m obviously not a big fan of the 4% Rule, certainly not for a 60-year retirement horizon. Dr. Graham’s methodology makes me even more nervous about the 4% Rule: The 60/40 portfolio runs out of money after a little more than 40 years, even without any return volatility and thus without Sequence Risk. The Monte Carlo Simulation success probabilities are also woefully low. And this already assumes aggressive return expectations for the non-U.S. equity portion. So, while I can understand when people criticize historical simulations (“past results are no guarantee for future results”) the alternative, i.e., using forward-looking return estimates looks almost worse if you take into account today’s valuations.]
We hope you enjoyed today’s post. Looking forward to your comments below!