Executive Summary
As retirees and their planners adjust to the 'new normal' - a world of lower-than-average returns for the foreseeable future, many have questioned whether the historical safe withdrawal rate research is still valid. After all, if returns will be below average in the coming years, doesn't that imply safe withdrawal rates must be below average as well? In point of fact, though, safe withdrawal rates do not depend on average returns in the first place; the worst safe withdrawal rates in history that we rely upon are actually associated with 15-year real returns of less than 1%/year from a balanced portfolio! Accordingly, given current bond yields, dividend yields, and inflation, if the current environment for today's retirees will result in a "new record low" safe withdrawal rate, the S&P 500 would still have to be no higher in 2027 than it was in 2007 or even 2000! On the other hand, merely projecting equities to recover to new highs by the end of the decade or generating a mid-single-digits return would actually represent an upside surprise, allowing for higher retirement spending than 4.5% safe withdrawal rates!
The inspiration for today's blog post is some recent conversations I've had with other planners, who have questioned whether the safe withdrawal rate research is still relevant in today's low return environment. "In the 'new normal'", the planner usually states, "returns are likely to be lower than historical averages for both stocks and bonds. Doesn't that mean historical safe withdrawal rates are unrealistic?"
"Not at all," I reply, "Because historical safe withdrawal rates aren't based on historical averages. They're based on historical worst case scenarios."
Understanding Safe Withdrawal Rates
Safe withdrawal rates are determined by looking at the initial withdrawal rate, with that dollar amount adjusted each subsequent year for inflation, that would have worked given the actual sequence of stock, bond, and inflation that occurred during the historical time period. For instance, the chart below shows the safe withdrawal rate that would have worked for a 30-year time horizon for each rolling 30-year time period in the past 140 years (using publicly available data from Professor Shiller at Yale); the chart assumes a 60/40 portfolio (invested in large-cap stocks and intermediate-term government bonds) that is rebalanced annually after the actual stock and bond returns for that year, where spending adjusts each year based on the actual rate of inflation for the 30-year time horizon, and returns are modeled based on the actual sequence that markets delivered in that particular rolling 30-year time period.
As the chart reveals, the safe withdrawal rate for any particular 30 year time period has varied tremendously over the decades. The point of the chart, though, is not that the average safe withdrawal rate has been about 6.5% over the years (which is almost the same as the safe withdrawal rate you get simply assuming long-term average returns), nor that the safe withdrawal rate has peaked in the double digits a few times over the years. The point of safe withdrawal rates is the low points - the worst case scenarios. Whether the retiree started at the beginning of the Great Depression, entering into the stagflationary recession of the late 1960s and 1970s, or the credit crisis and panic of 1907 and the slow economic growth that followed, an initial withdrawal rate below 4.5% was both necessary and sufficient for the portfolio to survive for 30 years.
And of course, if a 4.5% withdrawal rate is used in virtually any other environment, the client will simply have funds left over, or can raise spending along the way; in point of fact, the safe withdrawal rate actually has a 96% probability of leaving more than 100% of the original starting principal! On the other hand, if the client actually lives through one of those historical worst case scenarios, the withdrawal rate is just low enough to sustain withdrawals for the time period.
Quantifying "Worst Case Scenarios"
So if safe withdrawal rates are based not on average returns but "worst case scenario" returns (at least, the worst we've seen in history), just how bad are those returns? The chart below shows the 30-year average annual compound growth rate for stocks and bonds for the four worst historical safe withdrawal rate scenarios, which all produced safe withdrawal rates in the 4% to 4.5% (depending on exactly which data set is used).
| 30-Year Nominal Returns |
Starting 1907 |
Starting 1929 |
Starting 1937 |
Starting 1966 |
Average of
Bad Periods
|
Overall 140-year Average |
|
Stocks |
7.77% |
8.19% |
10.12% |
10.23% |
9.08% |
10.35% |
|
Bonds |
4.25% |
1.74% |
2.13% |
7.85% |
3.99% |
4.81% |
|
Rebalanced Portfolio |
7.02% |
6.28% |
7.24% |
9.56% |
7.53% |
8.13% |
|
Inflation |
1.64% |
1.69% |
2.82% |
5.38% |
2.88% |
2.23% |
Notably, there appears to be little relationship between the nominal returns of stocks and bonds over 30 years, relative to the long-term average, given that these are all worst-case scenarios for safe withdrawal rates. Some had average equity returns, others did not. Some had above-average bond returns, others had below-average results. Some had above-average inflation, and others had below-average inflation. In fact, over the entire 140-year time horizon, neither the 30-year return in stocks nor bonds had much of a relationship to the associated safe withdrawal rate; the correlation of 30-year stock returns to safe withdrawal rates is a not-significant -0.09, and the 30-year bond return isn't much better at -0.15.
In part, the lack of relationship between long-term returns and safe withdrawal rates is because ultimately, when taking withdrawals, it doesn't matter if returns average out in the end, because a period of 10-15 years of low returns means there may be little or no account balance remaining when the good returns finally arrive for the last half of retirement. Accordingly, while 30-year returns have little relationship to safe withdrawal rates, returns over the first 15 years of the retirement time horizon have a much stronger relationship; in fact, the 15-year real (inflation-adjusted) return of the portfolio actually has a whopping 0.91 correlation to the safe withdrawal rate, as shown in the graph below. (And notably, these low-return environments are actually predictable from market valuation, given a -0.65 correlation between Shiller P/E10 and safe withdrawal rates, but that's a discussion for another day!)
Accordingly, the reality is that what drove the worst safe withdrawal rates in history were especially poor real returns over the first half of retirement; whether driven by low returns and low inflation, or high returns and high inflation, if the compound inflation-adjusted return of the portfolio was weak, so was the safe withdrawal rate, as shown below.
| 15-Year Real Returns |
Starting 1907 |
Starting 1929 |
Starting 1937 |
Starting 1966 |
Average of
Bad Periods
|
Overall 140-year Average |
|
Stocks |
-1.47% |
1.07% |
3.45% |
-0.13% |
-0.73% |
8.39% |
|
Bonds |
-0.39% |
1.79% |
-3.07% |
1.08% |
-0.15% |
2.85% |
|
Rebalanced Portfolio |
-0.70% |
2.29% |
1.23% |
0.64% |
0.86% |
6.17% |
|
Inflation |
1.64% |
1.69% |
2.82% |
5.38% |
2.88% |
2.23% |
Putting Historical Bad Returns Environments In Current Context
Given that safe withdrawal rates are based on historical worst case scenarios, and given the information we have about how bad those historical scenarios have been, we can begin to understand how bad returns would really have to be, from here, to lead to a safe withdrawal rate that is worst than anything seen in history.
The average real return on a 60/40 (re-)balanced portfolio associated with the worst safe withdrawal rate scenarios in history was a mere 0.86% average annual compound growth rate over the first 15 years of retirement. In point of fact, this was actually driven by a slightly negative real return in bonds (at -0.15%) and a slightly positive real return in equities of 0.73% (the reason the rebalanced portfolio returns were slightly higher than the returns of stocks or bonds separately was due to the favorable market timing of some of the rebalancing trades).
So what would this look like in the current return environment? With nominal inflation trending just over 2% and a 15-year government bond yielding just below 2%, the projected 15-year real return on bonds is projected to be slightly negative - suggesting that bonds are currently in an environment similar to historical worst case safe withdrawal rate scenarios.
Similarly, the current dividend yield on the S&P 500 is approximately 2%, suggesting that any real return in equities from here must come from price appreciation. However, the historical record shows that the "worst case scenarios" typically generate less than 1%/year in real returns for 15 years. Thus, generating a comparable real return for equities would assume no more than approximately 16% cumulative price appreciation from here; with an S&P 500 price level of approximately 1,350, this would driven the S&P 500 back to 1,567 in 2027, putting it at almost the exact same level it was at its peak in 2007 before the onset of the financial crisis and recession.
Viewed another way, what the data shows is that if you expect the coming safe withdrawal rates from here to be worse than anything seen in history, you need to assume not just below-average returns; you need to assume that the stock market cannot generate more than 1% real returns between now and 2027 given a 15-year real bond return of 0% at todays rates, and that if inflation increases from here that equities will fail to increase dividends dollar payouts, grow earnings, or provide any effective hedge to inflation whatsoever.
Given this reality, it's notable that merely getting 'new normal' returns of low single digits would actually represent not a risk to historical safe withdrawal rates; it would actually be an upside surprise that would result in materially higher lifetime spending! It would only be appropriate to assume a safe withdrawal rate lower than the historical 4% - 4.5% rate if you believe that equities will fail to deliver even a 1% real return over the next 15 years, implying (given current dividend and inflation levels) that the S&P 500 price level will be lower in 2027 than it was in 2007 (which would also be lower than it was in 2000, resulting in no appreciation for 27 years!).
So what do you think? Do you expect that the market 15 years from now still won't be any higher than it was 12 years ago, and that equities will generate a real return of less than 1% between now and the second half of the next decade? Would it impact your recommendations given that safe withdrawal rates are based on not on average returns, but less-than-1% real returns for 15-year time periods?
Is there any dataset that attempts to reconstruct the economics of a hegemon in decline? The concern that sits in the back of my mind is that our dataset is actually very small and only looks at the performance of companies in a political superpower during its rise.
I know the US isn’t going to fall overnight, but I would imagine that a long decline in hegemony would have major implications for investment returns.
When you say, “safe withdrawal rates are based on historical worst case scenarios” which worst case scenarios should be considered? I wouldn’t presume to predict any market 15 years into the future, but if we’re looking at the spectrum of worst case financial markets, there are many examples throughout history of real returns far worse than +1% CAGR for a 15 year period.
In fact, history is being made right now. Greek stocks are down 90% over the last 13 years. The Portugese market is down 67% over the last 12 years. Japanese stocks are down 77% over the last 22 years. You can look all across southern Europe right now and make a plausible case for 15 year time periods that will produce highly negative real returns.
Personally, I wouldn’t trust any safe withdrawal rate determination that wasn’t integrated with a robust approach to risk management.
Michael:
Thanks as always for your insight. I think this also ties in to your recent blog on modeling/forecasting rates of return that are much lower than average over the next 10 years and possibly higher afterwards. Valuations and sequencing of returns really matter. (BTW, I am curious how you are accounting for this in your financial plans.)
I am not sure how your withdrawal rate research cited coincides with the research that Wade Pfau has done but he would suggest much lower safe withdrawal rates based on his global data.
The recent white paper by Morningstar (David Blanchett) comparing withdrawal rate approaches was also illuminating.
Great stuff, Michael. And when you combine this with your earlier observations regarding making adjustments in planned spending during prolonged periods of low returns, it suggests that all the hysteria about the new normal is overblown. Of course, there’s no guarantee of ABSOLUTE financial security no matter how rich you are. What this suggests is the best counsel we can give to people is to have a good plan (dreams of perfection are the enemy of good)and to stop worrying so much that they forget to live their precious lives.
Am I the first to tell you that there’s no “chart below” the first paragraph of “Understanding Safe Withdrawal Rates”??
Eric,
Oops, indeed you’re the first person to say anything. I don’t know why the charts/graphs didn’t post, but they’re there now.
Thanks!
– Michael
Hi Mike
I am wondering why is it necessary to forecast long term returns and volatilities while Bill Bengen showed us a check as you go approach, a sort of a Dead Reckoning, for safe withdrawals?
I am sure that you’re familiar with the following:
Monitoring Current Annual Withdrawal Rate vs. Life Expectancy
(as I see it in William Bengen’s reports, and notice that it is for more than 30 years horizon. Bengen backested it, on rolling 30 years periods, from 1926 till 2005 on: Tax-advantaged acc with 40% Lrg caps, 20% Sm caps, 40% intermediate Gov bonds)
Age…….A….…B……C….…..D…..….E
50……….34….44..4.1%..5.1%..65%
55……….30….40….4.2…..5.3…..65%
60……….25……35….4.3….…5.4…….60%
65……….21….31….4.4…….5.5…..50%
70……….17….27….4.5…..5.6….…45%
75……….13….23….5.0…..6.3…..35%
80…………10….20……5.2…..6.5…..30%
85………….8……18……5.3…..6.6…..30%
90………….6……16……6.0…..7.5…..30%
95………….4……14….6.5…..8.1…..30%
100……….3….13……7.5…..9.4..…30%
A = Approximate unisex life expectancy (LE).
Note: I have updated column A according to tables in the current IRS tables in Pub 590.
B = Adjust LE by adding 10 years.
C = SWR% for the adjusted LE.
SWR%= safe withdrawal rate (SAFEMAX®).
D = Add 25% to C and get the RED FLAG%
E = equity suggested with three asset classes: LCS, SCS, and intermediate term Gov bonds, at the ratio of : SCS allocation = ½ LCS.
Note: I have added this column taken from Bill Bengen’s other reports.
Current Withdrawal Rate: end of year dollar withdrawal divided by beginning of year account value both in nominal terms. Don’t go over the RED FLAG.
Bengen, in his recent report from May 2012 added:
“One reason that “SAFEMAX” (the maximum “safe” historical withdrawal rate) is only 4.5% is that I assumed in my research the investor would ride every bear market down to its ultimate conclusion. Since 2000, that has been a painful experience for investors, with two declines of 50% or more to deal with in the S&P 500 index.”
And
“In summary, the 4.5% rule (and its infinite variations for time horizon, tax bracket, current market valuations, etc.) may be challenged in coming years. However, it appears to be working now. Rather than discard it willy-nilly, I recommend that advisors consider other weapons in their arsenal, such as the ones I have mentioned, to extend the life of a client’s portfolio.”
Source:
http://www.fa-mag.com/component/content/article/10772.html?issue=190&magazineID=1&Itemid=73
Going by Bengen’s above numbers I could see that withdrawing 4% COLAd from each of the following two Vanguard funds, since the beginning of 1972 (till now), showed no account depletion, but did show in certain periods Current Withdrawal Rates (CWR) above those marked as Red Flags by Bengen.
See it here:
A 4% COLAd withdrawal at the beg of 1972:
Year CWR CWR
……………….Wellesley Income……Wellington
………………………….VWINX…………….…VWELX
1978 5.6% 7.4%
1979 6.2 7.8
1980 6.8 7.9
1981 6.7 9.1
1984 5.9 7.7
CWR = End of year $ withdrawal divided be beginning of year $ account value. Both in nominal terms.
No wonder that even without depleting the accounts by the end of 2009, these relatively hi CWRs caused failure rates of above 20% when carrying out Monte Carlo Simulations (MCS) in these cases.
Mike, I assume that you’ll straighten out those tables in my above post. Also, since you added those illustrations I can see from the first illustration that it is safe to extend Bill Bengen’s data to 1871 regarding the 4% safe initial withdrawal.
As to clients with annuities, pensions, rental real estates, etc., Bill Bengen, covered it in his 2006 book, by providing a spreadsheet similar to that in the web-based program “MoneyGuidePro”.
Michael, your article makes me think of something related that I haven’t seen written about — namely, what safe withdrawal rates are appropriate for various ages? For example, you get a new client, a married couple in their upper 70’s. What is a safe withdrawal rate for them at their ages? I’m talking about people who never determined a SWR when they first retired and, therefore, never had a spending plan from year to year, but now have decided it’s time to see a financial advisor.
Mike – please specify the dates to which you refer in this part of your article . . .
“The average real return on a 60/40 (re-)balanced portfolio associated with the worst safe withdrawal rate scenarios in history was a mere 0.86% average annual compound growth rate over the first 15 years of retirement. In point of fact, this was actually driven by a slightly negative real return in bonds (at -0.15%) and a slightly positive real return in equities of 0.73% (the reason the rebalanced portfolio returns were slightly higher than the returns of stocks or bonds separately was due to the favorable market timing of some of the rebalancing trades).”
In addition, what was counter balance CAGR during the 15 year period that followed?
Many thanks!
Scott
Scott,
Thanks for your comment.
I’ll have to dig back through the data to see exactly which time periods these were from the data set. I’ll try to post a follow-up when I can go back to the numbers.
Regarding your final comment, the growth rates in the subsequent periods were all dramatically higher – in fact, as memory serves, they were all not just much higher than these low data points, but were much higher than the long-term average as well, as the end-point of these time periods generally resulted in extremely compressed market valuations.
– Michael
Michael, excellent—as everyone else has pointed out here!
Two items:
a) Did you ever or would you be willing to now, provide the dates of the 15-year period under inspection (if you just updated the article long ago, apologies if I missed it—perhaps reply with which paragraph and where you elaborated on the dates or timeframe for the data you were using)?
b) Given we’re roughly 10 years out now from this article—who’d a known we’d continue to have one of the greatest bull markets in history continue for another decade (congrats to those who retired in 2012 and enjoyed their lives w/o worrying about withdrawal rates!)?!!—I for one would love to see an updated analysis, since all of the prognostications of a coming long decline seem even more pronounced now and Joe Elsasser’s comment about a hegemon in decline seems to quite well fit the USA even more so now, would you consider re-publishing and updating this analysis for 2022?
Cheers and thank you very kindly!
I appreciate this excellent article but have to admit I’m increasingly astonished by all the forecasts by knowledgeable people that omit taxes and fees … but especially taxes. This omission makes all the conclusions suspect.